August 2018, Volume 24, Number 3

ENVIRONMENTAL & ENGINEERING GEOSCIENCE

Environmental & Engineering Geoscience AUGUST 2018

VOLUME XXIV, NUMBER 3

Volume XXIV, Number 3, August 2018

THE JOINT PUBLICATION OF THE ASSOCIATION OF ENVIRONMENTAL AND ENGINEERING GEOLOGISTS AND THE GEOLOGICAL SOCIETY OF AMERICA SERVING PROFESSIONALS IN ENGINEERING GEOLOGY, ENVIRONMENTAL GEOLOGY, AND HYDROGEOLOGY

Environmental & Engineering Geoscience

Environmental & Engineering Geoscience

Volume 24, Number 3, August 2018

Volume 24, Number 3, August 2018

Table of Contents

Table of Contents

251

Using GIS-Based Spatial Analysis to Determine Factors Influencing the Formation of Sinkholes in Greene County, Missouri Shishay T. Kidanu, Neil L. Anderson, and J. David Rogers

251

Using GIS-Based Spatial Analysis to Determine Factors Influencing the Formation of Sinkholes in Greene County, Missouri Shishay T. Kidanu, Neil L. Anderson, and J. David Rogers

263

Simulation of Tailings Flow Resulting from a Dam Breach Using Smoothed Particle Hydrodynamics Poulad Daneshvar and Attila M. Zsaki

263

Simulation of Tailings Flow Resulting from a Dam Breach Using Smoothed Particle Hydrodynamics Poulad Daneshvar and Attila M. Zsaki

281

Technical Note: Development of a Rapid System to Diagnose Ground Settlement Jenny L. Wolicki, Paul M. Santi, Benjamin D. Haugen, Jacquelyn A. Hagbery, Ethan J. Faber, Stephen N. Semmens, and Hayden E. Brown

281

Technical Note: Development of a Rapid System to Diagnose Ground Settlement Jenny L. Wolicki, Paul M. Santi, Benjamin D. Haugen, Jacquelyn A. Hagbery, Ethan J. Faber, Stephen N. Semmens, and Hayden E. Brown

293

Groundwater Vulnerability Mapping in Urbanized Hydrological System Using Modified DRASTIC Model and Sensitivity Analysis Ismail Chenini, Adel Zghibi, Mohamed Haythem Msaddek, and Mahmoud Dlala

293

Groundwater Vulnerability Mapping in Urbanized Hydrological System Using Modified DRASTIC Model and Sensitivity Analysis Ismail Chenini, Adel Zghibi, Mohamed Haythem Msaddek, and Mahmoud Dlala

305

Water-Mortar Interaction in a Tunnel Located in Southern Calabria (Southern Italy) Giovanni Vespasiano, Pasqualino Notaro, and Giuseppe Cianflone

305

Water-Mortar Interaction in a Tunnel Located in Southern Calabria (Southern Italy) Giovanni Vespasiano, Pasqualino Notaro, and Giuseppe Cianflone

317

An Analysis of the Layered Failure of Coal: New Insights into the Flow Process of Outburst Coal Qingyi Tu, Yuanping Cheng, Qingquan Liu, Liang Wang, Wei Zhao, Wei Li, Jun Dong, and Pinkun Guo

317

An Analysis of the Layered Failure of Coal: New Insights into the Flow Process of Outburst Coal Qingyi Tu, Yuanping Cheng, Qingquan Liu, Liang Wang, Wei Zhao, Wei Li, Jun Dong, and Pinkun Guo

333

Effect of Oil-Degrading Bacteria on Geotechnical Properties of Crude Oil–Contaminated Sand Hossein Soltani-Jigheh, Hamed Vafaei Molamahmood, Taghi Ebadi, and Ali Abolhasani Soorki

333

Effect of Oil-Degrading Bacteria on Geotechnical Properties of Crude Oil–Contaminated Sand Hossein Soltani-Jigheh, Hamed Vafaei Molamahmood, Taghi Ebadi, and Ali Abolhasani Soorki

343

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole Norbert H. Maerz, Kenneth J. Boyko, Gary J. Pendergrass, and Justin W. Brown

343

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole Norbert H. Maerz, Kenneth J. Boyko, Gary J. Pendergrass, and Justin W. Brown

349

Technical Note: The El Indio Mine Closure Plan Effects over the Water Quality of the Upper Elqui Basin Jorge Oyarzún, Jorge Nú ñez, Hugo Maturana, and Ricardo Oyarzún

349

Technical Note: The El Indio Mine Closure Plan Effects over the Water Quality of the Upper Elqui Basin Jorge Oyarzún, Jorge Nú ñez, Hugo Maturana, and Ricardo Oyarzún

Using GIS-Based Spatial Analysis to Determine Factors Influencing the Formation of Sinkholes in Greene County, Missouri

Using GIS-Based Spatial Analysis to Determine Factors Influencing the Formation of Sinkholes in Greene County, Missouri

SHISHAY T. KIDANU1 NEIL L. ANDERSON J. DAVID ROGERS

SHISHAY T. KIDANU1 NEIL L. ANDERSON J. DAVID ROGERS

Department of Geosciences and Geological and Petroleum Engineering, Missouri University of Science and Technology, Rolla, MO 65409

Department of Geosciences and Geological and Petroleum Engineering, Missouri University of Science and Technology, Rolla, MO 65409

Key Terms: Geographic Information Systems (GIS), Sinkhole, Geographic Weighted Regression (GWR), Ordinary Least Squares Regression (OLS), Greene County

weighting in models of sinkhole susceptibility or hazard mapping.

Key Terms: Geographic Information Systems (GIS), Sinkhole, Geographic Weighted Regression (GWR), Ordinary Least Squares Regression (OLS), Greene County

INTRODUCTION ABSTRACT Sinkholes are inherent features of the karst terrain of Greene County, Missouri, that present hazards and engineering challenges to construction/infrastructure development. Analysis of relationships between the spatial distribution of sinkholes and possible influencing factors can help in understanding the controls involved in the formation of sinkholes. The spatial analysis outlined herein can aid in the assessment of potential sinkhole hazards.

In this research, Geographic Information System–based ordinary least squares regression (OLS) and geographically weighted regression (GWR) methods were used to determine and evaluate principal factors appearing to influence the formation and distribution of karst sinkholes. From the OLS result, seven out of 12 possible influencing factors were found to exert significant control on sinkhole formation processes in the study area.

These factors are overburden thickness, depth to groundwater, slope of the ground surface, distance to the nearest surface drainage line, distance to the nearest geological structure (such as faults or folds), distance to the nearest road, and distance to the nearest spring. These factors were then used as independent variables in the GWR model. The GWR model examined the spatial nonstationarity among the various factors and demonstrated better performance over OLS. GWR model coefficient estimates for each variable were mapped. These maps provide spatial insights into the influence of the variables on sinkhole densities throughout the study area. GWR spatial analysis appears to be an effective approach to understand sinkhole-influencing factors. The results could be useful to provide an objective means of parameter

1 Corresponding

author email: stkq7f@mst.edu.

INTRODUCTION

Karst topography develops on carbonate and evaporitic rocks, primarily by dissolution of soluble minerals. It is usually characterized by numerous sinkholes, caves, losing streams, springs, and preferential seepage pathways often influenced by geologic structure, stratigraphy, and watershed area. Karst is often a challenging environment when dealing with groundwater, engineering, and environmental issues (Chalikakis et al., 2011). Sinkholes are one of the most significant hazards in karst areas (Waltham et al., 2005; Gutiérrez, 2010). Sinkholes that suddenly collapse can result in loss of human life and property, and ground deformation associated with subsidence often damages infrastructure, such as highways and utilities (Carbonel et al., 2014). Thousands of sinkholes have been identified in the state of Missouri; Greene County, in particular, is one of the counties in the state most known for the occurrence of sinkholes.

The formation of sinkholes is influenced by a combination of interacting geologic, geomorphologic, hydrologic, and anthropogenic factors (Kaufmann, 2008; Galve et al., 2009; and Doctor and Doctor, 2012).

Ascertaining the main influencing factors and understanding the nature of their interactions can enable researchers to better understand where and how individual sinkholes may appear. The interactions between influencing factors are frequently not obvious and are often hidden from direct observation, making it unlikely to predict the occurrence of an individual sinkhole at a specific site. Nevertheless, the analysis of the spatial statistical relationships between sinkhole density and the potential influencing factors could help determine the principal causal factors influencing the formation and general spatial distribution and density of sinkholes in a particular area. Identifying the major influencing factors and the interactive

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

weighting in models of sinkhole susceptibility or hazard mapping.

251

ABSTRACT Sinkholes are inherent features of the karst terrain of Greene County, Missouri, that present hazards and engineering challenges to construction/infrastructure development. Analysis of relationships between the spatial distribution of sinkholes and possible influencing factors can help in understanding the controls involved in the formation of sinkholes. The spatial analysis outlined herein can aid in the assessment of potential sinkhole hazards.

In this research, Geographic Information System–based ordinary least squares regression (OLS) and geographically weighted regression (GWR) methods were used to determine and evaluate principal factors appearing to influence the formation and distribution of karst sinkholes. From the OLS result, seven out of 12 possible influencing factors were found to exert significant control on sinkhole formation processes in the study area.

These factors are overburden thickness, depth to groundwater, slope of the ground surface, distance to the nearest surface drainage line, distance to the nearest geological structure (such as faults or folds), distance to the nearest road, and distance to the nearest spring. These factors were then used as independent variables in the GWR model. The GWR model examined the spatial nonstationarity among the various factors and demonstrated better performance over OLS. GWR model coefficient estimates for each variable were mapped. These maps provide spatial insights into the influence of the variables on sinkhole densities throughout the study area. GWR spatial analysis appears to be an effective approach to understand sinkhole-influencing factors. The results could be useful to provide an objective means of parameter

1 Corresponding

author email: stkq7f@mst.edu.

Karst topography develops on carbonate and evaporitic rocks, primarily by dissolution of soluble minerals. It is usually characterized by numerous sinkholes, caves, losing streams, springs, and preferential seepage pathways often influenced by geologic structure, stratigraphy, and watershed area. Karst is often a challenging environment when dealing with groundwater, engineering, and environmental issues (Chalikakis et al., 2011). Sinkholes are one of the most significant hazards in karst areas (Waltham et al., 2005; Gutiérrez, 2010). Sinkholes that suddenly collapse can result in loss of human life and property, and ground deformation associated with subsidence often damages infrastructure, such as highways and utilities (Carbonel et al., 2014). Thousands of sinkholes have been identified in the state of Missouri; Greene County, in particular, is one of the counties in the state most known for the occurrence of sinkholes.

The formation of sinkholes is influenced by a combination of interacting geologic, geomorphologic, hydrologic, and anthropogenic factors (Kaufmann, 2008; Galve et al., 2009; and Doctor and Doctor, 2012).

Ascertaining the main influencing factors and understanding the nature of their interactions can enable researchers to better understand where and how individual sinkholes may appear. The interactions between influencing factors are frequently not obvious and are often hidden from direct observation, making it unlikely to predict the occurrence of an individual sinkhole at a specific site. Nevertheless, the analysis of the spatial statistical relationships between sinkhole density and the potential influencing factors could help determine the principal causal factors influencing the formation and general spatial distribution and density of sinkholes in a particular area. Identifying the major influencing factors and the interactive

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

251

Kidanu, Anderson, and Rogers

relationships between them should provide an objective means of parameter weighting that would be useful in any Geographic Information System (GIS)–driven model examining sinkhole susceptibility for hazard mapping.

GIS techniques have been employed in various types of geohazard zonation analyses for land use planning and landscape management (Rogers, 1997). GIS spatial data processing and analysis techniques can be used to facilitate handling and processing of large data sets for sinkhole susceptibility modeling and to determine and evaluate the factors influencing the formation of sinkholes. Researchers have used different spatial analytical approaches to model sinkhole susceptibility.

The most commonly used approaches are those that use proximity of neighboring sinkholes (i.e., Drake and Ford, 1972; Magdalene and Alexander, 1995) or sinkhole density (i.e., Brook and Allison, 1986; Orndorff et al., 2000). These approaches have sought to make a qualitative evaluation of the relation between sinkhole occurrence and the primary influencing factors (geologic, geomorphologic, hydrologic, or anthropologic effects). Another approach is a heuristic model in which weights are assigned to the factors that influence sinkhole susceptibility for risk assessment (Kaufmann, 2008). The main limitation of the heuristic approach is the subjectivity related to expert evaluation and the difficulty of reproducing the method for different geologic areas. Galve et al. (2009) found that nearest neighbor and sinkhole density methods performed better than other techniques when identifying areas of sinkhole susceptibility, but those methods do not include sinkhole formation explanatory variables. Their ability to measure the influence of various factors on sinkhole development was limited (Doctor and Doctor, 2012). The methods based on density and proximity may not satisfactorily identify sinkhole alignments; for instance, a sinkhole-prone belt determined by a fracture or a lithologic boundary may be missed in such susceptibility maps. The other classes of susceptibility modeling are probabilistic or statistical methodologies that derive the susceptibility models from the analysis of spatial statistical relationships between known sinkholes and a group of influencing factors (Galve et al., 2009).

Geographically weighted regression (GWR) is a relatively recent and sophisticated method of spatial statistical analysis that seeks to measure spatially varying relationships, such as the influence of controlling factors on sinkhole formation. GWR is a local regression version of the global ordinary least squares (OLS) regression method. GWR can be an effective tool to study spatial data relationships with spatial non-stationarity (Fotheringham et al., 2002). In this research, GIS-based global (OLS) and spatial (GWR) 252

Kidanu, Anderson, and Rogers

multivariate regression methods were applied to evaluate and assess the variables controlling the formation of sinkholes in Greene County. The results suggest that there are seven variables that appear to be the principal sinkhole influencing factors. Moreover, coefficient surface maps for each influencing factor were generated to observe how each relationship between sinkhole occurrence and the influencing factors varied across the study area.

LOCATION AND GEOLOGY OF THE STUDY AREA Greene County is located in southwestern Missouri (Figure 1) and is underlain mainly by Mississippian age Burlington-Keokuk Limestone (Figure 2). This bedrock underlies more than 70% of the county. About 98% of the sinkholes in Greene County are formed on Burlington-Keokuk Limestone bedrock. The study area encompasses about 1,336 km2 .

Burlington-Keokuk Limestone is characterized by layers of limestone interbedded with thin layers of chert and the presence of chert nodules within the limestone layers. The limestone is a light gray, coarsely crystalline, and nearly pure calcite. Uneven dissolution of Burlington-Keokuk Limestone has resulted in highly irregular bedrock-overburden interface (Fellows, 1970) and is characterized by the formation of prominent knobs (pinnacles) of bedrock bounded by deep troughs (grikes or “cutters”) caused by dissolution along fractures.

DATA SETS AND METHODOLOGY Data Sets A set of relevant Environmental Systems Research Institute (ESRI) data sets and digital maps of the study area were gathered from a variety of open sources, including the Missouri Geological Survey GeoSTRAT program (Missouri Department of Natural Resources, 2016), the Missouri Spatial Data Information Service (2016), and the U.S. Department of Agriculture (2016).

Further refinements, processing, and conversions were then made on the gathered data sets using ArcGIS 10.2 to derive a set of variables. The derived variables that were implemented in the multivariate regression modeling are sinkhole density (dependent variable) and a set of potential sinkhole-influencing factors (independent variables). The independent variables consist of geological, geomorphic, hydrogeologic, and anthropogenic raster data sets. The ESRI data sets and digital maps, together with the corresponding derived variables, are summarized in Table 1.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

relationships between them should provide an objective means of parameter weighting that would be useful in any Geographic Information System (GIS)–driven model examining sinkhole susceptibility for hazard mapping.

GIS techniques have been employed in various types of geohazard zonation analyses for land use planning and landscape management (Rogers, 1997). GIS spatial data processing and analysis techniques can be used to facilitate handling and processing of large data sets for sinkhole susceptibility modeling and to determine and evaluate the factors influencing the formation of sinkholes. Researchers have used different spatial analytical approaches to model sinkhole susceptibility.

The most commonly used approaches are those that use proximity of neighboring sinkholes (i.e., Drake and Ford, 1972; Magdalene and Alexander, 1995) or sinkhole density (i.e., Brook and Allison, 1986; Orndorff et al., 2000). These approaches have sought to make a qualitative evaluation of the relation between sinkhole occurrence and the primary influencing factors (geologic, geomorphologic, hydrologic, or anthropologic effects). Another approach is a heuristic model in which weights are assigned to the factors that influence sinkhole susceptibility for risk assessment (Kaufmann, 2008). The main limitation of the heuristic approach is the subjectivity related to expert evaluation and the difficulty of reproducing the method for different geologic areas. Galve et al. (2009) found that nearest neighbor and sinkhole density methods performed better than other techniques when identifying areas of sinkhole susceptibility, but those methods do not include sinkhole formation explanatory variables. Their ability to measure the influence of various factors on sinkhole development was limited (Doctor and Doctor, 2012). The methods based on density and proximity may not satisfactorily identify sinkhole alignments; for instance, a sinkhole-prone belt determined by a fracture or a lithologic boundary may be missed in such susceptibility maps. The other classes of susceptibility modeling are probabilistic or statistical methodologies that derive the susceptibility models from the analysis of spatial statistical relationships between known sinkholes and a group of influencing factors (Galve et al., 2009).

Geographically weighted regression (GWR) is a relatively recent and sophisticated method of spatial statistical analysis that seeks to measure spatially varying relationships, such as the influence of controlling factors on sinkhole formation. GWR is a local regression version of the global ordinary least squares (OLS) regression method. GWR can be an effective tool to study spatial data relationships with spatial non-stationarity (Fotheringham et al., 2002). In this research, GIS-based global (OLS) and spatial (GWR) 252

multivariate regression methods were applied to evaluate and assess the variables controlling the formation of sinkholes in Greene County. The results suggest that there are seven variables that appear to be the principal sinkhole influencing factors. Moreover, coefficient surface maps for each influencing factor were generated to observe how each relationship between sinkhole occurrence and the influencing factors varied across the study area.

LOCATION AND GEOLOGY OF THE STUDY AREA Greene County is located in southwestern Missouri (Figure 1) and is underlain mainly by Mississippian age Burlington-Keokuk Limestone (Figure 2). This bedrock underlies more than 70% of the county. About 98% of the sinkholes in Greene County are formed on Burlington-Keokuk Limestone bedrock. The study area encompasses about 1,336 km2 .

Burlington-Keokuk Limestone is characterized by layers of limestone interbedded with thin layers of chert and the presence of chert nodules within the limestone layers. The limestone is a light gray, coarsely crystalline, and nearly pure calcite. Uneven dissolution of Burlington-Keokuk Limestone has resulted in highly irregular bedrock-overburden interface (Fellows, 1970) and is characterized by the formation of prominent knobs (pinnacles) of bedrock bounded by deep troughs (grikes or “cutters”) caused by dissolution along fractures.

DATA SETS AND METHODOLOGY Data Sets A set of relevant Environmental Systems Research Institute (ESRI) data sets and digital maps of the study area were gathered from a variety of open sources, including the Missouri Geological Survey GeoSTRAT program (Missouri Department of Natural Resources, 2016), the Missouri Spatial Data Information Service (2016), and the U.S. Department of Agriculture (2016).

Further refinements, processing, and conversions were then made on the gathered data sets using ArcGIS 10.2 to derive a set of variables. The derived variables that were implemented in the multivariate regression modeling are sinkhole density (dependent variable) and a set of potential sinkhole-influencing factors (independent variables). The independent variables consist of geological, geomorphic, hydrogeologic, and anthropogenic raster data sets. The ESRI data sets and digital maps, together with the corresponding derived variables, are summarized in Table 1.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Determining Factors in Sinkhole Formation

Determining Factors in Sinkhole Formation

Figure 1. Location map of the study area.

Figure 1. Location map of the study area.

Dependent and Independent Variables The sinkhole data set of the study area was extracted from the publicly available Missouri sinkhole database of the Missouri Geological Survey in June 2016 and shows the point locations of 1,419 sinkholes (Figure 3).

The nearest neighbor analysis (NNA) tool in ArcGIS 10.2 was used to analyze the spatial distribution pattern of known sinkholes to ascertain whether the distribution is random. NNA provides p-values and nearest neighbor ratios as indicators of predictive patterns.

The p-value is the probability that some random process created the observed spatial pattern. If the NNA on the sinkhole distribution shows a clustered pattern, it is very unlikely that the observed pattern is the result of random processes. Rather, it implies that there is an underlying process with a set of controlling factors

responsible for the formation and distribution of the sinkholes in the study area.

The dependent variable used in the OLS and GWR analysis consisted of sinkhole density values of each sinkhole location, extracted from sinkhole density map.

The sinkhole density map was generated using the kernel density tool in ArcGIS 10.2. A buffer size of 2,500 m was ascribed around each sinkhole location and used to calculate the kernel density. This size (2,500 m) was determined by using multi-distance spatial cluster analysis (Ripley’s K-function) tool in ArcGIS, which is useful in assessing possible scale effects that may be influencing spatially clustered sinkhole arrays.

The independent variables are the raster data layers encompassing the potential sinkhole-influencing factors. Twelve independent variables were considered for input into the model. These variables are overburden

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

253

Dependent and Independent Variables The sinkhole data set of the study area was extracted from the publicly available Missouri sinkhole database of the Missouri Geological Survey in June 2016 and shows the point locations of 1,419 sinkholes (Figure 3).

The nearest neighbor analysis (NNA) tool in ArcGIS 10.2 was used to analyze the spatial distribution pattern of known sinkholes to ascertain whether the distribution is random. NNA provides p-values and nearest neighbor ratios as indicators of predictive patterns.

The p-value is the probability that some random process created the observed spatial pattern. If the NNA on the sinkhole distribution shows a clustered pattern, it is very unlikely that the observed pattern is the result of random processes. Rather, it implies that there is an underlying process with a set of controlling factors

responsible for the formation and distribution of the sinkholes in the study area.

The dependent variable used in the OLS and GWR analysis consisted of sinkhole density values of each sinkhole location, extracted from sinkhole density map.

The sinkhole density map was generated using the kernel density tool in ArcGIS 10.2. A buffer size of 2,500 m was ascribed around each sinkhole location and used to calculate the kernel density. This size (2,500 m) was determined by using multi-distance spatial cluster analysis (Ripley’s K-function) tool in ArcGIS, which is useful in assessing possible scale effects that may be influencing spatially clustered sinkhole arrays.

The independent variables are the raster data layers encompassing the potential sinkhole-influencing factors. Twelve independent variables were considered for input into the model. These variables are overburden

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

253

Kidanu, Anderson, and Rogers

Figure 2. Geological and sinkhole locations map of Greene County (ESRI data source: Missouri Department of Natural Resources, 2016).

thickness, depth to groundwater, distance to the nearest geological structure (faults, folds, and other related tectonic structures), distance to the nearest drainage line, distance to the nearest spring, groundwater elevation, ground surface elevation (altitude), bedrock elevation, slope of the existing ground surface, distance to the nearest road, and ground surface curvature (planar and

Kidanu, Anderson, and Rogers

Figure 3. Sinkhole distribution map of the study area (ESRI data source: Missouri Department of Natural Resources, 2016).

Figure 3. Sinkhole distribution map of the study area (ESRI data source: Missouri Department of Natural Resources, 2016).

profile). Some potential influencing factors, including soil type, rate of groundwater drawdown, and geochemical and climatic processes, could not be included in the sinkhole formation model due to a paucity of data, insufficient data form, and model criteria.

profile). Some potential influencing factors, including soil type, rate of groundwater drawdown, and geochemical and climatic processes, could not be included in the sinkhole formation model due to a paucity of data, insufficient data form, and model criteria.

Multivariate Regression Methods Multivariate spatial regression analysis is a statistical technique that can examine, model, and explore spatial relationships among given variables across any designated area. In this research, it was used to evaluate possible relationships between sinkhole density and the

Figure 2. Geological and sinkhole locations map of Greene County (ESRI data source: Missouri Department of Natural Resources, 2016).

thickness, depth to groundwater, distance to the nearest geological structure (faults, folds, and other related tectonic structures), distance to the nearest drainage line, distance to the nearest spring, groundwater elevation, ground surface elevation (altitude), bedrock elevation, slope of the existing ground surface, distance to the nearest road, and ground surface curvature (planar and

Table 1. Gathered ESRI data sets and digital maps, data sources, and the derived variables.

Main Data Set/Map

Data Sources

Sinkhole locations (ESRI data) Bedrock type (ESRI data) Geologic structures (ESRI data) Overburden thickness contour lines (ESRI data) Depth-to-groundwater contour lines (ESRI data) Ground elevation contour lines (ESRI data) Soil type map Drainage lines map Spring locations (ESRI data) Missouri highway/road feature data

254

Missouri Department of Natural Resources (2016) Missouri Department of Natural Resources (2016) Missouri Department of Natural Resources (2016) Missouri Spatial Data Information Service (2016) Missouri Department of Natural Resources (2016) Missouri Spatial Data Information Service (2016) U.S. Department of Agriculture (2016) U.S. Department of Agriculture (2016) Missouri Department of Natural Resources (2016) http://www.mapcruzin.com

Derived Variables Sinkhole density raster Bedrock type map Distance to nearest geological structure raster Overburden thickness raster

Multivariate spatial regression analysis is a statistical technique that can examine, model, and explore spatial relationships among given variables across any designated area. In this research, it was used to evaluate possible relationships between sinkhole density and the

Table 1. Gathered ESRI data sets and digital maps, data sources, and the derived variables.

Cell Size (m × m) 50 × 50 —

Main Data Set/Map

Data Sources

Sinkhole locations (ESRI data) Bedrock type (ESRI data)

50 × 50

Geologic structures (ESRI data)

50 × 50

Overburden thickness contour lines (ESRI data) Depth-to-groundwater contour lines (ESRI data) Ground elevation contour lines (ESRI data)

Depth-to-groundwater raster

50 × 50

Digital elevation model Ground surface slope Curvature (planar and profile) Soil type map Distance to nearest drainage line raster Distance to nearest spring raster

50 × 50 50 × 50 50 × 50 — 50 × 50 50 × 50

Distance to nearest road raster

50 × 50

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Multivariate Regression Methods

Soil type map Drainage lines map Spring locations (ESRI data) Missouri highway/road feature data

254

Missouri Department of Natural Resources (2016) Missouri Department of Natural Resources (2016) Missouri Department of Natural Resources (2016) Missouri Spatial Data Information Service (2016) Missouri Department of Natural Resources (2016) Missouri Spatial Data Information Service (2016) U.S. Department of Agriculture (2016) U.S. Department of Agriculture (2016) Missouri Department of Natural Resources (2016) http://www.mapcruzin.com

Derived Variables Sinkhole density raster Bedrock type map Distance to nearest geological structure raster Overburden thickness raster

Cell Size (m × m) 50 × 50 — 50 × 50 50 × 50

Depth-to-groundwater raster

50 × 50

Digital elevation model Ground surface slope Curvature (planar and profile) Soil type map Distance to nearest drainage line raster Distance to nearest spring raster

50 × 50 50 × 50 50 × 50 — 50 × 50 50 × 50

Distance to nearest road raster

50 × 50

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Determining Factors in Sinkhole Formation

Determining Factors in Sinkhole Formation

physical factors believed to be influencing the sinkhole formation processes at the current time. When dealing with spatial data relationships, regression methods may assume that these relationships are consistent geographically (stationarity) or take into account the spatial locations of features, permitting the estimated parameters to vary locally (non-stationarity). The latter assumption better reflects spatially varying relationships between dependent and independent (explanatory) variables and usually results in improved model performance.

OLS regression is the most commonly employed regression technique and is usually the starting point for all spatial regression analysis (ESRI, 2014a). OLS provides a global model of the variable or process that one needs to understand or predict by creating a single regression equation to represent that process (ESRI, 2014b). The OLS regression model, with k the number of independent variables, is of the form (Charlton and Fotheringham, 2009) y = 0 + 1 x1 + 2 x2+ 3 x3 + . . . + k xk + ε,

where y is the dependent variable, x1 is the explanatory/independent variable, 0 is the intercept of the line on the y-axis, 1 represents the slope coefficient for the independent variable x1 , and ε is a mean zero random error term with constant (but unknown) variance and is normally distributed.

The OLS regression method assumes that the spatial relationships between dependent and independent variables are static and will not be efficient if there exists spatial non-stationarity in the relationships between the variables. When the relationship between variables under study exhibit non-stationarity (spatially varying) behavior, GWR, a local regression technique, is normally preferred.

GWR is one of the most sophisticated applied methodologies for local regression analysis (Brunsdon et al., 1996, 1998; Kalogirou and Hatzichristos, 2007).

GWR allows for local (spatial) variables to be estimated (Fotheringham and Brunsdon, 1999) and for examination of spatial non-stationarity of the factors influencing the formation and distribution of sinkholes. The GWR version of the OLS regression model extends the traditional regression framework by allowing parameters to be estimated locally (Charlton and Fotheringham, 2009) and can be expressed as Yi = 0(ui, vi ) +

d

k=1

k(ui, vi ) Xi k + εi

i = 1, 2, . . . , n,

where Yi is the dependent variable in spatial location with the coordinate (ui , vi ), X1 , X2 , . . .,Xd are explanatory (independent) variables, Xik means the kth explanatory variable in spatial location with coordinate

physical factors believed to be influencing the sinkhole formation processes at the current time. When dealing with spatial data relationships, regression methods may assume that these relationships are consistent geographically (stationarity) or take into account the spatial locations of features, permitting the estimated parameters to vary locally (non-stationarity). The latter assumption better reflects spatially varying relationships between dependent and independent (explanatory) variables and usually results in improved model performance.

OLS regression is the most commonly employed regression technique and is usually the starting point for all spatial regression analysis (ESRI, 2014a). OLS provides a global model of the variable or process that one needs to understand or predict by creating a single regression equation to represent that process (ESRI, 2014b). The OLS regression model, with k the number of independent variables, is of the form (Charlton and Fotheringham, 2009) Figure 4. Flowchart showing the procedures and methods used in the study.

(ui , vi ), 0 (ui , vi ) represents the intercept value, and k (ui , vi ) is a set of values of coefficients at spatial location i.

Several researchers have used GWR to model spatially varying relationships or processes. Some examples include (1) the exploration of the relations between riverbank erosion and geomorphological controls (Atkinson et al., 2003), (2) analysis of the relationship between geologic and hydrologic features and sinkhole occurrence (Doctor and Doctor, 2012), (3) the spatial simulation of regional land use patterns (Liao et al., 2010), (4) assessing risk factors for malaria hot spots (Ndiath et al., 2015), (5) assessment of land subsidence potential (Blachowski, 2016), (6) landslide susceptibility mapping (Arzu and Sebnem, 2010), and (7) the exploration of spatial non-stationarity of fisheries survey data (Windle et al., 2010). In this research OLS followed by GWR were employed to analyze the influencing factors for the formation and distribution of sinkholes across the study area. A flowchart outlining the procedures and methods used in this study are summarized in Figure 4.

where y is the dependent variable, x1 is the explanatory/independent variable, 0 is the intercept of the line on the y-axis, 1 represents the slope coefficient for the independent variable x1 , and ε is a mean zero random error term with constant (but unknown) variance and is normally distributed.

The OLS regression method assumes that the spatial relationships between dependent and independent variables are static and will not be efficient if there exists spatial non-stationarity in the relationships between the variables. When the relationship between variables under study exhibit non-stationarity (spatially varying) behavior, GWR, a local regression technique, is normally preferred.

GWR is one of the most sophisticated applied methodologies for local regression analysis (Brunsdon et al., 1996, 1998; Kalogirou and Hatzichristos, 2007).

GWR allows for local (spatial) variables to be estimated (Fotheringham and Brunsdon, 1999) and for examination of spatial non-stationarity of the factors influencing the formation and distribution of sinkholes. The GWR version of the OLS regression model extends the traditional regression framework by allowing parameters to be estimated locally (Charlton and Fotheringham, 2009) and can be expressed as Yi = 0(ui, vi ) +

RESULTS AND DISCUSSIONS Sinkhole Density and Cluster Analysis Several authors (e.g., Zhou et al., 2003; Brezinski, 2004) have mentioned that in areas where active sinkholes have developed, there is a greater chance that

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

y = 0 + 1 x1 + 2 x2+ 3 x3 + . . . + k xk + ε,

255

d

k=1

k(ui, vi ) Xi k + εi

i = 1, 2, . . . , n,

where Yi is the dependent variable in spatial location with the coordinate (ui , vi ), X1 , X2 , . . .,Xd are explanatory (independent) variables, Xik means the kth explanatory variable in spatial location with coordinate

Figure 4. Flowchart showing the procedures and methods used in the study.

(ui , vi ), 0 (ui , vi ) represents the intercept value, and k (ui , vi ) is a set of values of coefficients at spatial location i.

Several researchers have used GWR to model spatially varying relationships or processes. Some examples include (1) the exploration of the relations between riverbank erosion and geomorphological controls (Atkinson et al., 2003), (2) analysis of the relationship between geologic and hydrologic features and sinkhole occurrence (Doctor and Doctor, 2012), (3) the spatial simulation of regional land use patterns (Liao et al., 2010), (4) assessing risk factors for malaria hot spots (Ndiath et al., 2015), (5) assessment of land subsidence potential (Blachowski, 2016), (6) landslide susceptibility mapping (Arzu and Sebnem, 2010), and (7) the exploration of spatial non-stationarity of fisheries survey data (Windle et al., 2010). In this research OLS followed by GWR were employed to analyze the influencing factors for the formation and distribution of sinkholes across the study area. A flowchart outlining the procedures and methods used in this study are summarized in Figure 4.

RESULTS AND DISCUSSIONS Sinkhole Density and Cluster Analysis Several authors (e.g., Zhou et al., 2003; Brezinski, 2004) have mentioned that in areas where active sinkholes have developed, there is a greater chance that

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

255

Kidanu, Anderson, and Rogers

Figure 5. Map showing sinkhole locations (dots) and densities (number of sinkholes per km2 ).

new sinkholes will form. Therefore, sinkhole density is an important factor in determining the areas most prone to sinkhole development. Kemmerly (1982) has asserted that cluster analysis may be applied to evaluate whether the generation of new sinkholes is influenced by the location of the pre-existing sinkhole population.

Other authors (Hyatt et al., 1999; Gutiérrez-Santolalla et al., 2005) have studied sinkhole distributions centered on manipulation of statistical values for prediction of future sinkholes. In this research, a sinkhole density data set (Figure 5) was generated using the kernel density tool in ArcGIS along with cluster analysis using the NNA tool to see whether any spatial patterns of sinkholes were discernable. The NNA result (Figure 6) suggests that the sinkholes are significantly clustered, with a p-value of 0.000000 and a nearest neighbor ratio (observed mean distance/expected mean distance) of 0.52. The significant clustering implies that it is very unlikely that some random process created the observed distributions. Rather, the clustering indicates that there is an underlying process with a set of key influencing factors that is likely responsible for the formation and distribution of the sinkholes in the study area.

After ascertaining the clustered nature of sinkhole distribution, the next logical question was what are the main factors controlling this observed clustered pattern? OLS followed by GWR analysis was employed to explore the spatial relationship between sinkhole density and the explanatory variables so that we could extract the significant controlling variables.

OLS Model Results As mentioned previously, 12 potential sinkhole influencing factors were considered in the OLS analysis.

A series of model checks were performed to evaluate 256

the reliability of the OLS regression model. According to the robust probability significant test results, ground surface curvature (both planar and profile), elevation to top of rock, and groundwater elevation were not significantly correlated with the dependent variable (sinkhole density) and were therefore removed from the model.

Another test was a multicollinearity test in which the variance inflation factor (VIF) was employed to make sure that none of the explanatory variables were redundant. The rule of thumb for interpreting VIF values was that they should be less than 7.5, with smaller values representing better correlations. Variables with VIF values greater than 7.5 are generally removed from the model. The test results showed that the VIF values of all the variables were less than 7.5, except for ground surface elevation (altitude), elevation to top of rock, and groundwater elevation, so the ground surface elevation (altitude) variable was also removed from the model. After performing all these model tests, seven of the 12 variables were selected as significant explanatory variables likely influencing the formation of sinkholes in the study area. These variables are overburden thickness, distance to the nearest drainage line, depth to groundwater, slope of the ground surface, distance to the nearest geological structure, distance to the nearest road, and distance to the nearest spring (Table 2), and their thematic maps are presented in Figure 7.

The explanatory variables were selected on the basis of exhibiting robust probability statistics with low VIF values in the range of 1.08 to 4.0 (Table 2). They also exhibit theoretically justifiable coefficient signs on a global scale. Overburden thickness, distance to the nearest road, slope of the ground surface, and distance to the nearest geological structure exhibit negative correlations with the occurrence of sinkholes, suggesting that, for example, areas closer to geological structures have a higher incidence of sinkhole occurrence than areas farther away. The remaining factors, depth to groundwater, distance to the nearest spring, and distance to the nearest drainage line exhibit positive coefficient signs.

The result from the OLS model showed that the adjusted R2 value is 0.570 and that the Akaike information criterion (AIC) value is 4853. This suggests that the OLS global model can explain about 57 percent (adjusted R2 = 0.570) of the variation in sinkhole density, with AIC = 4,853. The adjusted R2 and the AICc are statistics derived from the regression equation to quantify model performance (ESRI, 2014b).

Analysis of variance returned a statistically significant F-statistic value of 157.84, and the Wald statistic has a significant chi-square value of 2,443.37. These results indicate that the model formulation was statistically significant. The Jarque–Bera statistic returned a nonsignificant chi-square value of 3.42, indicating that the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Kidanu, Anderson, and Rogers

Figure 5. Map showing sinkhole locations (dots) and densities (number of sinkholes per km2 ).

new sinkholes will form. Therefore, sinkhole density is an important factor in determining the areas most prone to sinkhole development. Kemmerly (1982) has asserted that cluster analysis may be applied to evaluate whether the generation of new sinkholes is influenced by the location of the pre-existing sinkhole population.

Other authors (Hyatt et al., 1999; Gutiérrez-Santolalla et al., 2005) have studied sinkhole distributions centered on manipulation of statistical values for prediction of future sinkholes. In this research, a sinkhole density data set (Figure 5) was generated using the kernel density tool in ArcGIS along with cluster analysis using the NNA tool to see whether any spatial patterns of sinkholes were discernable. The NNA result (Figure 6) suggests that the sinkholes are significantly clustered, with a p-value of 0.000000 and a nearest neighbor ratio (observed mean distance/expected mean distance) of 0.52. The significant clustering implies that it is very unlikely that some random process created the observed distributions. Rather, the clustering indicates that there is an underlying process with a set of key influencing factors that is likely responsible for the formation and distribution of the sinkholes in the study area.

After ascertaining the clustered nature of sinkhole distribution, the next logical question was what are the main factors controlling this observed clustered pattern? OLS followed by GWR analysis was employed to explore the spatial relationship between sinkhole density and the explanatory variables so that we could extract the significant controlling variables.

OLS Model Results As mentioned previously, 12 potential sinkhole influencing factors were considered in the OLS analysis.

A series of model checks were performed to evaluate 256

the reliability of the OLS regression model. According to the robust probability significant test results, ground surface curvature (both planar and profile), elevation to top of rock, and groundwater elevation were not significantly correlated with the dependent variable (sinkhole density) and were therefore removed from the model.

Another test was a multicollinearity test in which the variance inflation factor (VIF) was employed to make sure that none of the explanatory variables were redundant. The rule of thumb for interpreting VIF values was that they should be less than 7.5, with smaller values representing better correlations. Variables with VIF values greater than 7.5 are generally removed from the model. The test results showed that the VIF values of all the variables were less than 7.5, except for ground surface elevation (altitude), elevation to top of rock, and groundwater elevation, so the ground surface elevation (altitude) variable was also removed from the model. After performing all these model tests, seven of the 12 variables were selected as significant explanatory variables likely influencing the formation of sinkholes in the study area. These variables are overburden thickness, distance to the nearest drainage line, depth to groundwater, slope of the ground surface, distance to the nearest geological structure, distance to the nearest road, and distance to the nearest spring (Table 2), and their thematic maps are presented in Figure 7.

The explanatory variables were selected on the basis of exhibiting robust probability statistics with low VIF values in the range of 1.08 to 4.0 (Table 2). They also exhibit theoretically justifiable coefficient signs on a global scale. Overburden thickness, distance to the nearest road, slope of the ground surface, and distance to the nearest geological structure exhibit negative correlations with the occurrence of sinkholes, suggesting that, for example, areas closer to geological structures have a higher incidence of sinkhole occurrence than areas farther away. The remaining factors, depth to groundwater, distance to the nearest spring, and distance to the nearest drainage line exhibit positive coefficient signs.

The result from the OLS model showed that the adjusted R2 value is 0.570 and that the Akaike information criterion (AIC) value is 4853. This suggests that the OLS global model can explain about 57 percent (adjusted R2 = 0.570) of the variation in sinkhole density, with AIC = 4,853. The adjusted R2 and the AICc are statistics derived from the regression equation to quantify model performance (ESRI, 2014b).

Analysis of variance returned a statistically significant F-statistic value of 157.84, and the Wald statistic has a significant chi-square value of 2,443.37. These results indicate that the model formulation was statistically significant. The Jarque–Bera statistic returned a nonsignificant chi-square value of 3.42, indicating that the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Determining Factors in Sinkhole Formation

Determining Factors in Sinkhole Formation

Figure 6. Results of NNA showing the sinkhole distribution pattern.

Figure 6. Results of NNA showing the sinkhole distribution pattern.

model’s prediction is free from bias (i.e., the residuals have a normal distribution). All of these diagnostic tests suggests a fairly strong model, although one statistic, the Koenker test, was found to be statistically signif-

icant, indicating that the relationship between some or perhaps all of the explanatory variables and the dependent variable is non-stationary (spatially varying) across the study area. The reason for this is that

model’s prediction is free from bias (i.e., the residuals have a normal distribution). All of these diagnostic tests suggests a fairly strong model, although one statistic, the Koenker test, was found to be statistically signif-

Table 2. Summary statistics for OLS (significant at ** 0.01% level; * 5% level).

Variables Overburden thickness Distance to the nearest drainage line Depth to groundwater Distance to the nearest road Slope of the ground surface Distance to the nearest geological structure Distance to the nearest spring

Table 2. Summary statistics for OLS (significant at ** 0.01% level; * 5% level).

Coefficient Values

Robust_Pr [b]

VIF

−0.044726 0.002405 0.023164 −0.000845 −0.153766 −0.000041 0.000207

0.00** 0.00** 0.00** 0.00** 0.00** 0.040* 0.00**

2.192135 1.321380 4.004964 1.086326 1.270746 1.156041 2.301522

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

icant, indicating that the relationship between some or perhaps all of the explanatory variables and the dependent variable is non-stationary (spatially varying) across the study area. The reason for this is that

257

Variables Overburden thickness Distance to the nearest drainage line Depth to groundwater Distance to the nearest road Slope of the ground surface Distance to the nearest geological structure Distance to the nearest spring

Coefficient Values

Robust_Pr [b]

VIF

−0.044726 0.002405 0.023164 −0.000845 −0.153766 −0.000041 0.000207

0.00** 0.00** 0.00** 0.00** 0.00** 0.040* 0.00**

2.192135 1.321380 4.004964 1.086326 1.270746 1.156041 2.301522

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

257

258

Kidanu, Anderson, and Rogers

Kidanu, Anderson, and Rogers

Figure 7. Thematic maps of seven independent variables in our regression (OLS and GWR) model.

Figure 7. Thematic maps of seven independent variables in our regression (OLS and GWR) model.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

258

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Determining Factors in Sinkhole Formation

Determining Factors in Sinkhole Formation

Figure 8. Slope and depth-to-groundwater coefficient surface maps derived from the GWR analysis.

Figure 8. Slope and depth-to-groundwater coefficient surface maps derived from the GWR analysis.

some explanatory variables may be important for predicting the formation of the sinkhole in some locations but not in other areas. The spatial autocorrelation test run on the OLS model’s residuals with Moran’s I tool exhibited a clustered pattern. The presence of spatially clustered residuals, as well as the statistically significant value of the Koenker statistic in the OLS model, suggests the presence of spatial non-stationarity in the data. This supports the premise that a local regression method can better explain the process than can a global regression model (OLS). For these reasons, GWR was applied, and it is evident that the model’s fitness will likely be improved by incorporating GWR, which takes into account the spatial variability of factors.

GWR Model Results As mentioned earlier, OLS appears to be the best starting point in building GWR models, so the GWR model was run using the same dependent variable and seven independent variables selected from the OLS analysis (Table 2).

The results of the GWR analysis showed that 86 percent (adjusted R2 = 0.8557) of the variance in sinkhole density can be explained by the model, which is much higher than that of OLS (57 percent). The AICc value for the GWR model was 758, whereas that derived from the OLS was 4,853. Greater adjusted R2 and smaller AICc values indicate that the GWR model (local regression) is superior to the OLS model (global regression) and has captured the spatial non-stationarity of variables.

GWR calculates different regression parameter values (e.g., coefficient) for each cell that can be mapped so that the spatial variations of parameters can be examined and observed visually. Coefficient surface maps for each explanatory variable were generated to ascer-

tain how the relationship between sinkhole occurrence and the influencing factors varies across the study area.

For example, coefficient maps of two variables (slope and depth to groundwater) are shown in Figure 8.

These maps help us understand which of the influencing factors were most important in the sinkhole formation process and how the relations vary spatially. For instance, results derived from the global OLS model indicated that the slope variable has a negative relationship with sinkhole occurrence across the study area; however, according to the GWR analysis, the contribution of this variable to sinkhole occurrence spatially varies across the study area, with coefficients ranging from −0.44 to 0.23. The range of coefficient values suggests that the nature (positive or negative) and strength of the relationship vary spatially across the study area (Figure 8). Similarly, the coefficients of the other variables also vary across the study area, including depth to groundwater (−0.02 to 0.06), overburden thickness (−1.66 to 0.11), distance to the nearest spring (−0.00040 to 0.0014), distance to the nearest geological structures (−0.00026 to 0.00050), distance to the nearest drainage line (0.0000027 to 0.0033), and distance to the nearest road (−0.0018 to 0.0012).

CONCLUSIONS AND RECOMMENDATIONS Burlington-Keokuk Limestone bedrock underlies more than 70 percent of Greene County and 98 percent of the identified sinkholes in the county (Missouri Department of Natural Resources, 2016) formed in this unit. Analysis of the sinkholes’ spatial distribution and patterns suggests that the sinkholes are not randomly distributed but are spatially clustered. This implies that there is a process controlled by a finite set of factors that promote the formation and development of karst sinkholes.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

259

some explanatory variables may be important for predicting the formation of the sinkhole in some locations but not in other areas. The spatial autocorrelation test run on the OLS model’s residuals with Moran’s I tool exhibited a clustered pattern. The presence of spatially clustered residuals, as well as the statistically significant value of the Koenker statistic in the OLS model, suggests the presence of spatial non-stationarity in the data. This supports the premise that a local regression method can better explain the process than can a global regression model (OLS). For these reasons, GWR was applied, and it is evident that the model’s fitness will likely be improved by incorporating GWR, which takes into account the spatial variability of factors.

GWR Model Results As mentioned earlier, OLS appears to be the best starting point in building GWR models, so the GWR model was run using the same dependent variable and seven independent variables selected from the OLS analysis (Table 2).

The results of the GWR analysis showed that 86 percent (adjusted R2 = 0.8557) of the variance in sinkhole density can be explained by the model, which is much higher than that of OLS (57 percent). The AICc value for the GWR model was 758, whereas that derived from the OLS was 4,853. Greater adjusted R2 and smaller AICc values indicate that the GWR model (local regression) is superior to the OLS model (global regression) and has captured the spatial non-stationarity of variables.

GWR calculates different regression parameter values (e.g., coefficient) for each cell that can be mapped so that the spatial variations of parameters can be examined and observed visually. Coefficient surface maps for each explanatory variable were generated to ascer-

tain how the relationship between sinkhole occurrence and the influencing factors varies across the study area.

For example, coefficient maps of two variables (slope and depth to groundwater) are shown in Figure 8.

These maps help us understand which of the influencing factors were most important in the sinkhole formation process and how the relations vary spatially. For instance, results derived from the global OLS model indicated that the slope variable has a negative relationship with sinkhole occurrence across the study area; however, according to the GWR analysis, the contribution of this variable to sinkhole occurrence spatially varies across the study area, with coefficients ranging from −0.44 to 0.23. The range of coefficient values suggests that the nature (positive or negative) and strength of the relationship vary spatially across the study area (Figure 8). Similarly, the coefficients of the other variables also vary across the study area, including depth to groundwater (−0.02 to 0.06), overburden thickness (−1.66 to 0.11), distance to the nearest spring (−0.00040 to 0.0014), distance to the nearest geological structures (−0.00026 to 0.00050), distance to the nearest drainage line (0.0000027 to 0.0033), and distance to the nearest road (−0.0018 to 0.0012).

CONCLUSIONS AND RECOMMENDATIONS Burlington-Keokuk Limestone bedrock underlies more than 70 percent of Greene County and 98 percent of the identified sinkholes in the county (Missouri Department of Natural Resources, 2016) formed in this unit. Analysis of the sinkholes’ spatial distribution and patterns suggests that the sinkholes are not randomly distributed but are spatially clustered. This implies that there is a process controlled by a finite set of factors that promote the formation and development of karst sinkholes.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

259

Kidanu, Anderson, and Rogers

In this study, GIS-based multivariate regression methods (OLS and GWR) were applied to evaluate the spatial relationships between potential sinkholeinfluencing factors (explanatory variables) and sinkhole density (dependent variable) with the aim of evaluating the significant controlling factors. The OLS analysis revealed that seven of the 12 possible influencing factors considered in the analysis likely play important roles in triggering the formation of sinkholes.

These factors are overburden thickness, slope of the ground surface, depth to groundwater, distance to the nearest drainage line, distance to the nearest road, distance to the nearest geological structure, and distance to the nearest spring.

The OLS results also indicated that the relationship between some or perhaps all of the explanatory variables and the dependent variable is non-stationary across the study area. Hence, GWR emerged as being more appropriate for analyzing those relationships because it has the capability of capturing the spatial non-stationarity of the influencing factors. GWR improved the model and explained 86 percent (better than OLS = 57 percent) of the sinkhole density variability. The GWR model coefficient values for each explanatory variable provide visual insight into the influence of these variables on localized sinkhole density and patterns, and the values can be used to provide an objective means of parameter weighting in models of sinkhole susceptibility or hazard mapping/zoning.

Due to a paucity of data, insufficient data form, and model criteria, there are some potential influencing factors that were not included in the model (this may include falling or rising depth to groundwater, soil type, geochemical processes, and so on). The OLS and GWR models were able to explain only 57 and 86 percent of the processes responsible for the formation of mapped sinkholes, respectively. Therefore, further research incorporating more data with better resolution is recommended to improve the model.

REFERENCES ARZU, E. H. AND SEBNEM, B. D., 2010, Improvement of statistical landslide susceptibility mapping by using spatial and global regression methods in the case of More and Romsdal (Norway): Landslides, Vol. 7, pp. 55–68. doi:10.1007/s10346-009-0188-x.

ATKINSON, P. M.; GERMAN, S. E.; SEAR, D. A.; AND CLARK, M. J., 2003, Exploring the relations between river bank erosion and geomorphological controls using geographically weighted logistic regression: Geographical Analysis, Vol. 35, No. 1, pp. 58–82.

BLACHOWSKI., J., 2016, Application of GIS spatial regression methods in assessment of land subsidence in complicated mining conditions: Case study of the Walbrzych coal mine (SW Poland): Natural Hazards, Vol. 84, No. 2, pp. 997–1014. doi:10.1007/s11069-016-2470-2.

260

BREZINSKI, D. K., 2004, Stratigraphy of the Frederick Valley and Its Relationship to Karst: Report No. 75, Maryland Department of Natural Resources, Annapolis.

BROOK, G. A. AND ALLISON, T. L., 1986, Fracture mapping and ground susceptibility modeling in covered karst terrain—The example of Dougherty County, Georgia. In Land Subsidence: IAHSAISH Publication, Vol. 151, pp. 595–606.

BRUNSDON, C.; CHARLTON, M.; AND FOTHERINGHAM, S., 1996, Geographically weighted regression: A method for exploring spatial nonstationarity: Geographical Analysis, Vol. 28, no. 4, pp. 281–298.

BRUNSDON, C.; FOTHERINGHAM, S.; AND CHARLTON, M., 1998, Geographically weighted regression—Modelling spatial non-stationarity: The Statistician, Vol. 47, pp. 431–443. doi:10.1111/1467-9884.00145.

CARBONEL, D.; RODRÍGUEZ, V.; GUTIÉRREZ, F.; MCCALPIN, J. P.; LINARES, R.; ROQUE, C.; ZARROCA, M.; GUERRERO, J.; AND SASOWSKY, I., 2014, Evaluation of trenching, ground penetrating radar (GPR) and electrical resistivity tomography (ERT) for sinkhole characterization: Earth Surface Processes and Landforms, Vol. 39, pp. 214–227.

CHALIKAKIS, K.; PLAGNES, V.; GUERIN, R.; VALOIS, R.; AND BOSCH, F. P., 2011, Contribution of geophysical methods to karst-system exploration: An overview: Hydrogeology Journal, Vol. 19, pp. 1169–1180.

CHARLTON, M. AND FOTHERINGHAM, S., 2009, Geographically Weighted Regression: White paper, National Centre for Geocomputation, National University of Ireland Maynooth.

DOCTOR, D. AND DOCTOR, K., 2012, Spatial analysis of geologic and hydrologic features relating to sinkhole occurrence in Jefferson County, West Virginia: Carbonates and Evaporites, Vol. 27, No. 2, pp. 143–152.

DRAKE, J. J. AND FORD, D. C., 1972, The analysis of growth patterns of two generation populations—The examples of karst sinkholes: The Canadian Geographer, Vol. 16, pp. 381–384.

ENVIRONMENTAL SYSTEMS RESOURCE INSTITUTE (ESRI), 2014a, How OLS Regression Works: Electronic document, available at http://resources.arcgis.com/en/help/main ENVIRONMENTAL SYSTEMS RESOURCE INSTITUTE (ESRI), 2014b, Interpreting OLS Results: Electronic document, available at http://resources.arcgis.com/en/help/main FELLOWS, L. D., 1970, Geology of Galloway Quadrangle Greene County Missouri: Missouri Geological Survey and Water Resources, pp. 3–14.

FOTHERINGHAM, A. S. AND BRUNSDON, C., 1999, Local forms of spatial analysis: Geographical Analysis, Vol. 31, No. 4, pp. 340–358.

FOTHERINGHAM, A. S.; BRUNSDON, C.; AND CHARLTON, M., 2002, Geographically Weighted Regression—The Analysis of Spatially Varying Relationships: Wiley, Chichester.

GALVE, P.; GUTIÉRREZ, F.; REMONDO, J.; BONACHEA, J.; LUCHA, P.; AND CENDRERO, A., 2009, Evaluating and comparing methods of sinkhole susceptibility mapping in the Ebro Valley evaporite karst (NE Spain): Geomorphology, Vol. 111, pp. 160–172.

GUTIÉRREZ, F., 2010, Hazards associated with karst. In AlcántaraAyala, I. and Goudie, A. (Editors), Geomorphological Hazards and Disaster Prevention: Cambridge University Press, Cambridge, pp. 161–173.

GUTIÉRREZ-SANTOLALLA, F.; GUTIÉRREZ-ELORZA, M.; MARÍN, C.; DESIR, G.; AND MALDONADO, C., 2005, Spatial distribution, morphometry and activity of La Puebla de Alfindén sinkhole field in the Ebro River valley (NE Spain) applied aspects for hazard zonation: Environmental Geology, Vol. 48, pp. 360–369.

HYATT, J.; WILKES, H.; AND JACOBS, P., 1999. Spatial relationship between new and old sinkholes in covered karst, Albany, Georgia, USA. In Beck, B. F. (Editor), Hydrogeology and

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Kidanu, Anderson, and Rogers

In this study, GIS-based multivariate regression methods (OLS and GWR) were applied to evaluate the spatial relationships between potential sinkholeinfluencing factors (explanatory variables) and sinkhole density (dependent variable) with the aim of evaluating the significant controlling factors. The OLS analysis revealed that seven of the 12 possible influencing factors considered in the analysis likely play important roles in triggering the formation of sinkholes.

These factors are overburden thickness, slope of the ground surface, depth to groundwater, distance to the nearest drainage line, distance to the nearest road, distance to the nearest geological structure, and distance to the nearest spring.

The OLS results also indicated that the relationship between some or perhaps all of the explanatory variables and the dependent variable is non-stationary across the study area. Hence, GWR emerged as being more appropriate for analyzing those relationships because it has the capability of capturing the spatial non-stationarity of the influencing factors. GWR improved the model and explained 86 percent (better than OLS = 57 percent) of the sinkhole density variability. The GWR model coefficient values for each explanatory variable provide visual insight into the influence of these variables on localized sinkhole density and patterns, and the values can be used to provide an objective means of parameter weighting in models of sinkhole susceptibility or hazard mapping/zoning.

Due to a paucity of data, insufficient data form, and model criteria, there are some potential influencing factors that were not included in the model (this may include falling or rising depth to groundwater, soil type, geochemical processes, and so on). The OLS and GWR models were able to explain only 57 and 86 percent of the processes responsible for the formation of mapped sinkholes, respectively. Therefore, further research incorporating more data with better resolution is recommended to improve the model.

REFERENCES ARZU, E. H. AND SEBNEM, B. D., 2010, Improvement of statistical landslide susceptibility mapping by using spatial and global regression methods in the case of More and Romsdal (Norway): Landslides, Vol. 7, pp. 55–68. doi:10.1007/s10346-009-0188-x.

ATKINSON, P. M.; GERMAN, S. E.; SEAR, D. A.; AND CLARK, M. J., 2003, Exploring the relations between river bank erosion and geomorphological controls using geographically weighted logistic regression: Geographical Analysis, Vol. 35, No. 1, pp. 58–82.

BLACHOWSKI., J., 2016, Application of GIS spatial regression methods in assessment of land subsidence in complicated mining conditions: Case study of the Walbrzych coal mine (SW Poland): Natural Hazards, Vol. 84, No. 2, pp. 997–1014. doi:10.1007/s11069-016-2470-2.

260

BREZINSKI, D. K., 2004, Stratigraphy of the Frederick Valley and Its Relationship to Karst: Report No. 75, Maryland Department of Natural Resources, Annapolis.

BROOK, G. A. AND ALLISON, T. L., 1986, Fracture mapping and ground susceptibility modeling in covered karst terrain—The example of Dougherty County, Georgia. In Land Subsidence: IAHSAISH Publication, Vol. 151, pp. 595–606.

BRUNSDON, C.; CHARLTON, M.; AND FOTHERINGHAM, S., 1996, Geographically weighted regression: A method for exploring spatial nonstationarity: Geographical Analysis, Vol. 28, no. 4, pp. 281–298.

BRUNSDON, C.; FOTHERINGHAM, S.; AND CHARLTON, M., 1998, Geographically weighted regression—Modelling spatial non-stationarity: The Statistician, Vol. 47, pp. 431–443. doi:10.1111/1467-9884.00145.

CARBONEL, D.; RODRÍGUEZ, V.; GUTIÉRREZ, F.; MCCALPIN, J. P.; LINARES, R.; ROQUE, C.; ZARROCA, M.; GUERRERO, J.; AND SASOWSKY, I., 2014, Evaluation of trenching, ground penetrating radar (GPR) and electrical resistivity tomography (ERT) for sinkhole characterization: Earth Surface Processes and Landforms, Vol. 39, pp. 214–227.

CHALIKAKIS, K.; PLAGNES, V.; GUERIN, R.; VALOIS, R.; AND BOSCH, F. P., 2011, Contribution of geophysical methods to karst-system exploration: An overview: Hydrogeology Journal, Vol. 19, pp. 1169–1180.

CHARLTON, M. AND FOTHERINGHAM, S., 2009, Geographically Weighted Regression: White paper, National Centre for Geocomputation, National University of Ireland Maynooth.

DOCTOR, D. AND DOCTOR, K., 2012, Spatial analysis of geologic and hydrologic features relating to sinkhole occurrence in Jefferson County, West Virginia: Carbonates and Evaporites, Vol. 27, No. 2, pp. 143–152.

DRAKE, J. J. AND FORD, D. C., 1972, The analysis of growth patterns of two generation populations—The examples of karst sinkholes: The Canadian Geographer, Vol. 16, pp. 381–384.

ENVIRONMENTAL SYSTEMS RESOURCE INSTITUTE (ESRI), 2014a, How OLS Regression Works: Electronic document, available at http://resources.arcgis.com/en/help/main ENVIRONMENTAL SYSTEMS RESOURCE INSTITUTE (ESRI), 2014b, Interpreting OLS Results: Electronic document, available at http://resources.arcgis.com/en/help/main FELLOWS, L. D., 1970, Geology of Galloway Quadrangle Greene County Missouri: Missouri Geological Survey and Water Resources, pp. 3–14.

FOTHERINGHAM, A. S. AND BRUNSDON, C., 1999, Local forms of spatial analysis: Geographical Analysis, Vol. 31, No. 4, pp. 340–358.

FOTHERINGHAM, A. S.; BRUNSDON, C.; AND CHARLTON, M., 2002, Geographically Weighted Regression—The Analysis of Spatially Varying Relationships: Wiley, Chichester.

GALVE, P.; GUTIÉRREZ, F.; REMONDO, J.; BONACHEA, J.; LUCHA, P.; AND CENDRERO, A., 2009, Evaluating and comparing methods of sinkhole susceptibility mapping in the Ebro Valley evaporite karst (NE Spain): Geomorphology, Vol. 111, pp. 160–172.

GUTIÉRREZ, F., 2010, Hazards associated with karst. In AlcántaraAyala, I. and Goudie, A. (Editors), Geomorphological Hazards and Disaster Prevention: Cambridge University Press, Cambridge, pp. 161–173.

GUTIÉRREZ-SANTOLALLA, F.; GUTIÉRREZ-ELORZA, M.; MARÍN, C.; DESIR, G.; AND MALDONADO, C., 2005, Spatial distribution, morphometry and activity of La Puebla de Alfindén sinkhole field in the Ebro River valley (NE Spain) applied aspects for hazard zonation: Environmental Geology, Vol. 48, pp. 360–369.

HYATT, J.; WILKES, H.; AND JACOBS, P., 1999. Spatial relationship between new and old sinkholes in covered karst, Albany, Georgia, USA. In Beck, B. F. (Editor), Hydrogeology and

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

Determining Factors in Sinkhole Formation Engineering Geology of Sinkholes and Karst: Proceedings of the 7th Multidisciplinary Conference on Sinkholes and the Engineering and Environmental Impacts of Karst: A.A. Balkema Harrisburg, PA, pp. 37–44.

KALOGIROU, S. AND HATZICHRISTOS, T., 2007, A spatial modelling framework for income estimation: Spatial Economic Analysis, Vol. 2, No. 3, pp. 297–316.

KAUFMANN, J., 2008, A statistical approach to karst collapse hazard analysis in Missouri: In Yuhr, L. B.; Alexander, E. Calvin, Jr.; and Beck, Barry F. (Editors), Proceedings of the 11th Multidisciplinary Conference on Sinkholes and the Engineering and Environmental Impacts of Karst: American Society of Civil Engineers, Reston, VA, pp. 257–268.

KEMMERLY, P. R., 1982. Spatial analysis of a karst depression population: Clues to genesis. Geological Society of America Bulletin, Vol. 93, pp. 1078–1086.

LIAO, Q.; LI, M.; CHEN, Z.; SHAO, Y.; AND YANG, K., 2010, Spatial simulation of regional land use patterns based on GWR and CLUE-S model: In XXXX (Editor), Proceedings of the 18th International Conference on Geoinformatics: Yu Liu and Aijun Chen, IEEE, Beijing, China pp. 1–6, doi:10.1109/GEOINFORMATICS.2010.5567963.

MISSOURI DEPARTMENT OF NATURAL RESOURCES, 2016, Missouri Geological Survey Geosciences Technical Resource Assessment Tool (GeoSTRAT): Electronic geospatial data, available at https://dnr.mo.gov MISSOURI SPATIAL DATA INFORMATION SERVICE, 2016, Electronic geospatial data, available at http://msdis.missouri.edu MAGDALENE, S. AND ALEXANDER, E.C., 1995, Sinkhole distribution in Winona County, Minnesota revisited: in: Beck, Barry F. and Person, Felicity M. eds., Karst Geohazards, Proceedings of the Fifth Multidisciplinary Conference on Sinkholes and the Engineering and Environmental Impact of Karst:

Determining Factors in Sinkhole Formation

Gatlinburg, Tenn., 2–5 April, 1995, A.A. Balkema, Rotterdam, p. 43–51.

NDIATH, M. N.; CISSE, B.; NDIAYE, J. L.; GOMIS, J. F.; BATHIERY, O.; DIA, A. T.; GAYE, O.; AND FAYE, B., 2015, Application of geographically-weighted regression analysis to assess risk factors for malaria hotspots in Keur Soce health and demographic surveillance site: Malaria Journal, Vol. 14, p. 463.

ORNDORFF, R. C.; WEARY, D. J.; AND LAGUEUX, K. M., 2000, Geographic information systems analysis of geologic controls on the distribution of dolines in the Ozarks of southcentral Missouri, USA: Acta Carsologica, Vol. 29, No. 2, pp. 161–175.

ROGERS, J. D., 1997, Spatial Geologic Hazard Analysis in Practice: Spatial Analysis in Soil Dynamics and Earthquake Engineering: ASCE Geotechnical Special Publication 67, pp. 15–28.

UNITED STATES DEPARTMENT OF AGRICULTURE (USDA), 2016, Natural Resources Conservation Service: Electronic geospatial data, available at https://websoilsurvey. sc.egov.usda.gov/App/WebSoilSurvey.aspx WALTHAM, T.; BELL, F.; AND CULSHAW, M. G., 2005, Sinkholes and Subsidence: Karst and Cavernous Rocks in Engineering and Construction: Springer Praxis, Chichester, 382 pp.

WINDLE, M. J. S.; ROSE, G. A.; DEVILLERS, R.; AND MARIEJOSÉE, F., 2010, Exploring spatial non-stationarity of fisheries survey data using geographically weighted regression (GWR): An example from the Northwest Atlantic. ICES Journal of Marine Science, Vol. 67, No. 1, pp. 145–154. doi:10.1093/icesjms/fsp224.

ZHOU, W.; BECK, B. F.; AND ADAMS, A. L., 2003, Application of matrix analysis in delineating sinkhole risk areas along Highway (I-70 near Fredrick, Maryland): Environmental Geology, No. 44, pp. 834–842.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

261

Engineering Geology of Sinkholes and Karst: Proceedings of the 7th Multidisciplinary Conference on Sinkholes and the Engineering and Environmental Impacts of Karst: A.A. Balkema Harrisburg, PA, pp. 37–44.

KALOGIROU, S. AND HATZICHRISTOS, T., 2007, A spatial modelling framework for income estimation: Spatial Economic Analysis, Vol. 2, No. 3, pp. 297–316.

KAUFMANN, J., 2008, A statistical approach to karst collapse hazard analysis in Missouri: In Yuhr, L. B.; Alexander, E. Calvin, Jr.; and Beck, Barry F. (Editors), Proceedings of the 11th Multidisciplinary Conference on Sinkholes and the Engineering and Environmental Impacts of Karst: American Society of Civil Engineers, Reston, VA, pp. 257–268.

KEMMERLY, P. R., 1982. Spatial analysis of a karst depression population: Clues to genesis. Geological Society of America Bulletin, Vol. 93, pp. 1078–1086.

LIAO, Q.; LI, M.; CHEN, Z.; SHAO, Y.; AND YANG, K., 2010, Spatial simulation of regional land use patterns based on GWR and CLUE-S model: In XXXX (Editor), Proceedings of the 18th International Conference on Geoinformatics: Yu Liu and Aijun Chen, IEEE, Beijing, China pp. 1–6, doi:10.1109/GEOINFORMATICS.2010.5567963.

MISSOURI DEPARTMENT OF NATURAL RESOURCES, 2016, Missouri Geological Survey Geosciences Technical Resource Assessment Tool (GeoSTRAT): Electronic geospatial data, available at https://dnr.mo.gov MISSOURI SPATIAL DATA INFORMATION SERVICE, 2016, Electronic geospatial data, available at http://msdis.missouri.edu MAGDALENE, S. AND ALEXANDER, E.C., 1995, Sinkhole distribution in Winona County, Minnesota revisited: in: Beck, Barry F. and Person, Felicity M. eds., Karst Geohazards, Proceedings of the Fifth Multidisciplinary Conference on Sinkholes and the Engineering and Environmental Impact of Karst:

Gatlinburg, Tenn., 2–5 April, 1995, A.A. Balkema, Rotterdam, p. 43–51.

NDIATH, M. N.; CISSE, B.; NDIAYE, J. L.; GOMIS, J. F.; BATHIERY, O.; DIA, A. T.; GAYE, O.; AND FAYE, B., 2015, Application of geographically-weighted regression analysis to assess risk factors for malaria hotspots in Keur Soce health and demographic surveillance site: Malaria Journal, Vol. 14, p. 463.

ORNDORFF, R. C.; WEARY, D. J.; AND LAGUEUX, K. M., 2000, Geographic information systems analysis of geologic controls on the distribution of dolines in the Ozarks of southcentral Missouri, USA: Acta Carsologica, Vol. 29, No. 2, pp. 161–175.

ROGERS, J. D., 1997, Spatial Geologic Hazard Analysis in Practice: Spatial Analysis in Soil Dynamics and Earthquake Engineering: ASCE Geotechnical Special Publication 67, pp. 15–28.

UNITED STATES DEPARTMENT OF AGRICULTURE (USDA), 2016, Natural Resources Conservation Service: Electronic geospatial data, available at https://websoilsurvey. sc.egov.usda.gov/App/WebSoilSurvey.aspx WALTHAM, T.; BELL, F.; AND CULSHAW, M. G., 2005, Sinkholes and Subsidence: Karst and Cavernous Rocks in Engineering and Construction: Springer Praxis, Chichester, 382 pp.

WINDLE, M. J. S.; ROSE, G. A.; DEVILLERS, R.; AND MARIEJOSÉE, F., 2010, Exploring spatial non-stationarity of fisheries survey data using geographically weighted regression (GWR): An example from the Northwest Atlantic. ICES Journal of Marine Science, Vol. 67, No. 1, pp. 145–154. doi:10.1093/icesjms/fsp224.

ZHOU, W.; BECK, B. F.; AND ADAMS, A. L., 2003, Application of matrix analysis in delineating sinkhole risk areas along Highway (I-70 near Fredrick, Maryland): Environmental Geology, No. 44, pp. 834–842.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 251–261

261

Simulation of Tailings Flow Resulting from a Dam Breach Using Smoothed Particle Hydrodynamics

Simulation of Tailings Flow Resulting from a Dam Breach Using Smoothed Particle Hydrodynamics

POULAD DANESHVAR ATTILA M. ZSAKI1

POULAD DANESHVAR ATTILA M. ZSAKI1

Department of Building, Civil and Environmental Engineering, Concordia University, Montreal, Quebec H3G 2W1, Canada

Department of Building, Civil and Environmental Engineering, Concordia University, Montreal, Quebec H3G 2W1, Canada

Key Terms: Dam Breach, Tailings Dam, SPH Model, Geotechnical, Engineering Geology ABSTRACT Failure of tailings dams often results in the release of substantial amounts of tailings into the environment, causing considerable damage. The flow of tailings presents a complex modeling challenge due to the freesurface flow and large deformations involved, rendering it intractable by conventional finite element or finite difference methods. A mesh-free formulation, based on smoothed particle hydrodynamics (SPH), was utilized to back-analyze documented tailings dam failures. As with any numerical model, the calibration of model parameters to corresponding physical quantities is a requirement prior to any application of the model. In addition to the effect of model parameters, such as the roughness of terrain, that are hard to quantify, the capabilities and limitations of the SPH model itself were investigated using a simple experimental flume flow setup. In this paper, the calibrated model was applied to literature-reported tailings dam failures. The outflow of tailings interacting with the terrain resulted in good agreement between the simulation results and the reported cases, enabling use of the modeling approach to assess the potential damage caused by tailings dam breaches and to predict flow paths of tailings.

INTRODUCTION Recently, the failure of tailings dams has received considerable attention; the breach of an alumina retaining tailings dam in Hungary (in 2010) released almost a million cubic meters of toxic waste and caused an environmental disaster (Ruyters et al., 2011), while a breach at the non-operating Opemiska tailings facility (in 2008) in Quebec, Canada, contaminated a river ecosystem and a drinking water resource for the nearby population. Thus, the failure of tailings facilities poses an ongoing risk, even after the closure of the mining op1 Corresponding

email: am.zsaki@concordia.ca.

eration (Kossof et al., 2014). There has been a considerable amount of research done on the stability of tailings dams subjected to both static (Ozcan et al., 2013) and dynamic loads (Psarropoulos and Tsompanakis, 2008; Ishihara et al., 2015); however, in comparison, the fate of flowing tailings from dam breach has received relatively modest attention (Jeyapalan et al., 1983; Han and Wang, 1996; and Lin and Li, 2012). This is partially due to the complexity of predicting the behavior of flowing fluid-like matter that interacts with the terrain on which it flows. Models of the flow with either finite elements or a finite difference formulation have to resolve the issue of a free surface and moving boundaries with frequent re-meshing and updating of the simulation.

The class of numerical methods based on mesh-free formulation relatively easily resolves the issues of free surfaces and large deformation of the flowing body.

Smoothed particle hydrodynamics (SPH), in particular, is a mature method used in developing fluid-flow models (Liu and Liu, 2003, 2010; Violeau, 2012). This paper investigates the modeling of tailings flow using an SPH formulation. Since model parameters such as the roughness of terrain are hard to quantify, a simple experimental setup using flume flow was modeled to test the SPH model and to explore its capabilities and limitations, as applied to the flow of tailings. The model was applied to two literature-reported tailings dam failure cases, and simulations were carried out, showing results comparable to the reported extent and coverage of tailings flow.

Key Terms: Dam Breach, Tailings Dam, SPH Model, Geotechnical, Engineering Geology

SMOOTHED PARTICLE HYDRODYNAMICS MODEL FOR TAILINGS FLOW

Recently, the failure of tailings dams has received considerable attention; the breach of an alumina retaining tailings dam in Hungary (in 2010) released almost a million cubic meters of toxic waste and caused an environmental disaster (Ruyters et al., 2011), while a breach at the non-operating Opemiska tailings facility (in 2008) in Quebec, Canada, contaminated a river ecosystem and a drinking water resource for the nearby population. Thus, the failure of tailings facilities poses an ongoing risk, even after the closure of the mining op-

The ability of mesh-free methods, including the SPH method, to treat large deformations without dealing with the difficulties of a highly deformed mesh or frequent re-meshing has made such methods popular in the recent years (Liu and Liu, 2003, 2010). This paper only summarizes the salient features of SPH; for a general treatment references, see publications by Liu and Liu (2003) and, more recently, Liu and Liu (2010) and Violeau (2012), which provide an excellent coverage. According to Rodriguez-Paz and Bonet (2003), in

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

263

ABSTRACT Failure of tailings dams often results in the release of substantial amounts of tailings into the environment, causing considerable damage. The flow of tailings presents a complex modeling challenge due to the freesurface flow and large deformations involved, rendering it intractable by conventional finite element or finite difference methods. A mesh-free formulation, based on smoothed particle hydrodynamics (SPH), was utilized to back-analyze documented tailings dam failures. As with any numerical model, the calibration of model parameters to corresponding physical quantities is a requirement prior to any application of the model. In addition to the effect of model parameters, such as the roughness of terrain, that are hard to quantify, the capabilities and limitations of the SPH model itself were investigated using a simple experimental flume flow setup. In this paper, the calibrated model was applied to literature-reported tailings dam failures. The outflow of tailings interacting with the terrain resulted in good agreement between the simulation results and the reported cases, enabling use of the modeling approach to assess the potential damage caused by tailings dam breaches and to predict flow paths of tailings.

INTRODUCTION

1 Corresponding

email: am.zsaki@concordia.ca.

eration (Kossof et al., 2014). There has been a considerable amount of research done on the stability of tailings dams subjected to both static (Ozcan et al., 2013) and dynamic loads (Psarropoulos and Tsompanakis, 2008; Ishihara et al., 2015); however, in comparison, the fate of flowing tailings from dam breach has received relatively modest attention (Jeyapalan et al., 1983; Han and Wang, 1996; and Lin and Li, 2012). This is partially due to the complexity of predicting the behavior of flowing fluid-like matter that interacts with the terrain on which it flows. Models of the flow with either finite elements or a finite difference formulation have to resolve the issue of a free surface and moving boundaries with frequent re-meshing and updating of the simulation.

The class of numerical methods based on mesh-free formulation relatively easily resolves the issues of free surfaces and large deformation of the flowing body.

Smoothed particle hydrodynamics (SPH), in particular, is a mature method used in developing fluid-flow models (Liu and Liu, 2003, 2010; Violeau, 2012). This paper investigates the modeling of tailings flow using an SPH formulation. Since model parameters such as the roughness of terrain are hard to quantify, a simple experimental setup using flume flow was modeled to test the SPH model and to explore its capabilities and limitations, as applied to the flow of tailings. The model was applied to two literature-reported tailings dam failure cases, and simulations were carried out, showing results comparable to the reported extent and coverage of tailings flow.

SMOOTHED PARTICLE HYDRODYNAMICS MODEL FOR TAILINGS FLOW The ability of mesh-free methods, including the SPH method, to treat large deformations without dealing with the difficulties of a highly deformed mesh or frequent re-meshing has made such methods popular in the recent years (Liu and Liu, 2003, 2010). This paper only summarizes the salient features of SPH; for a general treatment references, see publications by Liu and Liu (2003) and, more recently, Liu and Liu (2010) and Violeau (2012), which provide an excellent coverage. According to Rodriguez-Paz and Bonet (2003), in

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

263

Daneshvar and Zsaki

the SPH method, particles having properties such as mass, velocity, density, and position are used to discretize a continuum. Assuming particles move with the material’s velocity in the continuum, Lagrangiantype governing equations can be constructed. The partial differential equations are transformed into integral equations using a reproducing kernel approximation to express the spatial derivatives at sample points and particles. The main scheme behind this method is the interpolation function, such as f(x) in Eq. 1, where it is evaluated with the help of a kernel function, W(x, h),

f xďż˝ W x − xďż˝ , h d xďż˝ .

(1) f (x) = D

In terms of discrete particles a and b, this is f (x) =

N

b

f (xb ) W (x − xb ) Vb = fh (x),

(2)

where h is the smoothing length, indicating the area around a particle in which the dynamics of the first particle are considerably affected by a second particle.

In addition to Crespo et al. (2008), who suggested that the kernel function has to satisfy consistency and positivity while monotonically decreasing, Swegle et al.

(1995) also suggested that the chosen kernel function has to satisfy the following requirements:

• it has the property of a delta function, lim W (x, h) = (x);

h→0

(3)

• it satisfies the normalization (or unity) condition,

W (x, h) d x = 1; (4) • it is zero everywhere except in the smoothing domain (compactness) W (x, h) = 0

for

|x| ≼ 2h.

A general form for a kernel function is x

W (x, h) = d f ( ) = , m h h

(5)

(6)

where dm is the number of dimensions, and is a factor that ensures the satisfaction of the consistency condition âˆŤ W(x)d x = 1.

D

Equations in the SPH method are used to simulate the motion of interpolated points in the fluid, which are assumed to be particles, using the properties of each particle such as mass and velocity. The SPH formulation is based on conservation law equations, material constitutive models, and equations of state. Following is a brief description of each of these.

264

Daneshvar and Zsaki

Conservation of Momentum A general form for the momentum conservation equation is 1 dv = − ∇ p + g + ďż˝, dt

(7)

where g is the acceleration due to gravity, p is pressure, is density, and ďż˝ stands for the diffusive terms.

Among different diffusive equations used to develop a momentum equation, the “artificial viscosity� approach has been used by Monaghan (1989) and Crespo et al. (2008), which results in the following equation for particle a, with respect to particle b,

pa pb dva =− mb + + ďż˝ ab ∇a Wab + g, (8) dt a2 b2 b

where ∇a Wab is the gradient of the kernel, mb is the mass of particle b, and

2 â&#x2C6;&#x2019; cab ab + ab vab xab < 0

ab .

(9) �ab = 0 vab xab > 0 Here, ab =

hvab xab , 2 xab + 2

(10)

where x and v represent the position and velocity of a particle; xab and vab are the relative displacements and velocities; = 0.01h2 ; and c is the speed of sound. The constants and are characteristic of the problem, and Rodriguez-Paz and Bonet (2003) and Monaghan (1989) give examples for these constants.

Density and Continuity The SPH method is based on laws of conservation.

According to the law of conservation of mass, where dV is an infinitesimal volume:

d d V = 0.

(11) dt

The objective of using the SPH method is to reach a rate of change for density and then to find a smoothed density for each particle and sum over the particles in the effective radius of a sample point; thus, density is expressed as

= mb Wab .

(12) b

While the fluid is considered incompressible, it easier to solve the problem if a slight compressibility is assumed. For fluids such as water, the density on a free surface falls to zero. Thus, to be able to avoid this artificial and abrupt density decrease near the interfaces,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

the SPH method, particles having properties such as mass, velocity, density, and position are used to discretize a continuum. Assuming particles move with the materialâ&#x20AC;&#x2122;s velocity in the continuum, Lagrangiantype governing equations can be constructed. The partial differential equations are transformed into integral equations using a reproducing kernel approximation to express the spatial derivatives at sample points and particles. The main scheme behind this method is the interpolation function, such as f(x) in Eq. 1, where it is evaluated with the help of a kernel function, W(x, h),

f xďż˝ W x â&#x2C6;&#x2019; xďż˝ , h d xďż˝ .

(1) f (x) = D

In terms of discrete particles a and b, this is f (x) =

N

b

f (xb ) W (x â&#x2C6;&#x2019; xb ) Vb = fh (x),

(2)

where h is the smoothing length, indicating the area around a particle in which the dynamics of the first particle are considerably affected by a second particle.

In addition to Crespo et al. (2008), who suggested that the kernel function has to satisfy consistency and positivity while monotonically decreasing, Swegle et al.

(1995) also suggested that the chosen kernel function has to satisfy the following requirements:

â&#x20AC;˘ it has the property of a delta function, lim W (x, h) = (x);

hâ&#x2020;&#x2019;0

(3)

â&#x20AC;˘ it satisfies the normalization (or unity) condition,

W (x, h) d x = 1; (4) â&#x20AC;˘ it is zero everywhere except in the smoothing domain (compactness) W (x, h) = 0

for

|x| â&#x2030;Ľ 2h.

A general form for a kernel function is x

W (x, h) = d f ( ) = , m h h

(5)

(6)

where dm is the number of dimensions, and is a factor that ensures the satisfaction of the consistency condition â&#x2C6;Ť W(x)d x = 1.

D

Equations in the SPH method are used to simulate the motion of interpolated points in the fluid, which are assumed to be particles, using the properties of each particle such as mass and velocity. The SPH formulation is based on conservation law equations, material constitutive models, and equations of state. Following is a brief description of each of these.

264

Conservation of Momentum A general form for the momentum conservation equation is 1 dv = â&#x2C6;&#x2019; â&#x2C6;&#x2021; p + g + ďż˝, dt

(7)

where g is the acceleration due to gravity, p is pressure, is density, and ďż˝ stands for the diffusive terms.

Among different diffusive equations used to develop a momentum equation, the â&#x20AC;&#x153;artificial viscosityâ&#x20AC;? approach has been used by Monaghan (1989) and Crespo et al. (2008), which results in the following equation for particle a, with respect to particle b,

pa pb dva =â&#x2C6;&#x2019; mb + + ďż˝ ab â&#x2C6;&#x2021;a Wab + g, (8) dt a2 b2 b

where â&#x2C6;&#x2021;a Wab is the gradient of the kernel, mb is the mass of particle b, and

2 â&#x2C6;&#x2019; cab ab + ab vab xab < 0

ab .

(9) �ab = 0 vab xab > 0 Here, ab =

hvab xab , 2 xab + 2

(10)

where x and v represent the position and velocity of a particle; xab and vab are the relative displacements and velocities; = 0.01h2 ; and c is the speed of sound. The constants and are characteristic of the problem, and Rodriguez-Paz and Bonet (2003) and Monaghan (1989) give examples for these constants.

Density and Continuity The SPH method is based on laws of conservation.

According to the law of conservation of mass, where dV is an infinitesimal volume:

d d V = 0.

(11) dt

The objective of using the SPH method is to reach a rate of change for density and then to find a smoothed density for each particle and sum over the particles in the effective radius of a sample point; thus, density is expressed as

= mb Wab .

(12) b

While the fluid is considered incompressible, it easier to solve the problem if a slight compressibility is assumed. For fluids such as water, the density on a free surface falls to zero. Thus, to be able to avoid this artificial and abrupt density decrease near the interfaces,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

SPH Tailings Dam Breach Simulation

to achieve correct pressure, and to simplify the calculations, Crespo et al. (2008) suggested using a modified equation, which assigns the same initial density to all particles. This state only changes when the particles are in motion (Monagan, 1989),

d a = mb vab ∇a Wab .

(13) dt b Particle Motion The rate of changes in the position of a particle can be expressed by d xa = va . dt

(14)

However, to ensure more orderly movement of particles and to avoid particle penetration in high-speed flow, in which, if not corrected, one fluid’s particles can penetrate into the other fluid, Monaghan (1994) proposed an XSPH variant �a, which was used in other situations such as those modeled by Monaghan (1994) and Crespo et al. (2008), and which will be added to the right side of Eq. 15, such that mb (vb − va ) d xa = va + ε Wab , dt ab b

(15)

where ε is suggested to be taken as 0.5, as proposed by Monaghan (1994).

The SPH was successfully applied to model both low-viscosity liquids, like water (Liu and Liu, 2003; Violeau, 2012), and high-viscosity ones, like tailings (Rodriguez-Paz and Bonet, 2003; Ho-Minh et al., 2016).

This general overview of SPH is not intended to be a comprehensive one. Nevertheless, it introduces the reader to the basic SPH concepts that will be referred to in the following modeling of tailings flow.

SPH MODEL CALIBRATION The governing equations of the SPH model were presented in the preceding section. The corresponding implementation used in the research was RealFlow (Botella et al., 2006). Since implementation details, other than the supplied user’s manual, are not published, calibration tests were performed to understand the capabilities and limitations of the simulation tools.

A set of key variables was identified in the SPH model, and their influence on the simulation, in RealFlow, in particular, was investigated. These parameters were the initial density, viscosity, and the friction and roughness of a terrain, which are descriptors of both the fluid and the solid surface on which the

SPH Tailings Dam Breach Simulation

flow occurs. Although it is generally accepted that deposited tailings stratify and segregate to some extent, the SPH model can only consider them as a homogeneous isotropic material.

Boundary conditions of the model are defined by three-dimensional objects, e.g., the terrain geometry, to restrict the movement of the flow, representing a noflow boundary condition. All deformations and movements of the tailings are due to gravity. Initial density governs the distribution of particles in a threedimensional model domain at the beginning of simulation. The default value in RealFlow is representative of water. However, the values for viscosity in RealFlow have no physical units, and so great care was taken to explore the behavior of the simulation for any change in the value of viscosity. Both friction and roughness were parameters describing the quality of the surfaces on which the fluid was flowing. Similar to viscosity, these parameters had no direct correlation to physical quantities, and so their variation and the ensuing effect on model behavior were carefully observed.

For the purpose of simulation, the same value was assigned to both friction and roughness; thus, the two variables will be treated as one variable, named terrain roughness.

As a characteristic, yet simple test, an experimental setup was modeled in which the tailings flow was examined in a flume test (Crowder, 2004; Henriquez and Simms, 2009). As a benchmark, published results such as those performed by Crowder (2004) were considered. Crowder carried out experiments on three different tailings pastes retrieved from Bulyanhulu gold mine. Generally, the tailings flow behavior is characterized by a final runout length and angle of repose of tailings, which were suddenly released from a container by lifting a gate. The tailings flow into the flume was modeled, and measurements were taken of the length covered and the angle of the tailings surface relative to horizontal. A schematic of the experimental setup is shown in Figure 1.

The numerical model of the experimental setup was simulated using RealFlow, and values were computed for the runout length, angle of repose, and the speed of tailings flow. The speed was recorded because it was anticipated that the tailings would not come to a full rest in a reasonable time, if at all, due to the finite precision and numerical nature of the simulation. To calculate the angle of repose at “rest,” the equation proposed by Sofra and Boger (2002) was used with the assumption that the surface profile is linear:

H1 − H2 −1 , (16) r = tan L where �r stands for the angle of repose, H1 and H2 are the heights of the material at the gate and toe,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

265

to achieve correct pressure, and to simplify the calculations, Crespo et al. (2008) suggested using a modified equation, which assigns the same initial density to all particles. This state only changes when the particles are in motion (Monagan, 1989),

d a = mb vab ∇a Wab .

(13) dt b Particle Motion The rate of changes in the position of a particle can be expressed by d xa = va . dt

(14)

However, to ensure more orderly movement of particles and to avoid particle penetration in high-speed flow, in which, if not corrected, one fluid’s particles can penetrate into the other fluid, Monaghan (1994) proposed an XSPH variant �a, which was used in other situations such as those modeled by Monaghan (1994) and Crespo et al. (2008), and which will be added to the right side of Eq. 15, such that mb (vb − va ) d xa = va + ε Wab , dt ab b

(15)

where ε is suggested to be taken as 0.5, as proposed by Monaghan (1994).

The SPH was successfully applied to model both low-viscosity liquids, like water (Liu and Liu, 2003; Violeau, 2012), and high-viscosity ones, like tailings (Rodriguez-Paz and Bonet, 2003; Ho-Minh et al., 2016).

This general overview of SPH is not intended to be a comprehensive one. Nevertheless, it introduces the reader to the basic SPH concepts that will be referred to in the following modeling of tailings flow.

SPH MODEL CALIBRATION The governing equations of the SPH model were presented in the preceding section. The corresponding implementation used in the research was RealFlow (Botella et al., 2006). Since implementation details, other than the supplied user’s manual, are not published, calibration tests were performed to understand the capabilities and limitations of the simulation tools.

A set of key variables was identified in the SPH model, and their influence on the simulation, in RealFlow, in particular, was investigated. These parameters were the initial density, viscosity, and the friction and roughness of a terrain, which are descriptors of both the fluid and the solid surface on which the

flow occurs. Although it is generally accepted that deposited tailings stratify and segregate to some extent, the SPH model can only consider them as a homogeneous isotropic material.

Boundary conditions of the model are defined by three-dimensional objects, e.g., the terrain geometry, to restrict the movement of the flow, representing a noflow boundary condition. All deformations and movements of the tailings are due to gravity. Initial density governs the distribution of particles in a threedimensional model domain at the beginning of simulation. The default value in RealFlow is representative of water. However, the values for viscosity in RealFlow have no physical units, and so great care was taken to explore the behavior of the simulation for any change in the value of viscosity. Both friction and roughness were parameters describing the quality of the surfaces on which the fluid was flowing. Similar to viscosity, these parameters had no direct correlation to physical quantities, and so their variation and the ensuing effect on model behavior were carefully observed.

For the purpose of simulation, the same value was assigned to both friction and roughness; thus, the two variables will be treated as one variable, named terrain roughness.

As a characteristic, yet simple test, an experimental setup was modeled in which the tailings flow was examined in a flume test (Crowder, 2004; Henriquez and Simms, 2009). As a benchmark, published results such as those performed by Crowder (2004) were considered. Crowder carried out experiments on three different tailings pastes retrieved from Bulyanhulu gold mine. Generally, the tailings flow behavior is characterized by a final runout length and angle of repose of tailings, which were suddenly released from a container by lifting a gate. The tailings flow into the flume was modeled, and measurements were taken of the length covered and the angle of the tailings surface relative to horizontal. A schematic of the experimental setup is shown in Figure 1.

The numerical model of the experimental setup was simulated using RealFlow, and values were computed for the runout length, angle of repose, and the speed of tailings flow. The speed was recorded because it was anticipated that the tailings would not come to a full rest in a reasonable time, if at all, due to the finite precision and numerical nature of the simulation. To calculate the angle of repose at “rest,” the equation proposed by Sofra and Boger (2002) was used with the assumption that the surface profile is linear:

H1 − H2 −1 , (16) r = tan L where �r stands for the angle of repose, H1 and H2 are the heights of the material at the gate and toe,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

265

Daneshvar and Zsaki

Daneshvar and Zsaki

Figure 1. Side view of the flume experimental setup (after Crowder, 2004) and the SPH model using the same dimensions. Note: the depth of the flume was 0.2 m.

Figure 1. Side view of the flume experimental setup (after Crowder, 2004) and the SPH model using the same dimensions. Note: the depth of the flume was 0.2 m.

respectively, and L is the distance the paste covered from the gate to the toe. The experiment was aimed at establishing the general behavior of the model. Representative values for the previously identified key parameters were selected, and the observations of numerical results were compared to the published results of flume experiments of Crowder (2004). Table 1 summarizes the input parameters of the model. The tailings viscosity in the simulation is of a relatively high viscosity fluid, comparable to the one reported by Crowder (2004). The value for viscosity was selected based on reported cases using RealFlow (Botella et al., 2006), with a value of 3 for a water-like substance.

Since the dynamic viscosity of water is 1.002 × 10−3 N·s/m2 at about 20◦ C, the value taken for viscosity of the tailings was 20, which corresponds to a viscosity value of approximately 7 × 10−3 N·s/m2 , based on the parameters indicated by Vick (2005) and subsequently by Villavicencio (2011). The parameter for coefficient of terrain roughness was selected to represent an average, neither fully smooth, nor excessively rough surface.

The numerical model was created in RealFlow to the dimensions shown in Figure 1, and simulated particles were given an initial density computed using the unit weight from Table 1. Enough time was allowed for the particles to settle under gravity, forming a uniform distribution of tailings behind the gate. Once the gate was lifted, the model simulated tailings runout, and the simulation was terminated once the tailings came to an “apparent” rest (velocity cutoff of 0.0005 m/s used as a stopping criterion). The distance covered and velocity

respectively, and L is the distance the paste covered from the gate to the toe. The experiment was aimed at establishing the general behavior of the model. Representative values for the previously identified key parameters were selected, and the observations of numerical results were compared to the published results of flume experiments of Crowder (2004). Table 1 summarizes the input parameters of the model. The tailings viscosity in the simulation is of a relatively high viscosity fluid, comparable to the one reported by Crowder (2004). The value for viscosity was selected based on reported cases using RealFlow (Botella et al., 2006), with a value of 3 for a water-like substance.

Since the dynamic viscosity of water is 1.002 × 10−3 N·s/m2 at about 20◦ C, the value taken for viscosity of the tailings was 20, which corresponds to a viscosity value of approximately 7 × 10−3 N·s/m2 , based on the parameters indicated by Vick (2005) and subsequently by Villavicencio (2011). The parameter for coefficient of terrain roughness was selected to represent an average, neither fully smooth, nor excessively rough surface.

The numerical model was created in RealFlow to the dimensions shown in Figure 1, and simulated particles were given an initial density computed using the unit weight from Table 1. Enough time was allowed for the particles to settle under gravity, forming a uniform distribution of tailings behind the gate. Once the gate was lifted, the model simulated tailings runout, and the simulation was terminated once the tailings came to an “apparent” rest (velocity cutoff of 0.0005 m/s used as a stopping criterion). The distance covered and velocity

Table 1. Input data for the flume test.

Unit Weight (kN/m3 ) 18.0

266

Viscosity (N·s/m2 )

Coefficient of Terrain Roughness

20

0.4

profiles of the front of the tailings flow area are shown in Figures 2 and 3.

A few key observations can be made about the tailings flow simulation; initially, a relatively high velocity was attained due to the vertical front of tailings falling under gravity at the gate. Correspondingly, a relatively large distance was covered fairly fast. Subsequently, the effect of viscosity and surface roughness prevailed, and the flow slowed down. In the final stages of simulation, there was very little advance in the tailings front with time. Therefore, to address the issue that a numerical simulation will never reach a complete stop, an argument can be made that below a certain velocity threshold, one can consider the flow as reaching a halt. Based on a number of trial simulations, it was confirmed that a minimum velocity cutoff of 0.0005 m/s was a reasonable cutoff to define the atrest condition and end the tailings runout simulation.

This stopping criterion is a balance between computational accuracy of flow and computation time. The time associated with the tailings flow seems quite long in comparison to real experiments, where the tailings could reach the end of the flume in a matter of seconds or minutes. Nevertheless, the general physics of tailings flow appears to be captured by the numerical simulation. There can be a discrepancy between the simulation results and observed real tailings flows. Either the simulation time or the runout distance might not be the same. Parameters such as the cutoff velocity, viscosity of tailings, or the roughness of the terrain can influence a simulation, as is evident from Figure 3, where the choice of cutoff velocity greatly affected the simulation time. Since tailings flows resulting from dam breaches can easily cover great distances in a matter of minutes, as will be seen from the case studies, it is our opinion that the extent of area covered by tailings is more important. Although emergency response times are critical, the failure usually occurs in a time span much less than the first responders can reach the scene.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

Table 1. Input data for the flume test.

Unit Weight (kN/m3 ) 18.0

266

Viscosity (N·s/m2 )

Coefficient of Terrain Roughness

20

0.4

profiles of the front of the tailings flow area are shown in Figures 2 and 3.

A few key observations can be made about the tailings flow simulation; initially, a relatively high velocity was attained due to the vertical front of tailings falling under gravity at the gate. Correspondingly, a relatively large distance was covered fairly fast. Subsequently, the effect of viscosity and surface roughness prevailed, and the flow slowed down. In the final stages of simulation, there was very little advance in the tailings front with time. Therefore, to address the issue that a numerical simulation will never reach a complete stop, an argument can be made that below a certain velocity threshold, one can consider the flow as reaching a halt. Based on a number of trial simulations, it was confirmed that a minimum velocity cutoff of 0.0005 m/s was a reasonable cutoff to define the atrest condition and end the tailings runout simulation.

This stopping criterion is a balance between computational accuracy of flow and computation time. The time associated with the tailings flow seems quite long in comparison to real experiments, where the tailings could reach the end of the flume in a matter of seconds or minutes. Nevertheless, the general physics of tailings flow appears to be captured by the numerical simulation. There can be a discrepancy between the simulation results and observed real tailings flows. Either the simulation time or the runout distance might not be the same. Parameters such as the cutoff velocity, viscosity of tailings, or the roughness of the terrain can influence a simulation, as is evident from Figure 3, where the choice of cutoff velocity greatly affected the simulation time. Since tailings flows resulting from dam breaches can easily cover great distances in a matter of minutes, as will be seen from the case studies, it is our opinion that the extent of area covered by tailings is more important. Although emergency response times are critical, the failure usually occurs in a time span much less than the first responders can reach the scene.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

SPH Tailings Dam Breach Simulation

SPH Tailings Dam Breach Simulation

Figure 2. Runout distance of tailings for initial test.

Figure 2. Runout distance of tailings for initial test.

The uncertainty in the exact physical meaning of parameters such as terrain roughness prompted the expansion of the testing program to evaluate the sensitivity and response of the model for changes in this parameter. The range of values used for terrain roughness were from 0.2 to 0.7 in increments of 0.1. In comparison to the default values in RealFlow, this would characterize the terrain as somewhat smooth to fairly rough (perhaps vegetation covered, thus considerably retarding flow). Other key parameters were kept the same as in the previous test.

The resulting plots of angle of repose and runout distance, as a function of terrain roughness, are shown on Figures 4 and 5, respectively. It was observed that the relationship between the angle of repose and terrain roughness is generally linear, and with increasing roughness, the angle of repose increases as well. However, the runout distance exhibits a non-linear relationship; for values of terrain roughness greater than 0.4, the curve levels off. Thus, values of terrain roughness in excess of 0.5 should only be used where a really rough terrain is present.

Figure 3. Velocity of tailings for initial test.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

The uncertainty in the exact physical meaning of parameters such as terrain roughness prompted the expansion of the testing program to evaluate the sensitivity and response of the model for changes in this parameter. The range of values used for terrain roughness were from 0.2 to 0.7 in increments of 0.1. In comparison to the default values in RealFlow, this would characterize the terrain as somewhat smooth to fairly rough (perhaps vegetation covered, thus considerably retarding flow). Other key parameters were kept the same as in the previous test.

The resulting plots of angle of repose and runout distance, as a function of terrain roughness, are shown on Figures 4 and 5, respectively. It was observed that the relationship between the angle of repose and terrain roughness is generally linear, and with increasing roughness, the angle of repose increases as well. However, the runout distance exhibits a non-linear relationship; for values of terrain roughness greater than 0.4, the curve levels off. Thus, values of terrain roughness in excess of 0.5 should only be used where a really rough terrain is present.

Figure 3. Velocity of tailings for initial test.

267

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

267

Daneshvar and Zsaki

Daneshvar and Zsaki

Figure 4. Angle of repose as a function of terrain roughness.

Figure 4. Angle of repose as a function of terrain roughness.

MODELING OF TAILINGS DAM FAILURES USING SPH—TAPO CANYON AND MERRIESPRUIT TAILINGS DAM FAILURES Tapo Canyon Tailings Dam The Tapo Canyon dam was a 24-m-high tailings dam retaining sand- and gravel-sized aggregate min-

ing waste, located in Northridge, CA. The tailings dam was constructed over weak sedimentary rocks such as marine sandstones, conglomerates, and shales (Harder and Stewart, 1996). The embankments itself was composed of tailings and leftover material of aggregate mining using the “upstream” construction method (Harder and Stewart, 1996). This method is inherently prone to

MODELING OF TAILINGS DAM FAILURES USING SPH—TAPO CANYON AND MERRIESPRUIT TAILINGS DAM FAILURES Tapo Canyon Tailings Dam The Tapo Canyon dam was a 24-m-high tailings dam retaining sand- and gravel-sized aggregate min-

Figure 5. Runout distance as a function of terrain roughness.

268

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

ing waste, located in Northridge, CA. The tailings dam was constructed over weak sedimentary rocks such as marine sandstones, conglomerates, and shales (Harder and Stewart, 1996). The embankments itself was composed of tailings and leftover material of aggregate mining using the “upstream” construction method (Harder and Stewart, 1996). This method is inherently prone to

Figure 5. Runout distance as a function of terrain roughness.

268

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

SPH Tailings Dam Breach Simulation

instability due to any subsequent raising of the embankment being founded on the tailings below. The dam failed as a result of the Northridge earthquake in 1994 (Harder and Stewart, 1996). The tailings pond and the embankment were constructed within a hill, which had been excavated prior to the construction of the pond. Since the ground-water level had made the excavation difficult, mining was stopped, and this pond was converted for settling out fines washed out of the sand and gravel aggregates during mining. Most of the tailings were smaller than the No. 140 sieve size (0.106 mm). The tailings were composed of stratified layers covering a wide range of materials, from fat clays with plasticity indices as high as 30–50 to non-plastic sandy silts and silty sands (Harder and Stewart, 1996).

From 1992, for 2 years, the pond was used by a nearby concrete batch plant for waste concrete, which resulted in a discontinuous layer of concrete with a depth of 1.2 to 1.8 m on the surface of the pond.

The western half still contained some water, which Harder and Stewart (1996) believed to be a result of leakage through the conveyance ditches and ponds located on the immediate northern side. Earthquakeinduced liquefaction of tailings and parts of the dam during the Northridge earthquake caused reductions in soil strength and stiffness, resulting in large and relatively intact blocks of dam sliding over 60 m downstream, which caused the impounded tailings to flow out through the breach and travel several hundred meters downstream (Harder and Stewart, 1996). The retained tailings behind the breach had formed a slope toward the breach in a shape that Harder and Stewart (1996) described as the tailings going through a cone.

The relatively high-viscosity tailings passed through the breach and traveled up to 180 m downstream, reaching an adjacent hillside and entering a creek bed (Harder and Stewart, 1996). The tailings then traveled in the creek channel, reaching a perhaps a few thousand meters (Harder and Stewart, 1996).

Merriespruit Tailings Dam The Merriespruit gold tailings dam was a 38-m-high dam located 250 km from Johannesburg, South Africa, occupying 154 hectares. The dams of the tailings facility were constructed using the “paddock” system of construction, which is, according to Davies (2002), an upstream-type construction method, where subsequent dam raises are built on tailings. It was reported

by Davies (2002) that the operation practices led to little-to-none freeboard, and the dam shells were saturated. By 1978, after 16 years of operation, the embankment was retaining approximately 260 million cubic meters (about 390 million tons) of mining waste, and the capacity was subsequently increased at a rate

SPH Tailings Dam Breach Simulation

of 10 million tons per year (Van Niekerk and Viljoen, 2005). The northern part of the embankment dam was as close as 320 m to houses nearby, which were located in the suburbs of Merriespruit. An almost northsouth–extending valley was located between the houses and the northern portion of the embankment dam.

This shallow drainage valley underlay the center of the northern part of the dam. In 1993, some seepage was reported in the northern section of the embankment dam above the drainage exit and in the portion of the dam separating the center and northern parts of the tailings facility, between ponds 4A and 4B as shown on Figure 6, allowing tailings to flow into the northern pond. These pre-existing conditions, coupled with the reduction of freeboard, which was reduced from 500 mm to 150 mm in some parts of the northern part of the dam, resulted in the accumulation of the tailings closer to the northern portion of the embankment dam and farther from the decant towers (Van Niekerk and Viljoen, 2005). Due to unexpected heavy rainfall in February 1994, water started to flow over the northern section of the embankment dam, followed by sudden failure of the dam with a final breach width of 150 m. Approximately 600,000 cubic meters of tailings (2.5 million tons; Davies, 2001) traveled a distance of 1,960 m through the valley and into an adjacent urban area within 5 minutes (Wagener, 1997), which resulted in 17 deaths and serious damages to properties (Wagener, 1997). The general distribution of tailings after the breach in relation to the surroundings is shown on Figure 6.

Considerations in Model Implementation of Tailings Flows There are two key parameters in defining a model for use in simulation of tailings flows: the terrain and the properties of tailings. Characteristics of the terrain influence tailings runout and speed of advance. In particular, the terrain affects the tailings’ velocity due to the gradient of terrain; the direction and channeling of flow due to large-scale topographic variations; and the runout distance and area the tailings will cover.

The terrain’s roughness and existing irregularities, on a small scale, influence the velocity of tailings via a resisting force against the tailings’ advancement. Depending on their sizes, the local irregularities can affect the direction of flow by acting as obstacles in the flow path or by trapping the tailings in pits or accelerating movement by introducing steep slopes. The incorporation of large-scale features in the model was achieved using digital elevation models (DEMs) of the terrain affected by the tailings dam failures. These DEM models were selected with the highest available degree of resolution, which was often not better than 30 m, as

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

269

instability due to any subsequent raising of the embankment being founded on the tailings below. The dam failed as a result of the Northridge earthquake in 1994 (Harder and Stewart, 1996). The tailings pond and the embankment were constructed within a hill, which had been excavated prior to the construction of the pond. Since the ground-water level had made the excavation difficult, mining was stopped, and this pond was converted for settling out fines washed out of the sand and gravel aggregates during mining. Most of the tailings were smaller than the No. 140 sieve size (0.106 mm). The tailings were composed of stratified layers covering a wide range of materials, from fat clays with plasticity indices as high as 30–50 to non-plastic sandy silts and silty sands (Harder and Stewart, 1996).

From 1992, for 2 years, the pond was used by a nearby concrete batch plant for waste concrete, which resulted in a discontinuous layer of concrete with a depth of 1.2 to 1.8 m on the surface of the pond.

The western half still contained some water, which Harder and Stewart (1996) believed to be a result of leakage through the conveyance ditches and ponds located on the immediate northern side. Earthquakeinduced liquefaction of tailings and parts of the dam during the Northridge earthquake caused reductions in soil strength and stiffness, resulting in large and relatively intact blocks of dam sliding over 60 m downstream, which caused the impounded tailings to flow out through the breach and travel several hundred meters downstream (Harder and Stewart, 1996). The retained tailings behind the breach had formed a slope toward the breach in a shape that Harder and Stewart (1996) described as the tailings going through a cone.

The relatively high-viscosity tailings passed through the breach and traveled up to 180 m downstream, reaching an adjacent hillside and entering a creek bed (Harder and Stewart, 1996). The tailings then traveled in the creek channel, reaching a perhaps a few thousand meters (Harder and Stewart, 1996).

Merriespruit Tailings Dam The Merriespruit gold tailings dam was a 38-m-high dam located 250 km from Johannesburg, South Africa, occupying 154 hectares. The dams of the tailings facility were constructed using the “paddock” system of construction, which is, according to Davies (2002), an upstream-type construction method, where subsequent dam raises are built on tailings. It was reported

by Davies (2002) that the operation practices led to little-to-none freeboard, and the dam shells were saturated. By 1978, after 16 years of operation, the embankment was retaining approximately 260 million cubic meters (about 390 million tons) of mining waste, and the capacity was subsequently increased at a rate

of 10 million tons per year (Van Niekerk and Viljoen, 2005). The northern part of the embankment dam was as close as 320 m to houses nearby, which were located in the suburbs of Merriespruit. An almost northsouth–extending valley was located between the houses and the northern portion of the embankment dam.

This shallow drainage valley underlay the center of the northern part of the dam. In 1993, some seepage was reported in the northern section of the embankment dam above the drainage exit and in the portion of the dam separating the center and northern parts of the tailings facility, between ponds 4A and 4B as shown on Figure 6, allowing tailings to flow into the northern pond. These pre-existing conditions, coupled with the reduction of freeboard, which was reduced from 500 mm to 150 mm in some parts of the northern part of the dam, resulted in the accumulation of the tailings closer to the northern portion of the embankment dam and farther from the decant towers (Van Niekerk and Viljoen, 2005). Due to unexpected heavy rainfall in February 1994, water started to flow over the northern section of the embankment dam, followed by sudden failure of the dam with a final breach width of 150 m. Approximately 600,000 cubic meters of tailings (2.5 million tons; Davies, 2001) traveled a distance of 1,960 m through the valley and into an adjacent urban area within 5 minutes (Wagener, 1997), which resulted in 17 deaths and serious damages to properties (Wagener, 1997). The general distribution of tailings after the breach in relation to the surroundings is shown on Figure 6.

Considerations in Model Implementation of Tailings Flows There are two key parameters in defining a model for use in simulation of tailings flows: the terrain and the properties of tailings. Characteristics of the terrain influence tailings runout and speed of advance. In particular, the terrain affects the tailings’ velocity due to the gradient of terrain; the direction and channeling of flow due to large-scale topographic variations; and the runout distance and area the tailings will cover.

The terrain’s roughness and existing irregularities, on a small scale, influence the velocity of tailings via a resisting force against the tailings’ advancement. Depending on their sizes, the local irregularities can affect the direction of flow by acting as obstacles in the flow path or by trapping the tailings in pits or accelerating movement by introducing steep slopes. The incorporation of large-scale features in the model was achieved using digital elevation models (DEMs) of the terrain affected by the tailings dam failures. These DEM models were selected with the highest available degree of resolution, which was often not better than 30 m, as

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

269

Daneshvar and Zsaki

Daneshvar and Zsaki

Figure 6. The flow path and distribution of tailings in the Merriespruit tailings dam breach (Van Niekerk and Viljoen, 2005).

Figure 6. The flow path and distribution of tailings in the Merriespruit tailings dam breach (Van Niekerk and Viljoen, 2005).

shown on Figure 7. Therefore, the large-scale features of DEM maps needed to the supplemented with relatively accurate small-scale parameters such as terrain roughness.

The properties of tailings used in the modelling affect the simulation on a much finer scale; the travel distance of tailings, apart from the effect of topography, depends on parameters such as the viscosity and density of tailings. The more viscous the tailings are, the greater are the driving forces needed to travel longer distances; therefore, under the same circumstances, tailings with higher viscosity and density come to rest in a shorter distance from the impoundment. Therefore, for the Tapo Canyon scenario, as reported by Harder and Stewart (1996), the viscosity and density were relatively large, representing a more thick, viscous flow. On the contrary, the Merriespruit tailings exhibited a more fluid behavior and were represented by relatively lower values of viscosity and density. The viscosities and densities of the tailings are described on a relative scale, since there is not enough published information available to assign a discrete value. Therefore, both density and viscosity were varied in the simulation to obtain 270

the literature-reported flow characteristics of their respective tailings, as discussed in the next section. The approach taken is similar to that described in Haddad et al. (2010, 2016) and Munari et al. (2013).

Although both the properties of the terrain and tailings are key factors affecting the tailings flow, geometric considerations of the tailings dam set the initial conditions for the release of tailings. Table 2 summarizes the tailings dam geometry and reach of tailings flow for the two case studies, summarizing the information reported by Van Niekerk and Viljoen (2005) for the Merriespruit and Harder and Stewart (1996) for the Tapo Canyon tailings facility.

Tailings Dam Model Construction The representation of tailings facilities in the simulation consisted of a triangulated DEM model for the terrain and a set of simple geometric shapes for the tailings dam and zone of breach. For the Tapo Canyon facility, the tailings impoundment dam had a trapezoidal cross section with two 270-m-long sides, with the third side being 300 m in length. Knowing that the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

shown on Figure 7. Therefore, the large-scale features of DEM maps needed to the supplemented with relatively accurate small-scale parameters such as terrain roughness.

The properties of tailings used in the modelling affect the simulation on a much finer scale; the travel distance of tailings, apart from the effect of topography, depends on parameters such as the viscosity and density of tailings. The more viscous the tailings are, the greater are the driving forces needed to travel longer distances; therefore, under the same circumstances, tailings with higher viscosity and density come to rest in a shorter distance from the impoundment. Therefore, for the Tapo Canyon scenario, as reported by Harder and Stewart (1996), the viscosity and density were relatively large, representing a more thick, viscous flow. On the contrary, the Merriespruit tailings exhibited a more fluid behavior and were represented by relatively lower values of viscosity and density. The viscosities and densities of the tailings are described on a relative scale, since there is not enough published information available to assign a discrete value. Therefore, both density and viscosity were varied in the simulation to obtain 270

the literature-reported flow characteristics of their respective tailings, as discussed in the next section. The approach taken is similar to that described in Haddad et al. (2010, 2016) and Munari et al. (2013).

Although both the properties of the terrain and tailings are key factors affecting the tailings flow, geometric considerations of the tailings dam set the initial conditions for the release of tailings. Table 2 summarizes the tailings dam geometry and reach of tailings flow for the two case studies, summarizing the information reported by Van Niekerk and Viljoen (2005) for the Merriespruit and Harder and Stewart (1996) for the Tapo Canyon tailings facility.

Tailings Dam Model Construction The representation of tailings facilities in the simulation consisted of a triangulated DEM model for the terrain and a set of simple geometric shapes for the tailings dam and zone of breach. For the Tapo Canyon facility, the tailings impoundment dam had a trapezoidal cross section with two 270-m-long sides, with the third side being 300 m in length. Knowing that the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

SPH Tailings Dam Breach Simulation

SPH Tailings Dam Breach Simulation

Figure 7. A DEM of South Africa, showing the location of interest (USGS, 2013).

Figure 7. A DEM of South Africa, showing the location of interest (USGS, 2013).

breach was developed close to the southwestern part of the embankment dam, a separate object was created to represent the breached zone, which was removed at the beginning of the simulation to let the tailings flow out of the impoundment. Similarly, the Merriespruit facility had five sides, with the longest side (the western part of the embankment dam) being approximately 1,687 m.

Since the dam was a substantial structure, retaining tailings in more than one tailings impoundment pond covering a vast area, only the northern part of the dam, where the breach developed, was included in the simulation. To form the breach, similar to the Tapo Canyon case, a retaining object was embedded in the dam where the breach developed, and it was subsequently removed at the beginning of simulation when the area behind the

dam was filled, in essence representing a dam-breach situation, as often referenced in fluid dynamics.

The simulations were not intended to reproduce the breach development process, but rather the extent of tailings flow through the final breach. Hence, the known final breach dimensions were used in the model, and the breach was assumed to form instantaneously. Based on the reported rapidity of the inundation at Merriespruit, and the reported speed of the dam breach at Tapo Canyon, this is considered a reasonable approximation for both cases studied.

The properties of tailings and terrain-specific parameters in the simulation program for the two case studies needed to be explored to represent the as-reported characteristics. Since there are multiple objectives to achieve

breach was developed close to the southwestern part of the embankment dam, a separate object was created to represent the breached zone, which was removed at the beginning of the simulation to let the tailings flow out of the impoundment. Similarly, the Merriespruit facility had five sides, with the longest side (the western part of the embankment dam) being approximately 1,687 m.

Since the dam was a substantial structure, retaining tailings in more than one tailings impoundment pond covering a vast area, only the northern part of the dam, where the breach developed, was included in the simulation. To form the breach, similar to the Tapo Canyon case, a retaining object was embedded in the dam where the breach developed, and it was subsequently removed at the beginning of simulation when the area behind the

Table 2. Key parameters for tailings dam failure simulations.

Dam Height (m)

Breach Length (m)

Tapo Canyon tailings dam 24 Merriespruit tailings dam 38

Runout Distance (m)

dam was filled, in essence representing a dam-breach situation, as often referenced in fluid dynamics.

The simulations were not intended to reproduce the breach development process, but rather the extent of tailings flow through the final breach. Hence, the known final breach dimensions were used in the model, and the breach was assumed to form instantaneously. Based on the reported rapidity of the inundation at Merriespruit, and the reported speed of the dam breach at Tapo Canyon, this is considered a reasonable approximation for both cases studied.

The properties of tailings and terrain-specific parameters in the simulation program for the two case studies needed to be explored to represent the as-reported characteristics. Since there are multiple objectives to achieve

Table 2. Key parameters for tailings dam failure simulations.

Approximate Runout Time (min)

Nature of Tailings

60

180

—

High viscosity

150

1,960

5

Low viscosity

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

271

Dam Height (m)

Breach Length (m)

Tapo Canyon tailings dam 24 Merriespruit tailings dam 38

Runout Distance (m)

Approximate Runout Time (min)

Nature of Tailings

60

180

—

High viscosity

150

1,960

5

Low viscosity

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

271

Daneshvar and Zsaki

Daneshvar and Zsaki

Figure 8. Runout distance as a function of terrain roughness for various values of viscosity for Tapo Canyon tailings flow simulation.

Figure 8. Runout distance as a function of terrain roughness for various values of viscosity for Tapo Canyon tailings flow simulation.

in a simulation, such as the runout length and the associated time it takes to reach it or the area that becomes inundated with the tailings, it might not be possible to satisfy all of these in the same simulation due to the number of unknown or lesser known factors. Thus, our goal was to match the literature-reported runout length and area of tailings coverage and relax the condition of closely matching the time aspect. The reason for this is that most tailings flows occur in a matter of minutes, giving little or no warning and a very short reaction time, while most effort is taken up by remedial measures, which depend on the extent of tailings outflow.

As mentioned earlier, the lesser-known parameters in the SPH model are the viscosity of tailings, and the coefficient of terrain roughness. Thus, for both case studies, the simulation was executed multiple times for a range or viscosities and coefficients of terrain roughness, and the runout length was determined from each model by measuring the front of the tailings flow. For the Tapo Canyon case, as summarized in Figure 8, the range of coefficient of terrain roughness was from 0.0 to 1.2, while the dimensionless viscosity ranged from 3 272

to 150. The resulting set of curves show that to match the runout length of 180 m reported by Harder and Stewart (1996), both viscosity and the coefficient of terrain roughness had to be relatively high. From the plots, interpolated values were 110 for viscosity and 0.98 for the terrain roughness. It has to be noted that there exists a range of combinations of these parameters that could match the runout length of 180 m, but we sought to select a physically plausible set. For example, from Figure 8, a viscosity of 150 and terrain roughness of 0 yield the correct runout length, yet, physically, a terrain roughness of zero is unlikely. Thus, using the same strategy for the Merriespruit model, as shown in Figure 9, values of 3 for viscosity and 0.1 for the coefficient of terrain roughness were selected. To confirm the validity of these parameters, satellite imagery was consulted to qualitatively estimate the terrain roughness. For Tapo Canyon, the terrain appeared to be arid and rough, with shrubby vegetation, while for Merriespruit, it seemed to be smooth and covered with fine-grained soil, confirming the choice of parameter values.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

in a simulation, such as the runout length and the associated time it takes to reach it or the area that becomes inundated with the tailings, it might not be possible to satisfy all of these in the same simulation due to the number of unknown or lesser known factors. Thus, our goal was to match the literature-reported runout length and area of tailings coverage and relax the condition of closely matching the time aspect. The reason for this is that most tailings flows occur in a matter of minutes, giving little or no warning and a very short reaction time, while most effort is taken up by remedial measures, which depend on the extent of tailings outflow.

As mentioned earlier, the lesser-known parameters in the SPH model are the viscosity of tailings, and the coefficient of terrain roughness. Thus, for both case studies, the simulation was executed multiple times for a range or viscosities and coefficients of terrain roughness, and the runout length was determined from each model by measuring the front of the tailings flow. For the Tapo Canyon case, as summarized in Figure 8, the range of coefficient of terrain roughness was from 0.0 to 1.2, while the dimensionless viscosity ranged from 3 272

to 150. The resulting set of curves show that to match the runout length of 180 m reported by Harder and Stewart (1996), both viscosity and the coefficient of terrain roughness had to be relatively high. From the plots, interpolated values were 110 for viscosity and 0.98 for the terrain roughness. It has to be noted that there exists a range of combinations of these parameters that could match the runout length of 180 m, but we sought to select a physically plausible set. For example, from Figure 8, a viscosity of 150 and terrain roughness of 0 yield the correct runout length, yet, physically, a terrain roughness of zero is unlikely. Thus, using the same strategy for the Merriespruit model, as shown in Figure 9, values of 3 for viscosity and 0.1 for the coefficient of terrain roughness were selected. To confirm the validity of these parameters, satellite imagery was consulted to qualitatively estimate the terrain roughness. For Tapo Canyon, the terrain appeared to be arid and rough, with shrubby vegetation, while for Merriespruit, it seemed to be smooth and covered with fine-grained soil, confirming the choice of parameter values.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

SPH Tailings Dam Breach Simulation

SPH Tailings Dam Breach Simulation

Figure 9. Runout distance as a function of terrain roughness for various values of viscosity for Merriespruit tailings flow simulation.

Figure 9. Runout distance as a function of terrain roughness for various values of viscosity for Merriespruit tailings flow simulation.

Thus, the following discussion of the simulation results used these values, which closely correspond to the reported runout and spread of tailings flow.

DISCUSSION OF SIMULATION RESULTS The following discussion of the results of the simulations is presented using optimized parameters that achieved the literature-reported behavior of tailings flow. In particular, parameters such as terrain roughness, which are difficult to quantify on a larger scale, required considerable adjustment, as seen from the previous section. Nevertheless, it appears that with optimized parameters, the mesh-free particle-based simulation captured the physics of tailings flow with an instantaneous breach.

Tapo Canyon Tailings Dam Failure Simulation For the more viscous tailings of the Tapo Canyon simulation, the recommended value for viscosity (in the range of 1 to 5 as a parameter in RealFlow; Botella

et al., 2006) had to be considerably increased to capture a slow-flowing viscous behavior. It was estimated that the tailings had a general density of about 2,400 kg/m3 . In addition to viscosity, the runout distance was also influenced by terrain roughness. These values were estimated to be quite large in order to resist the flow of tailings, in conjunction with viscosity, and thus bring the tailings to a rest in a relatively short distance from the dam. The major parameters of the simulation are summarized in Table 3.

A time versus distance plot of the flow history is shown in Figure 10, and the corresponding time versus velocity profile is shown in Figure 11. Both plots were measured at the front of the tailings flow, which was tracked throughout the simulation.

Table 3. Final model parameters for Tapo Canyon tailings dam failure simulation.

Density (kg/m3 ) 2,400

Viscosity (N·s/m2 )

Coefficient of Terrain Roughness

110

0.98

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

273

Thus, the following discussion of the simulation results used these values, which closely correspond to the reported runout and spread of tailings flow.

DISCUSSION OF SIMULATION RESULTS The following discussion of the results of the simulations is presented using optimized parameters that achieved the literature-reported behavior of tailings flow. In particular, parameters such as terrain roughness, which are difficult to quantify on a larger scale, required considerable adjustment, as seen from the previous section. Nevertheless, it appears that with optimized parameters, the mesh-free particle-based simulation captured the physics of tailings flow with an instantaneous breach.

Tapo Canyon Tailings Dam Failure Simulation For the more viscous tailings of the Tapo Canyon simulation, the recommended value for viscosity (in the range of 1 to 5 as a parameter in RealFlow; Botella

et al., 2006) had to be considerably increased to capture a slow-flowing viscous behavior. It was estimated that the tailings had a general density of about 2,400 kg/m3 . In addition to viscosity, the runout distance was also influenced by terrain roughness. These values were estimated to be quite large in order to resist the flow of tailings, in conjunction with viscosity, and thus bring the tailings to a rest in a relatively short distance from the dam. The major parameters of the simulation are summarized in Table 3.

A time versus distance plot of the flow history is shown in Figure 10, and the corresponding time versus velocity profile is shown in Figure 11. Both plots were measured at the front of the tailings flow, which was tracked throughout the simulation.

Table 3. Final model parameters for Tapo Canyon tailings dam failure simulation.

Density (kg/m3 ) 2,400

Viscosity (N·s/m2 )

Coefficient of Terrain Roughness

110

0.98

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

273

Daneshvar and Zsaki

Daneshvar and Zsaki

Figure 10. Runout distance versus time for Tapo Canyon tailings flow simulation.

Figure 10. Runout distance versus time for Tapo Canyon tailings flow simulation.

As evident from Figure 10, in the few seconds after the dam breach, the tailings covered a relatively large distance due to collapse of the tailings mass under gravity and acceleration down the sloping terrain at the base of embankment. The same effect can be observed on the velocity versus time plot, where the maximum velocity of about 0.38 m/s is reached about 80–100 seconds after the initiation of movement. Subsequently, the velocity drops off sharply and becomes almost asymptotic to about 0.05 m/s until the runout distance reaches 180 m. The simulation was terminated at 3000 seconds, when the tailings had reached the final runout distance corresponding to the literaturereported 180 m, with general characteristics of the flow closely matching what Harder and Stewart (1996) had described.

A sequence of images was captured from the simulation and is shown in Figure 12 to depict the advancing front and the shape of flowing tailings mass at a few key locations along the flow path.

One aspect of the simulation has to be noted; even though the velocity slowed down considerably, to about 0.05 m/s when the tailings have reached the 180 m mark, it did not reach zero, or the previously established cutoff velocity. Other than the inherent property of the simulation where the particles’ velocity never reaches zero due to computation round-off and other numerical anomalies, the shortcoming of the model was to a considerable degree due to the lack of details on the terrain and methods for estimating terrain roughness.

As the tailings flowed on the terrain, the only resisting factors against the flow of the tailings were the

As evident from Figure 10, in the few seconds after the dam breach, the tailings covered a relatively large distance due to collapse of the tailings mass under gravity and acceleration down the sloping terrain at the base of embankment. The same effect can be observed on the velocity versus time plot, where the maximum velocity of about 0.38 m/s is reached about 80–100 seconds after the initiation of movement. Subsequently, the velocity drops off sharply and becomes almost asymptotic to about 0.05 m/s until the runout distance reaches 180 m. The simulation was terminated at 3000 seconds, when the tailings had reached the final runout distance corresponding to the literaturereported 180 m, with general characteristics of the flow closely matching what Harder and Stewart (1996) had described.

Figure 11. Velocity versus time for Tapo Canyon tailings flow simulation.

274

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

A sequence of images was captured from the simulation and is shown in Figure 12 to depict the advancing front and the shape of flowing tailings mass at a few key locations along the flow path.

One aspect of the simulation has to be noted; even though the velocity slowed down considerably, to about 0.05 m/s when the tailings have reached the 180 m mark, it did not reach zero, or the previously established cutoff velocity. Other than the inherent property of the simulation where the particles’ velocity never reaches zero due to computation round-off and other numerical anomalies, the shortcoming of the model was to a considerable degree due to the lack of details on the terrain and methods for estimating terrain roughness.

As the tailings flowed on the terrain, the only resisting factors against the flow of the tailings were the

Figure 11. Velocity versus time for Tapo Canyon tailings flow simulation.

274

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

SPH Tailings Dam Breach Simulation

SPH Tailings Dam Breach Simulation

Figure 12. Sequence of tailings flow profiles at 25 m (127 seconds after breach), 120 m (1,417 seconds after breach), and 180 m (2,948 seconds after breach) distance from the dam, Tapo Canyon tailings flow simulation.

Figure 12. Sequence of tailings flow profiles at 25 m (127 seconds after breach), 120 m (1,417 seconds after breach), and 180 m (2,948 seconds after breach) distance from the dam, Tapo Canyon tailings flow simulation.

terrain’s topography and terrain roughness, whereas in the physical event, both natural and constructed obstacles also contributed to the process of retarding the tailings advance.

terrain’s topography and terrain roughness, whereas in the physical event, both natural and constructed obstacles also contributed to the process of retarding the tailings advance.

Merriespruit Tailings Dam Failure Simulation For the case of low-viscosity tailings, representing the Merriespruit tailings, both the values of viscosity and terrain roughness were considerably lower than those of the Tapo Canyon case. As suggested by literature, the tailings pond had a higher than usual accumulation of ponded free water (Van Niekerk and Viljoen, 2005).

Thus, as the breach developed, this ponded free water mixed with the tailings, and thus the flowing tailings had characteristics similar to water. In order to match the tailings runout, the viscosity used in the model was selected to be more representative of a watery material.

The key model parameters are summarized in Table 4.

The terrain roughness, representing the terrain, was reduced to simulate a condition where tailings can flow unimpeded, attaining a large runout length. As discussed earlier, the terrain-specific variables are perhaps the weakest link in establishing the model for simula-

tion, since they represent the averaged small-scale topographic features, which were not available with the resolution of DEM data available for this study.

After the removal of a block representing the dam, the low-viscosity tailings flowed out of the impoundment and traveled a distance of 1,960 m in approximately 12 minutes in the model, spending the last 3–4 minutes in the terminal zone due to the velocity slowly reaching the cutoff velocity. The actual observed travel time to reach the terminus point of tailings flow was reported by Wagener (1997) to be about 5 minutes.

Thus, according to our simulation, it took about twice as long. However, the terminal zone of a bird sanctuary pond was reached in 8.65 seconds. In addition, the approximately 10 km/hr average velocity of flow appears to be within literature-reported values of <1 km/hr to Table 4. Final model parameters for Merriespruit tailings dam failure simulations.

Density (kg/m3 ) 1,500

Viscosity (N·s/m2 )

Coefficient Terrain Roughness

3

0.1

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

275

Merriespruit Tailings Dam Failure Simulation For the case of low-viscosity tailings, representing the Merriespruit tailings, both the values of viscosity and terrain roughness were considerably lower than those of the Tapo Canyon case. As suggested by literature, the tailings pond had a higher than usual accumulation of ponded free water (Van Niekerk and Viljoen, 2005).

Thus, as the breach developed, this ponded free water mixed with the tailings, and thus the flowing tailings had characteristics similar to water. In order to match the tailings runout, the viscosity used in the model was selected to be more representative of a watery material.

The key model parameters are summarized in Table 4.

The terrain roughness, representing the terrain, was reduced to simulate a condition where tailings can flow unimpeded, attaining a large runout length. As discussed earlier, the terrain-specific variables are perhaps the weakest link in establishing the model for simula-

tion, since they represent the averaged small-scale topographic features, which were not available with the resolution of DEM data available for this study.

After the removal of a block representing the dam, the low-viscosity tailings flowed out of the impoundment and traveled a distance of 1,960 m in approximately 12 minutes in the model, spending the last 3–4 minutes in the terminal zone due to the velocity slowly reaching the cutoff velocity. The actual observed travel time to reach the terminus point of tailings flow was reported by Wagener (1997) to be about 5 minutes.

Thus, according to our simulation, it took about twice as long. However, the terminal zone of a bird sanctuary pond was reached in 8.65 seconds. In addition, the approximately 10 km/hr average velocity of flow appears to be within literature-reported values of <1 km/hr to Table 4. Final model parameters for Merriespruit tailings dam failure simulations.

Density (kg/m3 ) 1,500

Viscosity (N·s/m2 )

Coefficient Terrain Roughness

3

0.1

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

275

Daneshvar and Zsaki

Daneshvar and Zsaki

Figure 13. Runout distance versus time for Merriespruit tailings flow simulation.

Figure 13. Runout distance versus time for Merriespruit tailings flow simulation.

40 km/hr (Blight et al., 1981). The plots of distance versus time and velocity versus time for the advancing front of tailings movement are shown in Figures 13 and 14, respectively.

As evident from Figures 13 and 14, within the first 100 seconds, the tailings flow advanced rapidly through an instantaneous breach, and the velocity continuously increased. During approximately the first 150 seconds, the velocity continued to increase (probably due to the head of the tailings behind the dam under the influence of gravity). The effects of terrain roughness and irregularities of the terrain became greater, along with

the flatness of the terrain, resulting in decrease in the momentum of flow, as observable from the decline in velocity.

Similar to the Tapo Canyon simulation, the velocity would never actually reach zero due to the numerical nature of the simulation. However, according to reports and observations from satellite imagery, the tailings had entered a pond upon reaching their terminus on land. Establishing the termination velocity of the runout was complicated by the presence of the pond.

It was assumed that once the tailings had entered the water, they became diluted and dispersed. Key frames

40 km/hr (Blight et al., 1981). The plots of distance versus time and velocity versus time for the advancing front of tailings movement are shown in Figures 13 and 14, respectively.

As evident from Figures 13 and 14, within the first 100 seconds, the tailings flow advanced rapidly through an instantaneous breach, and the velocity continuously increased. During approximately the first 150 seconds, the velocity continued to increase (probably due to the head of the tailings behind the dam under the influence of gravity). The effects of terrain roughness and irregularities of the terrain became greater, along with

Figure 14. Velocity versus time for Merriespruit tailings flow simulation.

276

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

the flatness of the terrain, resulting in decrease in the momentum of flow, as observable from the decline in velocity.

Similar to the Tapo Canyon simulation, the velocity would never actually reach zero due to the numerical nature of the simulation. However, according to reports and observations from satellite imagery, the tailings had entered a pond upon reaching their terminus on land. Establishing the termination velocity of the runout was complicated by the presence of the pond.

It was assumed that once the tailings had entered the water, they became diluted and dispersed. Key frames

Figure 14. Velocity versus time for Merriespruit tailings flow simulation.

276

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

SPH Tailings Dam Breach Simulation

SPH Tailings Dam Breach Simulation

Figure 15. Key frames of Merriespruit tailings dam failure simulation showing advancement of tailings at 80 m, 1,650 m and 1,960 m from the dam.

Figure 15. Key frames of Merriespruit tailings dam failure simulation showing advancement of tailings at 80 m, 1,650 m and 1,960 m from the dam.

of the tailings advancement from the simulation are shown on Figure 15. The upper image shows the front of the tailings flow at 80 m from the breach. The middle image was taken at 1,650 m from the breach, while the bottom one is at the terminus, reaching the pond at 1,960 m from the tailings facility. Due to the relative flatness of the terrain, where the housing development was located, the tailings flow became widely dispersed, as shown on the middle and bottom images on Figure 15.

Although the tailings flow exhibited the expected behavior of thinning and fanning out on flat terrain, an important shortcoming was identified; the location and arrangement of houses forming streets in the subdivision could act as channels to guide and divert the

of the tailings advancement from the simulation are shown on Figure 15. The upper image shows the front of the tailings flow at 80 m from the breach. The middle image was taken at 1,650 m from the breach, while the bottom one is at the terminus, reaching the pond at 1,960 m from the tailings facility. Due to the relative flatness of the terrain, where the housing development was located, the tailings flow became widely dispersed, as shown on the middle and bottom images on Figure 15.

Although the tailings flow exhibited the expected behavior of thinning and fanning out on flat terrain, an important shortcoming was identified; the location and arrangement of houses forming streets in the subdivision could act as channels to guide and divert the

flow and reduce the velocity of tailings. This was not captured in the simulations.

In conclusion, if the results of the simulation are superimposed on the satellite imagery of the terrain and the outline of the flow (from Van Niekerk and Viljoen, 2005), substantial agreement is evident, as seen from Figure 16. The distribution of flow, in comparison to what the literature reported (cf. Figure 6), encompasses the residential area, where most of the damage occurred, and terminates at the pond of a bird sanctuary.

If accurate model parameters can be derived independently, the SPH model may be able to predict runout from failure of a tailings dam. The current shortcoming with this method is the accurate prediction of model parameters.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

277

flow and reduce the velocity of tailings. This was not captured in the simulations.

In conclusion, if the results of the simulation are superimposed on the satellite imagery of the terrain and the outline of the flow (from Van Niekerk and Viljoen, 2005), substantial agreement is evident, as seen from Figure 16. The distribution of flow, in comparison to what the literature reported (cf. Figure 6), encompasses the residential area, where most of the damage occurred, and terminates at the pond of a bird sanctuary.

If accurate model parameters can be derived independently, the SPH model may be able to predict runout from failure of a tailings dam. The current shortcoming with this method is the accurate prediction of model parameters.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263â&#x20AC;&#x201C;279

277

Daneshvar and Zsaki

Daneshvar and Zsaki

tive viscosity of tailings. Additionally, the value for the roughness was confirmed by visually assessing satellite imagery and estimating the terrain roughness from it as well. Simulations using this value yielded a close match to the reported values of runout for both case studies.

As a practical predictive analysis tool for assessing the potential runout of tailings in case of a dam breach, the terrain roughness may be estimated from satellite imagery and used to perform a set of simulations. The results could then be employed to assess the risk and damage and map out the affected area in the event of a tailings facility failure.

tive viscosity of tailings. Additionally, the value for the roughness was confirmed by visually assessing satellite imagery and estimating the terrain roughness from it as well. Simulations using this value yielded a close match to the reported values of runout for both case studies.

As a practical predictive analysis tool for assessing the potential runout of tailings in case of a dam breach, the terrain roughness may be estimated from satellite imagery and used to perform a set of simulations. The results could then be employed to assess the risk and damage and map out the affected area in the event of a tailings facility failure.

ACKNOWLEDGMENTS

ACKNOWLEDGMENTS

The research presented in this paper was supported by an Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant held by the second author and by a graduate student award from Faculty of Engineering and Computer Science (ENCS)–Concordia to the first author.

The research presented in this paper was supported by an Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant held by the second author and by a graduate student award from Faculty of Engineering and Computer Science (ENCS)–Concordia to the first author.

REFERENCES Figure 16. The result of Merriespruit tailings dam failure simulation superimposed on satellite imagery (Google Earth, 2014),

CONCLUSIONS The damage caused by tailings dam failures can be a costly and often fatal result of many mining operations around the globe. Upon failure, the interaction of fluidlike tailings with the terrain generally dictates the direction and velocity of flow. Characteristics of the terrain, such as large-scale topographic features and small-scale roughness, can retard and direct the flow of tailings. To simulate the tailings flow, RealFlow, a SPH-based tool, was used after a set of calibration tests to model a flume flow experiment with considerable success at capturing the characteristics of the flow. The SPH model was used to re-create two literature-reported tailings dam failure cases: one with a fairly viscous tailings (Tapo Canyon) and one with low-viscosity tailings (Merriespruit). Although the results generated by the simulation exhibit strong correspondence with the actual observed flow pattern and distribution, the most important and least understood parameter in the present simulation is associated with the roughness of terrain. Therefore, a set of simulations was performed to explore the influence of terrain roughness on the runout distance for both case studies. A group of curves was obtained for different viscosities of tailings. From this, knowing the observed runout distance, a value for the terrain roughness was selected for use in a simulation for a given rela278

BLIGHT, G. E.; ROBINSON, M. J.; AND DIERING, J. A. C., 1981, The flow of slurry from a breached tailings dam: Journal of the South African Institute of Mining and Metallurgy, Vol. 1981, No. 1, pp. 1–8.

BOTELLA, B. L.; ALIX, J. M.; MORA, L. M.; AND STASIUK, M., 2006, RF4 User Guide v1.1: Next Limit S.L., Barcelona, Spain.

CRESPO, A. J. C.; GOMEZ-GESTEIRA, M.; AND DALRYMPLE, R. A., 2008, Modeling dam break behavior over a wet bed by a SPH technique: Journal of Waterway, Port, Coastal and Ocean Engineering, Vol. 134, No. 6, pp. 313–320.

CROWDER, J. J., 2004, Deposition, Consolidation, and Strength of a Non-Plastic Tailings Paste for Surface Disposal: Unpublished Ph.D. Thesis, Department of Civil Engineering, University of Toronto, Toronto, Canada.

DAVIES, M. P., 2001, Impounded mine tailings: What are the failures telling us?: The Canadian Mining and Metallurgical Bulletin, Vol. 94, pp. 53–59.

DAVIES, M. P., 2002, Tailings impoundment failures: Are geotechnical engineers listening?: Waste Geotechnics, Geotechnical News, pp. 31–36.

GOOGLE EARTH, 2014, Google Earth v.5.2.1: Electronic document, available at http://www.oldversion.com/windows/googleearth-5-2-1-1588 HADDAD, B.; PALACIOS, D.; PASTOR, M.; AND ZAMORANOD, J. J., 2016, Smoothed particle hydrodynamic modeling of volcanic debris flows: Application to Huiloac Gorge lahars (Popocatépetl volcano, Mexico): Journal of Volcanology and Geothermal Research, Vol. 324, pp. 73–87.

HADDAD, B.; PASTOR, M.; PALACIOS, D.; AND MUÑOZ-SALINAS, E., 2010, A SPH depth integrated model for Popocatépetl 2001 lahar (Mexico): Sensitivity analysis and runout simulation: Engineering Geology, Vol. 114, pp. 312–329.

HAN G. AND WANG, D., 1996, Numerical modeling of Anhui debris flow: Journal of Hydraulic Engineering, Vol. 122, No. 5, pp.

262–265.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

REFERENCES Figure 16. The result of Merriespruit tailings dam failure simulation superimposed on satellite imagery (Google Earth, 2014),

CONCLUSIONS The damage caused by tailings dam failures can be a costly and often fatal result of many mining operations around the globe. Upon failure, the interaction of fluidlike tailings with the terrain generally dictates the direction and velocity of flow. Characteristics of the terrain, such as large-scale topographic features and small-scale roughness, can retard and direct the flow of tailings. To simulate the tailings flow, RealFlow, a SPH-based tool, was used after a set of calibration tests to model a flume flow experiment with considerable success at capturing the characteristics of the flow. The SPH model was used to re-create two literature-reported tailings dam failure cases: one with a fairly viscous tailings (Tapo Canyon) and one with low-viscosity tailings (Merriespruit). Although the results generated by the simulation exhibit strong correspondence with the actual observed flow pattern and distribution, the most important and least understood parameter in the present simulation is associated with the roughness of terrain. Therefore, a set of simulations was performed to explore the influence of terrain roughness on the runout distance for both case studies. A group of curves was obtained for different viscosities of tailings. From this, knowing the observed runout distance, a value for the terrain roughness was selected for use in a simulation for a given rela278

BLIGHT, G. E.; ROBINSON, M. J.; AND DIERING, J. A. C., 1981, The flow of slurry from a breached tailings dam: Journal of the South African Institute of Mining and Metallurgy, Vol. 1981, No. 1, pp. 1–8.

BOTELLA, B. L.; ALIX, J. M.; MORA, L. M.; AND STASIUK, M., 2006, RF4 User Guide v1.1: Next Limit S.L., Barcelona, Spain.

CRESPO, A. J. C.; GOMEZ-GESTEIRA, M.; AND DALRYMPLE, R. A., 2008, Modeling dam break behavior over a wet bed by a SPH technique: Journal of Waterway, Port, Coastal and Ocean Engineering, Vol. 134, No. 6, pp. 313–320.

CROWDER, J. J., 2004, Deposition, Consolidation, and Strength of a Non-Plastic Tailings Paste for Surface Disposal: Unpublished Ph.D. Thesis, Department of Civil Engineering, University of Toronto, Toronto, Canada.

DAVIES, M. P., 2001, Impounded mine tailings: What are the failures telling us?: The Canadian Mining and Metallurgical Bulletin, Vol. 94, pp. 53–59.

DAVIES, M. P., 2002, Tailings impoundment failures: Are geotechnical engineers listening?: Waste Geotechnics, Geotechnical News, pp. 31–36.

GOOGLE EARTH, 2014, Google Earth v.5.2.1: Electronic document, available at http://www.oldversion.com/windows/googleearth-5-2-1-1588 HADDAD, B.; PALACIOS, D.; PASTOR, M.; AND ZAMORANOD, J. J., 2016, Smoothed particle hydrodynamic modeling of volcanic debris flows: Application to Huiloac Gorge lahars (Popocatépetl volcano, Mexico): Journal of Volcanology and Geothermal Research, Vol. 324, pp. 73–87.

HADDAD, B.; PASTOR, M.; PALACIOS, D.; AND MUÑOZ-SALINAS, E., 2010, A SPH depth integrated model for Popocatépetl 2001 lahar (Mexico): Sensitivity analysis and runout simulation: Engineering Geology, Vol. 114, pp. 312–329.

HAN G. AND WANG, D., 1996, Numerical modeling of Anhui debris flow: Journal of Hydraulic Engineering, Vol. 122, No. 5, pp.

262–265.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

SPH Tailings Dam Breach Simulation HARDER, L. F., JR. AND STEWART, J. P., 1996, Failure of Tapo Canyon tailings dam: Journal of Performance of Constructed Facilities, Vol. 10, No. 3, pp. 109–114.

HENRIQUEZ, J. AND SIMMS, P., 2009, Dynamic imaging and modelling of multilayer deposition of gold paste tailings: Minerals Engineering, Vol. 22, No. 2, pp. 128–139.

HO-MINH, D; TONG, J.; AND TAN, D. S., 2016, A smoothed particle hydrodynamics model of sediment-water mixture dispersing in quiescent water. In Ivey, G., Jones, N. and Zhou, T. (Editors), 20th Australasian Fluid Mechanics Conference, December 5–8, 2016, Australian Fluid Mechanics Society, Perth, Australia.

ISHIHARA, K.; KENNOSUKE, U.; YAMADA, S.; YASUDA, S.; AND YONEOKA, T., 2015, Breach of a tailings dam in the 2011 earthquake in Japan: Soil Dynamics and Earthquake Engineering, Vol. 68, pp. 3–22.

JEYAPALAN, J. K.; DUNCAN, J. M.; AND SEED, H. B., 1983, Analyses of flow failures of mine tailings dams: Journal of Geotechnical Engineering, Vol. 109, No. 2, pp. 150–171.

KOSSOFF, D.; DUBBIN, W. E.; ALFREDSSON, M.; EDWARDS, S. J.; MACKLIN, M. G.; AND HUDSON-EDWARDS, K. A., 2014, Mine tailings dams: Characteristics, failure, environmental impacts, and remediation: Applied Geochemistry, Vol. 51, pp. 229–245.

LIN J. AND LI, J., 2012, Tailings dam break flow and sediment numerical simulation. In Qu, S. and Lin, S. (Editors) International Conference on Civil Engineering and Urban Planning 2012, August 18–20, 2012, Yantai, China, Construction Institute of ASCE, Reston, VA. pp. 677–683.

LIU, G. R. AND LIU, M. B., 2003, Smoothed Particle Hydrodynamics; A Meshfree Particle Method: World Scientific, Hackensack, NJ. LIU, M. B. AND LIU, G. R., 2010, Smoothed particle hydrodynamics (SPH): An overview and recent developments: Archives of Computational Methods in Engineering, Vol. 17, pp. 25–76.

MONAGHAN, J. J., 1989, On the problem of penetration in particle methods: Journal of Computational Physics, Vol. 82, pp. 1–5.

MONAGHAN, J. J., 1994, Simulating free surface flows with SPH: Journal of Computational Physics, Vol. 110, pp. 399–406.

MUNARI, S.; BOSSI, G.; D’AGOSTINO, V.; POZZA, E.; BETTELLA, A.; AND COLA, S., 2013, Calibration of a SPH model for the numerical analysis of mud-flow run-out: The case of the Rotolon torrent in Italy. In Reichart, G. J., Tinti, S. and Wasowski, J.

SPH Tailings Dam Breach Simulation

(Editors) EGU General Assembly 2013, April 7–12, 2013: European Geosciences Union, Munich, Germany. p. 14189.

OZCAN, N. T.; ULUSAY, R.; AND ISIK, N. S., 2013, A study on geotechnical characterization and stability of downstream slope of a tailings dam to improve its storage capacity (Turkey): Environmental Earth Sciences, Vol. 69, No. 6, pp. 1871–1890.

PSARROPOULOS, P. N. AND TSOMPANAKIS, Y., 2008, Stability of tailings dams under static and seismic loading: Canadian Geotechnical Journal, Vol. 45, pp. 663–675.

RODRIGUEZ-PAZ, M. AND BONET, J., 2003, A corrected smooth particle hydrodynamics method for the simulation of debris flows: Journal of Numerical Methods for Partial Differential Equations, Vol. 20, No. 1, pp. 140–163.

RUYTERS, S.; MERTENS, J.; VASSILIEVA, E.; DEHANDSCHUTTER, B.; POFFIJN, A.; AND SMOLDERS, E., 2011, The red mud accident in Ajka (Hungary): Plant toxicity and trace metal bioavailability in red mud contaminated soil: Environmental Science & Technology, Vol. 45, pp. 1616–1622.

SOFRA, F. AND BOGER, D. V., 2002, Rheology for waste minimization in the minerals industry: Chemical Engineering Journal, Vol. 86, No. 3, pp. 319–330.

SWEGLE, J. W.; HICKS, D. L.; AND ATTAWAY, S. W., 1995, Smoothed particle hydrodynamics stability analysis: Journal of Computational Physics, Vol. 116, pp. 123–134.

U.S. GEOLOGICAL SURVEY (USGS), 2013, USGS DEM repository: Electronic document, available at http://edc2.usgs.gov/ geodata/index.php VAN NIEKERK, H. J. AND VILJOEN, M. J., 2005, Causes and consequences of the Merriespruit and other tailings dams failures: Journal of Land Degradation and Development, Vol. 16, pp.

201–212.

VICK, S. G., 2005, Planning, Design, and Analysis of Tailings Dams: Wiley-Interscience, Toronto, Ontario, Canada.

VILLAVICENCIO, A. G.; BREUL, P.; BACCONNET, C.; BOISSIER, D.; AND ESPINACE, A. R., 2011, Estimation of the variability of tailings dams properties in order to perform probabilistic assessment: Geotechnical and Geological Engineering, Vol. 29, No. 6, pp. 1073–1084.

VIOLEAU, D., 2012, Fluid Mechanics and the SPH Method, Theory and Applications: Oxford University Press, Oxford, U.K. WAGENER, F., 1997, The Merriespruit slimes dam failure: Overview and lessons learnt: Technical paper: Journal of South African Civil Engineering, Vol. 39, No. 3, pp. 11–15.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

279

HARDER, L. F., JR. AND STEWART, J. P., 1996, Failure of Tapo Canyon tailings dam: Journal of Performance of Constructed Facilities, Vol. 10, No. 3, pp. 109–114.

HENRIQUEZ, J. AND SIMMS, P., 2009, Dynamic imaging and modelling of multilayer deposition of gold paste tailings: Minerals Engineering, Vol. 22, No. 2, pp. 128–139.

HO-MINH, D; TONG, J.; AND TAN, D. S., 2016, A smoothed particle hydrodynamics model of sediment-water mixture dispersing in quiescent water. In Ivey, G., Jones, N. and Zhou, T. (Editors), 20th Australasian Fluid Mechanics Conference, December 5–8, 2016, Australian Fluid Mechanics Society, Perth, Australia.

ISHIHARA, K.; KENNOSUKE, U.; YAMADA, S.; YASUDA, S.; AND YONEOKA, T., 2015, Breach of a tailings dam in the 2011 earthquake in Japan: Soil Dynamics and Earthquake Engineering, Vol. 68, pp. 3–22.

JEYAPALAN, J. K.; DUNCAN, J. M.; AND SEED, H. B., 1983, Analyses of flow failures of mine tailings dams: Journal of Geotechnical Engineering, Vol. 109, No. 2, pp. 150–171.

KOSSOFF, D.; DUBBIN, W. E.; ALFREDSSON, M.; EDWARDS, S. J.; MACKLIN, M. G.; AND HUDSON-EDWARDS, K. A., 2014, Mine tailings dams: Characteristics, failure, environmental impacts, and remediation: Applied Geochemistry, Vol. 51, pp. 229–245.

LIN J. AND LI, J., 2012, Tailings dam break flow and sediment numerical simulation. In Qu, S. and Lin, S. (Editors) International Conference on Civil Engineering and Urban Planning 2012, August 18–20, 2012, Yantai, China, Construction Institute of ASCE, Reston, VA. pp. 677–683.

LIU, G. R. AND LIU, M. B., 2003, Smoothed Particle Hydrodynamics; A Meshfree Particle Method: World Scientific, Hackensack, NJ. LIU, M. B. AND LIU, G. R., 2010, Smoothed particle hydrodynamics (SPH): An overview and recent developments: Archives of Computational Methods in Engineering, Vol. 17, pp. 25–76.

MONAGHAN, J. J., 1989, On the problem of penetration in particle methods: Journal of Computational Physics, Vol. 82, pp. 1–5.

MONAGHAN, J. J., 1994, Simulating free surface flows with SPH: Journal of Computational Physics, Vol. 110, pp. 399–406.

MUNARI, S.; BOSSI, G.; D’AGOSTINO, V.; POZZA, E.; BETTELLA, A.; AND COLA, S., 2013, Calibration of a SPH model for the numerical analysis of mud-flow run-out: The case of the Rotolon torrent in Italy. In Reichart, G. J., Tinti, S. and Wasowski, J.

(Editors) EGU General Assembly 2013, April 7–12, 2013: European Geosciences Union, Munich, Germany. p. 14189.

OZCAN, N. T.; ULUSAY, R.; AND ISIK, N. S., 2013, A study on geotechnical characterization and stability of downstream slope of a tailings dam to improve its storage capacity (Turkey): Environmental Earth Sciences, Vol. 69, No. 6, pp. 1871–1890.

PSARROPOULOS, P. N. AND TSOMPANAKIS, Y., 2008, Stability of tailings dams under static and seismic loading: Canadian Geotechnical Journal, Vol. 45, pp. 663–675.

RODRIGUEZ-PAZ, M. AND BONET, J., 2003, A corrected smooth particle hydrodynamics method for the simulation of debris flows: Journal of Numerical Methods for Partial Differential Equations, Vol. 20, No. 1, pp. 140–163.

RUYTERS, S.; MERTENS, J.; VASSILIEVA, E.; DEHANDSCHUTTER, B.; POFFIJN, A.; AND SMOLDERS, E., 2011, The red mud accident in Ajka (Hungary): Plant toxicity and trace metal bioavailability in red mud contaminated soil: Environmental Science & Technology, Vol. 45, pp. 1616–1622.

SOFRA, F. AND BOGER, D. V., 2002, Rheology for waste minimization in the minerals industry: Chemical Engineering Journal, Vol. 86, No. 3, pp. 319–330.

SWEGLE, J. W.; HICKS, D. L.; AND ATTAWAY, S. W., 1995, Smoothed particle hydrodynamics stability analysis: Journal of Computational Physics, Vol. 116, pp. 123–134.

U.S. GEOLOGICAL SURVEY (USGS), 2013, USGS DEM repository: Electronic document, available at http://edc2.usgs.gov/ geodata/index.php VAN NIEKERK, H. J. AND VILJOEN, M. J., 2005, Causes and consequences of the Merriespruit and other tailings dams failures: Journal of Land Degradation and Development, Vol. 16, pp.

201–212.

VICK, S. G., 2005, Planning, Design, and Analysis of Tailings Dams: Wiley-Interscience, Toronto, Ontario, Canada.

VILLAVICENCIO, A. G.; BREUL, P.; BACCONNET, C.; BOISSIER, D.; AND ESPINACE, A. R., 2011, Estimation of the variability of tailings dams properties in order to perform probabilistic assessment: Geotechnical and Geological Engineering, Vol. 29, No. 6, pp. 1073–1084.

VIOLEAU, D., 2012, Fluid Mechanics and the SPH Method, Theory and Applications: Oxford University Press, Oxford, U.K. WAGENER, F., 1997, The Merriespruit slimes dam failure: Overview and lessons learnt: Technical paper: Journal of South African Civil Engineering, Vol. 39, No. 3, pp. 11–15.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 263–279

279

Technical Note

Technical Note

Development of a Rapid System to Diagnose Ground Settlement

Development of a Rapid System to Diagnose Ground Settlement

JENNY L. WOLICKI1

JENNY L. WOLICKI1

7229 Dove Court, Parker, CO 80134

7229 Dove Court, Parker, CO 80134

PAUL M. SANTI

PAUL M. SANTI

Department of Geology and Geological Engineering, Colorado School of Mines, Golden, CO 80401

Department of Geology and Geological Engineering, Colorado School of Mines, Golden, CO 80401

BENJAMIN D. HAUGEN

BENJAMIN D. HAUGEN

14146 Denver West Parkway, Suite 200, Golden, CO 80401

14146 Denver West Parkway, Suite 200, Golden, CO 80401

JACQUELYN A. HAGBERY

JACQUELYN A. HAGBERY

9800 Mount Pyramid Court, Suite 330, Englewood, CO 80112

9800 Mount Pyramid Court, Suite 330, Englewood, CO 80112

ETHAN J. FABER

ETHAN J. FABER

5625 South Sycamore Street, Apartment 307, Littleton, CO 80120

5625 South Sycamore Street, Apartment 307, Littleton, CO 80120

STEPHEN N. SEMMENS

STEPHEN N. SEMMENS

1221 Illinois Street, Apartment 2E, Golden, CO 80401

1221 Illinois Street, Apartment 2E, Golden, CO 80401

HAYDEN E. BROWN

HAYDEN E. BROWN

912 12th Street, Golden, CO 80401

912 12th Street, Golden, CO 80401

Key Terms: Land Subsidence, Geologic Hazards, Engineering Geology, Geotechnical ABSTRACT Ground settlement causes extensive damage worldwide and has a significant impact on engineering design and mitigation. Identification of settlement origins can be challenging and expensive. The ability to identify the likely cause(s) of settlement using a rapid forensic analysis can reduce field time and constrain site investigation strategies, thereby reducing costs. The Rapid Settlement Diagnostic System (RSDS) has been developed as a preliminary tool for efficiently identifying likely sources of geology-related settlement. The four-step questionnaire in Microsoft Excel format enables personnel to quickly isolate potential causes of settlement using initial site observations and basic site knowledge. The first step distinguishes whether the observations involve geology-related settlement or another issue, such as structural instability. The second and third steps narrow down the list of 1 Corresponding

author email: jenny.wolicki@gmail.com.

likely settlement causes for the area of concern through a series of questions regarding the site conditions with answer options of “yes,” “no,” or “unknown.” The fourth step provides a final calculation and rank indicating the most likely cause(s) of settlement that should be investigated further through a detailed site investigation. The RSDS has been tested using historic cases of settlement with known causes to evaluate the effectiveness and ease of use of the tool. The results from the validation process have been used to revise the questions and scoring system in the RSDS in order to improve the usefulness and clarity of the tool.

INTRODUCTION Differential ground settlement, commonly referred to as “settlement,” is a process that causes extensive damage worldwide and has a significant impact on engineering design and mitigation. Identification of settlement origins (e.g., karst, groundwater withdrawal, liquefaction and so on) can be challenging and expensive, so the ability to diagnose the likely cause(s) of settlement using a rapid forensic analysis can reduce field

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

281

Key Terms: Land Subsidence, Geologic Hazards, Engineering Geology, Geotechnical ABSTRACT Ground settlement causes extensive damage worldwide and has a significant impact on engineering design and mitigation. Identification of settlement origins can be challenging and expensive. The ability to identify the likely cause(s) of settlement using a rapid forensic analysis can reduce field time and constrain site investigation strategies, thereby reducing costs. The Rapid Settlement Diagnostic System (RSDS) has been developed as a preliminary tool for efficiently identifying likely sources of geology-related settlement. The four-step questionnaire in Microsoft Excel format enables personnel to quickly isolate potential causes of settlement using initial site observations and basic site knowledge. The first step distinguishes whether the observations involve geology-related settlement or another issue, such as structural instability. The second and third steps narrow down the list of 1 Corresponding

author email: jenny.wolicki@gmail.com.

likely settlement causes for the area of concern through a series of questions regarding the site conditions with answer options of “yes,” “no,” or “unknown.” The fourth step provides a final calculation and rank indicating the most likely cause(s) of settlement that should be investigated further through a detailed site investigation. The RSDS has been tested using historic cases of settlement with known causes to evaluate the effectiveness and ease of use of the tool. The results from the validation process have been used to revise the questions and scoring system in the RSDS in order to improve the usefulness and clarity of the tool.

INTRODUCTION Differential ground settlement, commonly referred to as “settlement,” is a process that causes extensive damage worldwide and has a significant impact on engineering design and mitigation. Identification of settlement origins (e.g., karst, groundwater withdrawal, liquefaction and so on) can be challenging and expensive, so the ability to diagnose the likely cause(s) of settlement using a rapid forensic analysis can reduce field

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

281

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

time and constrain site investigation strategies, thereby reducing costs. On the identification of the most likely causes, a more detailed site investigation can be completed to further characterize the site, then specific mitigation techniques can be implemented to improve the site. The Rapid Settlement Diagnostic System (RSDS) has been developed as a preliminary tool for efficiently identifying likely sources of apparent ground settlement. The RSDS is a four-step questionnaire in Microsoft Excel format that can be completed with initial site observations and other existing information for the area of concern, including geologic maps and soil survey maps. Those who would benefit most from the RSDS are personnel trained in aspects of settlement hazards, but they do not need to be experts. The RSDS has been validated by testing the tool against cases of settlement with known causes and making adjustments to the questionnaire until acceptable results are obtained for each case. The purpose of this article is to provide a simple explanation of the format and scoring system of the tool, discuss the validation procedure and specific changes made, and provide recommendations to the user to optimize the tool.

FORMAT OF RSDS The RSDS was created by a team at the Colorado School of Mines. The tool considers 15 potential causes of settlement that were chosen based on their unique inherent properties, failure modes, and mitigation strategies: 1. Water withdrawal 2. Hydrocarbon withdrawal 3. Mines 4. Karst 5. Seismic liquefaction 6. Seismic settlement 7. Oxidation of peat 8. Dissolution 9. Piping 10. Hydrocompaction 11. Permafrost 12. Frost action 13. Clay consolidation 14. Expansive soil 15. Heaving bedrock Some of these geologic processes involve expansion rather than compaction but can cause the differential settlement of structures and are therefore included in the tool.

The RSDS comprises four systematic steps with a series of questions that may be answered “yes,” “no,” or 282

“unknown.” The first step (“Other Potential Types of Ground Movement”) distinguishes whether the observations at a site involve geology-related settlement or an alternative cause, such as civil structural instability or fault rupture, that results in deformation similar to settlement. The second step (“Settlement Type Ranking”) includes 13 general questions related to one or more of the 15 causes of geology-related settlement included in the tool. Once all of the questions in step 2 have been answered, a likelihood score is calculated for each settlement type based on the user’s answers. The third step (“Refining Settlement Cause”) yields correction factors for each likely cause of settlement that are multiplied by the score from step 2 to give the “Final Score” in the fourth step (“Final List of Expected Settlement Causes in Order of Likelihood”). The highest-ranked causes in step 4 should be considered when planning future site investigation strategies. Extensive technical literature is available for guidance in conducting additional analyses to confirm the findings, such as laboratory testing and subsurface investigations, found under the “Recommended Site Investigations” tab at the end of the form. Selection of mitigation techniques is an important final step that is outside the scope of the tool.

Figure 1A–D show the four steps of the questionnaire with example answers to illustrate how inputs and results are presented in the tool. Inputs were selected to show the variety of potential outputs that the user can expect to see, and they are not representative of a specific site.

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

time and constrain site investigation strategies, thereby reducing costs. On the identification of the most likely causes, a more detailed site investigation can be completed to further characterize the site, then specific mitigation techniques can be implemented to improve the site. The Rapid Settlement Diagnostic System (RSDS) has been developed as a preliminary tool for efficiently identifying likely sources of apparent ground settlement. The RSDS is a four-step questionnaire in Microsoft Excel format that can be completed with initial site observations and other existing information for the area of concern, including geologic maps and soil survey maps. Those who would benefit most from the RSDS are personnel trained in aspects of settlement hazards, but they do not need to be experts. The RSDS has been validated by testing the tool against cases of settlement with known causes and making adjustments to the questionnaire until acceptable results are obtained for each case. The purpose of this article is to provide a simple explanation of the format and scoring system of the tool, discuss the validation procedure and specific changes made, and provide recommendations to the user to optimize the tool.

FORMAT OF RSDS The RSDS was created by a team at the Colorado School of Mines. The tool considers 15 potential causes of settlement that were chosen based on their unique inherent properties, failure modes, and mitigation strategies:

EXPLANATION OF RSDS SCORING The scoring structure in the RSDS is different in each of the four steps. Step 1 is divided into five sections, each of which has 1–10 questions. For “Fault Rupture,” both questions must be answered as “yes” to be considered a possible cause. These two questions confirm that a fault exists and that it passes directly through the structure, both of which must be true in order for a structure to experience differential settlement due to fault rupture.

For “Construction/Structural Issues,” at least one of the two answers must be “yes” to be a possible cause.

Although the concerns addressed in these questions are not always associated with civil structural or construction issues, they can be indicators, so the civil structure should be investigated. Objective identification of mass movement is difficult to do from a surficial forensic analysis and thus is judged from the probability of observed features attributed to mass movements, such as visible toe bulges, scarps, and tension cracks. For “Mass Movement,” the number of answered “yes” questions must be greater than or equal to five to be considered a likely cause of deformation. The authors expect that at

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

1. Water withdrawal 2. Hydrocarbon withdrawal 3. Mines 4. Karst 5. Seismic liquefaction 6. Seismic settlement 7. Oxidation of peat 8. Dissolution 9. Piping 10. Hydrocompaction 11. Permafrost 12. Frost action 13. Clay consolidation 14. Expansive soil 15. Heaving bedrock Some of these geologic processes involve expansion rather than compaction but can cause the differential settlement of structures and are therefore included in the tool.

The RSDS comprises four systematic steps with a series of questions that may be answered “yes,” “no,” or 282

“unknown.” The first step (“Other Potential Types of Ground Movement”) distinguishes whether the observations at a site involve geology-related settlement or an alternative cause, such as civil structural instability or fault rupture, that results in deformation similar to settlement. The second step (“Settlement Type Ranking”) includes 13 general questions related to one or more of the 15 causes of geology-related settlement included in the tool. Once all of the questions in step 2 have been answered, a likelihood score is calculated for each settlement type based on the user’s answers. The third step (“Refining Settlement Cause”) yields correction factors for each likely cause of settlement that are multiplied by the score from step 2 to give the “Final Score” in the fourth step (“Final List of Expected Settlement Causes in Order of Likelihood”). The highest-ranked causes in step 4 should be considered when planning future site investigation strategies. Extensive technical literature is available for guidance in conducting additional analyses to confirm the findings, such as laboratory testing and subsurface investigations, found under the “Recommended Site Investigations” tab at the end of the form. Selection of mitigation techniques is an important final step that is outside the scope of the tool.

Figure 1A–D show the four steps of the questionnaire with example answers to illustrate how inputs and results are presented in the tool. Inputs were selected to show the variety of potential outputs that the user can expect to see, and they are not representative of a specific site.

EXPLANATION OF RSDS SCORING The scoring structure in the RSDS is different in each of the four steps. Step 1 is divided into five sections, each of which has 1–10 questions. For “Fault Rupture,” both questions must be answered as “yes” to be considered a possible cause. These two questions confirm that a fault exists and that it passes directly through the structure, both of which must be true in order for a structure to experience differential settlement due to fault rupture.

For “Construction/Structural Issues,” at least one of the two answers must be “yes” to be a possible cause.

Although the concerns addressed in these questions are not always associated with civil structural or construction issues, they can be indicators, so the civil structure should be investigated. Objective identification of mass movement is difficult to do from a surficial forensic analysis and thus is judged from the probability of observed features attributed to mass movements, such as visible toe bulges, scarps, and tension cracks. For “Mass Movement,” the number of answered “yes” questions must be greater than or equal to five to be considered a likely cause of deformation. The authors expect that at

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

Development of a Rapid System to Diagnose Ground Settlement

Development of a Rapid System to Diagnose Ground Settlement

Figure 1. (A) Rapid Settlement Diagnostic System (RSDS) step 1: example inputs and results. (B) RSDS step 2: example inputs and results.

(C) RSDS step 3: example inputs and results. (D) RSDS step 4: example inputs and results.

Figure 1. (A) Rapid Settlement Diagnostic System (RSDS) step 1: example inputs and results. (B) RSDS step 2: example inputs and results.

(C) RSDS step 3: example inputs and results. (D) RSDS step 4: example inputs and results.

least five of the 10 listed features will likely be observed at a site if significant mass movement has occurred. For “Landfill Subsidence” and “Poorly Compacted Soils,” the question must be answered as “yes” to be a possible cause. If a red “yes” appears next to the potential cause of settlement-like observations, then it should be

least five of the 10 listed features will likely be observed at a site if significant mass movement has occurred. For “Landfill Subsidence” and “Poorly Compacted Soils,” the question must be answered as “yes” to be a possible cause. If a red “yes” appears next to the potential cause of settlement-like observations, then it should be

investigated further. If a green “no” appears, then it is not a likely cause.

The questions in step 2 focus on ranking the likelihood of each cause of geology-related settlement. This step consists of 13 questions associated with one or more of the 15 different types of settlement. Answering

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

283

investigated further. If a green “no” appears, then it is not a likely cause.

The questions in step 2 focus on ranking the likelihood of each cause of geology-related settlement. This step consists of 13 questions associated with one or more of the 15 different types of settlement. Answering

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

283

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

Figure 1. (continued)

Figure 1. (continued)

“yes” or “unknown” to a question increases the score for certain types, while answering “no” does not affect it. Weighting of the questions is needed because each question applies to a different number of settlement types. Questions applying to fewer settlement types are 284

worth more because they are a stronger indicator for those specific settlement types. For each question, the total number of settlement types that are associated with an answer of “yes” are summed, then the reciprocal of this sum is calculated. This reciprocal represents

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

“yes” or “unknown” to a question increases the score for certain types, while answering “no” does not affect it. Weighting of the questions is needed because each question applies to a different number of settlement types. Questions applying to fewer settlement types are 284

worth more because they are a stronger indicator for those specific settlement types. For each question, the total number of settlement types that are associated with an answer of “yes” are summed, then the reciprocal of this sum is calculated. This reciprocal represents

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

Development of a Rapid System to Diagnose Ground Settlement

Development of a Rapid System to Diagnose Ground Settlement

Figure 1. (continued)

Figure 1. (continued)

the amount that each “yes” for the particular question is worth. In addition, some questions have the potential to fully eliminate the settlement type. For example, one question in step 2 addresses the occurrence of an earthquake. If this question is answered “no,” then the scores for seismic settlement and seismic liquefaction are reduced to 0.0. After all of the 13 questions are answered, a score for each settlement type is calculated by summing the weighted score for each. Each score is then normalized by dividing the earned score by the maximum potential score for that settlement type so that all of the scores can be directly compared. Each weighted and normalized total score is then multiplied by 100 to give the “Score” in step 2 a value ranging from 0.0 to 100.0. The various settlement types are automatically ranked in the “Rank” column based on the total “Score” values relative to one another.

Step 3 focuses on refining the likelihood of the causes of settlement. The top five ranked settlement types from step 2 are automatically highlighted in red, and their associated questions should be answered in step 3. Other settlement types not ranked in the top five can still be answered and analyzed. Questions in step 3 are more detailed and applicable only to their respective settle-

ment types. Again, “yes” and “unknown” answers are awarded one point, and a “no” answer earns zero points (distinguishing between the “yes” and “unknown” answers should be done in a later sensitivity analysis, described below). Each set of questions, ranging from one to seven in number, falls specifically under one of the 15 identified settlement types. For each potential cause, the number of “yes” or “unknown” answers are summed and divided by the number of questions for that cause to give a correction factor between 0.00 and 1.00. Again, some questions have the potential to fully eliminate the settlement type (e.g., no presence of freezing temperatures in regard to the likelihood of frost action). Thus, the correction factor is reduced to 0.00 when these types of questions are answered “no.”

The score from step 2 is multiplied by the correction factor from step 3 to give the “Final Score” between 0.0 and 100.0 in step 4. The ranks indicate the most likely cause(s) of settlement that should be explored further through a detailed site investigation (e.g., drilling, geophysics, or mapping). A more detailed explanation of the scoring scheme is included in the RSDS user’s manual (available from the authors). The user’s manual

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

285

the amount that each “yes” for the particular question is worth. In addition, some questions have the potential to fully eliminate the settlement type. For example, one question in step 2 addresses the occurrence of an earthquake. If this question is answered “no,” then the scores for seismic settlement and seismic liquefaction are reduced to 0.0. After all of the 13 questions are answered, a score for each settlement type is calculated by summing the weighted score for each. Each score is then normalized by dividing the earned score by the maximum potential score for that settlement type so that all of the scores can be directly compared. Each weighted and normalized total score is then multiplied by 100 to give the “Score” in step 2 a value ranging from 0.0 to 100.0. The various settlement types are automatically ranked in the “Rank” column based on the total “Score” values relative to one another.

Step 3 focuses on refining the likelihood of the causes of settlement. The top five ranked settlement types from step 2 are automatically highlighted in red, and their associated questions should be answered in step 3. Other settlement types not ranked in the top five can still be answered and analyzed. Questions in step 3 are more detailed and applicable only to their respective settle-

ment types. Again, “yes” and “unknown” answers are awarded one point, and a “no” answer earns zero points (distinguishing between the “yes” and “unknown” answers should be done in a later sensitivity analysis, described below). Each set of questions, ranging from one to seven in number, falls specifically under one of the 15 identified settlement types. For each potential cause, the number of “yes” or “unknown” answers are summed and divided by the number of questions for that cause to give a correction factor between 0.00 and 1.00. Again, some questions have the potential to fully eliminate the settlement type (e.g., no presence of freezing temperatures in regard to the likelihood of frost action). Thus, the correction factor is reduced to 0.00 when these types of questions are answered “no.”

The score from step 2 is multiplied by the correction factor from step 3 to give the “Final Score” between 0.0 and 100.0 in step 4. The ranks indicate the most likely cause(s) of settlement that should be explored further through a detailed site investigation (e.g., drilling, geophysics, or mapping). A more detailed explanation of the scoring scheme is included in the RSDS user’s manual (available from the authors). The user’s manual

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

285

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

Figure 1. (continued)

Figure 1. (continued)

provides additional information regarding the creation of the tool, its purpose, and the justification for each question in the tool.

VALIDATION Questions in the RSDS are based on engineering geology principles and published literature on the causes and mechanisms of settlement; however, the tool required validation. The RSDS was tested against historic cases of settlement with known causes to evaluate the effectiveness and ease of use of the tool. Cases were selected from published books, journals, and conference proceedings that had sufficient site information to answer the questions in the form. Most cases were specific sites, such as a single building or development, 286

but some regional-scale cases were used when specific sites did not have sufficient information. For example, west-central Florida was selected as the case for karst settlement, and the questions were answered based on regional observations. In some cases, it was necessary to refer to additional documents to find general site information that was not included in the main reference.

Also, in rare cases, it was necessary to use the “unknown” answer option or conduct a sensitivity analysis for questions that could not be answered confidently.

The accuracy of the tool to properly identify each historic case was scored with a letter grade of A, B, or C. An A grade was awarded if the known cause of settlement was given a final rank of 1 in the RSDS. A B grade was awarded if the known cause of settlement was given a final rank of 2 in the RSDS. A C grade was

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

provides additional information regarding the creation of the tool, its purpose, and the justification for each question in the tool.

VALIDATION Questions in the RSDS are based on engineering geology principles and published literature on the causes and mechanisms of settlement; however, the tool required validation. The RSDS was tested against historic cases of settlement with known causes to evaluate the effectiveness and ease of use of the tool. Cases were selected from published books, journals, and conference proceedings that had sufficient site information to answer the questions in the form. Most cases were specific sites, such as a single building or development, 286

but some regional-scale cases were used when specific sites did not have sufficient information. For example, west-central Florida was selected as the case for karst settlement, and the questions were answered based on regional observations. In some cases, it was necessary to refer to additional documents to find general site information that was not included in the main reference.

Also, in rare cases, it was necessary to use the “unknown” answer option or conduct a sensitivity analysis for questions that could not be answered confidently.

The accuracy of the tool to properly identify each historic case was scored with a letter grade of A, B, or C. An A grade was awarded if the known cause of settlement was given a final rank of 1 in the RSDS. A B grade was awarded if the known cause of settlement was given a final rank of 2 in the RSDS. A C grade was

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

287

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

*The correct cause of settlement was not in the top five ranks after step 2 questions were answered; however, all questions were answered in step 3 to calculate the final ranks, allowing the final score to rank higher.

If two grades were awarded (i.e., B/A), this indicates that a sensitivity analysis was performed for one or more question(s), so the two grades correspond to the two iterations. This was done when a question could not be answered “yes” or “no” with certainty.

Tihansky (1999); Tihansky and Knochenmus (2001) Coplin and Galloway (1999) Coplin and Galloway (1999); Mace et al. (2006) Quigley et al. (2013) Castle et al. (1974) Noe (1997) Ingebritsen and Ikehara (1999) Siekmeier et al. (1998) Lovekin and Higgings (2003); White (2003) White and Greenman (2008) Thompson and Mechling (2016) Alfaro et al. (2009) Wang et al. (2015) Page (1998) Vessely et al. (2003) A A B A A A A A A B*/A A B A B A A A B A A A A A A C* B B A B B* West-central Florida Goose Creek oil field, Houston, TX Brownwood subdivision, Baytown, TX 2010 M 7.1 Darfield Earthquake, Christchurch, New Zealand San Fernando Earthquake of February 9, 1971, San Fernando, CA Colorado Front Range Sacramento-San Joaquin Delta, CA Cass County, NE Mid Valley Baptist Church, Roaring Fork Corridor, CO Southwestern Colorado Glenwood Springs, CO Northern Manitoba, Canada (NLCY) of Qinghai-Tibet Railway, China Sacramento Valley, CA Denver area, CO Karst Hydrocarbon withdrawal Water withdrawal Seismic liquefaction Seismic settlement Heaving bedrock Oxidation of peat Mines Dissolution Piping Hydrocompaction Permafrost Frost action Clay consolidation Expansive soil

Initial Grade Final Grade Case History Location Origin of Settlement

Table 1. Summary of historic cases.

awarded if the known cause of settlement was given a rank worse than 2 in the RSDS. These results were used to identify which questions in the tool needed to be removed or revised and if any aspects of the score system needed to be adjusted. After these changes were made, all the cases were tested again and given a second grade. This process was repeated until all cases earned a grade of A or B. Table 1 summarizes the 15 historic cases that were selected, the site locations, the initial grades prior to making any changes to the RSDS, the final grades after making all changes, and the references for each case. One case was selected for each of the 15 unique causes of settlement included in the tool. The answers to all step 2 and step 3 questions for all of the tested cases are summarized in Table 2.

Changes were based primarily on cases that earned lower initial grades; however, some changes were made simply to improve clarity and reduce user error. Many questions in the original version of the RSDS were not changed because they proved to be useful on completing the validation procedure. The rationale for these questions is not given in this article because the RSDS user’s manual contains this information. The changes that were made to the original version of the tool and the basis for these changes are described as follows.

An additional alternative cause was added to step 1 to account for apparent settlement due to poorly compacted fills. The compaction of granular fills under an applied load is different than clay consolidation and hydrocompaction, so it was necessary to include this as a different cause. Settlement due to poorly compacted soils may be considered a geologic process; however, it was included in step 1 because the process is a result of poor construction techniques and not a natural occurrence.

Some questions in steps 2 and 3 were revised to be either more specific or more general in order to reduce user error. For example, it was specified that shallow bedrock is less than or equal to 50 m deep in the fourth question in step 2 for the purpose of the tool. “Shallow” may be a subjective term, and it can be helpful to users to have a more objective value to compare to the conditions at their site of interest. The purpose of this question in the RSDS is to assess the likelihood of settlement due to mines, karst, and heaving bedrock. Settlement due to karst and mines can still occur when overburden thicknesses are approximately 50 m, so this was chosen as a guideline for the RSDS user (Turney, 1985; Siekmeier et al., 1998; and Coplin and Galloway, 1999). In step 3, the first question under clay consolidation–type settlement originally asked whether a new load was applied more than 6 months ago. Clay consolidation is a slow process that occurs for months or years following the application of a new load; however, the rate of consolidation is depen-

References

Development of a Rapid System to Diagnose Ground Settlement *The correct cause of settlement was not in the top five ranks after step 2 questions were answered; however, all questions were answered in step 3 to calculate the final ranks, allowing the final score to rank higher.

If two grades were awarded (i.e., B/A), this indicates that a sensitivity analysis was performed for one or more question(s), so the two grades correspond to the two iterations. This was done when a question could not be answered “yes” or “no” with certainty.

Tihansky (1999); Tihansky and Knochenmus (2001) Coplin and Galloway (1999) Coplin and Galloway (1999); Mace et al. (2006) Quigley et al. (2013) Castle et al. (1974) Noe (1997) Ingebritsen and Ikehara (1999) Siekmeier et al. (1998) Lovekin and Higgings (2003); White (2003) White and Greenman (2008) Thompson and Mechling (2016) Alfaro et al. (2009) Wang et al. (2015) Page (1998) Vessely et al. (2003) A A B A A A A A A B*/A A B A B A A A B A A A A A A C* B B A B B* West-central Florida Goose Creek oil field, Houston, TX Brownwood subdivision, Baytown, TX 2010 M 7.1 Darfield Earthquake, Christchurch, New Zealand San Fernando Earthquake of February 9, 1971, San Fernando, CA Colorado Front Range Sacramento-San Joaquin Delta, CA Cass County, NE Mid Valley Baptist Church, Roaring Fork Corridor, CO Southwestern Colorado Glenwood Springs, CO Northern Manitoba, Canada (NLCY) of Qinghai-Tibet Railway, China Sacramento Valley, CA Denver area, CO Karst Hydrocarbon withdrawal Water withdrawal Seismic liquefaction Seismic settlement Heaving bedrock Oxidation of peat Mines Dissolution Piping Hydrocompaction Permafrost Frost action Clay consolidation Expansive soil

Initial Grade Final Grade Case History Location Origin of Settlement

Table 1. Summary of historic cases.

awarded if the known cause of settlement was given a rank worse than 2 in the RSDS. These results were used to identify which questions in the tool needed to be removed or revised and if any aspects of the score system needed to be adjusted. After these changes were made, all the cases were tested again and given a second grade. This process was repeated until all cases earned a grade of A or B. Table 1 summarizes the 15 historic cases that were selected, the site locations, the initial grades prior to making any changes to the RSDS, the final grades after making all changes, and the references for each case. One case was selected for each of the 15 unique causes of settlement included in the tool. The answers to all step 2 and step 3 questions for all of the tested cases are summarized in Table 2.

Changes were based primarily on cases that earned lower initial grades; however, some changes were made simply to improve clarity and reduce user error. Many questions in the original version of the RSDS were not changed because they proved to be useful on completing the validation procedure. The rationale for these questions is not given in this article because the RSDS user’s manual contains this information. The changes that were made to the original version of the tool and the basis for these changes are described as follows.

An additional alternative cause was added to step 1 to account for apparent settlement due to poorly compacted fills. The compaction of granular fills under an applied load is different than clay consolidation and hydrocompaction, so it was necessary to include this as a different cause. Settlement due to poorly compacted soils may be considered a geologic process; however, it was included in step 1 because the process is a result of poor construction techniques and not a natural occurrence.

Some questions in steps 2 and 3 were revised to be either more specific or more general in order to reduce user error. For example, it was specified that shallow bedrock is less than or equal to 50 m deep in the fourth question in step 2 for the purpose of the tool. “Shallow” may be a subjective term, and it can be helpful to users to have a more objective value to compare to the conditions at their site of interest. The purpose of this question in the RSDS is to assess the likelihood of settlement due to mines, karst, and heaving bedrock. Settlement due to karst and mines can still occur when overburden thicknesses are approximately 50 m, so this was chosen as a guideline for the RSDS user (Turney, 1985; Siekmeier et al., 1998; and Coplin and Galloway, 1999). In step 3, the first question under clay consolidation–type settlement originally asked whether a new load was applied more than 6 months ago. Clay consolidation is a slow process that occurs for months or years following the application of a new load; however, the rate of consolidation is depen-

References

Development of a Rapid System to Diagnose Ground Settlement

287

288

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

Table 2. Summary of step 2 and step 3 answers for historic cases.

Table 2. Summary of step 2 and step 3 answers for historic cases.

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281â&#x20AC;&#x201C;291

288

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281â&#x20AC;&#x201C;291

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281â&#x20AC;&#x201C;291

Table 2. Continued.

Development of a Rapid System to Diagnose Ground Settlement

Table 2. Continued.

Development of a Rapid System to Diagnose Ground Settlement

289

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281â&#x20AC;&#x201C;291

289

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

dent on specific soil properties, and some degree of consolidation is likely to occur sooner than 6 months after construction (Salem and El-Sherbiny, 2014).

Therefore, the question was revised to ask if a construction load had been applied to the site. In this case, it was appropriate to be less specific due to the variable behavior of clay soils.

Two questions were combined in step 3 because the individual questions seemed to have too much of an impact on the final scores. One question under hydrocompaction-type settlement asked if bedrock or soil is clay or silt rich, and another question asked if there are mine tailings in the vicinity. Both of these conditions can contribute to hydrocompaction, but it is not necessary that both conditions exist at the site.

Therefore, the two questions were combined into a single question that asks if either condition exists.

One question in step 3 was removed because it did not encompass the range of properties contributing to a particular type of settlement. The original version of the tool included a question under dissolution-type settlement asking if the shallow soils are composed of silts or sands with minimal clay content. Gypsiferous soils are often categorized as silty sands or sandy silts, but this is not always the case. The dissolution case that was tested involved a site located on an alluvial fan composed of material classified as clayey silty sand and clayey sand (Lovekin and Higgings, 2003). This example of a gypsiferous soil with significant clay content provided reason to remove the question from the tool.

Finally, some aspects of the scoring scheme were adjusted. In at least one of the tested cases, seismic settlement was ranked as one of the most likely causes of settlement, even if the site had not experienced an earthquake. The weight of the question in step 2 regarding the occurrence of earthquakes was revised so that if the question is answered “no,” then the total score for seismic settlement and seismic liquefaction will automatically reduce to 0.00 because these two causes of settlement require a seismic event in order to occur.

The RSDS was tested against cases of settlement with known causes in order to validate the tool, and adjustments were made to questions and weightings to improve its usefulness. Further validation is important as feedback from users becomes available.

LIMITATIONS AND SENSITIVITY ANALYSIS The RSDS is a tool that can be used to characterize the most likely causes of settlement at a site based on preliminary site observations and information regarding the site geology, hydrology, and construction history. The tool may not be suitable for sites with very limited information because its effectiveness is depen290

dent on the accuracy and completeness of the answers to questions in each step. In addition, the tool is a cursory diagnostic system; thus, the final scores and ranks are not a definitive diagnosis for the settlement problem on-site.

Users should also be careful with the use of the “unknown” answer. An “unknown” answer has the same value as a “yes” answer, which produces a conservative answer, but can lead to false-positive and false-negative results. While the “unknown” option serves as a temporary placeholder for uncertain questions, it should not be used when calculating the final scores and ranks.

It is recommended that the user conduct a sensitivity analysis for questions that cannot be answered with certainty by answering “yes” in one iteration and “no” in a second iteration of the RSDS instead of using the “unknown” option. The importance of this sensitivity analysis is shown in the case of settlement caused by piping, where the correct cause of settlement was not initially identified in the top five answers after completing the questions in step 2. The identification hinges on the question in step 2 that asks, “Are sand boils, sand volcanoes, or other vented sediment features present?”

This question was originally answered “no” because this observation was not discussed in the references used for the case. However, this lack of information does not necessarily mean that vented features were not present at the site. Therefore, a sensitivity analysis was conducted by completing a second iteration of the questionnaire for each case and answering “yes” for the question regarding vented sediment features, as seen in Table 2. This increased the total scores for piping in step 2 so that it ranked in the top five potential causes, and after answering the questions in step 3, piping earned the highest final score. This example highlights the importance of visiting the site of interest in order to make observations that cannot always be gathered from maps and historic documents. It also shows the benefit of conducting a sensitivity analysis for questions that cannot be answered with certainty.

CONCLUSIONS Initial trials and calibration have shown the RSDS to be an accurate, rapid, and versatile tool for preliminary identification of the likely causes of ground settlement at a site that has experienced movement or structural damage. The tool was tested against a number of sites with known settlement causes as a first validation step, and then the tool was used by a group of technical personnel to further refine its accuracy. Users are cautioned to recognize the tool’s limits as a preliminary diagnostic evaluation and to conduct sensitivity analysis for questions that cannot be answered with confidence.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

Wolicki, Santi, Haugen, Hagbery, Faber, Semmens, and Brown

dent on specific soil properties, and some degree of consolidation is likely to occur sooner than 6 months after construction (Salem and El-Sherbiny, 2014).

Therefore, the question was revised to ask if a construction load had been applied to the site. In this case, it was appropriate to be less specific due to the variable behavior of clay soils.

Two questions were combined in step 3 because the individual questions seemed to have too much of an impact on the final scores. One question under hydrocompaction-type settlement asked if bedrock or soil is clay or silt rich, and another question asked if there are mine tailings in the vicinity. Both of these conditions can contribute to hydrocompaction, but it is not necessary that both conditions exist at the site.

Therefore, the two questions were combined into a single question that asks if either condition exists.

One question in step 3 was removed because it did not encompass the range of properties contributing to a particular type of settlement. The original version of the tool included a question under dissolution-type settlement asking if the shallow soils are composed of silts or sands with minimal clay content. Gypsiferous soils are often categorized as silty sands or sandy silts, but this is not always the case. The dissolution case that was tested involved a site located on an alluvial fan composed of material classified as clayey silty sand and clayey sand (Lovekin and Higgings, 2003). This example of a gypsiferous soil with significant clay content provided reason to remove the question from the tool.

Finally, some aspects of the scoring scheme were adjusted. In at least one of the tested cases, seismic settlement was ranked as one of the most likely causes of settlement, even if the site had not experienced an earthquake. The weight of the question in step 2 regarding the occurrence of earthquakes was revised so that if the question is answered “no,” then the total score for seismic settlement and seismic liquefaction will automatically reduce to 0.00 because these two causes of settlement require a seismic event in order to occur.

The RSDS was tested against cases of settlement with known causes in order to validate the tool, and adjustments were made to questions and weightings to improve its usefulness. Further validation is important as feedback from users becomes available.

LIMITATIONS AND SENSITIVITY ANALYSIS The RSDS is a tool that can be used to characterize the most likely causes of settlement at a site based on preliminary site observations and information regarding the site geology, hydrology, and construction history. The tool may not be suitable for sites with very limited information because its effectiveness is depen290

dent on the accuracy and completeness of the answers to questions in each step. In addition, the tool is a cursory diagnostic system; thus, the final scores and ranks are not a definitive diagnosis for the settlement problem on-site.

Users should also be careful with the use of the “unknown” answer. An “unknown” answer has the same value as a “yes” answer, which produces a conservative answer, but can lead to false-positive and false-negative results. While the “unknown” option serves as a temporary placeholder for uncertain questions, it should not be used when calculating the final scores and ranks.

It is recommended that the user conduct a sensitivity analysis for questions that cannot be answered with certainty by answering “yes” in one iteration and “no” in a second iteration of the RSDS instead of using the “unknown” option. The importance of this sensitivity analysis is shown in the case of settlement caused by piping, where the correct cause of settlement was not initially identified in the top five answers after completing the questions in step 2. The identification hinges on the question in step 2 that asks, “Are sand boils, sand volcanoes, or other vented sediment features present?”

This question was originally answered “no” because this observation was not discussed in the references used for the case. However, this lack of information does not necessarily mean that vented features were not present at the site. Therefore, a sensitivity analysis was conducted by completing a second iteration of the questionnaire for each case and answering “yes” for the question regarding vented sediment features, as seen in Table 2. This increased the total scores for piping in step 2 so that it ranked in the top five potential causes, and after answering the questions in step 3, piping earned the highest final score. This example highlights the importance of visiting the site of interest in order to make observations that cannot always be gathered from maps and historic documents. It also shows the benefit of conducting a sensitivity analysis for questions that cannot be answered with certainty.

CONCLUSIONS Initial trials and calibration have shown the RSDS to be an accurate, rapid, and versatile tool for preliminary identification of the likely causes of ground settlement at a site that has experienced movement or structural damage. The tool was tested against a number of sites with known settlement causes as a first validation step, and then the tool was used by a group of technical personnel to further refine its accuracy. Users are cautioned to recognize the tool’s limits as a preliminary diagnostic evaluation and to conduct sensitivity analysis for questions that cannot be answered with confidence.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

Development of a Rapid System to Diagnose Ground Settlement

The RSDS Microsoft Excel file and user’s manual are available by contacting the authors.

REFERENCES ALFARO, M. C.; CIRO, G. A.; THIESSEN, K. J.; AND NG, T., 2009, Case study of degrading permafrost beneath a road embankment: Journal of Cold Regions Engineering, Vol. 23, No. 3, pp. 93–111.

CASTLE, R. O.; ALT, J. N.; SAVAGE, J. C.; AND BALAZS, E. I., 1974, Elevation changes preceding the San Fernando Earthquake of February 9, 1971: Geology, Vol. 2, pp. 61–66.

COPLIN, L. S. AND GALLOWAY, D., 1999, Houston-Galveston, Texas: Land Subsidence in the United States: U.S. Geological Survey Circular 1182, pp. 35–48.

INGEBRITSEN, S. E. AND IKEHARA, M. E., 1999, Sacramento-San Joaquin Delta: Land Subsidence in the United States: U.S. Geological Survey Circular 1182, pp. 83–94.

LOVEKIN, J. R. AND HIGGINGS, J. D., 2003, Major Geologic Hazards Along the Roaring Fork River Near Glenwood Springs, Colorado: Association of Engineering Geologists, Special Publication No. 15, Colorado Geological Survey Special Publication 55.

MACE, R. E.; DAVIDSON, S. C.; ANGLE, E. S.; AND MULLICAN, W. F., III, 2006, Aquifers of the Gulf Coast of Texas: Texas Water Development Board Report 365.

NOE, D. C., 1997, Heaving-bedrock hazards, mitigation, and landuse policy: Front Range Piedmont, Colorado: Environmental Geosciences, Vol. 4, No. 2, pp. 48–57.

PAGE, R. W., 1998, A compressible diatomaceous clay, Sacramento Valley, CA. In Borchers, J. W. (Editor), Land Subsidence Case Studies and Current Research: Proceedings of The Dr. Joseph F. Poland Symposium on Land Subsidence: U.S. Geological Survey, Reston, VA, pp. 81–88.

QUIGLEY, M. C.; BASTIN, S.; AND BRADLEY, B. A., 2013, Recurrent liquefaction in Christchurch, New Zealand, during the Canterbury Earthquake sequence: Geology, Vol. 41, pp. 419–422.

SALEM, M. AND EL-SHERBINY, R., 2014, Comparison of measured and calculated consolidation settlements of thick undercon-

Development of a Rapid System to Diagnose Ground Settlement

solidated clay: Alexandria Engineering Journal, Vol. 53, pp.

107–117.

SIEKMEIER, J. A.; POWELL, L. R.; AND TRIPLETT, T. L., 1998, Case study of sinkholes over a Nebraska limestone mine. In Borchers, J. W. (Editor), Land Subsidence Case Studies and Current Research: Proceedings of The Dr. Joseph F. Poland Symposium on Land Subsidence: U.S. Geological Survey, Reston, VA, pp. 303–308.

THOMPSON, R. AND MECHLING, J., 2016, Geotechnical Engineering for the Remediation of Structures on Collapsing Soils: Geotechnical and Structural Engineering Congress 2016, p.

735–746.

TIHANSKY, A. B., 1999, Sinkholes, West-Central Florida: Land Subsidence in the United States: U.S. Geological Survey Circular 1182, pp. 121–140.

TIHANSKY, A. B. AND KNOCHENMUS, L. A., 2001, Karst features and hydrogeology in west-central Florida—A field perspective. In Kuniansky, E. L. (Editor), U.S. Geological Survey Karst Interest Group Proceedings, pp. 198–211.

THOMPSON, R. AND MECHLING, J., 2016, Geotechnical Engineering for the Remediation of Structures on Collapsing Soils: Geotechnical and Structural Engineering Congress 2016, p.

735–746.

TURNEY, J. E., 1985, Subsidence Above Inactive Coal Mines: Information for the Homeowner: Colorado Geological Survey Special Publication 26.

VESSELY, M. J.; CUSHMAN, D. A.; AND FLANAGAN, K. P., 2003, Case Study of the Effect of Transient Wetting for Structures on Expansive Soil and Bedrock: Association of Engineering Geologists, Special Publication No. 15, Colorado Geological Survey Special Publication 55.

WANG, T., LIU, Y., YAN, H., AND XU, L., 2015, An experimental study on the mechanical properties of silty soils under repeated freeze–thaw cycles: Cold Regions Science and Technology, Vol. 112, pp. 51–65.

WHITE, J. L., 2003, Evaporite Karst Subsidence Hazards in Colorado: Association of Engineering Geologists, Special Publication No. 15, Colorado Geological Survey Special Publication 55.

WHITE, J. L., AND GREENMAN, C., 2008, Collapsible Soils in Colorado: Colorado Geological Survey, Engineering Geology 14.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

291

The RSDS Microsoft Excel file and user’s manual are available by contacting the authors.

REFERENCES ALFARO, M. C.; CIRO, G. A.; THIESSEN, K. J.; AND NG, T., 2009, Case study of degrading permafrost beneath a road embankment: Journal of Cold Regions Engineering, Vol. 23, No. 3, pp. 93–111.

CASTLE, R. O.; ALT, J. N.; SAVAGE, J. C.; AND BALAZS, E. I., 1974, Elevation changes preceding the San Fernando Earthquake of February 9, 1971: Geology, Vol. 2, pp. 61–66.

COPLIN, L. S. AND GALLOWAY, D., 1999, Houston-Galveston, Texas: Land Subsidence in the United States: U.S. Geological Survey Circular 1182, pp. 35–48.

INGEBRITSEN, S. E. AND IKEHARA, M. E., 1999, Sacramento-San Joaquin Delta: Land Subsidence in the United States: U.S. Geological Survey Circular 1182, pp. 83–94.

LOVEKIN, J. R. AND HIGGINGS, J. D., 2003, Major Geologic Hazards Along the Roaring Fork River Near Glenwood Springs, Colorado: Association of Engineering Geologists, Special Publication No. 15, Colorado Geological Survey Special Publication 55.

MACE, R. E.; DAVIDSON, S. C.; ANGLE, E. S.; AND MULLICAN, W. F., III, 2006, Aquifers of the Gulf Coast of Texas: Texas Water Development Board Report 365.

NOE, D. C., 1997, Heaving-bedrock hazards, mitigation, and landuse policy: Front Range Piedmont, Colorado: Environmental Geosciences, Vol. 4, No. 2, pp. 48–57.

PAGE, R. W., 1998, A compressible diatomaceous clay, Sacramento Valley, CA. In Borchers, J. W. (Editor), Land Subsidence Case Studies and Current Research: Proceedings of The Dr. Joseph F. Poland Symposium on Land Subsidence: U.S. Geological Survey, Reston, VA, pp. 81–88.

QUIGLEY, M. C.; BASTIN, S.; AND BRADLEY, B. A., 2013, Recurrent liquefaction in Christchurch, New Zealand, during the Canterbury Earthquake sequence: Geology, Vol. 41, pp. 419–422.

SALEM, M. AND EL-SHERBINY, R., 2014, Comparison of measured and calculated consolidation settlements of thick undercon-

solidated clay: Alexandria Engineering Journal, Vol. 53, pp.

107–117.

SIEKMEIER, J. A.; POWELL, L. R.; AND TRIPLETT, T. L., 1998, Case study of sinkholes over a Nebraska limestone mine. In Borchers, J. W. (Editor), Land Subsidence Case Studies and Current Research: Proceedings of The Dr. Joseph F. Poland Symposium on Land Subsidence: U.S. Geological Survey, Reston, VA, pp. 303–308.

THOMPSON, R. AND MECHLING, J., 2016, Geotechnical Engineering for the Remediation of Structures on Collapsing Soils: Geotechnical and Structural Engineering Congress 2016, p.

735–746.

TIHANSKY, A. B., 1999, Sinkholes, West-Central Florida: Land Subsidence in the United States: U.S. Geological Survey Circular 1182, pp. 121–140.

TIHANSKY, A. B. AND KNOCHENMUS, L. A., 2001, Karst features and hydrogeology in west-central Florida—A field perspective. In Kuniansky, E. L. (Editor), U.S. Geological Survey Karst Interest Group Proceedings, pp. 198–211.

THOMPSON, R. AND MECHLING, J., 2016, Geotechnical Engineering for the Remediation of Structures on Collapsing Soils: Geotechnical and Structural Engineering Congress 2016, p.

735–746.

TURNEY, J. E., 1985, Subsidence Above Inactive Coal Mines: Information for the Homeowner: Colorado Geological Survey Special Publication 26.

VESSELY, M. J.; CUSHMAN, D. A.; AND FLANAGAN, K. P., 2003, Case Study of the Effect of Transient Wetting for Structures on Expansive Soil and Bedrock: Association of Engineering Geologists, Special Publication No. 15, Colorado Geological Survey Special Publication 55.

WANG, T., LIU, Y., YAN, H., AND XU, L., 2015, An experimental study on the mechanical properties of silty soils under repeated freeze–thaw cycles: Cold Regions Science and Technology, Vol. 112, pp. 51–65.

WHITE, J. L., 2003, Evaporite Karst Subsidence Hazards in Colorado: Association of Engineering Geologists, Special Publication No. 15, Colorado Geological Survey Special Publication 55.

WHITE, J. L., AND GREENMAN, C., 2008, Collapsible Soils in Colorado: Colorado Geological Survey, Engineering Geology 14.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 281–291

291

Groundwater Vulnerability Mapping in Urbanized Hydrological System Using Modified DRASTIC Model and Sensitivity Analysis

Groundwater Vulnerability Mapping in Urbanized Hydrological System Using Modified DRASTIC Model and Sensitivity Analysis

ISMAIL CHENINI* ADEL ZGHIBI MOHAMED HAYTHEM MSADDEK MAHMOUD DLALA

ISMAIL CHENINI* ADEL ZGHIBI MOHAMED HAYTHEM MSADDEK MAHMOUD DLALA

UR13ES26, Department of Geology, Faculty of Mathematical, Physical and Natural Sciences of Tunis, University of Tunis–El Manar, El Manar, 2092 Tunisia

UR13ES26, Department of Geology, Faculty of Mathematical, Physical and Natural Sciences of Tunis, University of Tunis–El Manar, El Manar, 2092 Tunisia

Key Terms: Groundwater Vulnerability, Urban Watershed, DRASTIC, Urban Hydrogeology, Sensitivity Analysis ABSTRACT The groundwater vulnerability assessment is normally applied to rural watersheds. However, urbanization modifies the hydrogeological processes. A modified DRASTIC model was adopted to establish a groundwater vulnerability map in an urbanized watershed.

The modified DRASTIC model incorporated a landuse map, and net recharge was calculated taking into account the specificity of the urban hydrogeological system. The application of the proposed approach to the Mannouba watershed demonstrates that the groundwater vulnerability indexes range from 80 to 165. The study’s results shows that 30 percent of the Mannouba watershed area has a high vulnerability index, 45 percent of the area has a medium index, and 25 percent of the study area has a low vulnerability index. To specify the effect of each DRASTIC factor on the calculated vulnerability index, sensitivity analyses were performed.

Land use, topography, and soil media have an important theoretical weight greater than the effective weight. The impact of the vadose zone factor has the most important effective weight and affects the vulnerability index. The sensitivity assessment explored the variation in vulnerability after thematic layer removal. In this analysis, the removal of hydraulic conductivity and impact of vadose zone modified the vulnerability index. Groundwater vulnerability assessment in urbanized watersheds is difficult and has to consider the impact of urbanization in the hydrogeological parameters.

*Corresponding author email: ismail.chenini@fst.utm.tn

INTRODUCTION The impact of urbanization on the hydrological and hydrogeological system is currently a major issue (Zilberbrand et al., 2001; Trowsdale and Lerner, 2007; and Chae et al., 2008). Urban development and pollution threaten the groundwater resources (Foster, 2007).

Recent research has focused on the evaluation of potential urban groundwater contamination, involving many hydrogeologists and researchers in environmental issues. Researchers have proposed different approaches for assessing the impact of urbanization on the groundwater quality. They have discussed various management tools and modeling of urban water resources (Murray and McCray, 2005). Various local studies have discussed the protection and contamination of urban water.

Initially developed for natural and rural hydrogeological systems, the groundwater vulnerability evaluation is a key tool for groundwater protection. The available methods are based on several parameters, such as topography, lithology, vadose zone, aquifer lithology, and hydrogeology. The vulnerability index takes into account the capacity of the natural environment to eliminate, dilute, or degrade a potential contaminant from the surface (Vrba and Zaporozec, 1994; Gogu and Dassargues, 2000). These methods were developed for natural hydrogeological contexts (Doerfliger and Zwahlen, 1998). However, the urbanization process induces a series of quantitative and qualitative impacts on the groundwater (Hwang et al., 2015), such as: (1) impermeabilization of the urban aquifer, (2) changes in the spatial configuration of the aquifer and variations in the recharge rate, (3) modification of flow regime, and (4) groundwater quality degradation.

Compared to natural hydrogeological systems, urbanized areas have specific hydrogeological characteristics. It is necessary to integrate the nature of the urban aquifer when mapping the vulnerability of groundwater and evaluating the contamination hazards caused by urbanization.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

293

Key Terms: Groundwater Vulnerability, Urban Watershed, DRASTIC, Urban Hydrogeology, Sensitivity Analysis ABSTRACT The groundwater vulnerability assessment is normally applied to rural watersheds. However, urbanization modifies the hydrogeological processes. A modified DRASTIC model was adopted to establish a groundwater vulnerability map in an urbanized watershed.

The modified DRASTIC model incorporated a landuse map, and net recharge was calculated taking into account the specificity of the urban hydrogeological system. The application of the proposed approach to the Mannouba watershed demonstrates that the groundwater vulnerability indexes range from 80 to 165. The study’s results shows that 30 percent of the Mannouba watershed area has a high vulnerability index, 45 percent of the area has a medium index, and 25 percent of the study area has a low vulnerability index. To specify the effect of each DRASTIC factor on the calculated vulnerability index, sensitivity analyses were performed.

Land use, topography, and soil media have an important theoretical weight greater than the effective weight. The impact of the vadose zone factor has the most important effective weight and affects the vulnerability index. The sensitivity assessment explored the variation in vulnerability after thematic layer removal. In this analysis, the removal of hydraulic conductivity and impact of vadose zone modified the vulnerability index. Groundwater vulnerability assessment in urbanized watersheds is difficult and has to consider the impact of urbanization in the hydrogeological parameters.

*Corresponding author email: ismail.chenini@fst.utm.tn

INTRODUCTION The impact of urbanization on the hydrological and hydrogeological system is currently a major issue (Zilberbrand et al., 2001; Trowsdale and Lerner, 2007; and Chae et al., 2008). Urban development and pollution threaten the groundwater resources (Foster, 2007).

Recent research has focused on the evaluation of potential urban groundwater contamination, involving many hydrogeologists and researchers in environmental issues. Researchers have proposed different approaches for assessing the impact of urbanization on the groundwater quality. They have discussed various management tools and modeling of urban water resources (Murray and McCray, 2005). Various local studies have discussed the protection and contamination of urban water.

Initially developed for natural and rural hydrogeological systems, the groundwater vulnerability evaluation is a key tool for groundwater protection. The available methods are based on several parameters, such as topography, lithology, vadose zone, aquifer lithology, and hydrogeology. The vulnerability index takes into account the capacity of the natural environment to eliminate, dilute, or degrade a potential contaminant from the surface (Vrba and Zaporozec, 1994; Gogu and Dassargues, 2000). These methods were developed for natural hydrogeological contexts (Doerfliger and Zwahlen, 1998). However, the urbanization process induces a series of quantitative and qualitative impacts on the groundwater (Hwang et al., 2015), such as: (1) impermeabilization of the urban aquifer, (2) changes in the spatial configuration of the aquifer and variations in the recharge rate, (3) modification of flow regime, and (4) groundwater quality degradation.

Compared to natural hydrogeological systems, urbanized areas have specific hydrogeological characteristics. It is necessary to integrate the nature of the urban aquifer when mapping the vulnerability of groundwater and evaluating the contamination hazards caused by urbanization.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

293

Chenini, Zghibi, Msaddek, and Dlala

The objective of the paper was first to establish a link between available methods for characterization of natural hydrogeological system vulnerability and urban hydrogeology by a methodological adaptation of the vulnerability mapping to urban areas.

Second, we aimed to analyze the impact of urbanization in a Tunisian aquifer. The first aim was addressed by reviewing the most used method for vulnerability mapping, with special care to the specificity of each method. The second objective was addressed by applying a modified DRASTIC model to establish a groundwater vulnerability map in the Mannouba catchment.

The novelty of this study was the use of a multi-method approach for vulnerability mapping of an urban hydrogeological system with specific modifications. This research paper attempts to optimize the vulnerability mapping by taking in account the hydrogeological heterogeneity induced by the urbanization of aquifer systems.

METHODS Literature Review The aim of this review is to present the current state of the art in research methods applied to groundwater vulnerability mapping. To evaluate the groundwater vulnerability, an assessment of the impact of urbanization on the hydrogeological parameters is required.

Concept of Groundwater Vulnerability There is no standard definition associated with groundwater vulnerability, which describes the relative facility for contamination of groundwater. Naturally, the physical environment provides the water resource with a protection from pollutant contamination. The concept of vulnerability in hydrogeology was used for the first time in 1960, and since 1980, it has been used frequently (Haertle, 1983; Aller et al., 1987; and Foster and Hirata, 1988). Actually, the term is commonly used all over the world. Several definitions of groundwater vulnerability have been proposed.

Related to the potential for water resource contamination, groundwater vulnerability is a non-measurable property of an aquifer that is based on the concept that some zones are more susceptible to pollution than others (Musekiwa and Majola, 2013). Always, vulnerability is associated with groundwater contamination.

Groundwater vulnerability has been defined in many ways (Haertle, 1983; Foster, 1987; and Foster and Hirata, 1988). Groundwater vulnerability is “an intrinsic property of a groundwater system that depends on the 294

sensitivity of that system to human and/or natural impacts” (Vrba and Zaporozec, 1994 p 121).

Popescu et al. (2008) defined vulnerability as the relative degree of perturbation and harm caused by human or environment. In hydrogeology, the groundwater vulnerability is described as the susceptibility of an aquifer to be contaminated by a pollutant induced by human anthropic activity (Liggett and Talwar, 2009).

Vulnerability maps generated by hydrogeologists are used by planners and decision makers. They are useful for groundwater resources management and land-use management. However, vulnerability maps are not a direct representation of water resources and their associated elements (piezometric levels, aquifer extension, aquifer hydrodynamic). The vulnerability is rather a set of derived parameters and an index calculated from elements such as slope and soil type. Thus, vulnerability maps integrate data from different sources. These maps are time-dependent because of the parameters’ variability over time, such as the piezometric level and the vadose zone.

The vulnerability map is a spatial two-dimensional (2-D) representation of the degree of vulnerability regarding the potential contamination based on hydrogeological parameters (Foster, 2007). The vulnerability map is a homogeneous module showing a vulnerability index, based on several parameters. Therefore, available methods of determining vulnerability do not provide absolute index values.

The groundwater vulnerability concept is modeled as a source–percolation–receptor model (Liggett and Talwar, 2009). A pollutant infiltrates through the vadose zone along a percolation flow towards the aquifer’s saturated zone. When the pollutant reaches the aquifer’s saturated zone, it becomes the receptor. This describes the facility with which contaminants percolate from the source to the receptor.

The vadose zone acts as an important parameter against groundwater pollution. This zone is important for the transport and the refining of the contaminant. Therefore, it is essential to understand the impact of the vadose zone on the vulnerability mapping. In the fissured formations, penetration rates of the pollutant increase compared to other formations (Foster, 1987). Therefore, the vulnerability will be important.

Aquifer vulnerability involves some particular concepts, such as: (1) intrinsic vulnerability, which is based on aquifer properties such as lithology, hydrology, and hydrogeology (Civita, 1994), and (2) specific vulnerability, which is related to each pollutant. It is used to define the sensitivity of groundwater to a specific contaminant (Schnebelen et al., 2002).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Chenini, Zghibi, Msaddek, and Dlala

The objective of the paper was first to establish a link between available methods for characterization of natural hydrogeological system vulnerability and urban hydrogeology by a methodological adaptation of the vulnerability mapping to urban areas.

Second, we aimed to analyze the impact of urbanization in a Tunisian aquifer. The first aim was addressed by reviewing the most used method for vulnerability mapping, with special care to the specificity of each method. The second objective was addressed by applying a modified DRASTIC model to establish a groundwater vulnerability map in the Mannouba catchment.

The novelty of this study was the use of a multi-method approach for vulnerability mapping of an urban hydrogeological system with specific modifications. This research paper attempts to optimize the vulnerability mapping by taking in account the hydrogeological heterogeneity induced by the urbanization of aquifer systems.

METHODS Literature Review The aim of this review is to present the current state of the art in research methods applied to groundwater vulnerability mapping. To evaluate the groundwater vulnerability, an assessment of the impact of urbanization on the hydrogeological parameters is required.

Concept of Groundwater Vulnerability There is no standard definition associated with groundwater vulnerability, which describes the relative facility for contamination of groundwater. Naturally, the physical environment provides the water resource with a protection from pollutant contamination. The concept of vulnerability in hydrogeology was used for the first time in 1960, and since 1980, it has been used frequently (Haertle, 1983; Aller et al., 1987; and Foster and Hirata, 1988). Actually, the term is commonly used all over the world. Several definitions of groundwater vulnerability have been proposed.

Related to the potential for water resource contamination, groundwater vulnerability is a non-measurable property of an aquifer that is based on the concept that some zones are more susceptible to pollution than others (Musekiwa and Majola, 2013). Always, vulnerability is associated with groundwater contamination.

Groundwater vulnerability has been defined in many ways (Haertle, 1983; Foster, 1987; and Foster and Hirata, 1988). Groundwater vulnerability is “an intrinsic property of a groundwater system that depends on the 294

sensitivity of that system to human and/or natural impacts” (Vrba and Zaporozec, 1994 p 121).

Popescu et al. (2008) defined vulnerability as the relative degree of perturbation and harm caused by human or environment. In hydrogeology, the groundwater vulnerability is described as the susceptibility of an aquifer to be contaminated by a pollutant induced by human anthropic activity (Liggett and Talwar, 2009).

Vulnerability maps generated by hydrogeologists are used by planners and decision makers. They are useful for groundwater resources management and land-use management. However, vulnerability maps are not a direct representation of water resources and their associated elements (piezometric levels, aquifer extension, aquifer hydrodynamic). The vulnerability is rather a set of derived parameters and an index calculated from elements such as slope and soil type. Thus, vulnerability maps integrate data from different sources. These maps are time-dependent because of the parameters’ variability over time, such as the piezometric level and the vadose zone.

The vulnerability map is a spatial two-dimensional (2-D) representation of the degree of vulnerability regarding the potential contamination based on hydrogeological parameters (Foster, 2007). The vulnerability map is a homogeneous module showing a vulnerability index, based on several parameters. Therefore, available methods of determining vulnerability do not provide absolute index values.

The groundwater vulnerability concept is modeled as a source–percolation–receptor model (Liggett and Talwar, 2009). A pollutant infiltrates through the vadose zone along a percolation flow towards the aquifer’s saturated zone. When the pollutant reaches the aquifer’s saturated zone, it becomes the receptor. This describes the facility with which contaminants percolate from the source to the receptor.

The vadose zone acts as an important parameter against groundwater pollution. This zone is important for the transport and the refining of the contaminant. Therefore, it is essential to understand the impact of the vadose zone on the vulnerability mapping. In the fissured formations, penetration rates of the pollutant increase compared to other formations (Foster, 1987). Therefore, the vulnerability will be important.

Aquifer vulnerability involves some particular concepts, such as: (1) intrinsic vulnerability, which is based on aquifer properties such as lithology, hydrology, and hydrogeology (Civita, 1994), and (2) specific vulnerability, which is related to each pollutant. It is used to define the sensitivity of groundwater to a specific contaminant (Schnebelen et al., 2002).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Groundwater Vulnerability Mapping in Urbanized Watershed

Overview of Intrinsic Vulnerability Assessment Approaches To assess the contamination process through the aquifer system, hydrogeological information is analyzed. Based on punctual data, some parameters, such as depth to water table, aquifer thickness, and aquifer conductivity, are extrapolated. Several vulnerability models integrate such information to evaluate the groundwater vulnerability. The appropriate approach with which to evaluate the groundwater vulnerability depends on various factors, including data availability, scope of the study, scale, time, and cost (Liggett and Talwar, 2009).

Several approaches for assessing the groundwater vulnerability have been developed (Stempvoort et al., 1993; Baalousha, 2016; Baki et al., 2017; Kihumba et al., 2017; and Shrestha et al., 2017). Some vulnerability methods are presented as complex models taking into account the physical, chemical, and biological processes in the vadose zone (Beaujean et al., 2014; Huang et al., 2017; and Hussain et al., 2017). The most applied method used is a weighting process of different criteria affecting vulnerability (Gogu and Dassargues, 2000; Ghazavi and Ebrahimi, 2015; Singh et al., 2015; and Hamza et al., 2017).

The vulnerability assessment methods can be classified into two categories:

Statistical methods: These are the category of

vulnerability assessment used worldwide in recent years. They are based on a variable that influences the contaminant concentration or a contamination probability (Antonakos and Lambrakis, 2007).

Mapping methods with an index: This approach is based on the combination of lithological, geological, and hydrogeological maps using geographic information system software. This method considers the attribution of a numerical index or a value to each parameter.

Vulnerability assessment methods with an index consider a variety of parameters. The most widely used approaches are: (1) DRASTIC, which stands for Depth to water table, Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone, and Conductivity (Aller et al., 1987; Krogulec and Trzeciak, 2017), (2) AVI: Aquifer Vulnerability Index (Stempvoort et al., 1993), (3) GOD: Groundwater occurrence, Overall aquifer class, and Depth to water table (Foster, 1987), (4) SYNTACS this model is defined by seven parameters used which are: Depth to water table (S), Effective infiltration (I), Unsaturated zone (N), Soil media (T), Aquifer media (A), Hydraulic conductivity (C), Slope (S) (Civita, 1994), (5) PI The P factor indicates the Protective cover as a function of

Groundwater Vulnerability Mapping in Urbanized Watershed

the thickness and hydraulic properties of strata between surface and groundwater hydraulic head The I factor Infiltration process (Goldscheider et al., 2000), (6) EPIK Epikarst (E) Protective cover (P), Infiltration condition (I), and Karst network (K) (Doerfliger and Zwahlen, 1998), (7) GLA Groundwater occurrence; groundwater Level compared to sea level; Aquifer hydraulic Conductivity (Hölting et al., 1995), and (8) GALDIT Groundwater occurrence; groundwater Level compared to sea level; Aquifer hydraulic Conductivity; Distance from the shore; Impact of seawater intrusion; Thickness of the aquifer (Kardan Moghaddam et al., 2017).

Intrinsic Vulnerability Assessment in an Urban Area The impacts of urbanization on the groundwater flow regime and water quality have been discussed by many authors (Klimas, 1995; Cronin et al., 2003; Ragab et al., 2003; Khan, 2005; Taniguchi et al., 2007; Rao, 2008; Hayashi et al., 2009; Kazemi, 2011; Ramier et al., 2011; and Flechter et al., 2013). The urbanization influences the water balance (Fletcher et al., 2013).

The imperviousness caused by urbanization alters the hydrological system functioning by increasing the volume of runoff. Buildings, streets, and pavements are considered impermeable, and this property affects the duration and intensity of rainfall (Sutherland, 1995; Ragab et al., 2003; and Ramier et al., 2011).

Compared to natural systems, urban watersheds have a specific hydrological water cycle characterized by its complexity. The impermeable surfaces modify the runoff rate and the rainwater infiltration. Therefore, urban groundwater recharge is difficult to evaluate. Underground pipes can locally increase the urban aquifer recharge in cases of water loss. The impermeabilization decreases the infiltration rate (Karpf and Krebs, 2004). Initially, hydrogeologists found that urbanization reduces the recharge of groundwater due to the increasing rates of impermeabilization (Srinivasan et al., 2013). The direct reduction in recharge from infiltration was not significant considering the water budget in an urbanized aquifer (Lerner, 2002).

However, according to Foster and Hirata (1988), urban groundwater recharge was found to increase.

In addition to the impacts of urbanization on the runoff and aquifer recharge rate, this process significantly modifies the groundwater flow in the vadose zone and in the aquifer saturated zone. The exploitation of an aquifer, which is variable over the time, can also change groundwater flow in urban areas. Urbanization largely modifies the soil properties and the infiltration capacity. Various types of foundations can considerably alter the hydraulic conductivity of the unsaturated zone. The underground pipes can modify the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

295

Overview of Intrinsic Vulnerability Assessment Approaches To assess the contamination process through the aquifer system, hydrogeological information is analyzed. Based on punctual data, some parameters, such as depth to water table, aquifer thickness, and aquifer conductivity, are extrapolated. Several vulnerability models integrate such information to evaluate the groundwater vulnerability. The appropriate approach with which to evaluate the groundwater vulnerability depends on various factors, including data availability, scope of the study, scale, time, and cost (Liggett and Talwar, 2009).

Several approaches for assessing the groundwater vulnerability have been developed (Stempvoort et al., 1993; Baalousha, 2016; Baki et al., 2017; Kihumba et al., 2017; and Shrestha et al., 2017). Some vulnerability methods are presented as complex models taking into account the physical, chemical, and biological processes in the vadose zone (Beaujean et al., 2014; Huang et al., 2017; and Hussain et al., 2017). The most applied method used is a weighting process of different criteria affecting vulnerability (Gogu and Dassargues, 2000; Ghazavi and Ebrahimi, 2015; Singh et al., 2015; and Hamza et al., 2017).

The vulnerability assessment methods can be classified into two categories:

Statistical methods: These are the category of

vulnerability assessment used worldwide in recent years. They are based on a variable that influences the contaminant concentration or a contamination probability (Antonakos and Lambrakis, 2007).

Mapping methods with an index: This approach is based on the combination of lithological, geological, and hydrogeological maps using geographic information system software. This method considers the attribution of a numerical index or a value to each parameter.

Vulnerability assessment methods with an index consider a variety of parameters. The most widely used approaches are: (1) DRASTIC, which stands for Depth to water table, Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone, and Conductivity (Aller et al., 1987; Krogulec and Trzeciak, 2017), (2) AVI: Aquifer Vulnerability Index (Stempvoort et al., 1993), (3) GOD: Groundwater occurrence, Overall aquifer class, and Depth to water table (Foster, 1987), (4) SYNTACS this model is defined by seven parameters used which are: Depth to water table (S), Effective infiltration (I), Unsaturated zone (N), Soil media (T), Aquifer media (A), Hydraulic conductivity (C), Slope (S) (Civita, 1994), (5) PI The P factor indicates the Protective cover as a function of

the thickness and hydraulic properties of strata between surface and groundwater hydraulic head The I factor Infiltration process (Goldscheider et al., 2000), (6) EPIK Epikarst (E) Protective cover (P), Infiltration condition (I), and Karst network (K) (Doerfliger and Zwahlen, 1998), (7) GLA Groundwater occurrence; groundwater Level compared to sea level; Aquifer hydraulic Conductivity (Hölting et al., 1995), and (8) GALDIT Groundwater occurrence; groundwater Level compared to sea level; Aquifer hydraulic Conductivity; Distance from the shore; Impact of seawater intrusion; Thickness of the aquifer (Kardan Moghaddam et al., 2017).

Intrinsic Vulnerability Assessment in an Urban Area The impacts of urbanization on the groundwater flow regime and water quality have been discussed by many authors (Klimas, 1995; Cronin et al., 2003; Ragab et al., 2003; Khan, 2005; Taniguchi et al., 2007; Rao, 2008; Hayashi et al., 2009; Kazemi, 2011; Ramier et al., 2011; and Flechter et al., 2013). The urbanization influences the water balance (Fletcher et al., 2013).

The imperviousness caused by urbanization alters the hydrological system functioning by increasing the volume of runoff. Buildings, streets, and pavements are considered impermeable, and this property affects the duration and intensity of rainfall (Sutherland, 1995; Ragab et al., 2003; and Ramier et al., 2011).

Compared to natural systems, urban watersheds have a specific hydrological water cycle characterized by its complexity. The impermeable surfaces modify the runoff rate and the rainwater infiltration. Therefore, urban groundwater recharge is difficult to evaluate. Underground pipes can locally increase the urban aquifer recharge in cases of water loss. The impermeabilization decreases the infiltration rate (Karpf and Krebs, 2004). Initially, hydrogeologists found that urbanization reduces the recharge of groundwater due to the increasing rates of impermeabilization (Srinivasan et al., 2013). The direct reduction in recharge from infiltration was not significant considering the water budget in an urbanized aquifer (Lerner, 2002).

However, according to Foster and Hirata (1988), urban groundwater recharge was found to increase.

In addition to the impacts of urbanization on the runoff and aquifer recharge rate, this process significantly modifies the groundwater flow in the vadose zone and in the aquifer saturated zone. The exploitation of an aquifer, which is variable over the time, can also change groundwater flow in urban areas. Urbanization largely modifies the soil properties and the infiltration capacity. Various types of foundations can considerably alter the hydraulic conductivity of the unsaturated zone. The underground pipes can modify the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

295

Chenini, Zghibi, Msaddek, and Dlala

recharge rate and can modify the vertical water flow in the vadose zone. The qualitative impact of urbanization on the groundwater flow and the vadose zone lithology significantly modifies the vulnerability of the urban aquifers. Therefore, all the vulnerability-related parameters must be analyzed carefully by considering urban aquifer specificities and the spatial and temporal variability of each parameter.

Urban Aquifer Vulnerability Mapping Several questions arise when vulnerability mapping is undertaken for an urbanized aquifer: (1) What is the impact of urbanization on the groundwater recharge processes? (2) How does the urbanization impact the groundwater flow? (3) How can available models be adapted for vulnerability mapping by the integration of urban aquifer specificities?

In addition to these issues, related to the impact of urbanization on the intrinsic vulnerability, the adaptation of the vulnerability mapping to an urbanized aquifer must take into account the mapping scale and the contamination risk caused by the urbanization.

The extent of the urban area is mapped before the evaluation of the urbanization impact on the vulnerability mapping. If the aquifer surface largely exceeds the urbanized area, it is reasonable to apply the common approach for vulnerability modeling. In this case, the urban area impact on the vulnerability is negligible.

However, the urban environment has a specific hydrogeology (Mair and El-Kadi, 2013). The water cycle is disturbed and characterized by its complexity.

The presence of impermeable surfaces modifies the runoff coefficient and the infiltration rate. In addition, the high density of pipes underground can locally increase the recharge in the saturated zone, called seepage. In some case studies, pipes have been used to drain groundwater from the urban hydrogeological system to its outlet (Karpf and Krebs, 2004). The groundwater recharge in an urban hydrogeological system must be analyzed in more detail.

The impact of urbanization processes on surface runoff, on water infiltration, and on groundwater vertical flow in the unsaturated zone are key factors in the assessment of groundwater vulnerability in an urban area.

Because of the complexity of the urban environment and the variable impacts of urbanization on the hydrogeological system, it is therefore difficult to propose a general methodology adapted to vulnerability mapping in an urban area. The groundwater flow and the impact of urbanization are widely related to the aquifer configuration and make it difficult to define a general guideline for vulnerability mapping in an urban area.

296

Chenini, Zghibi, Msaddek, and Dlala

Case Study in Tunisia Located in northern of Tunisia, the Mannouba watershed has an elliptic shape that covers 235 km2 (Figure 1). Mannouba catchment is limited by hills with lithological outcrops of Cretaceous to Quaternary strata. The lithology and the structures of the watershed are given in the Figure 2. About the surface water flow, the most important rivers are oued Gueriana and oued Melah. Many rivers are located in the surrounding hills. Located in the central eastern part of the watershed, Sebkhat Sijoumi is the principal surface water collector.

The hydrogeological system of Mannouba catchment is formed by two aquifer levels: (1) the phreatic aquifer made up by Mio-Plio-Quaternary deposits, and (2) the deep aquifer defined essentially by Cretaceous limestone. For the phreatic aquifer, the reservoir is made up by alluvial deposits (Figure 2). The substratum is defined by marl and clay deposits. Sebkhat Sijoumi is the outlet of the phreatic aquifer. The general groundwater flow in this aquifer is from watershed border to the Sijoumi.

Adopted Methodology In this study, a modified DRASTIC model was applied to establish the groundwater vulnerability map.

The proposed methodology is dedicated to the study of an aquifer located in an urban area (Figure 3).

The main objective is to demonstrate the effect of urbanization on the vulnerability map characterization.

The hydrogeological parameters used in the DRASTIC model are widely influenced by urbanization. The DRASTIC model parameters are: (1) depth to water table (D), (2) net recharge (R), (3) aquifer media (A), (4) soil media (S), (5) topography (slope; T), (6) impact of the vadose zone (I), and (7) hydraulic conductivity (C).

A weight was assigned for each parameter (Table 1). All considered parameters were divided into classes with a specific rate according to the pollutant infiltration facility to the aquifer (Table 2).

The DRASTIC model is normally used to establish vulnerability maps in rural watersheds. It is not appropriate for urbanized watersheds. In this study, a calibrated DRASTIC model is proposed considering the land-use parameter, L. The method is called DRASTIC/L. Land use is commonly modified by urbanization, and so the land-use map was used to evaluate the impact of urbanization. The land-use parameter was assigned a relative weight reflecting this impact on the vulnerability. In a scale from 1 to 5, we assigned a 2 to this parameter. The rate for all related classes was assigned as in Table 2.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

recharge rate and can modify the vertical water flow in the vadose zone. The qualitative impact of urbanization on the groundwater flow and the vadose zone lithology significantly modifies the vulnerability of the urban aquifers. Therefore, all the vulnerability-related parameters must be analyzed carefully by considering urban aquifer specificities and the spatial and temporal variability of each parameter.

Urban Aquifer Vulnerability Mapping Several questions arise when vulnerability mapping is undertaken for an urbanized aquifer: (1) What is the impact of urbanization on the groundwater recharge processes? (2) How does the urbanization impact the groundwater flow? (3) How can available models be adapted for vulnerability mapping by the integration of urban aquifer specificities?

In addition to these issues, related to the impact of urbanization on the intrinsic vulnerability, the adaptation of the vulnerability mapping to an urbanized aquifer must take into account the mapping scale and the contamination risk caused by the urbanization.

The extent of the urban area is mapped before the evaluation of the urbanization impact on the vulnerability mapping. If the aquifer surface largely exceeds the urbanized area, it is reasonable to apply the common approach for vulnerability modeling. In this case, the urban area impact on the vulnerability is negligible.

However, the urban environment has a specific hydrogeology (Mair and El-Kadi, 2013). The water cycle is disturbed and characterized by its complexity.

The presence of impermeable surfaces modifies the runoff coefficient and the infiltration rate. In addition, the high density of pipes underground can locally increase the recharge in the saturated zone, called seepage. In some case studies, pipes have been used to drain groundwater from the urban hydrogeological system to its outlet (Karpf and Krebs, 2004). The groundwater recharge in an urban hydrogeological system must be analyzed in more detail.

The impact of urbanization processes on surface runoff, on water infiltration, and on groundwater vertical flow in the unsaturated zone are key factors in the assessment of groundwater vulnerability in an urban area.

Because of the complexity of the urban environment and the variable impacts of urbanization on the hydrogeological system, it is therefore difficult to propose a general methodology adapted to vulnerability mapping in an urban area. The groundwater flow and the impact of urbanization are widely related to the aquifer configuration and make it difficult to define a general guideline for vulnerability mapping in an urban area.

296

Case Study in Tunisia Located in northern of Tunisia, the Mannouba watershed has an elliptic shape that covers 235 km2 (Figure 1). Mannouba catchment is limited by hills with lithological outcrops of Cretaceous to Quaternary strata. The lithology and the structures of the watershed are given in the Figure 2. About the surface water flow, the most important rivers are oued Gueriana and oued Melah. Many rivers are located in the surrounding hills. Located in the central eastern part of the watershed, Sebkhat Sijoumi is the principal surface water collector.

The hydrogeological system of Mannouba catchment is formed by two aquifer levels: (1) the phreatic aquifer made up by Mio-Plio-Quaternary deposits, and (2) the deep aquifer defined essentially by Cretaceous limestone. For the phreatic aquifer, the reservoir is made up by alluvial deposits (Figure 2). The substratum is defined by marl and clay deposits. Sebkhat Sijoumi is the outlet of the phreatic aquifer. The general groundwater flow in this aquifer is from watershed border to the Sijoumi.

Adopted Methodology In this study, a modified DRASTIC model was applied to establish the groundwater vulnerability map.

The proposed methodology is dedicated to the study of an aquifer located in an urban area (Figure 3).

The main objective is to demonstrate the effect of urbanization on the vulnerability map characterization.

The hydrogeological parameters used in the DRASTIC model are widely influenced by urbanization. The DRASTIC model parameters are: (1) depth to water table (D), (2) net recharge (R), (3) aquifer media (A), (4) soil media (S), (5) topography (slope; T), (6) impact of the vadose zone (I), and (7) hydraulic conductivity (C).

A weight was assigned for each parameter (Table 1). All considered parameters were divided into classes with a specific rate according to the pollutant infiltration facility to the aquifer (Table 2).

The DRASTIC model is normally used to establish vulnerability maps in rural watersheds. It is not appropriate for urbanized watersheds. In this study, a calibrated DRASTIC model is proposed considering the land-use parameter, L. The method is called DRASTIC/L. Land use is commonly modified by urbanization, and so the land-use map was used to evaluate the impact of urbanization. The land-use parameter was assigned a relative weight reflecting this impact on the vulnerability. In a scale from 1 to 5, we assigned a 2 to this parameter. The rate for all related classes was assigned as in Table 2.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Groundwater Vulnerability Mapping in Urbanized Watershed

Groundwater Vulnerability Mapping in Urbanized Watershed

Figure 1. Location of the study area.

Figure 1. Location of the study area.

The impermeabilization caused by urbanization can decrease the infiltration process and reduces the vulnerability. Urbanization also affects the groundwater recharge of the aquifer. In this study, we also discuss the groundwater recharge process as widely modified by urbanization.

The DRASTIC index Di was then calculated using a linear combination as shown in the following equation: Di = DwDr + RwRr + AwAr + SwSr + T wT r + IwIr + CwCr + LwLr, where D, R, A, S, T, I, C, and L are the eight parameters of the modified model, w is the weight assigned to the parameter, and r is the rate assigned to each class.

All resulting maps were combined to establish the vulnerability map.

A sensitivity test was performed using a singleparameter sensitivity analysis. The purpose of the test was to establish the degree of a parameter’s sensitivity in the vulnerability mapping.

RESULTS The depth to water table (D) plays a key role in the groundwater vulnerability qualification. It is the measurement of the water table depth in piezometers. As shown in Figure 4, a depth to water table map was established using GIS utilities. Table 2 summarizes the classes defined for this parameter.

The net recharge (R) is the water rate that reaches the aquifer saturated zone by the infiltration process.

The urban infrastructure reduces the recharge process; therefore, the estimation of this parameter is difficult.

Many human activities have an impact on the aquifer recharge, such as domestic and municipal irrigation, and possible leakage and losses from water mains and sewers. A multi-factor approach was applied to quantify the recharge rate in an urbanized study area. The water table fluctuation was used to calculate the net recharge between 2013 and 2014.

The recharge induced by the urban environment was estimated according to Hibbs and Sharp (2012).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

297

The impermeabilization caused by urbanization can decrease the infiltration process and reduces the vulnerability. Urbanization also affects the groundwater recharge of the aquifer. In this study, we also discuss the groundwater recharge process as widely modified by urbanization.

The DRASTIC index Di was then calculated using a linear combination as shown in the following equation: Di = DwDr + RwRr + AwAr + SwSr + T wT r + IwIr + CwCr + LwLr, where D, R, A, S, T, I, C, and L are the eight parameters of the modified model, w is the weight assigned to the parameter, and r is the rate assigned to each class.

All resulting maps were combined to establish the vulnerability map.

A sensitivity test was performed using a singleparameter sensitivity analysis. The purpose of the test was to establish the degree of a parameter’s sensitivity in the vulnerability mapping.

RESULTS The depth to water table (D) plays a key role in the groundwater vulnerability qualification. It is the measurement of the water table depth in piezometers. As shown in Figure 4, a depth to water table map was established using GIS utilities. Table 2 summarizes the classes defined for this parameter.

The net recharge (R) is the water rate that reaches the aquifer saturated zone by the infiltration process.

The urban infrastructure reduces the recharge process; therefore, the estimation of this parameter is difficult.

Many human activities have an impact on the aquifer recharge, such as domestic and municipal irrigation, and possible leakage and losses from water mains and sewers. A multi-factor approach was applied to quantify the recharge rate in an urbanized study area. The water table fluctuation was used to calculate the net recharge between 2013 and 2014.

The recharge induced by the urban environment was estimated according to Hibbs and Sharp (2012).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

297

Chenini, Zghibi, Msaddek, and Dlala

Chenini, Zghibi, Msaddek, and Dlala

Figure 2. (A) Lithological map of Mannouba catchment (after El May et al., 2010). (B) Cross section showing the structure of the studied aquifer. (C) Synthetic log of the study area.

Figure 2. (A) Lithological map of Mannouba catchment (after El May et al., 2010). (B) Cross section showing the structure of the studied aquifer. (C) Synthetic log of the study area.

298

298

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293â&#x20AC;&#x201C;304

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293â&#x20AC;&#x201C;304

2

5

Agriculture zone Forest zone 6

6 Alluvium

8

3

5

2

>10% 1

3

Sand silt and clay Sand and conglomerates

6

4×10−4 –5×10−4 5 Limestone 5 5%–10% 2

Halomorphic soil Calcimorphic soil Calcimorphic red soil with Lithosols Fluvial soil and sand Salty halomorphic soil 6

4 Unbuilt land 4 2×10−4 –4×10−4 9 2%–5% 3 Soil crust 7

8

Sand silt and conglomerates Alluvium 6

5

2

9–15

>23

10–17.5 7 4.5–9

17.5–25.5

3 9 1.5–4.5

5–10

3

Limestone

6

Urbanized area

2

<2%

10

Silt and evaporites Clay and sands

6

7 2 5×10 –2×10

Urban area

Rate Classes Rate

−5

−4

Rate

Classes

Rate

Classes

Rate

Classes

Rate

Classes

Rate

Hydraulic Conductivity Impact of Vadose Zone Topography (Slope) Soil Media Aquifer Media Net Recharge

Classes Rate Classes

2 >23

Depth to Groundwater

3

5

2

Halomorphic soil Calcimorphic soil Calcimorphic red soil with Lithosols Fluvial soil and sand Salty halomorphic soil 6 8 5 9–15

17.5–25.5

299

5 4 3 2 1 5 3 2

Table 2. Rate of classes defined for each DRASTIC parameter (Aller et al., 1987) and land-use factor.

6 Alluvium

8

6

Sand silt and clay Sand and conglomerates 3 >10% 1

Soil crust 7

Sand silt and conglomerates Alluvium 6 10–17.5 7 4.5–9

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Depth to groundwater (D) Net recharge (R) Aquifer media (A) Soil media (S) Topography (slope, T) Impact of the vadose zone (I) Hydraulic conductivity (C) Land-use map (L)

Weight

Classes

Land Use

2

5

Agriculture zone Forest zone 6 4×10−4 –5×10−4 5 5%–10% 2

5

9 2%–5% 3

Silt and evaporites Clay and sands 10 9

5–10

3

Urbanized area 6 Limestone

2

<2%

Limestone

4 4 2×10−4 –4×10−4 6

Unbuilt land

7 Urban area 2 5×10 –2×10 3

Classes Rate

−4 −5

Classes Rate

Classes

Rate

Classes

Rate

Classes

Rate

Classes

Rate

Hydraulic Conductivity Impact of Vadose Zone Topography (Slope) Soil Media Aquifer Media

The recharge net map was then established (Figure 4). The rate assigned for each class is shown in Table 2.

The aquifer media (A) variable refers to the aquifer lithology. In the study area, the variation of this parameter was mapped based on the geological map of Mannouba watershed and using the lithological data in well cross sections. The map showing the spatial variability of the aquifer media was established (Figure 4). The rating for each lithological class of the aquifer media is summarized in Table 2.

The soil media (S) considerably affects groundwater contamination processes. It can reduce or accelerate these processes. The soil map of Mannouba watershed was established. The rating values were determined according to the classes reported in Table 2.

DRASTIC Parameters

1.5–4.5

5 4 3 2 1 5 3 2

Classes

Depth to groundwater (D) Net recharge (R) Aquifer media (A) Soil media (S) Topography (slope, T) Impact of the vadose zone (I) Hydraulic conductivity (C) Land-use map (L)

Weight

Figure 3. Flowchart showing steps of the adopted methodology.

Table 1. Adopted weight for DRASTIC/L parameters (modified after Aller et al., 1987).

Rate

DRASTIC Parameters

Groundwater Vulnerability Mapping in Urbanized Watershed

Classes

Table 1. Adopted weight for DRASTIC/L parameters (modified after Aller et al., 1987).

Net Recharge

The recharge net map was then established (Figure 4). The rate assigned for each class is shown in Table 2.

The aquifer media (A) variable refers to the aquifer lithology. In the study area, the variation of this parameter was mapped based on the geological map of Mannouba watershed and using the lithological data in well cross sections. The map showing the spatial variability of the aquifer media was established (Figure 4). The rating for each lithological class of the aquifer media is summarized in Table 2.

The soil media (S) considerably affects groundwater contamination processes. It can reduce or accelerate these processes. The soil map of Mannouba watershed was established. The rating values were determined according to the classes reported in Table 2.

Depth to Groundwater

Figure 3. Flowchart showing steps of the adopted methodology.

Table 2. Rate of classes defined for each DRASTIC parameter (Aller et al., 1987) and land-use factor.

Land Use

Rate

Groundwater Vulnerability Mapping in Urbanized Watershed

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

299

Chenini, Zghibi, Msaddek, and Dlala

Chenini, Zghibi, Msaddek, and Dlala

The slope or topography (T) parameter was assigned a weight equal to 1. It is a parameter with a small role in vulnerability mapping. In the Mannouba watershed, the slope map shows the variability of slope classes (Figure 4).

The impact of the vadose zone (I) is an important parameter to be evaluated in the vulnerability mapping. It affects the propagation of pollutants from the surface to the saturate zone. This impact is influenced by the lithology and the thickness of the vadose zone. We considered the lithological map and the interpretation of well logging to establish the map related to this parameter (Figure 4).

The hydraulic conductivity (C) provides information about pollutant migration in the aquifer media.

The hydraulic conductivity values used to calculate the degrees of vulnerability were collected from pumping tests. The spatial variation of this parameter is shown in Figure 2.

The land-use parameter (L was incorporated into this model to characterize urbanization in the watershed. This parameter considers the land-use variation over time due to the human activities caused by the urbanization process. The land-use map was established using remotely sensed data. It represents the land-use variability. In industrial and agricultural zones, the vulnerability increases. The land-use map for the Mannouba watershed shows four classes (Table 2). This parameter was added to the DRASTIC model to demonstrate the effect of land-use modification on the groundwater vulnerability mapping.

The slope or topography (T) parameter was assigned a weight equal to 1. It is a parameter with a small role in vulnerability mapping. In the Mannouba watershed, the slope map shows the variability of slope classes (Figure 4).

The impact of the vadose zone (I) is an important parameter to be evaluated in the vulnerability mapping. It affects the propagation of pollutants from the surface to the saturate zone. This impact is influenced by the lithology and the thickness of the vadose zone. We considered the lithological map and the interpretation of well logging to establish the map related to this parameter (Figure 4).

The hydraulic conductivity (C) provides information about pollutant migration in the aquifer media.

The hydraulic conductivity values used to calculate the degrees of vulnerability were collected from pumping tests. The spatial variation of this parameter is shown in Figure 2.

The land-use parameter (L was incorporated into this model to characterize urbanization in the watershed. This parameter considers the land-use variation over time due to the human activities caused by the urbanization process. The land-use map was established using remotely sensed data. It represents the land-use variability. In industrial and agricultural zones, the vulnerability increases. The land-use map for the Mannouba watershed shows four classes (Table 2). This parameter was added to the DRASTIC model to demonstrate the effect of land-use modification on the groundwater vulnerability mapping.

Sensitivity Analysis

Sensitivity Analysis

In the first step of the study, a vulnerability map was established based on the modified DRASTIC model (Figure 5). The classes of vulnerability are summarized in Table 3. The Land use parameter was integrated into the DRASTIC model in order to consider the impact of urbanization in the vulnerability mapping. On the other side, the recharge net of the aquifer was analyzed because the urbanization modifies this process.

In the study area, the vulnerability index was classified as low, medium, and high in different areas. This index is sensitive to the parameter weight, the rate, and the values assigned for each class. Vulnerability indexes between 80 and 119 are considered low-vulnerability areas. Medium- and high-vulnerability areas covered, respectively, 45 percent and 35 percent of the total area of Mannouba watershed. The high-vulnerability area was located in the urban zone.

In this study, two sensitivity analyses were carried out: (1) single-parameter sensitivity analysis, and (2) map removal sensitivity analysis. The sensitivity analyses provide knowledge on the impact of the weight as-

In the first step of the study, a vulnerability map was established based on the modified DRASTIC model (Figure 5). The classes of vulnerability are summarized in Table 3. The Land use parameter was integrated into the DRASTIC model in order to consider the impact of urbanization in the vulnerability mapping. On the other side, the recharge net of the aquifer was analyzed because the urbanization modifies this process.

In the study area, the vulnerability index was classified as low, medium, and high in different areas. This index is sensitive to the parameter weight, the rate, and the values assigned for each class. Vulnerability indexes between 80 and 119 are considered low-vulnerability areas. Medium- and high-vulnerability areas covered, respectively, 45 percent and 35 percent of the total area of Mannouba watershed. The high-vulnerability area was located in the urban zone.

In this study, two sensitivity analyses were carried out: (1) single-parameter sensitivity analysis, and (2) map removal sensitivity analysis. The sensitivity analyses provide knowledge on the impact of the weight as-

300

Figure 4. Maps of the DRASTIC/L model (D: depth to water table; R: net recharge; A: aquifer media; S: soil media; T: topography [slope]; I: impact of vadose zone; C: hydraulic conductivity; L: land use).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

300

Figure 4. Maps of the DRASTIC/L model (D: depth to water table; R: net recharge; A: aquifer media; S: soil media; T: topography [slope]; I: impact of vadose zone; C: hydraulic conductivity; L: land use).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Groundwater Vulnerability Mapping in Urbanized Watershed

Groundwater Vulnerability Mapping in Urbanized Watershed

Table 3. Intervals of DRASTIC index values and corresponding vulnerability classes (Aller at al., 1987).

Table 3. Intervals of DRASTIC index values and corresponding vulnerability classes (Aller at al., 1987).

DRASTIC Index

DRASTIC Index

Vulnerability Class

80–119 120–159 >160

Low Medium high

80–119 120–159 >160

where Ew is the effective weight for a considered parameter; Pw is the weight of a parameter; Pr is the a rate of a class; and Vi is the vulnerability index. Table 4 summarizes the statistical analysis of modified DRASTIC model parameters.

The second part of the sensibility analysis is done using a map removal process. This method consists of thematic map extraction in the same time. The analysis evaluates the vulnerability map sensitivity when one or more maps are removed from the model. Table 5 illustrates the analysis and statistics based on the map removal sensitivity. The formula from Farjad et al. (2012) was used:

S = V/N − V � /n | /N × 100,

Figure 5. Classes of groundwater vulnerability.

signed to each parameter and the rate assigned to each class. It allows us to assess the importance of each parameter in the modified DRASTIC model.

The single-parameter sensitivity analysis was performed to evaluate the effect of each factor in the groundwater vulnerability index. It assesses the individual parameter impact and the effective weight for the groundwater vulnerability evaluation. The effective weight was calculated based on the following equation: Ew = ((Pw × Pr ) /Vi ) × 100,

where S is the sensitivity assessment expressed as a variation index (%); V is the DRASTIC/L vulnerability index; V� is the vulnerability index of a parameter; N is the eight parameters included in the calculation of V, and n is the number of parameters used in the calculation of V� .

The sensitivity assessment by map removal process explains the relationship between V and V� . Statistics about the vulnerability variation after thematic layer removal are shown in Table 5.

DISCUSSION The study was performed in the shallow aquifer of Mannouba catchment, using a model originally developed for rural watersheds. To adapt the DRASTIC model to urbanized watersheds, the land-use param-

Low Medium high

where Ew is the effective weight for a considered parameter; Pw is the weight of a parameter; Pr is the a rate of a class; and Vi is the vulnerability index. Table 4 summarizes the statistical analysis of modified DRASTIC model parameters.

The second part of the sensibility analysis is done using a map removal process. This method consists of thematic map extraction in the same time. The analysis evaluates the vulnerability map sensitivity when one or more maps are removed from the model. Table 5 illustrates the analysis and statistics based on the map removal sensitivity. The formula from Farjad et al. (2012) was used:

S = V/N − V � /n | /N × 100,

Figure 5. Classes of groundwater vulnerability.

signed to each parameter and the rate assigned to each class. It allows us to assess the importance of each parameter in the modified DRASTIC model.

The single-parameter sensitivity analysis was performed to evaluate the effect of each factor in the groundwater vulnerability index. It assesses the individual parameter impact and the effective weight for the groundwater vulnerability evaluation. The effective weight was calculated based on the following equation: Ew = ((Pw × Pr ) /Vi ) × 100,

Table 4. Sensitivity analysis for single parameters and statistics.

where S is the sensitivity assessment expressed as a variation index (%); V is the DRASTIC/L vulnerability index; V� is the vulnerability index of a parameter; N is the eight parameters included in the calculation of V, and n is the number of parameters used in the calculation of V� .

The sensitivity assessment by map removal process explains the relationship between V and V� . Statistics about the vulnerability variation after thematic layer removal are shown in Table 5.

DISCUSSION The study was performed in the shallow aquifer of Mannouba catchment, using a model originally developed for rural watersheds. To adapt the DRASTIC model to urbanized watersheds, the land-use param-

Table 4. Sensitivity analysis for single parameters and statistics.

Effective Weight (%) Parameter D R A S T I C L

Vulnerability Class

Effective Weight (%)

Theoretical Weight

Theoretical Weight (%)

Mean

Minimum

Maximum

Standard Deviation

5 4 3 2 1 5 3 2

20 16 12 8 4 20 12 8

16 17.33 17.62 4.3 4.95 21.65 9.29 6.85

12 14.8 13.04 6.2 03.7 18.5 7.4 5

20 19.87 22.2 2.4 6.2 24.8 11.18 8.7

2.5 2.05 0.47 1.17 2.86 3.75 1.63 1.8

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Parameter D R A S T I C L

301

Theoretical Weight

Theoretical Weight (%)

Mean

Minimum

Maximum

Standard Deviation

5 4 3 2 1 5 3 2

20 16 12 8 4 20 12 8

16 17.33 17.62 4.3 4.95 21.65 9.29 6.85

12 14.8 13.04 6.2 03.7 18.5 7.4 5

20 19.87 22.2 2.4 6.2 24.8 11.18 8.7

2.5 2.05 0.47 1.17 2.86 3.75 1.63 1.8

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

301

Chenini, Zghibi, Msaddek, and Dlala Table 5. Index variation of the map removal sensitivity analysis.

Index Variation (%) Removed Parameter D R A S T I C L

Mean

Minimum

Maximum

Standard Deviation

1.01 1.65 3.05 3.03 2.24 4.4 5.65 2.1

0.3 0.8 1.11 2.31 2.08 2.56 3.8 2.45

1.73 2.5 5 3.75 2.4 3.25 7.5 1.75

1.24 1.25 1.3 1.47 1.2 0.7 0.02 0.56

eter was included in the model. The groundwater vulnerability index varied from low to high: 25 percent of the study catchment showed low vulnerability, located in the north and southwest sides of Mannouba catchment. Areas with medium vulnerability were more dominant and covered 45 percent of the catchment.

The high-vulnerability index covered 30 percent of the Mannouba catchment and was located in the urbanized area.

Results of the sensitivity analysis showed that the impact of aquifer media, vadose zone, net recharge, and depth to water table had the most important weights. They are the most dominant parameters that impact vulnerability mapping. For the land use, topography, and soil media, the theoretical weight was more important than the effective weight. The impact of vadose zone was the most important effective weight and had an important impact in the vulnerability evaluation process.

The sensitivity assessment by map removal process showed that the hydraulic conductivity and the impact of vadose zone had the highest mean variation index.

In this analysis, the vulnerability index is sensitive to the removal of aquifer media and soil media.

The modified DRASTIC model was carried out in urbanized watershed using GIS utilities to evaluate the groundwater vulnerability. Special attention was given to incorporate the impact of urbanization on the hydrogeological parameters such as net recharge calculation, land use, and soil media. The established vulnerability maps have a great importance in urban water management.

CONCLUSION Assessment of groundwater vulnerability in an urban area is an important element for groundwater protection and management. The objective of this study was the evaluation of groundwater vulnerability in Mannouba catchment using a modified DRASTIC model. the land-use parameter was integrated into the 302

Chenini, Zghibi, Msaddek, and Dlala

DRASTIC index calculation. Eight parameters were integrated in a GIS environment, and a vulnerability index map was computed. Low vulnerability covered about 25 percent of the watershed area, and medium vulnerability covered 45 percent, while high vulnerability was indicated over 30 percent of the study area. The single-parameter sensitive analysis confirmed that theoretical weight was different from the calculated effective weight for each parameter. The DRASTIC index was influenced by impact of the vadose zone, aquifer media, and net recharge. The impact of vadose zone controlled the vulnerability index, with a theoretical weight of 20 percent and an effective weight around 22 percent. The map removal sensitivity indicated that the hydraulic conductivity and the vadose zone impact were the most important parameters that modify the sensitivity of the vulnerability index. In the two sensitivity analyses, vadose zone impact was the key parameter that influences the vulnerability index.

REFERENCES Aller, L.; Bennet, T.; Lehr, J. H.; and Petty, R. J., 1987, DRASTIC: A Standardized System for Evaluating Ground Water Pollution Potential Using Hydro Geologic Settings: U.S. Environmental Protection Agency Report EPA/600/2-85/018.

Antonakos, A. K. and Lambrakis, N. J., 2007, Development and testing of three hybrid methods for the assessment of aquifer vulnerability to nitrates, based on the DRASTIC model, an example from NE Korinthia, Greece: Journal of Hydrology, Vol. 333, pp. 288–304.

Baalousha, H. M., 2016, Groundwater vulnerability mapping of Qatar aquifers: Journal of African Earth Sciences, Vol. 124, pp.

75–93.

Baki, S.; Hilali, M.; Kacimi, I.; Kassou, N.; Nouiyti, N.; and Bahassi, A., 2017, Assessment of groundwater intrinsic vulnerability to pollution in the pre-Saharan Areas—The case of the Tafilalet Plain (southeast Morocco): Procedia Earth and Planetary Science, Vol. 17, pp. 590–593.

Beaujean, J.; Lemieux, J. M.; Dassargues, A.; Therrien, R.; and Brouyère, S., 2014, Physically based groundwater vulnerability assessment using sensitivity analysis methods: Groundwater, Vol. 52, No. 6, pp. 864–874.

Chae, G. T.; Yuna, S. T.; Choi, B. Y.; Yuc, S. Y.; Jo, H. Y.; Mayer, B.; Kim, Y. J.; and Lee, J. Y., 2008, Hydrochemistry of urban groundwater, Seoul, Korea. The impact of subway tunnels on groundwater quality: Journal of Contaminant Hydrology, Vol. 101, pp. 42–52.

Civita, M., 1994, Aquifer Vulnerability Map to Pollution: Theory and Application: Pitagora Editrice, Bologna, Italy, 325 p.

Cronin, A. A.; Taylor, R. G.; Powell, K. L.; Barrett, M. H.; Trowsdale, S. A.; and Lerner, D. N., 2003, Temporal variations in the depth-specific hydrochemistry and sewagerelated microbiology of an urban sandstone aquifer, Nottingham, United Kingdom: Hydrogeology Journal, Vol. 11, pp. 205–216.

Doerfliger, N. and Zwahlen, F., 1998, Groundwater Vulnerability Mapping in Karstic Regions (EPIK), Practical Guide: Swiss Agency for the Environment, Forests and Landscape (SAEFL), Bern, Switzerland, 56 p.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Table 5. Index variation of the map removal sensitivity analysis.

Index Variation (%) Removed Parameter D R A S T I C L

Mean

Minimum

Maximum

Standard Deviation

1.01 1.65 3.05 3.03 2.24 4.4 5.65 2.1

0.3 0.8 1.11 2.31 2.08 2.56 3.8 2.45

1.73 2.5 5 3.75 2.4 3.25 7.5 1.75

1.24 1.25 1.3 1.47 1.2 0.7 0.02 0.56

eter was included in the model. The groundwater vulnerability index varied from low to high: 25 percent of the study catchment showed low vulnerability, located in the north and southwest sides of Mannouba catchment. Areas with medium vulnerability were more dominant and covered 45 percent of the catchment.

The high-vulnerability index covered 30 percent of the Mannouba catchment and was located in the urbanized area.

Results of the sensitivity analysis showed that the impact of aquifer media, vadose zone, net recharge, and depth to water table had the most important weights. They are the most dominant parameters that impact vulnerability mapping. For the land use, topography, and soil media, the theoretical weight was more important than the effective weight. The impact of vadose zone was the most important effective weight and had an important impact in the vulnerability evaluation process.

The sensitivity assessment by map removal process showed that the hydraulic conductivity and the impact of vadose zone had the highest mean variation index.

In this analysis, the vulnerability index is sensitive to the removal of aquifer media and soil media.

The modified DRASTIC model was carried out in urbanized watershed using GIS utilities to evaluate the groundwater vulnerability. Special attention was given to incorporate the impact of urbanization on the hydrogeological parameters such as net recharge calculation, land use, and soil media. The established vulnerability maps have a great importance in urban water management.

CONCLUSION Assessment of groundwater vulnerability in an urban area is an important element for groundwater protection and management. The objective of this study was the evaluation of groundwater vulnerability in Mannouba catchment using a modified DRASTIC model. the land-use parameter was integrated into the 302

DRASTIC index calculation. Eight parameters were integrated in a GIS environment, and a vulnerability index map was computed. Low vulnerability covered about 25 percent of the watershed area, and medium vulnerability covered 45 percent, while high vulnerability was indicated over 30 percent of the study area. The single-parameter sensitive analysis confirmed that theoretical weight was different from the calculated effective weight for each parameter. The DRASTIC index was influenced by impact of the vadose zone, aquifer media, and net recharge. The impact of vadose zone controlled the vulnerability index, with a theoretical weight of 20 percent and an effective weight around 22 percent. The map removal sensitivity indicated that the hydraulic conductivity and the vadose zone impact were the most important parameters that modify the sensitivity of the vulnerability index. In the two sensitivity analyses, vadose zone impact was the key parameter that influences the vulnerability index.

REFERENCES Aller, L.; Bennet, T.; Lehr, J. H.; and Petty, R. J., 1987, DRASTIC: A Standardized System for Evaluating Ground Water Pollution Potential Using Hydro Geologic Settings: U.S. Environmental Protection Agency Report EPA/600/2-85/018.

Antonakos, A. K. and Lambrakis, N. J., 2007, Development and testing of three hybrid methods for the assessment of aquifer vulnerability to nitrates, based on the DRASTIC model, an example from NE Korinthia, Greece: Journal of Hydrology, Vol. 333, pp. 288–304.

Baalousha, H. M., 2016, Groundwater vulnerability mapping of Qatar aquifers: Journal of African Earth Sciences, Vol. 124, pp.

75–93.

Baki, S.; Hilali, M.; Kacimi, I.; Kassou, N.; Nouiyti, N.; and Bahassi, A., 2017, Assessment of groundwater intrinsic vulnerability to pollution in the pre-Saharan Areas—The case of the Tafilalet Plain (southeast Morocco): Procedia Earth and Planetary Science, Vol. 17, pp. 590–593.

Beaujean, J.; Lemieux, J. M.; Dassargues, A.; Therrien, R.; and Brouyère, S., 2014, Physically based groundwater vulnerability assessment using sensitivity analysis methods: Groundwater, Vol. 52, No. 6, pp. 864–874.

Chae, G. T.; Yuna, S. T.; Choi, B. Y.; Yuc, S. Y.; Jo, H. Y.; Mayer, B.; Kim, Y. J.; and Lee, J. Y., 2008, Hydrochemistry of urban groundwater, Seoul, Korea. The impact of subway tunnels on groundwater quality: Journal of Contaminant Hydrology, Vol. 101, pp. 42–52.

Civita, M., 1994, Aquifer Vulnerability Map to Pollution: Theory and Application: Pitagora Editrice, Bologna, Italy, 325 p.

Cronin, A. A.; Taylor, R. G.; Powell, K. L.; Barrett, M. H.; Trowsdale, S. A.; and Lerner, D. N., 2003, Temporal variations in the depth-specific hydrochemistry and sewagerelated microbiology of an urban sandstone aquifer, Nottingham, United Kingdom: Hydrogeology Journal, Vol. 11, pp. 205–216.

Doerfliger, N. and Zwahlen, F., 1998, Groundwater Vulnerability Mapping in Karstic Regions (EPIK), Practical Guide: Swiss Agency for the Environment, Forests and Landscape (SAEFL), Bern, Switzerland, 56 p.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Groundwater Vulnerability Mapping in Urbanized Watershed El May, M.; Dlala, M.; and Chenini, I., 2010, Urban geological mapping: Geotechnical data analysis for rational development planning: Engineering Geology, Vol. 116, pp. 129–138.

Farjad, B.; Mohd Shafri, H. Z.; Mohamed, T. A.; Pirasteh, S.; and Wijesekara, N., 2012, Groundwater intrinsic vulnerability and risk mapping: Water Management, Vol. 165, No. 8, pp.

441–450.

Fletcher, T. D.; Andrieu, H.; and Hamel, P., 2013. Understanding, management and modeling of urban hydrology and its consequences for receiving waters: A state of the art: Advances in Water Resources, Vol. 51, pp. 261–279.

Foster, S. S. D., 1987, Fundamental concepts in aquifer vulnerability, pollution risk and protection strategy. In van Duijvenbooden, W. and van Waegeningh, H. (Editors) the Netherlands Vulnerability of Soil and Groundwater to Pollutants, The Hague, Netherlands, 21, pp. 68–73.

Foster, S. S. D., 2007, Aquifer pollution vulnerability concept and tools—Use, benefits and constraints. In Witkowski, A. J.; Kowalczyk, A.; and Vrba, J. (Editors), Groundwater Vulnerability Assessment and Mapping: Taylor & Francis Group, London, U.K. 11, pp. 3–9.

Foster, S. S. D. and Hirata, R., 1988, Groundwater Pollution Risk Assessment: A Methodology Using Available Data: WHOPAHO/HPE-CEPIS Technical Manual, World Health Organization, Lima, Peru, 81 p.

Ghazavi, R. and Ebrahimi, Z., 2015, Assessing groundwater vulnerability to contamination in an arid environment using DRASTIC and GOD models: International Journal of Environmental Science and Technology, Vol. 12, No. 9, pp.

2909–2918.

Gogu, R. C. and Dassargues, A., 2000, Current trends and future challenges in groundwater vulnerability assessment using overlay and index methods: Environmental Geology, Vol. 39, No. 6, pp. 549–559.

Goldscheider, N.; Klute, M.; Sturm, S.; and Hotzl, H., 2000, The PI-method—A GIS based approach to mapping groundwater vulnerability with special consideration of karst aquifers: Zeitschrift Angewandte Geologie, Hannover, Vol. 46, No. 3, pp. 157–166.

Haertle, A., 1983, Method of Working and Employment of EDP during the Preparation of Groundwater Vulnerability Maps: IAHS Publication 142, International Association of Hydrological Sciences, London, U.K., pp. 1073–1085.

Hamza, S. M.; Ahsan, A.; Imteaz, M. A.; Ghazali, A. H.; and Mohammed, T. A., 2017, GIS-based FRASTIC model for pollution vulnerability assessment of fractured-rock aquifer systems: Environmental Earth Sciences, Vol. 76, No. 5, pp. 197.

Hayashi, T.; Tokunaga, T.; Aichi, M.; Shimada, J.; and Taniguchi, M., 2009, Effect of human activities and urbanization on groundwater environments: An example from the aquifer system of Tokyo and the surrounding area: Science of the Total Environment, Vol. 407, pp. 3165–3172.

Hibbs, B. J. and Sharp, J. M., 2012, Hydrogeological impacts of urbanization: Environmental and Engineering Geoscience, Vol. 18, No. 1, pp. 3–24.

Hölting, B.; Haertle, T.; Hohberger, K. H.; Nachtigall, K.; Villinger, E.; Weinzierl, W.; and Wrobel, J.-P., 1995, Concept to assess the protective function of the layers above the groundwater surface: Geologisches Jahrbuch der BGR Hannover, Ser. C, Vol. 63, pp. 5–24.

Huang, L.; Zeng, G.; Liang, J.; Hua, S.; Yuan, Y.; Li, X.; and Liu, J., 2017, Combined impacts of land use and climate change in the modeling of future groundwater vulnerability: Journal of Hydrologic Engineering, Vol. 22, No. 7, pp. 05017007. doi: 10.1061/(ASCE)HE.19435584.0001493#sthash.XiYpDY1J.

Groundwater Vulnerability Mapping in Urbanized Watershed

Hussain, Y.; Ullah, S. F.; Hussain, M. B.; Aslam, A. Q.; Akhter, G.; Martinez-Carvajal, H.; and Cárdenas-Soto, M., 2017, Modelling the vulnerability of groundwater to contamination in an unconfined alluvial aquifer in Pakistan: Environmental Earth Sciences, Vol. 76, No. 2, pp. 84.

Hwang, H. H.; Panno, S. V.; and Hackley, K. C., 2015, Sources and changes in groundwater quality with increasing urbanization, northeastern Illinois: Environmental & Engineering Geoscience, Vol. 21, No. 2, pp. 75–90.

Kardan Moghaddam, H.; Jafari, F.; and Javadi, S., 2017, Vulnerability evaluation of a coastal aquifer via GALDIT model and comparison with DRASTIC index using quality parameters: Hydrological Sciences Journal, Vol. 62, No. 1, pp. 137–146.

Karpf, C. and Krebs, P., 2004, Sewers as drainage systems— Quantification of groundwater infiltration. In Proceedings of the NOVATECH Conference:, IWA Pub. Lyon, France, pp.

969–975.

Kazemi G. A., 2011, Impact of urbanization on the groundwater resources in Shahrood, northeastern Iran: Comparison with other Iranian and Asian cities: Physics and Chemistry of the Earth, Vol. 36, pp. 150–159.

Khan, S. D., 2005, Urban development and flooding in Houston Texas: Inferences from remote sensing data using neural network technique: Environmental Geology, Vol. 47, pp. 1120–1127.

Kihumba, A. M.; Vanclooster, M.; and Longo, J. N., 2017, Assessing groundwater vulnerability in the Kinshasa region, DR Congo, using a calibrated DRASTIC model: Journal of African Earth Sciences, Vol. 126, pp. 13–22.

Klimas, A. A., 1995, Impacts of urbanization and protection of water resources in the Vilnius District, Lithuania: Hydrogeology Journal, Vol. 3, pp. 24–35.

Krogulec, E. and Trzeciak, J., 2017, DRASTIC assessment of groundwater vulnerability to pollution in the Vistula floodplain in central Poland: Hydrology Research, 48(4), 1088–1099.

Lerner, D. N., 2002, Identifying and quantifying urban recharge: A review: Hydrogeology Journal, Vol. 10, No. 1, pp. 143–152.

Liggett, J. E. and Talwar, S., 2009, Groundwater vulnerability assessments and integrated water resource management: Watershed Management Bulletin, Vol. 13, pp. 18–29.

Mair, A. and El-Kadi, A. I., 2013, Logistic regression modeling to assess groundwater vulnerability to contamination in Hawaii, USA: Journal of Contaminant Hydrology, Vol. 153, pp. 1–23.

Murray, K. E. and McCray, J. E., 2005, Development and application of a regional-scale pesticide transport and groundwater vulnerability model. Environmental & Engineering Geoscience, Vol. 11, No. 3, pp. 271–284.

Musekiwa, C. and Majola, K., 2013, Groundwater vulnerability map for South Africa: South African Journal of Geomatics, Vol. 2, pp. 152–163.

Popescu, I. C.; Gardin, N.; Brouyére, S.; and Dassargues, A., 2008, Groundwater vulnerability assessment using physicallybased modelling: From challenges to pragmatic solutions. In Refsgaard, J. C.; Kovar, K.; Haarder, E.; and Nygaard, E. (Editors), ModelCARE 2007 Proceedings, Calibration and Reliability in Groundwater Modelling Denmark: IAHS Publication 320, p. 83–88 International Association of Hydrological Sciences, London, U.K. Ragab, R.; Bromley, J.; Rosier, P.; Cooper, J. D.; and Gash, J. H. C., 2003, Experimental study of water fluxes in a residential area: 1. Rainfall, roof runoff and evaporation: The effect of slope and aspect. Hydrological Processes, Vol. 17, pp. 2409–2422.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

303

El May, M.; Dlala, M.; and Chenini, I., 2010, Urban geological mapping: Geotechnical data analysis for rational development planning: Engineering Geology, Vol. 116, pp. 129–138.

Farjad, B.; Mohd Shafri, H. Z.; Mohamed, T. A.; Pirasteh, S.; and Wijesekara, N., 2012, Groundwater intrinsic vulnerability and risk mapping: Water Management, Vol. 165, No. 8, pp.

441–450.

Fletcher, T. D.; Andrieu, H.; and Hamel, P., 2013. Understanding, management and modeling of urban hydrology and its consequences for receiving waters: A state of the art: Advances in Water Resources, Vol. 51, pp. 261–279.

Foster, S. S. D., 1987, Fundamental concepts in aquifer vulnerability, pollution risk and protection strategy. In van Duijvenbooden, W. and van Waegeningh, H. (Editors) the Netherlands Vulnerability of Soil and Groundwater to Pollutants, The Hague, Netherlands, 21, pp. 68–73.

Foster, S. S. D., 2007, Aquifer pollution vulnerability concept and tools—Use, benefits and constraints. In Witkowski, A. J.; Kowalczyk, A.; and Vrba, J. (Editors), Groundwater Vulnerability Assessment and Mapping: Taylor & Francis Group, London, U.K. 11, pp. 3–9.

Foster, S. S. D. and Hirata, R., 1988, Groundwater Pollution Risk Assessment: A Methodology Using Available Data: WHOPAHO/HPE-CEPIS Technical Manual, World Health Organization, Lima, Peru, 81 p.

Ghazavi, R. and Ebrahimi, Z., 2015, Assessing groundwater vulnerability to contamination in an arid environment using DRASTIC and GOD models: International Journal of Environmental Science and Technology, Vol. 12, No. 9, pp.

2909–2918.

Gogu, R. C. and Dassargues, A., 2000, Current trends and future challenges in groundwater vulnerability assessment using overlay and index methods: Environmental Geology, Vol. 39, No. 6, pp. 549–559.

Goldscheider, N.; Klute, M.; Sturm, S.; and Hotzl, H., 2000, The PI-method—A GIS based approach to mapping groundwater vulnerability with special consideration of karst aquifers: Zeitschrift Angewandte Geologie, Hannover, Vol. 46, No. 3, pp. 157–166.

Haertle, A., 1983, Method of Working and Employment of EDP during the Preparation of Groundwater Vulnerability Maps: IAHS Publication 142, International Association of Hydrological Sciences, London, U.K., pp. 1073–1085.

Hamza, S. M.; Ahsan, A.; Imteaz, M. A.; Ghazali, A. H.; and Mohammed, T. A., 2017, GIS-based FRASTIC model for pollution vulnerability assessment of fractured-rock aquifer systems: Environmental Earth Sciences, Vol. 76, No. 5, pp. 197.

Hayashi, T.; Tokunaga, T.; Aichi, M.; Shimada, J.; and Taniguchi, M., 2009, Effect of human activities and urbanization on groundwater environments: An example from the aquifer system of Tokyo and the surrounding area: Science of the Total Environment, Vol. 407, pp. 3165–3172.

Hibbs, B. J. and Sharp, J. M., 2012, Hydrogeological impacts of urbanization: Environmental and Engineering Geoscience, Vol. 18, No. 1, pp. 3–24.

Hölting, B.; Haertle, T.; Hohberger, K. H.; Nachtigall, K.; Villinger, E.; Weinzierl, W.; and Wrobel, J.-P., 1995, Concept to assess the protective function of the layers above the groundwater surface: Geologisches Jahrbuch der BGR Hannover, Ser. C, Vol. 63, pp. 5–24.

Huang, L.; Zeng, G.; Liang, J.; Hua, S.; Yuan, Y.; Li, X.; and Liu, J., 2017, Combined impacts of land use and climate change in the modeling of future groundwater vulnerability: Journal of Hydrologic Engineering, Vol. 22, No. 7, pp. 05017007. doi: 10.1061/(ASCE)HE.19435584.0001493#sthash.XiYpDY1J.

Hussain, Y.; Ullah, S. F.; Hussain, M. B.; Aslam, A. Q.; Akhter, G.; Martinez-Carvajal, H.; and Cárdenas-Soto, M., 2017, Modelling the vulnerability of groundwater to contamination in an unconfined alluvial aquifer in Pakistan: Environmental Earth Sciences, Vol. 76, No. 2, pp. 84.

Hwang, H. H.; Panno, S. V.; and Hackley, K. C., 2015, Sources and changes in groundwater quality with increasing urbanization, northeastern Illinois: Environmental & Engineering Geoscience, Vol. 21, No. 2, pp. 75–90.

Kardan Moghaddam, H.; Jafari, F.; and Javadi, S., 2017, Vulnerability evaluation of a coastal aquifer via GALDIT model and comparison with DRASTIC index using quality parameters: Hydrological Sciences Journal, Vol. 62, No. 1, pp. 137–146.

Karpf, C. and Krebs, P., 2004, Sewers as drainage systems— Quantification of groundwater infiltration. In Proceedings of the NOVATECH Conference:, IWA Pub. Lyon, France, pp.

969–975.

Kazemi G. A., 2011, Impact of urbanization on the groundwater resources in Shahrood, northeastern Iran: Comparison with other Iranian and Asian cities: Physics and Chemistry of the Earth, Vol. 36, pp. 150–159.

Khan, S. D., 2005, Urban development and flooding in Houston Texas: Inferences from remote sensing data using neural network technique: Environmental Geology, Vol. 47, pp. 1120–1127.

Kihumba, A. M.; Vanclooster, M.; and Longo, J. N., 2017, Assessing groundwater vulnerability in the Kinshasa region, DR Congo, using a calibrated DRASTIC model: Journal of African Earth Sciences, Vol. 126, pp. 13–22.

Klimas, A. A., 1995, Impacts of urbanization and protection of water resources in the Vilnius District, Lithuania: Hydrogeology Journal, Vol. 3, pp. 24–35.

Krogulec, E. and Trzeciak, J., 2017, DRASTIC assessment of groundwater vulnerability to pollution in the Vistula floodplain in central Poland: Hydrology Research, 48(4), 1088–1099.

Lerner, D. N., 2002, Identifying and quantifying urban recharge: A review: Hydrogeology Journal, Vol. 10, No. 1, pp. 143–152.

Liggett, J. E. and Talwar, S., 2009, Groundwater vulnerability assessments and integrated water resource management: Watershed Management Bulletin, Vol. 13, pp. 18–29.

Mair, A. and El-Kadi, A. I., 2013, Logistic regression modeling to assess groundwater vulnerability to contamination in Hawaii, USA: Journal of Contaminant Hydrology, Vol. 153, pp. 1–23.

Murray, K. E. and McCray, J. E., 2005, Development and application of a regional-scale pesticide transport and groundwater vulnerability model. Environmental & Engineering Geoscience, Vol. 11, No. 3, pp. 271–284.

Musekiwa, C. and Majola, K., 2013, Groundwater vulnerability map for South Africa: South African Journal of Geomatics, Vol. 2, pp. 152–163.

Popescu, I. C.; Gardin, N.; Brouyére, S.; and Dassargues, A., 2008, Groundwater vulnerability assessment using physicallybased modelling: From challenges to pragmatic solutions. In Refsgaard, J. C.; Kovar, K.; Haarder, E.; and Nygaard, E. (Editors), ModelCARE 2007 Proceedings, Calibration and Reliability in Groundwater Modelling Denmark: IAHS Publication 320, p. 83–88 International Association of Hydrological Sciences, London, U.K. Ragab, R.; Bromley, J.; Rosier, P.; Cooper, J. D.; and Gash, J. H. C., 2003, Experimental study of water fluxes in a residential area: 1. Rainfall, roof runoff and evaporation: The effect of slope and aspect. Hydrological Processes, Vol. 17, pp. 2409–2422.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

303

Chenini, Zghibi, Msaddek, and Dlala Ramier, D.; Berthier, E.; and Andrieu, H., 2011, The hydrological behaviour of urban streets: Long-term observations and modelling of runoff losses and rainfall, runoff transformation: Hydrological Processes, Vol. 25, pp. 2161–2178.

Rao, N. S., 2008, Groundwater quality monitoring in an urban area for sustainable development: Environmental Geosciences, Vol. 15, No. 2, pp. 63–73.

Schnebelen, N.; Platel, J. P.; LeNidre, Y.; and Baudry, D., 2002, Gestion des eaux Souterraines en Aquitaine Année 5.

Opération Sectorielle. Protection de la Nappe de l’Oligocène en Region Bordelaise: ERIC Document Reproduction Service 98– 01; 51178, France.

Shrestha, S.; Kafle, R.; and Pandey, V. P., 2017, Evaluation of index-overlay methods for groundwater vulnerability and risk assessment in Kathmandu Valley, Nepal: Science of the Total Environment, Vol. 575, pp. 779–790.

Singh, A.; Srivastav, S. K.; Kumar, S.; and Chakrapani, G. J., 2015, A modified-DRASTIC model (DRASTICA) for assessment of groundwater vulnerability to pollution in an urbanized environment in Lucknow, India: Environmental Earth Sciences, Vol. 74, No. 7, pp. 5475–5490.

Srinivasan, V.; Seto, K. C.; Emerson, R.; and Gorelick, S. M., 2013, The impact of urbanization on water vulnerability: A coupled human–environment system approach for

304

Chennai, India: Global Environmental Change, Vol. 23, No. 1, pp. 229–239.

Stempvoort, D. V.; Ewert, L.; and Wassenaar, L., 1993, Aquifer vulnerability index: A GIS-compatible method for groundwater vulnerability mapping: Canadian Water Resources Journal, Vol. 18, No. 1, pp. 25–37.

Sutherland, R. C., 1995, Methodology for estimating effective impervious area of urban watersheds: Watershed Protection Techniques, Vol. 2, pp. 282–293.

Taniguchi, M.; Uemura, T.; and Jago-on, K., 2007, Combined effects of urbanization and global warming on subsurface temperature in four Asian cities: Vadose Zone Journal, Vol. 6, pp.

591–596.

Trowsdale, S. A. and Lerner, D. N., 2007, A modelling approach to determine the origin of urban ground water: Journal of Contaminant Hydrology, Vol. 91, pp. 171–183.

Vrba, J. and Zaporozec, A., 1994, Guidebook on Mapping Groundwater Vulnerability: International Association of Hydrogeologists International Contributions to Hydrogeology 16, H. Heise, Hannover, Germany, 131 p.

Zilberbrand, M.; Rosenthal, E.; and Shachnai, E., 2001, Impact of urbanization on hydrochemical evolution of groundwater and on unsaturated-zone gas composition in the coastal city of Tel Aviv, Israel: Journal of Contaminant Hydrology, Vol. 50, pp. 175–208.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Chenini, Zghibi, Msaddek, and Dlala Ramier, D.; Berthier, E.; and Andrieu, H., 2011, The hydrological behaviour of urban streets: Long-term observations and modelling of runoff losses and rainfall, runoff transformation: Hydrological Processes, Vol. 25, pp. 2161–2178.

Rao, N. S., 2008, Groundwater quality monitoring in an urban area for sustainable development: Environmental Geosciences, Vol. 15, No. 2, pp. 63–73.

Schnebelen, N.; Platel, J. P.; LeNidre, Y.; and Baudry, D., 2002, Gestion des eaux Souterraines en Aquitaine Année 5.

Opération Sectorielle. Protection de la Nappe de l’Oligocène en Region Bordelaise: ERIC Document Reproduction Service 98– 01; 51178, France.

Shrestha, S.; Kafle, R.; and Pandey, V. P., 2017, Evaluation of index-overlay methods for groundwater vulnerability and risk assessment in Kathmandu Valley, Nepal: Science of the Total Environment, Vol. 575, pp. 779–790.

Singh, A.; Srivastav, S. K.; Kumar, S.; and Chakrapani, G. J., 2015, A modified-DRASTIC model (DRASTICA) for assessment of groundwater vulnerability to pollution in an urbanized environment in Lucknow, India: Environmental Earth Sciences, Vol. 74, No. 7, pp. 5475–5490.

Srinivasan, V.; Seto, K. C.; Emerson, R.; and Gorelick, S. M., 2013, The impact of urbanization on water vulnerability: A coupled human–environment system approach for

304

Chennai, India: Global Environmental Change, Vol. 23, No. 1, pp. 229–239.

Stempvoort, D. V.; Ewert, L.; and Wassenaar, L., 1993, Aquifer vulnerability index: A GIS-compatible method for groundwater vulnerability mapping: Canadian Water Resources Journal, Vol. 18, No. 1, pp. 25–37.

Sutherland, R. C., 1995, Methodology for estimating effective impervious area of urban watersheds: Watershed Protection Techniques, Vol. 2, pp. 282–293.

Taniguchi, M.; Uemura, T.; and Jago-on, K., 2007, Combined effects of urbanization and global warming on subsurface temperature in four Asian cities: Vadose Zone Journal, Vol. 6, pp.

591–596.

Trowsdale, S. A. and Lerner, D. N., 2007, A modelling approach to determine the origin of urban ground water: Journal of Contaminant Hydrology, Vol. 91, pp. 171–183.

Vrba, J. and Zaporozec, A., 1994, Guidebook on Mapping Groundwater Vulnerability: International Association of Hydrogeologists International Contributions to Hydrogeology 16, H. Heise, Hannover, Germany, 131 p.

Zilberbrand, M.; Rosenthal, E.; and Shachnai, E., 2001, Impact of urbanization on hydrochemical evolution of groundwater and on unsaturated-zone gas composition in the coastal city of Tel Aviv, Israel: Journal of Contaminant Hydrology, Vol. 50, pp. 175–208.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 293–304

Water-Mortar Interaction in a Tunnel Located in Southern Calabria (Southern Italy)

Water-Mortar Interaction in a Tunnel Located in Southern Calabria (Southern Italy)

GIOVANNI VESPASIANO* PASQUALINO NOTARO GIUSEPPE CIANFLONE

GIOVANNI VESPASIANO* PASQUALINO NOTARO GIUSEPPE CIANFLONE

E3 (Environment, Earth, Engineering) soc.coop., University of Calabria, P. Bucci Street, 87036, Arcavacata di Rende (CS), Italy

E3 (Environment, Earth, Engineering) soc.coop., University of Calabria, P. Bucci Street, 87036, Arcavacata di Rende (CS), Italy

Key Terms: Tunnel, Water-Mortar Interaction, Calabria, Southern Italy ABSTRACT In this work, we analyzed the results of a geochemical analysis aimed to define the origin of pH anomalies (pH > 11) in water samples collected inside a tunnel located in southern Calabria (southern Italy). We also analyzed the precipitates found close to the main drainage pipes. The hydrogeochemical study allowed us to identify a main NaOH water facies for the many samples collected close to the tunnel. In addition, the correlation diagrams highlighted high concentrations of Na, K, and Al, unrelated to simple water-rock interaction.

Further evaluation excluded the possibility that interaction between the water and the outcropping lithologies was the only cause of the ongoing processes. This consideration is supported by the high Na and K concentrations, which cannot be accounted for by interaction between water and calcareous marl. Excluding a natural origin and some anthropogenic factors, one possible explanation is an interaction between the groundwater and the mortars used for consolidation during the excavation phase of the tunnel. Mortar and concrete degradation in aqueous environments produces a great increase in pH, initially deriving from interstitial fluids containing strong alkali (NaOH and KOH) and non-negligible K and Na concentrations, such as we observed in the collected samples.

INTRODUCTION Some natural and/or anthropogenic chemicals inside geologic terrains and water can produce the degradation of cementitious grouts used during the consolidation and construction phases of underground cement structures, in particular, when these chemicals interact with the constituents of the cement. Cement paste is primarily made of relatively soluble calcium * Corresponding

author email: giovanni.vespasiano@unical.it

hydroxides, which are stable in strongly basic environments due to the presence of OH− ions, and alkalis, which are dissolved in the aqueous phase inside the capillary pores (Gascoyne, 2002; Soler and Mäder, 2010). Therefore, contact between the grout and an acidic environment (geologic terrains and water) can compromise this equilibrium, producing a degradation of their structural elements.

These chemical reactions (Prospect 1, UNI 11104:2016 Standard) are very common when cement structures come into contact with terrains and/or water containing chemicals that can react with the cement constituents. In the investigated area, as part of the renewal and upgrade works of the A2 Salerno–Reggio Calabria highway, southern Italy, environmental monitoring during the operational phase highlighted anomalous pH values in the waters, which drained through the permanent lining and reached the southern entrance of the tunnel via a drainage pipe.

These anomalous pH values are unexplainable solely in the context of water-rock interaction processes.

The aim of this work is to supply a hydrogeochemical model able to explain the contamination processes affecting these waters. We collected numerous water samples inside and outside the tunnel, and we sampled solids derived from precipitation processes. The final goals were to: (1) verify if the pH values measured in the waters inside the tunnel are due to anthropogenic alteration (i.e., mortars of the lining) or natural geochemical and/or hydrogeological processes, such as the interaction with the outcropping deposits; and (2) define the composition and the origin of the precipitates near the main drainage pipes.

GEOLOGICAL SETTING The study area is located along the western side of the Aspromonte Massif, which represents, together with the Peloritan Mountains, the southern termination of the Calabrian Arc. The origin of the arc was related to episodic Neogene roll-back of a NWdipping subduction zone with the associated opening of backarc basins in the western Mediterranean

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

305

Key Terms: Tunnel, Water-Mortar Interaction, Calabria, Southern Italy ABSTRACT In this work, we analyzed the results of a geochemical analysis aimed to define the origin of pH anomalies (pH > 11) in water samples collected inside a tunnel located in southern Calabria (southern Italy). We also analyzed the precipitates found close to the main drainage pipes. The hydrogeochemical study allowed us to identify a main NaOH water facies for the many samples collected close to the tunnel. In addition, the correlation diagrams highlighted high concentrations of Na, K, and Al, unrelated to simple water-rock interaction.

Further evaluation excluded the possibility that interaction between the water and the outcropping lithologies was the only cause of the ongoing processes. This consideration is supported by the high Na and K concentrations, which cannot be accounted for by interaction between water and calcareous marl. Excluding a natural origin and some anthropogenic factors, one possible explanation is an interaction between the groundwater and the mortars used for consolidation during the excavation phase of the tunnel. Mortar and concrete degradation in aqueous environments produces a great increase in pH, initially deriving from interstitial fluids containing strong alkali (NaOH and KOH) and non-negligible K and Na concentrations, such as we observed in the collected samples.

INTRODUCTION Some natural and/or anthropogenic chemicals inside geologic terrains and water can produce the degradation of cementitious grouts used during the consolidation and construction phases of underground cement structures, in particular, when these chemicals interact with the constituents of the cement. Cement paste is primarily made of relatively soluble calcium * Corresponding

author email: giovanni.vespasiano@unical.it

hydroxides, which are stable in strongly basic environments due to the presence of OH− ions, and alkalis, which are dissolved in the aqueous phase inside the capillary pores (Gascoyne, 2002; Soler and Mäder, 2010). Therefore, contact between the grout and an acidic environment (geologic terrains and water) can compromise this equilibrium, producing a degradation of their structural elements.

These chemical reactions (Prospect 1, UNI 11104:2016 Standard) are very common when cement structures come into contact with terrains and/or water containing chemicals that can react with the cement constituents. In the investigated area, as part of the renewal and upgrade works of the A2 Salerno–Reggio Calabria highway, southern Italy, environmental monitoring during the operational phase highlighted anomalous pH values in the waters, which drained through the permanent lining and reached the southern entrance of the tunnel via a drainage pipe.

These anomalous pH values are unexplainable solely in the context of water-rock interaction processes.

The aim of this work is to supply a hydrogeochemical model able to explain the contamination processes affecting these waters. We collected numerous water samples inside and outside the tunnel, and we sampled solids derived from precipitation processes. The final goals were to: (1) verify if the pH values measured in the waters inside the tunnel are due to anthropogenic alteration (i.e., mortars of the lining) or natural geochemical and/or hydrogeological processes, such as the interaction with the outcropping deposits; and (2) define the composition and the origin of the precipitates near the main drainage pipes.

GEOLOGICAL SETTING The study area is located along the western side of the Aspromonte Massif, which represents, together with the Peloritan Mountains, the southern termination of the Calabrian Arc. The origin of the arc was related to episodic Neogene roll-back of a NWdipping subduction zone with the associated opening of backarc basins in the western Mediterranean

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

305

Vespasiano, Notaro, and Cianflone

Figure 1. (a) Satellite view (from Google Earth) of the investigated area with location of the sampled wells. (b) Cross section of the tunnel with location of the exploratory boreholes, drain pipes, and water table (blue and dashed line). Section type refers to the consolidation method.

(Malinverno and Ryan, 1986; Faccenna et al., 1997).

This kind of mountain belt is primarily characterized by extensional tectonics acting along two principal directions, both parallel and orthogonal to the retreat direction of the subsiding lithosphere (Doglioni, 1995; Lenci et al., 2004).

Since the early Pleistocene, widespread uplift has characterized the Calabrian Arc (Westaway, 1993), while locally subsiding areas are also observed (Cianflone et al., 2015a, 2015b). The Aspromonte Massif is considered a southeast-verging nappe edifice, bounded to the north by a strike-slip fault system (Palmi Line) consisting of three stacked crystalline basement units (Ortolano et al., 2015). Extensional tectonics and normal fault systems, running along the Tyrrhenian coast, characterize the western massif side and are responsible for regional uplift, causing the development of widespread marine terraces (Dumas et al., 1999, 2005; Miyauchi et al., 1994). The nappes are unconformably covered by Upper Miocene to Pleistocene terrigenous deposits (Monaco et al., 1996).

Close to the investigated area, numerous boreholes and geological data collected during the tunnel excavation allow us to detail the local geology. We observed syn-sedimentary faults, which create depozones and abrupt lateral thickness variations in the depositional units, and recent tectonics responsible for the deformation of the original depositional features.

From bottom to top, we recognized the following lithostratigraphic units (Figure 1): (1) Pezzo Conglomerate, consisting of a texturally immature conglomerate derived from the physical/chemical alteration of the granitoid bedrock; (2) Trubi Formation, composed 306

Figure 2. Example of excavation and consolidation scheme, along longitudinal and transverse section, of the C1 section type. Table lists the different section types and the relative consolidation methods used during tunnel excavation. For each type section, the relative length for both northern and southern tunnel sections is reported.

by the alternation of marls and clays characterized by lateral variations in thickness due to tectonics; and (3) Messina Gravels Formation, made up of conglomerates with rounded, heterometric and polygenic pebbles alternating with layers of medium-coarse sand with light gray-yellowish color.

Tunnel Technical Characteristics The investigated tunnel is part of the renewal and upgrade works of the A2 Salerno–Reggio Calabria highway. Here, we briefly describe the techniques used during tunnel excavation. The excavation was carried out by proceeding from both tunnel entrances, along the four fronts, using truncated cone consolidated sections. The consolidation, both of the ground cavity and the core-face, consisted of mortar injections of different length, number, and kind (Figure 2), depending on the section type used. The tunnel excavation required several reinforcements and mortar injections, which, considering both the type and the technology used, can be included in the common digging practices by means of traditional boring machines.

MATERIALS AND METHODS After a preliminary study of all available geological data, a geochemical survey was carried out from May to July 2016. Ten water samples were collected and analyzed to determine the concentration of major dissolved constituents and select minor dissolved constituents.

After a preliminary purge operation, two wells, located outside and upgradient of the tunnel (background value), were sampled, whereas the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

Vespasiano, Notaro, and Cianflone

Figure 1. (a) Satellite view (from Google Earth) of the investigated area with location of the sampled wells. (b) Cross section of the tunnel with location of the exploratory boreholes, drain pipes, and water table (blue and dashed line). Section type refers to the consolidation method.

(Malinverno and Ryan, 1986; Faccenna et al., 1997).

This kind of mountain belt is primarily characterized by extensional tectonics acting along two principal directions, both parallel and orthogonal to the retreat direction of the subsiding lithosphere (Doglioni, 1995; Lenci et al., 2004).

Since the early Pleistocene, widespread uplift has characterized the Calabrian Arc (Westaway, 1993), while locally subsiding areas are also observed (Cianflone et al., 2015a, 2015b). The Aspromonte Massif is considered a southeast-verging nappe edifice, bounded to the north by a strike-slip fault system (Palmi Line) consisting of three stacked crystalline basement units (Ortolano et al., 2015). Extensional tectonics and normal fault systems, running along the Tyrrhenian coast, characterize the western massif side and are responsible for regional uplift, causing the development of widespread marine terraces (Dumas et al., 1999, 2005; Miyauchi et al., 1994). The nappes are unconformably covered by Upper Miocene to Pleistocene terrigenous deposits (Monaco et al., 1996).

Close to the investigated area, numerous boreholes and geological data collected during the tunnel excavation allow us to detail the local geology. We observed syn-sedimentary faults, which create depozones and abrupt lateral thickness variations in the depositional units, and recent tectonics responsible for the deformation of the original depositional features.

From bottom to top, we recognized the following lithostratigraphic units (Figure 1): (1) Pezzo Conglomerate, consisting of a texturally immature conglomerate derived from the physical/chemical alteration of the granitoid bedrock; (2) Trubi Formation, composed 306

Figure 2. Example of excavation and consolidation scheme, along longitudinal and transverse section, of the C1 section type. Table lists the different section types and the relative consolidation methods used during tunnel excavation. For each type section, the relative length for both northern and southern tunnel sections is reported.

by the alternation of marls and clays characterized by lateral variations in thickness due to tectonics; and (3) Messina Gravels Formation, made up of conglomerates with rounded, heterometric and polygenic pebbles alternating with layers of medium-coarse sand with light gray-yellowish color.

Tunnel Technical Characteristics The investigated tunnel is part of the renewal and upgrade works of the A2 Salerno–Reggio Calabria highway. Here, we briefly describe the techniques used during tunnel excavation. The excavation was carried out by proceeding from both tunnel entrances, along the four fronts, using truncated cone consolidated sections. The consolidation, both of the ground cavity and the core-face, consisted of mortar injections of different length, number, and kind (Figure 2), depending on the section type used. The tunnel excavation required several reinforcements and mortar injections, which, considering both the type and the technology used, can be included in the common digging practices by means of traditional boring machines.

MATERIALS AND METHODS After a preliminary study of all available geological data, a geochemical survey was carried out from May to July 2016. Ten water samples were collected and analyzed to determine the concentration of major dissolved constituents and select minor dissolved constituents.

After a preliminary purge operation, two wells, located outside and upgradient of the tunnel (background value), were sampled, whereas the

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

Water-Mortar Interaction in a Tunnel

remaining eight samples were collected inside the tunnel by the same number of exploratory boreholes and from external drain pipes. The locations of exploratory boreholes were chosen in relation to lithology and tunnel consolidation techniques carried out during excavation.

Furthermore, four solid samples were collected: three samples represented by a white precipitate found inside exploratory boreholes and drain pipes (samples PL5, PL6_7, and PLfi1, respectively), and one sample representative of the Trubi Formation. The latter was taken due to the high dissolution rates characterizing some of its phases, for example, calcite, which were much higher than the other phases characterizing crystalline basement and sedimentary sequences (Critelli et al., 2015; Vespasiano and Apollaro, 2016; and Vespasiano et al., 2016a). Phases characterized by high dissolution rates control the geochemistry of the fluids.

In the laboratories of the University of Calabria (Department of Biology, Ecology and Earth Science), rock samples were analyzed by X-ray diffraction (XRD) techniques. The methodology of water sampling and analysis was described by Apollaro et al. (2015, 2016) and Vespasiano et al. (2015a, 2015b, 2016b, 2016c, 2016d), but it is reported here below to make this contribution self-consistent. For each sample, temperature, pH, Eh, electrical conductivity (EC), total dissolved solids (TDS), total alkalinity, and reactive silica were measured in the field by means of portable instruments. Three pH buffers, with nominal pH values of 7, 10, and 12, at 25°C, were used for pH calibration at each site. The Eh equipment was tested against the ZoBell’s standard solution (Nordstrom, 1977). The Eh values were obtained by inserting the platinum electrode, coupled with an Ag/AgCl reference electrode, into the aqueous solution sample. The resulting oxidation-reduction potential was corrected for the difference between the standard potential of the Ag/AgCl reference electrode at the solution temperature and the potential of the standard hydrogen electrode (Nordstrom and Wilde, 2005).

Total alkalinity was determined by acidimetric titration using 0.05 N HCl as the titrating agent and methyl orange as the indicator. A specially designed microdosimeter was employed to minimize the amount of reactants, facilitating the analysis in the field. Reactive silica was measured by means of the colorimetric method involving the silico-molybdate complex (yellow form) using a portable colorimeter. Samples for the analysis of major anions, major cations, and minor dissolved constituents were filtered in situ through a membrane filter with 0.45 μm pore size, acidified by the addition of 1:1 HCl (only for major cations and minor dissolved constituents), and stored in cold (4°C) and dark conditions.

Water-Mortar Interaction in a Tunnel

In the laboratories of the University of Calabria (Department of Biology, Ecology and Earth Science), the concentrations of F− , Cl− , Br− , SO4 2− , NO3 − , Na+ , K+ , Mg2+ , and Ca2+ were measured by high-performance liquid chromatography (DIONEX ICS1100), whereas minor dissolved constituents were measured by inductively coupled plasma–mass spectrometry (ELAN DRC, Perkin Elmer SCIEX). Data quality for major components was estimated by charge balance. Deviations between the sum of concentrations of major cations and the sum of concentrations of major anions, both in equivalent units, varied between −5 percent and +5 percent. Data quality for minor dissolved constituents was checked running the NIST1643e standard reference solution. Deviations from the certified concentrations were less than 10 percent.

Unfortunately, it was impossible to obtain samples of the mortars used for cement injections. To cope with the lack of data, technical specifications listed on the conformity certificates of the various mixtures were examined. The documents showed that the mixtures used for several injections and consolidation work through columnar treatments (jet grouting) were constituted exclusively by the CEM II/AS 42.5R cement type (portland cement) and CEM III/AS 42.5R type, with the addition of additives such as Flowcable and bentonite. As reported in the technical features, CEM III/AS is a blast-furnace cement with strength class 42.5, with high initial resistance (R), which contains a clinker mass percentage (K) between 35 percent and 64 percent, and a total blast furnace slag content between 36 percent and 65 percent.

The slag shows a composition similar to Portland cement, generally characterized by a mass of at least 2/3 the sum of calcium oxide (CaO), magnesium oxide (MgO), and silicon dioxide (SiO2 ); the last fraction consists of aluminum oxide (Al2 O3 ) and small amounts of other compounds. Finally, CEM III has a percentage lower than 5 percent of selected secondary constituents. Analyses carried out on the bentonite samples revealed the following composition (percent v/v): silicon dioxide (52.9); aluminum oxide (12.8); calcium oxide (10.4); magnesium oxide (2.2); sodium oxide (3.4); potassium oxide (0.8); ferric oxide (3.2); titanium oxide (0.3); sulfur trioxide (0.23); manganese oxide (12.09); and phenol formaldehyde (13.1), with a pH value of 9.5. The results of the laboratory analyses and field data are shown in Tables 1, 2, and 3.

RESULTS The mineralogical compositions (percent v/v) of Trubi Formation samples and precipitates (PL6_7, PL5 e PLfi1) are reported in Table 1. Microscopically,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

307

remaining eight samples were collected inside the tunnel by the same number of exploratory boreholes and from external drain pipes. The locations of exploratory boreholes were chosen in relation to lithology and tunnel consolidation techniques carried out during excavation.

Furthermore, four solid samples were collected: three samples represented by a white precipitate found inside exploratory boreholes and drain pipes (samples PL5, PL6_7, and PLfi1, respectively), and one sample representative of the Trubi Formation. The latter was taken due to the high dissolution rates characterizing some of its phases, for example, calcite, which were much higher than the other phases characterizing crystalline basement and sedimentary sequences (Critelli et al., 2015; Vespasiano and Apollaro, 2016; and Vespasiano et al., 2016a). Phases characterized by high dissolution rates control the geochemistry of the fluids.

In the laboratories of the University of Calabria (Department of Biology, Ecology and Earth Science), rock samples were analyzed by X-ray diffraction (XRD) techniques. The methodology of water sampling and analysis was described by Apollaro et al. (2015, 2016) and Vespasiano et al. (2015a, 2015b, 2016b, 2016c, 2016d), but it is reported here below to make this contribution self-consistent. For each sample, temperature, pH, Eh, electrical conductivity (EC), total dissolved solids (TDS), total alkalinity, and reactive silica were measured in the field by means of portable instruments. Three pH buffers, with nominal pH values of 7, 10, and 12, at 25°C, were used for pH calibration at each site. The Eh equipment was tested against the ZoBell’s standard solution (Nordstrom, 1977). The Eh values were obtained by inserting the platinum electrode, coupled with an Ag/AgCl reference electrode, into the aqueous solution sample. The resulting oxidation-reduction potential was corrected for the difference between the standard potential of the Ag/AgCl reference electrode at the solution temperature and the potential of the standard hydrogen electrode (Nordstrom and Wilde, 2005).

Total alkalinity was determined by acidimetric titration using 0.05 N HCl as the titrating agent and methyl orange as the indicator. A specially designed microdosimeter was employed to minimize the amount of reactants, facilitating the analysis in the field. Reactive silica was measured by means of the colorimetric method involving the silico-molybdate complex (yellow form) using a portable colorimeter. Samples for the analysis of major anions, major cations, and minor dissolved constituents were filtered in situ through a membrane filter with 0.45 μm pore size, acidified by the addition of 1:1 HCl (only for major cations and minor dissolved constituents), and stored in cold (4°C) and dark conditions.

In the laboratories of the University of Calabria (Department of Biology, Ecology and Earth Science), the concentrations of F− , Cl− , Br− , SO4 2− , NO3 − , Na+ , K+ , Mg2+ , and Ca2+ were measured by high-performance liquid chromatography (DIONEX ICS1100), whereas minor dissolved constituents were measured by inductively coupled plasma–mass spectrometry (ELAN DRC, Perkin Elmer SCIEX). Data quality for major components was estimated by charge balance. Deviations between the sum of concentrations of major cations and the sum of concentrations of major anions, both in equivalent units, varied between −5 percent and +5 percent. Data quality for minor dissolved constituents was checked running the NIST1643e standard reference solution. Deviations from the certified concentrations were less than 10 percent.

Unfortunately, it was impossible to obtain samples of the mortars used for cement injections. To cope with the lack of data, technical specifications listed on the conformity certificates of the various mixtures were examined. The documents showed that the mixtures used for several injections and consolidation work through columnar treatments (jet grouting) were constituted exclusively by the CEM II/AS 42.5R cement type (portland cement) and CEM III/AS 42.5R type, with the addition of additives such as Flowcable and bentonite. As reported in the technical features, CEM III/AS is a blast-furnace cement with strength class 42.5, with high initial resistance (R), which contains a clinker mass percentage (K) between 35 percent and 64 percent, and a total blast furnace slag content between 36 percent and 65 percent.

The slag shows a composition similar to Portland cement, generally characterized by a mass of at least 2/3 the sum of calcium oxide (CaO), magnesium oxide (MgO), and silicon dioxide (SiO2 ); the last fraction consists of aluminum oxide (Al2 O3 ) and small amounts of other compounds. Finally, CEM III has a percentage lower than 5 percent of selected secondary constituents. Analyses carried out on the bentonite samples revealed the following composition (percent v/v): silicon dioxide (52.9); aluminum oxide (12.8); calcium oxide (10.4); magnesium oxide (2.2); sodium oxide (3.4); potassium oxide (0.8); ferric oxide (3.2); titanium oxide (0.3); sulfur trioxide (0.23); manganese oxide (12.09); and phenol formaldehyde (13.1), with a pH value of 9.5. The results of the laboratory analyses and field data are shown in Tables 1, 2, and 3.

RESULTS The mineralogical compositions (percent v/v) of Trubi Formation samples and precipitates (PL6_7, PL5 e PLfi1) are reported in Table 1. Microscopically,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

307

308 43 67 100 100

Trubi (TR2) PL5 PL6_7 PLfi1

18 7 — —

SiO2 (Quartz) 17 — — —

K0,5 (Al;Fe;Mg)3 (Si;Al)4 O10 (OH)2 (Illite) 4 — — —

NaAlSi3 O8 (Albite) 11 — — —

(Ca;Na)(Si;Al)4 O8 (Anorthite) 7 — — —

Mg2 Al3 (Si3 Al)O10 (O)8 (Chlorite) — 21 — —

(MgFe)9 (SiAl)8 O20 OH10 *H2 O (Corrensite)

— 5 — —

KAlSiO3 O11 (Mica)

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

308 43 67 100 100

Trubi (TR2) PL5 PL6_7 PLfi1

18 7 — —

—

—

0.48 0.49 1.77 1.77 1 0.42 0.38 0.28 0.41 0.45

—

—

673 695 2356 2352 1368 595 542 401 584 638

—

—

9.16 11.49 60.83 59.63 38.29 16.67 23.46 12.92 57.67 83.68

—

—

5.91 4.65 0.90 0.01 0.16 8.94 11.13 3.59 19.39 16.19

—

—

24.78 26.15 64.79 63.08 57.38 20.02 15.25 17.10 4.89 3.41

—

—

73.01 76.30 190.01 187.09 160.94 63.95 57.06 58.20 47.03 40.17

—

—

0.34 0.39 4.21 3.46 2.71 0.68 0.73 0.37 0.34 0.82

—

—

39.09 39.29 53.14 50.97 48.90 40.76 33.13 41.17 77.18 69.25

0.53 0.47 2.16 2.01 1.37 0.54

0.28

250

250 —

0.5

74.45 74.52 52.58 50.72 60.39 72.04 63.26 78.59 101.50 125.00

500

250

17 — — —

4 — — —

NaAlSi3 O8 (Albite)

11 — — —

(Ca;Na)(Si;Al)4 O8 (Anorthite)

7 — — —

Mg2 Al3 (Si3 Al)O10 (O)8 (Chlorite)

Table 1. Mineralogical composition (percent v/v) of samples TR2, PL5, PL6_7, and PLfi1.

19.13 18.9 20.13 19.66 19.69 20.07 18.72 18.94 19.77 18.88

—

—

K0,5 (Al;Fe;Mg)3 (Si;Al)4 O10 (OH)2 (Illite)

963 994 3366 3359 1952 850 774 572 835 911

SiO2 (Quartz)

74.3 90 91.7 96.9 91.1 86.4 100.3 83.6 88.5 105.2

—

—

1.56 1.52 3.35 1.69 1.66 1.54 1.43 1.92 2.18 1.63

1.5

1.5

35.51 34.14 11.41 13.93 26.19 33.91 28.03 46.09 35.45 35.63

50

50

— 21 — —

(MgFe)9 (SiAl)8 O20 OH10 *H2 O (Corrensite)

128.14 146.44 784.07 793.22 575.09 129.66 138.81 76.27 213.56 242.54

—

—

0.32

—

—

— 5 — —

KAlSiO3 O11 (Mica)

7.68 8.83 51.74 53.14 41.61 10.15 7.65 0.49

3

0.5

11.14 11.18 11.75 11.67 11.58 10.89 8.7 9.65 7.79 7.62

—

WHO guideline value PLFi1 PLFi2 PL1 PL2 PL5 PL7 PL8 PL9 Pozzo5 PozzoX

—

74.3 90 91.7 96.9 91.1 86.4 100.3 83.6 88.5 105.2

—

—

963 994 3366 3359 1952 850 774 572 835 911

—

—

19.13 18.9 20.13 19.66 19.69 20.07 18.72 18.94 19.77 18.88

—

—

0.48 0.49 1.77 1.77 1 0.42 0.38 0.28 0.41 0.45

—

—

673 695 2356 2352 1368 595 542 401 584 638

—

—

9.16 11.49 60.83 59.63 38.29 16.67 23.46 12.92 57.67 83.68

—

—

5.91 4.65 0.90 0.01 0.16 8.94 11.13 3.59 19.39 16.19

—

—

24.78 26.15 64.79 63.08 57.38 20.02 15.25 17.10 4.89 3.41

—

—

73.01 76.30 190.01 187.09 160.94 63.95 57.06 58.20 47.03 40.17

—

—

0.34 0.39 4.21 3.46 2.71 0.68 0.73 0.37 0.34 0.82

—

—

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

39.09 39.29 53.14 50.97 48.90 40.76 33.13 41.17 77.18 69.25

0.53 0.47 2.16 2.01 1.37 0.54

0.28

250

250 —

0.5

74.45 74.52 52.58 50.72 60.39 72.04 63.26 78.59 101.50 125.00

500

250

128.14 146.44 784.07 793.22 575.09 129.66 138.81 76.27 213.56 242.54

—

—

1.56 1.52 3.35 1.69 1.66 1.54 1.43 1.92 2.18 1.63

1.5

1.5

35.51 34.14 11.41 13.93 26.19 33.91 28.03 46.09 35.45 35.63

50

50

7.68 8.83 51.74 53.14 41.61 10.15 7.65 0.49

3

0.5

0.32

—

—

Eh EC T Salinity TDS Ca Mg K Na Sr NH4 Cl SO4 Alkt (mg/L) F− NO3 NO2 Br pH (mV) (mS/cm) (°C) (p.s.u.) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) as HCO− 3 (mg/L) (mg/L) (mg/L) (mg/L) Thresholds for groundwater in Italian law D. Lgs. 152/2006

Sample

Table 2. Concentrations of chemical-physical parameters and major chemical components in water samples from the study area, where HCO3 represents total alkalinity (Alkt) in mg HCO3 /L. In italic and bold are highlighted the thresholds for groundwater in Italian law and values over threshold respectively.

CaCO3 (Calcite) ID

11.14 11.18 11.75 11.67 11.58 10.89 8.7 9.65 7.79 7.62

—

WHO guideline value PLFi1 PLFi2 PL1 PL2 PL5 PL7 PL8 PL9 Pozzo5 PozzoX

—

Thresholds for groundwater in Italian law D. Lgs. 152/2006

Sample

Eh EC T Salinity TDS Ca Mg K Na Sr NH4 Cl SO4 Alkt (mg/L) F− NO3 NO2 Br pH (mV) (mS/cm) (°C) (p.s.u.) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) as HCO− 3 (mg/L) (mg/L) (mg/L) (mg/L)

Table 2. Concentrations of chemical-physical parameters and major chemical components in water samples from the study area, where HCO3 represents total alkalinity (Alkt) in mg HCO3 /L. In italic and bold are highlighted the thresholds for groundwater in Italian law and values over threshold respectively.

CaCO3 (Calcite)

ID

Table 1. Mineralogical composition (percent v/v) of samples TR2, PL5, PL6_7, and PLfi1.

Vespasiano, Notaro, and Cianflone Vespasiano, Notaro, and Cianflone

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

309

0.21 0.21 0.88 0.36 0.37 0.37 0.18 0.46 0.18 0.22 n.d. n.d.

7.52 0.14 4.05 0.08 0.07 0.21 0.05 0.14 60.86 78.28 50.81 52.65 72.58 48.13 54.53 29.21 42.22 78.34 73.80 68.97 226.58 223.62 155.28 66.05 101.65 33.78 29.98 75.00 n.d.

0.02 0.10 n.d.

0.03 n.d.

0.01 n.d.

0.05 0.05 0.30 0.28 0.51 0.53 0.50 0.42 0.52 0.62 0.56 0.82 0.16 0.29 0.77 0.11 0.21 0.12 0.03 0.43 0.63 0.29 0.89 0.70 0.01 0.01 0.03 0.37 1.88 0.31 0.78 1.14 65.76 7.54 71.20 7.91 161.38 13.32 175.72 12.46 166.24 10.76 54.12 4.22 37.49 3.68 48.69 6.05 3.19 2.42 5.15 0.96 1.49 1.53 2.68 2.68 2.44 1.54 1.54 1.39 1.26 1.82 878.2 919.7 5120.6 4628.8 3446.7 1074.5 1184.2 399.0 406.3 838.4 12.83 n.d.

16.33 n.d.

24.05 14.87 n.d. n.d.

67.69 13.10 1.48 1.51 7.27 2.83 3.17 0.67 1.39 0.91 0.10 1.81 2.08 1.36 3.34 2.22 2.28 1.02 2.14 0.26 1.67 3.29 0.27 0.27 0.53 0.48 0.43 0.16 0.22 0.13 0.11 0.24 0.03 0.03 1.00 0.02 0.16 0.05 n.d.

0.77 5.69 7.60 3.32 3.57 2.31 2.36 3.17 2.11 3.05 2.83 0.37 3.63 6.45 6.32 6.58 6.86 6.18 2.04 3.90 7.34 0.34 5.80 47.48 23.75 48.39 7.43 31.91 810.53 35.25 678.04 50.02 244.97 39.00 31.24 53.77 47.18 54.70 51.16 85.10 n.d.

67.40 3.76 n.d. = no data.

11.12 12.27 34.27 37.01 37.23 18.27 11.69 9.64 14.77 9.14 PLFi1 PLFi2 PL1 PL2 PL5 PL7 PL8 PL9 Pozzo5 PozzoX

20 — — — 3 10 10 — — — 40 — 4000 2000 70 — 400 50 — 200 — WHO guideline value

—

5 — — — 5 10 10 — — — 10 — 3000 1000 20 50 50 50 — 200 1000 — Thresholds for groundwater in Italian law D. Lgs. 152/2006

Li B Al V Cr Mn Co Ni Cu Zn Sr Se Rb Mo U Pb As Cd Ba Ti Sn Sb (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb)

precipitates showed a granular appearance, incoherent when hydrated, and solid with low resistance and greasy to the touch once dried. The greasiness to the touch falls within the peculiar characteristics of the slaked lime formed from the dispersion in water of Ca(OH)2 (Marchese, 1972).

As reported in Table 1, samples PL6_7 and PLfi1 are constituted exclusively by calcite. In contrast, sample PL5 is constituted also by different secondary phases. Calcite remains the main constituent, followed by quartz, phyllosilicates, and corrensite. Finally, sample TR2 (Trubi), as expected, is made up of calcite (∼43 percent), followed by quartz, illite, albite, anorthite, and chlorite.

Successively, water characterization was carried out.

Different flow rates were identified in the examined boreholes. PL1 and PL2 showed no or very low flow rates with rising values moving towards the drain pipes. Unfortunately, it was not possible to measure flow-rate values. In relation to chemical-physical parameters, water samples showed an average temperature of 19.4 ± 0.5°C (1σ), with little variability. Comparable Eh was measured, with values between 74 mV and 105 mV. High variabilities were found for pH and EC. Waters sampled from exploratory boreholes showed values ranging from 572 μS/cm (PL9) to 3,366 μS/cm (PL1), whereas waters sampled from drain pipes (PLfi1 e PLfi2) and private wells (P5 and PX) showed similar values to each other, with an average of 978 μS/cm and 873 μS/cm, respectively. High EC values in samples PL1 and PL2 (3,366 and 3,359 μS/cm, respectively) could be attributable to evaporative effects linked to stagnant water resulting from independent flow conditions or to a long residence time that promotes prolonged water-rock interaction, whereas, for downstream boreholes, less residence time and calcite precipitation could represent the main cause of decreases of salinity and EC. Finally, pH values showed the highest anomalies. All exploratory boreholes showed values higher than 11. Only samples PL8 and PL9 showed lower values, 8.7 and 9.6, respectively.

Private well values never surpassed 7.8, and chemicalphysical characteristics were generally lower than the Italian law threshold (Prospect 1, UNI 11104:2016 Standard).

Water chemistry was examined by means of: (1) triangular plots involving major cations and major anions (Figure 3), both prepared starting from the concentrations in equivalent units, as this approach is more suitable than the one involving concentrations in weight units (e.g., Zaporozec, 1972); (2) the correlation graph of SO4 + Cl vs. HCO3 , in which iso-salinity lines are drawn for reference (Figure 4); (3) the correlation diagram pH vs. PCO2 (Figure 5); and (4) the correlation graphs of Cl vs. other major and trace constituents

Sample

n.d. = no data.

Water-Mortar Interaction in a Tunnel Table 3. Concentrations of trace chemical components in water samples from the study area. In italic and bold are highlighted the thresholds for groundwater in Italian law and values over threshold respectively.

0.21 0.21 0.88 0.36 0.37 0.37 0.18 0.46 0.18 0.22 n.d. n.d.

7.52 0.14 4.05 0.08 0.07 0.21 0.05 0.14 60.86 78.28 50.81 52.65 72.58 48.13 54.53 29.21 42.22 78.34 73.80 68.97 226.58 223.62 155.28 66.05 101.65 33.78 29.98 75.00 n.d.

0.02 0.10 n.d.

0.03 n.d.

0.01 n.d.

0.05 0.05 0.30 0.28 0.51 0.53 0.50 0.42 0.52 0.62 0.56 0.82 0.16 0.29 0.77 0.11 0.21 0.12 0.03 0.43 0.63 0.29 0.89 0.70 0.01 0.01 0.03 0.37 1.88 0.31 0.78 1.14 65.76 7.54 71.20 7.91 161.38 13.32 175.72 12.46 166.24 10.76 54.12 4.22 37.49 3.68 48.69 6.05 3.19 2.42 5.15 0.96 1.49 1.53 2.68 2.68 2.44 1.54 1.54 1.39 1.26 1.82 878.2 919.7 5120.6 4628.8 3446.7 1074.5 1184.2 399.0 406.3 838.4 12.83 n.d.

16.33 n.d.

24.05 14.87 n.d. n.d.

67.69 13.10 1.48 1.51 7.27 2.83 3.17 0.67 1.39 0.91 0.10 1.81 2.08 1.36 3.34 2.22 2.28 1.02 2.14 0.26 1.67 3.29 0.27 0.27 0.53 0.48 0.43 0.16 0.22 0.13 0.11 0.24 0.03 0.03 1.00 0.02 0.16 0.05 n.d.

0.77 5.69 7.60 3.32 3.57 2.31 2.36 3.17 2.11 3.05 2.83 0.37 3.63 6.45 6.32 6.58 6.86 6.18 2.04 3.90 7.34 0.34 5.80 47.48 23.75 48.39 7.43 31.91 810.53 35.25 678.04 50.02 244.97 39.00 31.24 53.77 47.18 54.70 51.16 85.10 n.d.

67.40 3.76 11.12 12.27 34.27 37.01 37.23 18.27 11.69 9.64 14.77 9.14 PLFi1 PLFi2 PL1 PL2 PL5 PL7 PL8 PL9 Pozzo5 PozzoX

20 — — — 3 10 10 — — — 40 — 4000 2000 70 — 400 50 — 200 — WHO guideline value

—

5 — — — 5 10 10 — — — 10 — 3000 1000 20 50 50 50 — 200 1000 — Thresholds for groundwater in Italian law D. Lgs. 152/2006

Li B Al V Cr Mn Co Ni Cu Zn Sr Se Rb Mo U Pb As Cd Ba Ti Sn Sb (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) (ppb) Sample

Table 3. Concentrations of trace chemical components in water samples from the study area. In italic and bold are highlighted the thresholds for groundwater in Italian law and values over threshold respectively.

Water-Mortar Interaction in a Tunnel

precipitates showed a granular appearance, incoherent when hydrated, and solid with low resistance and greasy to the touch once dried. The greasiness to the touch falls within the peculiar characteristics of the slaked lime formed from the dispersion in water of Ca(OH)2 (Marchese, 1972).

As reported in Table 1, samples PL6_7 and PLfi1 are constituted exclusively by calcite. In contrast, sample PL5 is constituted also by different secondary phases. Calcite remains the main constituent, followed by quartz, phyllosilicates, and corrensite. Finally, sample TR2 (Trubi), as expected, is made up of calcite (∼43 percent), followed by quartz, illite, albite, anorthite, and chlorite.

Successively, water characterization was carried out.

Different flow rates were identified in the examined boreholes. PL1 and PL2 showed no or very low flow rates with rising values moving towards the drain pipes. Unfortunately, it was not possible to measure flow-rate values. In relation to chemical-physical parameters, water samples showed an average temperature of 19.4 ± 0.5°C (1σ), with little variability. Comparable Eh was measured, with values between 74 mV and 105 mV. High variabilities were found for pH and EC. Waters sampled from exploratory boreholes showed values ranging from 572 μS/cm (PL9) to 3,366 μS/cm (PL1), whereas waters sampled from drain pipes (PLfi1 e PLfi2) and private wells (P5 and PX) showed similar values to each other, with an average of 978 μS/cm and 873 μS/cm, respectively. High EC values in samples PL1 and PL2 (3,366 and 3,359 μS/cm, respectively) could be attributable to evaporative effects linked to stagnant water resulting from independent flow conditions or to a long residence time that promotes prolonged water-rock interaction, whereas, for downstream boreholes, less residence time and calcite precipitation could represent the main cause of decreases of salinity and EC. Finally, pH values showed the highest anomalies. All exploratory boreholes showed values higher than 11. Only samples PL8 and PL9 showed lower values, 8.7 and 9.6, respectively.

Private well values never surpassed 7.8, and chemicalphysical characteristics were generally lower than the Italian law threshold (Prospect 1, UNI 11104:2016 Standard).

Water chemistry was examined by means of: (1) triangular plots involving major cations and major anions (Figure 3), both prepared starting from the concentrations in equivalent units, as this approach is more suitable than the one involving concentrations in weight units (e.g., Zaporozec, 1972); (2) the correlation graph of SO4 + Cl vs. HCO3 , in which iso-salinity lines are drawn for reference (Figure 4); (3) the correlation diagram pH vs. PCO2 (Figure 5); and (4) the correlation graphs of Cl vs. other major and trace constituents

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

309

Vespasiano, Notaro, and Cianflone

Vespasiano, Notaro, and Cianflone

Figure 4. Correlation diagram for SO4 + Cl vs. HCO3 prepared from concentrations in equivalent units.

Figure 4. Correlation diagram for SO4 + Cl vs. HCO3 prepared from concentrations in equivalent units.

In relation of the high pH values measured in the exploratory boreholes, it is fair to assume that HCO3 does not represent the dominant anion. For this reason, a speciation calculation was carried out in the EQ3/6 8.0a software package using analytical data and constraining the alkalinity value with calcite saturation. This constraint is reasonable since the waters

In relation of the high pH values measured in the exploratory boreholes, it is fair to assume that HCO3 does not represent the dominant anion. For this reason, a speciation calculation was carried out in the EQ3/6 8.0a software package using analytical data and constraining the alkalinity value with calcite saturation. This constraint is reasonable since the waters

Figure 3. Triangular plots of (a) major cations and (b) major anions, both prepared from concentrations in equivalent units.

Figure 3. Triangular plots of (a) major cations and (b) major anions, both prepared from concentrations in equivalent units.

(Figures 6 and 7). These plots indicate that the following.

All samples showed a Na-HCO3 composition, where Na largely prevailed over K, with a Na/K ratio of 5.2 ± 0.5 in equivalent units, and where HCO3 prevailed over SO4 and Cl. The only exceptions were found for wells P5 and PX and exploratory borehole PL9, which showed Ca-HCO3 and Na-SO4 compositions, respectively. HCO3 waters are typical for natural water-rock interaction processes where CO2 converts into HCO3 . They represent the typical natural groundwater composition. Figure 3b depicts an evident linear distribution starting from the HCO3 vertex.

This represents a constant SO4 /Cl ratio and different HCO3 /Cl and HCO3 /SO4 ratios. The trend is easily explicable by calcite precipitation processes, confirmed

by XRD analysis.

(Figures 6 and 7). These plots indicate that the following.

All samples showed a Na-HCO3 composition, where Na largely prevailed over K, with a Na/K ratio of 5.2 ± 0.5 in equivalent units, and where HCO3 prevailed over SO4 and Cl. The only exceptions were found for wells P5 and PX and exploratory borehole PL9, which showed Ca-HCO3 and Na-SO4 compositions, respectively. HCO3 waters are typical for natural water-rock interaction processes where CO2 converts into HCO3 . They represent the typical natural groundwater composition. Figure 3b depicts an evident linear distribution starting from the HCO3 vertex.

This represents a constant SO4 /Cl ratio and different HCO3 /Cl and HCO3 /SO4 ratios. The trend is easily explicable by calcite precipitation processes, confirmed

by XRD analysis.

310

Figure 5. Correlation diagram for pH vs. PCO2 for the water samples collected in the study area. Figure reports the mean concentration values for soil and atmospheric PCO2 (Brook et al., 1983).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

310

Figure 5. Correlation diagram for pH vs. PCO2 for the water samples collected in the study area. Figure reports the mean concentration values for soil and atmospheric PCO2 (Brook et al., 1983).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

Water-Mortar Interaction in a Tunnel

Figure 6. Correlation diagram for Cl vs. Ca and Mg for the water samples collected in the study area.

have shown precipitation in place. Speciation calculations revealed that the dominant anion for waters from the exploratory boreholes and drain pipes appears to be OH− , highlighting a sodium hydroxide geochemical composition. The only exceptions were the PL8 and PL9 exploratory boreholes, for which the dominant anion was HCO3 . The absolute concentration based on the correlation diagram of SO4 + Cl vs. HCO3 is shown in Figure 4.

The diagram shows different total ionic salinity (TIS) intervals for all chemical types.

Moreover, it is useful to observe the correlation diagram pH vs. PCO2 in Figure 5. The partial pressure of CO2 was calculated by EQ3/6 v.8.0a (Wolery and Jarek, 2003) using the latest thermodynamic database (Wolery and Jove-Colon, 2007; Wolery, T. J., 2013). As shown in the figure, it is evident that most of samples fall below the fields of soil PCO2 (0.002–0.04 bar; Brook et al., 1983) and mean atmospheric PCO2 (10−3.5 bar).

This is supported by the high pH values and allows us to assume a state of isolation from the atmosphere.

Samples PL8 and PL9 fall in the proximity of the mean atmospheric line, likely due to partial exchange with the atmosphere. The latter, as previously mentioned, is situated in the southern sector of the tunnel and has shown the highest flow rates. Only samples P5 and PX (Ca-HCO3 chemical type) were positioned in the proximity of the soil field, showing PCO2 values between 0.003 and 0.005 bar and pH values between 7.62

Figure 7. Correlation diagram for Cl vs. Na, K, and Al for the water samples collected in the study area.

Water-Mortar Interaction in a Tunnel

and 7.79. These represent typical values of water that is constantly replenished by surface contributions linked to biological activity.

Correlation diagrams for Cl vs. other constituents represent a useful instrument to improve the knowledge of processes in place and those that occur during water evolution. The diagrams report the mean seawater dilution line (data from Nordstrom et al., 1979), which is representative of the meteoric water composition. On the correlation diagram Cl vs. Ca (Figure 6), all samples fall above the seawater dilution line, which reflects the primordial meteoric component. In particular, samples PL1 and PL2 show Ca concentrations comparable with those of private wells (PX and P5), whereas lower concentrations were detected for the other exploratory boreholes.

The decrease in Ca concentration could be linked to calcite precipitation. Calcium intakes may be attributable to interactions with the Trubi Formation upgradient and/or to the interaction with the portlandite characterizing the cement mortars. This latter process could also explain the high pH. The diagram Cl vs. Mg (Figure 6) shows a decrease in magnesium concentration. This phenomenon is easily explained by the extremely basic conditions of the solution. Mg2+ is stable in solution until pH values of 11. Beyond this threshold, the solid phase dominates, decreasing its concentration in solution. Interesting data are provided by the XRD analysis, which revealed the presence of Mg-corrensite in the precipitate sampled at exploratory borehole PL5.

Regarding the correlation diagrams of Cl vs. Na and K (Figure 7), it is important to emphasize the different concentrations encountered between the private wells and the exploratory boreholes (including drain pipes).

In fact, private wells fall close to the seawater dilution line, pointing out the absence of important water-rock interaction processes that can release these two elements in solution. In contrast, water sampled in the exploratory boreholes and along drain pipes showed concentration peaks that would be unjustifiable from mere water-rock interaction processes, and due to the small extension of the examined area. It is conceivable that sodium and potassium were derived from nonnatural factors. Possible sources could be the cement mortars and the percentages of sodium hydroxide and potassium hydroxide within.

Finally, the correlation diagram for Cl vs. Al (Figure 7) highlights the high Al concentrations. In the case of aluminum, the pH affects the mobility of the element. At extreme pH values (low or high), the mobility of the element increases, and, consequently, so do the concentrations found in the solution. The availability of the element can be linked to aluminum oxides (Al2 O3 ) characterizing the cement mixtures used.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

311

Figure 6. Correlation diagram for Cl vs. Ca and Mg for the water samples collected in the study area.

have shown precipitation in place. Speciation calculations revealed that the dominant anion for waters from the exploratory boreholes and drain pipes appears to be OH− , highlighting a sodium hydroxide geochemical composition. The only exceptions were the PL8 and PL9 exploratory boreholes, for which the dominant anion was HCO3 . The absolute concentration based on the correlation diagram of SO4 + Cl vs. HCO3 is shown in Figure 4.

The diagram shows different total ionic salinity (TIS) intervals for all chemical types.

Moreover, it is useful to observe the correlation diagram pH vs. PCO2 in Figure 5. The partial pressure of CO2 was calculated by EQ3/6 v.8.0a (Wolery and Jarek, 2003) using the latest thermodynamic database (Wolery and Jove-Colon, 2007; Wolery, T. J., 2013). As shown in the figure, it is evident that most of samples fall below the fields of soil PCO2 (0.002–0.04 bar; Brook et al., 1983) and mean atmospheric PCO2 (10−3.5 bar).

This is supported by the high pH values and allows us to assume a state of isolation from the atmosphere.

Samples PL8 and PL9 fall in the proximity of the mean atmospheric line, likely due to partial exchange with the atmosphere. The latter, as previously mentioned, is situated in the southern sector of the tunnel and has shown the highest flow rates. Only samples P5 and PX (Ca-HCO3 chemical type) were positioned in the proximity of the soil field, showing PCO2 values between 0.003 and 0.005 bar and pH values between 7.62

Figure 7. Correlation diagram for Cl vs. Na, K, and Al for the water samples collected in the study area.

and 7.79. These represent typical values of water that is constantly replenished by surface contributions linked to biological activity.

Correlation diagrams for Cl vs. other constituents represent a useful instrument to improve the knowledge of processes in place and those that occur during water evolution. The diagrams report the mean seawater dilution line (data from Nordstrom et al., 1979), which is representative of the meteoric water composition. On the correlation diagram Cl vs. Ca (Figure 6), all samples fall above the seawater dilution line, which reflects the primordial meteoric component. In particular, samples PL1 and PL2 show Ca concentrations comparable with those of private wells (PX and P5), whereas lower concentrations were detected for the other exploratory boreholes.

The decrease in Ca concentration could be linked to calcite precipitation. Calcium intakes may be attributable to interactions with the Trubi Formation upgradient and/or to the interaction with the portlandite characterizing the cement mortars. This latter process could also explain the high pH. The diagram Cl vs. Mg (Figure 6) shows a decrease in magnesium concentration. This phenomenon is easily explained by the extremely basic conditions of the solution. Mg2+ is stable in solution until pH values of 11. Beyond this threshold, the solid phase dominates, decreasing its concentration in solution. Interesting data are provided by the XRD analysis, which revealed the presence of Mg-corrensite in the precipitate sampled at exploratory borehole PL5.

Regarding the correlation diagrams of Cl vs. Na and K (Figure 7), it is important to emphasize the different concentrations encountered between the private wells and the exploratory boreholes (including drain pipes).

In fact, private wells fall close to the seawater dilution line, pointing out the absence of important water-rock interaction processes that can release these two elements in solution. In contrast, water sampled in the exploratory boreholes and along drain pipes showed concentration peaks that would be unjustifiable from mere water-rock interaction processes, and due to the small extension of the examined area. It is conceivable that sodium and potassium were derived from nonnatural factors. Possible sources could be the cement mortars and the percentages of sodium hydroxide and potassium hydroxide within.

Finally, the correlation diagram for Cl vs. Al (Figure 7) highlights the high Al concentrations. In the case of aluminum, the pH affects the mobility of the element. At extreme pH values (low or high), the mobility of the element increases, and, consequently, so do the concentrations found in the solution. The availability of the element can be linked to aluminum oxides (Al2 O3 ) characterizing the cement mixtures used.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

311

Vespasiano, Notaro, and Cianflone

Figure 8. Calcite dissolution in pure water. Modified from Langmuir (1971).

DISCUSSION In the study area, there is no evaporitic sequence that could release alkali into the water cycle. Under these conditions, a possible source of alkali may be the cement mortars, in particular, to the interstitial waters enriched in Na(OH) and K(OH).

Regarding water quality, values of a few major and trace elements were compared with the Italian legal limit values of D. Lgs 152/2006 and the World Health Organization (WHO) drinking water guidelines, which establish the lowest threshold of concentration for groundwater. Only samples PL1, PL2, and PL5 showed an Al content higher than the lowest threshold provided for by Italian law. In this case, pH could play a fundamental role on the mobility of Al. Concerning major elements, anomalous concentrations were detected for NH4 (law limit of 0.5 ppm), F (law limit of 1.5 ppm), and NO2 (law limit of 0.5 ppm), where, for samples PL1 and PL2, concentrations higher than 50 ppm were observed. These concentrations reflect intermediate stages of the nitrification and the denitrification processes by aerobic or anaerobic bacteria, respectively.

Based on geochemical and geological data, a possible cause for the pH anomalies and calcite precipitation could be due to the interaction between groundwater and the Trubi Formation, which consists of clay and limestone marl. As shown in Figure 8 (modified from Langmuir, 1971), which shows the theoretical compositional evolution linked to an interaction 312

between pure water and calcite, assuming open and closed systems with respect to CO2 , calcite dissolution alone cannot explain the measured pH, since that would be achieved before saturation.

This allows us to dismiss the claim that water-rock interaction with the Trubi Formation represents the only cause of these processes, and this is also confirmed by the high Na and K concentrations, which are not justifiable from mere water-rock interaction processes. Thus, excluding both the natural and anthropogenic factors such as stabilization lime or the presence of cineritic deposits, this suggests that the cause is interaction between groundwater and cement mortars used for the tunnel consolidation phases.

From a general point of view, cement consists of four materials formed during the calcining process: tricalcium silicate (C3 S), dicalcium silicate (C2 S), tricalcium aluminate (C3 A), and tetracalcium aluminoferrite (C4 AF), with C = CaO, S = SiO2 , A = Al2 O3 , and F = Fe2 O3 . These constituents react with water to produce 40–60 percent calcium silicate hydrate (commonly referred to as CSH, but with variable stoichiometry), 20–25 percent portlandite (CaOH2 or CH), 10– 20 percent hydrated aluminates, ferrites, and sulfates, 10–20 percent of pore fluids, and 0-5 percent of NaOH and KOH, which are generally dissolved in pore fluids (Gascoyne, 2002; Soler and Mäder, 2010). Cement powder reacts with water during the hydration. The main products are: calcium silicate hydrates (CSH) and portlandite (CaOH2 or CH). The chemistries of these components must be considered when determining the stability of cement and concrete.

Hardened cement is generally slow to react with water, unless it is porous, in which case, large amounts of water may be able to flow through it and dissolve the sparingly soluble components. The breakdown of cement and concrete in aqueous environments produces a high-pH plume, which is derived, initially, from pore fluids in the cement containing strong alkali (NaOH, KOH) and non-negligible Na and K concentrations (Glasser et al., 1985; Lundén and Andersson, 1989; Gascoyne, 2002; Soler and Mäder, 2010).

The pH of water leachate from hardened cement follows a number of well-defined stages, of which the initial two are very important (Lea, 1971; Glasser and Marr, 1983; Askarieh et al., 1997; Oscarson et al., 1997): (1) Initially, the pH of the small amounts of strong alkali (NaOH, KOH) present in the pore fluids dominates and can give values as high as 13.5 to 14. In adjacent groundwater, this often creates a highpH plume that spreads out from the grouted area in response to the flow conditions and mass/volume of grout present. In our system, this step could explain the high K and Na concentrations and high pH value.

(2) Once the alkalis have been leached out, the pH is

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

Vespasiano, Notaro, and Cianflone

Figure 8. Calcite dissolution in pure water. Modified from Langmuir (1971).

DISCUSSION In the study area, there is no evaporitic sequence that could release alkali into the water cycle. Under these conditions, a possible source of alkali may be the cement mortars, in particular, to the interstitial waters enriched in Na(OH) and K(OH).

Regarding water quality, values of a few major and trace elements were compared with the Italian legal limit values of D. Lgs 152/2006 and the World Health Organization (WHO) drinking water guidelines, which establish the lowest threshold of concentration for groundwater. Only samples PL1, PL2, and PL5 showed an Al content higher than the lowest threshold provided for by Italian law. In this case, pH could play a fundamental role on the mobility of Al. Concerning major elements, anomalous concentrations were detected for NH4 (law limit of 0.5 ppm), F (law limit of 1.5 ppm), and NO2 (law limit of 0.5 ppm), where, for samples PL1 and PL2, concentrations higher than 50 ppm were observed. These concentrations reflect intermediate stages of the nitrification and the denitrification processes by aerobic or anaerobic bacteria, respectively.

Based on geochemical and geological data, a possible cause for the pH anomalies and calcite precipitation could be due to the interaction between groundwater and the Trubi Formation, which consists of clay and limestone marl. As shown in Figure 8 (modified from Langmuir, 1971), which shows the theoretical compositional evolution linked to an interaction 312

between pure water and calcite, assuming open and closed systems with respect to CO2 , calcite dissolution alone cannot explain the measured pH, since that would be achieved before saturation.

This allows us to dismiss the claim that water-rock interaction with the Trubi Formation represents the only cause of these processes, and this is also confirmed by the high Na and K concentrations, which are not justifiable from mere water-rock interaction processes. Thus, excluding both the natural and anthropogenic factors such as stabilization lime or the presence of cineritic deposits, this suggests that the cause is interaction between groundwater and cement mortars used for the tunnel consolidation phases.

From a general point of view, cement consists of four materials formed during the calcining process: tricalcium silicate (C3 S), dicalcium silicate (C2 S), tricalcium aluminate (C3 A), and tetracalcium aluminoferrite (C4 AF), with C = CaO, S = SiO2 , A = Al2 O3 , and F = Fe2 O3 . These constituents react with water to produce 40–60 percent calcium silicate hydrate (commonly referred to as CSH, but with variable stoichiometry), 20–25 percent portlandite (CaOH2 or CH), 10– 20 percent hydrated aluminates, ferrites, and sulfates, 10–20 percent of pore fluids, and 0-5 percent of NaOH and KOH, which are generally dissolved in pore fluids (Gascoyne, 2002; Soler and Mäder, 2010). Cement powder reacts with water during the hydration. The main products are: calcium silicate hydrates (CSH) and portlandite (CaOH2 or CH). The chemistries of these components must be considered when determining the stability of cement and concrete.

Hardened cement is generally slow to react with water, unless it is porous, in which case, large amounts of water may be able to flow through it and dissolve the sparingly soluble components. The breakdown of cement and concrete in aqueous environments produces a high-pH plume, which is derived, initially, from pore fluids in the cement containing strong alkali (NaOH, KOH) and non-negligible Na and K concentrations (Glasser et al., 1985; Lundén and Andersson, 1989; Gascoyne, 2002; Soler and Mäder, 2010).

The pH of water leachate from hardened cement follows a number of well-defined stages, of which the initial two are very important (Lea, 1971; Glasser and Marr, 1983; Askarieh et al., 1997; Oscarson et al., 1997): (1) Initially, the pH of the small amounts of strong alkali (NaOH, KOH) present in the pore fluids dominates and can give values as high as 13.5 to 14. In adjacent groundwater, this often creates a highpH plume that spreads out from the grouted area in response to the flow conditions and mass/volume of grout present. In our system, this step could explain the high K and Na concentrations and high pH value.

(2) Once the alkalis have been leached out, the pH is

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

Water-Mortar Interaction in a Tunnel

controlled, at about 12.5, by CaOH2 . This pH is maintained for a long while after hardening because of the relatively high content of unreacted CaOH2 in the cement. When all unreacted CaOH2 has been removed, CSH dissolution begins, resulting in gradual pH decrease (Gascoyne, 2002).

In the process of dissolution of the CSH matrix, silica and alumina tend to be left as hydrated residual grains. At high pH, these become sparingly soluble and are able to migrate from the leaching interface. In the case of cement or grout injected into permeable fractures in bedrock, any dissolved silica and alumina will migrate until pH decreases due to dilution or water-rock interaction. These components will then precipitate from solution and may coat fracture surfaces and block the narrower aperture fractures and matrix pores. The timing of the development of these processes is difficult to predict.

Engkvist et al. (1996) analyzed major ions and pH in replenished solutions of fresh and saline groundwater in contact with crushed ordinary Portland cement for about 1.5 years. Initially, high concentrations of Na and K were seen (due to leaching of alkaline pore fluids), but these rapidly decreased, while Ca remained constant and at a pH of about 12.5. Successively, a consistent value was observed until about 1/3 of the time horizon (6 months). From this point, pH and concentrations of OH and Ca started to decrease, indicating the removal of portlandite. That study highlighted that the first steps advance very quickly, followed by CSH dissolution characterized by prolonged development.

An interesting aspect to be clarified is the aggressiveness against cement mortars (UNI EN 206-1; Unicemento, 2006) of theoretically slightly aggressive water, such as those sampled in P5 and PX, which represent the blank of the site. A possible, although unlikely, explanation is given by the boundary conditions that characterize the groundwater evolution. In fact, Italian law reports threshold values of water characterized by temperature between 5 and 25°C, where low flow rate approximates to a steady groundwater level.

In the studied area, a high and variable flow rate along drain pipes PLfi1 and PLfi2 was detected. This condition permits us to consider the absence of static conditions in the interaction area and implies the presence of constantly renovated waters, thus maintaining a relatively high degree of aggressiveness.

Another possible cause is the water/cement used in the preparation of the cement mortars. In fact, the amount of water added to the cement mix is usually in the range of water/cement ratio (w/c) = 0.35 to 0.5, and this is sufficient to allow the cement to flow.

However, the amount of water required to chemically hydrate the cement is less (w/c = 0.22 to 0.25), and

Water-Mortar Interaction in a Tunnel

so hardened cement contains an aqueous pore fluid, which contains most of the highly soluble alkalis, e.g., NaOH and KOH (Gascoyne, 2002; Soler and Mäder, 2010). This condition could lead to high pH and high concentrations of K and Na in solution (Gascoyne, 2002; Soler and Mäder, 2010).

Regardless of the factors that control and promote the aggressiveness of waters, interestingly the XRD analysis revealed the presence of Mg-corrensite in the precipitate (PL5). Savage (2013) reported that as the migration of interstitial fluid cement with a high quantity of alkali is certainly attenuated in clay formations, in fractured crystalline units, as in the case of the examined tunnel, the process is greatly accelerated. In these environments, the presence or loss of Mg in solution is particularly linked to the formation of secondary solid phases such as Mg-corrensite. The decrease of Mg concentration associated with the presence of phases such as Mg-corrensite in the examined samples confirms the hypothesis of an interaction among the crystalline basement, pore water, and cement mortars.

CONCLUSIONS In this work, we analyzed the results a geochemical analysis aimed to define the origin of the pH anomalies (pH > 11) in the water samples collected inside a tunnel located in southern Calabria (southern Italy).

In addition, we analyzed the precipitates found close to the main drainage pipes.

The hydrogeochemical study allowed us to identify a main NaOH composition for the many samples collected close to the tunnel. The correlation diagrams highlighted high concentrations of Na, K, and Al, which are unlikely to be due to simple water-rock interaction processes, and a Mg concentration decrease associated with Mg-corrensite precipitation detected in the collected precipitates.

Our analysis permitted us to exclude the possibility that the interaction between water and the outcropping lithologies (crystalline bedrock, Trubi Formation, Pezzo Conglomerate, and Messina Gravels) represents the only cause of these ongoing processes. This consideration is supported by the high Na and K concentrations, which cannot be justified by the interaction between water and calcareous marl.

Excluding natural origins and some anthropogenic factors (such as lime stabilizations) and the absence of ash accumulation, a possible explanation is the interaction between the groundwater and the mortars used for the consolidation during the excavation phase. Indeed, the presence of widespread porosity and fractures favors the dissolution of scarcely soluble phases, such as the cementitious ones, if abundant amounts of water are available. The grout and concrete

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

313

controlled, at about 12.5, by CaOH2 . This pH is maintained for a long while after hardening because of the relatively high content of unreacted CaOH2 in the cement. When all unreacted CaOH2 has been removed, CSH dissolution begins, resulting in gradual pH decrease (Gascoyne, 2002).

In the process of dissolution of the CSH matrix, silica and alumina tend to be left as hydrated residual grains. At high pH, these become sparingly soluble and are able to migrate from the leaching interface. In the case of cement or grout injected into permeable fractures in bedrock, any dissolved silica and alumina will migrate until pH decreases due to dilution or water-rock interaction. These components will then precipitate from solution and may coat fracture surfaces and block the narrower aperture fractures and matrix pores. The timing of the development of these processes is difficult to predict.

Engkvist et al. (1996) analyzed major ions and pH in replenished solutions of fresh and saline groundwater in contact with crushed ordinary Portland cement for about 1.5 years. Initially, high concentrations of Na and K were seen (due to leaching of alkaline pore fluids), but these rapidly decreased, while Ca remained constant and at a pH of about 12.5. Successively, a consistent value was observed until about 1/3 of the time horizon (6 months). From this point, pH and concentrations of OH and Ca started to decrease, indicating the removal of portlandite. That study highlighted that the first steps advance very quickly, followed by CSH dissolution characterized by prolonged development.

An interesting aspect to be clarified is the aggressiveness against cement mortars (UNI EN 206-1; Unicemento, 2006) of theoretically slightly aggressive water, such as those sampled in P5 and PX, which represent the blank of the site. A possible, although unlikely, explanation is given by the boundary conditions that characterize the groundwater evolution. In fact, Italian law reports threshold values of water characterized by temperature between 5 and 25°C, where low flow rate approximates to a steady groundwater level.

In the studied area, a high and variable flow rate along drain pipes PLfi1 and PLfi2 was detected. This condition permits us to consider the absence of static conditions in the interaction area and implies the presence of constantly renovated waters, thus maintaining a relatively high degree of aggressiveness.

Another possible cause is the water/cement used in the preparation of the cement mortars. In fact, the amount of water added to the cement mix is usually in the range of water/cement ratio (w/c) = 0.35 to 0.5, and this is sufficient to allow the cement to flow.

However, the amount of water required to chemically hydrate the cement is less (w/c = 0.22 to 0.25), and

so hardened cement contains an aqueous pore fluid, which contains most of the highly soluble alkalis, e.g., NaOH and KOH (Gascoyne, 2002; Soler and Mäder, 2010). This condition could lead to high pH and high concentrations of K and Na in solution (Gascoyne, 2002; Soler and Mäder, 2010).

Regardless of the factors that control and promote the aggressiveness of waters, interestingly the XRD analysis revealed the presence of Mg-corrensite in the precipitate (PL5). Savage (2013) reported that as the migration of interstitial fluid cement with a high quantity of alkali is certainly attenuated in clay formations, in fractured crystalline units, as in the case of the examined tunnel, the process is greatly accelerated. In these environments, the presence or loss of Mg in solution is particularly linked to the formation of secondary solid phases such as Mg-corrensite. The decrease of Mg concentration associated with the presence of phases such as Mg-corrensite in the examined samples confirms the hypothesis of an interaction among the crystalline basement, pore water, and cement mortars.

CONCLUSIONS In this work, we analyzed the results a geochemical analysis aimed to define the origin of the pH anomalies (pH > 11) in the water samples collected inside a tunnel located in southern Calabria (southern Italy).

In addition, we analyzed the precipitates found close to the main drainage pipes.

The hydrogeochemical study allowed us to identify a main NaOH composition for the many samples collected close to the tunnel. The correlation diagrams highlighted high concentrations of Na, K, and Al, which are unlikely to be due to simple water-rock interaction processes, and a Mg concentration decrease associated with Mg-corrensite precipitation detected in the collected precipitates.

Our analysis permitted us to exclude the possibility that the interaction between water and the outcropping lithologies (crystalline bedrock, Trubi Formation, Pezzo Conglomerate, and Messina Gravels) represents the only cause of these ongoing processes. This consideration is supported by the high Na and K concentrations, which cannot be justified by the interaction between water and calcareous marl.

Excluding natural origins and some anthropogenic factors (such as lime stabilizations) and the absence of ash accumulation, a possible explanation is the interaction between the groundwater and the mortars used for the consolidation during the excavation phase. Indeed, the presence of widespread porosity and fractures favors the dissolution of scarcely soluble phases, such as the cementitious ones, if abundant amounts of water are available. The grout and concrete

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

313

Vespasiano, Notaro, and Cianflone

degradation in aqueous environments produces a great increase of pH, initially deriving from the interstitial fluids containing strong alkali (NaOH and KOH) components and non-negligible K and Na concentrations (Gascoyne, 2002; Soler and Mäder, 2010), such as those observed in the collected samples.

ACKNOWLEDGMENTS The authors wish to thank Ing. Enrico Cecere of the T.O. of G.C. group (Salini, Impregilo/Condotte). The authors are particularly grateful to Dr. Luigi Marini, Dr. Sandro Martinetti, and Dr. Dario Burza for their support and advice. We express our gratitude to Dr. David Mc Enroe and Dr. Marco Cianflone for the English review.

REFERENCES Apollaro, C.; Vespasiano, G.; De Rosa, R.; and Marini, L., 2015, Use of mean residence time and flowrate of thermal waters to evaluate the volume of reservoir water contributing to the natural discharge and the related geothermal reservoir volume. Application to northern Thailand hot springs: Geothermics, Vol. 58, pp. 62–74.

Apollaro, C.; Vespasiano, G.; Muto, F.; De Rosa, R.; Barca, D.; and Marini, L., 2016, Use of mean residence time of water, flowrate, and equilibrium temperature indicated by water geothermometers to rank geothermal resources. Application to the thermal water circuits of northern Calabria: Journal Volcanology Geothermal Research, Vol. 328, pp. 147–158.

Askarieh, M. M.; Harris, A. W.; Haworth, A.; Heath, T. G.; and Tweed, C. J., 1997, Chemical Environment of a Cementitious Repository: Nirex Science Report S/98/008.

Brook, G. A.; Folkoff, M. E.; and Box, E. O., 1983, A world model of soil carbon dioxide: Earth Surface Processes Landforms, Vol. 8, pp. 79–88.

Cianflone, G.; Tolomei, C.; Brunori, C. A.; and Dominici, R., 2015a, InSAR time series analysis of natural and anthropogenic coastal plain subsidence: The case of Sibari (southern Italy): Remote Sensing, Vol. 7, pp. 16004–16023.

Cianflone, G.; Tolomei, C.; Brunori, C. A.; and Dominici, R., 2015b, Preliminary study of the surface ground displacements in the Crati Valley (Calabria) by means of InSAR data: Rendiconti Online della Società Geologica Italiana, Vol. 33, pp. 20–23.

Critelli, T.; Vespasiano, G.; Apollaro, C.; Muto, F.; Marini, L; and De Rosa, R., 2015, Hydrogeochemical study of an ophiolitic aquifer: A case study of Lago (southern Italy, Calabria): Environmental Earth Sciences, Vol. 74, pp. 533–543.

D.L. 152/2006, Legislative Decree 3 April 2006, n. 152 "Environmental regulations". Official Gazette n. 88 of 14-4- 2006, Ordinary Supplement n. 96.

Doglioni, C., 1995, Geological remarks on the relationships between extension and convergent geodynamic settings: Tectonophysics, Vol. 252, pp. 253–267.

Dumas, B.; Guérémy, P.; and Raffy, J., 1999, Essai de corrélations entre des étagements de lignes de rivage soulevées en Calabre méridionale (Italie) et des courbes isotopiques à haute résolution entre 130 et 40 ka: Quaternaire, Vol. 2/3, pp. 107–119.

Dumas, B.; Guérémy, P.; and Raffy, J., 2005, Evidence for sealevel oscillations by the “characteristic thickness” of marine deposits from raised terraces of southern Calabria (Italy), Quaternary Science Reviews, Vol. 24, pp. 2120–2136.

314

Engkvist, I.; Albinsson, Y.; and Engkvist, W. J., 1996, The LongTerm Stability of Cement—Leaching Tests: Swedish Nuclear and Waste Management Company Technical Report 96-09, 40 p.

Faccenna, C.; Mattei, M.; Funiciello, R.; and Jolivet, L., 1997, Styles of back-arc extension in the central Mediterranean: Terra Nova, Vol. 9, pp. 126–130.

Gascoyne, M., 2002, Influence of Grout and Cement on Groundwater Composition: Posiva Working Report 2002-07, 44 p.

Glasser, F. P.; Angus, M. J.; McCulloch, C. E.; Macphee, D.; and Rahman, A. A., 1985, The chemical environment in cements. In Materials Research Society Symposium Proceedings, Vol. 44: Materials Research Society, Scientific Basis for Nuclear Waste Management, VIII. Boston, MA, USA. pp. 849– 858.

Glasser, F. P. and Marr, J., 1983, Effect of silica, PFA and slag additives on the composition of cement pore fluids. In Alkalis in Concrete—6th International Symposium on Alkali-Aggregate Reactions: Ed. G.M. Idorn and S. Rostam: pub. DBF, Copenhagen, Denmark, pp. 239–242.

Langmuir, D., 1971, Aqueous Environmental Geochemistry: Prentice Hall, Upper Saddle River, NJ, 600 p.

Lea, F. M., 1971, The Chemistry of Cement and Concrete: Chemical Publishing Company, Los Angeles, CA, 727 p.

Lenci, F.; Carminati, E.; Doglioni, C.; and Scrocca, D., 2004.

Basal décollement and subduction depth vs. topography in the Apennines–Calabrian arc: Bollettino della Società Geologica Italiana, Vol. 123, pp. 497–502.

Lundén, I. and Andersson, K., 1989, Modelling of the mixing of cement pore water and groundwater using the PHREEQE code. In Lutz, W. and Ewing, R. C. (Editors), Materials Research Society Symposium Proceedings, Vol. 127: Materials Research Society, Pittsburgh, PA, pp. 949–956.

Malinverno, A. and Ryan, W. B. F., 1986, Extension in Tyrrhenian Sea & shortening in the Apennines as result of arc migration driven by sinking of the lithosphere: Tectonics, Vol. 5, pp.

227–254.

Marchese, B., 1972, Tecnologia dei Materiali e Chimica Applicate: Liguori Editore, Napoli, Italy, 532 p.

Miyauchi, T.; Dai Pra, G.; and Sylos Labini, S., 1994, Geochronology of Pleistocene marine terraces and regional tectonics in the Tyrrhenian coast of south Calabria, Italy: Il Quaternario, Vol. 7, No. 1, pp. 17–34.

Monaco, C.; Tortorici, L.; Nicolich, R.; Cernobi, L.; and Costa, M., 1996, From collisional to rifted basins: An example from the southern Calabrian Arc (Italy): Tectonophysics, Vol. 266, pp. 233–249.

Nordstrom, D. K., 1977, Thermochemical redox equilibria of ZoBell’s solution. Geochimica et Cosmochimica Acta, Vol. 41, pp.

1835–1841.

Nordstrom, D. K.; Plummer, L. N.; Wigley, T. M. L.; Wolery, T. J.; Ball, J. W.; Jenne, E. A.; Bassett, R. L.; Crerar, D. A.; Florence, T. M.; Fritz, B.; Hoffman, M.; Holdren, G. R., Jr.; Lafon, G. M.; Mattigod, S. V.; McDuff, R. E.; Morel, F.; Reddy, M. M.; Sposito, G.; and Thrailkill, J., 1979, A comparison of computerized chemical models for equilibrium calculations in aqueous systems. In Jenne, E. A. (Editor), Chemical Modeling in Aqueous Systems. Speciation, Sorption, Solubility, and Kinetics. American Chemical Society Symposium Series 93: American Chemical Society, Washington, D.C., pp. 857–892.

Nordstrom, D. K. and Wilde, F. D., 2005, Reduction oxidation potential (electrode method). In Wilde, F. D. (Editor), Field Measurements. National Field Manual for the Collection of Water-Quality Data: U.S. Geological Survey TWRI

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

Vespasiano, Notaro, and Cianflone

degradation in aqueous environments produces a great increase of pH, initially deriving from the interstitial fluids containing strong alkali (NaOH and KOH) components and non-negligible K and Na concentrations (Gascoyne, 2002; Soler and Mäder, 2010), such as those observed in the collected samples.

ACKNOWLEDGMENTS The authors wish to thank Ing. Enrico Cecere of the T.O. of G.C. group (Salini, Impregilo/Condotte). The authors are particularly grateful to Dr. Luigi Marini, Dr. Sandro Martinetti, and Dr. Dario Burza for their support and advice. We express our gratitude to Dr. David Mc Enroe and Dr. Marco Cianflone for the English review.

REFERENCES Apollaro, C.; Vespasiano, G.; De Rosa, R.; and Marini, L., 2015, Use of mean residence time and flowrate of thermal waters to evaluate the volume of reservoir water contributing to the natural discharge and the related geothermal reservoir volume. Application to northern Thailand hot springs: Geothermics, Vol. 58, pp. 62–74.

Apollaro, C.; Vespasiano, G.; Muto, F.; De Rosa, R.; Barca, D.; and Marini, L., 2016, Use of mean residence time of water, flowrate, and equilibrium temperature indicated by water geothermometers to rank geothermal resources. Application to the thermal water circuits of northern Calabria: Journal Volcanology Geothermal Research, Vol. 328, pp. 147–158.

Askarieh, M. M.; Harris, A. W.; Haworth, A.; Heath, T. G.; and Tweed, C. J., 1997, Chemical Environment of a Cementitious Repository: Nirex Science Report S/98/008.

Brook, G. A.; Folkoff, M. E.; and Box, E. O., 1983, A world model of soil carbon dioxide: Earth Surface Processes Landforms, Vol. 8, pp. 79–88.

Cianflone, G.; Tolomei, C.; Brunori, C. A.; and Dominici, R., 2015a, InSAR time series analysis of natural and anthropogenic coastal plain subsidence: The case of Sibari (southern Italy): Remote Sensing, Vol. 7, pp. 16004–16023.

Cianflone, G.; Tolomei, C.; Brunori, C. A.; and Dominici, R., 2015b, Preliminary study of the surface ground displacements in the Crati Valley (Calabria) by means of InSAR data: Rendiconti Online della Società Geologica Italiana, Vol. 33, pp. 20–23.

Critelli, T.; Vespasiano, G.; Apollaro, C.; Muto, F.; Marini, L; and De Rosa, R., 2015, Hydrogeochemical study of an ophiolitic aquifer: A case study of Lago (southern Italy, Calabria): Environmental Earth Sciences, Vol. 74, pp. 533–543.

D.L. 152/2006, Legislative Decree 3 April 2006, n. 152 "Environmental regulations". Official Gazette n. 88 of 14-4- 2006, Ordinary Supplement n. 96.

Doglioni, C., 1995, Geological remarks on the relationships between extension and convergent geodynamic settings: Tectonophysics, Vol. 252, pp. 253–267.

Dumas, B.; Guérémy, P.; and Raffy, J., 1999, Essai de corrélations entre des étagements de lignes de rivage soulevées en Calabre méridionale (Italie) et des courbes isotopiques à haute résolution entre 130 et 40 ka: Quaternaire, Vol. 2/3, pp. 107–119.

Dumas, B.; Guérémy, P.; and Raffy, J., 2005, Evidence for sealevel oscillations by the “characteristic thickness” of marine deposits from raised terraces of southern Calabria (Italy), Quaternary Science Reviews, Vol. 24, pp. 2120–2136.

314

Engkvist, I.; Albinsson, Y.; and Engkvist, W. J., 1996, The LongTerm Stability of Cement—Leaching Tests: Swedish Nuclear and Waste Management Company Technical Report 96-09, 40 p.

Faccenna, C.; Mattei, M.; Funiciello, R.; and Jolivet, L., 1997, Styles of back-arc extension in the central Mediterranean: Terra Nova, Vol. 9, pp. 126–130.

Gascoyne, M., 2002, Influence of Grout and Cement on Groundwater Composition: Posiva Working Report 2002-07, 44 p.

Glasser, F. P.; Angus, M. J.; McCulloch, C. E.; Macphee, D.; and Rahman, A. A., 1985, The chemical environment in cements. In Materials Research Society Symposium Proceedings, Vol. 44: Materials Research Society, Scientific Basis for Nuclear Waste Management, VIII. Boston, MA, USA. pp. 849– 858.

Glasser, F. P. and Marr, J., 1983, Effect of silica, PFA and slag additives on the composition of cement pore fluids. In Alkalis in Concrete—6th International Symposium on Alkali-Aggregate Reactions: Ed. G.M. Idorn and S. Rostam: pub. DBF, Copenhagen, Denmark, pp. 239–242.

Langmuir, D., 1971, Aqueous Environmental Geochemistry: Prentice Hall, Upper Saddle River, NJ, 600 p.

Lea, F. M., 1971, The Chemistry of Cement and Concrete: Chemical Publishing Company, Los Angeles, CA, 727 p.

Lenci, F.; Carminati, E.; Doglioni, C.; and Scrocca, D., 2004.

Basal décollement and subduction depth vs. topography in the Apennines–Calabrian arc: Bollettino della Società Geologica Italiana, Vol. 123, pp. 497–502.

Lundén, I. and Andersson, K., 1989, Modelling of the mixing of cement pore water and groundwater using the PHREEQE code. In Lutz, W. and Ewing, R. C. (Editors), Materials Research Society Symposium Proceedings, Vol. 127: Materials Research Society, Pittsburgh, PA, pp. 949–956.

Malinverno, A. and Ryan, W. B. F., 1986, Extension in Tyrrhenian Sea & shortening in the Apennines as result of arc migration driven by sinking of the lithosphere: Tectonics, Vol. 5, pp.

227–254.

Marchese, B., 1972, Tecnologia dei Materiali e Chimica Applicate: Liguori Editore, Napoli, Italy, 532 p.

Miyauchi, T.; Dai Pra, G.; and Sylos Labini, S., 1994, Geochronology of Pleistocene marine terraces and regional tectonics in the Tyrrhenian coast of south Calabria, Italy: Il Quaternario, Vol. 7, No. 1, pp. 17–34.

Monaco, C.; Tortorici, L.; Nicolich, R.; Cernobi, L.; and Costa, M., 1996, From collisional to rifted basins: An example from the southern Calabrian Arc (Italy): Tectonophysics, Vol. 266, pp. 233–249.

Nordstrom, D. K., 1977, Thermochemical redox equilibria of ZoBell’s solution. Geochimica et Cosmochimica Acta, Vol. 41, pp.

1835–1841.

Nordstrom, D. K.; Plummer, L. N.; Wigley, T. M. L.; Wolery, T. J.; Ball, J. W.; Jenne, E. A.; Bassett, R. L.; Crerar, D. A.; Florence, T. M.; Fritz, B.; Hoffman, M.; Holdren, G. R., Jr.; Lafon, G. M.; Mattigod, S. V.; McDuff, R. E.; Morel, F.; Reddy, M. M.; Sposito, G.; and Thrailkill, J., 1979, A comparison of computerized chemical models for equilibrium calculations in aqueous systems. In Jenne, E. A. (Editor), Chemical Modeling in Aqueous Systems. Speciation, Sorption, Solubility, and Kinetics. American Chemical Society Symposium Series 93: American Chemical Society, Washington, D.C., pp. 857–892.

Nordstrom, D. K. and Wilde, F. D., 2005, Reduction oxidation potential (electrode method). In Wilde, F. D. (Editor), Field Measurements. National Field Manual for the Collection of Water-Quality Data: U.S. Geological Survey TWRI

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

Water-Mortar Interaction in a Tunnel Book 9 (TWRI Book 9), electronic document available at http://water.usgs.gov/owq/FieldManual/Chapter6/6.5_ contents.html.

Ortolano, G.; Cirrincione, R.; Pezzino, A.; Tripodi, V.; and Zappalà, L., 2015, Petro-structural geology of the eastern Aspromonte Massif crystalline basement (southern ItalyCalabria): An example of interoperable geo-data management from thin section– to field scale: Journal of Maps, Vol. 11, No. 1, pp. 181–200.

Oscarson, D. W.; Dixon, D. A.; and Onofrei, M., 1997, Aspects of Day/Concrete Interactions: Atomic Energy of Canada Limited Report AECL-11715/COG-96-562-1.

Savage, D., 2013, Constraints on cement-clay interaction: Procedia Earth and Planetary Science, Vol. 7, pp. 770–773.

Soler, J. M. and Mäder, U. K., 2010, Cement-rock interaction: Infiltration of a high-pH solution into a fractured granite core: Geologica Acta, Vol. 8, pp. 221–233.

Unicemento, 2006, Concrete—Part 1: Specification, Performance, Production and Conformity: Technical Standard UNI EN 2061:2006. UNI 11104, 2016, Concrete - Specification, performance, production and conformity - Complementary specifications for the application of EN206.

Vespasiano, G. and Apollaro, C., 2016, Preliminary geochemical characterization of a carbonate aquifer: The case of Pollino Massif (Calabria, south Italy): Rendiconti Online della Società Geologica Italiana, Vol. 38, pp. 109–112.

Vespasiano, G.; Apollaro, C.; De Rosa, R.; Muto, F.; Larosa, S.; Fiebig, J.; Mulch, A.; and Marini, L., 2015a, The Small Spring Method (SSM) for the definition of stable isotope– elevation relationships in northern Calabria (southern Italy): Applied Geochemistry, Vol. 63, pp. 333–346.

Vespasiano, G.; Apollaro, C.; Muto, F.; De Rosa, R.; and Critelli, T., 2015b, Preliminary geochemical and geological characterization of the thermal site of Spezzano Albanese (Calabria, south Italy): Rendiconti Online della Società Geologica Italiana., Vol. 33, pp. 108–110.

Vespasiano, G.; Apollaro, C.; Muto, F.; De Rosa, R.; Dotsika, E.; and Marini, L., 2016a, Preliminary geochemical charac-

Water-Mortar Interaction in a Tunnel

terization of the warm waters of the Grotta delle Ninfe near Cerchiara di Calabria (south Italy): Rendiconti Online della Società Geologica Italiana, Vol. 39, pp. 130–133.

Vespasiano, G.; Apollaro, C.; Marini, L.; Dominici, R.; Cianflone, G.; Romanazzi, A.; Polemio, M.; and De Rosa, R., 2016b, Hydrogeological and isotopic study of the multiaquifer system of the Sibari Plain (Calabria, southern Italy): Rendiconti Online della Società Geologica Italiana, Vol. 39, pp. 134–137.

Vespasiano, G.; Cianflone, G.; Cannata, C. B.; Apollaro, C.; Dominici, R.; and De Rosa, R., 2016c, Analysis of groundwater pollution in the Sant’Eufemia Plain (Calabria – south Italy): Italian Journal of Engineering Geology and Environment.

Vol. 2, pp 1–15.

Vespasiano, G.; Marini, L.; Apollaro, C.; and De Rosa, R., 2016d, Preliminary geochemical characterization of the thermal waters of Caronte SPA springs (Calabria, south Italy): Rendiconti Online della Società Geologica Italiana, Vol. 39, pp. 138–141.

Westaway, R., 1993, Quaternary uplift of southern Italy: Journal of Geophysical Research, Vol. 98, No. B12, pp.

21741–21772. WHO, 2008. Guidelines for Drinking-Water Quality: Recommendations. Incorporating 1st and 2nd Addenda, vol. 1, third ed. World Health Organization, Geneva. <http://www.who.int>. WHO, 2011. Guidelines for Drinking-Water Quality, fourth ed. <http://www.who.int>.

Wolery, T. J., 2013, EQ3/6—Software for Geochemical Modeling, Version 8.0a: Lawrence Livermore National Laboratory Report LLNL-CODE-2013-683958.

Wolery, T. J. and Jove-Colon, C. F., 2007, Qualification of Thermodynamic Data for Geochemical Modeling of Mineral-Water Interactions in Dilute Systems: Sandia National Laboratories Report ANL-WIS-GS-000003, REV 01.

Wolery, T. W. and Jarek, R. L., 2003. Software User’s Manual. EQ3/6, Version 8.0: Sandia National Laboratories Report. Albuquerque, New Mexico, p. 376.

Zaporozec, A., 1972, Graphical interpretation of water-quality data: Ground Water, Vol. 10, pp. 32–43.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

315

Book 9 (TWRI Book 9), electronic document available at http://water.usgs.gov/owq/FieldManual/Chapter6/6.5_ contents.html.

Ortolano, G.; Cirrincione, R.; Pezzino, A.; Tripodi, V.; and Zappalà, L., 2015, Petro-structural geology of the eastern Aspromonte Massif crystalline basement (southern ItalyCalabria): An example of interoperable geo-data management from thin section– to field scale: Journal of Maps, Vol. 11, No. 1, pp. 181–200.

Oscarson, D. W.; Dixon, D. A.; and Onofrei, M., 1997, Aspects of Day/Concrete Interactions: Atomic Energy of Canada Limited Report AECL-11715/COG-96-562-1.

Savage, D., 2013, Constraints on cement-clay interaction: Procedia Earth and Planetary Science, Vol. 7, pp. 770–773.

Soler, J. M. and Mäder, U. K., 2010, Cement-rock interaction: Infiltration of a high-pH solution into a fractured granite core: Geologica Acta, Vol. 8, pp. 221–233.

Unicemento, 2006, Concrete—Part 1: Specification, Performance, Production and Conformity: Technical Standard UNI EN 2061:2006. UNI 11104, 2016, Concrete - Specification, performance, production and conformity - Complementary specifications for the application of EN206.

Vespasiano, G. and Apollaro, C., 2016, Preliminary geochemical characterization of a carbonate aquifer: The case of Pollino Massif (Calabria, south Italy): Rendiconti Online della Società Geologica Italiana, Vol. 38, pp. 109–112.

Vespasiano, G.; Apollaro, C.; De Rosa, R.; Muto, F.; Larosa, S.; Fiebig, J.; Mulch, A.; and Marini, L., 2015a, The Small Spring Method (SSM) for the definition of stable isotope– elevation relationships in northern Calabria (southern Italy): Applied Geochemistry, Vol. 63, pp. 333–346.

Vespasiano, G.; Apollaro, C.; Muto, F.; De Rosa, R.; and Critelli, T., 2015b, Preliminary geochemical and geological characterization of the thermal site of Spezzano Albanese (Calabria, south Italy): Rendiconti Online della Società Geologica Italiana., Vol. 33, pp. 108–110.

Vespasiano, G.; Apollaro, C.; Muto, F.; De Rosa, R.; Dotsika, E.; and Marini, L., 2016a, Preliminary geochemical charac-

terization of the warm waters of the Grotta delle Ninfe near Cerchiara di Calabria (south Italy): Rendiconti Online della Società Geologica Italiana, Vol. 39, pp. 130–133.

Vespasiano, G.; Apollaro, C.; Marini, L.; Dominici, R.; Cianflone, G.; Romanazzi, A.; Polemio, M.; and De Rosa, R., 2016b, Hydrogeological and isotopic study of the multiaquifer system of the Sibari Plain (Calabria, southern Italy): Rendiconti Online della Società Geologica Italiana, Vol. 39, pp. 134–137.

Vespasiano, G.; Cianflone, G.; Cannata, C. B.; Apollaro, C.; Dominici, R.; and De Rosa, R., 2016c, Analysis of groundwater pollution in the Sant’Eufemia Plain (Calabria – south Italy): Italian Journal of Engineering Geology and Environment.

Vol. 2, pp 1–15.

Vespasiano, G.; Marini, L.; Apollaro, C.; and De Rosa, R., 2016d, Preliminary geochemical characterization of the thermal waters of Caronte SPA springs (Calabria, south Italy): Rendiconti Online della Società Geologica Italiana, Vol. 39, pp. 138–141.

Westaway, R., 1993, Quaternary uplift of southern Italy: Journal of Geophysical Research, Vol. 98, No. B12, pp.

21741–21772. WHO, 2008. Guidelines for Drinking-Water Quality: Recommendations. Incorporating 1st and 2nd Addenda, vol. 1, third ed. World Health Organization, Geneva. <http://www.who.int>. WHO, 2011. Guidelines for Drinking-Water Quality, fourth ed. <http://www.who.int>.

Wolery, T. J., 2013, EQ3/6—Software for Geochemical Modeling, Version 8.0a: Lawrence Livermore National Laboratory Report LLNL-CODE-2013-683958.

Wolery, T. J. and Jove-Colon, C. F., 2007, Qualification of Thermodynamic Data for Geochemical Modeling of Mineral-Water Interactions in Dilute Systems: Sandia National Laboratories Report ANL-WIS-GS-000003, REV 01.

Wolery, T. W. and Jarek, R. L., 2003. Software User’s Manual. EQ3/6, Version 8.0: Sandia National Laboratories Report. Albuquerque, New Mexico, p. 376.

Zaporozec, A., 1972, Graphical interpretation of water-quality data: Ground Water, Vol. 10, pp. 32–43.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 305–315

315

An Analysis of the Layered Failure of Coal: New Insights into the Flow Process of Outburst Coal

An Analysis of the Layered Failure of Coal: New Insights into the Flow Process of Outburst Coal

QINGYI TU YUANPING CHENG1 QINGQUAN LIU LIANG WANG WEI ZHAO WEI LI JUN DONG

QINGYI TU YUANPING CHENG1 QINGQUAN LIU LIANG WANG WEI ZHAO WEI LI JUN DONG

Key Laboratory of Coal Methane and Fire Control, Ministry of Education, China University of Mining and Technology, Xuzhou 221116, China; and Faculty of Safety Engineering, China University of Mining and Technology, Xuzhou 221116, China

Key Laboratory of Coal Methane and Fire Control, Ministry of Education, China University of Mining and Technology, Xuzhou 221116, China; and Faculty of Safety Engineering, China University of Mining and Technology, Xuzhou 221116, China

PINKUN GUO

PINKUN GUO

State Key Laboratory of Coal Mine Disaster Dynamics and Control, Chongqing University, Chongqiong 400044, China

State Key Laboratory of Coal Mine Disaster Dynamics and Control, Chongqing University, Chongqiong 400044, China

Key Terms: Coal and Gas Outburst, Coal Flow, Layered Failure, Zonal Distribution, Flow Velocity ABSTRACT Coal and gas outbursts are among the most destructive dynamic disasters. It is very important to study the failure process and mechanism of coal and gas outbursts in coal mines. Based on outburst cases and experiments, the outburst process and failure types of coal are discussed, and the flow process of outburst coal is studied in depth. This article concludes that the whole process of outburst generally involves four stages (preparation, trigger, development, and termination), which include both time and spatial dimensions. There are two types of coal failure: shear failure and tensile failure. Under the effect of the gas pressure gradient, the coal exhibits tensile failure, and the coal ultimately fails in the form of multiple similar layers. After the layer is separated, the pressure difference between the two sides of the layer is found to be stable for a period of time, and a pressure difference coefficient, i , is introduced to characterize the change of this pressure difference before and after the separation of the layer. Furthermore, the flow process of outburst coal is divided into a quasi-static phase, a constant acceleration phase, a freefall phase, and a deceleration phase.

Accordingly, a simplified model of the coal flow is established, and the model is calculated using the data from outburst experiments. Based on the layered failure of outburst coal, the flow velocity of outburst coal is obtained. The distribution features of outburst coal in both

1

Corresponding author email: ypccumt2015@outlook.com.

the experiments and the outburst cases are explained as well.

Key Terms: Coal and Gas Outburst, Coal Flow, Layered Failure, Zonal Distribution, Flow Velocity ABSTRACT

INTRODUCTION With the depletion of shallow coal resources, the mining depth of coal mines has increased year by year. In recent years, the depth of some coal mines in Mid-Eastern China has reached 800–1,200 m (Kang et al., 2010). Under high stress, high temperatures, and high coal gas, coal and gas outbursts (hereinafter referred to as “outbursts”) are becoming more and more serious (Wang et al., 2012), and the probability of large/extra-large outbursts is increasing significantly (Cheng, 2010). An outburst is a complex dynamic process in which a large amount of coal and coal gas enter a mining space in a short period of time (Song and Cheng, 2012; Fan et al., 2016), causing blockage of the mining space and inducing secondary disasters such as gas explosions (Kang, 2011; Chen et al., 2013).

Over the years, the mechanism of these outbursts and their evolution process have been hotly contested and difficult issues for many scholars (Choi and Wold, 2004; Guan et al., 2009; Wei et al., 2010; and Jin et al., 2011). Bodziony and Lama (1998) studied literature in different languages, established a multi-factor model, and argued that an outburst is a comprehensive effect of several factors (e.g., coal gas, stress, tectonic environment, mining activities, and the strength of coal). Jiang and Yu (1994) proposed a “spherical shell destabilization” hypothesis; the dynamic process of an outburst is explained by the viewpoint of “spherical shell destabilization.” Based on the statistical characteristics, the process of outburst is divided into four stages by Jin

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

317

Coal and gas outbursts are among the most destructive dynamic disasters. It is very important to study the failure process and mechanism of coal and gas outbursts in coal mines. Based on outburst cases and experiments, the outburst process and failure types of coal are discussed, and the flow process of outburst coal is studied in depth. This article concludes that the whole process of outburst generally involves four stages (preparation, trigger, development, and termination), which include both time and spatial dimensions. There are two types of coal failure: shear failure and tensile failure. Under the effect of the gas pressure gradient, the coal exhibits tensile failure, and the coal ultimately fails in the form of multiple similar layers. After the layer is separated, the pressure difference between the two sides of the layer is found to be stable for a period of time, and a pressure difference coefficient, i , is introduced to characterize the change of this pressure difference before and after the separation of the layer. Furthermore, the flow process of outburst coal is divided into a quasi-static phase, a constant acceleration phase, a freefall phase, and a deceleration phase.

Accordingly, a simplified model of the coal flow is established, and the model is calculated using the data from outburst experiments. Based on the layered failure of outburst coal, the flow velocity of outburst coal is obtained. The distribution features of outburst coal in both

1

Corresponding author email: ypccumt2015@outlook.com.

the experiments and the outburst cases are explained as well.

INTRODUCTION With the depletion of shallow coal resources, the mining depth of coal mines has increased year by year. In recent years, the depth of some coal mines in Mid-Eastern China has reached 800–1,200 m (Kang et al., 2010). Under high stress, high temperatures, and high coal gas, coal and gas outbursts (hereinafter referred to as “outbursts”) are becoming more and more serious (Wang et al., 2012), and the probability of large/extra-large outbursts is increasing significantly (Cheng, 2010). An outburst is a complex dynamic process in which a large amount of coal and coal gas enter a mining space in a short period of time (Song and Cheng, 2012; Fan et al., 2016), causing blockage of the mining space and inducing secondary disasters such as gas explosions (Kang, 2011; Chen et al., 2013).

Over the years, the mechanism of these outbursts and their evolution process have been hotly contested and difficult issues for many scholars (Choi and Wold, 2004; Guan et al., 2009; Wei et al., 2010; and Jin et al., 2011). Bodziony and Lama (1998) studied literature in different languages, established a multi-factor model, and argued that an outburst is a comprehensive effect of several factors (e.g., coal gas, stress, tectonic environment, mining activities, and the strength of coal). Jiang and Yu (1994) proposed a “spherical shell destabilization” hypothesis; the dynamic process of an outburst is explained by the viewpoint of “spherical shell destabilization.” Based on the statistical characteristics, the process of outburst is divided into four stages by Jin

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

317

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

et al. (2011)—a preparation stage, a trigger stage, a development stage, and a termination stage—and these authors noted that the flow process of outburst coal is attributed to the development stage.

With respect to the flow process of outburst coal, limited research exists. Considerable controversies still exist in regard to the flow state, the flow velocity, and the explanations of some outburst phenomena. One view is that a powerful air shock, which is formed by the compression of a rapid coal-gas mixture flow, is the reason outbursts cause great destruction (Zhou et al., 2014, 2015). Chen and Sun et al. (Chen, 2011; Sun et al., 2011) suggested that coal-gas flows can reach sonic speed under critical conditions. Furthermore, they noted that the flow inside the fractured zone becomes choked in these critical conditions, and a loud sound occurs. However, Zhao et al. (2016) explained that a gas-solid plug flow forms in the flow process of coal. The pushing force is applied not just once, and a flow of a great mass can be completed at a slow speed.

In addition, the pulse-shaped crash sound of an outburst is explained using plug flow theory.

Previous studies on outburst coal flow focus only on outburst coal transport without considering the outburst process and coal failure types; in addition, the evolution of outburst parameters and their impact on the flow state of outburst coal are ignored. It is difficult to study how the flow state changes in the flow process.

However, based on outburst process analysis, there is a close relationship between outburst parameters and the flow state of outburst coal, and large changes in the flow state occur in the flow process. Therefore, it is worthwhile to study the outburst process and then determine how the outburst coal flow is transported. Furthermore, it is necessary to establish a relationship between the flow state of outburst coal and the outburst parameters, and some parameters (such as the flow distance and flow velocity) can be calculated quantitatively.

OUTBURST PROCESS AND CASE ANALYSIS

Trigger Stage Disturbed by external activities, coal in a critical state enters rapid failure, a small amount of coal is stripped, and an initial outburst hole forms. This stage provides a space condition for the subsequent development of the outburst. Under critical conditions, outburst experiments were carried out by Guo (2014) and Tu et al.

(2016), and an initial outburst hole was obtained, as shown in Figure 2.

Development Stage This stage is a complex transfer process (Hu et al., 2015) involving the gradual deepening of the damage surface of an outburst into the internal side. It can be subdivided into three periods (growth period, stable period, and decay period) (Guo, 2014), and the coal may experience two kinds of failure: shear failure and tensile failure (Xu et al., 2006). Evidently, the cycle process of coal failure can be described as follows: (1) When the coal is suddenly exposed, the stress quickly transfers, resulting in concentrated stress forms in front of the exposed surface. (2) Under concentrated stress, the coal is in shear failure, and large numbers of coal fractures develop, causing a great weakening of the coal strength (Alonso et al., 2003). (3) A high gas pressure gradient forms because of the quick migration of coal gas. This has a tensile effect on the coal body.

The macroscopic crack extends in the direction perpendicular to the gas pressure gradient, and a “spherical shell” spallation appears (Song and Cheng, 2012). (4) As a result of the supply of desorption gas, a high gas pressure can be maintained in the macroscopic crack, and the spallation is broken and separated from the original coal. (5) The broken coal is consistently transported by airflow to the working space. (6) The damage surface of outburst goes deeper into the internal side.

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

et al. (2011)—a preparation stage, a trigger stage, a development stage, and a termination stage—and these authors noted that the flow process of outburst coal is attributed to the development stage.

With respect to the flow process of outburst coal, limited research exists. Considerable controversies still exist in regard to the flow state, the flow velocity, and the explanations of some outburst phenomena. One view is that a powerful air shock, which is formed by the compression of a rapid coal-gas mixture flow, is the reason outbursts cause great destruction (Zhou et al., 2014, 2015). Chen and Sun et al. (Chen, 2011; Sun et al., 2011) suggested that coal-gas flows can reach sonic speed under critical conditions. Furthermore, they noted that the flow inside the fractured zone becomes choked in these critical conditions, and a loud sound occurs. However, Zhao et al. (2016) explained that a gas-solid plug flow forms in the flow process of coal. The pushing force is applied not just once, and a flow of a great mass can be completed at a slow speed.

In addition, the pulse-shaped crash sound of an outburst is explained using plug flow theory.

Previous studies on outburst coal flow focus only on outburst coal transport without considering the outburst process and coal failure types; in addition, the evolution of outburst parameters and their impact on the flow state of outburst coal are ignored. It is difficult to study how the flow state changes in the flow process.

However, based on outburst process analysis, there is a close relationship between outburst parameters and the flow state of outburst coal, and large changes in the flow state occur in the flow process. Therefore, it is worthwhile to study the outburst process and then determine how the outburst coal flow is transported. Furthermore, it is necessary to establish a relationship between the flow state of outburst coal and the outburst parameters, and some parameters (such as the flow distance and flow velocity) can be calculated quantitatively.

OUTBURST PROCESS AND CASE ANALYSIS

Division of Outburst Process

Division of Outburst Process

The dynamic process for coal and gas outburst is outlined in Figure 1: an outburst generally involves the following stages (Jin et al., 2011), which consider the time and spatial dimensions.

The dynamic process for coal and gas outburst is outlined in Figure 1: an outburst generally involves the following stages (Jin et al., 2011), which consider the time and spatial dimensions.

Preparation Stage All kinds of outburst factors (such as outburst energy, stress, coal gas, etc.) continuously accumulate with the quasi-static failure of coal, and the critical condition or critical threshold is satisfied gradually (Tu et al., 2016).

318

Termination Stage The outburst energy, stress, and coal gas cannot afford the maintenance of the outburst. The outburst ceases, while the coal returns to a static balance. In addition, in the spatial dimension, the outburst process can be divided into two parts, before and after the separation of the outburst, as shown in Figure 2. The “after” part refers to the flow process of outburst coal in the mining space, including coal acceleration, coal deceleration, and coal resting.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Preparation Stage All kinds of outburst factors (such as outburst energy, stress, coal gas, etc.) continuously accumulate with the quasi-static failure of coal, and the critical condition or critical threshold is satisfied gradually (Tu et al., 2016).

318

Trigger Stage Disturbed by external activities, coal in a critical state enters rapid failure, a small amount of coal is stripped, and an initial outburst hole forms. This stage provides a space condition for the subsequent development of the outburst. Under critical conditions, outburst experiments were carried out by Guo (2014) and Tu et al.

(2016), and an initial outburst hole was obtained, as shown in Figure 2.

Development Stage This stage is a complex transfer process (Hu et al., 2015) involving the gradual deepening of the damage surface of an outburst into the internal side. It can be subdivided into three periods (growth period, stable period, and decay period) (Guo, 2014), and the coal may experience two kinds of failure: shear failure and tensile failure (Xu et al., 2006). Evidently, the cycle process of coal failure can be described as follows: (1) When the coal is suddenly exposed, the stress quickly transfers, resulting in concentrated stress forms in front of the exposed surface. (2) Under concentrated stress, the coal is in shear failure, and large numbers of coal fractures develop, causing a great weakening of the coal strength (Alonso et al., 2003). (3) A high gas pressure gradient forms because of the quick migration of coal gas. This has a tensile effect on the coal body.

The macroscopic crack extends in the direction perpendicular to the gas pressure gradient, and a “spherical shell” spallation appears (Song and Cheng, 2012). (4) As a result of the supply of desorption gas, a high gas pressure can be maintained in the macroscopic crack, and the spallation is broken and separated from the original coal. (5) The broken coal is consistently transported by airflow to the working space. (6) The damage surface of outburst goes deeper into the internal side.

Termination Stage The outburst energy, stress, and coal gas cannot afford the maintenance of the outburst. The outburst ceases, while the coal returns to a static balance. In addition, in the spatial dimension, the outburst process can be divided into two parts, before and after the separation of the outburst, as shown in Figure 2. The “after” part refers to the flow process of outburst coal in the mining space, including coal acceleration, coal deceleration, and coal resting.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317â&#x20AC;&#x201C;331

Figure 1. Dynamic process for coal and gas outburst.

Analysis of the Layered Failure of Coal

Figure 1. Dynamic process for coal and gas outburst.

Analysis of the Layered Failure of Coal

319

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317â&#x20AC;&#x201C;331

319

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Figure 2. Initial outburst hole for outburst experiment. (a) Top view; (b) Magnification of local area; and (c) Elevation view from the outburst mouth.

Figure 2. Initial outburst hole for outburst experiment. (a) Top view; (b) Magnification of local area; and (c) Elevation view from the outburst mouth.

Case Analysis Outburst of the Zhongliangshan Coal Mine An outburst occurred in the Zhongliangshan coal mine in Chongqing, China, on November 4, 1977 (Zhao et al., 2016). This outburst occurred in coal uncovered by a crosscut. The coal (rock) ejected measured 817 tons, the whole process lasted for 39 seconds, and 38,540 m3 of coal gas was released. The evolutions of the gas pressure, temperature, sounds, and gas emission speed were recorded, and the synthetic curve in the outburst parameters for the Zhongliangshan outburst is shown in Figure 3. These data were valuable for the study of the outburst mechanism.

As shown in Figure 3, for the closer borehole #2, the distance from the free surface is 5 m. There, the gas pressure drop begins 1.5 to 2.5 seconds after outburst; in the next 1.3 seconds, the gas pressure quickly drops from 1.65 to 0.3 MPa and remains at 0.3 MPa for approximately 3 to 4 seconds. In contrast, for borehole #1, the distance from borehole #2 is 14 m. There the gas pressure drop begins 6 seconds after outburst. The gas pressure drop is slow in the first 3 seconds; then the gas pressure quickly drops from 1.8 to 0.4 MPa, and it remains at 0.4 MPa for a second.

These data reflect the evolution of outburst parameters in the outburst process. The time difference in gas pressure variation, recorded by boreholes #1 and #2, explains that an outburst is a gradual process from the outside to the inside. After the coal is separated, the 320

inner space volume of coal increases sharply, causing a rapid drop in the gas pressure in a short period of time.

However, because of the supply of desorption gas, the gas pressure is almost constant at a small value for a period of time. Therefore, it can be considered that the gas pressure difference between the inner and outer sides of the coal (hereinafter referred to as pressure difference) is almost constant after the coal is separated.

Furthermore, compared with the initial pressure difference (p0 − pa ), this pressure difference is very small.

The pressure difference is 0.3 MPa for borehole #1, which is only 17.6 percent of the initial pressure difference (1.7 MPa). In addition, the pressure difference is 0.2 MPa for borehole #2, only 12.9 percent of the initial pressure difference (1.55 MPa).

Outburst of Xinxing Coal Mine On November 21, 2009, an extra-large outburst occurred in the Xinxing coal mine (Heilongjiang Province, China), in which a violent gas explosion was induced, and 108 people died. The regional structure and coal distributing feature for the Xinxing outburst is shown in Figure 4; this accident occurred in the exploring roadway of the No. 15 coal seam, and the outburst point was located in the geological fracture area, which is controlled by large faults. Near the outburst point, there is a magma intrusion area in the floor of the No. 15 coal seam, with a distance of approximately 7–18 m, and this magma intrusion area has a good sealing effect

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Case Analysis Outburst of the Zhongliangshan Coal Mine An outburst occurred in the Zhongliangshan coal mine in Chongqing, China, on November 4, 1977 (Zhao et al., 2016). This outburst occurred in coal uncovered by a crosscut. The coal (rock) ejected measured 817 tons, the whole process lasted for 39 seconds, and 38,540 m3 of coal gas was released. The evolutions of the gas pressure, temperature, sounds, and gas emission speed were recorded, and the synthetic curve in the outburst parameters for the Zhongliangshan outburst is shown in Figure 3. These data were valuable for the study of the outburst mechanism.

As shown in Figure 3, for the closer borehole #2, the distance from the free surface is 5 m. There, the gas pressure drop begins 1.5 to 2.5 seconds after outburst; in the next 1.3 seconds, the gas pressure quickly drops from 1.65 to 0.3 MPa and remains at 0.3 MPa for approximately 3 to 4 seconds. In contrast, for borehole #1, the distance from borehole #2 is 14 m. There the gas pressure drop begins 6 seconds after outburst. The gas pressure drop is slow in the first 3 seconds; then the gas pressure quickly drops from 1.8 to 0.4 MPa, and it remains at 0.4 MPa for a second.

These data reflect the evolution of outburst parameters in the outburst process. The time difference in gas pressure variation, recorded by boreholes #1 and #2, explains that an outburst is a gradual process from the outside to the inside. After the coal is separated, the 320

inner space volume of coal increases sharply, causing a rapid drop in the gas pressure in a short period of time.

However, because of the supply of desorption gas, the gas pressure is almost constant at a small value for a period of time. Therefore, it can be considered that the gas pressure difference between the inner and outer sides of the coal (hereinafter referred to as pressure difference) is almost constant after the coal is separated.

Furthermore, compared with the initial pressure difference (p0 − pa ), this pressure difference is very small.

The pressure difference is 0.3 MPa for borehole #1, which is only 17.6 percent of the initial pressure difference (1.7 MPa). In addition, the pressure difference is 0.2 MPa for borehole #2, only 12.9 percent of the initial pressure difference (1.55 MPa).

Outburst of Xinxing Coal Mine On November 21, 2009, an extra-large outburst occurred in the Xinxing coal mine (Heilongjiang Province, China), in which a violent gas explosion was induced, and 108 people died. The regional structure and coal distributing feature for the Xinxing outburst is shown in Figure 4; this accident occurred in the exploring roadway of the No. 15 coal seam, and the outburst point was located in the geological fracture area, which is controlled by large faults. Near the outburst point, there is a magma intrusion area in the floor of the No. 15 coal seam, with a distance of approximately 7–18 m, and this magma intrusion area has a good sealing effect

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Analysis of the Layered Failure of Coal

Analysis of the Layered Failure of Coal

Figure 3. Location and synthetic curve in the outburst parameters for the Zhongliangshan outburst. (a) Location map; (b) Synthetic curve.

Figure 3. Location and synthetic curve in the outburst parameters for the Zhongliangshan outburst. (a) Location map; (b) Synthetic curve.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317â&#x20AC;&#x201C;331

321

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317â&#x20AC;&#x201C;331

321

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Figure 4. Regional structure and coal distributing feature in Xinxing outburst.

Figure 4. Regional structure and coal distributing feature in Xinxing outburst.

on coal gas; thus, a high coal gas content is maintained in this area. In addition, the occurrence of the coal seam greatly changes near the outburst point. The No. 15 coal seam is close to the exploring roadway, within a distance of only 10–15 m.

According to the scene investigation, a flat outburst hole with a small mouth and a large cavity is found

322

at the upper-right side of the working surface, and the mouth acreage of the outburst hole is approximately 2.5 m2 . The coal (rock) ejected is 3,845 tons, including 1,697 tons of coal and 2,148 tons of rock. The packing length (distribution length) of the coal (rock) in the roadway is 317 m, including a length of 278 m in the exploring roadway. There is obvious zonal distribution

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

on coal gas; thus, a high coal gas content is maintained in this area. In addition, the occurrence of the coal seam greatly changes near the outburst point. The No. 15 coal seam is close to the exploring roadway, within a distance of only 10–15 m.

According to the scene investigation, a flat outburst hole with a small mouth and a large cavity is found

322

at the upper-right side of the working surface, and the mouth acreage of the outburst hole is approximately 2.5 m2 . The coal (rock) ejected is 3,845 tons, including 1,697 tons of coal and 2,148 tons of rock. The packing length (distribution length) of the coal (rock) in the roadway is 317 m, including a length of 278 m in the exploring roadway. There is obvious zonal distribution

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Analysis of the Layered Failure of Coal

Analysis of the Layered Failure of Coal

Table 1. Experimental scheme of coal and gas outburst.

Table 1. Experimental scheme of coal and gas outburst.

Triaxial Stress

Experiment No.

Gas Pressure (MPa)

Ambient Temperature (◦ C)

z (MPa)

x (MPa)

y (MPa)

1 2 3 4 5 6

0.25 0.30 0.35 0.40 0.45 0.50

25 25 25 25 25 25

4.8 4.8 4.8 4.8 4.8 4.8

5.5 5.5 5.5 5.5 5.5 5.5

5.5 5.5 5.5 5.5 5.5 5.5

Note: The axial direction of the outburst mouth is defined as the x axis, the lateral direction as the y axis, and the vertical direction as the z axis.

for this coal (rock) in the exploring roadway. Interestingly, the coal is concentrated in an area far from the outburst point, although a 10-m rock pillar exists in front of the coal seam. As shown in Figure 4, in the range from 278 m (Section A-A) to 113 m (Section FF), the coal is mainly packed. Then, in the range from 113 m (Section F-F) to 11 m (Section J-J), the roadway is filled with rock. However, in the range inside of 11 m (Section J-J), the roadway is piled up with large rocks. The largest rock, with dimensions of 3.6 × 3.0 × 2.6 m, is found at a distance of more than 10 m from the working surface. In addition, the packing height (distribution height) of coal reaches 1.5 m at Section A-A. Coal and Gas Outburst Experiments A coal and gas outburst experiment is a direct and effective method for reconstructing the outburst process and can reflect the features of an outburst (Geng et al., 2017). A large number of outburst experiments were conducted by Jiang and Yu (1994), and spherical shell layers were found in those experiments.

Similar experiments were conducted by Guo (2014), and the layer characteristics of coal after an outburst were obtained. In addition, using a transparent shock tube and a high-speed camera, Guan et al. (2009) observed the failure process of coal.

Using a triaxial coal and gas outburst simulation system, series of outburst experiments were conducted by the authors (the experimental scheme is shown in Table 1). For the experiments, outburst occurred, a large amount of outburst coal was ejected, and there was an obvious spherical outburst hole in the residual coal after the outburst ended. Behind the outburst hole, multi-layer coal layers, which are listed layer by layer, were found. As previously noted, these layers were spherical, and the coal maintained good integrity in the radial direction, as shown in Figure 5.

Figure 5. Features of outburst coal in experiment chamber.

The flow of outburst coal in different periods (including early, middle, and later stages) and the final distribution were recorded by a high-speed camera. The flow and distribution features of outburst coal in Experiment 6 are shown in Figure 6. These three periods correspond to the outburst growth period, the stable period, and the decay period (Guo, 2014). In the outburst growth period (Figure 6a), the ejection mass and ejection velocity of outburst coal increase gradually. In the outburst stable period (Figure 6b), a large amount of outburst coal is ejected at a high speed. However, a clear sign of decline is found in the outburst decay period (Figure 6c), and the ejection mass and ejection velocity of outburst coal decrease with a lack of coal gas. A stopwatch was used to record the time. The entire process of the outburst is 1.69 seconds, and the time of the decay period is 0.19 seconds, only 11.2 percent of the total time. Comparing Figure 6b and c, the coal distributed near the outburst mouth is ejected in the decay period, and the mass of coal is very small. As shown in Figure 6d, the distribution of the outburst coal is spindle shaped, and most of the coal is concentrated in an area far from the outburst mouth.

Some information about the outburst features is obtained in these experiments (Table 2) (Tu et al., 2015), including the outburst results, maximum flow distance of outburst coal (Lmax ), total mass of outburst coal (M), length of the outburst hole (D), and average thickness of the coal layer (d). Simplifying these outburst holes into a regular half-circle in the horizontal plane, the radius (ra ) and half of the field angle ( /2) of the outburst hole are calculated as well. Furthermore, based on the distance from the outburst mouth, the distributions of outburst coal in the ground are equally divided into five areas, marked as areas I through V, as shown in Figure 6. The mass of outburst coal in each area is recorded, and the mass ratio is calculated, as shown in Table 3.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

323

Triaxial Stress

Experiment No.

Gas Pressure (MPa)

Ambient Temperature (◦ C)

z (MPa)

x (MPa)

y (MPa)

1 2 3 4 5 6

0.25 0.30 0.35 0.40 0.45 0.50

25 25 25 25 25 25

4.8 4.8 4.8 4.8 4.8 4.8

5.5 5.5 5.5 5.5 5.5 5.5

5.5 5.5 5.5 5.5 5.5 5.5

Note: The axial direction of the outburst mouth is defined as the x axis, the lateral direction as the y axis, and the vertical direction as the z axis.

for this coal (rock) in the exploring roadway. Interestingly, the coal is concentrated in an area far from the outburst point, although a 10-m rock pillar exists in front of the coal seam. As shown in Figure 4, in the range from 278 m (Section A-A) to 113 m (Section FF), the coal is mainly packed. Then, in the range from 113 m (Section F-F) to 11 m (Section J-J), the roadway is filled with rock. However, in the range inside of 11 m (Section J-J), the roadway is piled up with large rocks. The largest rock, with dimensions of 3.6 × 3.0 × 2.6 m, is found at a distance of more than 10 m from the working surface. In addition, the packing height (distribution height) of coal reaches 1.5 m at Section A-A. Coal and Gas Outburst Experiments A coal and gas outburst experiment is a direct and effective method for reconstructing the outburst process and can reflect the features of an outburst (Geng et al., 2017). A large number of outburst experiments were conducted by Jiang and Yu (1994), and spherical shell layers were found in those experiments.

Similar experiments were conducted by Guo (2014), and the layer characteristics of coal after an outburst were obtained. In addition, using a transparent shock tube and a high-speed camera, Guan et al. (2009) observed the failure process of coal.

Using a triaxial coal and gas outburst simulation system, series of outburst experiments were conducted by the authors (the experimental scheme is shown in Table 1). For the experiments, outburst occurred, a large amount of outburst coal was ejected, and there was an obvious spherical outburst hole in the residual coal after the outburst ended. Behind the outburst hole, multi-layer coal layers, which are listed layer by layer, were found. As previously noted, these layers were spherical, and the coal maintained good integrity in the radial direction, as shown in Figure 5.

Figure 5. Features of outburst coal in experiment chamber.

The flow of outburst coal in different periods (including early, middle, and later stages) and the final distribution were recorded by a high-speed camera. The flow and distribution features of outburst coal in Experiment 6 are shown in Figure 6. These three periods correspond to the outburst growth period, the stable period, and the decay period (Guo, 2014). In the outburst growth period (Figure 6a), the ejection mass and ejection velocity of outburst coal increase gradually. In the outburst stable period (Figure 6b), a large amount of outburst coal is ejected at a high speed. However, a clear sign of decline is found in the outburst decay period (Figure 6c), and the ejection mass and ejection velocity of outburst coal decrease with a lack of coal gas. A stopwatch was used to record the time. The entire process of the outburst is 1.69 seconds, and the time of the decay period is 0.19 seconds, only 11.2 percent of the total time. Comparing Figure 6b and c, the coal distributed near the outburst mouth is ejected in the decay period, and the mass of coal is very small. As shown in Figure 6d, the distribution of the outburst coal is spindle shaped, and most of the coal is concentrated in an area far from the outburst mouth.

Some information about the outburst features is obtained in these experiments (Table 2) (Tu et al., 2015), including the outburst results, maximum flow distance of outburst coal (Lmax ), total mass of outburst coal (M), length of the outburst hole (D), and average thickness of the coal layer (d). Simplifying these outburst holes into a regular half-circle in the horizontal plane, the radius (ra ) and half of the field angle ( /2) of the outburst hole are calculated as well. Furthermore, based on the distance from the outburst mouth, the distributions of outburst coal in the ground are equally divided into five areas, marked as areas I through V, as shown in Figure 6. The mass of outburst coal in each area is recorded, and the mass ratio is calculated, as shown in Table 3.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

323

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Figure 6. Flow and distribution features of outburst coal. (a) Earlier stage; (b) Middle stage; (c) Later stage; and (d) Stop.

Figure 6. Flow and distribution features of outburst coal. (a) Earlier stage; (b) Middle stage; (c) Later stage; and (d) Stop.

FLOW PROCESS OF OUTBURST COAL Simplifying of the Flow Process The outburst cases and the outburst experiment show that an outburst is a transfer process, and the coal is broken into multi-layer coal layers with a certain thickness. For a layer, the flow process after the separation can be simplified as follows: 1. The move velocity of outburst coal is ignored before it is separated. That is, it is quasi-static in the process from coal crack growth to coal separation (Pan and Li, 2013).

2. After the coal layer is separated, the gas pressure in the inner side of the coal layer drops rapidly. However, as a result of the supply of desorption gas, the gas pressure is almost constant at a small value, and

a constant pressure difference is maintained for a period of time. The coal layer moves forward, driven by the constant pressure difference, and this motion can be attributed to a constant acceleration motion.

A typical outburst hole is pear-shaped, with a small mouth and a large cavity; the contracting mouth has a definite inhibitory effect on coal gas emissions, ensuring the maintenance of a gas difference. Therefore, the outburst coal motion in an outburst hole can be simplified to a constant acceleration motion (Sun et al., 2011).

3. After the outburst coal enters the mining space, which is relatively open, the gas expands rapidly, and the remaining pressure difference disappears in an instant. However, for many outburst cases and outburst experiments, there is a distance between the outburst mouth and the ground, and this allows

FLOW PROCESS OF OUTBURST COAL Simplifying of the Flow Process The outburst cases and the outburst experiment show that an outburst is a transfer process, and the coal is broken into multi-layer coal layers with a certain thickness. For a layer, the flow process after the separation can be simplified as follows: 1. The move velocity of outburst coal is ignored before it is separated. That is, it is quasi-static in the process from coal crack growth to coal separation (Pan and Li, 2013).

2. After the coal layer is separated, the gas pressure in the inner side of the coal layer drops rapidly. However, as a result of the supply of desorption gas, the gas pressure is almost constant at a small value, and

Table 2. Results and parameters for outburst experiments.

a constant pressure difference is maintained for a period of time. The coal layer moves forward, driven by the constant pressure difference, and this motion can be attributed to a constant acceleration motion.

A typical outburst hole is pear-shaped, with a small mouth and a large cavity; the contracting mouth has a definite inhibitory effect on coal gas emissions, ensuring the maintenance of a gas difference. Therefore, the outburst coal motion in an outburst hole can be simplified to a constant acceleration motion (Sun et al., 2011).

3. After the outburst coal enters the mining space, which is relatively open, the gas expands rapidly, and the remaining pressure difference disappears in an instant. However, for many outburst cases and outburst experiments, there is a distance between the outburst mouth and the ground, and this allows

Table 2. Results and parameters for outburst experiments.

Experiment No.

Results of Outburst

M (kg)

Lmax (m)

D (m)

d (m)

(◦ )

R (m)

Experiment No.

Results of Outburst

M (kg)

Lmax (m)

D (m)

d (m)

(◦ )

R (m)

1 2 3 4 5 6

No outburst Stripping Stripping Outburst Outburst Outburst

0 0.009 0.026 4.126 5.295 6.366

0 0.10 0.21 14.47 16.98 18.14

0 0.012 0.023 0.109 0.132 0.167

— — — 0.0128 0.0118 0.0092

— — 169.76 193.76 205.12 216.82

— — 0.0251 0.0972 0.1086 0.1270

1 2 3 4 5 6

No outburst Stripping Stripping Outburst Outburst Outburst

0 0.009 0.026 4.126 5.295 6.366

0 0.10 0.21 14.47 16.98 18.14

0 0.012 0.023 0.109 0.132 0.167

— — — 0.0128 0.0118 0.0092

— — 169.76 193.76 205.12 216.82

— — 0.0251 0.0972 0.1086 0.1270

324

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

324

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Analysis of the Layered Failure of Coal

Analysis of the Layered Failure of Coal

Figure 7. Simplified flow process of outburst coal.

Figure 7. Simplified flow process of outburst coal.

the coal to move a distance in the air. If the air friction resistance is ignored, this process is similar to a freefall motion.

4. When the outburst coal traveling at a high speed falls to the ground, it will continue to move forward.

However, a strong friction resistance exists, and the coal slows down and then comes to a rest.

2. Before ejection from the outburst mouth, the coal is crushed, but the original layered structure is maintained.

3. Coal gas is an ideal gas, the coal is considered as a continuous isotropic elastic medium, and the friction coefficient is constant.

4. Other energy losses, including particle/particle collision, heat loss, and sound energy, are neglected.

5. The roadway is horizontal without a dip angle.

According to the spatial dimension, the flow process of outburst coal can be divided into a quasi-static phase, a constant acceleration phase, a freefall phase, and a deceleration phase, as shown in Figure 7.

Constant Acceleration Phase A pressure difference coefficient is introduced to characterize the change of the pressure difference before and after the separation of the coal layer. It is worth noting that this pressure difference coefficient changes with the development of an outburst. In the outburst growth period and stable period (Figure 6a and b), the gas supply is sufficient, and the coefficient may maintain relative stability. In contrast, in the outburst decay period (Figure 6c) the gas supply is limited and the coefficient will decrease rapidly. Therefore, the pressure difference, after the coal layer is separated, can be expressed as follows:

Establishment of Simplified Model Basic Assumptions Some assumptions are needed for the establishment of the simplified outburst coal flow model: 1. Outburst stably develops in an infinite space. The coal behind the initial exposed surface is broken into multi-layer coal layers with a spherical shell shape.

These layers are similar, and the field angles are the same.

� pi = i � p0 ,

(1)

the coal to move a distance in the air. If the air friction resistance is ignored, this process is similar to a freefall motion.

4. When the outburst coal traveling at a high speed falls to the ground, it will continue to move forward.

However, a strong friction resistance exists, and the coal slows down and then comes to a rest.

According to the spatial dimension, the flow process of outburst coal can be divided into a quasi-static phase, a constant acceleration phase, a freefall phase, and a deceleration phase, as shown in Figure 7.

I

II

III

Some assumptions are needed for the establishment of the simplified outburst coal flow model: 1. Outburst stably develops in an infinite space. The coal behind the initial exposed surface is broken into multi-layer coal layers with a spherical shell shape.

These layers are similar, and the field angles are the same.

V

I

II

III

(1)

Experiment No. and Area 6

IV

� pi = i � p0 ,

Table 3. Mass and mass ratio of outburst coal in different areas.

5 IV

A pressure difference coefficient is introduced to characterize the change of the pressure difference before and after the separation of the coal layer. It is worth noting that this pressure difference coefficient changes with the development of an outburst. In the outburst growth period and stable period (Figure 6a and b), the gas supply is sufficient, and the coefficient may maintain relative stability. In contrast, in the outburst decay period (Figure 6c) the gas supply is limited and the coefficient will decrease rapidly. Therefore, the pressure difference, after the coal layer is separated, can be expressed as follows:

Basic Assumptions

Experiment No. and Area

Parameter

Constant Acceleration Phase

Establishment of Simplified Model

Table 3. Mass and mass ratio of outburst coal in different areas.

4

2. Before ejection from the outburst mouth, the coal is crushed, but the original layered structure is maintained.

3. Coal gas is an ideal gas, the coal is considered as a continuous isotropic elastic medium, and the friction coefficient is constant.

4. Other energy losses, including particle/particle collision, heat loss, and sound energy, are neglected.

5. The roadway is horizontal without a dip angle.

V

I

II

III

4 IV

V

Mass (kg) 0.223 0.755 0.825 1.002 1.320 0.228 0.630 0.911 1.557 1.970 0.401 1.229 1.190 1.426 2.120 Mass ratio (%) 5.4 18.3 20.0 24.3 32.0 4.3 11.9 17.2 29.4 37.2 6.3 19.3 18.7 22.4 33.3

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

325

Parameter

I

II

III

5 IV

V

I

II

III

6 IV

V

I

II

III

IV

V

Mass (kg) 0.223 0.755 0.825 1.002 1.320 0.228 0.630 0.911 1.557 1.970 0.401 1.229 1.190 1.426 2.120 Mass ratio (%) 5.4 18.3 20.0 24.3 32.0 4.3 11.9 17.2 29.4 37.2 6.3 19.3 18.7 22.4 33.3

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

325

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

where ďż˝pi is the pressure difference of layer i (i â&#x2030;Ľ 1) in MPa; i is the pressure difference coefficient of layer i; and ďż˝p0 is the initial pressure difference, ďż˝p0 = p0 â&#x2C6;&#x2019; pa , in MPa. Based on Eq. 1, with the development of the outburst, the flow distance of each layer in the outburst hole increases gradually, and this distance is expressed as follows: Li = l0 + R0 +

n

di ,

(2)

where Li is the flow distance in the outburst hole of layer i (in meters), and l0 is the distance from the curvature center of initial exposed surface to the outburst mouth. The value is negative if the curvature center is out of the outburst mouth; otherwise, it is positive. R0 is the curvature radius of the initial exposed surface (in meters).

Through the analysis of coal stress, the resultant stress of a layer is obtained.

Fi = ďż˝ pi Ă&#x2014; Ai â&#x2C6;&#x2019; fm1 ;

(3) (4)

where Fi is the resultant stress of layer i (in Newtons); Ai is the equivalent area (in m2 ); fm1 is the frictional resistance (in Newtons); is the density (in kg/m3 ); g is the gravity acceleration (in m/s2 ); 1 is the friction coefficient; and Vi is the volume of layer i. Noting that there is simplified calculation for the equivalent area and volume of layer, the expressions are as follows:

2 Ai = 2 Ri Ă&#x2014; 1 â&#x2C6;&#x2019; cos ; (5) 2 Vi = Ai di ;

(6)

where is the field angle of layer (in degrees) and Ri is the radius of layer i, approximately Ri = Riâ&#x2C6;&#x2019;1 + di (in meters).

A simple equation is used to approximately calculate the mass of each layer, and the total mass of outburst coal is also obtained. mi = Ai di ; M=

n

mi ;

(7) (8)

i =1

where mi is the mass of layer i (in kg) and M is the total mass (in kg).

The motion of coal in the outburst hole follows the Newtonâ&#x20AC;&#x2122;s second law and energy theorem. Thus, after acceleration, the kinetic energy (Ei ďż˝ ) and velocity (vi ďż˝ ) of each layer are obtained.

326

Ei ďż˝ = l0 + R0 +

n

i =1

di

Ă&#x2014; (ďż˝ pi Ă&#x2014; Ai â&#x2C6;&#x2019; fm1 );

(9)

n

ďż˝ pi ďż˝

â&#x2C6;&#x2019; g 1 . (10) vi = 2 l0 + R0 + di Ă&#x2014; di i =1

where ďż˝pi is the pressure difference of layer i (i â&#x2030;Ľ 1) in MPa; i is the pressure difference coefficient of layer i; and ďż˝p0 is the initial pressure difference, ďż˝p0 = p0 â&#x2C6;&#x2019; pa , in MPa. Based on Eq. 1, with the development of the outburst, the flow distance of each layer in the outburst hole increases gradually, and this distance is expressed as follows:

Freefall Phase

i =1

fm1 = g 1 Vi ;

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

In this stage, after the coal is ejected from the outburst mouth, the coal continues to flow in the air for some time, and the flow distance can be calculated as follows (Zhao et al., 2016): h=

1 2 gt ; 2

Li ďż˝ = vi t;

(11) (12)

where h is the distance between the outburst mouth and the ground (in meters), and Li ďż˝ is the flow distance in this stage (in meters).

Li = l0 + R0 +

Li �� =

Ei , Vi g 2

(13)

where 2 is the friction coefficient in this stage.

Using Eqs. 8 and 9, the flow distance of ejecting coal in mining space can be obtained: Li = Li � + Li �� .

(14)

Calculation and Analysis As described in Section 3, those outburst experiments not only obtain the distribution features of the outburst coal (Tables 2 and 3) but they also provide the relevant parameters (e.g., length of the outburst hole, average thickness of coal layer, the radius and half of the field angle of the outburst hole) for the calculation of the simplified model. Beyond that, some other parameters are needed, and those parameters come from the literature or experimental measurements, ensuring that the values are reasonable (Table 4).

However, the pressure difference coefficient in Eq. 1 is difficult to obtain, and this coefficient changes with the state of outburst. For the case of the Zhongliangshan coal mine, the values of this coefficient, which are monitored by two monitoring boreholes, are 0.176 and 0.129. However, for the outburst experiment, the coal

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317â&#x20AC;&#x201C;331

di ,

(2)

i =1

Fi = ďż˝ pi Ă&#x2014; Ai â&#x2C6;&#x2019; fm1 ; fm1 = g 1 Vi ;

(3) (4)

where Fi is the resultant stress of layer i (in Newtons); Ai is the equivalent area (in m2 ); fm1 is the frictional resistance (in Newtons); is the density (in kg/m3 ); g is the gravity acceleration (in m/s2 ); 1 is the friction coefficient; and Vi is the volume of layer i. Noting that there is simplified calculation for the equivalent area and volume of layer, the expressions are as follows:

2 Ai = 2 Ri Ă&#x2014; 1 â&#x2C6;&#x2019; cos ; (5) 2 Vi = Ai di ;

(6)

where is the field angle of layer (in degrees) and Ri is the radius of layer i, approximately Ri = Riâ&#x2C6;&#x2019;1 + di (in meters).

A simple equation is used to approximately calculate the mass of each layer, and the total mass of outburst coal is also obtained. mi = Ai di ; M=

n

mi ;

(7) (8)

i =1

where mi is the mass of layer i (in kg) and M is the total mass (in kg).

The motion of coal in the outburst hole follows the Newtonâ&#x20AC;&#x2122;s second law and energy theorem. Thus, after acceleration, the kinetic energy (Ei ďż˝ ) and velocity (vi ďż˝ ) of each layer are obtained.

326

Ei ďż˝ = l0 + R0 +

n

i =1

di

Ă&#x2014; (ďż˝ pi Ă&#x2014; Ai â&#x2C6;&#x2019; fm1 );

(9)

n

ďż˝ pi ďż˝

â&#x2C6;&#x2019; g 1 . (10) vi = 2 l0 + R0 + di Ă&#x2014; di i =1

Freefall Phase

where Li is the flow distance in the outburst hole of layer i (in meters), and l0 is the distance from the curvature center of initial exposed surface to the outburst mouth. The value is negative if the curvature center is out of the outburst mouth; otherwise, it is positive. R0 is the curvature radius of the initial exposed surface (in meters).

Through the analysis of coal stress, the resultant stress of a layer is obtained.

Deceleration Phase Under the effect of friction resistance, the coal slows and then comes to a rest. Therefore, the flow distance in this stage is as follows:

n

In this stage, after the coal is ejected from the outburst mouth, the coal continues to flow in the air for some time, and the flow distance can be calculated as follows (Zhao et al., 2016): h=

1 2 gt ; 2

Li ďż˝ = vi t;

(11) (12)

where h is the distance between the outburst mouth and the ground (in meters), and Li ďż˝ is the flow distance in this stage (in meters).

Deceleration Phase Under the effect of friction resistance, the coal slows and then comes to a rest. Therefore, the flow distance in this stage is as follows: Li �� =

Ei , Vi g 2

(13)

where 2 is the friction coefficient in this stage.

Using Eqs. 8 and 9, the flow distance of ejecting coal in mining space can be obtained: Li = Li � + Li �� .

(14)

Calculation and Analysis As described in Section 3, those outburst experiments not only obtain the distribution features of the outburst coal (Tables 2 and 3) but they also provide the relevant parameters (e.g., length of the outburst hole, average thickness of coal layer, the radius and half of the field angle of the outburst hole) for the calculation of the simplified model. Beyond that, some other parameters are needed, and those parameters come from the literature or experimental measurements, ensuring that the values are reasonable (Table 4).

However, the pressure difference coefficient in Eq. 1 is difficult to obtain, and this coefficient changes with the state of outburst. For the case of the Zhongliangshan coal mine, the values of this coefficient, which are monitored by two monitoring boreholes, are 0.176 and 0.129. However, for the outburst experiment, the coal

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317â&#x20AC;&#x201C;331

Analysis of the Layered Failure of Coal Table 4. Magnitudes of parameters used in the calculations.

Analysis of the Layered Failure of Coal

Table 5. Total mass of outburst coal in different experiments.

Table 4. Magnitudes of parameters used in the calculations.

Parameter

Value

Source

Density, (kg/m3 ) Distance between the outburst mouth and the ground, h (m) Frictional resistance, 1 Frictional resistance, 2 Gravity acceleration, g (m/s2 )

1.4 × 103 0.9

Measurement

Parameter

4

5

6 6.295 6.846

Hodot, 1966

4.125 4.152 4.27

5.296 5.116

0.4 0.4 10

Experimental value (kg) Theoretical value (kg) MRE (%)

gas source in space is limited, and the supply of coal gas is insufficient. The decay period is short, and the outburst abruptly stops. Therefore, the pressure difference coefficient of the outburst experiment will be smaller, and the values are considered to be relative stability for most of the outburst time. Based on the experiment parameters (Lmax , D, and d) with Eqs. 1, 10, 1, and 14, the range of pressure difference coefficient in different experiments can be obtained preliminarily, and the range is 0.011 to 0.023.

Then, according to the curvature radius of the initial exposed surface in Experiment 3, and using the average thickness of coal layer (d), the number of layers in each experiment can be calculated. Furthermore, the pressure difference coefficient is simplified to a constant in the whole outburst process, and a middle value i = 0.016 is selected. The flow distance of each layer is obtained by using the simplified model and the related parameters in Tables 2 and 4. Finally, Eqs. 7 and 8 are used to calculate the mass of each layer and the total mass, respectively. Refering to the division method of the outburst experiment, the distribution of outburst coal is equally divided into five areas, marked as areas I through V. The theoretical mass ratio in different areas can also be found, and the theoretical and experimental values are compared, as shown in Figure 8. An average relative error (RE) is used to calculate the error between the theoretical and experimental values (Table

Experiment No.

5), thus (Jackowski and Woźniakowski, 1987):

n 100 (Mi e − Mi c ) MRE =

, n Mi e e

c

i =1

where Mi and Mi are the experimental and theoretical values of total mass in the different experiments, respectively.

The theoretical value of total mass in Table 5 shows good agreement with the experimental value, and the average relative error of three outburst experiments is 4.27. Otherwise, as shown in Figure 8, the distribution features of outburst coal based on the theoretical calculation are consistent with experimental statistics. The outburst coal is concentrated in areas IV and V, and the mass ratio of these two areas is more than 60 percent, but the mass ratio of the area I is less than 7 percent. We also find that the theoretical mass ratio in the different area is close to the experimental value, which verifies that the theoretical value is reasonable. Therefore, the simplified model of the outburst coal flow, based on the assumption that the coal is broken into multi-layer coal layers with a spherical shell shape, can predict the flow process of outburst coal well.

DISCUSSION Layered Failure of Outburst Coal Hodot (1966) noted that an outburst is the combined result of stress, coal gas, and coal strength, and it has a close relationship with the high gas pressure gradient near the exposed surface. In the process of outburst,

Figure 8. Mass ratio of outburst coal in different areas. (1) Experiment 4; (2) Experiment 5; and (3) Experiment 6.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Table 5. Total mass of outburst coal in different experiments.

Parameter

Value

Source

Density, (kg/m3 ) Distance between the outburst mouth and the ground, h (m) Frictional resistance, 1 Frictional resistance, 2 Gravity acceleration, g (m/s2 )

1.4 × 103 0.9

Measurement

Parameter

4

5

6 6.295 6.846

Hodot, 1966

4.125 4.152 4.27

5.296 5.116

0.4 0.4 10

Experimental value (kg) Theoretical value (kg) MRE (%)

gas source in space is limited, and the supply of coal gas is insufficient. The decay period is short, and the outburst abruptly stops. Therefore, the pressure difference coefficient of the outburst experiment will be smaller, and the values are considered to be relative stability for most of the outburst time. Based on the experiment parameters (Lmax , D, and d) with Eqs. 1, 10, 1, and 14, the range of pressure difference coefficient in different experiments can be obtained preliminarily, and the range is 0.011 to 0.023.

Then, according to the curvature radius of the initial exposed surface in Experiment 3, and using the average thickness of coal layer (d), the number of layers in each experiment can be calculated. Furthermore, the pressure difference coefficient is simplified to a constant in the whole outburst process, and a middle value i = 0.016 is selected. The flow distance of each layer is obtained by using the simplified model and the related parameters in Tables 2 and 4. Finally, Eqs. 7 and 8 are used to calculate the mass of each layer and the total mass, respectively. Refering to the division method of the outburst experiment, the distribution of outburst coal is equally divided into five areas, marked as areas I through V. The theoretical mass ratio in different areas can also be found, and the theoretical and experimental values are compared, as shown in Figure 8. An average relative error (RE) is used to calculate the error between the theoretical and experimental values (Table

Experiment No.

5), thus (Jackowski and Woźniakowski, 1987):

n 100 (Mi e − Mi c ) MRE =

, n Mi e e

c

i =1

where Mi and Mi are the experimental and theoretical values of total mass in the different experiments, respectively.

The theoretical value of total mass in Table 5 shows good agreement with the experimental value, and the average relative error of three outburst experiments is 4.27. Otherwise, as shown in Figure 8, the distribution features of outburst coal based on the theoretical calculation are consistent with experimental statistics. The outburst coal is concentrated in areas IV and V, and the mass ratio of these two areas is more than 60 percent, but the mass ratio of the area I is less than 7 percent. We also find that the theoretical mass ratio in the different area is close to the experimental value, which verifies that the theoretical value is reasonable. Therefore, the simplified model of the outburst coal flow, based on the assumption that the coal is broken into multi-layer coal layers with a spherical shell shape, can predict the flow process of outburst coal well.

DISCUSSION Layered Failure of Outburst Coal Hodot (1966) noted that an outburst is the combined result of stress, coal gas, and coal strength, and it has a close relationship with the high gas pressure gradient near the exposed surface. In the process of outburst,

Figure 8. Mass ratio of outburst coal in different areas. (1) Experiment 4; (2) Experiment 5; and (3) Experiment 6.

327

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

327

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

Figure 9. Distribution of stress and gas pressure in front of the exposed surface. (a) Shear failure; (b) Tensile failure.

Figure 9. Distribution of stress and gas pressure in front of the exposed surface. (a) Shear failure; (b) Tensile failure.

the instantaneous exposure of coal results in a rapid transfer of stress, and three stress areas form in front of the exposed surface: the unloading stress area, the concentration stress area, and the original state area.

Then, controlled by stress state and coal gas migration, there are two types of coal failure: shear failure and tensile failure. Figure 9 describes the distribution of the stress and gas pressure in front of the exposed surface, and the failure types of coal are displayed as well.

According to the magnitude of stress and gas pressure in different areas, it is easy to find that shear failure occurs in the concentration stress area when the leading factor is stress. Tensile failure occurs under the effect of a high gas pressure gradient in the unloading stress area, and the crack extends along the direction perpendicular to the gas pressure gradient (Paterson, 1986).

The coal is broken into a coal layer with a certain thickness, and the shape of the layer is related to the shape of the exposed surface. As in the experimental results, there are multi-layer coal layers behind the outburst hole, which are listed layer by layer (Figure 5). Therefore, it can be considered that the outburst coal fails in the form of multiple similar layers.

As described in the “spherical shell destabilization” hypothesis proposed by Jiang and Yu (1994), the coal is broken into spherical shell layers. This hypothesis con328

sidered that the occurrence of outbursts is related to a high gas pressure gradient near the exposed surface.

Once the mechanical condition is satisfied, the spherical shell layer forms and is then ejected. In addition, the hypothesis holds that the outburst develops layer by layer. After the previous layer is ejected, the space condition is provided, and the failure appears in the next layer.

In total, it is not difficult to conclude that the outburst coal fails in the form of multiple similar layers, and these layers fail gradually layer by layer. This special failure type is caused by the tensile effect of the gas pressure gradient, and the shape of each layer is similarly related to the shape of the initial exposed surface.

Flow Velocity of Outburst Coal As mentioned above, after the coal is separated, the coal is accelerated as a result of the pressure difference, and a high flow velocity is obtained when the coal is ejected from the outburst mouth. However, there is a debate regarding the flow velocity of outburst coal.

Based on energy analysis, Zhao et al. (2016) calculated the flow velocity of outburst coal in the case of the Zhongliangshan outburst, and the flow velocity range was 2.0–12.2 m/s. Wang (2012) used a shock tube

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

the instantaneous exposure of coal results in a rapid transfer of stress, and three stress areas form in front of the exposed surface: the unloading stress area, the concentration stress area, and the original state area.

Then, controlled by stress state and coal gas migration, there are two types of coal failure: shear failure and tensile failure. Figure 9 describes the distribution of the stress and gas pressure in front of the exposed surface, and the failure types of coal are displayed as well.

According to the magnitude of stress and gas pressure in different areas, it is easy to find that shear failure occurs in the concentration stress area when the leading factor is stress. Tensile failure occurs under the effect of a high gas pressure gradient in the unloading stress area, and the crack extends along the direction perpendicular to the gas pressure gradient (Paterson, 1986).

The coal is broken into a coal layer with a certain thickness, and the shape of the layer is related to the shape of the exposed surface. As in the experimental results, there are multi-layer coal layers behind the outburst hole, which are listed layer by layer (Figure 5). Therefore, it can be considered that the outburst coal fails in the form of multiple similar layers.

As described in the “spherical shell destabilization” hypothesis proposed by Jiang and Yu (1994), the coal is broken into spherical shell layers. This hypothesis con328

sidered that the occurrence of outbursts is related to a high gas pressure gradient near the exposed surface.

Once the mechanical condition is satisfied, the spherical shell layer forms and is then ejected. In addition, the hypothesis holds that the outburst develops layer by layer. After the previous layer is ejected, the space condition is provided, and the failure appears in the next layer.

In total, it is not difficult to conclude that the outburst coal fails in the form of multiple similar layers, and these layers fail gradually layer by layer. This special failure type is caused by the tensile effect of the gas pressure gradient, and the shape of each layer is similarly related to the shape of the initial exposed surface.

Flow Velocity of Outburst Coal As mentioned above, after the coal is separated, the coal is accelerated as a result of the pressure difference, and a high flow velocity is obtained when the coal is ejected from the outburst mouth. However, there is a debate regarding the flow velocity of outburst coal.

Based on energy analysis, Zhao et al. (2016) calculated the flow velocity of outburst coal in the case of the Zhongliangshan outburst, and the flow velocity range was 2.0–12.2 m/s. Wang (2012) used a shock tube

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Analysis of the Layered Failure of Coal Table 6. Flow velocity of outburst coal.

Experiment No. Parameter

4

5

6

Gas pressure (MPa) Flow velocity (m/s)

0.40 5.1–7.8

0.45 6.4–9.4

0.50 8.6–12.1

apparatus to test the failure process of a cylinder coal sample driven by CO2 and obtained the flow velocity of a coal-gas mixture under different gas pressures, and the ejection velocity of the coal with the gas pressure is shown in Figure 10. The flow velocity of ejecting coal increases as the gas pressure increases, and the flow velocity can reach 20–60 m/s when the gas pressure is 4–10 MPa. Based on the data from the outburst experiment and using Eq. 10, the flow velocity of each layer is calculated, and the flow velocity range of the different experiments is recorded in Table 6. For the outburst experiment, when the maximum flow distance of the outburst coal is less than 20 m, the flow velocity for the outburst coal is approximately 10 m/s. As the gas pressure increases, the maximum flow distance and flow velocity for outburst coal increase correspondingly. However, for many outburst cases, the maximum flow distance is tens to hundreds of meters, and the ejection velocity may reach tens of meters per second.

Distribution Features of Outburst Coal According to the scene investigation at the Xinxing coal mine outburst, there was obvious zonal distribu-

Analysis of the Layered Failure of Coal

tion for the outburst coal (rock) in the exploring roadway, and the coal was concentrated in an area far from the outburst mouth (Figure 6). A similar phenomenon was found in outburst experiments, with the outburst coal concentrated in areas IV and V, but the mass ratio of area I was less than 7 percent.

If we use the layered failure of outburst coal to explain the distribution features, a reasonable answer can be achieved. In the process of outburst, the coal is broken into layers with a certain thickness. For the same layer, after undergoing acceleration, the flow velocities of the different coal bodies are similar, and the flow distances for those coal bodies are similar. Therefore, an obvious zonal distribution for the outburst coal appears.

As for the previous analysis, the pressure difference between the two sides of the coal layer is a key factor for the acceleration motion of the layer, and this factor changes with the development of an outburst. In the outburst growth period and stable period, the pressure difference is relatively stable. As the outburst develops layer by layer, the flow velocity and mass of outburst coal increase gradually. Finally, a large amount of coal is concentrated in an area far from the outburst mouth.

In the outburst decay period, the gas supply is limited and the pressure difference decrease rapidly. The flow velocity and mass of the outburst coal, which can only accumulate in the area near the outburst mouth, are significantly reduced. For the outburst experiment, the decay period is extremely short approximately 10 percent of the total time; the mass ratios of areas IV and V are more than 60 percent.

Figure 10. Ejection velocity of the coal with the gas pressure.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Table 6. Flow velocity of outburst coal.

Experiment No. Parameter

4

5

6

Gas pressure (MPa) Flow velocity (m/s)

0.40 5.1–7.8

0.45 6.4–9.4

0.50 8.6–12.1

apparatus to test the failure process of a cylinder coal sample driven by CO2 and obtained the flow velocity of a coal-gas mixture under different gas pressures, and the ejection velocity of the coal with the gas pressure is shown in Figure 10. The flow velocity of ejecting coal increases as the gas pressure increases, and the flow velocity can reach 20–60 m/s when the gas pressure is 4–10 MPa. Based on the data from the outburst experiment and using Eq. 10, the flow velocity of each layer is calculated, and the flow velocity range of the different experiments is recorded in Table 6. For the outburst experiment, when the maximum flow distance of the outburst coal is less than 20 m, the flow velocity for the outburst coal is approximately 10 m/s. As the gas pressure increases, the maximum flow distance and flow velocity for outburst coal increase correspondingly. However, for many outburst cases, the maximum flow distance is tens to hundreds of meters, and the ejection velocity may reach tens of meters per second.

Distribution Features of Outburst Coal According to the scene investigation at the Xinxing coal mine outburst, there was obvious zonal distribu-

tion for the outburst coal (rock) in the exploring roadway, and the coal was concentrated in an area far from the outburst mouth (Figure 6). A similar phenomenon was found in outburst experiments, with the outburst coal concentrated in areas IV and V, but the mass ratio of area I was less than 7 percent.

If we use the layered failure of outburst coal to explain the distribution features, a reasonable answer can be achieved. In the process of outburst, the coal is broken into layers with a certain thickness. For the same layer, after undergoing acceleration, the flow velocities of the different coal bodies are similar, and the flow distances for those coal bodies are similar. Therefore, an obvious zonal distribution for the outburst coal appears.

As for the previous analysis, the pressure difference between the two sides of the coal layer is a key factor for the acceleration motion of the layer, and this factor changes with the development of an outburst. In the outburst growth period and stable period, the pressure difference is relatively stable. As the outburst develops layer by layer, the flow velocity and mass of outburst coal increase gradually. Finally, a large amount of coal is concentrated in an area far from the outburst mouth.

In the outburst decay period, the gas supply is limited and the pressure difference decrease rapidly. The flow velocity and mass of the outburst coal, which can only accumulate in the area near the outburst mouth, are significantly reduced. For the outburst experiment, the decay period is extremely short approximately 10 percent of the total time; the mass ratios of areas IV and V are more than 60 percent.

Figure 10. Ejection velocity of the coal with the gas pressure.

329

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

329

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

CONCLUSIONS In this work, two outburst cases are analyzed and six outburst experiments are conducted by the authors.

Based on this, the evolution process of outbursts and the failure types of coal are discussed in detail, and the law of pressure difference changes is obtained. Then, the flow process of outburst coal is studied deeply. The major conclusions are summarized as follows: 1. The whole process of an outburst generally involves four stages (preparation, trigger, development, and termination) and includes the time and spatial dimensions. There are two types of coal failure: shear failure and tensile failure. The coal ultimately fails in the form of multiple similar layers.

2. Based on the law of pressure difference change, a pressure difference coefficient, i , is introduced to characterize the change of the pressure difference. The flow process of outburst coal is divided into a quasi-static phase, a constant acceleration phase, a freefall phase, and a deceleration phase.

Accordingly, a simplified model of coal flow is established. Using the data from the outburst experiments, the mode is calculated, and the theoretical values are in good agreement with the experimental values.

3. Based on the layered failure of coal and using the simplified model, the flow velocity of outburst coal is obtained. For the outburst experiments, the flow velocity is approximately 10 m/s. However, for many outburst cases, the flow velocity may reach tens of meters per second.

4. The distribution features of ejecting coal in both the experiments and outburst cases are explained as well. In the outburst growth period and stable period, the flow velocity and mass of each layer increase gradually, causing a large amount of coal that is concentrated in the area far from the outburst mouth. For the same layer, the flow distances for different coal bodies are similar, and an obvious zonal distribution appears for the outburst coal.

ACKNOWLEDGMENT The authors are grateful for the support from the Fundamental Research Funds for the Central Universities (No. 2017BSCXB14).

REFERENCES ALONSO, E.; ALEJANO, L. R.; VARAS, F.; FDEZ-MANIN, G.; AND CARRANZA-TORRES, C., 2003, Ground response curves for rock masses exhibiting strain—Softening behaviour: International Journal Numerical Analytical Methods Geomechanics, Vol. 27, pp. 1153–1185.

330

CHEN, K. P., 2011, A new mechanistic model for prediction of instantaneous coal outbursts—Dedicated to the memory of Prof. Daniel D. Joseph: International Journal Coal Geology, Vol. 87, pp. 72–79.

CHEN, P.; WANG, E.; OU, J.; LI, Z.; WEI, M.; AND LI, X., 2013, Fractal characteristics of surface crack evolution in the process of gas-containing coal extrusion: International Journal Mining Science Technology, Vol. 23, pp. 121–126.

CHENG, Y., 2010, Theories and Engineering Applications on Coal Mine Gas Control: China University of Mining and Technology Press, Xuzhou, China.

CHOI, X. AND WOLD, M. B., 2004, Study of the Mechanisms of Coal and Gas Outbursts Using a New Numerical Modeling Approach. 2004 Coal Operators Conference, University of Wollongong Research Online, pp. 181–194.

FAN, C.; LI, S.; LUO, M.; DU, W.; AND YANG, Z., 2016, Coal and gas outburst dynamic system: International Journal Mining Science Technology, Vol. 27, pp. 49–55.

GENG, J.; XU, J.; NIE, W.; PENG, S.; ZHANG, C.; AND LUO, X., 2017, Regression analysis of major parameters affecting the intensity of coal and gas outbursts in laboratory: International Journal Mining Science Technology. Vol. 27. pp. 327–332.

GUAN, P.; WANG, H.; AND ZHANG, Y., 2009, Mechanism of instantaneous coal outbursts: Geology, Vol. 37, pp. 915–918.

GUO, P. K., 2014, Research on Laminar Spallation Mechanism of Coal and Gas Outburst Propagation: China University of Mining and Technology, Xuzhou, China.

HODOT, B. B., 1966, Outburst of Coal and Coalbed Gas [Chinese Translation]: China Coal Industry Press, Beijing, China.

HU, Q.; ZHANG, S.; WEN, G.; DAI, L.; AND WANG, B., 2015, Coallike material for coal and gas outburst simulation tests: International Journal Rock Mechanics Mining Sciences, Vol. 74, pp.

151–156.

JACKOWSKI, T. AND WOŹNIAKOWSKI, H., 1987, Complexity of approximation with relative error criterion in worst, average, and probabilistic settings: Journal Complexity, Vol. 3, pp. 114–135.

JIANG, C. AND YU, Q., 1994, Analyses on the developing process and mechanical conditions of coal gas outburst front: Journal China University Mining Technology, Vol. 23, pp. 1–9.

JIN, H.; HU, Q.; AND LIU, Y., 2011, Failure mechanism of coal and gas outburst initiation: First International Symposium on Mine Safety Science and Engineering, Procedia Engineering. Vol. 26.

Elsevier Ltd. pp. 1352–1360.

KANG, H.; ZHANG, X.; SI, L.; WU, Y.; AND GAO, F., 2010, In-situ stress measurements and stress distribution characteristics in underground coal mines in China: Engineering Geology, Vol. 116, pp. 333–345.

LAMA, R. D. AND BODZIONY, J., 1998, Management of outburst in underground coal mines: International Journal Coal Geology, Vol. 35, pp. 83–115.

PAN, Y. AND LI, A., 2013, Rapid iterative incremental model of the intermittent chaos of deep hole developing in coal-gas outburst: International Journal Mining Science Technology, Vol. 23, pp. 287–292.

PATERSON, L., 1986, A model for outburst in coal: International Journal Rock Mechanics Mining Science Geomechanics Abstracts, Vol. 23, pp. 327–332.

SONG, Y. AND CHENG, G., 2012, The mechanism and numerical experiment of spalling phenomena in one-dimensional coal and gas outburst: Procedia Environmental Sciences, Vol. 12, pp. 885–890.

SUN, D.; HU, Q.; AND MIAO, F., 2011, A mathematical model of coal-gas flow conveying in the process of coal and gas outburst and its application: Procedia Engineering, Vol. 26, pp. 147–153.

TU, Q.; CHENG, Y.; GUO, P.; JIANG, J.; WANG, L.; AND ZHANG, R., 2016, Experimental study of coal and gas outbursts related

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Tu, Cheng, Liu, Guo, Wang, Zhao, Li, and Dong

CONCLUSIONS In this work, two outburst cases are analyzed and six outburst experiments are conducted by the authors.

Based on this, the evolution process of outbursts and the failure types of coal are discussed in detail, and the law of pressure difference changes is obtained. Then, the flow process of outburst coal is studied deeply. The major conclusions are summarized as follows: 1. The whole process of an outburst generally involves four stages (preparation, trigger, development, and termination) and includes the time and spatial dimensions. There are two types of coal failure: shear failure and tensile failure. The coal ultimately fails in the form of multiple similar layers.

2. Based on the law of pressure difference change, a pressure difference coefficient, i , is introduced to characterize the change of the pressure difference. The flow process of outburst coal is divided into a quasi-static phase, a constant acceleration phase, a freefall phase, and a deceleration phase.

Accordingly, a simplified model of coal flow is established. Using the data from the outburst experiments, the mode is calculated, and the theoretical values are in good agreement with the experimental values.

3. Based on the layered failure of coal and using the simplified model, the flow velocity of outburst coal is obtained. For the outburst experiments, the flow velocity is approximately 10 m/s. However, for many outburst cases, the flow velocity may reach tens of meters per second.

4. The distribution features of ejecting coal in both the experiments and outburst cases are explained as well. In the outburst growth period and stable period, the flow velocity and mass of each layer increase gradually, causing a large amount of coal that is concentrated in the area far from the outburst mouth. For the same layer, the flow distances for different coal bodies are similar, and an obvious zonal distribution appears for the outburst coal.

ACKNOWLEDGMENT The authors are grateful for the support from the Fundamental Research Funds for the Central Universities (No. 2017BSCXB14).

REFERENCES ALONSO, E.; ALEJANO, L. R.; VARAS, F.; FDEZ-MANIN, G.; AND CARRANZA-TORRES, C., 2003, Ground response curves for rock masses exhibiting strain—Softening behaviour: International Journal Numerical Analytical Methods Geomechanics, Vol. 27, pp. 1153–1185.

330

CHEN, K. P., 2011, A new mechanistic model for prediction of instantaneous coal outbursts—Dedicated to the memory of Prof. Daniel D. Joseph: International Journal Coal Geology, Vol. 87, pp. 72–79.

CHEN, P.; WANG, E.; OU, J.; LI, Z.; WEI, M.; AND LI, X., 2013, Fractal characteristics of surface crack evolution in the process of gas-containing coal extrusion: International Journal Mining Science Technology, Vol. 23, pp. 121–126.

CHENG, Y., 2010, Theories and Engineering Applications on Coal Mine Gas Control: China University of Mining and Technology Press, Xuzhou, China.

CHOI, X. AND WOLD, M. B., 2004, Study of the Mechanisms of Coal and Gas Outbursts Using a New Numerical Modeling Approach. 2004 Coal Operators Conference, University of Wollongong Research Online, pp. 181–194.

FAN, C.; LI, S.; LUO, M.; DU, W.; AND YANG, Z., 2016, Coal and gas outburst dynamic system: International Journal Mining Science Technology, Vol. 27, pp. 49–55.

GENG, J.; XU, J.; NIE, W.; PENG, S.; ZHANG, C.; AND LUO, X., 2017, Regression analysis of major parameters affecting the intensity of coal and gas outbursts in laboratory: International Journal Mining Science Technology. Vol. 27. pp. 327–332.

GUAN, P.; WANG, H.; AND ZHANG, Y., 2009, Mechanism of instantaneous coal outbursts: Geology, Vol. 37, pp. 915–918.

GUO, P. K., 2014, Research on Laminar Spallation Mechanism of Coal and Gas Outburst Propagation: China University of Mining and Technology, Xuzhou, China.

HODOT, B. B., 1966, Outburst of Coal and Coalbed Gas [Chinese Translation]: China Coal Industry Press, Beijing, China.

HU, Q.; ZHANG, S.; WEN, G.; DAI, L.; AND WANG, B., 2015, Coallike material for coal and gas outburst simulation tests: International Journal Rock Mechanics Mining Sciences, Vol. 74, pp.

151–156.

JACKOWSKI, T. AND WOŹNIAKOWSKI, H., 1987, Complexity of approximation with relative error criterion in worst, average, and probabilistic settings: Journal Complexity, Vol. 3, pp. 114–135.

JIANG, C. AND YU, Q., 1994, Analyses on the developing process and mechanical conditions of coal gas outburst front: Journal China University Mining Technology, Vol. 23, pp. 1–9.

JIN, H.; HU, Q.; AND LIU, Y., 2011, Failure mechanism of coal and gas outburst initiation: First International Symposium on Mine Safety Science and Engineering, Procedia Engineering. Vol. 26.

Elsevier Ltd. pp. 1352–1360.

KANG, H.; ZHANG, X.; SI, L.; WU, Y.; AND GAO, F., 2010, In-situ stress measurements and stress distribution characteristics in underground coal mines in China: Engineering Geology, Vol. 116, pp. 333–345.

LAMA, R. D. AND BODZIONY, J., 1998, Management of outburst in underground coal mines: International Journal Coal Geology, Vol. 35, pp. 83–115.

PAN, Y. AND LI, A., 2013, Rapid iterative incremental model of the intermittent chaos of deep hole developing in coal-gas outburst: International Journal Mining Science Technology, Vol. 23, pp. 287–292.

PATERSON, L., 1986, A model for outburst in coal: International Journal Rock Mechanics Mining Science Geomechanics Abstracts, Vol. 23, pp. 327–332.

SONG, Y. AND CHENG, G., 2012, The mechanism and numerical experiment of spalling phenomena in one-dimensional coal and gas outburst: Procedia Environmental Sciences, Vol. 12, pp. 885–890.

SUN, D.; HU, Q.; AND MIAO, F., 2011, A mathematical model of coal-gas flow conveying in the process of coal and gas outburst and its application: Procedia Engineering, Vol. 26, pp. 147–153.

TU, Q.; CHENG, Y.; GUO, P.; JIANG, J.; WANG, L.; AND ZHANG, R., 2016, Experimental study of coal and gas outbursts related

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

Analysis of the Layered Failure of Coal to gas-enriched areas: Rock Mechanics Rock Engineering, pp.

1–13. Vol. 49. pp. 3769–3781.

TU, Q.; CHENG, Y.; WANG, L.; AND ZHANG, R., 2015, Study on dynamic process of coal and gas outburst: Coal Science Technology, Vol. 43, pp. 71–75.

WANG, L.; CHENG, Y. P.; WANG, L.; GUO, P. K.; AND LI, W., 2012, Safety line method for the prediction of deep coal-seam gas pressure and its application in coal mines: Safety Science, Vol. 50, pp. 523–529.

WANG, S., 2012, Gas Transport, Sorption, and Mechanical Response of Fractured Coal: The Pennsylvania State University, University Park, PA. 193 p.

WEI, H.; CHEN, Z.; YUE, J.; YU, Z.; AND MIN, Y., 2010, Failure modes of coal containing gas and mechanism of gas outbursts: International Journal Mining Science Technology, Vol. 20, pp.

504–509.

Analysis of the Layered Failure of Coal

XU, T.; TANG, C. A.; YANG, T. H.; ZHU, W. C.; AND LIU, J., 2006, Numerical investigation of coal and gas outbursts in underground collieries: International Journal Rock Mechanics Mining Sciences, Vol. 43, pp. 905–919.

ZHAO, W.; CHENG, Y.; GUO, P.; JIN, K.; TU, Q.; AND WANG, H., 2016, An analysis of the gas-solid plug flow formation: New insights into the coal failure process during coal and gas outbursts: Powder Technology, Vol. 305, pp. 39–47.

ZHOU, A.; WANG, K.; WANG, L.; DU, F.; AND LI, Z., 2015, Numerical simulation for propagation characteristics of shock wave and gas flow induced by outburst intensity: International Journal Mining Science Technology, Vol. 25, pp. 107–112.

ZHOU, A.; WANG, K.; AND WU, Z., 2014, Propagation law of shock waves and gas flow in cross roadway caused by coal and gas outburst: International Journal Mining Science Technology, Vol. 24, pp. 23–29.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

331

to gas-enriched areas: Rock Mechanics Rock Engineering, pp.

1–13. Vol. 49. pp. 3769–3781.

TU, Q.; CHENG, Y.; WANG, L.; AND ZHANG, R., 2015, Study on dynamic process of coal and gas outburst: Coal Science Technology, Vol. 43, pp. 71–75.

WANG, L.; CHENG, Y. P.; WANG, L.; GUO, P. K.; AND LI, W., 2012, Safety line method for the prediction of deep coal-seam gas pressure and its application in coal mines: Safety Science, Vol. 50, pp. 523–529.

WANG, S., 2012, Gas Transport, Sorption, and Mechanical Response of Fractured Coal: The Pennsylvania State University, University Park, PA. 193 p.

WEI, H.; CHEN, Z.; YUE, J.; YU, Z.; AND MIN, Y., 2010, Failure modes of coal containing gas and mechanism of gas outbursts: International Journal Mining Science Technology, Vol. 20, pp.

504–509.

XU, T.; TANG, C. A.; YANG, T. H.; ZHU, W. C.; AND LIU, J., 2006, Numerical investigation of coal and gas outbursts in underground collieries: International Journal Rock Mechanics Mining Sciences, Vol. 43, pp. 905–919.

ZHAO, W.; CHENG, Y.; GUO, P.; JIN, K.; TU, Q.; AND WANG, H., 2016, An analysis of the gas-solid plug flow formation: New insights into the coal failure process during coal and gas outbursts: Powder Technology, Vol. 305, pp. 39–47.

ZHOU, A.; WANG, K.; WANG, L.; DU, F.; AND LI, Z., 2015, Numerical simulation for propagation characteristics of shock wave and gas flow induced by outburst intensity: International Journal Mining Science Technology, Vol. 25, pp. 107–112.

ZHOU, A.; WANG, K.; AND WU, Z., 2014, Propagation law of shock waves and gas flow in cross roadway caused by coal and gas outburst: International Journal Mining Science Technology, Vol. 24, pp. 23–29.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 317–331

331

Effect of Oil-Degrading Bacteria on Geotechnical Properties of Crude Oil–Contaminated Sand

Effect of Oil-Degrading Bacteria on Geotechnical Properties of Crude Oil–Contaminated Sand

HOSSEIN SOLTANI-JIGHEH1 HAMED VAFAEI MOLAMAHMOOD

HOSSEIN SOLTANI-JIGHEH1 HAMED VAFAEI MOLAMAHMOOD

Department of Civil Engineering, Azarbaijan Shahid Madani University, Tabriz, Iran

Department of Civil Engineering, Azarbaijan Shahid Madani University, Tabriz, Iran

TAGHI EBADI

TAGHI EBADI

Faculty of Civil and Environmental Engineering, Amirkabir University of Technology, Tehran, Iran

Faculty of Civil and Environmental Engineering, Amirkabir University of Technology, Tehran, Iran

ALI ABOLHASANI SOORKI

ALI ABOLHASANI SOORKI

ACECR-Research Institute of Applied Sciences, Tehran, Iran

ACECR-Research Institute of Applied Sciences, Tehran, Iran

Key Terms: Geotechnical Properties, Soil Stabilization, Bio-Treatment, Crude Oil Contamination, Silty Sand ABSTRACT Various materials have conventionally been utilized to improve the geotechnical properties of soils. Recently, utilization of microorganisms to this end has attracted much attention because of its high capabilities. Various oil-degrading bacteria have proven to be effective in degrading crude oil pollution from the soil. However, their effect on geotechnical properties of the soil has not been thoroughly investigated. In this research, various oil-degrading bacteria were employed to degrade crude oil contamination from crude oil–contaminated sandy soil. Then alterations on geotechnical properties of biotreated sand were ascertained. The results showed that by employing these microorganisms, more than 50 and 80 percent of the crude oil was degraded after 10 and 30 days of bio-treatment, respectively. Moreover, biotreated samples, which are slightly crude oil contaminated, acquire larger values of maximum dry density and cohesion in comparison to untreated contaminated samples. As treatment time increases, the value of the internal friction angle remains almost constant, and unconfined compression strength slightly increases. Generally, bio-treatment alters the geotechnical properties of the contaminated soil, and the application of bio-treated soil as a road basement in road construction and erosion controller is suggested.

1

Corresponding author email: hsoltani@azaruniv.ac.ir

INTRODUCTION In January 1991, a massive oil spill in Kuwait attracted the attention of many people around the world.

About 92 percent of the producing oil wells were detonated and ignited, causing severe environmental and engineering impacts (Mukhopadhyay et al., 2008; Samani et al., 2016). In addition to the environmental threats, the geotechnical properties of the contaminated soil were altered and caused enormous damage to the existing structures. All of this attracted the attention of many researchers to work intensively on the influence of crude oil contamination on engineering properties of the soils.

Several studies have covered the alteration of geotechnical properties in petroleum-contaminated soils. Evgin and Das (1992) studied the shear behavior of crude oil–contaminated sand by employing a triaxial test. Al-Sanad et al. (1995) and Al-Sanad and Ismael (1997) published their findings on the effect of crude oil contamination and aging impacts on Kuwaiti sand.

In similar studies, Shin et al. (1999), Puri (2000), Shin and Das (2001), Khamehchiyan et al. (2007), and Nasr (2009) studied the impacts of crude oil contamination on sandy soils. In addition, Elisha (2012), Kermani and Ebadi (2012), Khosravi et al. (2013), Obeta and EzeUzomaka (2013), Ota (2013), Estabragh et al. (2016), and Nasehi et al. (2016) investigated the alteration on geotechnical properties of clayey soil due to petroleum contamination. The results of these researchers vary with each other; however, these statements could be mentioned as dominant impacts of crude oil contamination in soil. The values of optimum water content, permeability, unconfined compression strength, internal friction angle, and cohesion decrease, and the value of maximum dry density increases as crude oil contamination increase, in all soil types. However, there is also

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

333

Key Terms: Geotechnical Properties, Soil Stabilization, Bio-Treatment, Crude Oil Contamination, Silty Sand ABSTRACT Various materials have conventionally been utilized to improve the geotechnical properties of soils. Recently, utilization of microorganisms to this end has attracted much attention because of its high capabilities. Various oil-degrading bacteria have proven to be effective in degrading crude oil pollution from the soil. However, their effect on geotechnical properties of the soil has not been thoroughly investigated. In this research, various oil-degrading bacteria were employed to degrade crude oil contamination from crude oil–contaminated sandy soil. Then alterations on geotechnical properties of biotreated sand were ascertained. The results showed that by employing these microorganisms, more than 50 and 80 percent of the crude oil was degraded after 10 and 30 days of bio-treatment, respectively. Moreover, biotreated samples, which are slightly crude oil contaminated, acquire larger values of maximum dry density and cohesion in comparison to untreated contaminated samples. As treatment time increases, the value of the internal friction angle remains almost constant, and unconfined compression strength slightly increases. Generally, bio-treatment alters the geotechnical properties of the contaminated soil, and the application of bio-treated soil as a road basement in road construction and erosion controller is suggested.

1

Corresponding author email: hsoltani@azaruniv.ac.ir

INTRODUCTION In January 1991, a massive oil spill in Kuwait attracted the attention of many people around the world.

About 92 percent of the producing oil wells were detonated and ignited, causing severe environmental and engineering impacts (Mukhopadhyay et al., 2008; Samani et al., 2016). In addition to the environmental threats, the geotechnical properties of the contaminated soil were altered and caused enormous damage to the existing structures. All of this attracted the attention of many researchers to work intensively on the influence of crude oil contamination on engineering properties of the soils.

Several studies have covered the alteration of geotechnical properties in petroleum-contaminated soils. Evgin and Das (1992) studied the shear behavior of crude oil–contaminated sand by employing a triaxial test. Al-Sanad et al. (1995) and Al-Sanad and Ismael (1997) published their findings on the effect of crude oil contamination and aging impacts on Kuwaiti sand.

In similar studies, Shin et al. (1999), Puri (2000), Shin and Das (2001), Khamehchiyan et al. (2007), and Nasr (2009) studied the impacts of crude oil contamination on sandy soils. In addition, Elisha (2012), Kermani and Ebadi (2012), Khosravi et al. (2013), Obeta and EzeUzomaka (2013), Ota (2013), Estabragh et al. (2016), and Nasehi et al. (2016) investigated the alteration on geotechnical properties of clayey soil due to petroleum contamination. The results of these researchers vary with each other; however, these statements could be mentioned as dominant impacts of crude oil contamination in soil. The values of optimum water content, permeability, unconfined compression strength, internal friction angle, and cohesion decrease, and the value of maximum dry density increases as crude oil contamination increase, in all soil types. However, there is also

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

333

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

Figure 1. Location of soil samples.

Figure 1. Location of soil samples.

an exception to these results. For example, Khamehchiyan et al. (2007) and Kermani and Ebadi (2012) showed that with an increase in crude oil content, the internal friction angle increases in clayey soils, and maximum dry density decreases in silty sand soil.

Furthermore, some scientists have investigated novel techniques for improving the engineering properties of crude oil–contaminated soil. The impact of various materials, such as cement and cement bypass dust (Al-Rawas et al., 2005), nanoscale zero-valent iron and nanoscale hydrated lime (Nasehi et al., 2016) and incinerator ash (Mohamedzein et al., 2006), on geotechnical properties has been investigated previously. Recently, great interest has been devoted to the utilization of microorganisms to stabilize the soil. The techniques of microbial-induced calcite precipitation (Soon et al., 2013), bio-cementation (Soon et al., 2014) and bio-clogging (Yasuhara et al., 2012) are some of the methods that rely on microorganisms to enhance the engineering properties of the soils.

The previously mentioned literature indicates that there have been outstanding investigations on the impact of crude oil contamination on geotechnical properties of contaminated soil as well as some research on the improvement of the geotechnical properties of crude oil–contaminated soil. To the best knowledge of the authors, no significant study has investigated the effect of any oil-degrading bacteria on the geotechnical properties of crude oil–contaminated soil. Since degradation of crude oil is carried out via these bacteria, an appropriate evaluation of their impact on geotechnical properties is essential. Thus, the present study attempts to investigate this critical issue and struggles to find novel applications for bio-treated soil in the field of civil engineering.

MATERIALS AND METHODS The soil used throughout this study was a noncontaminated soil collected from the vicinity of crude 334

oil storage tanks (45◦ 15� N, 38◦ 50� E) in the Tabriz oil refinery site (Figure 1). Soil samples were taken from 40 to 50 centimeters below the ground surface to prevent entering the upper organic soil. They were dried by an oven at 105◦ C for 24 hours and then kept in plastic containers at room temperature for further experiments. Basic engineering properties of the soil are listed in Table 1, and the particle size distribution curve is displayed in Figure 2. The soil is classified as nonplastic silty sand based on the Unified Soil Classification System (D422, ASTM, 2007).

Crude oil was provided by the Tabriz oil refinery.

It was selected because it is known to be present at many contaminated sites and to have a high frequency of occurrence in soil (Mohammadi and Moharamzade, 2014). Table 2 summarizes the most important properties of the crude oil used in this study.

Oil-degrading bacteria that were isolated from oilcontaminated sites in southern areas of Iran were provided by the Research Institute of Applied Sciences, Tehran, Iran. The bacterial strains were introduced into the soil. Table 3 demonstrates the scientific names and the information regarding these microorganisms.

Mineral nutrients for enhancing bacterial growth were monopotassium phosphate (KH2 PO4 ) and ammonium chloride (NH4 Cl), which were employed as

Table 1. Some basic properties of soil.

Parameter Silt and clay Sand Gravel Specific gravity Liquid limit Plastic limit Water content Organic materials content pH (in 27◦ C)

Description 26% 62% 12% 2.47 NP NP 2.84% 3.38% 6.9

NP = non-plastic.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Standard Designation D422 (ASTM, 2007) D854 (ASTM, 2016) D4318 (ASTM, 2016) D2216 (ASTM, 2016) D2974 (ASTM, 2016) D4972 (ASTM, 2016)

an exception to these results. For example, Khamehchiyan et al. (2007) and Kermani and Ebadi (2012) showed that with an increase in crude oil content, the internal friction angle increases in clayey soils, and maximum dry density decreases in silty sand soil.

Furthermore, some scientists have investigated novel techniques for improving the engineering properties of crude oil–contaminated soil. The impact of various materials, such as cement and cement bypass dust (Al-Rawas et al., 2005), nanoscale zero-valent iron and nanoscale hydrated lime (Nasehi et al., 2016) and incinerator ash (Mohamedzein et al., 2006), on geotechnical properties has been investigated previously. Recently, great interest has been devoted to the utilization of microorganisms to stabilize the soil. The techniques of microbial-induced calcite precipitation (Soon et al., 2013), bio-cementation (Soon et al., 2014) and bio-clogging (Yasuhara et al., 2012) are some of the methods that rely on microorganisms to enhance the engineering properties of the soils.

The previously mentioned literature indicates that there have been outstanding investigations on the impact of crude oil contamination on geotechnical properties of contaminated soil as well as some research on the improvement of the geotechnical properties of crude oil–contaminated soil. To the best knowledge of the authors, no significant study has investigated the effect of any oil-degrading bacteria on the geotechnical properties of crude oil–contaminated soil. Since degradation of crude oil is carried out via these bacteria, an appropriate evaluation of their impact on geotechnical properties is essential. Thus, the present study attempts to investigate this critical issue and struggles to find novel applications for bio-treated soil in the field of civil engineering.

MATERIALS AND METHODS The soil used throughout this study was a noncontaminated soil collected from the vicinity of crude 334

oil storage tanks (45◦ 15� N, 38◦ 50� E) in the Tabriz oil refinery site (Figure 1). Soil samples were taken from 40 to 50 centimeters below the ground surface to prevent entering the upper organic soil. They were dried by an oven at 105◦ C for 24 hours and then kept in plastic containers at room temperature for further experiments. Basic engineering properties of the soil are listed in Table 1, and the particle size distribution curve is displayed in Figure 2. The soil is classified as nonplastic silty sand based on the Unified Soil Classification System (D422, ASTM, 2007).

Crude oil was provided by the Tabriz oil refinery.

It was selected because it is known to be present at many contaminated sites and to have a high frequency of occurrence in soil (Mohammadi and Moharamzade, 2014). Table 2 summarizes the most important properties of the crude oil used in this study.

Oil-degrading bacteria that were isolated from oilcontaminated sites in southern areas of Iran were provided by the Research Institute of Applied Sciences, Tehran, Iran. The bacterial strains were introduced into the soil. Table 3 demonstrates the scientific names and the information regarding these microorganisms.

Mineral nutrients for enhancing bacterial growth were monopotassium phosphate (KH2 PO4 ) and ammonium chloride (NH4 Cl), which were employed as

Table 1. Some basic properties of soil.

Parameter Silt and clay Sand Gravel Specific gravity Liquid limit Plastic limit Water content Organic materials content pH (in 27◦ C)

Description 26% 62% 12% 2.47 NP NP 2.84% 3.38% 6.9

NP = non-plastic.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Standard Designation D422 (ASTM, 2007) D854 (ASTM, 2016) D4318 (ASTM, 2016) D2216 (ASTM, 2016) D2974 (ASTM, 2016) D4972 (ASTM, 2016)

Effect of bacteria on contaminated sand

Effect of bacteria on contaminated sand

Table 3. Details of employed bacterial strains.

Serial Number

Microorganism Rhodococcus ruber strain KE1 Microbacterium sp. strain K-2-4 Gracilibacillus dipsosauri strain MK-1 Thalassospira xianhensis strain PM01 Vibrio sp. strain PM-06

Figure 2. Particle size distribution curve of the soil.

phosphorous and nitrogen sources, respectively. The optimum conditions for petroleum bio-degradation by the microbial consortium were provided according to the given instructions from the Research Institute of Applied Sciences, which is, briefly, pH: 6.5–9.5; temperature: 20◦ C–50◦ C; salinity: 0–80,000 ppm; relative humidity: 30–50 percent; and ventilation: normal.

The experimental program was divided into two separate stages. During the first stage, after soil classification, soil samples were sterilized by autoclaving for 15 minutes at 115◦ C, then the amounts of 3, 6, and 9 percent crude oil by dry weight of the soil specimen were sprayed and mixed manually with the soil. The mixtures were then put into plastic containers and kept in laboratory condition (20◦ C and pressure of 1.019 atmosphere) up to the test date. Required distilled water for each test was added to the sample 3 days before the test date, and the sample was put back into the plastic container to fully reach an equilibrium with the new condition. During the second stage, bio-degraded samples were prepared and tested. Five hundred milliliters of basic medium containing 1 × 107 colony-forming units of degrading bacteria per milliliter of the medium were sprayed over 1 kg of the crude oil–contaminated sample. Then the mineral nutrients by the amount of Table 2. Properties of crude oil.

Parameter

Value

Viscosity (gm s ) Density (g/cm3 at 25◦ C) API gravity (at 60◦ F) Flash point (◦ C) Specific gravity (at 25◦ C) −1 −1

API = American Petroleum Institute.

41.2 0.895 26.8 44.2 0.89

JQ963338.1 JQ963328.1 JQ963330.1 HM587995.1 JQ963335.1

Table 3. Details of employed bacterial strains.

Isolated Location Khoozestan province, Iran Kharg Island, Iran Khoozestan province, Iran Oil-Based Mud Unit of National Iranian South Oil Company Khoozestan province, Iran

1:10 of soil’s crude oil content were dissolved in distilled water, sprayed, and blended manually with the mixture.

During treatment, the temperature was kept in 20◦ C– 25◦ C, and the pressure was 1.019 atmosphere. The values of relative humidity, pH, and salinity were also controlled and the ventilation was provided by mixing samples every 2 days. In addition, non-bio-degraded control samples were also maintained for 30 days in the laboratory.

In order to fully determine the geotechnical properties of uncontaminated, crude oil–contaminated and treated samples, compaction (D698, ASTM, 2016), direct shear (D3080, ASTM, 2016), unconfined compression (D2166, ASTM, 2016), and consolidation (D2435, ASTM, 2016) tests were conducted on all sample types.

The scanning electron microscopic (SEM) analysis was performed for each sample (model VP LEO 1455 at Amirkabir University of Technology, Iran). The crude oil content was determined by performing gravimetric analysis according to Thouand et al. (1999), dividing the weight of residual crude oil (after evaporation of the solvent) by the weight of the wet sample. Water content in contaminated samples was calculated based on the equation developed by Zheng et al. (2014), presented as follows:

mt (1 + n − n ) − 1 − n × 100%, ww = mr where ww is water content, mt is the total weight of porous media before oven drying, mr is the residual weigh of the porous media after oven drying, n is the crude oil content for each sample, and is the oil drying loss coefficient, which was determined experimentally as 0.676 for the used crude oil.

Initial Observations Figure 3 presents the comparative observations made on uncontaminated, crude oil–contaminated,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Figure 2. Particle size distribution curve of the soil.

phosphorous and nitrogen sources, respectively. The optimum conditions for petroleum bio-degradation by the microbial consortium were provided according to the given instructions from the Research Institute of Applied Sciences, which is, briefly, pH: 6.5–9.5; temperature: 20◦ C–50◦ C; salinity: 0–80,000 ppm; relative humidity: 30–50 percent; and ventilation: normal.

The experimental program was divided into two separate stages. During the first stage, after soil classification, soil samples were sterilized by autoclaving for 15 minutes at 115◦ C, then the amounts of 3, 6, and 9 percent crude oil by dry weight of the soil specimen were sprayed and mixed manually with the soil. The mixtures were then put into plastic containers and kept in laboratory condition (20◦ C and pressure of 1.019 atmosphere) up to the test date. Required distilled water for each test was added to the sample 3 days before the test date, and the sample was put back into the plastic container to fully reach an equilibrium with the new condition. During the second stage, bio-degraded samples were prepared and tested. Five hundred milliliters of basic medium containing 1 × 107 colony-forming units of degrading bacteria per milliliter of the medium were sprayed over 1 kg of the crude oil–contaminated sample. Then the mineral nutrients by the amount of Table 2. Properties of crude oil.

Parameter

Value

Viscosity (gm s ) Density (g/cm3 at 25◦ C) API gravity (at 60◦ F) Flash point (◦ C) Specific gravity (at 25◦ C) −1 −1

RESULTS AND DISCUSSION

335

Serial Number

Microorganism

API = American Petroleum Institute.

41.2 0.895 26.8 44.2 0.89

Rhodococcus ruber strain KE1 Microbacterium sp. strain K-2-4 Gracilibacillus dipsosauri strain MK-1 Thalassospira xianhensis strain PM01

JQ963338.1

Vibrio sp. strain PM-06

JQ963335.1

JQ963328.1 JQ963330.1 HM587995.1

Isolated Location Khoozestan province, Iran Kharg Island, Iran Khoozestan province, Iran Oil-Based Mud Unit of National Iranian South Oil Company Khoozestan province, Iran

1:10 of soil’s crude oil content were dissolved in distilled water, sprayed, and blended manually with the mixture.

During treatment, the temperature was kept in 20◦ C– 25◦ C, and the pressure was 1.019 atmosphere. The values of relative humidity, pH, and salinity were also controlled and the ventilation was provided by mixing samples every 2 days. In addition, non-bio-degraded control samples were also maintained for 30 days in the laboratory.

In order to fully determine the geotechnical properties of uncontaminated, crude oil–contaminated and treated samples, compaction (D698, ASTM, 2016), direct shear (D3080, ASTM, 2016), unconfined compression (D2166, ASTM, 2016), and consolidation (D2435, ASTM, 2016) tests were conducted on all sample types.

The scanning electron microscopic (SEM) analysis was performed for each sample (model VP LEO 1455 at Amirkabir University of Technology, Iran). The crude oil content was determined by performing gravimetric analysis according to Thouand et al. (1999), dividing the weight of residual crude oil (after evaporation of the solvent) by the weight of the wet sample. Water content in contaminated samples was calculated based on the equation developed by Zheng et al. (2014), presented as follows:

mt (1 + n − n ) − 1 − n × 100%, ww = mr where ww is water content, mt is the total weight of porous media before oven drying, mr is the residual weigh of the porous media after oven drying, n is the crude oil content for each sample, and is the oil drying loss coefficient, which was determined experimentally as 0.676 for the used crude oil.

RESULTS AND DISCUSSION Initial Observations Figure 3 presents the comparative observations made on uncontaminated, crude oil–contaminated,

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

335

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

cles but also fills pore spaces between the soil particles with microbial biomass. It is also evident that as biotreatment time increases, more microbial biomass is produced, and, consequently, less space is being left empty between soil particles (see Figure 4c and d).

Figure 3. Initial observations for (a) uncontaminated soil sample, (b) crude oil–contaminated soil after a 30 days, and (c) bio-treated sample after 30 days of treatment.

and bio-treated samples. In this picture, an increase in size of the particles is evident so that the biggest particle size is related to the bio-treated sample.

SEM Analysis In this research, the morphology of the samples was investigated using SEM instruments. Subsequent results are displayed in Figure 4. It can be inferred that the size of the soil particles increases due to the addition of crude oil (see Figure 4a and b). In addition, bio-treatment not only increases the size of the parti-

Crude Oil Degradation Figure 5 portrays crude oil concentration in biotreated samples versus time (day). It is understandable that as bio-treatment time increases, crude oil content decreases more markedly. For example, the crude oil content in bio-treated samples with an initial crude oil content of 90,000 mg/kg soil decreases to 53,700 and 28,800 mg/kg soil after 10 and 30 days of biotreatment, respectively. Since residual crude oil consists mostly of heavy components such as asphaltenes, it is not susceptible to leaching because of the low solubility of these components in water (Spiecker et al., 2003). Additionally, various studies have indicated that the presence of asphaltenes among soil particles increases sorption of organic compounds and does not influence water sorption (Pourmohammadbagher and

cles but also fills pore spaces between the soil particles with microbial biomass. It is also evident that as biotreatment time increases, more microbial biomass is produced, and, consequently, less space is being left empty between soil particles (see Figure 4c and d).

Figure 3. Initial observations for (a) uncontaminated soil sample, (b) crude oil–contaminated soil after a 30 days, and (c) bio-treated sample after 30 days of treatment.

and bio-treated samples. In this picture, an increase in size of the particles is evident so that the biggest particle size is related to the bio-treated sample.

SEM Analysis In this research, the morphology of the samples was investigated using SEM instruments. Subsequent results are displayed in Figure 4. It can be inferred that the size of the soil particles increases due to the addition of crude oil (see Figure 4a and b). In addition, bio-treatment not only increases the size of the parti-

Crude Oil Degradation Figure 5 portrays crude oil concentration in biotreated samples versus time (day). It is understandable that as bio-treatment time increases, crude oil content decreases more markedly. For example, the crude oil content in bio-treated samples with an initial crude oil content of 90,000 mg/kg soil decreases to 53,700 and 28,800 mg/kg soil after 10 and 30 days of biotreatment, respectively. Since residual crude oil consists mostly of heavy components such as asphaltenes, it is not susceptible to leaching because of the low solubility of these components in water (Spiecker et al., 2003). Additionally, various studies have indicated that the presence of asphaltenes among soil particles increases sorption of organic compounds and does not influence water sorption (Pourmohammadbagher and

Figure 4. SEM analysis on (a) uncontaminated soil sample, (b) 6 percent crude oil–contaminated samples after 30 days, (c) bio-treated sample with initial crude oil content of 6 percent after 10 days of treatment, and (d) bio-treated sample with initial crude oil content of 6 percent after 30 days of treatment.

Figure 4. SEM analysis on (a) uncontaminated soil sample, (b) 6 percent crude oil–contaminated samples after 30 days, (c) bio-treated sample with initial crude oil content of 6 percent after 10 days of treatment, and (d) bio-treated sample with initial crude oil content of 6 percent after 30 days of treatment.

336

336

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Effect of bacteria on contaminated sand

Figure 5. Crude oil concentration after 10 and 30 days of bio-treatment.

Shaw, 2016). In fact, the residual crude oil is trapped among soil particles due to the existence of various organic matter; consequently, leaching of the residual crude oil is very limited so that it does not pose any environmental threat.

Compaction Test Compaction characteristics are very important because of their importance in the road construction industry. Figure 6 displays compaction results in the form of dry density versus water content for crude oil– contaminated samples. In this figure, the shown numbers are defined as the crude oil content. Generally, a reduction in both maximum dry density and optimum water content is observed as crude oil content increases

Figure 6. Compaction curves for crude oil–contaminated samples after 30 days.

Effect of bacteria on contaminated sand

from 0 to 9 percent. However, the only exception was for the sample containing 3 percent crude oil, which showed an increase in maximum dry density in comparison to the natural soil. The reason of a decrease in maximum dry density could be attributed to the capillary action (Khamehchiyan et al., 2007). The capillarity tension is highly dependent on the angle of contact and the surface tension of the medium. Since crude oil is more hydrophobic than water, it prevents appropriate contact of the water with soil particles. Thus, with an increase in the crude oil content, the capillary tension decreases, resulting in lower values of maximum dry density for crude oil–contaminated samples.

Another reason could be the waste of compaction energy by crude oil. As the crude oil has a higher viscosity in comparison with water (40 times more than water), more compaction energy is consumed for increasing the tension between crude oil molecules. As a result, it absorbs more energy to rise in the soil texture. Furthermore, the reason of a decrease in optimum water content might be the presence of crude oil instead of water, which has the same effect as the water (Puri, 2000).

Figure 7 presents compaction curves for bio-treated samples. It can be understood that by an increase in the amounts of initial crude oil, the maximum dry density and optimum water content decrease. For example, for bio-treated samples after 10 days of treatment, the value of maximum dry density corresponding to the initial crude oil content of 3 percent is 19.12 kN/m3 , which is more than the values of uncontaminated soil and bio-treated samples with 6 percent and 9 initial crude oil content. However, an overall increase in the maximum dry density and a decrease in optimum water content by an increase in bio-treatment time is evident.

Although no accurate measurement is conducted, the reason for this trend might be in physical properties of microbial biomass. Since each bacterium in microbial biomass has a very small size, the production of a large amount of them fills pore spaces between the soil particles, which yields to a better compaction of bio-treated samples. However, since microbial biomass will not last for a long time, there might be some unexpected consequences if it is being used as a construction material.

Figure 5. Crude oil concentration after 10 and 30 days of bio-treatment.

Shaw, 2016). In fact, the residual crude oil is trapped among soil particles due to the existence of various organic matter; consequently, leaching of the residual crude oil is very limited so that it does not pose any environmental threat.

Compaction Test Compaction characteristics are very important because of their importance in the road construction industry. Figure 6 displays compaction results in the form of dry density versus water content for crude oil– contaminated samples. In this figure, the shown numbers are defined as the crude oil content. Generally, a reduction in both maximum dry density and optimum water content is observed as crude oil content increases

from 0 to 9 percent. However, the only exception was for the sample containing 3 percent crude oil, which showed an increase in maximum dry density in comparison to the natural soil. The reason of a decrease in maximum dry density could be attributed to the capillary action (Khamehchiyan et al., 2007). The capillarity tension is highly dependent on the angle of contact and the surface tension of the medium. Since crude oil is more hydrophobic than water, it prevents appropriate contact of the water with soil particles. Thus, with an increase in the crude oil content, the capillary tension decreases, resulting in lower values of maximum dry density for crude oil–contaminated samples.

Another reason could be the waste of compaction energy by crude oil. As the crude oil has a higher viscosity in comparison with water (40 times more than water), more compaction energy is consumed for increasing the tension between crude oil molecules. As a result, it absorbs more energy to rise in the soil texture. Furthermore, the reason of a decrease in optimum water content might be the presence of crude oil instead of water, which has the same effect as the water (Puri, 2000).

Figure 7 presents compaction curves for bio-treated samples. It can be understood that by an increase in the amounts of initial crude oil, the maximum dry density and optimum water content decrease. For example, for bio-treated samples after 10 days of treatment, the value of maximum dry density corresponding to the initial crude oil content of 3 percent is 19.12 kN/m3 , which is more than the values of uncontaminated soil and bio-treated samples with 6 percent and 9 initial crude oil content. However, an overall increase in the maximum dry density and a decrease in optimum water content by an increase in bio-treatment time is evident.

Although no accurate measurement is conducted, the reason for this trend might be in physical properties of microbial biomass. Since each bacterium in microbial biomass has a very small size, the production of a large amount of them fills pore spaces between the soil particles, which yields to a better compaction of bio-treated samples. However, since microbial biomass will not last for a long time, there might be some unexpected consequences if it is being used as a construction material.

Direct Shear

Direct Shear

One of the most essential properties of any type of soil is its shear characteristics. This property is important because it controls stability of the foundation system and bearing capacity of the soil. In this research, the samples of direct shear test were compacted to 0.95 times the value of maximum dry density with corresponding optimum water content for each sample. The

One of the most essential properties of any type of soil is its shear characteristics. This property is important because it controls stability of the foundation system and bearing capacity of the soil. In this research, the samples of direct shear test were compacted to 0.95 times the value of maximum dry density with corresponding optimum water content for each sample. The

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

337

Figure 6. Compaction curves for crude oil–contaminated samples after 30 days.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

337

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

Figure 7. Compaction curves for bio-treated samples with an initial crude oil content of (a) 3 percent, (b) 6 percent, and (c) 9 percent.

Figure 7. Compaction curves for bio-treated samples with an initial crude oil content of (a) 3 percent, (b) 6 percent, and (c) 9 percent.

values of maximum dry density and optimum water content were chosen because the soil particles for each sample have the most particle interaction with each other at their corresponding maximum dry density.

Figure 8 plots the effect of crude oil contamination and bio-treatment on internal friction angle and cohesion of the soil. In this figure, an increase in bio-treatment time is highlighted by using arrows. Generally, the values of internal friction angle and cohesion decrease for crude oil–contaminated samples by increasing the crude oil content. Shin et al. (2002) showed a reduction in internal friction angle due to oil contamination in sandy soils, and Ghaly (2001) reported a decrease in the values of internal friction angle with an increase in oil contamination level. The reason for such behavior could be found in viscosity differences of the crude oil and the water. As the viscosity of the pore fluid increases, the shear strength of the granular soil decreases (Ratnaweera and Meegoda, 2006). Moreover,

338

the lubricating effect of the crude oil causes a reduction in the inter-particle friction, decreasing the internal friction angle of the crude oil–contaminated soil (Khosravi et al., 2013).

Furthermore, a decrease in internal friction angle by an increase in initial crude oil content for bio-treated samples is evident. In these samples, the friction between particles is influenced by the presence of microbial biomass. Since microbial biomass acquires smaller size in comparison to the soil particles, they are placed between the soil particles. Thus, the surface interaction of soil particles with each other decreases. On the other hand, the bacterial colonies tend to slip over each other while a shear force is applied, and the ones that resist have a lower strength in comparison to the soil particles. As a result, the friction between soil particles decreases.

Additionally, the values of cohesion for bio-treated samples increase by an increase in the bio-treatment

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

values of maximum dry density and optimum water content were chosen because the soil particles for each sample have the most particle interaction with each other at their corresponding maximum dry density.

Figure 8 plots the effect of crude oil contamination and bio-treatment on internal friction angle and cohesion of the soil. In this figure, an increase in bio-treatment time is highlighted by using arrows. Generally, the values of internal friction angle and cohesion decrease for crude oil–contaminated samples by increasing the crude oil content. Shin et al. (2002) showed a reduction in internal friction angle due to oil contamination in sandy soils, and Ghaly (2001) reported a decrease in the values of internal friction angle with an increase in oil contamination level. The reason for such behavior could be found in viscosity differences of the crude oil and the water. As the viscosity of the pore fluid increases, the shear strength of the granular soil decreases (Ratnaweera and Meegoda, 2006). Moreover,

338

the lubricating effect of the crude oil causes a reduction in the inter-particle friction, decreasing the internal friction angle of the crude oil–contaminated soil (Khosravi et al., 2013).

Furthermore, a decrease in internal friction angle by an increase in initial crude oil content for bio-treated samples is evident. In these samples, the friction between particles is influenced by the presence of microbial biomass. Since microbial biomass acquires smaller size in comparison to the soil particles, they are placed between the soil particles. Thus, the surface interaction of soil particles with each other decreases. On the other hand, the bacterial colonies tend to slip over each other while a shear force is applied, and the ones that resist have a lower strength in comparison to the soil particles. As a result, the friction between soil particles decreases.

Additionally, the values of cohesion for bio-treated samples increase by an increase in the bio-treatment

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Effect of bacteria on contaminated sand

Effect of bacteria on contaminated sand

Figure 8. Effect of contamination and bio-treatment on (a) internal friction angle and (b) cohesion. Notations of BS-10 and BS-30 denote bio-treated samples after performing bio-treatment for 10 and 30 days, respectively. Additionally, CS-30 presents the results of crude oil– contaminated samples after 30 days.

Figure 8. Effect of contamination and bio-treatment on (a) internal friction angle and (b) cohesion. Notations of BS-10 and BS-30 denote bio-treated samples after performing bio-treatment for 10 and 30 days, respectively. Additionally, CS-30 presents the results of crude oil– contaminated samples after 30 days.

time. In these samples, microbial biomass fills the pore spaces between particles and increases the inter-particle surface area. This action, therefore, causes an increase in the cohesion values in bio-treated samples (Hemmat et al., 2010). In addition, the surface tension of pore fluid plays an important role in the cohesion of soil (Kemper and Rosenau, 1984). It could be deduced that the microbial biomass increases the surface tension of the pore fluid by producing bio-surfactants, increasing the cohesion of the soil.

time. In these samples, microbial biomass fills the pore spaces between particles and increases the inter-particle surface area. This action, therefore, causes an increase in the cohesion values in bio-treated samples (Hemmat et al., 2010). In addition, the surface tension of pore fluid plays an important role in the cohesion of soil (Kemper and Rosenau, 1984). It could be deduced that the microbial biomass increases the surface tension of the pore fluid by producing bio-surfactants, increasing the cohesion of the soil.

Unconfined Compression and Consolidation Tests The unconfined compression test was conducted on uncontaminated, crude oil–contaminated and biotreated samples by loading rate of 0.5 mm/min. All samples were prepared at 0.95 times their associated maximum dry density and corresponding optimum water content. The results are presented in Figure 9,

which indicates that the values of unconfined compression strength and failure strain for all samples are smaller and larger than the associated value of uncontaminated sample, respectively. In addition, in crude oil–contaminated samples, the general trend is to show a peak behavior for both unconfined compression strength and failure strain by an increase in the crude oil content. Additionally, for bio-treated samples, the unconfined compression strength decreases and the failure strain increases as the initial crude oil content increases. However, in these samples, both unconfined compression strength and failure strain increase by increasing the time of bio-treatment.

Consolidation is another important property that was studied in this research. The results (Figure 10) indicate that most of the experienced settlements for all samples are immediate settlements, which are very small. Generally, the settlement increases as the crude oil content increases, and the values of settlement in

Figure 9. Effect of contamination and bio-treatment on the values of (a) unconfined compression strength and (b) failure strain. Notations of BS-10 and BS-30 denote bio-treated samples after performing bio-treatment for 10 and 30 days, respectively. Additionally, CS-30 presents the results of crude oil–contaminated samples after 30 days.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

339

Unconfined Compression and Consolidation Tests The unconfined compression test was conducted on uncontaminated, crude oil–contaminated and biotreated samples by loading rate of 0.5 mm/min. All samples were prepared at 0.95 times their associated maximum dry density and corresponding optimum water content. The results are presented in Figure 9,

which indicates that the values of unconfined compression strength and failure strain for all samples are smaller and larger than the associated value of uncontaminated sample, respectively. In addition, in crude oil–contaminated samples, the general trend is to show a peak behavior for both unconfined compression strength and failure strain by an increase in the crude oil content. Additionally, for bio-treated samples, the unconfined compression strength decreases and the failure strain increases as the initial crude oil content increases. However, in these samples, both unconfined compression strength and failure strain increase by increasing the time of bio-treatment.

Consolidation is another important property that was studied in this research. The results (Figure 10) indicate that most of the experienced settlements for all samples are immediate settlements, which are very small. Generally, the settlement increases as the crude oil content increases, and the values of settlement in

Figure 9. Effect of contamination and bio-treatment on the values of (a) unconfined compression strength and (b) failure strain. Notations of BS-10 and BS-30 denote bio-treated samples after performing bio-treatment for 10 and 30 days, respectively. Additionally, CS-30 presents the results of crude oil–contaminated samples after 30 days.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

339

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

Figure 10. Effect of contamination and bio-treatment on final settlement. Notations of BS-10 and BS-30 denote bio-treated samples after performing bio-treatment for 10 and 30 days, respectively.

Additionally, CS-30 presents the results of crude oil–contaminated samples after 30 days.

crude oil–contaminated samples are larger than the corresponding values in bio-treated samples. Moreover, it can be seen that the values of settlement for all bio-treated samples are less than those for uncontaminated samples.

Potential Applications By determining some of the geotechnical properties of contaminated and bio-treated soils and comparing with the corresponding values of uncontaminated soil, several applications could be proposed. An increase in maximum dry density and reduction in optimum moisture content was observed by performing bio-treatment; therefore, bio-treated soil is suitable for road construction. It has the potential to be stabilized with cement and mixed with crushed stone aggregate as a substitute in hot-mix asphalt concrete in road construction. Moreover, results showed that the values of the cohesion increase as bio-treatment time increases. It is, therefore, recommended to use bio-treated soil as an erosion controller. Generally, the soils with small particle sizes are susceptible to erosion because of their low resistance to wind and water. However, bio-treatment increased the size of the soil particles and their cohesion, improving the stability of soil against erosion.

This application of bio-treated soil is also approved by the crude observations and the SEM analysis. Furthermore, the reduction in the values of internal friction angel indicated that the use of bio-treated soils as a 340

Soltani-Jigheh, Molamahmood, Ebadi, and Soorki

foundation should be limited to the lightweight structures with shallow foundations.

foundation should be limited to the lightweight structures with shallow foundations.

CONCLUSION

CONCLUSION

An extensive laboratory program was conducted to investigate the effect of bio-treatment on crude oil– contaminated soil. The following conclusions are based on the tested materials and experiments:

An extensive laboratory program was conducted to investigate the effect of bio-treatment on crude oil– contaminated soil. The following conclusions are based on the tested materials and experiments:

• Maximum dry density and optimum water content decrease for crude oil–contaminated samples by an increase in crude oil content. However, in bio-treated samples, the value of maximum dry density increases and the optimum water content decreases for the same range of alterations.

• The values of cohesion and internal friction angle decreased as the crude oil content increased in contaminated samples. However, for bio-treated samples, the values of cohesion are slightly more than corresponding values of contaminated samples and are somewhat lower than the value of uncontaminated soil.

• In general, oil contamination reduces the unconfined compression strength of the uncontaminated soil.

However, by an increase in bio-treatment time, unconfined compression strength increases, indicating a better performance of bio-treated soil in comparison with contaminated soil.

• The usage of bio-treated soil is recommended in road construction and erosion control projects. In addition, the construction of lightweight foundations on top of these soils does not pose any significant problem.

• Maximum dry density and optimum water content decrease for crude oil–contaminated samples by an increase in crude oil content. However, in bio-treated samples, the value of maximum dry density increases and the optimum water content decreases for the same range of alterations.

• The values of cohesion and internal friction angle decreased as the crude oil content increased in contaminated samples. However, for bio-treated samples, the values of cohesion are slightly more than corresponding values of contaminated samples and are somewhat lower than the value of uncontaminated soil.

• In general, oil contamination reduces the unconfined compression strength of the uncontaminated soil.

However, by an increase in bio-treatment time, unconfined compression strength increases, indicating a better performance of bio-treated soil in comparison with contaminated soil.

• The usage of bio-treated soil is recommended in road construction and erosion control projects. In addition, the construction of lightweight foundations on top of these soils does not pose any significant problem.

ACKNOWLEDGMENTS This research has been supported by Azarbaijan Shahid Madani University. The authors wish to thank the university authorities for their help in this research.

REFERENCES AL-RAWAS, A.; HASSAN, H. F.; TAHA, R.; HAGO, A.; ALSHANDOUDI, B.; AND AL-SULEIMANI, Y., 2005, Stabilization of oil-contaminated soils using cement and cement by-pass dust: Management of Environmental Quality: An International Journal, Vol. 16, No. 6, pp. 670-680.

AL-SANAD, H. A.; EID, W. K.; AND ISMAEL, N. F., 1995, Geotechnical properties of oil-contaminated Kuwaiti sand: Journal Geotechnical Engineering, Vol. 121, No. 5, pp. 407–412.

AL-SANAD, H. A. AND ISMAEL, N. F., 1997, Aging effects on oilcontaminated Kuwaiti sand: Journal oGeotechnical Geoenvironmental Engineering, Vol. 123, No. 3, pp. 290–293.

ASTM (AMERICAN SOCIETY FOR TESTING AND MATERIALS), 2007.

Annual Book of ASTM Standards: Section 4, Construction: Soil and Rock, Vol. 4.08. West Conshohocken, PA. http://www.astm.org.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Figure 10. Effect of contamination and bio-treatment on final settlement. Notations of BS-10 and BS-30 denote bio-treated samples after performing bio-treatment for 10 and 30 days, respectively.

Additionally, CS-30 presents the results of crude oil–contaminated samples after 30 days.

crude oil–contaminated samples are larger than the corresponding values in bio-treated samples. Moreover, it can be seen that the values of settlement for all bio-treated samples are less than those for uncontaminated samples.

Potential Applications By determining some of the geotechnical properties of contaminated and bio-treated soils and comparing with the corresponding values of uncontaminated soil, several applications could be proposed. An increase in maximum dry density and reduction in optimum moisture content was observed by performing bio-treatment; therefore, bio-treated soil is suitable for road construction. It has the potential to be stabilized with cement and mixed with crushed stone aggregate as a substitute in hot-mix asphalt concrete in road construction. Moreover, results showed that the values of the cohesion increase as bio-treatment time increases. It is, therefore, recommended to use bio-treated soil as an erosion controller. Generally, the soils with small particle sizes are susceptible to erosion because of their low resistance to wind and water. However, bio-treatment increased the size of the soil particles and their cohesion, improving the stability of soil against erosion.

This application of bio-treated soil is also approved by the crude observations and the SEM analysis. Furthermore, the reduction in the values of internal friction angel indicated that the use of bio-treated soils as a 340

ACKNOWLEDGMENTS This research has been supported by Azarbaijan Shahid Madani University. The authors wish to thank the university authorities for their help in this research.

REFERENCES AL-RAWAS, A.; HASSAN, H. F.; TAHA, R.; HAGO, A.; ALSHANDOUDI, B.; AND AL-SULEIMANI, Y., 2005, Stabilization of oil-contaminated soils using cement and cement by-pass dust: Management of Environmental Quality: An International Journal, Vol. 16, No. 6, pp. 670-680.

AL-SANAD, H. A.; EID, W. K.; AND ISMAEL, N. F., 1995, Geotechnical properties of oil-contaminated Kuwaiti sand: Journal Geotechnical Engineering, Vol. 121, No. 5, pp. 407–412.

AL-SANAD, H. A. AND ISMAEL, N. F., 1997, Aging effects on oilcontaminated Kuwaiti sand: Journal oGeotechnical Geoenvironmental Engineering, Vol. 123, No. 3, pp. 290–293.

ASTM (AMERICAN SOCIETY FOR TESTING AND MATERIALS), 2007.

Annual Book of ASTM Standards: Section 4, Construction: Soil and Rock, Vol. 4.08. West Conshohocken, PA. http://www.astm.org.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

Effect of bacteria on contaminated sand ASTM (AMERICAN SOCIETY FOR TESTING AND MATERIALS), 2016.

Annual Book of ASTM Standards: Section 4, Construction: Soil and Rock, Vol. 4.08. West Conshohocken, PA. http://www.astm.org.

ELISHA, A. T., 2012, Effect of crude oil contamination on the geotechnical properties of soft clay soils of Niger Delta region of Nigeria: Electronic Journal Geotechnical Engineering, Vol. 17, No. Bund M, pp. 1929–1938.

ESTABRAGH, A.; BEYTOLAHPOUR, I.; MORADI, M.; AND JAVADI, A., 2016, Mechanical behavior of a clay soil contaminated with glycerol and ethanol: European Journal Environmental Civil Engineering, Vol. 20, No. 5, pp. 503–519.

EVGIN, E. AND DAS, B., 1992, Mechanical behavior of an oil contaminated sand: In Usmen and Acar (Editors), Environmental Geotechnology Proceedings, Mediterranean Conference: Balkema Publishers, Rotterdam, Netherlands, pp. 101–108., Usman, H. and Acar, Y. (eds), Balkema, Rotterdam.

GHALY, A., 2001, Strength remediation of oil contaminated sands: The Seventeenth International Conference on Solid Waste Technology and Management: University of Pennsylvania, Widener University School of Engineering, pp. 289–298.

HEMMAT, A.; AGHILINATEGH, N.; AND SADEGHI, M., 2010, Shear strength of repacked remoulded samples of a calcareous soil as affected by long-term incorporation of three organic manures in central Iran: Biosystems Engineering, Vol. 107, No. 3, pp.

251–261.

KEMPER, W. AND ROSENAU, R., 1984, Soil cohesion as affected by time and water content: Soil Science Society America Journal, Vol. 48, No. 5, pp. 1001–1006.

KERMANI, M. AND EBADI, T., 2012, The effect of oil contamination on the geotechnical properties of fine-grained soils: Soil and Sediment Contamination: An International Journal, Vol. 21, No. 5, pp. 655–671.

KHAMEHCHIYAN, M.; CHARKHABI, A. H.; AND TAJIK, M., 2007, Effects of crude oil contamination on geotechnical properties of clayey and sandy soils: Engineering Geology, Vol. 89, No. 3, pp. 220–229.

KHOSRAVI, E.; GHASEMZADEH, H.; SABOUR, M. R.; AND YAZDANI, H., 2013, Geotechnical properties of gas oilcontaminated kaolinite: Engineering Geology, Vol. 166, No., pp. 11–16.

MOHAMEDZEIN, Y. E.-A.; AL-AGHBARI, M. Y.; AND TAHA, R. A., 2006, Stabilization of desert sands using municipal solid waste incinerator ash: Geotechnical Geological Engineering, Vol. 24, No. 6, pp. 1767–1780.

MOHAMMADI, S. D. AND MOHARAMZADE, S. KHALIL, 2014, Investigation of engineering geological behavior of surface oil hydrocarbons contaminated soils in tabriz oil refinery area: Engineering Geology, Vol. 7, No. 1–2, pp. 41–56.

MUKHOPADHYAY, A.; AL-AWADI, E.; QUINN, M.; AKBER, A.; ALSENAFY, M.; AND RASHID, T., 2008, Ground water contamination in Kuwait resulting from the 1991 Gulf War: a preliminary assessment: Groundwater Monitoring Remediation, Vol. 28, No. 2, pp. 81–93.

NASEHI, S. A.; UROMEIHY, A.; NIKUDEL, M. R.; AND MORSALI, A., 2016, Influence of gas oil contamination on geotechnical properties of fine and coarse-grained soils: Geotechnical Geological Engineering, Vol. 34, No. 1, pp. 333–345.

NASEHI, S. A.; UROMEIHY, A.; NIKUDEL, M. R.; AND MORSALI, A., 2016, Use of nanoscale zero-valent iron and nanoscale hydrated lime to improve geotechnical properties of gas oil contaminated

Effect of bacteria on contaminated sand

clay: a comparative study: Environmental Earth Sciences, Vol. 75, No. 9, pp. 1–20.

NASR, A. M., 2009, Experimental and theoretical studies for the behavior of strip footing on oil-contaminated sand: Journal of Geotechnical and Geoenvironmental Engineering, Vol. 135, No. 12, pp. 1814–1822.

OBETA, I. AND EZE-UZOMAKA, O., 2013, Geotechnical properties of waste engine oil contaminated laterites: Nigerian Journal Technology, Vol. 32, No. 2, pp. 203210.

OTA, J. O., 2013, The effect of light crude oil contamination on the geotechnical properties of kaolinite clay soil: Doctoral thesis, Anglia Ruskin University.

POURMOHAMMADBAGHER, A. AND SHAW, J. M., 2016, Probing the impact of asphaltene contamination on kaolinite and illite clay behaviors in water and organic solvents: a calorimetric study: Energy Fuels, Vol. 30, No. 8, pp. 6561–6569.

PURI, V. K., 2000, Geotechnical aspects of oil-contaminated sands: Soil Sediment Contamination, Vol. 9, No. 4, pp. 359–374.

RATNAWEERA, P. and MEEGODA, J. N., 2006, Shear strength and stress-strain behavior of contaminated soils: Geotechnical Testing Journal, Vol. 29, No. 2, pp. 1–8.

SAMANI, M. R.; EBRAHIMBABAIE, P.; AND MOLAMAHMOOD, H. V., 2016, Hexavalent chromium removal by using synthesis of polyaniline and polyvinyl alcohol: Water Science Technology, Vol. 74, No. 10, pp. 2305–2313.

SHIN, E. C. and DAS, B. M., 2001, Bearing capacity of unsaturated oil-contaminated sand: International Journal Offshore Polar Engineering, Vol. 11, No. 3, pp. 220–227.

SHIN, E.; LEE, J.; AND DAS, B., 1999, Bearing capacity of a model scale footing on crude oil-contaminated sand: Geotechnical Geological Engineering, Vol. 17, No. 2, pp. 123–132.

SHIN, E.; OMAR, M.; TAHMAZ, A.; DAS, B.; AND ATALAR, C., 2002, Shear strength and hydraulic conductivity of oil-contaminated sand: In de Mello, L.G. and Alemeida, M. (Editors), Proceedings of Environmental Geotechnics IV (ICEG): Balkema Publishers Rio de Janeiro, Brazil, pp. 11–15.

SOON, N. W.; LEE, L. M.; KHUN, T. C.; AND LING, H. S., 2013, Improvements in engineering properties of soils through microbial-induced calcite precipitation: KSCE Journal Civil Engineering, Vol. 17, No. 4, pp. 718–728.

SOON, N. W.; LEE, L. M.; KHUN, T. C.; AND LING, H. S., 2014, Factors affecting improvement in engineering properties of residual soil through microbial-induced calcite precipitation: Journal Geotechnical Geoenvironmental Engineering, Vol. 140, No. 5, 04014006. doi:10.1061/(ASCE)GT.1943-5606.0001089.

SPIECKER, P. M.; GAWRYS, K. L.; AND KILPATRICK, P. K., 2003, Aggregation and solubility behavior of asphaltenes and their subfractions: Journal Colloid Interface Science, Vol. 267, No. 1, pp. 178–193.

THOUAND, G.; BAUDA, P.; OUDOT, J.; KIRSCH, G.; SUTTON, C.; AND VIDALIE, J., 1999, Laboratory evaluation of crude oil biodegradation with commercial or natural microbial inocula: Canadian Journal Microbiology, Vol. 45, No. 2, pp. 106–115.

YASUHARA, H.; NEUPANE, D.; HAYASHI, K.; AND OKAMURA, M., 2012, Experiments and predictions of physical properties of sand cemented by enzymatically-induced carbonate precipitation: Soils Foundations, Vol. 52, No. 3, pp. 539–549.

ZHENG, X.; ZHANG, J.; ZHENG, T.; LIANG, C.; AND WANG, H., 2014, A developed technique for measuring water content in oil-contaminated porous media: Environmental Earth Sciences, Vol. 71, No. 3, pp. 1349–1356.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

341

ASTM (AMERICAN SOCIETY FOR TESTING AND MATERIALS), 2016.

Annual Book of ASTM Standards: Section 4, Construction: Soil and Rock, Vol. 4.08. West Conshohocken, PA. http://www.astm.org.

ELISHA, A. T., 2012, Effect of crude oil contamination on the geotechnical properties of soft clay soils of Niger Delta region of Nigeria: Electronic Journal Geotechnical Engineering, Vol. 17, No. Bund M, pp. 1929–1938.

ESTABRAGH, A.; BEYTOLAHPOUR, I.; MORADI, M.; AND JAVADI, A., 2016, Mechanical behavior of a clay soil contaminated with glycerol and ethanol: European Journal Environmental Civil Engineering, Vol. 20, No. 5, pp. 503–519.

EVGIN, E. AND DAS, B., 1992, Mechanical behavior of an oil contaminated sand: In Usmen and Acar (Editors), Environmental Geotechnology Proceedings, Mediterranean Conference: Balkema Publishers, Rotterdam, Netherlands, pp. 101–108., Usman, H. and Acar, Y. (eds), Balkema, Rotterdam.

GHALY, A., 2001, Strength remediation of oil contaminated sands: The Seventeenth International Conference on Solid Waste Technology and Management: University of Pennsylvania, Widener University School of Engineering, pp. 289–298.

HEMMAT, A.; AGHILINATEGH, N.; AND SADEGHI, M., 2010, Shear strength of repacked remoulded samples of a calcareous soil as affected by long-term incorporation of three organic manures in central Iran: Biosystems Engineering, Vol. 107, No. 3, pp.

251–261.

KEMPER, W. AND ROSENAU, R., 1984, Soil cohesion as affected by time and water content: Soil Science Society America Journal, Vol. 48, No. 5, pp. 1001–1006.

KERMANI, M. AND EBADI, T., 2012, The effect of oil contamination on the geotechnical properties of fine-grained soils: Soil and Sediment Contamination: An International Journal, Vol. 21, No. 5, pp. 655–671.

KHAMEHCHIYAN, M.; CHARKHABI, A. H.; AND TAJIK, M., 2007, Effects of crude oil contamination on geotechnical properties of clayey and sandy soils: Engineering Geology, Vol. 89, No. 3, pp. 220–229.

KHOSRAVI, E.; GHASEMZADEH, H.; SABOUR, M. R.; AND YAZDANI, H., 2013, Geotechnical properties of gas oilcontaminated kaolinite: Engineering Geology, Vol. 166, No., pp. 11–16.

MOHAMEDZEIN, Y. E.-A.; AL-AGHBARI, M. Y.; AND TAHA, R. A., 2006, Stabilization of desert sands using municipal solid waste incinerator ash: Geotechnical Geological Engineering, Vol. 24, No. 6, pp. 1767–1780.

MOHAMMADI, S. D. AND MOHARAMZADE, S. KHALIL, 2014, Investigation of engineering geological behavior of surface oil hydrocarbons contaminated soils in tabriz oil refinery area: Engineering Geology, Vol. 7, No. 1–2, pp. 41–56.

MUKHOPADHYAY, A.; AL-AWADI, E.; QUINN, M.; AKBER, A.; ALSENAFY, M.; AND RASHID, T., 2008, Ground water contamination in Kuwait resulting from the 1991 Gulf War: a preliminary assessment: Groundwater Monitoring Remediation, Vol. 28, No. 2, pp. 81–93.

NASEHI, S. A.; UROMEIHY, A.; NIKUDEL, M. R.; AND MORSALI, A., 2016, Influence of gas oil contamination on geotechnical properties of fine and coarse-grained soils: Geotechnical Geological Engineering, Vol. 34, No. 1, pp. 333–345.

NASEHI, S. A.; UROMEIHY, A.; NIKUDEL, M. R.; AND MORSALI, A., 2016, Use of nanoscale zero-valent iron and nanoscale hydrated lime to improve geotechnical properties of gas oil contaminated

clay: a comparative study: Environmental Earth Sciences, Vol. 75, No. 9, pp. 1–20.

NASR, A. M., 2009, Experimental and theoretical studies for the behavior of strip footing on oil-contaminated sand: Journal of Geotechnical and Geoenvironmental Engineering, Vol. 135, No. 12, pp. 1814–1822.

OBETA, I. AND EZE-UZOMAKA, O., 2013, Geotechnical properties of waste engine oil contaminated laterites: Nigerian Journal Technology, Vol. 32, No. 2, pp. 203210.

OTA, J. O., 2013, The effect of light crude oil contamination on the geotechnical properties of kaolinite clay soil: Doctoral thesis, Anglia Ruskin University.

POURMOHAMMADBAGHER, A. AND SHAW, J. M., 2016, Probing the impact of asphaltene contamination on kaolinite and illite clay behaviors in water and organic solvents: a calorimetric study: Energy Fuels, Vol. 30, No. 8, pp. 6561–6569.

PURI, V. K., 2000, Geotechnical aspects of oil-contaminated sands: Soil Sediment Contamination, Vol. 9, No. 4, pp. 359–374.

RATNAWEERA, P. and MEEGODA, J. N., 2006, Shear strength and stress-strain behavior of contaminated soils: Geotechnical Testing Journal, Vol. 29, No. 2, pp. 1–8.

SAMANI, M. R.; EBRAHIMBABAIE, P.; AND MOLAMAHMOOD, H. V., 2016, Hexavalent chromium removal by using synthesis of polyaniline and polyvinyl alcohol: Water Science Technology, Vol. 74, No. 10, pp. 2305–2313.

SHIN, E. C. and DAS, B. M., 2001, Bearing capacity of unsaturated oil-contaminated sand: International Journal Offshore Polar Engineering, Vol. 11, No. 3, pp. 220–227.

SHIN, E.; LEE, J.; AND DAS, B., 1999, Bearing capacity of a model scale footing on crude oil-contaminated sand: Geotechnical Geological Engineering, Vol. 17, No. 2, pp. 123–132.

SHIN, E.; OMAR, M.; TAHMAZ, A.; DAS, B.; AND ATALAR, C., 2002, Shear strength and hydraulic conductivity of oil-contaminated sand: In de Mello, L.G. and Alemeida, M. (Editors), Proceedings of Environmental Geotechnics IV (ICEG): Balkema Publishers Rio de Janeiro, Brazil, pp. 11–15.

SOON, N. W.; LEE, L. M.; KHUN, T. C.; AND LING, H. S., 2013, Improvements in engineering properties of soils through microbial-induced calcite precipitation: KSCE Journal Civil Engineering, Vol. 17, No. 4, pp. 718–728.

SOON, N. W.; LEE, L. M.; KHUN, T. C.; AND LING, H. S., 2014, Factors affecting improvement in engineering properties of residual soil through microbial-induced calcite precipitation: Journal Geotechnical Geoenvironmental Engineering, Vol. 140, No. 5, 04014006. doi:10.1061/(ASCE)GT.1943-5606.0001089.

SPIECKER, P. M.; GAWRYS, K. L.; AND KILPATRICK, P. K., 2003, Aggregation and solubility behavior of asphaltenes and their subfractions: Journal Colloid Interface Science, Vol. 267, No. 1, pp. 178–193.

THOUAND, G.; BAUDA, P.; OUDOT, J.; KIRSCH, G.; SUTTON, C.; AND VIDALIE, J., 1999, Laboratory evaluation of crude oil biodegradation with commercial or natural microbial inocula: Canadian Journal Microbiology, Vol. 45, No. 2, pp. 106–115.

YASUHARA, H.; NEUPANE, D.; HAYASHI, K.; AND OKAMURA, M., 2012, Experiments and predictions of physical properties of sand cemented by enzymatically-induced carbonate precipitation: Soils Foundations, Vol. 52, No. 3, pp. 539–549.

ZHENG, X.; ZHANG, J.; ZHENG, T.; LIANG, C.; AND WANG, H., 2014, A developed technique for measuring water content in oil-contaminated porous media: Environmental Earth Sciences, Vol. 71, No. 3, pp. 1349–1356.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 333–341

341

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole

NORBERT H. MAERZ1

NORBERT H. MAERZ1

Missouri University of Science and Engineering, 1006 Kingshighway, Rolla, MO 65401

Missouri University of Science and Engineering, 1006 Kingshighway, Rolla, MO 65401

KENNETH J. BOYKO

KENNETH J. BOYKO

Missouri University of Science and Engineering, 1006 Kingshighway, Rolla, MO 65401

Missouri University of Science and Engineering, 1006 Kingshighway, Rolla, MO 65401

GARY J. PENDERGRASS

GARY J. PENDERGRASS

GeoEngineers, 3050 South Delaware, Springfield, MO 65804

GeoEngineers, 3050 South Delaware, Springfield, MO 65804

JUSTIN W. BROWN

JUSTIN W. BROWN

GeoEngineers, 3050 South Delaware, Springfield, MO 65804

GeoEngineers, 3050 South Delaware, Springfield, MO 65804

Key Terms: LIDAR, Caverns, Borehole, Karst ABSTRACT A development in a karstic area of southwest Missouri initiated a geophysical survey as part of an initial site investigation. After conducting an extensive geophysical survey over the proposed site, a significant anomaly was discovered. Subsequently drilling of a 0.15 m (6 in.) borehole into the anomaly intersected a void between 16.2 m (53 ft) and 28.7 m (94 ft) of depth. The property is at the top of the watershed, so both the borehole and the void were dry.

In order to model the stability of the cavern in response to the proposed development, it was necessary to determine both the size and the shape of the void.

Key Terms: LIDAR, Caverns, Borehole, Karst

LIDAR SURVEY Terrestrial light detection and ranging (LIDAR) scanning is routinely used to map the inside of caverns (Idrees and Pradhan, 2016; Zlot and Bosse, 2016).

More recently, inexpensive LIDAR scanners have been used for this purpose (Higgins, 2016). Scanning voids through borehole is a more complex undertaking.

Commercial operators provide these kinds of services (Renishaw, 2017), but no systems are available for purchase.

The authors built a system to lower an inexpensive LeddarTech solid-state LIDAR scanner into the borehole. The resulting scan map revealed a void that was larger than expected from viewing the borehole photography. The void was imaged on April 28, 2015.

ABSTRACT A development in a karstic area of southwest Missouri initiated a geophysical survey as part of an initial site investigation. After conducting an extensive geophysical survey over the proposed site, a significant anomaly was discovered. Subsequently drilling of a 0.15 m (6 in.) borehole into the anomaly intersected a void between 16.2 m (53 ft) and 28.7 m (94 ft) of depth. The property is at the top of the watershed, so both the borehole and the void were dry.

In order to model the stability of the cavern in response to the proposed development, it was necessary to determine both the size and the shape of the void.

LeddarTech Scanner BOREHOLE CAMERA SURVEY A borehole camera survey was conducted on February 25, 2015. The survey confirmed the presence of a void and provided some clear and detailed images of the cavern walls (Figure 1), but in the end it was not possible to determine the size of the void; it was apparent to observers only that the void was elongated. In the absence of distance reference clues in the images, most observers were of the opinion that the void was not very large.

1

Corresponding author email: norbert@mst.edu.

Terrestrial light detection and ranging (LIDAR) scanning is routinely used to map the inside of caverns (Idrees and Pradhan, 2016; Zlot and Bosse, 2016).

More recently, inexpensive LIDAR scanners have been used for this purpose (Higgins, 2016). Scanning voids through borehole is a more complex undertaking.

Commercial operators provide these kinds of services (Renishaw, 2017), but no systems are available for purchase.

The authors built a system to lower an inexpensive LeddarTech solid-state LIDAR scanner into the borehole. The resulting scan map revealed a void that was larger than expected from viewing the borehole photography. The void was imaged on April 28, 2015.

LeddarTech Scanner

The LeddarTech M16 (Figure 2) scanner is a solidstate light-emitting diode (LED) sensor module manufactured by LeddarTech Inc. It has been used in a wide

Figure 1. Borehole camera image entering the top of the cavern (left) and viewing the side wall of the upper part of the cavern (right).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343â&#x20AC;&#x201C;347

LIDAR SURVEY

343

BOREHOLE CAMERA SURVEY A borehole camera survey was conducted on February 25, 2015. The survey confirmed the presence of a void and provided some clear and detailed images of the cavern walls (Figure 1), but in the end it was not possible to determine the size of the void; it was apparent to observers only that the void was elongated. In the absence of distance reference clues in the images, most observers were of the opinion that the void was not very large.

1

Corresponding author email: norbert@mst.edu.

The LeddarTech M16 (Figure 2) scanner is a solidstate light-emitting diode (LED) sensor module manufactured by LeddarTech Inc. It has been used in a wide

Figure 1. Borehole camera image entering the top of the cavern (left) and viewing the side wall of the upper part of the cavern (right).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343â&#x20AC;&#x201C;347

343

Maerz, Boyko, Pendergrass, and Brown

Maerz, Boyko, Pendergrass, and Brown

Figure 2. (Left) LeddarTech M16 Scanner (http://leddartech.com/modules/m16-multi-element-sensor-module) with dimensions of 114 mm (4.5 in.) high, 40 mm (2.75 in.) wide, and 38 mm (1.5 in.) deep. (Right) The scanner in a protective encasement.

Figure 2. (Left) LeddarTech M16 Scanner (http://leddartech.com/modules/m16-multi-element-sensor-module) with dimensions of 114 mm (4.5 in.) high, 40 mm (2.75 in.) wide, and 38 mm (1.5 in.) deep. (Right) The scanner in a protective encasement.

variety of applications, including collision avoidance systems, real-time speed measurements, drone altimeters, vehicle inspection, and other applications (LeddarTech, 2017). The device has a focused LED emitter that illuminates an area approximately 50◦ wide by 8◦ tall with a short pulse of infrared light. The reflected wave front is optically focused by a single lens on a 16-pixel linear array. For each pixel in this array, the sensor determines a distance based on time of flight.

The acquisition rate is adjustable from 6 Hz to over 1,000 Hz, although faster rates result in increased noise.

A good compromise between noise and speed can be achieved when using an acquisition rate of 50 Hz. Each pixel is 2.8◦ wide and 7.5◦ tall. The device has a measurement accuracy of ± 5 cm (2 in.) and a range of up to 15–50 m (40–164 ft), depending on the reflectivity of the surface being scanned. We note that according to the manufacturer, when scanning distances of less than 0.7 m (2.3 ft), the device may be unreliable; however, there was no indication of this happening during the scan conducted on April 28, 2015.

variety of applications, including collision avoidance systems, real-time speed measurements, drone altimeters, vehicle inspection, and other applications (LeddarTech, 2017). The device has a focused LED emitter that illuminates an area approximately 50◦ wide by 8◦ tall with a short pulse of infrared light. The reflected wave front is optically focused by a single lens on a 16-pixel linear array. For each pixel in this array, the sensor determines a distance based on time of flight.

The acquisition rate is adjustable from 6 Hz to over 1,000 Hz, although faster rates result in increased noise.

A good compromise between noise and speed can be achieved when using an acquisition rate of 50 Hz. Each pixel is 2.8◦ wide and 7.5◦ tall. The device has a measurement accuracy of ± 5 cm (2 in.) and a range of up to 15–50 m (40–164 ft), depending on the reflectivity of the surface being scanned. We note that according to the manufacturer, when scanning distances of less than 0.7 m (2.3 ft), the device may be unreliable; however, there was no indication of this happening during the scan conducted on April 28, 2015.

Mode of Deployment

Mode of Deployment

The M16 device was lowered into the ground on a series of 1.8-m (6-ft) rigid painter’s rods, marked at 0.3-m (1-ft) intervals (Figure 3). Measurement were started at the top of the void at 16.2 m (53 ft) to the bottom of the hole and void at 28.7 m (94 ft).

The M16 device was lowered into the ground on a series of 1.8-m (6-ft) rigid painter’s rods, marked at 0.3-m (1-ft) intervals (Figure 3). Measurement were started at the top of the void at 16.2 m (53 ft) to the bottom of the hole and void at 28.7 m (94 ft).

344

Figure 3. Deployment of the scanner. Painters rods act as a support and exact depth setting mechanism for the scanner. A custom-made clamp is used to hold the rods in place vertically while allowing the 360◦ of horizontal rotation.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

344

Figure 3. Deployment of the scanner. Painters rods act as a support and exact depth setting mechanism for the scanner. A custom-made clamp is used to hold the rods in place vertically while allowing the 360◦ of horizontal rotation.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole

Figure 4. 3-D rendering of the geometry of the void. The spacing of the blue dots is 1 m (3.28 ft).

Figure 4. 3-D rendering of the geometry of the void. The spacing of the blue dots is 1 m (3.28 ft).

At each 0.3-m (1-ft) interval, the rods were clamped into place, and the scanner was horizontally rotated through 360◦ at 45◦ intervals. The device was further supported by a steel cable and a rope, and a 100-ft (30.5-m) length of Ethernet cable transmitted the data to a computer on the surface. At each 0.3-m (1-ft) interval, 128 data points were measured around 360◦ (one point every 2.8◦ corresponding to the pixel width of the M16).

Data Acquisition A Linux program was written to facilitate data acquisition. Each time the device was lowered to a new depth, the program operator would enter the depth value. As the device was rotated to each of the eight subcardinal positions, the operator would hit the appropriate arrow key to collect 16 distances relating to that 45◦ segment. When all eight subcardinal directions were acquired, the device would be lowered 0.3 m (1 ft) to collect another round of radial distances. Each “ring” of 128 radial distances (eight subcardinal directions × 16 distances per subcardinal segment) was stored in a cylindrical data structure.

When the device reached the bottom of the cavity at 28.7 m (94 ft), 50 rings of radial distances had been collected. Knowing the depth of each ring and the azimuth of each of the 128 points in each ring, the (x, y, z) coordinates of all points were computed. These points

were then formatted into a Virtual Reality Modeling Language (VRML) file for interactive 3-D viewing.

The points were also formatted into a stereolithography (STL) file for printing on a 3-D printer.

RESULTS Results of the investigation show that the void was larger than would have been expected based on observing the borehole camera images. Results are presented as a 3-D solid model (Figure 4) or as cross sections (Figure 5). The results show that the void is up to 20 m (65.6 ft) long and up to 10 m (32.8 ft) wide.

SUMMARY AND CONCLUSIONS A simple, inexpensive method to map a boreholeaccessed air-filled cavern is described herein. This method is effective if crude. Relying on rods so that the device can be lowered to a known depth and orientation azimuth is time consuming. In addition, if the void were much deeper, the weight of the rods alone would be an issue.

Improvement of the methodology will consist of two advancements that would permit wire-line measurements. This would require a constantly spinning scanner with an integrated accelerometer/magnetic compass plus a wire-line depth measuring capability.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

345

At each 0.3-m (1-ft) interval, the rods were clamped into place, and the scanner was horizontally rotated through 360◦ at 45◦ intervals. The device was further supported by a steel cable and a rope, and a 100-ft (30.5-m) length of Ethernet cable transmitted the data to a computer on the surface. At each 0.3-m (1-ft) interval, 128 data points were measured around 360◦ (one point every 2.8◦ corresponding to the pixel width of the M16).

Data Acquisition A Linux program was written to facilitate data acquisition. Each time the device was lowered to a new depth, the program operator would enter the depth value. As the device was rotated to each of the eight subcardinal positions, the operator would hit the appropriate arrow key to collect 16 distances relating to that 45◦ segment. When all eight subcardinal directions were acquired, the device would be lowered 0.3 m (1 ft) to collect another round of radial distances. Each “ring” of 128 radial distances (eight subcardinal directions × 16 distances per subcardinal segment) was stored in a cylindrical data structure.

When the device reached the bottom of the cavity at 28.7 m (94 ft), 50 rings of radial distances had been collected. Knowing the depth of each ring and the azimuth of each of the 128 points in each ring, the (x, y, z) coordinates of all points were computed. These points

were then formatted into a Virtual Reality Modeling Language (VRML) file for interactive 3-D viewing.

The points were also formatted into a stereolithography (STL) file for printing on a 3-D printer.

RESULTS Results of the investigation show that the void was larger than would have been expected based on observing the borehole camera images. Results are presented as a 3-D solid model (Figure 4) or as cross sections (Figure 5). The results show that the void is up to 20 m (65.6 ft) long and up to 10 m (32.8 ft) wide.

SUMMARY AND CONCLUSIONS A simple, inexpensive method to map a boreholeaccessed air-filled cavern is described herein. This method is effective if crude. Relying on rods so that the device can be lowered to a known depth and orientation azimuth is time consuming. In addition, if the void were much deeper, the weight of the rods alone would be an issue.

Improvement of the methodology will consist of two advancements that would permit wire-line measurements. This would require a constantly spinning scanner with an integrated accelerometer/magnetic compass plus a wire-line depth measuring capability.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

345

Maerz, Boyko, Pendergrass, and Brown

Maerz, Boyko, Pendergrass, and Brown

Figure 5. 2-D cross section rendering of the geometry of the void. Maximum horizontal dimension of the void is about 20 m (65.6 ft.) at a depth of 25 m (82 ft.). Concentric rings are spaced at 0.3-m (1-ft) intervals.

Figure 5. 2-D cross section rendering of the geometry of the void. Maximum horizontal dimension of the void is about 20 m (65.6 ft.) at a depth of 25 m (82 ft.). Concentric rings are spaced at 0.3-m (1-ft) intervals.

Figure 6. Comparison of borehole video image (left) and LIDAR cross section (right) at a depth of 16.5 m (54 ft). The LIDAR image is oriented with north to the top; there is no orientation information available for the borehole video image.

Figure 6. Comparison of borehole video image (left) and LIDAR cross section (right) at a depth of 16.5 m (54 ft). The LIDAR image is oriented with north to the top; there is no orientation information available for the borehole video image.

REFERENCES HIGGINS, S., 2016, Scientist Builds His Own LiDAR for CaveScanning: Electronic document, available at https://www. spar3d.com/blogs/the-other-dimension/scientist-buildslidar-cave-scanning

346

IDREES, M. O. AND PRADHAN, B., 2016, A decade of modern cave surveying with terrestrial laser scanning: A review of sensors, method and application development: International Journal of Speleology, Vol. 45, No. 1, pp. 71–78.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

REFERENCES HIGGINS, S., 2016, Scientist Builds His Own LiDAR for CaveScanning: Electronic document, available at https://www. spar3d.com/blogs/the-other-dimension/scientist-buildslidar-cave-scanning

346

IDREES, M. O. AND PRADHAN, B., 2016, A decade of modern cave surveying with terrestrial laser scanning: A review of sensors, method and application development: International Journal of Speleology, Vol. 45, No. 1, pp. 71–78.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole LeddarTech, 2017, LeddarTech Blog: Electronic document, available at https://blog.leddartech.com/tag/m16 R Borehole-Deployable Laser Scanner for Renishaw, 2017, C-ALS Concealed Cavity and Void Scanning: Electronic document, available at http://www.renishaw.com/en/c-als-borehole-

LIDAR Scanning of an Air-Filled Cavern Accessed through a Borehole

deployable-laser-scanner-for-concealed-cavity-and-voidscanning–25590 ZLOT, R. AND BOSSE, M., 2014, Three dimensional mobile mapping of caves: Journal of Cave and Karst Studies, Vol. 76, No. 3, pp.

191–206.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

347

LeddarTech, 2017, LeddarTech Blog: Electronic document, available at https://blog.leddartech.com/tag/m16 R Borehole-Deployable Laser Scanner for Renishaw, 2017, C-ALS Concealed Cavity and Void Scanning: Electronic document, available at http://www.renishaw.com/en/c-als-borehole-

deployable-laser-scanner-for-concealed-cavity-and-voidscanning–25590 ZLOT, R. AND BOSSE, M., 2014, Three dimensional mobile mapping of caves: Journal of Cave and Karst Studies, Vol. 76, No. 3, pp.

191–206.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 343–347

347

Technical Note

Technical Note

The El Indio Mine Closure Plan Effects over the Water Quality of the Upper Elqui Basin

The El Indio Mine Closure Plan Effects over the Water Quality of the Upper Elqui Basin

JORGE OYARZÚN

JORGE OYARZÚN

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile

JORGE NÚÑEZ

JORGE NÚÑEZ

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile; and Centro del Agua para Zonas Aridas y Semiáridas de América Latina y el Caribe, Benavente 980, La Serena, Chile

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile; and Centro del Agua para Zonas Aridas y Semiáridas de América Latina y el Caribe, Benavente 980, La Serena, Chile

HUGO MATURANA

HUGO MATURANA

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile

RICARDO OYARZÚN*

RICARDO OYARZÚN*

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile; and Centro de Estudios Avanzados en Zonas Aridas Av. Ra’ul Bitr’an 1305, La Serena, Chile

Departamento Ingeniería de Minas/Facultad de Ingeniería/Universidad de La Serena, Benavente 980, La Serena, Chile; and Centro de Estudios Avanzados en Zonas Aridas Av. Ra’ul Bitr’an 1305, La Serena, Chile

Key Terms: Mining, Pollution, Reclamation, Geochemistry

INTRODUCTION In the Elqui Basin, volcano-sedimentary and plutonic rocks of calc-alkaline intermediate character dominate, along with numerous ore deposits and hydrothermal alteration zones (Oyarzún et al., 2003; Dittmar, 2004). There are three main sub-basins: those of the Turbio, Claro and Elqui rivers (Bodini and Araya, 1998). The Turbio River receives part of its flow from the Toro River, the latter being directly affected by the presence of the El Indio District, at the head of the Elqui watershed (Figure 1). El Indio is a mining district of Tertiary age, representing a special environmental concern because of its enargitic (Asrich) character and the important presence of extensive advanced hydrothermal alteration zones that are the main natural sources of As and acid drainage in the area (Oyarzún et al., 2003). The district was mainly mined between 1975 and 1997. However, natural pollution originating from the primary enargite-rich oxidized ore bodies (Cu3 AsS4 ) and from the extensive surrounding hydrothermal alteration zone began at least * Corresponding author e-mail: royarzun@auserena.cl

thousands of years before (Oyarzún et al., 2004). Thus, the combination of both natural and human activity– related factors has resulted in remarkably high levels of As, Cu, Fe, and SO4 , among other parameters, both in surface water and in sediments (Flores et al., 2017), which makes it a case of study of worldwide interest.

Between 2002 and 2005 approximately, a closure plan of the El Indio mine operations was implemented (Barrick, 2016). This plan included a settling pond in the Malo River course that directly drains the mineralized area, with the aim of lowering the As surface water content. With the same objective and given the low capacity of the pond, the use of the Pastos Largos tailings deposit, located in the same area, was later added. Galleguillos et al. (2008) assessed the effect that these works had on the abatement of metals in the Elqui River. In particular, that study conducted a preliminary evaluation of the consequences of the gradual cease of operations at the El Indio (1997–2002) and the implementation of the mine closure plan (2003–2005) on the water quality of the Elqui River and its main tributaries, considering As, SO4 2− , Cu, Fe concentrations, and pH, for the period 1996–2006. After nearly 10 years of that work, it is interesting to re-analyze the behavior of the upper Elqui River in terms of its water quality in order to assess the actual effectiveness of the closure measures adopted but in a longer time scale, which is the main objective of the present technical note. It is important to consider that this closure

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

349

Key Terms: Mining, Pollution, Reclamation, Geochemistry

INTRODUCTION In the Elqui Basin, volcano-sedimentary and plutonic rocks of calc-alkaline intermediate character dominate, along with numerous ore deposits and hydrothermal alteration zones (Oyarzún et al., 2003; Dittmar, 2004). There are three main sub-basins: those of the Turbio, Claro and Elqui rivers (Bodini and Araya, 1998). The Turbio River receives part of its flow from the Toro River, the latter being directly affected by the presence of the El Indio District, at the head of the Elqui watershed (Figure 1). El Indio is a mining district of Tertiary age, representing a special environmental concern because of its enargitic (Asrich) character and the important presence of extensive advanced hydrothermal alteration zones that are the main natural sources of As and acid drainage in the area (Oyarzún et al., 2003). The district was mainly mined between 1975 and 1997. However, natural pollution originating from the primary enargite-rich oxidized ore bodies (Cu3 AsS4 ) and from the extensive surrounding hydrothermal alteration zone began at least * Corresponding author e-mail: royarzun@auserena.cl

thousands of years before (Oyarzún et al., 2004). Thus, the combination of both natural and human activity– related factors has resulted in remarkably high levels of As, Cu, Fe, and SO4 , among other parameters, both in surface water and in sediments (Flores et al., 2017), which makes it a case of study of worldwide interest.

Between 2002 and 2005 approximately, a closure plan of the El Indio mine operations was implemented (Barrick, 2016). This plan included a settling pond in the Malo River course that directly drains the mineralized area, with the aim of lowering the As surface water content. With the same objective and given the low capacity of the pond, the use of the Pastos Largos tailings deposit, located in the same area, was later added. Galleguillos et al. (2008) assessed the effect that these works had on the abatement of metals in the Elqui River. In particular, that study conducted a preliminary evaluation of the consequences of the gradual cease of operations at the El Indio (1997–2002) and the implementation of the mine closure plan (2003–2005) on the water quality of the Elqui River and its main tributaries, considering As, SO4 2− , Cu, Fe concentrations, and pH, for the period 1996–2006. After nearly 10 years of that work, it is interesting to re-analyze the behavior of the upper Elqui River in terms of its water quality in order to assess the actual effectiveness of the closure measures adopted but in a longer time scale, which is the main objective of the present technical note. It is important to consider that this closure

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

349

Oyarzún, Núñez, Maturana, and Oyarzún

Oyarzún, Núñez, Maturana, and Oyarzún

Figure 1. Elqui River basin with water quality station locations, main tributaries, and El Indio mine.

Figure 1. Elqui River basin with water quality station locations, main tributaries, and El Indio mine.

plan faced several challenges, in particular related to the “sealing” of the mine, considering the high degree of rock fracturing coupled with an also high hydraulic potential for water entrance to the abandoned mine works.

Galleguillos et al. (2008) work were considered in the current contribution. The details regarding both time period covered and data availability are presented in Table 1.

plan faced several challenges, in particular related to the “sealing” of the mine, considering the high degree of rock fracturing coupled with an also high hydraulic potential for water entrance to the abandoned mine works.

Galleguillos et al. (2008) work were considered in the current contribution. The details regarding both time period covered and data availability are presented in Table 1.

METHODOLOGY

Table 1. Data availability (N) for each parameter and period considered.

METHODOLOGY

Table 1. Data availability (N) for each parameter and period considered.

Water quality data were downloaded from the Chilean Water Authority (DGA) website (http:// snia.dga.cl/BNAConsultas/reportes). For the sake of comparison with the study of Gallleguillos et al.

(2008), three stations located in the upper part of the Elqui Basin were considered in the current analysis: (1) the Toro River before its confluence with the La Laguna River (which is directly influenced by the El Indio District), (2) the La Laguna River before its confluence with the Toro River (a rather pristine river not affected by mining activity), and (3) the Turbio River after its confluence with the Toro and La Laguna rivers. Likewise, the same water quality parameters covered in the 350

1975–1995 Toro River La Laguna River Turbio River 1996–2006 Toro River La Laguna River Turbio River 2007–2016 Toro River La Laguna River Turbio River

pH

Cu

As

Fe

SO4 2−

208 207 176

181 142 138

193 188 152

174 170 135

93 81 77

110 109 109

94 93 93

92 91 91

94 93 92

42 41 42

58 59 58

58 61 58

56 59 55

57 58 57

47 48 47

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

Water quality data were downloaded from the Chilean Water Authority (DGA) website (http:// snia.dga.cl/BNAConsultas/reportes). For the sake of comparison with the study of Gallleguillos et al.

(2008), three stations located in the upper part of the Elqui Basin were considered in the current analysis: (1) the Toro River before its confluence with the La Laguna River (which is directly influenced by the El Indio District), (2) the La Laguna River before its confluence with the Toro River (a rather pristine river not affected by mining activity), and (3) the Turbio River after its confluence with the Toro and La Laguna rivers. Likewise, the same water quality parameters covered in the 350

1975–1995 Toro River La Laguna River Turbio River 1996–2006 Toro River La Laguna River Turbio River 2007–2016 Toro River La Laguna River Turbio River

pH

Cu

As

Fe

SO4 2−

208 207 176

181 142 138

193 188 152

174 170 135

93 81 77

110 109 109

94 93 93

92 91 91

94 93 92

42 41 42

58 59 58

58 61 58

56 59 55

57 58 57

47 48 47

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

El Indio and Elqui Water Quality

El Indio and Elqui Water Quality

Table 2. Average (and standard deviation) of pH and concentrations (mg/L) for Cu, As, Fe, and SO4 −2 as presented by Guevara et al. (2003) (1975–1995 period), determined in Galleguillos et al. (2008) (1996–2006), and derived from the current work (2007–2016) as well as Chilean regulations for irrigation (NCh 1333) and drinking water (NCh 409).

Table 2. Average (and standard deviation) of pH and concentrations (mg/L) for Cu, As, Fe, and SO4 −2 as presented by Guevara et al. (2003) (1975–1995 period), determined in Galleguillos et al. (2008) (1996–2006), and derived from the current work (2007–2016) as well as Chilean regulations for irrigation (NCh 1333) and drinking water (NCh 409).

1975–1995 Toro River La Laguna River Turbio River 1996–2006 Toro River La Laguna River Turbio River 2007–2016 Toro River La Laguna River Turbio River Chilean Regulation NCh 1333 NCh 409

pH

Cu

As

Fe

SO4 2−

5.1 (0.7) 7.6 (0.6) 7.2 (0.5)

5.6 (4.9) 0.1 (0.5) 1.4 (1.4)

0.8 (0.5) <0.1 (0.1) 0.2 (0.2)

21.9 (12.9) 1.0 (2.8) 6.1 (4.8)

843 (155.7) 96 (31.8) 300 (105.3)

4.8 (0.7) 8.0 (0.5) 7.5 (0.7)

17.0 (25.7) <0.1 (0.6) 2.9 (2.3)

0.5 (0.6) 0.1 (0.1) 0.1 (0.1)

20.4 (13.6) 1.7 (4.1) 6.2 (6.3)

943 (105.0) 115 (30.1) 296 (116.2)

4.6 (0.4) 8.3 (0.3) 7.5 (0.3)

10.5 (4.7) <0.1 (0.0) 2.0 (0.5)

0.5 (0.1) <0.1 (0.0) 0.1 (0.0)

18.5 (5.2) 1.2 (1.3) 5.2 (1.7)

1,101.5 (55.9) 125.4 (14.3) 322.9 (43.6)

5.5–9.0 6.0–8.5

0.2 1.0

0.1 0.05

5.0 0.3

250 500

The analysis included a simple statistical treatment (i.e., averages for each subperiod), a nonparametric trend analysis by the method of Mann-Kendall, and a sudden change detection test by the Petitt method.

The non-parametric rank-based Mann-Kendall (MK) test was chosen because it makes no assumption about the distribution of the data. Also, as the MK test is based on sign differences rather than values, it is recognized as robust (i.e., not affected by extreme values and outliers) and works well even with missing values (Helsel and Hirsch, 2002). The Pettit test is another non-parametric test used for the detection of a significant change in the mean of a time series when the exact time of the change is unknown (Gao et al., 2011). These tests were applied on the time series of annual mean values of the water quality parameters between 1975 and 2015 (Pohlert, 2016; R Core Team, 2016).

RESULTS AND DISCUSSION Table 2 shows, for the sake of comparison, the average figures obtained by Galleguillos et al. (2008) for 1996–2006 as well as those derived from the present study (2007–2016). Also, it includes results of a study

by Guevara et al. (2003), which considers the period between 1975 and 1995, that is, prior to and during the early stages of the exploitation of the El Indio mine.

Since there are no available records (DGA database) prior to 1974 for the rivers under study, the work of Guevara et al. (2003) may be considered as a proxy of a “baseline” (i.e., not mining affected) situation for the study area.

Among the main results of the study of Galleguillos et al. (2008), with respect to those initially pre-

sented by Guevara et al. (2003), are the important increase (ca. three times) of the Cu levels in the Toro River, along with a moderate decrease in pH and a slight increase (12%) of SO4 2− . This is a typical situation of an increase of pollutant loads linked to the generation of acid mine drainage in the district, “most likely because of the enhanced contact surface between rocks, water and air, a consequence of the network of galleries of the underground mining operation and of the degree of fracturing of host rocks enhanced by the blasting works” (Galleguillos et al., 2008). On the other hand, it is described a slight decrease of As, which could be associated with the works carried out in the closure plan of El Indio, in particular, the construction of the settling pond and the subsequent use of the Pastos Largos tailings dam with the same objective (i.e., as a sediment-settling facility). By extending the analysis for the period 2007–2016, a decrease in the level of Cu compared to the period 1996– 2006 is observed, which, in any case, still exhibits values above the “base level” (i.e., 1975–1995 records). A similar decline is observed for Fe, which in this case reaches lower figures than the “base level” values. For As and pH, no differences are observed when comparing the “current” situation (2007–2016) with the period covered in Galleguillos et al. (2008). Finally, SO4 2− exhibits a higher average for the 2007–2016 period when compared to that of 1996–2006 (1109 vs.

943 mg/L). As the La Laguna River area has not been affected by mining activities, the high sulfate values should be explained by other factors, such as a decrease in river discharge as a consequence of a drought period the region has experienced, which has been shown to be inversely related to SO4 2− concentrations (Flores et al., 2017).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

351

1975–1995 Toro River La Laguna River Turbio River 1996–2006 Toro River La Laguna River Turbio River 2007–2016 Toro River La Laguna River Turbio River Chilean Regulation NCh 1333 NCh 409

pH

Cu

As

Fe

SO4 2−

5.1 (0.7) 7.6 (0.6) 7.2 (0.5)

5.6 (4.9) 0.1 (0.5) 1.4 (1.4)

0.8 (0.5) <0.1 (0.1) 0.2 (0.2)

21.9 (12.9) 1.0 (2.8) 6.1 (4.8)

843 (155.7) 96 (31.8) 300 (105.3)

4.8 (0.7) 8.0 (0.5) 7.5 (0.7)

17.0 (25.7) <0.1 (0.6) 2.9 (2.3)

0.5 (0.6) 0.1 (0.1) 0.1 (0.1)

20.4 (13.6) 1.7 (4.1) 6.2 (6.3)

943 (105.0) 115 (30.1) 296 (116.2)

4.6 (0.4) 8.3 (0.3) 7.5 (0.3)

10.5 (4.7) <0.1 (0.0) 2.0 (0.5)

0.5 (0.1) <0.1 (0.0) 0.1 (0.0)

18.5 (5.2) 1.2 (1.3) 5.2 (1.7)

1,101.5 (55.9) 125.4 (14.3) 322.9 (43.6)

5.5–9.0 6.0–8.5

0.2 1.0

0.1 0.05

5.0 0.3

250 500

The analysis included a simple statistical treatment (i.e., averages for each subperiod), a nonparametric trend analysis by the method of Mann-Kendall, and a sudden change detection test by the Petitt method.

The non-parametric rank-based Mann-Kendall (MK) test was chosen because it makes no assumption about the distribution of the data. Also, as the MK test is based on sign differences rather than values, it is recognized as robust (i.e., not affected by extreme values and outliers) and works well even with missing values (Helsel and Hirsch, 2002). The Pettit test is another non-parametric test used for the detection of a significant change in the mean of a time series when the exact time of the change is unknown (Gao et al., 2011). These tests were applied on the time series of annual mean values of the water quality parameters between 1975 and 2015 (Pohlert, 2016; R Core Team, 2016).

RESULTS AND DISCUSSION Table 2 shows, for the sake of comparison, the average figures obtained by Galleguillos et al. (2008) for 1996–2006 as well as those derived from the present study (2007–2016). Also, it includes results of a study

by Guevara et al. (2003), which considers the period between 1975 and 1995, that is, prior to and during the early stages of the exploitation of the El Indio mine.

Since there are no available records (DGA database) prior to 1974 for the rivers under study, the work of Guevara et al. (2003) may be considered as a proxy of a “baseline” (i.e., not mining affected) situation for the study area.

Among the main results of the study of Galleguillos et al. (2008), with respect to those initially pre-

sented by Guevara et al. (2003), are the important increase (ca. three times) of the Cu levels in the Toro River, along with a moderate decrease in pH and a slight increase (12%) of SO4 2− . This is a typical situation of an increase of pollutant loads linked to the generation of acid mine drainage in the district, “most likely because of the enhanced contact surface between rocks, water and air, a consequence of the network of galleries of the underground mining operation and of the degree of fracturing of host rocks enhanced by the blasting works” (Galleguillos et al., 2008). On the other hand, it is described a slight decrease of As, which could be associated with the works carried out in the closure plan of El Indio, in particular, the construction of the settling pond and the subsequent use of the Pastos Largos tailings dam with the same objective (i.e., as a sediment-settling facility). By extending the analysis for the period 2007–2016, a decrease in the level of Cu compared to the period 1996– 2006 is observed, which, in any case, still exhibits values above the “base level” (i.e., 1975–1995 records). A similar decline is observed for Fe, which in this case reaches lower figures than the “base level” values. For As and pH, no differences are observed when comparing the “current” situation (2007–2016) with the period covered in Galleguillos et al. (2008). Finally, SO4 2− exhibits a higher average for the 2007–2016 period when compared to that of 1996–2006 (1109 vs.

943 mg/L). As the La Laguna River area has not been affected by mining activities, the high sulfate values should be explained by other factors, such as a decrease in river discharge as a consequence of a drought period the region has experienced, which has been shown to be inversely related to SO4 2− concentrations (Flores et al., 2017).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

351

Oyarzún, Núñez, Maturana, and Oyarzún

When the annual evolution of the contents of Cu, As and SO4 2− is analyzed, Galleguillos et al. (2008) noted for the Toro River that “while As shows a slight downward trend, both Cu and SO4 2− generally increase between 2000 and 2006, when the mining operations in the district already had ceased.” Similarly, these authors pointed out that “the so far observed behaviour suggests that the district preserves, and may even have increased its capacity to generate acid drainage, despite the sealing carried out in the inlet of the galleries, as developed in the closure plan.” Complementing this analysis, Figure 2 (left column) shows for the Toro River the time evolution of annual averages of the different parameters analyzed, considering (1) the period of the more intense exploitation of the deposit, (2) the period of the initial implementation of the closure plan, and (3) the subsequent period. In the case of pH, it is noted that lower levels occurred in the first years of the post-closure period (which is included in the analysis of Galleguillos et al., 2008), but this has shown an upward trend afterward. A similar but inverse situation is noted for Cu; that is, the highest values occurred in the early years of the post-closure period, a situation that apparently has begun to reverse later.

However, annual average concentrations are still far above thresholds of both NCH 409 and NCh 1333 regulations. In the case of As, an apparent upward trend is seen in the post-closure period, which confirms the behavior identified by Galleguillos et al. (2008). Also, given the lithological characteristics of the area, annual average values have historically exceeded the maximum levels defined by both NCh 409 and NCh 1333.

The element Fe, on the other hand, presents a rather permanent behavior toward a decrease in the concentrations in the post-closure period but again, similar to As, largely exceeding the referred Chilean regulation thresholds. Finally, SO4 2− has shown a marked upward trend, reaching in the past years the highest historical levels for this station. Again, like Fe, As, and Cu, SO4 2− concentrations are higher than Chilean regulation thresholds.

Certainly, part of the results described may be associated with the mining activity, the geological characteristics of the area, and the degree of effectiveness of measures implemented in the closure plan in order to mitigate potential pollution problems. However, it is also important to complement this analysis with that of a nearby watershed without mining activity during the same period, as is the case of the La Laguna River (Figure 2, right column). The pH and concentrations of the different parameters studied are markedly different (higher pH and lower concentrations of Cu, As, Fe, and SO4 2− in the La Laguna River). Its highlights the better water quality of the La Laguna River, as normally NCh 409 is not exceeded and where NCh 1333 352

thresholds are never surpassed. However, at the same time, it is interesting to note that certain behaviors described for the Rio Toro are also seen in the case of the La Laguna River. Indeed, As and SO4 2− show similar behavior in both cases, while Fe and Cu hardly show variations of consideration for the entire period (1975– 2015), with exceptions in specific years. Finally, Table 3 presents the results of a trend and break-point analysis (MK and Pettit tests) for the sub-basins, considering the entire period of records (1975–2016). While for the Toro River only Cu and SO4 2− show significant trends, pH, As, and SO4 2− exhibit this behavior in the case of the La Laguna River. From this observation, it can be further verified (especially for SO4 2− ) that along with the mine exploitation and closure activities, there are other factors, such as rainfall and flow regimes, that can influence the exhibited patterns. This is corroborated by the results of Pettit test, as it is observed that only in the case of SO4 2− , both sub-basins exhibit a point of change (in the temporal evolution) of annual average concentrations in the same year (1992). In other words, and in more general terms, the effectiveness of a closure plan will be affected by (and therefore should consider) a number of factors, that is, those related to the characteristics of the site and its operation (network of galleries and tunnels, presence of fractures and faults, mineralogy, etc.), as well as “external” ones, such as those associated with the climate, its possible fluctuations, and the influence on local hydrology. In fact, complementing this idea, a recent study by Flores et al. (2017) presents an estimate of the possible effects of climate and water regimes change over the water quality of the upper part of the Elqui River basin.

These authors show that a likely decline in flow (associated with projections of future climate related to lower level of rainfall) could favor lower dispersion and transport of Fe (and, to a lesser extent, of As and Cu) but at the same time could determine an increase in SO4 2− concentrations. Thus, a wide consideration of issues beyond geology, such as the hydrologic regime under which a mine has operated in the past (and, in relation to that, the actual hydrological data availability), along with the projected conditions for the future, should be done, especially when determining both remedial actions and the temporal extent of monitoring campaigns (e.g., for water quality) in the framework of closure plan definition and implementation.

Certainly this is a complex situation, and a thorough analysis goes beyond the scope of this technical note.

However, factors such as those briefly mentioned here deserve special attention in the case of mining or industrial activities that could develop into systems such as the upper Elqui, with tributaries with an important diluting role and with no mining activities such as the La Laguna River (diluting pollutants from the Toro

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

Oyarzún, Núñez, Maturana, and Oyarzún

When the annual evolution of the contents of Cu, As and SO4 2− is analyzed, Galleguillos et al. (2008) noted for the Toro River that “while As shows a slight downward trend, both Cu and SO4 2− generally increase between 2000 and 2006, when the mining operations in the district already had ceased.” Similarly, these authors pointed out that “the so far observed behaviour suggests that the district preserves, and may even have increased its capacity to generate acid drainage, despite the sealing carried out in the inlet of the galleries, as developed in the closure plan.” Complementing this analysis, Figure 2 (left column) shows for the Toro River the time evolution of annual averages of the different parameters analyzed, considering (1) the period of the more intense exploitation of the deposit, (2) the period of the initial implementation of the closure plan, and (3) the subsequent period. In the case of pH, it is noted that lower levels occurred in the first years of the post-closure period (which is included in the analysis of Galleguillos et al., 2008), but this has shown an upward trend afterward. A similar but inverse situation is noted for Cu; that is, the highest values occurred in the early years of the post-closure period, a situation that apparently has begun to reverse later.

However, annual average concentrations are still far above thresholds of both NCH 409 and NCh 1333 regulations. In the case of As, an apparent upward trend is seen in the post-closure period, which confirms the behavior identified by Galleguillos et al. (2008). Also, given the lithological characteristics of the area, annual average values have historically exceeded the maximum levels defined by both NCh 409 and NCh 1333.

The element Fe, on the other hand, presents a rather permanent behavior toward a decrease in the concentrations in the post-closure period but again, similar to As, largely exceeding the referred Chilean regulation thresholds. Finally, SO4 2− has shown a marked upward trend, reaching in the past years the highest historical levels for this station. Again, like Fe, As, and Cu, SO4 2− concentrations are higher than Chilean regulation thresholds.

Certainly, part of the results described may be associated with the mining activity, the geological characteristics of the area, and the degree of effectiveness of measures implemented in the closure plan in order to mitigate potential pollution problems. However, it is also important to complement this analysis with that of a nearby watershed without mining activity during the same period, as is the case of the La Laguna River (Figure 2, right column). The pH and concentrations of the different parameters studied are markedly different (higher pH and lower concentrations of Cu, As, Fe, and SO4 2− in the La Laguna River). Its highlights the better water quality of the La Laguna River, as normally NCh 409 is not exceeded and where NCh 1333 352

thresholds are never surpassed. However, at the same time, it is interesting to note that certain behaviors described for the Rio Toro are also seen in the case of the La Laguna River. Indeed, As and SO4 2− show similar behavior in both cases, while Fe and Cu hardly show variations of consideration for the entire period (1975– 2015), with exceptions in specific years. Finally, Table 3 presents the results of a trend and break-point analysis (MK and Pettit tests) for the sub-basins, considering the entire period of records (1975–2016). While for the Toro River only Cu and SO4 2− show significant trends, pH, As, and SO4 2− exhibit this behavior in the case of the La Laguna River. From this observation, it can be further verified (especially for SO4 2− ) that along with the mine exploitation and closure activities, there are other factors, such as rainfall and flow regimes, that can influence the exhibited patterns. This is corroborated by the results of Pettit test, as it is observed that only in the case of SO4 2− , both sub-basins exhibit a point of change (in the temporal evolution) of annual average concentrations in the same year (1992). In other words, and in more general terms, the effectiveness of a closure plan will be affected by (and therefore should consider) a number of factors, that is, those related to the characteristics of the site and its operation (network of galleries and tunnels, presence of fractures and faults, mineralogy, etc.), as well as “external” ones, such as those associated with the climate, its possible fluctuations, and the influence on local hydrology. In fact, complementing this idea, a recent study by Flores et al. (2017) presents an estimate of the possible effects of climate and water regimes change over the water quality of the upper part of the Elqui River basin.

These authors show that a likely decline in flow (associated with projections of future climate related to lower level of rainfall) could favor lower dispersion and transport of Fe (and, to a lesser extent, of As and Cu) but at the same time could determine an increase in SO4 2− concentrations. Thus, a wide consideration of issues beyond geology, such as the hydrologic regime under which a mine has operated in the past (and, in relation to that, the actual hydrological data availability), along with the projected conditions for the future, should be done, especially when determining both remedial actions and the temporal extent of monitoring campaigns (e.g., for water quality) in the framework of closure plan definition and implementation.

Certainly this is a complex situation, and a thorough analysis goes beyond the scope of this technical note.

However, factors such as those briefly mentioned here deserve special attention in the case of mining or industrial activities that could develop into systems such as the upper Elqui, with tributaries with an important diluting role and with no mining activities such as the La Laguna River (diluting pollutants from the Toro

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

El Indio and Elqui Water Quality

El Indio and Elqui Water Quality

Figure 2. Scatter plot and Lowess line of the annual averages of pH, Cu, As, Fe, and SO4 2− . (1) 1977–1997, period of intensive exploitation of the deposit. (2) 2003–2005, closure plan implementation period. (3) 2006–2016, post-closure period. Left column plots are for the Toro River, whereas right column plots correspond to the La Laguna River. For the sake of comparison, the same periods (shaded areas 1, 2, and 3) are retained in the case of La Laguna. Finally, threshold values defined by NCH 409 (solid line) and NCh 1333 (dashed line) are incorporated (when possible).

Figure 2. Scatter plot and Lowess line of the annual averages of pH, Cu, As, Fe, and SO4 2− . (1) 1977–1997, period of intensive exploitation of the deposit. (2) 2003–2005, closure plan implementation period. (3) 2006–2016, post-closure period. Left column plots are for the Toro River, whereas right column plots correspond to the La Laguna River. For the sake of comparison, the same periods (shaded areas 1, 2, and 3) are retained in the case of La Laguna. Finally, threshold values defined by NCH 409 (solid line) and NCh 1333 (dashed line) are incorporated (when possible).

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

353

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

353

Oyarzún, Núñez, Maturana, and Oyarzún

Oyarzún, Núñez, Maturana, and Oyarzún

Table 3. Statistical significance (p-values) of Mann-Kendall (MK) trend analysis and Pettit test. Significant values (5% level) in bold. For the Pettit test, when significant, the year of the point of change is indicated between brackets.

Table 3. Statistical significance (p-values) of Mann-Kendall (MK) trend analysis and Pettit test. Significant values (5% level) in bold. For the Pettit test, when significant, the year of the point of change is indicated between brackets.

Toro River MK Pettit La Laguna River MK Pettit Turbio River MK Pettit

pH

As

Cu

Fe

SO4 2−

0.29 0.07

0.25 0.01 (1994)

0.00 0.00 (1988)

0.84 0.80

0.00 0.00 (1992)

0.00 0.00 (1994)

0.01 0.02 (1984)

0.03 0.02

0.49 0.69

0.00 0.00 (1992)

0.24 0.26

0.65 0.15

0.00 0.00 (1987)

0.64 0.23

0.00 0.01 (1991)

River, avoiding major pollution problems in the middle and lower parts of the studied area). Of course, if such system is to be perturbed (e.g., by future mining activities in the La Laguna River), the effectiveness of a closure plan (such as at El Indio) would be affected, stressing the importance of actually keeping permanent and long-term monitoring and update of closure plans (beyond just economic issues).

CONCLUSION Almost 10 years after the official end (in terms of legal liability) of the El Indio closure plan, its effect on the water quality of the Toro River (draining the mine area) appears to have been effective in general terms, especially regarding the concentrations of Cu, Fe, and, to some extent, As. However, and as expected for these highly complex natural systems further modified by mining activities, there are processes associated with air/water/rock interactions that have continued to occur and have favored the generation of high levels of SO4 2− . Moreover, these phenomena may be influenced by natural factors as well (e.g., decrease in rainfall and river flows), which could explain a similar behavior (in terms of water quality) observed in the La Laguna River, adjacent to the Toro River, where no mining activity has been developed. Thus, both group of factors, that is, those specific to the mining site and abandoned mining operations as well as the climate context, must be considered in the designing, implementation, and monitoring activities for the medium and long term associated with a given closure plan of a mining project. In other words, despite how good a closure plan could be planned and developed, the existence of an ore deposit of the characteristics of El Indio (i.e., with advanced argillic alteration, high contents of As) as well as the existence of intense rock fracturing and with a steep hydraulic gradient present problems that are most likely impossible to completely solve and that could only be mitigated at best. Indeed, despite the location of El Indio in an arid to semi-arid 354

zone, the district presents a rather permanent groundwater flow. Also, the abundance of Fe3+ allows the continuous oxidation of the residual sulfur-rich masses of the mine, affecting (increasing) As, Cu, and SO4 2− contents, which most likely will be high for a long period.

ACKNOWLEDGMENTS This contribution was made within the framework of the Programs on Sustainable Mining (PROMIS) and Water Resources and Environment (PRHIMA) of the Mining Engineering Department of the University of La Serena and Fondecyt Project No. 1180153 (Conicyt). The feedback given by G. Galleguillos is also appreciated.

REFERENCES Barrick, 2016, Cierre El Indio: Electronic document, available at http://barricklatam.com/cierre-el-indio.

Bodini, H. and Araya, F., 1998, Visión geográfica global. La Región de Coquimbo. Espacios y recursos para un desarrollo sustentable: Centro de Estudios Regionales, Universidad de La Serena, La Serena, Chile, 39 p.

Dittmar, T., 2004, Hydrochemical processes controlling arsenic and heavy metal contamination in the Elqui river system (Chile): Science of the Total Environment, Vol. 325, No. 1–3, pp. 193–207.

Flores, M.; Nuñez, J.; Oyarzún, J.; Freixas, J.; Maturana, H.; and Oyarzún, R., 2017, Surface water quality in a sulfide mineral-rich arid zone in North-Central Chile: Learning from a complex past, addressing an uncertain future: Hydrological Processes, No. 31, pp. 498–513.

Galleguillos, G.; Oyarzún, J.; Maturana, H.; and Oyarzún, R., 2008, Retención de arsénico en embalses: El caso del río Elqui, Chile: Tecnología y Ciencias del Agua (formerly Ingeniería Hidráulica en México), Vol. 23, No. 3, pp. 29–36.

Gao, P.; Mu, X.M.; Wang, F.; and Li, R., 2011, Changes in streamflow and sediment discharge and the response to human activities in the middle reaches of the Yellow River: Hydrology Erath System Science, No. 15, pp. 1–10.

Guevara, S.; Oyarzún, J.; and Maturana, H., 2003, Geoquímica de las aguas del río Elqui y de sus tributarios en el período 1975–1995: Factores naturales y efecto de las explotaciones

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

Toro River MK Pettit La Laguna River MK Pettit Turbio River MK Pettit

pH

As

Cu

Fe

SO4 2−

0.29 0.07

0.25 0.01 (1994)

0.00 0.00 (1988)

0.84 0.80

0.00 0.00 (1992)

0.00 0.00 (1994)

0.01 0.02 (1984)

0.03 0.02

0.49 0.69

0.00 0.00 (1992)

0.24 0.26

0.65 0.15

0.00 0.00 (1987)

0.64 0.23

0.00 0.01 (1991)

River, avoiding major pollution problems in the middle and lower parts of the studied area). Of course, if such system is to be perturbed (e.g., by future mining activities in the La Laguna River), the effectiveness of a closure plan (such as at El Indio) would be affected, stressing the importance of actually keeping permanent and long-term monitoring and update of closure plans (beyond just economic issues).

CONCLUSION Almost 10 years after the official end (in terms of legal liability) of the El Indio closure plan, its effect on the water quality of the Toro River (draining the mine area) appears to have been effective in general terms, especially regarding the concentrations of Cu, Fe, and, to some extent, As. However, and as expected for these highly complex natural systems further modified by mining activities, there are processes associated with air/water/rock interactions that have continued to occur and have favored the generation of high levels of SO4 2− . Moreover, these phenomena may be influenced by natural factors as well (e.g., decrease in rainfall and river flows), which could explain a similar behavior (in terms of water quality) observed in the La Laguna River, adjacent to the Toro River, where no mining activity has been developed. Thus, both group of factors, that is, those specific to the mining site and abandoned mining operations as well as the climate context, must be considered in the designing, implementation, and monitoring activities for the medium and long term associated with a given closure plan of a mining project. In other words, despite how good a closure plan could be planned and developed, the existence of an ore deposit of the characteristics of El Indio (i.e., with advanced argillic alteration, high contents of As) as well as the existence of intense rock fracturing and with a steep hydraulic gradient present problems that are most likely impossible to completely solve and that could only be mitigated at best. Indeed, despite the location of El Indio in an arid to semi-arid 354

zone, the district presents a rather permanent groundwater flow. Also, the abundance of Fe3+ allows the continuous oxidation of the residual sulfur-rich masses of the mine, affecting (increasing) As, Cu, and SO4 2− contents, which most likely will be high for a long period.

ACKNOWLEDGMENTS This contribution was made within the framework of the Programs on Sustainable Mining (PROMIS) and Water Resources and Environment (PRHIMA) of the Mining Engineering Department of the University of La Serena and Fondecyt Project No. 1180153 (Conicyt). The feedback given by G. Galleguillos is also appreciated.

REFERENCES Barrick, 2016, Cierre El Indio: Electronic document, available at http://barricklatam.com/cierre-el-indio.

Bodini, H. and Araya, F., 1998, Visión geográfica global. La Región de Coquimbo. Espacios y recursos para un desarrollo sustentable: Centro de Estudios Regionales, Universidad de La Serena, La Serena, Chile, 39 p.

Dittmar, T., 2004, Hydrochemical processes controlling arsenic and heavy metal contamination in the Elqui river system (Chile): Science of the Total Environment, Vol. 325, No. 1–3, pp. 193–207.

Flores, M.; Nuñez, J.; Oyarzún, J.; Freixas, J.; Maturana, H.; and Oyarzún, R., 2017, Surface water quality in a sulfide mineral-rich arid zone in North-Central Chile: Learning from a complex past, addressing an uncertain future: Hydrological Processes, No. 31, pp. 498–513.

Galleguillos, G.; Oyarzún, J.; Maturana, H.; and Oyarzún, R., 2008, Retención de arsénico en embalses: El caso del río Elqui, Chile: Tecnología y Ciencias del Agua (formerly Ingeniería Hidráulica en México), Vol. 23, No. 3, pp. 29–36.

Gao, P.; Mu, X.M.; Wang, F.; and Li, R., 2011, Changes in streamflow and sediment discharge and the response to human activities in the middle reaches of the Yellow River: Hydrology Erath System Science, No. 15, pp. 1–10.

Guevara, S.; Oyarzún, J.; and Maturana, H., 2003, Geoquímica de las aguas del río Elqui y de sus tributarios en el período 1975–1995: Factores naturales y efecto de las explotaciones

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

El Indio and Elqui Water Quality mineras en sus contenidos de Fe, Cu y As: Chilean Journal of Agricultural Research (formerly Agricultura Técnica), Vol. 66, No. 1, pp. 57–69.

Helsel, D.R. and Hirsch, R.M., 2002, Statistical Methods in Water Resources Techniques of Water Resources Investigations.

Book 4, Chapter A3: U.S. Geological Survey, Washington, DC, 522 p.

Oyarzún, R.; Lillo, J.; Higueras, P.; Oyarzún, J.; and Maturana, H., 2004, Strong arsenic enrichment in sediments from the Elqui watershed, Northern Chile: Industrial (gold mining at El Indio–Tambo district) vs. geologic processes: Journal of Geochemical Exploration, No. 84, 53–64.

El Indio and Elqui Water Quality

Oyarzún, J.; Maturana, H.; Paulo, A.; and Pasieczna, A., 2003, Heavy metals in stream sediments from the Coquimbo Region (Chile): Effects of sustained mining and natural processes in a semi-arid Andean basin: Mine Water Environment, Vol. 22, No. 3, pp. 155–161.

Pohlert, T., 2016, Trend: Non-Parametric Trend Tests and Change-Point Detection. R Package Version 0.2.0: Electronic document, available at https://CRAN.R-project.org/ package=trend.

R Core Team, 2016, R: A Language and Environment for Statistical Computing: Electronic document, available at https://www.Rproject.org.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

355

mineras en sus contenidos de Fe, Cu y As: Chilean Journal of Agricultural Research (formerly Agricultura Técnica), Vol. 66, No. 1, pp. 57–69.

Helsel, D.R. and Hirsch, R.M., 2002, Statistical Methods in Water Resources Techniques of Water Resources Investigations.

Book 4, Chapter A3: U.S. Geological Survey, Washington, DC, 522 p.

Oyarzún, R.; Lillo, J.; Higueras, P.; Oyarzún, J.; and Maturana, H., 2004, Strong arsenic enrichment in sediments from the Elqui watershed, Northern Chile: Industrial (gold mining at El Indio–Tambo district) vs. geologic processes: Journal of Geochemical Exploration, No. 84, 53–64.

Oyarzún, J.; Maturana, H.; Paulo, A.; and Pasieczna, A., 2003, Heavy metals in stream sediments from the Coquimbo Region (Chile): Effects of sustained mining and natural processes in a semi-arid Andean basin: Mine Water Environment, Vol. 22, No. 3, pp. 155–161.

Pohlert, T., 2016, Trend: Non-Parametric Trend Tests and Change-Point Detection. R Package Version 0.2.0: Electronic document, available at https://CRAN.R-project.org/ package=trend.

R Core Team, 2016, R: A Language and Environment for Statistical Computing: Electronic document, available at https://www.Rproject.org.

Environmental & Engineering Geoscience, Vol. XXIV, No. 3, August 2018, pp. 349–355

355

Environmental & Engineering Geoscience (ISSN 1078-7275) is published quarterly by the Association of Environmental & Engineering Geologists (AEG) and the Geological Society of America (GSA). Periodicals postage paid at AEG, 201 East Main St., Suite 1405, Lexington, KY 40507 and additional mailing offices.

EDITORIAL OFFICE: Environmental & Engineering Geoscience journal, Department of Geology, Kent State University, Kent, OH 44242, U.S.A. phone: 330-672-2968, fax: 330-672-7949, ashakoor@kent.edu.

CLAIMS: Claims for damaged or not received issues will be honored for 6 months from date of publication. AEG members should contact AEG, 201 East Main St., Suite 1405, Lexington, KY 40507. Phone: 844-331-7867. GSA members who are not members of AEG should contact the GSA Member Service center. All claims must be submitted in writing.

POSTMASTER: Send address changes to AEG, 201 East Main St., Suite 1405, Lexington, KY 40507. Phone: 844-331-7867. Include both old and new addresses, with ZIP code. Canada agreement number PM40063731. Return undeliverable Canadian addresses to Station A P.O. Box 54, Windsor, ON N9A 6J5 Email: returnsil@imexpb.com.

DISCLAIMER NOTICE: Authors alone are responsible for views expressed in articles. Advertisers and their agencies are solely responsible for the content of all advertisements printed and also assume responsibility for any claims arising therefrom against the publisher. AEG and Environmental & Engineering Geoscience reserve the right to reject any advertising copy.

SUBSCRIPTIONS: Member subscriptions: AEG members automatically receive digital access to the journal as part of their AEG membership dues. Members may order print subscriptions for $60 per year.

GSA members who are not members of AEG may order for $60 per year on their annual GSA dues statement or by contacting GSA. Nonmember subscriptions are $295 and may be ordered from the subscription department of either organization. A postage differential of $10 may apply to nonmember subscribers outside the United States, Canada, and Pan America. Contact AEG at 844-331-7867; contact GSA Subscription Services, Geological Society of America, P.O. Box 9140, Boulder, CO 80301.

Single copies are $75.00 each. Requests for single copies should be sent to AEG, 201 East Main St., Suite 1405, Lexington, KY 40507. © 2018 by the Association of Environmental and Engineering Geologists All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from AEG. THIS PUBLICATION IS PRINTED ON ACID-FREE PAPER Abdul ShAkoor Department of Geology Kent State University Kent, OH 44242 330-672-2968 ashakoor@kent.edu

EDITORS

briAn G. kAtz Florida Department of Environmental Protection 2600 Blair Stone Rd. Tallahassee, FL 32399 850-245-8233 eegeditorbkatz@gmail.com

EDITORIAL BOARD Jerome V. DeGraff CSU Fresno Chester (skip) f. Watts Radford University thomas oommen Michigan Technological Univ. syeD e. hasan University of Missouri

thomas J. BurBey Virginia Polytechnic Institute aBDul shakoor Kent State University Brian G. katz Florida Department of Environmental Protection

ASSOCIATE EDITORS John W. Bell Nevada Bureau of Mines and Geology riCharD e. JaCkson Geofirma Engineering, Ltd. Jeffrey r. keaton AMEC Americas paul G. marinos National Technical University of Athens, Greece June e. mireCki U.S. Army Corps of Engineers peter pehme Waterloo Geophysics, Inc niCholas pinter Southern Illinois University

paul m. santi Colorado School of Mines roBert l. sChuster U.S. Geological Survey roy J. shlemon R. J. Shlemon & Associates, Inc. GreG m. stoCk National Park Service resat ulusay Hacettepe University, Turkey Chester f. “skip” Watts Radford University terry r. West Purdue University

SUBMISSION OF MANUSCRIPTS Environmental & Engineering Geoscience (E&EG), is a quarterly journal devoted to the publication of original papers that are of potential interest to hydrogeologists, environmental and engineering geologists, and geological engineers working in site selection, feasibility studies, investigations, design or construction of civil engineering projects or in waste management, groundwater, and related environmental fields. All papers are peer reviewed.

The editors invite contributions concerning all aspects of environmental and engineering geology and related disciplines. Recent abstracts can be viewed under “Archive” at the web site, “http://eeg. geoscienceworld.org”. Articles that report on research, case histories and new methods, and book reviews are welcome. Discussion papers, which are critiques of printed articles and are technical in nature, may be published with replies from the original author(s).

Discussion papers and replies should be concise.

To submit a manuscript go to http://eeg.allentrack.net. If you have not used the system before, follow the link at the bottom of the page that says New users should register for an account. Choose your own login and password. Further instructions will be available upon logging into the system. Please carefully read the “Instructions for Authors”.

Authors do not pay any charge for color figures that are essential to the manuscript. Manuscripts of fewer than 10 pages may be published as Technical Notes.

For further information, you may contact Dr. Abdul Shakoor at the editorial office.

Cover photo Photo highlighting the precipitate formed during the interaction of water with mortar in a tunnel, in the area of Southern Calabria, Southern Italy (photo courtesy of Giovanni Vespasiano). See article on page 305.