Croatian Journal of Forest Engineering (CROJFE) - 36-2_2015

2

2015

Original scientific paper

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry of the Future?

Ola Lindroos, Ola Ringdahl, Pedro La Hera, Peter Hohnloser, Thomas Hellström Abstract Modern harvesters are technologically sophisticated, with many useful features such as the ability to automatically measure stem diameters and lengths. This information is processed in real time to support value optimization when cutting stems into logs. It can also be transferred from the harvesters to centralized systems and used for wood supply management. Such information management systems have been available since the 1990s in Sweden and Finland, and are constantly being upgraded. However, data on the position of the harvester head relative to the machine are generally not recorded during harvesting. The routine acquisition and analysis of such data could offer several opportunities to improve forestry operations and related processes in the future. Here, we analyze the possible benefits of having this information, as well as the steps required to collect and process it. The benefits and drawbacks of different sensing technologies are discussed in terms of potential applications, accuracy and cost. We also present the results of preliminary testing using two of the proposed methods.

Our analysis indicates that an improved scope for mapping and controlling machine movement is the main benefit that is directly related to the conduct of forestry operations. In addition, there are important indirect benefits relating to ecological mapping. Our analysis suggests that both of these benefits can be realized by measuring the angles of crane joints or the locations of crane segments and using the resulting information to compute the head's position.

In keeping with our findings, two companies have recently introduced sensor equipped crane solutions.

Keywords: boom tip control, automation, ALS, sensors, harvester data

1. Introduction Forest machines used for fully mechanized cut-tolength (CTL) harvesting are technically advanced with the ability to perform complex operations such as automatic mechanical measurement of stem diameters and lengths during harvesting. These measurements are processed in real time to support value optimization when cutting stems into logs. They can also be transferred from the machines and used in central systems for wood supply management by the forest industry (e.g. Eriksson and Lindroos 2014). Information management systems of this kind have been used in Sweden and Finland since the 1990s, and have been refined over time (Anon. 2010). Whereas spatial data were initially only gathered at the stand level, it is inCroat. j. for. eng. 36(2015)2

creasingly common for this information to be disaggregated within stands.

The main factor driving this increase in resolution is the now common use of the global navigation satellite system (GNSS) to determine machine positions.

This positional information is used to link data for harvested trees to the machine's position when the tree was harvested (e.g. Bollandsås et al. 2011). However, the positional accuracy of current systems is limited by two key problems. First, GNSS data typically have a relatively low positional accuracy of ±10 meters in forest environments (Andersen 2009, Rodrigues-Pérez et al. 2007, Naesset and Jonmeister 2002). Second, the actual position of the harvester head relative to the machine's position is generally not known beyond the

147

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

fact that it must be somewhere within the crane’s reach, which is typically around 10 m. This introduces an accuracy error comparable in magnitude to that inherent to the GNSS data. Despite these limitations, current positioning systems have found some practical applications (e.g. Möller et al. 2012).

The potential of using measurements of the spatial position of the harvester head as a low cost method for forest mapping was outlined long ago (Stendahl and Dahlin 2002). It might well be that increased accuracy will unlock many intriguing possibilities for future forestry and related areas. However, there has been little progress towards this objective, and the full potential of existing data gathering methods has not yet been embraced due to the low accuracy of the positional data. While methods for gathering more accurate positional data could certainly be developed, it has not yet been demonstrated that the benefits of doing so would justify the costs. Ideally, the potential benefits of such undertakings should be analyzed before allocating scarce development resources (cf. Lindroos 2012, Berg et al. 2014).

Today’s technology makes it possible to gather data at various levels of accuracy and cost. Since different applications are likely to have different accuracy requirements, it is important to match the level of accuracy to the desired objectives. In this article, we investigate the possible benefits of gathering highly accurate spatial data on the position of the harvester head, and compare these benefits to the effort required to gather and process such data. This information is useful for understanding the benefits and drawbacks of different sensing technologies in terms of their applications, accuracy and costs.

To this end, we start by analyzing the possible benefits of knowing the harvester head's position, and the possible methods of acquiring such information. We then compare the features and limitations of these methods to the accuracy required to achieve each benefit. This leads to an evaluation of the benefits and methods by which they could be realized. To support our findings, we summarize results from feasibility tests on two of the proposed methods and then conclude with a discussion section.

2. The benefits of knowing the position of the harvester head Knowing the location of the harvester head confers a number of benefits that can be divided into two categories. Indirect benefits are those that relate to mapping of the forest environment. These benefits are real-

148

ized by using the position of the harvester head to estimate the positions of other objects. Direct benefits relate to mapping and controlling the movement of the machine; in such cases, information on the position of the harvester head is used to guide the harvesting work. Here we list and discuss benefits of both kinds.

While the list is quite extensive, the potential benefits are numerous and so we suspect that readers may be able to identify some that we have overlooked.

2.1 Defining accuracy For each benefit, we provide an estimate of its accuracy requirements. Naturally, excessively accurate positional information is not a problem, whereas insufficient accuracy may make certain benefits impossible to achieve. Thus, most applications benefit from having the highest possible accuracy. However, increased accuracy normally implies increased costs, so for each benefit we aim to identify the minimum level of accuracy needed to achieve the relevant functionality. For example, in some applications, it is crucial to know which stand a tree originates from. Accuracy at the mm level would provide such information, but would be excessive; the objective could be achieved equally well with an accuracy of 10 to 20 meters.

Therefore, we have tried to estimate a reasonable minimum level of accuracy for each application, while noting that the actual accuracy required in a given situation may be somewhat variable, depending on local conditions. Our aim is to broadly capture and contrast the levels of accuracy required to achieve different benefits, rather than to exactly specify the accuracy required to achieve a particular goal. Therefore, we have chosen to focus on positional accuracy in the horizontal plane, and accuracy requirements are defined in terms of a set of concentric circles centred on the true position of the harvester head. The radius of the largest circle containing the estimated position of the head is used to specify the level of accuracy required. Five accuracy categories are considered: >10m, m, dm, cm and mm. If a benefit requires mm accuracy, the estimated position of the head must be within several millimetres (with an upper limit of 1 cm) of its true position. The cm, dm, and m categories indicate that the estimated position must be within several centimetres, decimetres or meters of the true position, respectively. An accuracy requirement of >10 m indicates that the benefit in question can be achieved if the estimated position is within a circle having a radius of >10 m.

We did not define an upper limit for this category because it is used to characterize benefits whose accuracy requirements are relatively low, but naturally there is always some limit.

Croat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

2.2 Mapping the forest environment Information about the forest environment is required to support various forestry operations and the study or management of forest ecology. Such information is useful for answering questions relating to the nature of the standing forest (e.g. how much forest with specific properties do we have?) and operational questions (where should the machine drive, where have trees been harvested?). Here we focus on the benefits of knowing where trees are (or have been) spatially positioned and the location of the harvester head.

2.2.1 Reconstructing representative 3D models of forests The most obvious application for data on the location of the harvester head is that it could be used to update existing information on the positions of trees during harvesting. Analysis of such data would make it possible to virtually reconstruct the harvested trees based on the properties of their logs and thereby generate a 3D model of the harvested forest. In current systems, all trees harvested when the machine is in a given location are assumed to be located in the same place as the machine, which makes it appear as though all of the harvested trees were growing on top of each other (Anon. 2010). This is a well-known limitation, and tree-specific positional data could easily be incorporated into existing data gathering protocols if there were some method of gathering it. In this way, it would be possible to use data gathered during conventional harvesting to study the spatial distribution of tree species and sizes in the clearcut forest. Information obtained by such large scale destructive sampling would probably be useful in silvicultural applications and forest ecology. For instance, it would provide large amounts of representative data that could be used in spatial forest modelling (e.g. Arii et al. 2008, Thorpe et al. 2010, Fortin et al. 2013). Perhaps more interestingly, 3D forest models of this sort could be used to predict the features of unharvested forests by analyzing the harvester data in conjunction with information gathered by remote sensing (see below). »Estimated accuracy need«: m to dm. The most important information in such datasets is the positions of trees relative to one another. Accuracy errors of several meters could thus potentially be tolerated if they were systematic.

2.2.2 Training of ALS and other remote sensing methods Recent studies have demonstrated that combining automatically generated 3D forest maps with airborne laser scanning (ALS) could be extremely useful. ALS Croat. j. for. eng. 36(2015)2

O. Lindroos et al.

can provide very accurate geographical information for large areas of land. Height models of tree crowns and the ground can be generated from ALS data and processed using tree crown segmentation algorithms to produce global tree maps with complete coverage.

Several variables related to crown shape and size, such as the stem volume (Hyyppa et al. 2008), can then be estimated.

Today, manual field inventories conducted in geo referenced sample plots are still needed to establish the models used to predict stem attributes from ALS data (Naesset et al. 2004). In the future, local tree maps generated by harvesters during normal forestry operations could make it possible to collect more reference data to improve predictions and obtain more detailed stem data. In this way, normal harvesting would serve as a form of destructive sampling, yielding data that could be used to make inferences about similar forest areas. Gathering data in this way during continuous large scale forest operations would enable the creation of a system for the training and refinement of ALS based algorithms for large scale studies. Tree data collected by harvesters during harvesting operations have previously been used to train ALS data, but it has either been done based on the machine position, in which case the problem of having multiple trees with the same estimated position arises (Bollandsås et al. 2011), or based on manual positioning information that was linked to the tree data after harvesting (Holmgren et al. 2012, Barth and Holmgren 2013).

Local maps generated on the basis of information on the position of the harvester head relative to the machine must be combined with global ALS maps in order to enable the development of improved ALS models. In other words, the trees in the local maps generated from the harvester data must be matched to trees in the ALS map. Matching presents various challenges. First, a tree's top and stump may have different locations if the tree leans. Second, there may be limitations on the ability to identify individual trees within the ALS data. Third, matching may be challenging due to poor accuracy in machine positioning, although this problem could be alleviated by using matching algorithms that only take positional information from a GNSS as starting positions (Rossmann et al. 2009, Rossmann et al. 2010). Such an approach could be used as an alternative or complement to GNSS, to help improve its relatively poor positional accuracy in forest environments (Andersen et al. 2009, Rodríguez-Pérez et al. 2007, Naesset and Jonmeister 2002). ALS models could thus be refined on the basis of harvester head data and used to feed information back into GNSS systems in order to increase their positional accuracy.

149

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

Naturally, this is not the only way of improving the quality of the data on the machine's position. Several proposed solutions involve sensors mounted on forest machines to create local tree maps of their environments in real time (Hellström et al. 2009, Hellström and Ringdahl 2009, Öhman et al. 2008). This is typically achieved with a 2D laser scanner combined with SLAM (simultaneous localization and mapping) algorithms (Wang et al. 2005, Miettinen et al. 2007, Huang et al. 2008). To achieve high accuracy in the SLAM algorithms, the centre of the tree must be accurately determined (Dissanayake et al. 2001), which is quite challenging due to the irregular shapes of tree trunks and their variability over time (see e.g. Ringdahl et al.

2013). »Estimated accuracy need«: dm to m. The most important aspect of spatial data for individual trees is the accuracy of the relative position of trees. In this application, the positions of the harvested tree (as judged by the data from the harvester) are matched to the estimated location of trees from the ALS data. Consequently, as long as there is a good match between the two datasets, they can be used to support one another, reducing the need for high accuracy in the head position data. However, if the two datasets are poorly matched, for example if some harvested trees cannot be discerned in the ALS data, highly accurate harvested tree positions would be very important because they would be used as a basis for correcting the ALS algorithms. Overall, however, accuracy errors of several meters could probably be tolerated if they were systematic.

2.2.3 Timber traceability Forest owners must be able to demonstrate that their production and harvesting operations are conducted responsibly when supplying demanding end users. It is, therefore, important to ensure that all product components are traceable. For tree based products, this means having the ability to unambiguously determine where the trees used in a given product were growing before being harvested. Traceability requirements can be met with very low levels of accuracy, and the levels achieved by current machine positional data are generally sufficient. In fact, greater accuracy would be of little use and there are many harder challenges that should be addressed first to increase traceability.

The key missing link is the ability to link the data gathered for individual logs at different stages in the chain of custody, and to then trace the fate of the log parts as they are mixed and blended (this mixing and blending may be extensive – sheets of paper are made of dissolved fibres from thousands of trees). Several meth-

150

ods for tracing the fate of individual logs have been proposed (e.g. Hakli et al. 2010, Murphy et al. 2012, Seidel et al. 2012, Athanasiadis et al. 2013), but few are implemented on a large scale. »Estimated accuracy need«: >10 m. There is generally little need to provide anything more than the stand from which a given tree was harvested, and in fact even this low level of accuracy may be excessive.

The upper limit on the tolerable accuracy for this application depends on stand size, the demands of the end users, and potentially future legal requirements.

2.2.4 Improved characterization of product properties It is well known that the properties of a tree's wood depend on the conditions under which it is grown.

Silvicultural regimes are thus designed to optimize wood properties such as density, fibre angles, knot occurrence and taper (Yang 2002, Eriksson et al. 2006, Persson 1977). Technological developments have made it possible for harvesters to gather data while bucking trees into logs, including data on wood properties that had not previously been considered such as stiffness (Murphy 2014). Furthermore, while trees of the same species can also vary in their chemical composition (e.g. Arshadi et al. 2013), data on the chemical composition of wood is rarely used during industrial processing. However, the ongoing development of new wood products and processes, such as biorefining, mean that it may become increasingly important to identify trees with desirable chemical properties. Data collected by harvesters can already provide industry relevant information at the stand level (Nordström et al. 2010). However, some important properties are likely to be related to the tree's geo-spatial properties. It may, therefore, be important to determine the locations of individual trees on a global map that also records data on local spatial conditions such as stand density. Information of this sort, such as the slope of the land a tree is growing on, the nature of the soil at the site, or the density of the tree stand can also be captured manually by the operator. However, the need for manual recording could be reduced or eliminated if better positional data were available. As with traceability, this would require the ability to link data for individual logs from multiple points in the supply chain. However, sorting on the basis of product features could be done at a relatively late stage in the supply chain, thereby avoiding the need for costly sorting in the forest. »Estimated accuracy need«: >10 m to 1 m. There is generally little need for greater positional accuracy than that provided by existing systems which simply Croat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

record the machine's position. However, the ability to provide more detailed information on the properties of raw materials could potentially enable the development of fine-tuned industrial processes that require feedstock with tightly defined properties. It is, therefore, possible that the ability to deliver would create a requirement for greater accuracy.

2.2.5 Calculating the density of tree removal Thinnings are normally performed at an intensity that is chosen so as to leave a residual stand with the density required to provide the best possible conditions for future development. Traditionally, thinning intensities are defined using mean values for entire stands, often in terms of the total basal area that should be harvested. A given intensity can be achieved by felling several thin trees or fewer thicker trees. It can, therefore, be difficult for operators to decide which trees to fell to achieve the desired intensity. Moreover, the tree density normally varies somewhat within a stand, making it even more difficult to decide which trees to fell. In some countries, this problem is addressed by marking trees to be harvested. However, in Nordic countries, operators select which trees to harvest. While their selections are normally considered to be quite reasonable, the selection process requires recurring manual (and thus costly) calibration by the operators. Improvements would, therefore, be beneficial.

Various methods have, therefore, been developed to calculate the density of trees removed per area (Stendahl and Dahlin 2002, Möller et al. 2012). However, these techniques are based on the machine's position, with estimations of the area harvested for given numbers of trees. They are useful since they provide the operator with real time feedback on the harvested density during thinnings, which can be compared to the desired intensity. The ability to determine the harvester head's position would increase the accuracy of these calculations, enabling better thinning and documentation of the variation in removal intensity within a stand. This would be useful in supporting operators’ decision making. However, this approach would be a lot more powerful if it were complemented by methods for sensing or deducing the density of the residual stand. Machine mounted sensors could potentially provide information on the residual stand (e.g. Rossman et al. 2009), but it will be challenging to develop appropriate and affordable sensor technologies. Alternatively, and perhaps more practically in the near term, information on the intensity of removal could be integrated with ALS data on the properties of the original stand. Provided that ALS data can be Croat. j. for. eng. 36(2015)2

O. Lindroos et al.

used to reliably identify small trees in the lower parts of the canopy, spatially specific thinning regimes could be designed and implemented by calculating the density of the residual stand as the difference between that of the original stand (determined from the ALS data) and the harvested density (determined from the data supplied by the harvester). Harvester head positioning data could thus enable operators to implement desired thinning regimes more easily and accurately. »Estimated accuracy need«: >10 m to 1 m. Existing methods (Möller et al. 2012) are functional with positional accuracies in the >10 m range, since they use information on the position of the machine rather than the head. A higher positional resolution would be beneficial for the development of detailed thinning regimes, but would probably be most useful for training the ALS models used to estimate the initial stand density and spatially resolve the thinning intensity.

2.2.6 Virtual marking and constraining If the position of the harvester head was recorded, it could potentially be used as a virtual pen to mark the positions of various features on digital maps of the harvesting site. This would need to be combined with a means for the operator to record the identity of the mapped feature. The most obvious features to record in this way would be the locations of created snags, i.e. trees cut at a greater height than normal. In addition, objects of interest could be mapped by holding the harvester head above or next to them. In this way, the locations of ecologically and/or culturally interesting objects could be reliably recorded on digital maps. Borders of various kinds could be delineated in a similar way, allowing to introduce virtual obstacles to the motion of the harvester and its crane in order to avoid harvest of trees or machine driving in predefined areas. »Estimated accuracy need«: 1 m in general, but potentially 1 dm if dealing with legal issues such as property boundaries. However, like current digital and paper maps, the generated maps would probably just be used as indicative maps, and field inspections would be required to verify the locations of specific objects. Accuracy at sub meter levels would thus be unnecessary for current applications, although it is possible that accuracy demands would become more stringent as data of this sort became readily available and applications were developed.

2.3 Mapping and controlling machine movements Reliable and accurate information on the positions of individual machine parts is essential when mapping and controlling machine movements. Conse-

151

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

quently, the acquisition and processing of such information is being studied intensively around the world in order to support the development of machines capable of autonomous navigation and performing various other functions (semi) autonomously. In the context of forestry, there has been substantial recent progress in this area (e.g. Hellström et al. 2009, Mettin et al. 2009, Rossman et al. 2009, Rossman et al. 2010, Ringdahl 2011, Ortiz Morales et al. 2014), but industrial acceptance of such approaches has been less widespread than is the case in related fields such as agriculture. Here we discuss the general benefits that could be realized by knowing the position of the boom tip of any forest machine (e.g. harvester or forwarder) because they are not related to ecological data collected during harvesting. In fact, these benefits could in principle be achieved for any machine with a hydraulic crane.

2.3.1 Machine (semi) automation For most mechanical manipulators, it is important to be able to estimate the position (and orientation) of the end effector in order to implement any form of decision making concerning its actions. This information can be fed into computer algorithms to control the manipulator’s movements or to provide information for supervision. The required accuracy of the estimates usually depends on the application. For example, controlling the motion of a manipulator requires high accuracy and fast sampling because the estimates are used by computers as feedback information for motion control. However, if the information is only needed from time to time for supervisory purposes, a wider range of measurement resolutions and sampling speeds may be acceptable.

Ideas of this sort have been studied for some time within the forest machine industry. Recent developments in sensing technologies for cranes have led to the introduction of one of the first commercial products for controlling forestry cranes using computerized algorithms (John Deere 2013). Solutions of this sort are said to provide »boom tip control« (BTC) because the operator uses joysticks to control the tip's movements rather than independently manipulating the movements of individual sections of the crane arm (e.g. Gellerstedt 2002). This has a number of advantages, not least of which is that it reduces the difficulty of controlling the crane, making the process more intuitive and easier to learn while improving the machine's efficiency (Westerberg 2014).

Although the commercialization of BTC represents a milestone for the forest machine industry, various other computer controlled functions have been sug-

152

gested and developed over the years. For instance the works presented by Shiriaev et al. (2008), Mettin et al.

(2009), Westerberg (2014), and Ortiz Morales et al.

(2014) discuss future applications in which most crane operations are governed by a mixture of human commands and semi-autonomous functions. A simple example would be a system that allowed a harvester to autonomously approach a tree that has been selected for felling. It would then be possible to use the existing technology to control the gripping and felling of the tree, which could be performed manually or by taking advantage of other modes of interaction such as voice commands. The automation of forwarder cranes would also offer several advantages because it could reduce the operator’s mental fatigue by automating the movement of the crane between bunk and log piles. This is also beneficial in terms of efficiency because computers are significantly better than people at identifying optimal working conditions in terms of speed of work, fuel consumption, energy usage, and so on (Westerberg 2014). Additionally, this technology also offers the possibility to improve operator and machine safety by applying virtual restrictions to harmful crane movements.

Despite the various technological advances underpinning these solutions and their commercial success, there are still major challenges to overcome in the automation of forestry machines. A key challenge stems from the highly unstructured nature of the forest environment, which will necessitate the development of reliable sensing technologies. In addition, considerable further progress in robotics will be required to enable fully automated forestry operations. »Estimated accuracy need«: mm to cm. The required accuracy varies from application to application.

However, an accuracy in the centimetre range is essential for efficient autonomous crane movement; lower levels of accuracy would impair the functioning of the control algorithms and risk damaging the machine or trees. The greater the sophistication and autonomy of the control system, the greater the level of accuracy required.

2.3.2 Improving operators’ working methods Operating forest machines is known to be challenging because many tasks must be conducted simultaneously and at high speed (e.g. Gellerstedt 2002, Ovaskainen and Heikkilä 2007). Operators are trained to handle the machines, but as with most trades, there are often several ways of accomplishing a given task, particularly given the heterogeneity of forest environments. It is, therefore, difficult to identify the most efficient working method for a particular situation and Croat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

task. The development of screen based forest machine simulators has facilitated the training of machine operators, allowing education and evaluation to be conducted in virtual environments (e.g. Ovaskainen 2005). However, the scope for practical evaluation of specific working methods is currently limited. Studies on the time consumption associated with specific elements of a working method can be helpful but their results are not readily translated into assessments of the efficiency of a given approach. A key problem in such analyses is that, while experienced individuals can relatively easily assess the general efficiency of an operator after briefly observing their work, they generally cannot easily explain in detail the reasons for their judgement (Purfürst and Lindroos 2011). However, if crane movements could be monitored, the resulting data might enable the analysis of efficient working methods, which could then be taught to other operators. For instance, Ortiz Morales et al. (2014) have shown that operators’ working practices can be improved by monitoring their crane movements and using motion optimization to analyze their working patterns and suggest ways of increasing efficiency. The results of such studies have direct applications in areas such as the training of machine operators, automation, information management and interaction design. »Estimated accuracy need«: dm. A positional accuracy of a few decimetres should be sufficient to identify the most efficient crane movement patterns for most types of work. Given the great diversity of working conditions encountered in practical forestry and the high levels of variation in the way different operators perform different movements, greater accuracy is unlikely to be particularly useful in this context. Moreover, it is unlikely that an operator would be able to follow a given crane path with better accuracy than some decimetres, given the time constraints that apply during normal work.

3. Methods for estimating harvester head pose 3.1 Working conditions and required features Before discussing methods that could be used to monitor the position of the harvester head, it is important to define the conditions under which these methods must function. Forest work is conducted all year round in unstructured terrain, which implies a high variation in temperature, humidity and visibility.

Moreover, forest machines move and vibrate due to both the work they do and the movements of the machine through the rough terrain. In addition, the machine's engine makes the environment noisy. Since forCroat. j. for. eng. 36(2015)2

O. Lindroos et al.

estry work involves harvesting trees, the machine and its sensors are at risk of being hit by trees or branches.

Finally, the forest environment is dirty; the machine will be exposed to soil, sawdust, rain, and possibly also snow, all of which may accumulate on its sensors.

There are also two other factors that should be considered. First, the harvester head will typically be within 10 m of the machine because its position is limited by the reach of the crane. Second, forestry operations are cost sensitive, so any method used to monitor the head position must not greatly increase the cost of the work.

In most non forestry contexts, the most practical way of obtaining robust and high resolution data on the position of the manipulator is to use sensors that are physically attached to it. A wide range of suitable sensors of different prices and sizes are available. As a rule of thumb, more accurate sensors are more costly. A second factor to consider when mounting sensors on a manipulator relates to the extra hardware needed.

For instance, external mounting often requires additional mechanical support, holders, screws, etc., which will probably have to be customized because the relevant components will not necessarily be commercially available. This may also be the case for any additional cabling and electronic devices required for data acquisition. In addition, real time data processing is required to achieve most of the benefits discussed in the preceding section. Therefore, fast computers are required.

In summary, an ideal method should be relatively economical, robust enough to tolerate harsh and varied forest conditions, and instantaneously provide accurate data.

3.2 Overview of methods Methods for determining the location of a harvester head may be either local or global, depending on the location of the »base coordinate system« (CS). Local methods use a CS fixed to the harvester while global methods use a CS fixed to the ground. Methods can also be categorized according to the level of information they provide on the head's location. The full »pose« of an object in 3D is a set of six numbers defined relative to the chosen CS that specify the head's »position« in space and its »orientation«. The »orientation« or »attitude« is a set of three numbers describing the head's placement in terms of rotations around the three coordinate axes. The »position« is another set of three numbers that define the head's location in terms of translations or offsets along the coordinate axes.

Different applications may require different components of the full pose, which can be measured in either relative or absolute fashion. To identify a tree

153

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

selected for harvesting on a map, it is typically sufficient to provide only the position of the harvester head with respect to the x and y axes (in global coordinates).

In contrast, semi-autonomous control of a crane during harvesting operations requires the full pose, expressed relative to the harvester.

An overview of existing methods for local pose estimation is given below. The discussion is limited to local techniques because, in most cases relevant to harvesters, it is sufficient to know the pose relative to the machine. Furthermore, a local pose can be easily transformed into a global one if the global pose of the local CS is known. The methods presented are grouped into four different categories: »Angle and range-based methods« derive the pose by estimating the angles and/or ranges (distances) between a number of sensors and the harvester head. »Joint estimating methods« estimate the pose based on the geometry of the crane combined with direct measurements of joint angles and displacements. »Inertial techniques« use a combination of accelerometers, gyroscopes, and magnetometers, while »tilt sensors« estimate the static pose by sensing the head's orientation with respect to the earth’s gravitational field.

3.3 Angle and range based methods Poses can be estimated by measuring various angles, i.e. by triangulation. Wiklund et al. (1988) described the use of a rotating laser placed on the roof of an autonomous guided vehicle (AGV) to estimate the vehicle's pose by measuring the angles between the vehicle's long axis and assorted reflectors placed at known points in the environment. Alternatively, it is possible to have a similar system in which the fixed devices are the signal sources: Chunhan et al. (2003) used infra-red emitters placed at fixed positions and measured the incident angle of their emitted light on a sensor placed on the object whose position was to be determined.

One limitation of this approach is that it requires a clear line of sight to the harvester head. Given that the head rotates, this might be hard to achieve. Locating the reflectors on the boom tip could potentially solve this problem. However, even if the reflectors could always be pointed in the desired direction, the line of sight could be blocked by obstacles (e.g. trees, undergrowth, stones) or atmospheric conditions (e.g. fog, rain, snow or dust). Given a free line of sight to a visible reflector, triangulation should be able to determine the position of the head (relative to the harvester) with cm to dm accuracy.

Poses can also be estimated by various types of range estimations (trilateration). Satellite navigation systems such as GNSS utilize this approach to estimate

154

global poses from »time of flight« data for radio signals broadcast simultaneously from several satellites. The GNSS receiver is placed on the object to be localized.

Essential for this technique is exact time synchronization, which is achieved by using more satellites than would otherwise be necessary. Trilateration is also used in some local methods. Smith et al. (2004) presented a technique for indoor positioning that does not require synchronization. In this approach, range is estimated from »time of flight« data for ultrasonic pulses transmitted from beacons placed in the environment, and radio waves are used to synchronize transmission. It is also possible to place the ultrasonic transmitter on the object to be localized, as is done in the »Active Bat« location system (Harter et al. 1999). In such cases, the pose of the object is determined on the basis of signals picked up by several fixed receivers.

One potential solution based on this approach would be to simply equip the harvester head with a separate GNSS receiver. However, the resulting data would have the same limited accuracy as the GNSS positioning data for the machine. Thus, until better GNSS accuracy is achieved, it seems more suitable to estimate the harvester head's position relative to the machine.

If both angular and range data are available, a single measurement may be sufficient to specify the object's position. The object's position in Cartesian coordinates can be easily obtained by conversion from polar coordinates. Several types of sensors can be utilized to obtain combined angular and range measurements. For example, laser scanners emit laser beams and directly measure the angular coordinates and linear displacement of objects that reflect those beams into a 2D plane in front of the scanner. To determine the position of a specific object, it must be identified in the laser scan such that the relevant angle and range are determined. Depending on the type of laser used and the nature of the object, this may be quite challenging (for an example, see the section on field experiment 2).

Cameras can also be utilized as described by Davison and Kita (2002), who employed stereovision to detect a marker on the target object and thereby determine its 3D position and 2D orientation.

These methods require a (reasonably) clear line of sight and the ability to distinguish the desired object (i.e. the harvester head and/or the tree to be harvested) from other nearby objects. Provided that these conditions are satisfied, techniques such as laser scanning can provide very accurate (mm to cm) estimates of the harvester head's position relative to the machine.

Camera based techniques are typically less accurate but should still be capable of providing data with a resolution of a few centimetres. Many of these methCroat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

ods offer a tradeoff between accuracy and data processing time. For instance, contemporary 3D lasers can provide high accuracy and high resolution data for easy object identification, but not (yet) in real time.

3.4 Joint estimating methods The pose of an end effector can be estimated from joint values and the known geometry of the system on which the effector is mounted. The geometry can often be obtained with high accuracy from CAD/SolidWorks models, and various types of sensors can be attached to the system to monitor the rotational and translational motion of its individual joints. The resolution of an estimate obtained in this way is, according to the ISO 9283 standard, defined as the smallest incremental movement that can be sensed. For a serial manipulator of N joints, the resolution of the estimated end effector pose can be approximated as: N

Resolution = ∑di ( q ) × dqi

Where:

i =1

(1)

O. Lindroos et al.

cal joint or its displacement by exploiting the phenomenon of variable resistance: the voltage across the resistors in the sensor is proportional to the joint value.

These sensors can be placed inside the cylinder, outside the cylinder, or at the joints. Examples are shown in Fig. 1.

When sensors are mounted on the cylinders, both the joint angles and the end effector position have to be estimated based on the geometry of the machine.

The resolution of the measurement will depend on many factors, the most influential of which are the voltage range of the sensor and the number of bits in the Analog to Digital conversion (ADC). For example, only 1024 (210) different levels can be distinguished with a standard 10 bit ADC. A sensor with such an ADC, a 1 meter opening, and a voltage range of 0 to 5 volts would have a resolution of 4.88 mm according to the expression below:

Voltage range  Resolution =  × cylinder maxi 2 bits

Voltage range  Resolution =  × cylinder maximum opening  (2) di distance between the end effector endpoint 2 bits th and rotational axis of the i joint; Similarly, when the joint angle is being measured q vector of measured joint angles; the resolution in radians can be calculated as: dqi resolution of the ith sensor. Voltage range  Consequently, the resolution is non linearly depen (3) Resolution =  × 2 p dent on the joints and cannot be specified explicitly 2 bits using a single value or range. A number of sensor types can be used to estimate joint angles; some that 3.4.2 Quadrature optical encoders are widely used in robotics are discussed below.

Most mechanical manipulators have encoders 3.4.1 Variable resistance joint sensors mounted on their joints to sense their angles of rotation (or displacement in the case of prismatic joints; Sensors containing variable resistors or voltage disee Fig. 2). viders measure the angle of rotation about a cylindri-

Fig. 1 Position sensing for hydraulic cylinders: a) A linear position sensor; b) Linear position sensor on a machine (source: www.texashydraulics.com) Croat. j. for. eng. 36(2015)2

155

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

Fig. 2 Joint angle sensors with quadrature optical encoders: a) Encoder mounted on a joint; b) Rotary encoder with wire box for measuring linear motion (source: http://www.tfe.umu.se) The resolution of optical encoders depends on the number of counts they perform per revolution. Assuming a resolution of N counts per revolution, the accuracy of the measurement is given by:

1 × 2 p N

(4)

1 ×  maximumdistance  N 

(5)

Resolution =

for rotary motion, and:

Resolution =

for linear motion.

For example, for a resolution of 5000 counts/revolution, we obtain an accuracy of 0.0012 radians. The accuracy in the measurement of the end effector position can then be estimated using Eq. 1.

Fig. 3 Optoelectronic absolute position monitoring system (source: http://www.parker.com)

156

3.4.3 Opto electronic joint sensors Sensors of this kind are embedded in the joint's hydraulic cylinder (Fig. 3). A reading device is mounted on the head of the cylinder, which reads a pattern (barcode) on the piston rod. The pattern is recognized and the rod position determined by processing software. The pattern stamped on the piston rod is highly resistant to the effects of side loading, dust and rust.

Such robustness makes this technology attractive for heavy duty machinery.

The accuracy of such sensors is usually given in their specifications, and their resolution typically varies from 0.03 mm to 0.25 mm. This is more than sufficiently accurate for most applications. However, it is important to recall that the resolution of the joint angles and end effector poses determined using sensors of this type is also dependent on the geometry of the machine.

3.4.4 Optical fibre goniometer The technology used in fibre optic sensors for measuring joint angles was originally developed for the purposes of motion capture. It is therefore used extensively in biomechanics applications and the film industry. In recent years, the technology has matured and become more robust, leading to applications in more diverse contexts. Its wider uptake has been facilitated by its simplicity and low setup costs.

Electrogoniometers determine the angle of rotation about a cantilever joint by using a light sensor housed Croat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

Fig. 4 Fibre optic electrogoniometer sensors: a) Joint angle electrogoniometer sensor; b) Manipulator with built in electrogoniometer sensors (source: http://www.adinstruments.com/) in an enclosure to measure the amount of light passing through a pair of optic fibres running along the length of the cantilever (Fig. 4a). The resolution of the measurement device is typically specified by the product manufacturer, and usually varies from 0.01 to 0.1 degrees (0.000174 to 0.0017 radians).

Once again, the resolution for the ultimate measurement of the end effector position can be calculated using Eq. 1.

3.5 Inertial techniques and tilt sensors Inertial measurement units (IMUs) are used for measuring velocity, orientation, and gravitational forces. Usually an IMU is a box containing sensors of three types; accelerometers, gyroscopes and magnetometers. In modern machine applications, IMUs are attached directly to the links in a rugged robust box, as shown in Fig. 5. To avoid the overhead of installing external cables, they are often wireless.

These kinds of IMUs are typically based on mechanical principles and are equipped with electronics

O. Lindroos et al.

that generate measurements in the form of electrical signals. Assuming the measuring device incorporates an ADC, the resolution of its measurements can be estimated using Eq. 2 and Eq. 3.

A problem often encountered using IMUs is that, while velocity and acceleration are reliably measured, it is difficult to obtain reliable positional information.

This is because environmental factors (e.g. the presence of nearby metals, magnetic fields, radio signals, etc.) may cause the output of the magnetometers to fluctuate. Additional software estimation is required to cope with this problem. Observer based estimation techniques, such as Kalman filtering, are used to combine the information from the accelerometers and gyroscopes of the IMU and obtain statistically optimal estimates of the system's 3D position and angular orientation. Using these data and simple geometrical relationships, it is possible to estimate the pose of the end effector. It is difficult to quantify the resolution of these measurements because it depends on the software that is used. Algorithmic estimation can be very accurate, regardless of the sensor resolution, because it is under the programmer’s control. This is the main reason why IMU technology has been widely used for wireless sensing in diverse industries including motion capture, biomedicine, unmanned aerial vehicles and robotics.

4. Matching benefits to methods for estimating the location of the harvester head The accuracy required to realize a particular benefit is not by itself sufficient to determine which sensing technologies are best suited to deliver that benefit.

More important criteria are the time constraints and the number of positions at which sensors are required.

Therefore, we briefly outline the anticipated positioning process and then present a matching matrix that compares the various benefits achievable with head positioning data to the methods of acquiring such data.

4.1 The positioning process

Fig. 5 Joint angle sensing on a machine fitted with IMUs: a) Placement of IMUs on an excavator; b) A ruggedized IMU (source: http:// www.topconpositioning.com) Croat. j. for. eng. 36(2015)2

Different applications have different positional information requirements. In general, either point or continuous estimates are needed. Point estimates provide information on the position of the harvester head at specific points in time; they are useful in cases where it is necessary to know the position of a harvested tree, for example. Most indirect applications (e.g. those associated with ecological mapping) require point estimates. Conversely, direct benefits such

157

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

Table 1 Potential for the implementation of different position sensing methods, and the fit between each method's accuracy and that required to realize specific benefits Benefit Method

Implementation potential

Angle and/or range Joint sensors Tilt sensors IMU

Virtual Density of marking and tree removals constraining

Machine (semi) automation

Improving work methods

++

--

0

++

++

0

++

++

++

++

0

++

++

++

++

0

++

3D forest mapping

Training of ALS

Timber traceability

Product features

0

+

+

++

++

++

++

++

++

++

++

+

++

++

++

++

++

++

++

Note: ++ – highly likely to be implemented/much higher accuracy than needed; + – likely to be implemented/higher accuracy than needed; 0 – might be implemented/acceptable accuracy; - – unlikely to be implemented/less accuracy than needed; -- – very unlikely to be implemented/much less accuracy than needed.

as those associated with mapping and controlling machine movements require a continuous flow of information on the location of the head. In automation applications this flow must be supplied in real time, which inevitably requires high processing capacity.

Point estimates present less of a challenge but require precise control over which positions are estimated. This can be achieved by connecting the estimation process to specific machine commands. During harvesting, the harvester head grasps the tree when it is felled. Therefore, if the location and the direction of a point on the harvester head are known, the position of the tree's centre can be determined. The estimation of the head's position (and thus the tree's position) can be synchronized with the action of cutting the tree, during which the harvester head is held still for a second or two to enable the chainsaw to cut through the stem. In this case, the harvester head would be in the desired location during the few seconds between being moved towards the tree and the point at which the tree starts to fall. Once the tree is felled, the harvester head and tree are moved to enable subsequently cut logs to fall into separate assortment piles. The position measurements must, therefore, be acquired quickly because the head spends relatively little time in the desired position, and there is typically only 30 seconds or so between tree fellings. Moreover, once a given tree's position has been determined, it must be saved and linked to each of the logs that are subsequently cut from the tree (together with information on the tree's length, taper, species, and so on). The entire process must then be repeated when the next tree is felled.

158

4.2 Matching matrix To systemize the matching of benefits and sensing methods, we have constructed a matrix showing how suitable we consider different methods to be for the realization of specific benefits, assuming that the accuracy of the positional data depends only on the accuracy with which the position of the head is estimated relative to the machine (Table 1). This approach is adopted because we expect that the accuracy of global positioning data for machines will increase substantially in the near future. We also give our opinions on the likelihood that the different methods will be commercialized. For the sake of simplicity, we classify the methods and benefits using a five level scale, and present both in rather general terms. Nevertheless, the matrix clearly shows that there are a range of viable methods that could be used to realize most of the benefits. It also indicates that the methods that are most likely to be implemented commercially, i.e. those based on crane joint position sensors or IMUs, are capable of meeting the requirements of most benefits (Table 1).

5. Field experiment 1 – a 2D laser scanner mounted on a harvester cabin An alternative to trying to localize the harvester head is to determine the position of the tree that was just harvested. The first step in implementing such a method is to identify a practical way of creating a local map of the trees surrounding the harvester. When a Croat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

tree is harvested, we should be able to detect which tree is missing from this map. The reliability of this tree detection process can be improved by accurate estimations of tree diameter. In this section, we describe two different field experiments using a SICK LMS 221 2D laser scanner to detect trees. The angular resolution of a single laser beam emitted by the scanner was 0.25° and its field of view is 100°. Each scan consisted of 401 beams. According to the manufacturer, the laser scanner had a measurement range of up to 80 m, and a measurement accuracy of ±3.5 cm for ranges up to 20 m. However, to take advantage of this high accuracy, trees must be distinguished from other objects (e.g. brush, branches, rocks, etc.). Therefore, a key goal of the study was to evaluate the utility of data collected using a 2D laser mounted on a harvester.

In the first experiment (Hellström et al. 2012), the laser scanner was mounted on top of a harvester cabin.

Measurements were acquired at three different locations in the same forest, with varying degrees of visual obstruction due to branches, leaves, needles, and so on. To identify trees from laser scanning data, the first thing that must be done is to cluster the laser points. This was done using an algorithm developed

by Jutila et al. (2007) with minor modifications. To validate the clusters, the estimated diameter of each tree cluster has to be calculated and checked to ensure that it is within a reasonable range (between 15 and 80 cm in our study). We implemented and tested the accuracy of several different methods for calculating tree diameters from clusters in the laser scanning data.

The tree identification algorithm was found to work reasonably well even at a forest site with quite severe visual obstruction due to branches and needles.

However, a tree that was identified in one laser scan was not always detected in the next even though the scanner was not moved between scans. It also failed to detect all trees. Some trees were blocked by the harvester crane and some were blocked by other trees or branches. Several methods for enhancing tree detection were evaluated, such as using median values from several consecutive scans. Overall, however, it was concluded that 2D laser scanners are not suitable or reliable for detecting recently harvested trees due to uncertainties in tree detection efficiency. However, the detected trees did appear to have been positioned with at least cm level accuracy (although this was not investigated rigorously).

Another possible use for this tree detection method is to localize the vehicle relative to the surrounding trees. By generating a local tree map and matching it with a global map, which could be generated from Croat. j. for. eng. 36(2015)2

O. Lindroos et al.

ALS data, the machine's position can be determined more exactly than would be possible by using a GNSS (Rossman et.al 2010), which has obvious limitations in dense forests. In this scenario, it is not necessary to find all trees (or a specific one); we need only find enough trees to allow the accurate matching of local and global maps.

The results obtained using different methods for estimating tree diameter varied considerably between the three experimental sites and also between methods. The average error for the different methods varied between 40% and 90%. No method was best for all circumstances. Since estimated diameters are used to validate identified tree clusters, a more accurate diameter estimation method should increase the number of trees identified in a given laser scan and thus make it easier to filter out clusters that do not correspond to trees. To develop better methods for diameter estimation, we conducted a second experiment indoors (Ringdahl et al. 2013) using nine tree trunk sections with diameters of 6–50 cm. The tree sections were placed one at a time in an indoor corridor, at distances varying between 5 and 20 meters, with different sides facing the laser scanner. In total, we measured 172 combinations of stem section diameters and distances.

For each measurement, the tree's real diameter was manually measured with mm accuracy using a caliper at the spot where the laser beams hit. The same clustering algorithm as described above was used to identify trees in the laser scan. However, since there was only one tree in each scan, there was no need for validation of the clustering.

The algorithms for diameter estimation evaluated in the previous experiment were used again in this study. We also developed enhanced algorithms that compensate for the effect of the beam width and rely on multiple scans. The best existing algorithms overestimated the tree trunk diameters by ca. 40%. Our enhanced algorithms reduced this error to less than 12%.

The tested 2D laser scanner thus proved to be unsuitable for estimating the positions of either harvested trees or the harvester head. However, its output may be useful in improving the accuracy with which the machine's position is determined, thereby increasing the accuracy of other methods for estimating the head position relative to the machine.

6. Field experiment 2 – encoders on crane joints A very reliable method used for estimating the boom tip coordinates of forestry cranes is to monitor

159

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

Fig. 6 Forwarder crane with sensors attached at the joint level for measuring angular motion. This experimental platform has been operating for more than a decade. During this time the sensors have never been changed. Based on this demonstration of the sensing device's reliability and robustness, a commercial version of this solution has been developed (Cranab 2013b)

the angular motion of their joints. Below we present some results from a decade long series of studies (Ortiz Morales et al. 2014, Westerberg 2014), whose results contributed to the development and recent release of a commercial product (Cranab 2013b). The aim of this discussion is to illustrate the practicality of this approach to gathering data on the positions of crane booms. Previous studies (Westerberg 2014) have shown that the method's accuracy is high enough to support crane automation; using a similar approach to that embodied in Eq. 1, it was determined to be of mm level (Daniel Ortíz et al. 2014).

A Komatsu 830.3 forwarder (Komatsu Forest 2013) was used in the study. The forwarder was equipped with a CRANAB CRF 5.1 crane (Cranab 2013a), which has a reach of 9.3 m. The grapple used was a Komatsu G28, with an Indexator G121 rotator (Indexator 2013).

During the experiments, the machine was equipped with four sensors to measure its joint angles and the telescopic displacement (Fig. 6). These sensors were quadrature encoders (Heidenhain, ROD 426−5000)

160

with a resolution of 5000 pulses per revolution. As such, they provide a measuring accuracy of 0.072 degrees (0.0012 rad) for the angular joints and 0.0007 m (0.7 mm) for the telescope. The crane was also fitted with a real time data acquisition unit that can record incoming signals at a frequency of 1 KHz (1000 recordings every second).

The sensing system was used to study the working patterns of experienced forwarder operators. To this end, a group of five professional operators were asked to use the machine in their normal working routine.

Since the only additions to the machine are the sensors, the operators were free to use its standard computer to tune the machine settings to their liking. Positional data were recorded over the course of a week of scheduled working time for each operator.

The measurements provided by the joint sensors represent angular movement in radians, and displacement in meters in the case of the telescopic link. These quantities are recorded using the coordinate system sketched in Fig. 7.

Croat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

Fig. 7 Forwarder crane coordinate system and generalized joint coordinates After processing the data, a reconstruction of the boom-tip's motion was generated by considering the kinematic geometry of the machine (Fig. 8). The results clearly indicate that it is possible to fit standard forest cranes with joint estimating sensors that can withstand the challenges of forestry operations. Moreover, it also clearly shows the potential benefit of having the capacity to analyze and visualize crane work (Fig. 8). Indeed, this kind of visualization is likely to substantially improve the analysis and evaluation of experienced operators’ working methods, and the training of new operators (Daniel Ortíz et al. 2015).

7. Discussion Our analysis highlights the joint estimating principle and IMUs as the methods with the greatest potential for implementation. Both satisfy the accuracy

Fig. 8 Calculated boom tip movement patterns based on recordings of the joint angles and telescopic opening Croat. j. for. eng. 36(2015)2

O. Lindroos et al.

requirements for all of the benefits that could potentially be realized using positional data for harvester heads. The practical potential of these technologies is demonstrated by the fact that two leading forestry companies have presented solutions incorporating them since the initiation of this project. John Deere uses sensors in its Intelligent Boom Control solution to enable boom tip control, and to generate feedback on certain aspects of the working process such as the frequency at which trees are moved from one side of the machine to another (John Deere 2013). Similarly, Cranab uses sensors to provide information about the position of the crane boom relative to the machine (Cranab 2013b). As indicated in the analyses, many sensing methods could potentially be used but are less robust or well suited to heavy duty forest work. The ongoing development of recently introduced products for positioning the harvester head (i.e. boom tip) will enable the realization of most of the benefits discussed herein if end users wish to do so. However, the realization of some benefits (particularly Timber traceability and improved classification of product properties) will require the development of cost efficient methods for linking information about individual logs from different stages in the forest value chain.

To accurately determine the harvester head's position, we chose to focus on the positioning of the harvester head relative to the forest machine. However, during the analysis it was noted that the interactions between certain methods offered opportunities to simultaneously estimate local and global positioning.

Based on these findings, we believe that combined local and global approaches will ultimately be adopted to improve the accuracy of both. However, some of the benefits considered in this work will not gain from high global accuracy. For instance, the benefit that has the most stringent requirements in terms of relative accuracy, i.e. machine (semi) automation, is completely independent of the machine's global position. In contrast, ecological mapping benefits require global accuracy. However, the required level of local accuracy (i.e. accuracy in determining the head's position relative to the machine) for ecological mapping is much lower than that for (semi) automation. In the end, high accuracy in both global and local positioning will always be useful, as long as it does not come at too high a price.

Naturally, this kind of study cannot cover all of the potential benefits that may be realized by accurately measuring the position of the harvester head or a forest machine's boom tip; similarly, it would be impossible to discuss all of the methods that could be used to acquire such data. Indeed, this overview is already

161

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

lengthy enough. However, we hope that by highlighting some of these benefits and promising positioning methods, we will introduce a wider audience to the potential of this recent technical development. In future, we look forward to building on the results and ideas presented herein by examining new solutions for realizing some of the prospective benefits arising from having detailed information on the location of the harvester head.

Bollandsås, O.M., Maltamo, M., Gobakken, T., Lien, V., Naesset, E., 2011: Prediction of Timber Quality Parameters of Forest Stands by Means of Small Footprint Airborne Laser Scanner Data. International Journal of Forest Engineering 22(1): 14–23.

Acknowledgements

Cranab 2013a: Cranab - Products - Forwarder Cranes. Available at: http://www.cranab.info/www%5Ccranabcom.nsf/ pages/ProductsForwarderCranes.

This work was funded by the Royal Swedish Academy of Agriculture and Forestry (KSLA; H11-0085MEK, H11-0085-GBN). We would like to thank SeesEditing Ltd for revising the English language.

8. References Alam, M.M., Strandgard, M.N., Brown M.W., Fox, J.C., 2012: Improving the productivity of mechanised harvesting systems using remote sensing. Australian Forestry 75(4): 238– 245.

Andersen, H.E., Clarkin, T., Winterberger, K., Strunk, J., 2009: An accuracy assessment of positions obtained using survey and recreational grade global positioning system receivers across a range of forest conditions within the Tanana Valley of interior Alaska. Western Journal of Applied Forestry 24(3): 128–136.

Anon., 2010: StanForD 2010 – modern communication with forest machines. Uppsala, Sweden, Skogforsk. 16 p.

Arii, K., Caspersen, J.P., Jones, T.A., Thomas, S.C., 2008: A selection harvesting algorithm for use in spatially explicit individual based forest simulation models. Ecological modelling 211(3): 251–266.

Arshadi, M., Backlund, I., Geladi, P., Bergsten, U., 2013: Comparison of fatty and resin acid composition in boreal lodgepole pine and Scots pine for biorefinery applications.

Industrial Crops and Products 49: 535–541.

Athanasiadis, I.N., Anastasiadou, D., Koulinas, K., Kiourtsis, F., 2013: Identifying Smart Solutions for Fighting Illegal Logging and Timber Trade. In Environmental Software Systems.

Fostering Information Sharing, Springer, Berlin Heidelberg 143–153 p.

Barth, A., Holmgren, J., 2013: Stem taper estimates based on airborne laser scanning and cut-to-length harvester measurements for pre-harvest planning. International Journal of Forest Engineering 24(3): 161–169.

Berg, S., Bergström, D. Nordfjell, T., 2014: Simulating conventional and integrated stump and round wood harvesting systems: a comparison of productivity and costs. International Journal of Forest Engineering 25(2): 138–155.

162

Chunhan, L., Yushin, C., Gunhong, P., Jaeheon, R., SeungGweon, J., Seokhyun, P., Jae, W.P., Hee, C.L., Keum-Shik, H., Man, H.L., 2004: Indoor positioning system based on incident angles of infrared emitters Industrial Electronics Society. IECON, 30th Annual Conference of IEEE (Vol. 3).

Cranab 2013b: Cranab presents an entirely new generation of cranes. Available at: http://www.cranab.se/static/en/206/.

Davison, A.J., Kita, N., 2002: Active Visual Localization for Multiple Inspection Robots. Advanced Robotics 16(3): 281– 295.

Dissanayake, M.W.M.G., Newman, P., Clark, S., DurrantWhyte, H., Csorba, M.A., 2001: Solution to the simultaneous localization and map building (SLAM) problem. IEEE Trans.

Robot. Autom. 17(3): 229–241.

Eriksson, D., Lindberg, H., Bergsten, U., 2006: Influence of silvicultural regime on wood structure characteristics and mechanical properties of clear wood in Pinus sylvestris. Silva Fennica 40(4): 743–762.

Eriksson, M., Lindroos, O., 2014: Productivity of harvesters and forwarders in CTL operations in Northern Sweden based on large follow-up datasets. International Journal of Forest Engineering 25(3):179–200.

Fortin, M., Delisle-Boulianne, S., Pothier, D., 2013: Considering spatial correlations between binary response variables in forestry: an example applied to tree harvest modeling.

Forest Science 59(3): 253–260.

Gellerstedt, S., 2002: Operation of the single-grip harvester: motor-sensory and cognitive work. International Journal of Forest Engineering 13(2): 35–47.

Hakli, J., Jaakkola, K., Pursula, P., Huusko, M., Nummila, K., 2010: UHF RFID based tracking of logs in the forest industry.

IEEE International Conference on RFID. Orlando, FL, USA, 14–16 April, 245–251.

Harter, A., Hopper, A., Steggles, P., Ward, A., Webster, P., 1999: The Anatomy of a Context Aware Application. In Proc.

5th ACM MOBICOM Conf. Seattle, WA. Heidenhain, 2013: Measurement and Control Technology for Demanding Positioning Tasks. Available at: http://www. heidenhain.de/de_EN/home/.

Hellström, T., Lärkeryd, P., Nordfjell, T., Ringdahl, O., 2009: Autonomous Forest Vehicles: Historic, envisioned, and state-of-the-art. International Journal of Forest Engineering 20(1): 31–38.

Croat. j. for. eng. 36(2015)2

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

O. Lindroos et al.

Hellström, T., Ringdahl, O., 2009: Real time path planning using a simulator-in-the-loop. International Journal of Vehicle Autonomous Systems 7(1/2): 56–72.

Möller, J.J., Bhuiyan, N., Hannrup, B., 2012: Monitoring of thinning using harvester data. Resultat 1, Uppsala, Sweden, Skogforsk.

Hellström, T., Hohnloser, P., Ringdahl, O., 2012: Tree diameter estimation using laser scanner. Technical Report UMINF 12.20, Department of Computing Science, Umeå University, Umeå, Sweden.

Murphy, G., 2014: Priority list bucking on a mechanized harvester considering external properties and stiffness of Douglas-fir. International Journal of Forest Engineering 25(3): 214–221.

Holmgren, J., Barth, A., Larsson, H., Olsson, H., 2012: Prediction of stem attributes by combining airborne laser scanning and measurements from harvesters. Silva Fennica 46(2): 227–239.

Murphy, G., Clark, J.A., Pilkerton, S., 2012: Current and Potential Tagging and Tracking Systems for Logs Harvested from Pacific Northwest Forests. Western Journal of Applied Forestry 27(2): 84–91.

Huang, S., Wang, Z., Dissanayake, G., 2008: Sparse local submap joining filter for building large-scale maps. IEEE Trans.

Robot. 24(5): 1121–1130.

Naesset, E., Jonmeister, T., 2002: Assessing point accuracy of DGPS under forest canopy before data acquisition, in the field and after postprocessing. Scandinavian Journal of Forest Research 17(4): 351–358.

Hyyppa, J., Hyyppa, H., Leckie, D., Gougeon, F., Yu, X., Maltamo, M., 2008: Review of methods of small-footprint airborne laser scanning for extracting forest inventory data in boreal forests. Int. J. Remote Sens. 29(5): 1339–1366.

Indexator, 2013: Rotators and accessories – Indexator. Available at: http://www.indexator.se/en/rotator_systems/rotators_and_accessories.

John Deere, 2013: Intelligent Boom Control (IBC). Available at: http://www.deere.co.uk/wps/dcom/en_GB/industry/forestry/learn_more/ibc_en.page.

Jutila, J., Kannas, K., Visala, A., 2007: Tree Measurement in Forest by 2D Laser Scanning. International Symposium on Computational Intelligence in Robotics and Automation.

IEEE, 491–496.

Komatsu Forest, 2013: 830.3 – Sverige – Komatsu Forest .

Available at: http://www.komatsuforest.se/default. aspx?id=10876.

Lee, C., Chang, Y., Park, G,. Ryu, J., Jeong, S.G., Park, S., Park, J.W., Lee, H.C., Hong, K.S., Lee, M.H., 2004: Indoor positioning system based on incident angles of infrared emitters.

Industrial Electronics Society. IECON 2004. 30th Annual Conference of IEEE, vol. 3: 2218–2222.

Lindroos, O., 2012: Evaluation of technical and organizational approaches for directly loading logs in mechanized CTL harvesting. Forest Science 58(4): 326–341.

Mettin, U., La Hera, P.X., Morales, D.O., Shiriaev, A.S., Freidovich, L.B., Westerberg, S., 2009: Path-constrained trajectory planning and time-independent motion control: Application to a forestry crane. In Proceedings of 14th International Conference on Advanced Robotics (ICAR).

Miettinen, M., Ohman, M., Visala, A., Forsman, P., 2007: Simultaneous Localization and Mapping for Forest Harvesters. In Proceedings of the 2007 IEEE International Conference on Robotics and Automation. 10–14 April, Roma, Italy: 517–522.

Milne, B., Chen, X.Q., Hann, C.E., Parker, R., 2013: Robotisation of forestry harvesting in New Zealand – An overview.

The 10th IEEE International Conference on Control and Automation (ICCA), 12-14 Jun, Hangzhou, China, 1609–1614.

Croat. j. for. eng. 36(2015)2

Næsset, E., Gobakken, T., Holmgren, J., Hyyppä, H., Hyyppä, J., Maltamo, M., Nilsson, M., Olsson, H.K., Persson, A.S., Söderman, U., 2004: Laser scanning of forest resources: The nordic experience. Scand. J. For. Res. 19(6): 482–499.

Nordström, M., Wilhelmsson, L., Arlinger, J., Möller, J.J., 2010: Harvester data can provide important advance information to end users. Resultat 21, Uppsala, Sweden, Skogforsk.

Ortiz Morales, D., Westerberg, S., La Hera, P.X., Mettin, U., Freidovich, L., Shiriaev, A.S., 2014: Increasing the Level of Automation in the Forestry Logging Process with Crane Trajectory Planning and Control. Journal of Field Robotics, 31(3): 343–363. doi: 10.1002/rob.21496.

Ortiz Morales, D., La Hera, P., Westerberg, S., Mettin, U., Freidovich, L., Shiriaev, A., 2015: Path-constrained motion analysis. An algorithm to understand human performance on hydraulic manipulators. IEEE Transactions Journal on Human-Machine Systems 45(2): 187–199.

Ovaskainen, H., Heikkilä, M., 2007: Visuospatial cognitive abilities in cut-to-length single-grip timber harvester work.

International Journal of Industrial Ergonomics 37(9): 771– 780.

Persson, 1977: Quality development in young spacing trials with Scots pine. Swedish University of Agricultural Science, Department of Forest Yield Research, Report 45, 152 p.

Purfürst, T., Lindroos, O., 2011: The long-term productivity’s correlation with subjective and objective ratings of harvester operators. Croatian Journal of Forest Engineering 32(2): 509–519.

Ringdahl, O., 2011: Automation in Forestry – Development of Unmanned Forwarders. PhD thesis, Department of Computing Science, Umeå University.

Ringdahl, O., Hohnloser, P., Hellström, T., Holmgren, J., Lindroos, O., 2013: Enhanced Algorithms for Estimating Tree Trunk Diameter Using 2D Laser Scanner. Remote Sensing 5(10): 4839–4856.

Rodríguez-Pérez, J.R., Álvarez, M., Sanz-Ablanedo, E., 2007: Assessment of low-cost GPS receiver accuracy and precision

163

O. Lindroos et al.

Estimating the Position of the Harvester Head – a Key Step towards the Precision Forestry ... (147–164)

in forest environments. Journal of Surveying Engineering 133(4): 159–167.

Rossmann, J., Krahwinkler, P., Schlette, C., 2010: Navigation of mobile robots in natural environments: Using sensor fusion in forestry. J. Syst. Cybern. Inform. 8(3): 67–71.

Rossmann, J., Schluse, M., Waspe, R., Moshammer, R., 2011: Simulation in the woods: from remote sensing based data acquisition and processing to various simulation applications. In: Proceedings of the 2011 Winter Simulation Conference »Simulation for a Sustainable World« edited by S. Jain, R. R. Creasey, J. Himmelspach, K. P. White, and M. Fu, December 11–14, Phoenix, Arizona, 984–996.

Rossmann, J., Schluse, M., Schlette, C., Buecken, A., Krahwinkler, P., Emde, M., 2009: Realization of a highly accurate mobile robot system for multi purpose precision forestry applications. The 14th International Conference on Advanced Robotics, 22–26 June, Munich, 1–6.

Seidel, F., Fripp, E., Adams, A., Denty, I., 2012: Tracking Sustainability. Review of Electronic and Semi-Electronic Timber Tracking Technologies. ITTO Technical Series TS-40. The International Tropical Timber Organization (ITTO), Yokohama, Japan. 62 p.

Shiriaev, A., Freidovich, L., Manchester, I., Mettin, U., La Hera, P., Westerberg, S., 2008: Status of Smart Crane Lab Project: Modeling and Control for a Forwarder Crane; Technical Report; Department of Applied Physics and Electronics, Umeå University: Umeå, Sweden.

Stendahl, J., Dahlin, B., 2002: Possibilities for harvester-based forest inventory in thinnings. Scandinavian Journal of Forest Research 17(6): 548–555.

Thorpe, H.C., Vanderwel, M.C., Fuller, M.M., Thomas, S.C., Caspersen, J.P., 2010: Modelling stand development after partial harvests: An empirically based, spatially explicit analysis for lowland black spruce. Ecological Modelling 221(2): 256–267.

Yang, K.C., 2002: Impact of spacing on juvenile wood and mature wood properties of white spruce (Picea glauca). Taiwan Journal of Forest Science 17(1): 13–29.

Zheng, Y., Liu, J., Wang, D., Yang, R., 2012: Laser scanning measurements on trees for logging harvesting operations.

Sensors 12(7): 9273–9285.

Wang, Z., Huang, S., Dissanayake, G., 2005: D-SLAM: Decoupled localization and mapping for autonomous robots.

In Proceedings of the International Symposium of Robotics Research, ISRR 05, San Francisco, CA, USA, 12–15 October (Vol. 26): 203–213.

Westerberg, S., 2014: Semi-automating forest machines. Motion planning, system integration and Human-Machine interactions. PhD thesis, Dep. Applied Physics and Electronics, Umeå University, Sweden.

Wiklund, U., Andersson, U., Hyyppä, K., 1988: AGV navigation by angle measurements. Proc. 6th int. Conf. Automated Guided Vehicle System, Brussels, October, 199–212.

Smith, A., Balakrishnan, H., Goraczko, M., Priyantha, N.B., 2004: Tracking Moving Devices with the Cricket Location System, in Proc. of The 2nd Int. Conf. on Mobile Systems, Applications and Services.

Öhman, M., Miettinen, M., Kannas, K., Jutila, J., Visala, A., Forsman, P., 2008: Tree Measurement and Simultaneous Localization and Mapping System for Forest Harvesters. In Laugier, C. and Siegwart, R., Field and Service Robotics.

Springer, Berlin Heidelberg, 369–378.

Received: September 3, 2014 Accepted: January 9, 2015

Authors’ address: Assoc. Prof. Ola Lindroos, PhD.* e-mail: ola.lindroos@slu.se Pedro La Hera, PhD. e-mail: xavier.lahera@slu.se Department of Forest Biomaterials and Technology Swedish University of Agricultural Sciences SE-901 83 Umeå SWEDEN Ola Ringdahl, PhD. e-mail: ringdahl@cs.umu.se Peter Hohnloser e-mail: peterh@cs.umu.se Assoc. Prof. Thomas Hellström, PhD. e-mail: thomash@cs.umu.se Department of Computing Science Umeå University SE-901 87 Umeå SWEDEN * Corresponding author

164

Croat. j. for. eng. 36(2015)2

Original scientific paper

Survival Test of RFID UHF Tags in Timber Harvesting Operations Gianni Picchi, Martin K端hmaier, Juan de Dios Diaz Marques Abstract Traceability of wood products is more and more relying on high technology systems. Among them the Radio-Frequency IDentification with Ultra High Frequency (RFID UHF ) tags are probably the most flexible and promising tools. Several studies address their use in timber logistics, but the possibility to mark standing trees and maintain intact the information along a whole-tree extraction system is still not explored. Under this perspective one of the main challenges is the capacity of UHF RFID tags to survive the harsh conditions of timber harvesting. Different tag models and different placement positions on the tree may lead to diverse ratio of tags arriving intact up to the landing. Particularly extracting operations may play a major role in damaging or removing the tags from the trees. In the present study, two tag models and two fixing modalities were compared during three commercial hauling and one transport operation in mountain conditions. Over a total of 239 tracked tags, just 5 were lost, proving a good reliability for this traceability system. This preliminary result will serve for addressing the electronic tree/log marking method in the frame of the project SLOPE, cofunded by the EC. Keywords: RFID UHF, tree marking, hauling, survival test, cable yarder

1. Introduction In the near future, the productivity and profitability of forest timber supply chains are expected to increase rapidly by implementing the tools and methods of precision forestry (Holopainen et al. 2014). This trend is also expected in mountain forestry, where forest operations are mainly based on cable yarding systems, but the relatively low degree of mechanization of these systems represents a potential constraint to the application of high precision technologies (Cavalli 2012). This specific challenge is addressed by the project SLOPE, co-funded by the EC, with the goal to set up an integrated and innovative timber supply chain in mountain areas. In the planned work flow, different digital data sources, such as aerial sensors and terrestrial laser scanners (LiDAR), are used for the acquisition of georeferenced 3D data on the standing trees. This is elaborated with dedicated software returning the optimal tree bucking suggestions (Dassot et al. 2011, Murphy 2008), which are used by the processor for maximizing the value of the extracted timber according to the market demand. The transmission Croat. j. for. eng. 36(2015)2

of such data requires a traceability system that relates the digital information generated for the single standing tree to the actual item being harvested and transmits this data to the forest machines, making available in real time the bucking instructions for each tree. The same traceability architecture can be used to assign an ID to each log produced and link this to all the available information (measures, quality). For this purpose several solutions have been used for actual marking of trees and logs in forest operations or timber logistics, such as color marking, barcodes, QR codes and Radio-Frequency IDentification technology (RFID) (Tzoulis and Andreopoulou 2013). The latter has a wide range of applications in the field of logistics, livestock and warehouse management among others (Ferrer et al. 2010, Zhu et al. 2012) and seems to be the most promising tool for supply chain management, particularly for the capacity to be red at considerable distances: Ultra High Frequency (UHF) RFID, operating at a frequency of 868 MHz, can be detected at distances of over 10 m in optimal conditions. According to the EPC standard, RFID tags have an internal memory that al-

165

G. Picchi et al.

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

lows a content of 96 bits (EPC Class 1), included in a string of 14 characters, providing a unique ID to each tag. Such information can be used for automatically tagging an item (log or tree) in a database.

In forestry applications, RFID technology has been mostly tested for control and optimization of timber logistics (Korten and Kaul 2008). Another application area is the traceability of timber throughout the whole supply chain, where the use of RFID for log marking may increase the degree of automation of the log inventory, reduce the need for repeated measurements of the timber and increase the overall efficiency, with evident economic benefits (Hakli et al. 2010). Nevertheless, the potential of electronic marking of trees may be extended beyond the traceability of goods, which could be coupled with other contents aiming at the optimization of the forest operations.

The tests performed so far with UHF RFID tags in timber logistics prove that this technology is reliable, and that, in the appropriate conditions, tags may be read even at relatively long distances (2–4 meters).

Kaul (2010) tested the performance of several RFID tag models both in bulk reading (e.g. a full truckload of tagged logs) and single items identification. With the appropriate selection of RFID tag model and reader/ antenna layout, bulk reading can return 92% of read rates at mill gate entrance, while single log identification in optimal conditions (e.g. at the belt conveyor of the sawmill) provided 100% of read rate.

Björk (2011) claims that an economic benefit can be detected along the supply chain even with partial tagging of the whole load of logs, which still allows for a certain degree of traceability of the loads. This would also make less critical the impact of a low automated read rate and/or a low physical survival rate. Nevertheless, the full potential of RFID marking, particularly when used for transmitting instructions, can be expressed only if the reliability of the system is high.

This is related to the read rate, but mainly to the capacity of RFID tags to survive all along the production process maintaining intact its operational capacity. In this sense, forest operations are a very challenging environment. In the case of cable yarding operations, the frequent shocks, frictions and impacts of the extracted material against ground, standing or felled trees and piled timber make it reasonable to expect that part of the applied RFID tags could be lost or destroyed in the phases of bunching, extraction and landing of the marked whole trees and the subsequent logistics operations of logs.

Fig. 1 depicts the application planned, where UHF RFID tags are first placed at breast height on standing trees by the forester during the marking phase (1).

166

Only trees to be felled are marked and related to the digital forest database previously generated by LiDAR technology. The chainsaw operator in charge of felling the test site will be equipped with a light RFID reader connected via Bluetooth to the service smartphone (always carried for security reasons) and before felling the tree will read the tag on it (2). This will allow him to relate the felled tree, and the corresponding ID, to a new tag, to be placed on the cross section surface (the butt of the felled tree). This position is considered as optimal for the designed supply chain, in fact it is more protected against friction during skidding operations than the tangential surface (on the bark) and leads to an optimal reading angle (perpendicular) with antennas placed on the cable yarder carriage and on the processor head. By reading the RFID tag of the tree, this latter machine will gather in real time the cutting instructions from the digital forest database (3), guaranteeing a precise and fast bucking operation, and maximizing the value of timber assortments. Finally, the processor designed in the frame of the project will apply a new UHF RFID tag to each log produced (4), linking the ID to a new database featuring the measured dimensions (diameters and length) and the commercial quality class assessed with dedicated sensors (Sandak 2015). Timber would now be ready to make part of a dedicated management system, where the application of RFID tags may lead to cost reductions estimated in the range of 70% compared to the current systems of industrial wood procurement (Björk et al. 2011).

The actual capacity of UHF RFID tags to endure these conditions is still unexplored, particularly for the phases of concentration and hauling of trees in the forest. The aim of this study was to evaluate the performance of two tag models and two fixing options thorough the timber supply chain in steep terrain during cable yarding and logistic operations. The study also aimed at identifying the most critical stand or operational parameters that could have an influence on the survival rate of the applied tags (e.g. diameter or length of the extracted piece; distance of concentration; etc.) and possible mitigation actions for reducing its negative impact on the reliability of information flow and traceability system.

2. Material and Methods 2.1 RFID tags used The basic structure of RFID tags is the transponder, a metallic element built in different shapes, according to the purpose and the manufacturer, and the integrated circuit. These elements are fixed on a simple Croat. j. for. eng. 36(2015)2

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

G. Picchi et al.

Fig. 1 Schematic representation of the timber supply chain process with RFID tag based transmission of information

Croat. j. for. eng. 36(2015)2

well as resistant to UV, acids and salt solutions.

This tag model was available just in a limited number (n.=31), thus it was only used in a part of the survival tests; Þ Smartrac Shortdipole Monza 5. A logistic dedicated model with long reading range. This is a very simple tag model basically constituted by the transponder and the relative die cut applied in plastic colored stickers provided in reels. In

Trasponder size, mm

Overall size, including plastic cover, mm

Reading range in laboratory conditions with high power readers (2 W), m

Average unitary weight, g

Table 1 Main characteristics of the RFID tags tested

RFID tag model

plastic tape structure (possibly with one adhesive surface for application), or protected and supported with different type of cases, according to the final application. Clearly, the complexity of the final RFID structure is reflected in the size and the unitary cost.

In a previous screening phase, a large number of commercial UHF RFID tag models had been compared in order to identify the most suitable types for the specific purposes of the project, and in general for tree and log marking. A selection of 10 tag models had been tested in standing tree and marking operations, gathering the opinions and suggestions of expert professionals. Finally, two models were selected for the actual survival test over the whole production process (Table 1): Þ Wintag Flexytag UHF D7040S. This tag model was selected because it was considered particularly suitable for the forest environment. In fact it was originally designed for laundry uses, with a waterproof polyurethane cover (IP68) designed for maintaining the operative capacity at temperatures ranging from –40°C to 80°C as

Wintag Flexytag UHF D7040S

47x13

64x45

2–3

4.9

Smartrac Shortdipole Monza 5

93x11

120x25

10

1.1

167

G. Picchi et al.

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

order to make it suitable for the purposes of timber traceability this model was modified by applying a PET EVA plastic cover with thermal treatment (110 °C).

The selected UHF RFID tags are designed for being manually fixed on the trees or logs by means of a common mechanical stapler deploying 6 mm long aluminum staples. This system was chosen for its simplicity, cost effectiveness and the negligible effect of the aluminum staples in case of impact with the sharp tools at the sawmill. Furthermore, the selected tags have very low unitary weight, an important aspect considering that at the sawmill the tags are discarded together with the external parts of the log, which enters the stream of industrial residual wood. This raw material is commonly used for pulp or energy production, and in both cases the share of impurities (plastic and metal of the tags) should be reduced to a minimum if this system is to be accepted by end users.

Fig. 2 Marking method with Wintag RFID and double stapling (DW)

The second treatment was the method of RFID tag fixing on the logs: Þ A double stapling was regarded as the standard solution, being theoretically more reliable, given the double grip of the staples on the wood and the tags and the total adherence of the tag on the log surface, reducing the risk of being caught and ripped by branches and slash; Þ Single stapling was compared because it presents two potential advantages:  stapling the tag on a single spot, instead of two, simplifies greatly the application of the RFID tag on the log, particularly when this is automation system for tag placing on a prototype timber processor head,  apparently the preliminary tests conducted in more controlled conditions (timber yard) suggest that, by not adhering completely to the wood surface, the tag becomes more readable because the antenna results are less exposed to the loss due to the dielectric constant of wood (Ghahfarokhi et al. 2011). In this sense, it should be taken into consideration that, by adopting this solution, it will not be possible to fully control the reading angle in the following automatic operations, thus a circular antenna and direction insensitive transponders are to be used if this solution is preferred.

The combination of RFID tag model and fixing system resulted in three treatments: Wintag RFID tag with double stapling (DW), Shortdipole RFID tag with double stapling (DS) and single stapling (SS). Due to

168

Fig. 3 Marking method with Shortdipole RFID and single stapling (SS), note the uplifted position the limited number of Wintag RFIDs available, these were not tested with the single staple (Fig. 2–3).

Electronic reading was performed by means of a hand held UHF RFID reader CAEN qID 1240 with in-built double perpendicular antennas. The reader was interfaced via Bluetooth to a mobile phone, returning the ID of identified tags, and the average number of tags red per second. Power settings were adjusted to low level (140 mW) for avoiding the risk of multiple tags return signal. In case of difficult access to the log pile, it could be set to maximum power (500 mW), allowing in the better cases a reading distance of about 80–60 cm in operative environment. The same reader was used in laboratory for testing the tags to be used prior to the field study.

In the forest, the trees and logs were measured and marked before extraction. A progressive number was assigned to each tree and log painted with forest spray Croat. j. for. eng. 36(2015)2

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

Table 2 Parameters collected in the forest prior and during cable yarder operations Measured parameters

Description Tree/log parameters

Maximum diameter, cm

Tags were applied at the butt of the felled tree or at the main diameter side of logs and the corresponding diameter measured. If this was not accessible, measurement and tag application were performed on the minor diameter side

Length, m

Measured or estimated

Species, descriptor

The tree species were noted

Type of section descriptor

Log (processed) Top (with branches) Whole tree Load parameters

Choker position descriptor

Top of the tree or minor diameter of the log/tree section Butt of the tree or main diameter of the log/tree section

Hauling distance, m

Distance travelled by the carriage loaded

Concentration distance, m

Distance from the original position to the vertical of the carriage

Type of load descriptor

Single item per load Multiple item per load (two or more trees, sections or logs) Electronic marking parameters

Tag type descriptor

DW, Flexytag fixed with two staples DS, Shortdipole fixed with two staples SS, Shortdipole fixed with one staple

color and a RFID tag was applied. Tag model and application type was varied randomly. During extraction, a further set of parameters were recorded, describing each single load (Table 2). Notes were taken for any particular event, such as impact of the tagged area against standing trees or friction with slash and ground.

Such descriptive and numerical parameters were considered as the most representative for a following factorial analysis of the results.

G. Picchi et al.

direction almost opposite to the position of the cable line, with an angle facilitating the extraction of trees.

In this layout, the operator fixes the tree or the tree sections by applying the chokers to the butt of the plant. This position reduces the effort for the cable crane to concentrate by skidding or partial skidding the tree close to the line, where it is finally lifted and transported along the corridor. This work condition minimizes the damage to the remaining trees (Marchi et al. 2014) and, given the reduced impacts, should also be the less dangerous for the RFID tags survival.

All the hauling operations studied were performed in conifer dominated stands, but differed for a number of details, among which the dominant was related to species. In all cases the conditions were particularly challenging compared to a common operation, and a high number of factors affecting the reliability of the tagging system could be evaluated, while in case of high rate of survival, the robustness of the system could have been confirmed: Site 1: was a mixed forest dominated by fir (Abies alba Mill.) with occasional Austrian Pine (Pinus nigra Arnold) and broadleaves (beech, chestnut, maple, cherry). The silvicultural work was partially an emergency operation due to storm damage on an already planned selective cut, with the dominant layer of conifers (broadleaves were processed and extracted only in case of damage). Most of the trees were felled over two months before the extraction operations, which were postponed for weather conditions (snow). The terrain presented a regular profile and slope. All the activities were performed by a crew Table 3 Main characteristics of the cable yarding lines studied during the tests Site 1 Area Altitude above sea level, m Coordinates of the cable yarder tower Average slope, %

Site 3

953

843

1079

N44 06.719 N44 05.830 N44 08.887 E11 16.721 E11 09.572 E11 19.353 35–45

30–40

40–50

12

6-8

14

Direction of extraction

Uphill

Uphill

Downhill

2.2 Cable yarder extraction

Average extraction distance, m

185

235

215

The capacity to endure the operations of lateral yarding and extraction of RFID tags was tested in three commercial cable yarding operations in the Italian Apennines (Table 3). In a properly planned and conducted cable crane extraction, the trees are felled in a

Average concentration length, m

12.3

15.4

12.6

Average log/tree diameter, cm

41

37

42

Minimum log/tree diameter, cm

27

22

20

Maximum log/tree diameter, cm

56

57

80

Croat. j. for. eng. 36(2015)2

Average mainline height, m

Site 2

Firenzuola Montepiano Firenzuola (Firenze) (Prato) (Firenze)

169

G. Picchi et al.

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

Fig. 4 Landing of cable yarder with limited operational size. In such conditions post extraction handling may cause further impacts on the tagged surfaces experienced in forest harvesting and cable yarder extraction (also operating in site 3).

The unloading area was limited, posing a further challenge to tag survival, since the probability of collision among trees unloaded or handled and the piled timber is higher (Fig. 4). This also led to a more complicate control of RFID tags survival with the handheld reader. For this reason, it was not possible to check electronically all of the trees, even if the tag and the tree number was visible. In these cases, the presence of the tag, the apparent condition (damaged or not) and the relative tree number were recorded.

Clearly, this issue would not occur if forest machines equipped with RFID readers will be used, since each tree would be grabbed and read individually.

Site 2: was a pine dominated site (Pinus nigra Arnold) of a final regeneration cut for facilitating the natural renovation of broadleaves. Very few standards were left, reducing the interference with the concentration of trees. This area was not affected by storm damage. The terrain profile had an unfavorable convex slope, which led to a low average height of the skyline in spite of the three intermediate supports installed.

The crew was composed of professional chainsaw operators, but inexperienced in the use of cable yarders for timber extraction. For this reason, most of the loads were skidded for almost 50% of the distance, with logs completely lying on the ground, in conditions very similar to winch extraction. In several occasions, the butt of the tree impacted against standing trees, protruding stones, already processed logs and the abundant slash left on the ground.

Site 3: The site was dominated by fir (Abies alba Mill.) with occasional beech. The whole area had been affected by a wind storm, which damaged over 50% of the original standing trees (completely uprooted to partially damaged by the fall of other trees), and the forestry activity can be described as an emergency intervention. Terrain profile was quite irregular, with two changes of convexity/concavity over stretches. In such conditions, trees were hooked to the chokers either at the butt or the top, according to their position and the easiest solution for untangling trunks and crowns. This made the lateral yarding quite difficult, with frequent scratching against standing and lying trees (Fig. 5). For the same reason, the specific work conditions of this forest operation were considered as challenging for the survival of tags, which were exposed to additional frictions and shocks compared to a properly planned extraction.

2.3 Intermediate Transport of logs by tractor

Fig. 5 Cable yarder concentration in a storm damaged forest is very challenging for the survival of tags due to the abundance of debris, deadwood and uprooted stumps, which could rip or damage the transponder (the white arrow indicates the position of the RFID tag on the log)

170

In timber logistics, the most critical conditions for the survival of RFID tags are expected to occur during log handling, mainly represented by loading and unloading operations. The endurance of tags to these activities was observed in a single commercial operation at the previously mentioned site 1, where timber was transported from the landing area of the cable yarder to an intermediate storage area over a distance of approximately 300 meters on a steep downhill forest road. The tractor towed a small single axe forest trailer, with tilting unloading system, which was loaded at the cable yarder landing by the processor in grapple mode.

Before loading, the dimensions of each log were measured by the crew, involving additional handling.

On the timber piled at the landing, RFID tags were applied on the log side facing the road, while loading Croat. j. for. eng. 36(2015)2

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

G. Picchi et al.

Logs with a diameter lower than 24 cm were not considered for the study (not tagged) but counted on the load since their presence could be one of the factors influencing the survival of tags on the marked logs. In the study, 7 round trips of the tractor followed, transporting a total of 198 logs, of which 86 carried a UHF RFID tag.

3. Results 3.1 Extraction operations The general performance of RFID tags along the extraction process of trees or logs was positive (Table 4). Overall 5 RIFD tags were lost during the extraction of 153 marked items, resulting in a mortality of 3.3%.

None of the tags arrived at the landing had been made inoperative, even when the whole of the hosting crosscut area presented evidences of shocks, abrasion or impacts. In some cases, the tag was not visible for the adherence of mud or soil on the crosscut section, but still it was possible to detect it with the dedicated reader. A partial removal was observed for 2 DS tags, i.e. one of the stapling areas was ripped, while the other still kept the tag more or less firmly. In any case, these tags were successfully read at the landing, and thus accounted within the survived. Given these proofs of resistance, the 6 RFID tags that could be checked only visually by the operator, and presented no signs of damage, were added to the total of success-

Fig. 6 Unloading was performed by tilting the trailer against already piled logs, causing further impacts potentially lethal for the tags

on the trailer was made in order to optimize and make stable the load, thus the tagged sides were randomly distributed in front and on the back of the trailer. At the storage area, logs were unloaded by tilting the trailer and sliding the load against the timber piles.

Also in this case, the work conditions were particularly challenging for the survival of tags (Fig. 6).

In this test, only DS and SS tagging systems were used. Not all of the logs in the pile were marked, either because below the diameter of 24 cm, set as minimum threshold, or because not safely accessible.

Table 4 Results of the reading test at landing (cable yarding) and at the storage area (tractor transport) Cable yarding

RFID tag type

Marked

SS

Site 1

Site 2

Site 3

Operative

Removed

Destroyed

Not damaged

Not visible

19

16

0

0

3

–

DS

23

20

0

0

3

–

DW

16

16

0

0

–

–

SS

19

19

0

0

–

–

DS

18

18

0

0

–

–

DW

15

14

1

0

–

–

SS

23

21

2

0

–

–

DS

20

18

2

0

–

–

153

142

5

0

6

–

SS

44

32

0

0

8

4

DS

42

31

0

0

8

3

86

63

0

0

16

7

TOTAL Tractor transport TOTAL

Croat. j. for. eng. 36(2015)2

Just visual control1

Electronic/visual control

171

G. Picchi et al.

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

fully landed tags. It should be emphasized here that in the work system envisaged, each RFID tag will be read automatically by the processor when grabbing single trees, thus averting the case of tags out of reach.

In any case, even if the tags out of reach for manual reading are excluded from the analysis, the mortality rate increases just to 4.3%.

Due to the very low number of lost tags, it is not possible to perform a statistical analysis of the results against the recorded parameters. On the other hand, this is per se a result, proving the high reliability of the system.

3.2 Transport of logs The results for the transport test have to be considered as preliminary for the relatively reduced number of logs. Nevertheless, due to particularly harsh conditions, the results can be regarded as particularly encouraging. In fact, over the fully verified tags (visually and electronically), none had suffered damage or was removed. A part of the tags was visible under the pile, but not reachable by means of the RFID reader. Two tags had visibly suffered an impact during unloading.

Electronic check (by means of the RFID reader) was possible just for one of these (single stapling), which resulted unharmed and perfectly operating. It is thus reasonable to add the tags (visually identified) among the survived tags, since during all of the trials the event of a tag still attached but made inoperative by shocks never occurred, and it can be considered as extremely unlikely in tags that do not even show sign of impacts or scratches. In any case, even considering this last tag as destroyed, the overall survival rate would be above 98%.

4. Discussion 4.1 Extraction operations Considering the factors evaluated and the direct observations of the operators, it is possible to identify the main cause of tag removal: the position of choker with respect to the crosscut section, where the RFID tag is applied. Of the 5 missing tags, 4 were placed on the side opposite to the choker position. Furthermore, all of the missing tags were lost in site 3, the most affected by storm damage. This forced the operators to set the chokers at the most accessible position, rather than the most appropriate for the cable yarding extraction itself. As a result, over 31% of logs/trees were hooked from the top or minor diameter end, opposite to the RFID tag position. However, this factor alone cannot explain the tag removal since in sites 1 and 2

172

the hooking position opposite to tag placing occurred in 19 and 17% of cases, respectively. When the tags are lost, the choker setting position appears to combine with other factors, the main being the specific layout of the cable line. In fact, the extraction in Site 3 was performed in downhill direction, and the combination of skyline height and terrain profile caused the hanging trees/logs to brush violently and repeatedly against the ground with the crosscut surface hosting the RFID tags. This happened mostly at the landing, but also occurred in other spots, according to the length of the extracted trees/logs. This combination of factors (loads hooked from the minor diameter or the crown and a convex unloading area) can be regarded as quite unusual and particularly unfavorable to RFID survival.

Nevertheless, the survival rate for this specific site was still above 90%, confirming the reliability of the tag marking system.

The RFID tag model and the method of application seem to have no influence on the final survival rate. In fact 2, 2 and 1 were lost, respectively, of the types DW, SS and DS. In this scenario, the method SS, representing the simplest UHF RFID tags and the fastest application system, should be preferred.

4.2 Transport of logs The test suggests that no difference can be found in single or double stapling fixing of the tags, thus the first should be preferred. This result is particularly encouraging for the development of an automated RFID tag application system to be installed on the timber processor. In fact, the possibility to fix the tag with just a single staple simplifies greatly the design of the machine.

By observing the operations, the loading phase does not appear to be a dangerous operation for the tag survival in the specific tested position, even if accidents, such as impact against other logs or the trailer posts, do occasionally happen. On the contrary, in the given case unloading seems to be a particularly challenging operation, since logs are landed by tilting the trailer, thus sliding down the timber, which potentially hits the ground with the radial face or the pile of logs previously unloaded.

It is important to note that all along the logistics of logs, the position of the tagged extreme cannot be certain. The RFID tag is necessarily applied to one extreme of the log: during loading and unloading operations this extreme can be positioned on the loading cage of the trailer/forwarder/truck and on the pile in a casual position. As a result, the tags will be partly facing one side of the arrangement and partly the opposite side. This issue shall be taken into account when Croat. j. for. eng. 36(2015)2

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

designing the following traceability systems for automatic detection of UHF RFID tags in the load, and particularly when planning the position of fixed antennas for bulk reading.

5. Conclusions The results of the test are very positive. The overall survival rate of RFID tags in forest operations and timber logistics was close to 98%. Considering just the cable yarding operations, the survival rate decreased slightly to almost 97%. In the single cable yarding operation, where RFID tag casualties were recorded, the survival rate was 91%, while 100% of tags successfully arrived at the landing in the other two sites observed.

Tractor transportation and timber piling led to no detectable damage or losses to the RFID tags, nevertheless, due to the limits of the manual RFID reading system used for this test, 28% of the tags could be verified just visually, or the tagged area could not be found at all in the piled timber, leaving a degree of uncertainty. This would not happen in the fully mechanized reading system under development, since each tag would be automatically red during extraction and processing of the trees.

The study confirms that large RFID tag models, with longer read range, can be used without incurring a higher risk of information loss. Furthermore, a single stapling position is enough to assure a stable grip on the crosscut surface, simplifying the tagging operations in forest (manual) and in the subsequent automated processes.

It can be conclude that simple RFID tags fixed on the trees with common aluminum staples can be regarded as a reliable tool for tracing or transmitting information all along the extraction process of timber from the forest to the landing and from there to the end user. This system provides a powerful and relatively inexpensive device for control and optimization of forest operations and timber logistics and its applications should be further tested in different work conditions and extraction systems.

Acknowledgements This work has been conducted within the framework of the project SLOPE receiving funding from the European Union’s Seventh Framework Programme for research, technological development and demonstration under the NMP.2013.3.0-2 (Grant number 604129). The authors wish to thank Dr. Mirando Di Croat. j. for. eng. 36(2015)2

G. Picchi et al.

Prinzio for his essential role in the organization of the field study and Mrs. Carolina Lombardini for her valuable support and suggestions in data collection.

6. References Björk, A., Erlandsson, M., Häkli, J., Jaakkola, K., Nilsson, Å., Nummila, K., Puntanen, V., Sirkka, A., 2011: Monitoring environmental performance of the forestry supply chain using RFID. Comput. Ind. 62(8): 830–841.

Cavalli, R., 2012: Prospects of research on cable logging in forest engineering community. Croat. J. For. Eng. 33(2): 339– 356.

Dassot, M., Constant, T., Fournier, M., 2011: The use of terrestrial LiDAR technology in forest science: application fields, benefits and challenges. Ann. For. Sci. 68(5): 959–974.

Ferrer, G., Dew, N., Apte, U., 2010: When is RFID right for your service? Int. J. Prod. Econ. 124(2): 414–425.

Ghahfarokhi, S.S., Prasad, S., Tayari, D., 2011: Impact of Moisture Content on RFID Antenna Performance for WoodLog Monitoring, in: RFMTC11 October 4–5th. Gavle, Sweden, 3–5 p.

Hakli, J., Jaakkola, K., Pursula, P., Huusko, M., Nummila, K., 2010: UHF RFID based tracking of logs in the forest industry, in: RFID, IEEE International Conference on RFID, April 12– 14. Orlando, USA, 245–251 p. doi:10.1109/RFID.2010.5467272 Holopainen, M., Vastaranta, M., Hyyppä, J., 2014: Outlook for the next generation’s precision forestry in Finland. Forests 5(7): 1682–1694.

Kaul, C., 2010: Auto-ID in timber supply chain – identifying single logs using RFID tags. FORMEC Conference, July 11– 14. Padova, Italy, 1–7 p.

Korten, S., Kaul, C., 2008: Application of RFID (Radio Frequency Identification) in the Timber Supply Chain. Croat. J. For. Eng. 29(1): 85–94.

Marchi, E., Picchio, R., Spinelli, R., Verani, S., Venanzi, R., Certini, G., 2014: Environmental impact assessment of different logging methods in pine forests thinning. Ecol. Eng. 70: 429–436.

Murphy, G., 2008: Determining Stand Value and Log Product Yields Using Terrestrial Lidar and Optimal Bucking: A Case Study. J. For. 106(6): 317–324.

Sandak, J., 2015: Slope project approach to estimate log quality by measuring cutting forces. International Wood Machining Seminar. 14–17 June, Quebec City, Canada.

Tzoulis, I., Andreopoulou, Z., 2013: Emerging Traceability Technologies as a Tool for Quality Wood Trade. Procedia Technol. 8: 606–611.

Zhu, X., Mukhopadhyay, S.K., Kurata, H., 2012: A review of RFID technology and its managerial applications in different industries. J. Eng. Technol. Manag. 29(1): 152–167.

173

G. Picchi et al.

Survival Test of RFID UHF Tags in Timber Harvesting Operations (165–174)

Authors’ address: Gianni Picchi, PhD.* e-mail: picchi@ivalsa.cnr.it CNR-IVALSA via Madonna del Piano 10 50019 Sesto Fiorentino ITALY Martin Kühmaier, PhD.* e-mail: martin.kuehmaier@boku.ac.at University of Natural Resources and Life Sciences, Vienna Department of Forest and Soil Sciences Institute of Forest Engineering Peter Jordan Strasse 82 1190 Vienna AUSTRIA

Received: May 29, 2015 Accepted: July 31, 2015

174

Juan de Dios Díaz Marqués, PhD.* e-mail: juan.diaz@itene.com ITENE - Instituto Tecnológico del Embalaje, Transporte y Logística Area: Advanced Intelligent Systems C/ Albert Einstein, 1 46980 Paterna, Valencia SPAIN * Corresponding author Croat. j. for. eng. 36(2015)2

Original scientific paper

Automated Time Study of Forwarders using GPS and a vibration sensor Martin Strandgard, Rick Mitchell Abstract Manual time and motion studies are the most common method to collect forest harvesting machine performance data. However, manual methods require skilled observers and are generally limited in duration, making it difficult to obtain a sufficiently large sample for machines with long cycle times such as skidders and forwarders. Of the automated data capture techniques studied previously, few have the breadth and ease of application to conduct long term autonomous studies for a range of harvesting machines. Analysis of Global Positioning System (GPS) data has been successfully trialled previously to conduct time studies of comparable accuracy with skilled observers, however, these approaches have been limited by the need for a degree of manual data processing.

The current study trialled a fully automated system using analysis of GPS and vibration sensor data to estimate cycle times and time elements, and compare them with those determined using traditional time and motion studies for three forwarders at different sites. The mean difference between the cycle times estimated by the two methods was <1 second. This demonstrated the automated system’s ability to accurately determine each log landing location and extent and each work cycle start and end points. The correspondence between time elements using each approach was poorer. This was mainly caused by mislabelling of brief periods by the automated system as loading events when the forwarder slowed to negotiate steep areas at one study site. These errors may be able to be addressed by adding further rules to the automated system.

Keywords: forwarder, global positioning system, multidat, automated time study, vibration sensor

1. Introduction Time and motion studies of forest harvesting machines are an important component of forest operations research. In the last four years, over 20% of the articles published in the Croatian Journal of Forest Engineering and the International Journal of Forest Engineering were based on the results of time and motion studies. However, traditional manual time and motion studies of forest harvesting machines are typically time consuming, costly, limited in duration and involve potentially hazardous work in close proximity to heavy machinery. Direct observation of harvesting machines also requires skilled observers in order to minimise data collection errors (Nuutinen et al. 2008) and can bias study results by influencing the operaCroat. j. for. eng. 36(2015)2

tor’s performance (the »Hawthorne Effect« (Hogg 2009, Magagnotti and Spinelli 2012)).

Automated data capture approaches enable collection of long term, detailed machine performance data without bias caused by the presence of an observer.

The difficulty in using an automated approach for forest harvesting time and motion studies lies in the variability in the number and type of work elements making up a work cycle. For example, a forwarder may load logs from several locations in the harvesting area, may commence at one landing and finish at another, and may unload logs onto a log stack or onto a waiting truck or a combination of these operations. In recent times, on board computers (OBC) collecting data to the StanForD standard (Skogforsk 2012) have become almost ubiquitous on forest harvesters and a number

175

M. Strandgard and R. Mitchell

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

of studies have been published using harvester OBC data (e.g. Purfürst 2010, Strandgard et al. 2013). However, OBCs are still rarely installed on other harvesting machines. Other automated approaches that have been used to collect harvesting machine performance data include John Deere’s TimberLink system (Gerasimov et al. 2012), CANBus (Controller Area Network) signal monitors (Nuutinen et al. 2008), dataloggers such as FPInnovation’s MultiDat (Davis and Kellogg 2005), and Global Positioning System (GPS) data loggers (McDonald and Fulton 2005). McDonald and Fulton (2005) suggested that automated time study technology needs to meet a number of requirements before it can be used as a research tool. The technology must: Þb e simple to install to minimize downtime for participants in production studies; Þb e useful across the widest possible range of machinery systems without requiring extensive reconfiguration of the data collection system for every machine and site; Þ s urvive under harsh operating conditions; Þp roduce data that duplicates that produced by a skilled field crew working on site.

Dupré (2006) conducted a GPS time and motion study on skidders that involved completely manual interpretation of the GPS data, whereas McDonald and Fulton (2005), in their time and motion study of skidders, manually defined site specific features, such as a polygon defining the log deck boundaries, as an initial step prior to automated extraction of machine activities from the GPS data. Manual entry of site specific features, such as used by McDonald and Fulton (2005), may require repeated site visits by the researcher or data collection by a member of the harvesting crew or a supervisor. Ideally, an automated time and motion study system would dispense with these manual components and extract details of machine activities directly from analysis of the GPS data.

Most automated data collection approaches fail to meet all the requirements: the TimberLink system is limited to late model John Deere machines; CANBus signal monitors need reconfiguration for different machine types, brands and models, and; to achieve its full potential, the Multidat needs to be hardwired into the machine and can require operator input. GPS data loggers are theoretically able to meet all the requirements (at least for skidders and forwarders). However, to meet the last requirement, requires development of methods to automatically interpret machine activities from the GPS data.

Use of GPS data in forest harvesting research is particularly suited to monitoring the activities of primary transport machines, such as skidders and forwarders, because their ability to move rapidly and cover large distances makes them difficult subjects for traditional time and motion study techniques and their movements, location and speed largely define the activities they perform. In addition, their long cycle times relative to harvesting machines requires a longer period of time to collect a statistically sufficient sample size, particularly for forwarders. A number of studies have used GPS data to interpret the activities of harvesting machines, primarily skidders (e.g. Veal et al.

2001, McDonald and Fulton 2005, Cordero et al. 2006, de Hoop and Dupré 2006). The major limitation of these studies is that they have required one or more manual steps to analyse the GPS data. de Hoop and

Three sites were used in the study (Table 1). Two sites were in short rotation Eucalyptus globulus plantations being clearfelled for chiplogs (each studied for a part day) and one was in a thinned Pinus radiata plantation being clearfelled for sawlogs and pulplogs (studied over two consecutive days). Total observation time was 17.5 hours. The weather was fine and sunny for all four days. Different forwarders (Table 1) and operators were studied at each site. All the operators were experienced.

176

The objective of the current study was to determine whether a fully automated time study system (ATSS) could be created to analyse GPS and vibration sensor data from a forwarder to accurately estimate the forwarder total cycle time and the type and duration of individual time elements.

2. Material and methods

Cycle time started when the forwarder commenced travelling empty from a log landing, and ended when the forwarder had completed unloading and was about to start travelling empty. Cycles were divided into the following time elements: »Travel empty«, »Loading«, »Moving during loading«, »Travel loaded«, »Unloading«, »Movement during unloading« and »Delays« (Table 2).

2.1 Automated time study system Multidat data loggers equipped with an internal GPS receiver (Garmin GPS 15 (12 parallel channels, accuracy <15m 95% of the time)) were installed in the cabin of each forwarder, with a magnetic base antenna on the cabin roof, to record GPS data. The GPS was set to record a point every 30 seconds and every 20 metres.

During testing of the ATSS prior to the trial, it was found that the ATSS could not reliably detect delays Croat. j. for. eng. 36(2015)2

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

M. Strandgard and R. Mitchell

Table 1 Site and forwarder details Location

Central Victoria

South-west, Western Australia

South-west, Western Australia

Eucalyptus globulus

Eucalyptus globulus

Pinus radiata

0.16

0.14

1.4

Age

12

10

32

Slope, °

<5

<5

7–24

Valmet 890.2, >10.000 engine hours

Valmet 890.4, 3500 engine hours

Valmet 890.3, >9000 engine hours

Species 3

Mean tree volume, m

Forwarder

Table 2 Forwarder time element definitions »Travel empty«

Starts when forwarder commences travel into the harvest area from the log landing and ends with start of the first crane movement to collect logs

»Loading«

Starts with commencement of crane movement to collect logs and ends when the forwarder commences another element. Includes adjustments to the logs on the bunk

»Moving during Movement between log piles with no crane movement. Starts when the wheels begin to rotate and ends when crane recommences movement. Simultaneous crane and wheel movement is recorded as loading loading« »Travel loaded«

Starts when travel to the log landing with a load and ends when wheels cease to rotate or grapple commences to move at the log landing

»Unloading«

Starts with commencement of crane movement, with an empty grapple, towards the forwarder bunk and ends when the forwarder commences another element. Includes adjustments to the log stack

»Moving during Movement between log stacks at the log landing with no crane movement. Starts when the wheels begin to rotate and ends when the crane recommences movement to the forwarder bunk. Simultaneous crane and wheel movement is recorded as unloading unloading« »Delay«

Any interruption causing the forwarder to cease working during a shift

using only GPS data. To overcome this limitation, the output of the Multidat internal vibration sensor was used. For each forwarder, the duration of each work cycle and type, and duration of each work element were determined through analysis of the GPS and vibration sensor data by the ATSS. The methodology used by the ATSS to analyse the GPS and vibration sensor data was as follows: Step 1 (Determine the location and extent of each log landing) The GPS used in the study recorded data to 5 decimal places – equivalent to an on-ground resolution of approximately 1 metre. The ATSS tallied the number of times each pair of GPS coordinates was recorded within a GPS dataset. GPS points with a high tally count relative to the remainder of the harvesting area were primarily log landings or log pickup areas. The ATSS labelled GPS points as part of a log landing if their tally count exceeded a user defined limit. Adjacent log landing GPS points were flagged as part of the same log landing. The boundary of each log landing was defined by a four sided polygon generated by Croat. j. for. eng. 36(2015)2

the ATSS to encompass each region of high GPS point density. Polygon boundaries were extended by 5 m to allow for noise in the GPS data.

Step 2 (Find start of first work cycle) The forwarder was defined as being at a landing if its GPS coordinates were within one of the log landing polygons defined in »Step« 1 and its speed and distance travelled between consecutive GPS points fell below the user defined thresholds (1 kmh-1 and 8 m, respectively). This definition allowed for occasions when the forwarder travelled through log landings without stopping. Cycles began and ended at a log landing so the ATSS detected the start of the first work cycle as the first instance when the forwarder was at a log landing at a GPS point and then travelling at the next GPS point.

Step 3 (Identify forwarder work elements) »Travel empty/loaded« and »Moving during unloading/loading« were identified when the forwarder speed and distance travelled between GPS points exceeded user defined thresholds. »Travel empty« and

177

M. Strandgard and R. Mitchell

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

2.2 Time and motion studies At the two Eucalyptus globulus sites, forwarder elemental times (Table 2) were recorded by single observers using the TimerPro Professional software (www. acsco.com) installed on a Personal Digital Assistant (PDA). At the Pinus radiata site, the forwarder activities were captured using a digital video camera and elemental times were later recorded from the video recordings using the TimerPro Professional software.

2.3 Data analysis

Fig. 1 Example of GPS points representing a forwarder cycle showing »Travel empty«, »Loading«, »Travel loaded« and »Unloading« elements and a log landing

»Travel loaded« were distinguished by whether the forwarder had been identified as »Loading« since it was last at a log landing. »Moving during unloading« or loading was determined as travel between »Loading« or »Unloading« occasions, respectively. »Loading« was identified as periods when the forwarder was not at a landing and its speed and distance travelled fell below user defined thresholds. »Unloading« was identified as periods when the forwarder was within a log landing polygon and its speed and distance travelled between GPS points were below user defined thresholds. »Delays« were identified as periods when the Multidat vibration sensor recorded that the vibration had dropped below the user defined threshold for working.

The key to the operation of the ATSS is accurate identification of the location and extent of each log landing. As the ATSS was designed for long observation periods, the observation periods for the study may have been too short to accurately identify all the log landings.

Therefore, data from a longer time period than that used for the manual time and motion (T&M) studies at each site were analysed by the ATSS and the results corresponding to each T&M study period were extracted. Fig. 1 shows an example of the GPS points representing a forwarder cycle and a log landing.

Extraction distance in each cycle was also estimated by the ATSS by adding the estimated distance between consecutive GPS point coordinates (Mean extraction distance = 359 m, Range = 152–868 m).

A comparison with manual distance estimation approaches was not made in this study.

178

The ATSS and T&M data for forwarder cycle times and »Loading«, »Unloading«, »Travel empty«, »Travel loaded« and »Delay« elemental times were compared using the Bland Altman Method (Bland and Altman 1986). Mean bias (mean difference between the T&M and ATSS values), limits of agreement (bias ±1.96 x standard deviation of the bias (SD)) and percentage error (1.96 x SD divided by the mean ATSS and T&M cycle or elemental times) were calculated for cycle and elemental times. The acceptable percentage error limit was set at ±30% (Critchley and Critchley 1999). Root Mean Square Errors (RMSE) were also calculated. »Moving« during loading and »Moving« during unloading elements were excluded from the analysis as they were minor components of the forwarder work cycles and did not occur in every cycle.

3. Results 3.1 Cycle times Thirty one forwarder cycles were recorded by the ATSS and manual T&M (8 per Eucalyptus globulus site, 15 at the Pinus radiata site). At the Pinus radiata site, the GPS signal was lost for periods of approximately 5 hours on each study day (study data was collected prior to GPS signal loss on each day). Analysis using Trimble’s mission planning software (http://ww2.trimble.com/planningsoftware_ts.asp) suggested the signal was lost due to occlusion of several GPS satellites by the hill the forwarder was working on. When the GPS signal was available, the ATSS was able to detect 100% of the forwarder cycles.

Individual forwarder cycle times estimated from the ATSS and T&M were very close with a mean difference of less than one second (Fig. 2 and Table 3). The percentage error for the ATSS and T&M cycle time differences was well within the limit of acceptability.

3.2 Elemental time »Unloading« times and »Delay« times were the most consistent elemental times between the ATSS and Croat. j. for. eng. 36(2015)2

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

M. Strandgard and R. Mitchell

Fig. 2 a) Plot comparing ATSS and T&M cycle times (1:1 line shown); b) Difference in cycle times (T&M–ATSS) (%) against the mean of ATSS and T&M cycle times with mean difference (short dashes) and limits of agreement (long dashes) T&M estimates (Fig. 3a, Fig. 3e, Fig. 4a and Fig. 4e) and »Travel empty« and »Travel loaded« times (Fig. 3b, Fig. 3c, Fig. 4b and Fig. 4c) were the least consistent.

Percentage error values reflected the differences in consistency, with the percentage error values for »Unloading« and »Delay« times being within the limit of acceptability, whereas those for the »Travel empty« and »Travel loaded« times were well outside the limit of acceptability (Table 3). The major cause of the variation between the ATSS and T&M travel times were instances when the operator stopped or slowed the forwarder while travelling, which were interpreted by the ATSS as a Loading event, which caused the mislabelling of subsequent travel times. Although »Loading« times (Fig. 3d and Fig. 4d) were reasonably con-

sistent between the ATSS and T&M times, the percentage error for this element was outside the limit of acceptability. The outliers resulted from a number of »Movement« during loading events that were recorded in the T&M study, but was just below the ATSS speed threshold used in the study (1 km h–1).

4. Discussion The ATSS analysis detected all 31 forwarder cycles observed during the corresponding traditional T&M studies and accurately estimated cycle times compared with the results of the T&M studies, however the correspondence between individual time elements was poorer. This is comparable to the findings of Mc-

Table 3 Mean, limits of agreement, percentage error and RMSE of cycle times and elemental times for travel empty, loading, travel loaded, unloading and delays (minutes) Cycle or elemental time

Mean (ATSS)

Mean (T&M)

Limits of agreement

% error

RMSE

»Cycle time«

34.2

34.2

–1.0 to 1.0

2.9

0.48

»Travel empty«

2.9

3.5

–2.2 to 3.4

86.4

1.5

»Loading«

11.0

10.4

–4.9 to 2.16

39.4

2.2

»Travel loaded«

2.6

3.0

–3.4 to 4.2

130.8

1.9

»Unloading«

9.8

9.2

–3.2 to 2.2

27.9

1.4

»Delay«

5.6

5.9

–1.3 to 2.1

29.2

0.96

Croat. j. for. eng. 36(2015)2

179

M. Strandgard and R. Mitchell

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

Fig. 3 Plots comparing ATSS and T&M elemental times a) »Unloading time«; b) »Travel empty time«; c) »Travel loaded time«; d) »Loading time«; e) »Delays«

180

Croat. j. for. eng. 36(2015)2

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

M. Strandgard and R. Mitchell

Fig. 4 Difference in elemental times (T&M–ATSS) (%) against the mean of ATSS and T&M elemental times with mean difference (short dashes) and limits of agreement (long dashes) a) »Unloading time«; b) »Travel empty time«; c) »Travel loaded time«; d) »Loading time«; e) »Delays« Croat. j. for. eng. 36(2015)2

181

M. Strandgard and R. Mitchell

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

Donald and Fulton (2005) in their GPS based automated time study of skidders. The good correspondence between the ATSS and T&M cycle times in the current study probably reflected the well-defined cycle start/end point, which has a strong contrast in activity when the forwarder finishes unloading at a log landing and commences travelling empty (typically operating at its highest travel speed (Stankić et al.

2012)).

The good correspondence between the ATSS and T&M cycle times also suggested the ATSS was able to accurately identify the location and extent of the log landings at the three sites. As mentioned previously, this step is critical to the operation of the ATSS. Potential log landing identification errors are: Þ erroneously labelling an area as a log landing (false positive); Þ failing to detect a log landing (false negative).

However, for a false positive to impact the analysis, the forwarder speed and distance, when travelling through an incorrectly labelled area, would need to be below the threshold values. This is likely to be a rare occurrence, although it is theoretically possible that a terrain feature could concentrate the forwarder activity in an area, and slow it sufficiently for it to be both incorrectly labelled as a log landing and for travel to be recorded as unloading. A false negative could occur when a log landing is used infrequently or the GPS data were collected over part of the harvesting operation. For the latter reason, GPS data collected over a period greater than that for the T&M study were used to generate the ATSS results. There was no evidence of false positives or false negatives occurring during the study.

As noted above, the primary cause of the large percentage errors for the differences between the ATSS and T&M »Travel empty« and »Travel loaded« elemental times was the forwarder stopping or slowing during travel. This was interpreted by the ATSS as loading, resulting in mislabelling of subsequent travel as »Moving during loading or unloading«. These errors mainly occurred during the Pinus radiata study and were the result of the forwarder manoeuvring carefully on steep areas. Reanalysing the data without the Pinus radiata site results, reduced the »Travel empty« and »Travel loaded« RMSE values to 1.08 minutes and 1.1 minutes, respectively, and reduced the percentage error values, however they were still outside the limit of acceptability. McDonald and Fulton (2005) similarly found in their study that unusual events caused the poor correspondence between automated and manual time estimates for some skidder work el-

182

ements. They suggested implementing additional rules to detect these unusual events, which may be a possible solution for the ATSS. Although the percentage error for the differences in the ATSS and T&M »Delay« times was within the limit of acceptability, the mean T&M delay time was slightly higher than the mean ATSS delay time. This was caused by the Multidat being set to record delays of one minute or greater, which resulted in a number of minor delays being included in the T&M data but not the ATSS data. In future studies, the Multidat minimum delay length could be set to a smaller value to examine the impact on mean delay values.

The main deficiency of using GPS data analysis for automated productivity studies is that there is no means of determining the product types and load weight or volume being carried by the forwarder.

However, the use of a forwarder equipped with a set of grapple or bunk load scales could address this issue for single product harvest operations. Long term automated productivity studies using GPS can reveal trends that are not apparent in typical short term time and motion studies, such as the differences in productivity between days of the week noted by Cordero et al. (2006) and the potential areas for harvest system improvement suggested by McDonald and Rummer (2002). Absence of an observer, when using automated time study technology, is assumed to overcome the »Hawthorne Effect«, however, there may still be an effect on machine operator performance from the presence of the data collection technology, especially if it has been temporarily installed for the duration of a study.

GPS signal loss has been reported from a number of machine tracking studies (McDonald et al. 2000, Veal et al. 2001, McDonald and Fulton 2005, Hejazian et al. 2013), although only Veal et al. (2001) found a cause to the signal loss in their study (tree canopy).

The GPS signal loss that occurred on the Pinus radiata site in the current study was believed to result from the occlusion of satellites close to the horizon caused by the hill the forwarder was working on. Use of Global Navigation Satellite System (GNSS) receivers that combine signals from both the GPS and GLONASS constellations is likely to significantly reduce instances of signal loss and also to improve the positional accuracy as they can access signals from over 50 satellites. However, as mentioned previously, the GNSS data would need to be combined with a means of detecting delays, such as the Multidat vibration sensor or a link to the engine management system to record engine rpm and/or load.

Croat. j. for. eng. 36(2015)2

Automated Time Study of Forwarders using GPS and a vibration sensor (175–184)

5. Conclusion Time and motion studies of forest harvesting machines are an important component of forest operations research. However, traditional time and motion studies are generally impractical for long term studies.

In the current study, the mean forwarder cycle time estimated using automated analysis of GPS and vibration sensor data was less than 1 second from the mean cycle time determined from traditional time and motion studies. The percentage error was also well within the limit of acceptability. For harvest areas producing a single product, combining the cycle times estimated from the GPS and vibration sensor data with output from a forwarder grapple or bunk load scale could be used to conduct long term, autonomous forwarder productivity studies which would allow examination of long term trends in forwarder productivity. However, results for individual time elements were poorer, mainly due to mislabelling of brief periods, when the forwarder stopped or travelled slowly manoeuvring on a steep slope. Inclusion of additional rules in the automated GPS data analysis may address this issue.

Acknowledgements The authors would like to thank the forest managers and harvesting contractors, and in particular the forwarder operators at each of the three sites without whom the study would not have been possible.

6. References Bland, J.M., Altman, D.G., 1986: Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 327(8467): 307–310.

Cordero, R., Mardones, O., Marticorena, M., 2006: Evaluation of forestry machinery performance in harvesting operations using GPS technology. In: P.A. Ackerman, D.W. Längin, M.C. Antonides (Eds.), Precision Forestry in Plantations, Semi Natural and Natural Forests. Proceedings of the International Precision Forestry Symposium, Stellenbosch University, South Africa.

Critchley, L.A., Critchley, J.A., 1999: A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. Journal of Clinical Monitoring and computing 15(2): 85–91.

Davis, C.T., Kellogg, L.D., 2005: Measuring machine productivity with the MultiDAT datalogger: A demonstration on three forest machines. USDA Forest Service. Gen. Tech. Rep. PSW-GTR-194. de Hoop, C., Dupré, R.H., 2006: Using GPS to Document Skidder Motions – A Comparison with Manual Data CollecCroat. j. for. eng. 36(2015)2

M. Strandgard and R. Mitchell

tion. Council on Forest Engineering (COFE) Conference Proceedings: »Working Globally – Sharing Forest Engineering Challenges and Technologies Around the World«. Coeur d’Alene, July 22–2 August.

Gerasimov, Y., Senkin, V., Väätäinen, K., 2012: Productivity of single-grip harvesters in clear-cutting operations in the northern European part of Russia. European Journal of Forest Research 131(3): 647–654.

Hejazian, M., Hosseini, S., Lotfalian, M., Ahmadikoolaei, P., 2013: Possibility of global positioning system (GPS) application for time studies in forest machinery. European Journal of Experimental Biology 3(4): 93–98.

Hogg, G.A., 2009: Multistem mechanised harvesting operation analysis – Application of discrete-event simulation. Unpublished Master of Science in Forestry thesis. University of Stellenbosch, 2009.

Magagnotti, N., Spinelli, R., (eds.)., 2012: Good practice guidelines for biomass production studies. COST Action FP0902, WG 2 Operations research and measurement methodologies. 1–44.

McDonald, T.P., Fulton, J.P., 2005: Automated time study of skidders using global positioning system data. Computer and Electronics in Agriculture 48(1): 19–37.

McDonald, T., Rummer, B., 2002: Variation in skidder productivity over time in timber harvest. In: Proceedings of the 25th Annual COFE Meeting on Forest Engineering Challenges: A Global Perspective. Auburn, AL. Council of Forest Engineers, Portland, OR, 1–5.

McDonald, T.P., Rummer, R.B., Taylor, S.E., 2000: Automated time study of feller bunchers. In: Proceedings, 23rd annual meeting of the Council on Forest Engineering. 11–14 September, Kelowna, BC. Corvallis, OR: Council on Forest Engineering: 1–4.

Nuutinen, Y., Väätäinen, K., Heinonen, J., Asikainen, A., Röser, D., 2008: The accuracy of manually recorded time study data for harvester operation shown via simulator screen. Silva Fennica 42(1): 63–72.

Purfürst, F.T., 2010: Learning Curves of Harvester Operators.

Croatian Journal of Forest Engineering 31(2): 89–97.

Skogforsk, 2012: StanForD. Listing of variables by category.

Retrieved May 6, 2014, from http://www.skogforsk.se/PageFiles/60712/AllVarGrp_ENG_120418.pdf.

Stankić, I., Poršinsky, T., Tomašić, Ž., Tonković, I., Frntić M., 2012: Productivity Models for Operational Planning of Timber Forwarding in Croatia. Croatian Journal of Forest Engineering 33(1): 61–78.

Strandgard, M., Walsh, D., Acuna, M., 2013: Estimating harvester productivity in Pinus radiata plantations using StanForD stem files. Scandinavian Journal of Forest Research 28(1): 73–80.

Veal, M.W., Taylor, S.E., McDonald, T.P., McLemore, D.K., Dunn, M.R., 2001: Accuracy of Tracking Forest Machines with GPS. Transactions of the ASAE 44(6): 1903–1911.

183

M. Strandgard and R. Mitchell

Automated Time Study of Forwarders using GPS and a vibration sensor (175â&#x20AC;&#x201C;184)

Authorsâ&#x20AC;&#x2122; address: Martin Strandgard* e-mail: mstrandg@usc.edu.au Australia Forest Operations Research Alliance (AFORA) University of the Sunshine Coast 500 Yarra Boulevard 3121 Richmond AUSTRALIA

Received: October 09, 2014 Accepted: January 28, 2015

184

Rick Mitchell e-mail: rmitchel@usc.edu.au Australia Forest Operations Research Alliance (AFORA) University of the Sunshine Coast 35 Shorts Place 6330 Albany AUSTRALIA * Corresponding author Croat. j. for. eng. 36(2015)2

Original scientific paper

Efficiency of Topping Trees in Cable Yarding Operations Christoph Huber, Karl Stampfer Abstract The extraction of biomass and nutrients out of the forest is implicit to every harvest operation.

In cable yarding, whole-tree harvesting (WTH) has become more prevalent in the last few decades and processing takes place at the roadside. There is a concern that WTH impairs site productivity due to nutrient removal. One option to increase the amount of biomass remaining in the stand is to top the trees before extraction. In order to estimate the influence of topping on system productivity, time studies on a medium-sized tower yarder were carried out in three spruce dominated stands. Heart rate monitoring of the chainsaw operator was performed to examine the physiological workload. The analysis showed that topping only impacts system productivity if it takes place during the inhaul of the load as it leads to interruptions of the extraction progress. These interruptions took on average 13 seconds per turn. In addition, if topping was performed on already lifted trees, a reduction of line-speed during the lateral yarding of the loads was observed. This led to a reduction in productivity between 5 and 11%, assuming that all trees would have been topped during the lateral yarding process. Analyses of the physical workload of the chainsaw operator showed that the workload of topping trees is significantly lower than that of the felling process. Relative heart rate of the subject was lower at the cable corridors where topping was ordered. This confounding result may be a consequence of many additional factors like slope gradient or cycle time. Under both scenarios, the worker never surpassed the limit of a sustainable cardio-vascular workload for an 8 hour working day. Hence, recovery time for the chainsaw operator can be considered as adequate when topping is performed in a three-man crew.

Keywords: topping, cable yarding, productivity, workload

1. Introduction Efficient harvesting in steep terrain is usually linked to cable based harvesting systems. The use of whole-tree harvesting (WTH) has become more common in Central Europe, mainly due to the technological development of boom-mounted processors. Heinimann et al. (2001) estimated that the use of a processor in cable logging results in cost savings of about 40% compared to motor-manual cut-to-length (CTL) systems.

The change from CTL to WTH leads to a shift of the delimbing process from within the forest to the roadside. This results in a greater removal of biomass and nutrients from the forest stand. The increased removal of the nutrient richest parts of the trees (needles Croat. j. for. eng. 36(2015)2

and twigs) has raised concerns about the sustainability of WTH (e.g. Raulund-Rasmussen et al. 2008, Kaarakka et al. 2013, Tveite and Hanssen 2013).

One method to increase the amount of logging residues, remaining in the forest when using WTH, is to top the trees. Nutrient analyses show that young needles contain higher concentrations of most nutrients than older needles, branches or stem wood (Lick 1989). Hence, topping trees is an effective way to increase the amount of nutrients left in the stand as tops contain predominantly young needles and twigs. The advantage of this treatment is dependent on the topping diameter. On the one hand, the quantity of branch and needle residue is largely affected by the topping diameter, but on the other hand the selection of the topping diameter can be influenced by changes in

185

C. Huber and K. Stampfer

Efficiency of Topping Trees in Cable Yarding Operations (185–194)

price and demand of wood. The topping diameter may also affect the length of the stem section, which may influence optimal bucking and consequently impair the grade, and thus the value, of the resulting logs.

Motor-manual topping of trees in the forest stand using a chainsaw may increase the workload of the chainsaw operator. It is also likely that topping lowers productivity of the harvesting system, depending on the integration of topping into the working process.

However, studies on the economical and ergonomic effects of motor-manual topping in cable logging are lacking.

The purpose of this study was to examine the impact of topping on the productivity of a tower yarder operating in thinning operations of Norway spruce (Picea abies). The effect of topping on the workload of the chainsaw operator was also investigated.

2. Materials and methods 2.1 Study site The thinning experiments were set up in early autumn 2014 in three Norway spruce (Picea abies) dominated stands in the eastern part of the Austrian Alps.

The three stands differ from each other both by their age and silvicultural pre-treatment (Table 1). At the 34 year old study site »Bairhübl«, no prior thinning had been performed. As a consequence this stand is char-

acterized by a high number of stems (1667 stems/ha) in comparison with the 38 year old study site »Bergtal« (979 stems/ha), which had been thinned once in the thicket-life stage. The third stand »Klommegger« represents a 58 year old stand that had been thinned once.

All three stands were considered to be in need of thinning. The slope gradient of the three study sites ranged from 50 to 80%.

2.2 Experimental design The thinning harvesting operation was performed using a truck-mounted »Wanderfalke« yarder, developed by the company »Mayr-Melnhof«, extracting whole trees uphill to the forest road. The tower yarder was equipped with a Mayr-Melnhof Sherpa U3.0 carriage with a maximum load capacity of 3 tons. At the roadside, processing was done using a Woody H50 processor developed by »Konrad Forsttechnik«. The tower yarder worked in a three-line gravity system, where the third cable (haulback line) is used to pull slack on the mainline.

The crew, consisting of one yarder operator, one choker setter and one chainsaw operator, was constant over the whole experimental period. No job rotation occurred between the workers. At the study site »Bairhübl«, the choker-setter was replaced by another one at the second cable corridor.

During the operation, the chainsaw operator felled a few trees (ca. 5–15) in advance so that the choker-

Table 1 Site and stand characteristics of experiments Parameter

»Klommegger« Corridor 1

»Bergtal«

Corridor 2

Corridor 1

»Bairhübl« Corridor 2

Corridor 1

Corridor 2

47°19'33.1''N, 15°19'16.6''E

47°20'13.3''N, 15°20'51.9''E

845 m

628 m

47°19'11.1''N, 15°20'31.4''E 822 m

Second thinning

First thinning

First thinning

First thinning

Cleaning

-

Number of trees before operation, stems/ha

728

979

1667

Number of trees after operation, stems/ha

320

454

500

Average DBH before operation, cm

29.3

18.2

15.5

Average DBH after operation, cm

35.0

20.5

20.0

Coordinates (position of the yarder at corridor 1) Operation type Previous operation

Treatment

Without topping

Topping

Without topping

Topping

Without topping

Topping

Length of cable corridor, m

170

170

240

255

110

140

Average inclination, %

54

51

60

62

68

71

Average piece volume, m3

0.67

0.61

0.19

0.21

0.16

0.19

Average number of trees per load

1.69

1.69

2.15

2.12

2.63

2.44

186

Croat. j. for. eng. 36(2015)2

Efficiency of Topping Trees in Cable Yarding Operations (185â&#x20AC;&#x201C;194)

setter was not hindered in his job. Another advantage of this working method is that it helps the chokersetter to make an optimum load because he is able to choose between different stems to hook on.

At each study site, two cable corridors were analyzed. At the first corridor, full trees including branches and tops were extracted uphill to the roadside. At the second corridor, all trees were topped by the chainsaw operator at a diameter of approximately 6 cm.

Topping was usually performed during the lateral yarding process before the trees were fully pulled up to the skyline (Fig. 1), which led to an interruption in the extraction progress. During this delay period, the chainsaw operator moved to the head of the trees and topped them. The extraction process continued when the worker was standing in a safe position. This procedure provides the only way to top hung-up trees in the forest stand, but was also applied to facilitate the work of the chainsaw

C. Huber and K. Stampfer

Fig. 1 Topping of an already lifted tree during lateral yarding process operator. In some cases, at the convenience of the chainsaw operator, trees were topped before the chokers were set.

Table 2 Work task definitions used in three time studies Cable yarding system Outhaul

Carriage movement from the landing to the choker-setter

Hook-on

Starts when the rigging is lowered (carriage positioning is included) and ends when the load reaches the carriage

Inhaul

Carriage movement from the stand to the landing

Grounding the load

Time of lowering the rigging and positioning the load at the landing

Unhooking the load

Starts when the yarder-operator gets off his seat to detach the chain strops and ends when he takes a seat in the cab

Raise rigging at landing Time to raise the rigging until it reaches the carriage Waiting

Operational delay time Choker-setter

Lower rigging

Time to spool out the mainline until the worker grabs it

Hook-on

Required time to move to the trees, to attach strops to them and to retreat to a safe position afterwards

Break-out

Time to raise the load and pre-extract it to the cable line until it reaches the carriage

Waiting during topping Time the working progress of the choker-setter is interrupted by topping trees Waiting

Operational delay time within which the choker setter is both waiting for the carriage and planning the next load Chainsaw operator

Felling

Time to fell a tree (including the steps: select and assess trees, clear vegetation around the base, felling the tree, observe tree fall)

Topping

Time to move to the trees and to top them

Other PSH0

Other activities like crosscutting or fueling the chainsaw

Waiting

Operational delay times within the chainsaw operator is hindered by the choker-setter General work tasks used in all three studies

Delays<15 min

Delays shorter than 15 minutes

Delays>15 min

Delays longer than 15 minutes

Croat. j. for. eng. 36(2015)2

187

C. Huber and K. Stampfer

Efficiency of Topping Trees in Cable Yarding Operations (185–194)

2.3 Time studies In order to record process variables that may influence time or system productivity, all trees of each study site were recorded according to tree species and DBH (diameter at 130 cm height). The associated heights were measured randomly using a sample of at least 20 heights per tree species and stand. To allocate each felled tree to a specific extraction process, all trees were numbered consecutively using aerosol cans. To estimate the extraction distance of each load in the field, the distance to the landing was measured along the cable corridor and marked on some residual stand trees next to the corridor.

Time studies were performed to illustrate the effect of topping on system productivity. Data collections were carried out using three mobile tablet PCs »ALGIZ 7« running in a Windows 7® environment, using a continuous timing method. To analyze the time of each working process, the working cycle of each crewmember was split into several functional elements (Table 2). Parallel to the time consumption measurements, the number of each harvested tree, as well as the extraction distance, was noted during the observation of the choker setter.

2.4 Physiological measurements The workload of the chainsaw operator was estimated based on heart rate measurements because direct measurement methods, such as the maximum oxygen uptake of a worker (VO2 max), are difficult to obtain under field conditions (Kirk and Sullman 2001).

The heart rate of the subject was measured at 1 sec intervals for the entire working time using a Polar RS800 running computer. The computer consists of a heartbeat-capturing transmitting unit and a receiverstorage unit. Before data collection, the transmitter is attached to a strap and secured around the subject’s chest just below the chest muscles. The transmitter sends the heart rate signal directly to the running computer, which displays and records the data. The use of sport heart rate monitors has already been proved successfully in several ergonomic studies (Vogelaere et al.

1986, Stampfer et al. 2010, Magagnotti and Spinelli 2012).

At the start of each working day, the heart rate monitor was attached to the chainsaw operator and started simultaneously with the time study software on the hand-held field computer »ALGIZ 7« in order to be able to merge the working heart rate (HRw) data set with the time study data set. The resting heart rate (HRr) was obtained for the subject upon arrival at the work site. Therefore, the chainsaw operator was asked

188

to remain seated inside the car without moving, drinking or smoking for a minimum of 10 min. The minimum heart rate within this time period was selected as his resting heart rate. The maximum heart rate (HRmax) was estimated by using the standard formula (Rodahl 1989):

HRmax = 220 – Age

(1)

Relative heart rate at work (% HRR) was determined by applying the following equation (Rodahl 1989, Apud 1989):

% HRR =

Where:

HRw - HRr ´ 100

HRmax - HRr

(2)

% HRR relative heart rate at work; HRw heart rate at work; HRr heart rate at rest; HRmax maximum heart rate.

In total, 28 hours of chainsaw operator heart-rate data were collected. The pre-work resting heart rate of the 32 year old subject was 55 bt./min.

2.5 Statistical analysis Height curves were computed for each stand using the DBH as independent variable, with b being the coeffi­cient and a being the constant: 1

(3) DBH The height of each tree was predicted from the tree height calculated by the height curve, and the volume was calculated using the cubing formulas according to Pollanschütz (1974).

ln(h – 1.3) = a + b ×

Former productivity studies on tower yarders (e.g.

Stampfer 2002, Stampfer et al. 2010, Talbot et al. 2014) showed that the mean piece volumes, extraction distance and harvesting intensity are the main factors influencing system productivity. The following productivity hypothesis is used in this study: Yarding productivity = f (mean volume per piece (tree vol), Extraction distance (dist), cutting intensity (int), Slope gradient (grad), topping) As some of these variables only influence parts of the extraction cycle, the productivity of the harvesting system (m3/PSH15) was determined by calculating individual efficiency models of the main elements of the extraction cycle (cp. Nurminen 2006):

Prod yarder =

60

c ×(eff hook-on + eff carriage + eff landing )

(4)

Croat. j. for. eng. 36(2015)2

Efficiency of Topping Trees in Cable Yarding Operations (185–194)

C. Huber and K. Stampfer

Table 3 Variables used to describe yarding productivity Factors

Covariates

Topping

(0) Without topping; (1) With topping

2 levels

Mean volume per piece

Mean tree volume per load

m3

Extraction distance

Distance between tower yarder and stopping position of the carriage

m

Cutting intensity

Removed fraction of volume in a defined area during harvest operation

%

Slope gradient

Slope steepness within a defined area

%

Table 3 gives an overview of the variables used to calculate the productivity of the harvesting system.

The hook-on phase involves all activities at the felling site, starting with lowering the rigging and ending with the completed lateral inhaul of the load, and the carriage phase comprises the inhaul and outhaul of the carriage. All working tasks at the landing, including grounding and unhooking the load as well as raising the rigging, are summarized in the landing phase. In order to include delays of up to 15 minutes in the model, a conversion factor (c) of 1.3 was used.

Analysis of variance was used to analyze the influence of co-variables and factors, including analysis of interactions between the variables (Stampfer 2002).

Due to the nonlinear relationship between the average tree volume and efficiency, the co-variable »average tree volume per cycle« was transformed using power functions. The suitability of the exponents was evaluated by the coefficient of determination and the distribution of the residuals. In order to estimate the coefficients of the variables used in the models, regression analysis was made.

ANOVA techniques were used to check the statistical differences between the different study sites and treatments. All analyses were carried out using both Microsoft® Excel 2013 and PASW 18.0 for Windows.

3. Results 3.1 Productivity analysis Interruptions of the working process, which occurred when trees were topped during the lateral inhaul of the load, took on average 12.65±6.68 seconds (Fig. 2) and were significantly higher (Tukey-HSD, p=0.0236) at the first thinning stand »Bairhübl« (14.11±7.49 sec) than at the second thinning stand »Klommegger« (11.05±5.95 sec). The expenditure of time for topping the trees at the study site »Bergtal« was on average 12.05±5.98 seconds and did not differ significantly from the other two stands.

The study also showed that breaking out a load takes more time when topping takes place during the Croat. j. for. eng. 36(2015)2

Fig. 2 Delay durations resulting from topping trees during the lateral yarding phase lateral yarding process, even if the interruption times due to topping the trees are not included. This average prolongation of the break-out process, not including any interruption times during this phase, was greatest at »Bergtal« (23.79%), followed by »Klommegger« (15.74%) and »Bairhübl« (11.89%).

Topping the trees during the lateral yarding process was mainly performed at the first thinning stands due to a high number of hang ups. Interruptions due to topping trees occurred in 69.17% (»Bairhübl«) and 65.99% (»Bergtal«) of the extraction cycles. In the other cases, the trees were topped before they were attached. In contrast to the first thinning stands, at the study site »Klommegger« only 44.44% of the loads were topped during the break-out task. This result is directly linked to a high number of broken stems during felling, which resulted in fewer trees that had to be topped. At »Klommegger«, 42.20% of the felled trees broke during the felling and extraction operation at an average diameter of 9.77±3.74 cm. At the other

189

C. Huber and K. Stampfer

Efficiency of Topping Trees in Cable Yarding Operations (185–194)

Regression results for the efficiency models are reported in Table 4.

The average tree volume per load (tree volume) influenced significantly the efficiency of all three cycle elements. During both the hook-on phase and the carriage phase the slope gradient played a significant role. The equations show that a higher inclination (inc) influences the efficiency (min/m3) of the hook-on phase in a positive way, and of the carriage phase in a negative way. Topping the trees during the extraction process (topping) only affects the efficiency of the hook-on phase, while the extraction distance (dist) had a significant influence on the efficiency of the carriage phase.

The overall mean values of the variables (Table 3) were used to calculate productivity models (Fig. 3) of the logging systems using treatments with and without topping the trees, as a function of the average tree volume. According to the productivity model, topping leads to a decrease of the average system productivity

two sites, tree breakage occurred less frequently, being more frequent at »Bergtal« (16.60%) than at »Bairhübl« (1.55%). At these two stands, the average size of the broken pieces was 5.02±1.78 cm and 6.50±3.32 cm, respectively, for »Bergtal« and »Bairhübl«.

Based on the chronometry data of 692 cycles and inventory data, efficiency models (min/m3) for the main elements of the extraction cycle were calculated: effhook-on = b1 × tree volume–0.9 + b2 × inc + b3 × topping [R2adj. = 0.87; F (3,689) = 1571.55, p<0.001]

(5)

effcarriage = b1 × tree volume–0.9 + b2 × inc + b3 × dist [R2adj. = 0.79; F (3,688) = 882.31, p<0,001]

(6)

efflanding = b1 × tree volume–0.9 [R2adj. = 0.77; F (1,689) = 2332.61, p<0,001]

(7)

Table 4 Regression model parameters for the extraction cycle elements Hook-on phase

Carriage phase

Landing phase

ß1

ß2

ß3

ß1

ß2

ß3

ß1

Coefficients

0.84

0.01

0.50

0.60

-0.01

0.01

0.38

Standard error

0.03

0.00

0.16

0.03

0.00

0.00

0.01

t-stat

27.01

4.48

3.21

21.28

-3.15

8.71

48.30

p-value

<.001

<.001

.001

<.001

.002

<.001

<.001

Table 5 Comparison of relative heart rate, DBH of felled trees and number of felled trees per minute within the felling task. Interruptions >15 min. are not included Without topping trees

With topping trees

p valuea

% HRR

26.71

23.43

<.001

Av. DBH of felled trees

24.44

24.53

.895

Felled trees/min. felling time

1.73

1.36

.001

% HRR

27.79

27.51

.001

Av. DBH of felled trees

15.96

16.35

.308

Felled trees/min. felling time

2.04

1.84

.119

% HRR

35.26

31.91

<.001

Av. DBH of felled trees

11.98

15.45

<.001

Felled trees/min. felling time

4.52

2.57

<.001

Results for »Klommegger«

»Bergtal«

»Bairhübl«

a

Statistical significance for the equality of treatment means

190

Croat. j. for. eng. 36(2015)2

Efficiency of Topping Trees in Cable Yarding Operations (185–194)

C. Huber and K. Stampfer

(-5.06%), respectively, at »Bairhübl«, »Bergtal« and »Klommegger«.

It is very likely that topping trees not only affects system productivity, which results in higher harvesting costs, but also increases the workload of the chainsaw operator. Hence, the time of each work phase was determined for the chainsaw operator (Fig. 4). The diagram illustrates that felling trees was the longest work task within the working cycle consuming 30 to 45% of the entire working time. At the cable corridors where topping was required, it took the chainsaw operator between 4.76% (»Klommegger«) and 8.59%

Fig. 3 Productivity of cable yarding system depending on tree volume and integration of topping into the working process from 5.25 m3/PSH15 to 4.97 m3/PSH15 (-5.36%) and from 4.78 m3/PSH15 to 4.54 m3/PSH15 (-4.90%), respectively, for »Bergtal« and »Bairhübl«, assuming that all trees would have been topped during the lateral yarding phase. At »Klommegger«, topping all trees after attaching them to the mainline would reduce the productivity from 11.92 m3/PSH15 to 10.56 m3/PSH15 (-11.39%). However, productivity loss due to topping was much smaller because of tree breakage and numerous trees that had been topped before they got hooked on. Considering these factors, topping trees decreased system productivity in fact to 4.62 m3/PSH15 (-3.39%), 5.06 m3/PSH15 (-3.54%) and 11.32 m3/PSH15

Fig. 4 Time consumption of different working tasks of the chainsaw operator

Table 6 Relative heart rate (%HRR) by working tasks at different study sites and treatments. Pairwise comparisons were performed, and different statistically significant means (p<0.05) were marked with different letters Study site »Klommegger«

»Bergtal«

»Bairhübl« a

Treatment

Felling

Topping

Other PSH0

Operational delays

Delays >15 min

F value

p valuea

With topping

23.55a

21.9b

22.47b

22.69b

22.57b

17.90

<.001

Without topping

28.25a

-

27.99a

28.09a

18.94b

792.05

<.001

With topping

28.73a

27.24b

27.39b

27.54b

24.15c

166.52

<.001

Without topping

29.83a

-

30.33a

27.33b

18.32c

929.44

<.001

With topping

34.30a

31.11b

33.64c

31.57b

26.34d

423.38

<.001

Without topping

37.32a

-

37.00ab

36.96b

31.04c

1802.07

<.001

Statistical significance for the equality of treatment means

Croat. j. for. eng. 36(2015)2

191

C. Huber and K. Stampfer

Efficiency of Topping Trees in Cable Yarding Operations (185–194)

(»Bairhübl«) of his working time (PSH15) to top trees.

The working task »other PSH0« was mainly scrubcleaning and was most time consuming at the first thinning stands, especially at »Bairhübl«, where no previous silvicultural treatment had been performed.

3.2 Heart rate analysis Table 5 shows the relative heart rate of the chainsaw operator in conjunction with the average DBH of the felled trees and the average number of felled trees per minute within the felling task. The results are presented separately for the three study sites. Overall, the relative heart rate of the chainsaw operator was significantly lower at all three sites at the cable corridors where topping was performed. At the study site »Bairhübl«, the average DBH of the felled trees differed significantly between the two treatments. During the felling task, a significantly higher number of trees were felled by the chainsaw operator at study sites »Bairhübl« and »Klommegger«. However, the average resting heart rate never surpassed the 40% cardiovascular load mark, which is defined as the limit of a sustainable workload for an 8 hour working day (Stampfer 1996).

Table 6 shows the differences in relative heart rate between different work tasks of the chainsaw operator.

The highest physiological workloads were measured during the felling task. At 50% of the cable corridors, relative heart rate of the felling task was significantly higher than at any other working task. No significant differences were observed between the working tasks topping and operational delays. Both work steps were characterized by significantly lower relative heart rates compared to the felling task.

4. Discussion Topping seems to decrease system productivity only when it is performed during the break-out phase because the lateral yarding process has to be delayed.

These interruptions took on average 12.65 seconds, and were longer at the first thinning stands than at the second thinning stand »Klommegger«. It seems to be very likely that these differences in time demand are directly associated with the number of trees per load.

While at the study sites »Bairhübl« and »Bergtal«, an average load consisted of 2.52±0.78 and 2.14±0.86 stems, respectively, at »Klommegger« only 1.69±0.70 stems formed a load. If a load consists of more trees, it will take more time to top all the trees. Conversely, the average time to top a single tree will probably decrease if the load is formed by a higher number of stems.

192

The results show that the break-out phase is not only prolonged because of interruption times due to topping the trees. Also the time to spool in the mainline in order to pull the load to the carriage, not including any interruptions, required more time at all sites when trees were topped during the lateral yarding phase. It is very likely that factors like positioning the load next to the chainsaw operator or reducing the line speed to minimize hazards for the chainsaw operator may decrease the extraction speed.

At »Klommegger«, tree breakage was an important factor influencing the quantity of trees that had to be topped. In comparison to the other two study sites, the size of the broken pieces was much larger at Klommegger. This observation concurs well with Fitec (2000), reporting that trees usually break at 2/3 of their total height.

Topping trees during the break-out phase not only affects productivity and logging costs; it also poses hazards for the chainsaw operator. Partially suspended trees can have tension and compression forces within them that make the job of topping more difficult and dangerous. Unexpected release of stems or unplanned load movement during topping may also be a significant safety hazard for the worker.

Additionally, topping trees also influences the time consumption of the chainsaw operator, mainly by the need to walk longer distances. A large amount of broken or topped trees remaining at the forest site also affects the ease with which the workers can safely move at the cutover area. However, Fig. 3 shows that the operational delay times of the chainsaw operator was at least 40%. During this period he was mainly waiting until most of the felled trees were extracted by the choker-setter. Consequently, a large part of this period can be considered as recovery time.

Table 5 showed that an introduction of topping trees into the working process of the chainsaw operator resulted in a statistically significant reduction of relative heart rate, although topping was included as an additional work task into his working process. It is very likely that the reduced workload relates to the chainsaw operator working at a lower pace when topping was ordered, but also other factors like slope gradient or walking distances may have influenced the physical workload and covered other effects.

5. Conclusions Topping can be a useful treatment to increase the amount of logging residues remaining at the forest site. When working with a three-man crew, this study Croat. j. for. eng. 36(2015)2

Efficiency of Topping Trees in Cable Yarding Operations (185–194)

showed a decrease in system productivity by only 3.4 to 5.1%. Consequently, the harvesting costs increase by approximately 1.00–2.50 €/m3. Topping does affect the work time of the chainsaw operator. Especially in steep terrain, walking to the head of the trees to top them can be a time-consuming task. However, in this study, physiological workload measurements of the chainsaw operator showed that the average heart rate reserve never surpassed the endurance limit of performance at any working task. Therefore, recovery time for the chainsaw operator can be considered as adequate when working in a three-man crew.

Acknowledgements Financial support for this study was provided by the cooperation-platform FHP (Forst-Holz-Papier).

The authors would like to thank the forest administration »Pfannberg« of the forest company »Mayr-Melnhof-Saurau« for supporting the field work and for providing the study sites. The authors also wish to thank the harvesting company »Holzernte Gosch G.m.b.H« for the good cooperation.

6. References Apud, E., 1989: Human biological methods for ergonomics research in forestry. Guide-Lines on Ergonomic Study in Forestry. International Labour Office, Geneva. 242 p.

FITEC, 2000: Best Practice Guidelines for Cable Logging.

Forest Industry Training and Educational Council New Zealand. 129 p.

Heinimann, H.R., Stampfer, K., Loschek, J., Caminada, L., 2001: Perspectives on Central European Cable Yarding Systems. International Mountain Logging and 11th Pacific Northwest Skyline Symposium. Seattle, Washington, USA. p. 268–279.

Kaarakka, L., Tamminen, P., Saarsalmi, A., Kukkola, M., Helmisaari, H.-S., Burton, A.J., 2014: Effects of repeated whole-tree harvesting on soil properties and tree growth in a Norway spruce (Picea abies (L.) Karst.) stand. Forest Ecology and Management 313: 180–187.

Kirk, P. M., Sullman, M. J. M., 2001: Heart rate strain in cable hauler choker setters in New Zealand logging operations.

Applied Ergonomics 32(4): 389–398.

Croat. j. for. eng. 36(2015)2

C. Huber and K. Stampfer

Lick, E., 1989: Untersuchungen zur Problematik des Biomassen- und Nährelemententzuges bei der Erstdurchforstung eines zentralalpinen Fichtenbestandes. PhD. Thesis. University of Natural Resources and Life Sciences, Vienna, Austria.

256 p.

Magagnotti, N., Spinelli, R., 2012: Replacing Steel Cable with Synthetic Rope to Reduce Operator Workload During Log Winching Operations. Small Scale Forestry 11(2): 223–236.

Nurminen, T., Korpunen H., Uusitalo J., 2006: Time consumption analysis of the mechanized cut-to-length harvesting system. Silva Fennica 40(2): 335–363.

Pollanschütz, J., 1974: Formzahlfunktionen der Hauptbaumarten Österreichs. Allgemeine Forstzeitung 85: 341– 343.

Raulund-Rasmussen, K., Stupak, I., Clarke, N., Callesen I., Helmisaari, H.S., Karltun, E., Varnagiryte-Kabasinskiene, I., 2008: Effects of very intensive forest biomass harvesting on short and long term site productivity. In: Sustainable use of forest biomass for energy (Röser, D., Asikainen, A., RaulundRasmussen, K., Stupak, I., ed.) Managing forest ecosystems 12: Springer, Dordrecht, 29–78.

Rodahl, K., 1989: The physiology of work. Taylor & Francis, London, New York, Philadelphia.

Stampfer, K., 1996: Belastungs- und Beanspruchungsermittlung bei verschiedenen mechanisierten forstlichen Arbeitssystemen. In: Trzesniowski, A.: Schriftenreihe des Instituts für Forsttechnik 3. Wien.

Stampfer, K., 2002: Optimierung von Holzerntesystemen im Gebirge. Habilitation treatise, University of Natural Resources and Life Sciences, Vienna, Austria. 96 p.

Stampfer, K., Leitner, T., Visser R., 2010: Efficiency and Ergonomic Benefits of Using Radio Controlled Chokers in Cable Yarding. Croatian Journal of Forest Engineering 31(1): 1–9.

Talbot, B., Aalmo, G.O., Stampfer, K., 2014: Productivity analysis of an un-guyed integrated yarder-processor with running skyline. Croatian Journal of Forest Engineering 35(2): 201–210.

Tveite, B., Hanssen, K.H., 2013: Whole-tree thinnings in stands of Scots pine (Pinus sylvestris) and Norway spruce (Picea abies): short- and long-term growth results. Forest Ecology and Management 298: 52–61.

Vogelaere, P., De Meyer, F., Duquet W., Vanjdefelde P., 1986: Validation du Sport Tester PE 3000 en function de l’enregistrement Holter. Science and Sports 1(4): 321–329.

193

C. Huber and K. Stampfer

Efficiency of Topping Trees in Cable Yarding Operations (185â&#x20AC;&#x201C;194)

Authorsâ&#x20AC;&#x2122; address:

Received: July 01, 2015 Accepted: July 28, 2015

194

Christoph Huber, MSc.* e-mail: c.huber@boku.ac.at Prof. Karl Stampfer, PhD. e-mail: karl.stampfer@boku.ac.at University of Natural Resources and Life Sciences Department of Forest and Soil Sciences Peter-Jordan-Strasse 82/3 1190 Vienna AUSTRIA * Corresponding author Croat. j. for. eng. 36(2015)2

Original scientific paper

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability and Technical Constraints Raffaele Spinelli, Rien Visser, Oliver Thees, Udo Hans Sauter, Nike Krajnc, Catherine Riond, Natascia Magagnotti Abstract A survey of cable logging contracts was conducted in 5 of the 8 Alpine countries, namely: France, Germany, Italy, Slovenia and Switzerland. The goals of the survey were to set a general reference for cable logging rates, to identify eventual differences between countries and to determine the effect of technical work parameters (i.e. piece size, removal per hectare, line length) on actual contract rates. With a total sample size of 566 units, the mean removal and rate were 165 m3 ha–1 and 42.9 € m–3, respectively. Both removal per hectare and contract logging rates varied considerably and the study found significant differences between countries.

Switzerland stood out from the group with the highest removal (345 m3 ha–1), but also the high­ est contract rate (79.5 € m–3). Removal per hectare was lowest in Italy with just 58.3 m3 ha–1, and logging rate lowest in Slovenia at 29.3 € m–3. Logging rates were highly correlated with the average labour rate of each country. Technical factors such as tree size, line length and tract size explained about 40% of the variability in logging rates. Therefore, 60% of the variability is explained by other technical factors not included in our data and by non-technical factors, such as local market dynamics.

Keywords: harvesting, mountain, wood, forestry, yarder

1. Introduction The Alps are one of the great mountain range systems of Europe, stretching 1200 km across 8 different countries, namely: Austria, France, Germany, Italy, Liechtenstein, Monaco, Slovenia and Switzerland (Fig. 1). Despite national borders, the region is united by common ecological and cultural characteristics, which set it apart from the rest of Europe (Onida 2009). For this reason, trans-boundary regionalism has become a dominant trend in the economical and political integration of the alpine space. The process of regionalization offers new opportunities to innovation and sustainable development. This is best expressed by the Alpine Convention, signed in 1991 between the European Union and the eight states with territory in the Alps (Balsiger 2008).

Forest cover represents 40% of the Alpine landscape. Forests have always played an important role Croat. j. for. eng. 36(2015)2

Fig. 1 The Alpine region and the 8 countries sharing territory within it

195

R. Spinelli et al.

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

in supporting the alpine economy, and that is still true today with the boom of engineered wood products and energy biomass (Onida 2009). Alpine forests also have a protective function, as they prevent soil erosion and shield settlements from avalanches and rock fall (Dorren et al. 2004).

The need to balance cost effective wood production with careful protection makes alpine forestry particularly complex. Continuous cover forestry is popular, as it offers a good compromise between these two vital functions. However, continuous cover forestry reduces harvest removal and it may constrain operation profitability (Mason et al. 1999). Low removals and the typical access constraints of the Alpine territory hinder the introduction of modern mechanisation, which is the only solution to cost containment in the face of increasing fuel and labour cost (Spinelli and Magagnotti 2011). As a consequence, silviculture is often delayed and results in a skewed age distribution (Binder et al. 2004). That contributes to the high vulnerability of Alpine forests to the effects of climate change (Seidl et al. 2011). Therefore, it is crucial to optimize forest operations in order to guarantee timely regeneration and maximize resiliency.

A first step towards optimization is cost control (Mathews 1942). To this end, it is necessary to know accurately the current harvesting contract rates, as well as their sensitivity to market factors (e.g. competition, labour cost, cost of living, etc.) and technical factors (e. g. line length, tree size, removal, etc.).

Time and motion studies are often used to develop an understanding of how individual elements in a logging system respond to stand and terrain parameters (Visser and Stampfer 2003). These studies can be combined to provide system cost figures. Otherwise, complex system models can be developed, using advanced simulation software. Such models are capable of combining stand, terrain and system information in order to produce reliable cost figures (McDonagh et al. 2004).

However, all these tools provide information about the technical cost of harvesting, not the actual logging rate reported on the harvesting contract (Stuart and Grace 2011). The latter is influenced by many non-technical factors, related to market structure and contractor interaction (Visser 2009). For this reason, predictions based exclusively on technical factors may not reflect actual rates, weakening the ability to predict and manage logging costs.

In trans-boundary regions, further differences in logging rates can be expected to derive from the specific economic conditions encountered in the different states that share the common regional space. While the landscape and culture are indeed common, local dif-

196

ferences are encountered in market competition, labour cost, fuel cost, access to credit, government subsidies and fiscal pressure, among others. Knowledge of these differences is especially important when planning for trans-border activities. For example, when concentrating a large international task force to face a sudden local crisis, such as catastrophic forest damage (CTBA 2004). Potential interface problems should be explored and solved well in advance, because the frequency and severity of catastrophic forest damage has been predicted to increase in the future, due to the unfavourable combination of climate change and past silvilcultural choices (Schelhaas et al. 2010, Klaus et al.

2011). This will also affect lumber prices (Koch et al.

2013).

Therefore, the goals of this study were to: Þ quantify the contract cable logging rates currently applied across the Alpine space, Þ determine the extent and the significance of possible national differences, Þ gauge the amount of variability introduced by non-technical local market factors (e.g. labour cost, cost of living, competition, etc.) in order to assess the potential error of relying only on technical factors.

This study focuses on cable logging (Fig. 2), which is the backbone of steep slope harvesting world-wide (Bont and Heinimann 2012). In 2012, there were over 350 cable logging contractors in alpine Italy alone (Spinelli et al. 2013). On steep terrain, cable logging is the cost effective alternative to building an extensive network of skidding trails and results in a much lower site impact compared to ground based logging (Spinelli et al. 2010). On the other hand, cable logging is inherently expensive because it is normally de-

Fig. 2 Representative cable logging operation Croat. j. for. eng. 36(2015)2

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

ployed on difficult sites. For this reason, cable logging offers lower profit margins compared to ground based logging. This justifies a stronger optimization effort, supported by a deeper knowledge of technical cost and market rates.

2. Material and methods Actual contract data were obtained from forest owners in 5 of the 8 Alpine countries, namely: France, Germany, Italy, Slovenia and Switzerland. Owners were asked to provide information about the actual contract rate in € per m3, as indicated in each individual contract. They were also asked to provide the following technical information relative to each contract: contract number; place name; forest compartment; total harvest volume; harvest area; number of harvested trees; line length; occurrence and inclusion of two staging in the contract rate. This information allowed calculating average tree size, tract size and removal per hectare, which are among the most important parameters affecting technical cost (Hasel 1946), together with line length.

A total of 566 data entries were collected: 64 from northern Italy (Friuli Venezia Giulia and Trentino), 96 from Switzerland, 42 from south-eastern France (Alpes du Nord), 227 from Slovenia and 127 from southern Germany. Contracts included all main work steps required for turning standing trees into merchandised logs, stacked at the landing (Spinelli and Magagnotti 2011). In particular, all contracts covered felling, processing and extraction to a landing. Since yarder landings are often small and remote, contracts may also include a further work step, namely: moving the logs from the yarder landing to a larger log yard that is accessible to on-road transportation, an operation that is commonly known as two staging (Spinelli et al.

2014). French, Italian and Swiss contracts did indicate two staging, and the Italian contracts also reported the actual two staging distance. However, German and Slovenian contracts provided no indication about the possible occurrence of two staging. The largest majority of the French, Italian and Slovenian contracts concerned public forests.

All contracts were based on regeneration cuts of mature stands, which were conducted according to different techniques, and especially selection cut and patch cut. The study included no clear cuts, which are generally banned in all the countries spanned by our survey in order to contain hydro geological risk. The type of cut applied in each contract was inferred from the removal per hectare. All stands were pure softwood or softwood-dominated mixed forests. Norway Croat. j. for. eng. 36(2015)2

R. Spinelli et al.

spruce (Picea abies Karst.) was the most common species, followed by larch (Larix Europea L.), silver fir (Abies alba L.), beech (Fagus sylvatica L.) and pines (Pinus sylvestris L. and Pinus cembra L.).

Most contracts were dated between 2010 and 2013.

A few were older, and the oldest ones had been issued in 2002. For this reason, all rates were annualized to 2013 figures using the consumer price indices found in the respective national statistics. Figures of average hourly labour costs for each country were obtained from the Eurostat database (Eurostat 2013). Swiss rates were expressed in Swiss francs and they were converted into Euros using the official exchange rate of December 2013, equal to 0.81 Euro per 1 Swiss franc.

Volume figures were generally provided under bark: the over bark figures were converted into their under bark equivalents using appropriate coefficients.

The German data set contained no information about tree size and line length, and therefore it was impossible to calculate the effect of these factors on the German contract rates.

Data were analyzed statistically using SPSS. As a first step, descriptive statistics were determined for each country. Then, the significance of these differences was tested through the analysis of variance, after checking that the data pool met all statistical assumptions. Regression analysis of the data allowed testing the relationship between contract rate and the main technical parameters listed above, and namely: removal, tree size, tract size, line length and the eventual inclusion of two staging in the contract. The regression coefficient R2 was taken as an indicator of how strong the effect of the main technical parameters was on contract rate. Conversely, unexplained variability was considered to derive at least in part from non-technical parameters, such as: local competition, negotiation and general market dynamics.

3. Results and discussion Mean contract rate was 42.9 € m–3, while mean tree size was 1.3 m3 (Table 1). The mean tract contained 500 trees, or 670 m3. Line length ranged between 100 and 450 m (interquartile range). Variations in removals were even higher, ranging between 30 and 400 m3 ha–1 (interquartile range).

Two staging was noted in Italy (81% of cases) and Switzerland (100% of cases), but not in France, Germany and Slovenia. Only Italy recorded the actual two staging distance, and the average was 3.7 km.

The study highlighted strong national differences (Table 2). In particular, tract size in m3 was higher for

197

R. Spinelli et al.

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

Table 1 Summary of data submitted for all countries n

Average

5th percentile

95th percentile

Number of trees, n

428

509

76

1499

Harvest area, ha

453

5.24

0.79

17.10

Total volume, m3

553

671

107

1790

415

1.33

0.37

2.47

Stand density, m /ha

440

165

33

405

Average extraction distance, m

426

285

86

500

Harvesting rate, €/m3

566

42.9

17.6

86.5

3

Piece size, m

3

France and Germany, compared to the rest. Removal per hectare was over three times larger in Switzerland (345 m3 ha–1) than in all the other countries. Italy showed the lowest removal (58.3 m3 ha–1), half the value as the combined average. In contrast, tree size was relatively even, with the French mean slightly higher than the means of all other countries. Mean line length was 591 m in France, significantly higher than the 353 m in Switzerland, which in turn was significantly higher than Italy and Slovenia (360 m and 220 m, respectively). Mean contract rate varied from 29 to 79 € m–3, and it was significantly different between countries. It was lowest in Slovenia and highest in Switzerland. The mean contract rate recorded for Switzerland was twice as high as in all other countries combined. Contract rates showed a strong relationship with national labour cost data (Fig. 3).

Regression analysis showed a good relationship between contract rate and such technical work conditions as harvest area, piece size and line length (Eq. 1).

Removal per hectare also had a significant effect, but this variable was excluded from the regression equation because of its strong collinearity with the country

Fig. 3 Chart showing harvesting rate compared to average labour rate variable. Removal was much higher for Switzerland, which also had the highest contract rates. As a consequence, inclusion of removal in the regression would return a positive parameter, with contract rate increasing as a function of removal. Therefore, inclusion of this variable was contrary to sound statistical practice and it returned illogical results.

CR = 25.92 + 0.037 L – 0.253 A – 2.91 TS Sig.= 0.005; R2 = 0.41

Where: CR L A TS

Contract rate, € m–3; Line length, m; Harvest area, ha; Tree size, m3.

Table 2 Breakdown by country showing average values and standard deviation Harvest area, ha Country France

9.6a

Standard deviation 6.1

Germany

12.0b

6.9

Italy

9.9a

Slovenia

4.0c

Switzerland

1.6d

Mean

Harvest, m3

Removal, m3/ha

Piece size, m3/tree

997a

103a

Standard deviation 60.3

1.7a

Standard deviation 0.75

1173b

97.6a

35.4

no data

no data

7.2

578c

58.3b

45.2

1.3b

3.5

386d

97.5a

87.2

1.2b

0.8

538c

345c

81.4

1.4b

Mean

Mean

Mean

Harvesting rate 2013 €/m3

Line length, m

591a

Standard deviation 302

no data

no data

0.88

246c

98

42.6c

7.5

0.63

209d

87

29.3d

10.3

0.59

353b

83

79.5e

27.2

Mean

47.2a

Standard deviation 6.5

38.5b

7.7

Mean

Notes: Different letters along the same columns indicate statistically significant differences between country averages, checked by analysis of variance at the 5% level

198

Croat. j. for. eng. 36(2015)2

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

Eq. 1 shows that the contract rate increases with line length and decreases with tract (harvest area) and tree size. This relationship is logical, because increasing line lengths tend to decrease productivity and increase cost, whereas increasing tract and tree size tend to increase productivity and decrease cost.

A careful literature search found no other surveys of Alpine or European logging contract rates, which highlights the value of our study and denies the possibility of any direct comparison with similar studies.

Similar studies are available for the USA (Baker et al.

2013) and New Zealand (Visser 2012), but harvesting technology and work conditions are so different from the Alpine ones that it is virtually impossible to draw any meaningful comparisons (Smidt et al. 2009).

Cable logging is more complex and expensive than ground based logging, which places Alpine forestry at a general disadvantage in terms of pure harvesting cost. Even the lowest logging rates recorded in this study are much higher than the harvesting costs reported for the boreal forests of northern Europe (11 € m–3, Gerasimov et al. 2013), the eucalypt plantations of western Europe (<20 € m–3, Magagnotti et al.

2011) and South America (8 US$ t–1 or 6.5 € t–1, Olive­ira and Seixas 2012) or the pine plantations of the Southeastern USA (9–10 US$ m–3 or ca. 7 € m–3, Conrad et al.

2013). The main problem is steep terrain, which prevents using cost effective ground-based logging. On Alpine highland plateaus, it is possible to resort to mechanised cut-to-length harvesting by deploying the classic harvester-forwarder chain (Zambelli et al.

2012), and in that case harvesting cost is not much higher than reported for northern Europe (ca. 12 € m–3, Spinelli and Magagnotti 2013). Fortunately, the high cost of steep terrain logging is largely offset by the high value of slow growth alpine timber. In northern Italy, the current price of saw logs is higher than 75 € m–3 at the landing (IRE 2013). In Austria, saw log price often exceeds 90 € m–3, for product delivered to the sawmill (Federal Ministry of Agriculture 2012). In Slovenia the average saw log prices vary between 70 € m–3 and 76 € m–3, at landing (SI-Stat 2013).

Published data are available for the calculated technical cost of cable harvesting in a number of different Alpine countries. Stampfer et al. (2010) reported a cost of 32 € m–3 for a typical cable logging operation in Austria. Spinelli et al. (2008) reported figures between 28 and 40 € m–3 for northern Italy, based on two case studies. Pischedda et al. (2001) presented four case studies in the French and Italian Alps, with harvesting costs varying from 29 to 64 € m–3. Chagnon and Pischedda (2005) indicated costs between 42 and 58 € m–3, specifically for the French Alps. Finally, Thees et al. (2011) reported costs between 45 and 65 € m–3 for the Swiss Croat. j. for. eng. 36(2015)2

R. Spinelli et al.

Alps. Of course, these figures should first be annualized to 2013 values before making any comparisons.

However, the bracket is so wide that these data could not offer much more than a general confirmation. In particular, it is not reasonable to use these data for gauging the difference between technical cost and market rate. In principle, such an attempt could be made by introducing the technical information contained in the contracts in one of the many cost calculators developed for Alpine cable harvesting operations, such as the Swiss Hepromo (Frutig et al. 2009), the French Newfor (Magaud 2013) or the Italian HEKK (Stauder 2013). These software programs return a technical harvesting cost based on user entered input data.

However, all such programs can calculate a proper harvesting cost only if they receive accurate information about machine service life and annual use, which could not possibly be inferred from the content of the individual contracts. Such information is crucial to a reliable cost calculation, and it cannot be approximated without much prejudice for the accuracy of the eventual cost estimate (Spinelli et al. 2011).

For this reason, it is best to use the regression coefficient R2 as the main indicator for the effect of nontechnical parameters on logging rate. In fact, the main reason for developing such a general multi country model was to use its regression coefficient as a basis for discussing the relative weight of technical and nontechnical factors, not for predicting general contract rates that do not exist in reality, since every country has its own individual market rates. The regression equation estimated from the study data can only explain 40% of the variability in the data pool. This is the variability caused by such main technical factors as tree size, line length and tract size. Therefore, 60% of the variability in the data pool is caused by other factors, both technical and non-technical. There are many technical factors that were not included in the current regression, and especially lateral yarding distance and two staging. The latter was not included in the regression because information was available only for two of the five countries included in the survey, and therefore inclusion of two staging as an independent variable would have generated a large number of missing values, determining severe unbalance. The same is true for silvicultural treatment, which may significantly affect cable logging cost (Hartley and Han 2007). In this study, information about silvicultural treatment was not collected directly, but it could have been inferred from removal per hectare. An alternative measure for removal that relates more directly to the use of the cable yarding system is to use the volume per unit length of cable corridor (m3 m–1). For example, for the French data there was a very strong relation-

199

R. Spinelli et al.

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

ship between Contract Rate (€ m–3) and volume per metre of cable corridor. Unfortunately, for both of these measures collinearity prevented using removal as a meaningful independent variable.

However, the regression included both tree size and line length, which have long been recognized as the most important variables when predicting yarder productivity (LeDoux 1985). Tract size is also an important predictor of harvesting cost (Greene et al. 1997, Moldenhauer and Bolding 2009, Visser 2012) and it was included as an independent variable in our regression equation. For this reason, it is unlikely that the technical factors excluded from our equation may account for 60% of the variability in the data pool: a substantial part of that variability may be explained by non-technical factors, and especially local market dynamics and national economics.

National differences certainly play an important role, especially for Switzerland. The Swiss contract rates are significantly higher, which is in part explained by the much higher labour cost, and a low productivity caused by several circumstances (e.g. difficult terrain/working conditions). In fact, higher wages cause many foreign nationals to look for employment in Switzerland from the neighbouring countries and justify a sustained cross border commuting pattern (Banfi et al. 2005).

The French contract rates are the second highest, and that could be explained by the longer mean line length. Harvesting cost generally increases with line length, and such increase can become quite steep if the distance exceeds the capacity of modern tower yarders, as it often occurs in France. In that case, it is necessary to resort to traditional long distance cable systems, which incur much higher set up and dismantle costs compared with modern tower yarders (Stampfer et al. 2006). The high French rates may also be caused by the unbalance between the demand for yarding services and the actual availability. Grulois (2007) reports about the decline of cable yarding in France, which occurred over the past decades, when skidders became the exclusive choice of loggers in the French Alps. Efforts are being made to re-introduce cable yarding in southern France, but it will take several years before a large enough yarder fleet will be available in the area. In the meantime there are very few yarder operators in the French Alps, although the demand for cable yarding services is increasing. The longer line length and the larger tree size found in France may also depend on the preference for skidder operations in the past decades. French yarding operators are likely targeting those forests that are far from the road and have turned over-mature due to delayed regeneration harvests.

200

The study also highlights some important differences in the silvicultural practice applied in different countries. In particular, removals are much higher in Switzerland than in any other country included in this study. However, no contracts in our survey concerned clear cuts, and the very intense Swiss cuts were still selection cuts, applied to heavily overstocked stands.

Of course, this is only the case of those owners who joined the study: we did not conduct comprehensive national surveys, and therefore our results cannot lead to any conclusions about the prevalent silvicultural practice in any of the participating countries. However, our results may be indicative of trends, and they point to a very low removal in Italy and at a rather large one in Switzerland. It is certainly worth noticing that the Swiss contract rates are very high. This may further support the hypothesis of a large influence on non-technical market factors on establishing local cable logging rates.

4. Conclusion Contract rates differ significantly between countries, and are proportional to local labour cost. Differences are related to different work conditions, and especially removal per hectare, line length, tract size and tree size. However, technical factors explain only part of the difference in local contract rates, and nontechnical market factors also seem to have a strong effect. In turn, that may make cross border activity a good business for those loggers based where labour cost is lowest. Contract rate differences must be considered especially in the event of assembling a large international task force to face a sudden local crisis, such as catastrophic forest damage. Loggers cannot be drawn to countries with lower contract rates, unless rates are adjusted. While showing that actual contract rates are only partly related to technical factors, this study also highlights the lack of other similar studies, which contrasts with the large number of studies about technical cost. Future research should pay more attention to actual contract rates and expand this survey to cover a larger number of countries: this type of information is especially relevant to sound forest management, because actual contract rates may limit silvicultural choice.

Acknowledgements This study is a part of the project »Organization and rationalization of logging operations« funded by Regione Autonoma Friuli-Venezia Giulia, Servizio Gestione Forestale e Produzione Legnosa.

Croat. j. for. eng. 36(2015)2

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

5. References Baker, S., Greene, D., Harris, T., Mei, R., 2013: Regional Cost Analysis and Indices for Conventional Timber Harvesting Operations. Wood Supply Research Institute, University of Georgia, Athens, Georgia (USA), 49 p.

Balsiger, J., 2008: Regional Sustainable Development in the European Alps. European University Institute Working Paper MWP 2008/23. European University, San Domenico di Fiesole, Firenze (Italy). ISSN: 1830–7728, 35 p.

Banfi, S., Filippini, M., Hunt, L., 2005: Fuel tourism in border regions: the case of Switzerland. Energy Economics 27(5): 689–707.

Binder, C., Hofer, C., Wiek, A., Scholz, R., 2004: Transition towards improved regional wood flows by integrating material flux analysis and agent analysis: the case of Appenzell Ausserrhoden, Switzerland. Ecological Economics 49(1): 1–17.

Bont, L., Heinimann, H., 2012: Optimum geometric layout of a single cable road. European Journal of Forest Research 131(5): 1439–1448.

Chagnon, J., Pischedda, D., 2005: La forêt cable. Dossier spécial sur le débardage par câble. Le Bois International. Givors (France), 24 p.

Conrad, J., Bolding, C., Aust, M., Smith, R., Horcher, A., 2013: Harvesting productivity and costs when utilizing energywood from pine plantations of the southern Coastal Plain USA. Biomass and Bioenergy 52: 85–95.

CTBA, 2004: Technical guide on harvesting and conservation of storm damaged timber. CTBA, Paris (France). Available on line at: www.stodafor.org, 105 p.

Dorren, L., Berger, F., Imeson, A., Maier, B., Rey, F., 2004: Integrity, stability and management of protection forests in the European Alps. Forest Ecology and Management 195(1): 165–176.

Eurostat, 2013: Labour cost in the EU 27. Eurostat News Release 54. Downloaded from http://epp.eurostat.ec.europa.eu, 4 p.

Federal Ministry of Agriculture, 2012: Austrian market report. Vienna (Austria), 16 p.

Frutig, F., Thees, O., Lemm, R., Kostadinov, F., 2009: Holzernteproduktivitätsmodelle HeProMo – Konzeption, Realisierung, Nutzung und Weiterentwicklung. In: Thees O., Lemm R. (Ed.) Management zukunftsfähige Waldnutzung.

Grundlagen, Methoden und Instrumente. Birmensdorf, Eidg. Forschungsanstalt für Wald, Schnee und Landschaft WSL; Zürich, VDF.: 441–466.

Gerasimov, Y., Sokolov, A., Syunev, V., 2013: Development trends and future prospects of cut-to-length machinery. Advanced Materials Research 705: 468–473.

Greene, D., Harris, T., DeForest, C., Wang, J., 1997: Harvesting cost implications of changes in the size of timber sales in Croat. j. for. eng. 36(2015)2

R. Spinelli et al.

Georgia. Southern Journal of Applied Forestry 21(4): 193– 198.

Grulois, S., 2007: Cable yarding in France: past, present and perspectives. In: Sessions J., Havill Y. Proceedings of the International Mountain Logging and 13th Pacific Northwest Skyline Symposium. Corvallis, OR (USA), 299 p.

Hartley, D., Han, H.S., 2007: Effects of Alternative Silvicultural Treatments on Cable Harvesting Productivity and Cost in Western Washington. Western Journal of Applied Forestry 22(3): 204–212.

Hasel, A., 1946: Logging cost as related to tree size and intensity of cutting in Ponderosa pine. Journal of Forestry 44(8): 552–560.

Klaus, M., Holsten, A., Hostert, P., Kropp, J., 2011: Integrated methodology to assess windthrow impacts on forest stands under climate change. Forest Ecology and Management 261(11): 1799–1810.

Koch, S., Schwarzbauer, P., Stern, T., 2013: Monthly wood supply behavior of associated forest owners in Austria–Insights from the analysis of a micro-econometric panel. Journal of Forest Economics 19(3): 331–346.

IRE, 2013: Listino dei prezzi all’ingrosso. Chamber of Commerce, Industry, Crafts and Agriculture of Bolzano. Bullettin 12. Bolzano (Italy), 33 p.

Le Doux, C., 1985: Stump-to-mill timber production cost equations for cable logging eastern hardwoods. USDA Forest Service, Northeast Forest Experiment Station Res. Pap. NE 566. Bromall, PA (USA). p. 6.

Magagnotti, N., Nati, C., Pari, L., Spinelli, R., Visser, R., 2011: Assessing the cost of stump-site debarking in eucalypt plantations. Biosystems Engineering 110(4): 443–449.

Magaud, P., 2013: A methodological approach towards specific cost and productivity models for cable yarding in the French Alps. In: Talbot B. (Ed.) Proceedins of the IUFRO Conference: Forest operations in mountainous conditions.

June 2–5, Ås (Norway), 17 p.

Mason, B., Kerr, G., Simpson, J., 1999: What is continuous cover forestry? Forestry Commission Information Note 29.

Forestry Commission, Edinburgh (UK).

Mathews, D.M., 1942: Cost Control in the Logging Industry.

McGraw-Hill Book Company. New York (USA), 374 p.

McDonagh, K., Meller R., Visser, R., McDonald, T., 2004: Improving harvesting system efficiency through system and machine simulation. Southern Journal of Applied Forestry 28(2): 91–99.

Moldenhauer, M., Bolding, C., 2009: Parcelization of South Carolina’s private forestland: loggers’ reactions to a growing threat. Forest Products Journal 59(6): 37–43.

Onida, M., 2009: The Alps: eight countries, a single territory.

Permanent Secretariate of the Alpine Convention. Innsbruck (Austria), 89 p.

201

R. Spinelli et al.

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

Oliveira, E., Seixas, F., 2012: Energy analysis of two eucalyptus harvesting systems in Brazil. Proceedings of the 35th Council of Forest Engineers Annual Meeting. Sept. 9–12.

New Bern, North Carolina (USA), 7 p.

Pischedda, D., Chagnon, J.L., Leclerc, D., Millot, M., Mermin, E., Chauvin, C., 2001: Techniques for the sustainable, multifunctional management of mountain forests: the case of the French and Italian Alps. In: Arzberger U., Grimoldi M. (Ed.) Proceedings of the FAO/ECE/ILO Workshop: New trends in wood harvesting with cable systems for sustainable forest management in the mountains. June 18–24, Ossiach (Austria).

Schelhaas, M., Hengeveld, G., Moriondo, M., Reinds, G., Kundzewicz, Z., Maat, H., Bindi, M., 2010: Assessing risk and adaptation options to fires and windstorms in European forestry. Mitigation and Adaption Strategies for Global Change 15(7): 681–701.

Seidl, R., Rammer, W., Lexer, M., 2011: Climate change vulnerability of sustainable forest management in the Eastern Alps. Climatic Change 106(2): 225–254.

SI-Stat, 2013: Statistical data portal from Statistical office of the Republic of Slovenia, http://pxweb.stat.si/pxweb/dialog/ statfile1.asp. Accessed on August 8th, 2014.

Smidt, M., Tufts, R., Gallagher, T., 2009: Logging efficiency and cost. Alabama Cooperative Extension System Report ANR-1347. Auburn, Alabama (USA). Available on line at: www.aces.edu, 7 p.

Spinelli, R., Magagnotti, N., 2011: The effects of introducing modern technology on the financial, labour and energy performance of forest operations in the Italian Alps. Forest Policy and Economics 13(7): 520–524.

Spinelli, R., Magagnotti, N., 2013: The effect of harvest tree distribution on harvesting productivity in selection cuts.

Scandinavian Journal of Forest Research 28(7): 701–709.

Spinelli, R., Magagnotti, N., Della Giacoma, F., 2008: Meccanizzazione nelle fustaie alpine: due diversi sistemi di lavoro (mechanization in Alpine forests: two different harvesting systems). Sherwood 147: 45–49.

Spinelli, R., Magagnotti, N., Picchi, G., 2011: Annual use, economic life and residual value of cut-to-length harvesting machines. Journal of Forest Economics 17(4): 378–387.

Spinelli, R., Magagnotti, N., Nati, C., 2010: Benchmarking the impact of traditional small-scale logging systems used

202

in Mediterranean forestry. Forest Ecology and Management 260(11): 1997–2001.

Spinelli, R., Magagnotti, N., Facchinetti, D., 2013: A survey of logging enterprises in the Italian Alps: firm size and type, annual production, total workforce and machine fleet. International Journal of Forest Engineering 24: 109–120.

Spinelli, R., Di Gironimo, G., Esposito, G., Magagnotti, N., 2014: Alternative supply chains for logging residues under access constraints. Scandinavian Journal of Forest Research 29(3): 266–274.

Stampfer, K., Visser, R., Kanzian, C., 2006: Cable corridor installation times for European yarders. International Journal of Forest Engineering 17(2): 71–77.

Stampfer, K., Leitner, T., Visser, R., 2010: Efficiency and ergonomic benefits of using radio-controlled chokers in cable yarding. Croatian Journal of Forest Engineering 31(1): 1–9.

Stauder, M., 2013: Personal communication. TIS Innovation Park, Bolzano (Italy).

Stuart, W., Grace, L., 2011: The logging cost index. Forest Operations Review 5: Available on line at: www.forestoperationsreview.org/departments/features/item/79-the-logging-cost-index, 4 p.

Thees, O., Frutig, F., Fenner, P., 2011: Colheita de madeira em terrenos acidentados – recentes desenvolvimentos técnicos e seu uso na Suíça. In Malinovski J. R. et al. (Ed.) Anais XVI Seminário de atualizacão em sistemas de colheita de madeira e transporte florestal. 11 e 12 de Abril de 2011. Campinas, São Paulo, Brasil, 125–146 p.

Visser, R., Stampfer, K., 2003: Tree-length system evaluation of second thinning in loblolly pine plantations. Southern Journal of Applied Forestry 27(2): 77–82.

Visser, R., 2010: Methodologies for Setting timber Harvesting (Logging) Rates. New Zealand Journal of Forestry 55(3): 10–11.

Visser, R. 2012: Benchmarking to Improve Harvesting Cost and Productivity: 2011 Update. Future Forests Research Ltd, Rotorua, New Zealand. Harvesting Technical Note HTN0408, 2012: 7 p.

Zambelli, P., Lora, C., Spinelli, R., Tattoni, C., Vitti, A., Zatelli, P., Ciolli, M., 2012: A GIS decision support system for regional forest management to assess biomass availability for renewable energy production. Environmental Modelling and Software 38: 203–213.

Croat. j. for. eng. 36(2015)2

Cable Logging Contract Rates in the Alps: the Effect of Regional Variability ... (195–203)

R. Spinelli et al.

Authors’ address:

Received: March 25, 2015 Accepted: August 6, 2015 Croat. j. for. eng. 36(2015)2

Raffaele Spinelli, PhD.* e-mail: spinelli@ivalsa.cnr.it Natascia Magagnotti, PhD. e-mail: magagnotti@ivalsa.cnr.it CNR IVALSA Via Madonna del Piano 10 50019 Sesto Fiorentino (FI) ITALY Assoc. Prof. Rien Visser, PhD. e-mail: rien.visser@canterbury.ac.nz University of Canterbury College of Engineering Private Bag 480 Christchurch NEW ZEALAND Oliver Thees, PhD. e-mail: oliver.thees@wsl.ch Swiss Federal Research Institute WSL Zürcherstrasse 111 8903 Birmensdorf SWITZERLAND Udo Hans Sauter, PhD. e-mail: udo.sauter@forst.bwl.de Forest Research Institute Baden-Württemberg (FVA) Wonnhaldestraße 4 79100 Freiburg GERMANY Nike Krajnc, PhD. e-mail: nike.krajnc@gozdis.si Slovenian Forestry Institute Department of Forest Techniques and Economy Večna pot 2 1000 Ljubljana SLOVENIA Catherine Riond, PhD. e-mail: catherine.riond@onf.fr ONF Pôle R&D Montagne Quai Charles Roissard 42 73026 Chambery CEDEX FRANCE * Corresponding author

203

Original scientific paper

Damage Caused by Wheeled Skidders on Cambisols of Central Europe Michal Allman, Michal Ferenčík, Martin Jankovský, Miroslav Stanovský, Valéria Messingerová Abstract Machine traffic and timber skidding significantly affect the soil surface and soil properties.

The effects are mostly negative and result in soil erosion, worsening of soil properties and inhibition of the growth of roots and soil organisms. In this study, we evaluated forest soil damage caused by the HSM 805 HD wheeled skidder during timber skidding in selected forest stands in the School Forest Enterprise in Zvolen. We estimated the limits for operation of forest machines in the stands and evaluated the moisture content and bulk density of the soil, CO2 concentration in the upper layer of the soil, determined the soil texture, Atterberg limits and critical moisture using the Proctor test, CBR test and examining the depth of ruts on skid trails. The measurements were taken from undisturbed forest soil unaffected by skidder traffic, the ruts and between the ruts.

The results showed significant differences between the values of soil samples from undisturbed soil and the soil affected by the skidder. The exceeding of CO2 concentration limits and bulk density in the soil from the ruts were recorded in both stands. The methods used present the basic methodology for evaluating the effect of logging machinery on forest soil and for setting limits that will allow or prohibit the operation of logging machinery according to forest stand conditions. The moisture content of soil, when it changes from the solid to plastic phase, was chosen as the limit for machine operation. This value is also easy to measure.

Keywords: limits, soil damage, moisture, timber skidding

1. Introduction Machinery used for timber skidding causes various damage to the remaining trees and forest soil. This damage is primarily caused by using machinery in unsuitable conditions, or by using unsuitable machines in given conditions. For example, using large skidders in the first thinning, or skidding timber after heavy rainfall usually cause damage far beyond any limits. Determination of the limits is the most important way to decrease the damage caused by skidders.

Most methods evaluate the damage after the logging process is finished. Such methods are based on measurement of changes in soil properties caused by the machinery. The measured properties are usually penetration resistance, bulk density of soil, moisture content of soil and depth of ruts (Stanovský et al. 2012).

Croat. j. for. eng. 36(2015)2

Problems connected with the influence of skidding machinery on forest soil have been studied by many authors and connected with the rapid development of the machinery during the past 50–60 years (Ebel 2006).

The negative influences of machinery on the soil surface has been researched by many authors including Hildebrand (1981), Hildebrand and Wiebel (1982), Benecke (1982), Bredberg and Wästerlund (1983), Löffler (1983), Becker et al. (1986), Löffler (1986), and Zander (1988).

The movement of forest machinery on the soil surface causes changes of its shape. The changes are dependent on soil moisture, its physical properties and other characteristics. These changes usually result in erosion of skid trails (relocation of upper soil horizons and structural changes of the soil). Mechanisa-

205

M. Allman et al.

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205–215)

tion and modernisation of the logging process increase work costs with unchanged timber prices (Matthies et al. 1995, Hamberger 2002, Wehner 2002, Ziesak 2004). This results in a continued performance and dimensional growth of forest machinery (Forbrig 2000), especially the machinery weight. This is the most dangerous factor causing negative effects on forest stands after forest harvesting and timber transportation. The present paper focuses on skidders. In Slovakia more than 70% of harvested timber is transported from forest stands to forest landings by wheeled skidders.

Various authors, such as Löffler (1986), Becker et al. (1986), Hoffman and Becker (1990), came to the conclusion that even the first pass of heavy forest machinery on the surface of forest soil causes significant structural changes in its upper layers. There are many factors that affect the rate of soil damage after logging and only their detailed research will enable dealing with the damage. Soil moisture is one of the most important of them (Majnounian and Jourgholami 2013).

Determination of critical moisture content in soil is crucial for devising the limits for using harvesting and transporting machinery in unfavourable conditions.

Two basic limits of soil moisture content can be used in forestry practice: Þ Critical value of soil moisture for its compaction.

Soil with moisture content at this level will be the most susceptible to compaction caused by forest machinery traffic; Þ In case of further increase of the soil moisture, at a certain point it will reach the second critical value – plastic limit. Creation of tracks in the soil starts at this very moment if forest machinery is used in such conditions.

It is necessary to implement soil protection into forestry certification systems, such as the Forest Stewardship Council (FSC) Programme for the Endorsement of Forest Certification (PEFC). Selecting the suitable technology and season for harvesting and timber skidding is the basis for minimising the impacts of technology on the forest environment.

The present study aimed to estimate simple limits for operation of the HSM 805 HD skidder in given conditions. These limits should be easy to measure/ estimate for forestry practice without the need for special equipment. The next goal was to compare the properties of forest soil in the undisturbed forest stand and the soil in the skidding line to evaluate the level of damage caused by the operation of the HSM 805 HD wheeled skidder. We also assessed the efficiency of individual methods for forestry practice.

206

2. Material and methods The measurements were carried out in two forest stands with similar conditions (slope, exposition, soil type, shelterwood system). Motormanual felling and skidding with the HSM 805 HD skidder took place in both stands. Basic data about the forest stands is presented in Table 1 and data about the skidder in Table 2.

Felling in the forest stands was conducted in different seasons. Forest stand 588 was felled in August 2012 during a dry period, when the total recorded precipitation in the given area was approx. 10 mm per month (SHMU 2012). The second forest stand 574 was felled in October 2012, during a period with inten­ sive rainfall, when the total precipitation was almost 160 mm per month (SHMU 2012). The measurements in both stands were performed immediately after the Table 1 Basic data about forest stands Stand number

588

574B11

VŠLP Zvolen

VŠLP Zvolen

LS Budča

LS Budča

48.58348387N

48.590818386N

19.054429793E

19.04499979E

Area, ha

4.1

1.6

Soil type

Cambisol

Cambisol

Soil texture

Silt loam

Sandy loam

Slope, %

25

20

Place GPS coordinates

DB 30%, BO 23%, HB 17%, BK 9%,

Tree species*

JS 6%, SM 3%, CR 2%, OS 1%

Age, years

BK 78%, BO 11%, DB 7%, HB 3%, CR 1%

110

110

215

411

Treatment, m ha

52

257

Treatment type

Seed cut

Overstory removal cut

Exposition

SV

SV

Felling time

August 2012

October 2012

Number of plots/ number

6/12

–/22 (spacing 15 m)

Felled volume, m3 3

-1

of measuring lines * DB – oak, BO – pine, HB – hornbeam, BK – beech, JS – ash, SM – spruce, CR – cherry

Croat. j. for. eng. 36(2015)2

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205–215)

Soil samples were collected from both opposite edges of the plots, which were transversal to the skidding lines (two measuring lines on one sample plot); Þ The residual stand in stand 574 was not present because an overstory removal cut was carried out. Therefore, data on the remaining stand was not recorded and it was not necessary to establish sample plots. Hence, we only established lines transversal to the skidding line. The number of lines was estimated on the basis of the equation for setting the minimal sample size according to Scheer (2010) and Schürger (2012).

Table 2 Technical parameters of the HSM 805 HD skidder 11 700

Weight, kg

(including crane and grapple)

Tyre pressure, kPa

65 (Mellgreen 1980)

Engine

IVECO NEF

Displacement, cm3

3908 –1

Max. power/at revolutions, kW/n min

110/2500

Max. torque/at revolutions, Nm/n min–1

490/1400

Transmission

CLARK 2000; powershift, converter, 3+3 speeds, additional transmission (2 speeds)

Tyres

600/60–30.5

Axles

NAF

Winch

ADLER HY 20–double drum

Pulling force, kN

2x100

Winch speed, mmin–1

0–150

Max. cable length/at diameter, m/mm

80/14

Crane

LOGLIFT F 101 RT 72

Reach, m

7.2

Gross lift torque, kNm

125

Net lift torque, kNm

98

Slewing torque, kNm

34

logging process to avoid changes to the soil properties.

This work was based on the comparison of properties of undisturbed soil in the forest stands with properties of soil in the skid trails created during the logging process. The properties were only measured in newly created skid trails to avoid the influence of previous treatments. The intensity of traffic was described by means of treatment intensity (Table 1) and by the ground pressure of the machine (Table 2).

Determination of the sample size was conducted in two ways: Þ In stand 588 (after seed cut), we used square shaped sample plots due to the fact that we also recorded the damage of the remaining stand (not presented in this paper). The number of sample plots was determined on the basis of the Finnish method adjusted by Ulrich (2002). The sample plots (20x20 m) were set so that the centre of the plot was placed on the skidding line.

Croat. j. for. eng. 36(2015)2

M. Allman et al.

In both cases, the measurements were carried out on established lines. Following methods were used: Þ measurement of moisture content and bulk density of the soil (by using the Eijkelkamp sampling cylinder set), Þ measurement of the CO2 concentration in the upper layer of the soil, using the Vaisala MI70 device, Þ Proctor test (for determination of the critical moisture at which the soil is maximally compressible), Þ assessment of soil texture using the washing and densimetric method according to Casagrande, Þ determination of the plasticity index (Atterberg limits), Þ CBR test, Þ measurement of the transversal soil surface profile on skid trails (McMahon 1995).

For the determination of the soil moisture and bulk density, three soil samples were taken from each established line. One sample was taken from the soil in undisturbed forest, the second was from the rut, and the third was taken from between the ruts. The Eijkelkamp sampling set was used to take samples. The set contains cylinders with a capacity of 100 cm3. We used the standard sampling procedure and took the sample after removing the upper organic layer of the soil. The samples were weighed (mw) and we calculated the bulk density in the natural status. After this procedure, the samples were dried at 105°C, until they reached a constant weight – dry weight (m0). Then, the moisture content of the samples was calculated (Wr) according to the following equation:

Wr =

mw − m0 × 100,% m0

(1)

Where: mw weight of fresh sample, g; m0 weight of dry sample, g.

207

M. Allman et al.

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205–215)

The calculated bulk densities of fresh soil from individual forest stands were compared with critical values. The critical bulk density values for individual soil types inhibiting the growth of roots were estimated according to Table 3.

(1995) method. These measurements were carried out at the established lines, transversally to the skidding lines. One depth measurement was performed per every 20 cm of skid trail width. Damage levels were estimated according to Weise (2002) in Lukáč (2005).

Table 3 Critical soil bulk density values which inhibit the growth of roots (Arshad and Coen 1992) in Skoupy et al. (2011)

For the evaluation of physical characteristics of the soils, two soil samples were taken from both forest stands. One sample was from the undisturbed forest stand and the other was from the rut. Washing and densimetric methods (Casagrande) were used for assessing the soil texture.

Soil type (texture)

Critical bulk density of soil, g cm-3

Sand, loamy sand

1.80

Fine sands and loamy sands

1.77

Sandy loam

1.75

Loam, sandy clay loam

1.70

Clay loam

1.65

Sandy clay

1.60

Silt, silt loam

1.55

Silty clay loam

1.50

Silty clay

1.45

Clay

1.40

The concentration of CO2 in soil was measured at two locations on the measuring line. One measurement point was in the rut and the other was placed in undisturbed forest soil. Both measurements were conducted at the same time using the VAISALA MI70 device connected with two identical probes.

CO2 measurement probes were placed close to the points where we took the soil samples. The distance of the CO2 measuring points from the soil sampling points was as small as possible, but far enough to avoid mutual influence of the measurements (approx. 20 cm). Both probes were placed in previously drilled holes. The depth of the holes was 10 cm and the diameter was identical with the diameter of the probes. Gaps between the soil and the probe were sealed. The measuring range of both probes was between 0–5% CO2 in the soil atmosphere. The device recorded the values every 15 s during a 5 min period in every sampling place. Original Vaisala software was used for processing the recorded data. The critical value of the CO2 concentration in soil for the growth of roots and for the survival of soil microbiota is 0.6% (Güldner 2002).

We also determined the damage to the soil surface and measured the depth of the ruts from the undisturbed surface on the lines using the McMahon

208

The standard Proctor test was used for the assessment of critical moisture content (for compaction) in the soil. The plastic limit of individual soils was estimated using the method described in detail by Poršinsky et al. (2006).

The mechanical strength of soil was estimated using the California bearing ratio (CBR) test. The principle of the test is based on pressing a steel cylinder with a diameter of 50 mm into the soil sample at a speed 1.27 mm min-1 and recording the pressure at depths of 2.54 mm and 5.08 mm. The recorded pressure was compared with the standardised pressure for the given depth and the final CBR values were estimated according to the following formula:

Where:

CBR =

p × 100,% ps

(2)

p recorded cylinder pressure at given depth, kPa; ps standard pressure, 70.8 kPa at 2.54 mm and 106 kPa at 5.08 mm.

We tested the soil sample with the maximum moisture estimated in the Proctor test. A sample with increased moisture was tested after 96 hour saturation in water.

For processing and evaluating the collected data, we used the MS Excel, Statistica 10.0, EijkelkampPenetroViewer and Vaisala MI 70 software products.

3. Results As presented in Table 4, we recorded a higher average moisture content of all soil samples taken from stand 574 when compared with stand 588. The average moisture of samples from the stand, rut and between the ruts in stand 588 was almost identical. In stand 574, we also recorded only minor changes in soil moisture.

Croat. j. for. eng. 36(2015)2

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205â&#x20AC;&#x201C;215)

Table 4 Average soil moisture values in forest stands Stand

Average moisture of soil, %

588

574

Undisturbed

20.4

36.2

In rut

20.9

33.6

Between ruts

20.8

35.1

was used for the validation of the statistical confidence of differences. Test results are presented in Tables 5 and 6. Significant differences are marked with +.

Table 5 Results of the t-test for mean bulk density values of samples from individual forest stands Stand

Fig. 1 shows significant differences in the bulk densities of soil samples from individual locations. The critical bulk density value (1.55 g cm-3 according to Table 3) in forest stand 588 (silt loam) was exceeded in the rut.

The average soil moisture content was approx. 21% (Table 4). The soil texture in forest stand 574 is sandy loam with the addition of gravel. The critical bulk density value for such soil type is 1.75 g cm-3 (Table 3). In this case, the average bulk density from the rut samples exceeded the critical value (1.79 g cm-3) and the average bulk density of samples taken between ruts was 1.72 g cm-3. The average soil moisture content in this stand ranged from 33â&#x20AC;&#x201C;36%. The moisture content was considerably higher when compared with stand 588. The higher soil moisture content was caused by higher precipitation during skidding in stand 574.

We examined the differences between the mean bulk density values of individual soil samples from all of the localities (wet samples after sampling and the same samples after drying) from both stands. t-test

M. Allman et al.

588

574

Samples

Undisturbed vs. Undisturbed vs.

Rut vs.

Rut

Between ruts

Between ruts

Wet

+

+

-

After drying

+

+

-

Wet

+

+

-

After drying

+

+

-

Significant differences were recorded between the undisturbed soil samples against the rut samples as presented in Table 5. These differences proved to be present in all four cases (fresh and dried samples from both forest stands). Statistically significant difference was also found between the undisturbed soil samples and the samples taken from between the ruts. On the other hand, no statistically significant differences were found between the rut samples and the samples taken from between the ruts.

The differences were also tested between the mean dry sample values from individual forest stands. The results of the test are presented in Table 6.

Fig. 1 Average bulk densities of soil samples from individual locations: a) stand 588, b) stand 574 Croat. j. for. eng. 36(2015)2

209

M. Allman et al.

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205â&#x20AC;&#x201C;215)

Table 6 Results of the t-test for mean bulk density values of soil samples from both stands after drying Dry samples

588 vs. 574

Samples from undisturbed soil

+

Samples from ruts

-

Samples between ruts

-

The difference between the average bulk soil densities taken from the undisturbed area in both forest stands was found to be statistically significant. On the other hand, no significant differences were recorded between the stands in the case of rut samples or samples taken from between the ruts. Therefore, it can be concluded that the soils in these stands (with different parameters) were compacted to similar bulk densities.

The measured CO2 concentration values in the soil atmosphere from both stands are presented in Fig. 2.

A significant increase of CO2 content was recorded in the ruts of both forest stands compared to the undisturbed soil values. In stand 588 the CO2 level reached 190% of the value from the undisturbed soil (simultaneous measurements). CO2 concentration in the rut almost reached the critical value (0.6%) in this stand. We observed a similar situation in the other stand (574), where the CO2 level in the soil of the rut was 2.93 times higher than the level of the undisturbed soil. In this stand the CO2 concentration in undisturbed soil was near critical, and the concentration in the rut

exceeded the critical value 2.88 times. A comparison of values between the stands showed a significant difference in both values. The CO2 concentration of the undisturbed soil from stand 574 was approx. 193% higher when compared with stand 588. A similar situation is seen between the values measured in ruts, where the increment was approx. 304%. These results show a significant increment of CO2 concentration caused by machinery traffic in both forest stands. We also recorded a significant difference in the concentration between the undisturbed stands and the ruts.

These differences were caused by an increase of precipitation during skidding in stand 574.

Analyses of soil properties was carried out to determine the critical soil moisture content values in both stands (Tables 7 and 8). When the soil moisture content reaches or exceeds the critical value, all machinery traffic in forest stands should be prohibited.

The soil texture in stand 588 was determined as silt loam. The technical standard STN 72 1002 assigns this soil into category VII, VIII, or IX; the most frequent soil particle was silt. These soils are prone to ground freezing, and after water saturation their load capacity drops by approx. 40% compared to soil in optimal conditions. The critical soil moisture value, which limits the use of technology, ranges within 31â&#x20AC;&#x201C;33%. Intensive compaction of upper soil layers starts if the soil moisture reaches 24%.

According to our analyses, the soil in stand 574 is of the same type as the previous one and their properties are similar. The critical value for creation of ruts (plastic

Fig. 2 CO2 concentration in the soil atmosphere of the forest stand (P) and the rut (K) from both stands: a) stand 588, b) stand 574

210

Croat. j. for. eng. 36(2015)2

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205â&#x20AC;&#x201C;215)

Table 7 Soil texture and critical soil property values in stand 588 Soil texture

Undisturbed stand

Rut

Clay

5

5

Silt

68

Sand

Undisturbed

Rut

Moisture content, %

22.7

30.8

26.3

33.3

71

Bulk density, kg m-3

1480

1340

1480

1370

27

24

Porosity, %

44.6

49.8

44.6

48.7

Liquid limit

44

42

Saturation, %

75.4

82.7

87.5

93.7

Plastic limit

31

33

CBR depth ratio 2.54 mm

5.7/40

4.3/30

5.4/38

3.1/22

6.0/64

4.4/47

5.3/56

3.3/35

Max. compaction Plasticity index

24

31

13

9

Table 8 Critical soil property values in stand 574 Soil texture

Silt loam Undisturbed stand

Rut

Clay

18

11

Silt

47

52

Sand

35

37

Liquid limit

43

35

Plastic limit

30

23

22

22

13

13

Locality Particle size distribution %

Water content

After 96 hour of saturation Rut

%

%

Optimum moisture

Sample

Undisturbed

Particle size distribution

Water content

Table 9 Results of the CBR test samples from stand 588

Silt loam

Locality

M. Allman et al.

%

Max. compaction Plasticity index

limit) for the previously undisturbed soil of the forest stands is different than the limit for previously compacted soil (ruts, skidding lines, forwarding lines, etc.).

In this case it is 30% and 23%, respectively. The value for maximum soil compaction in this stand is 22%, both for the undisturbed soil and for the soil in ruts.

The load capacity of the soil was determined using the standard CBR test. Results of the test are presented in Tables 9 and 10. The test was divided into two parts.

In the first part, we tested the samples adjusted to moisture, when the soil is mostly susceptible to compaction (optimum moisture). We individually determined the optimum moisture for every sample using the Proctor standard test (Tables 8 and 9). In the second part, we tested soil samples, which were previously saturated for 96 hours with water.

Croat. j. for. eng. 36(2015)2

-2

% /Ncm

5.08 mm

The soil penetration resistance from the undisturbed stand 588 (Table 9) with optimum moisture was 40 N cm-2 (CBR index 5.7%) at a depth of approx.

2.54 mm, and it increased to 64 N cm-2 (CBR index 6.0%) at a depth of approx. 5.08 mm. After the test, the sample was put into water for 96 hours. The test results after saturation were as follows: the moisture content increased and the soil penetration resistance decreased. At a depth of approx. 2.54 mm, a resistance of approx. 38 N cm-2 was recorded (CBR index 5.4%), and at a depth of approx. 5.08 mm the resistance was 56 N cm-2 (CBR index 5.3%).

Penetration resistance is lower in the case of soil samples from the rut. Table 10 presents the results of the CBR soil tests from forest stand 574. In this case, even the little change (increase) of soil moisture content, after saturation with water, caused a significant decrease of the soil penetration resistance. Penetration

Table 10 Results of CBR test samples from stand 574 Optimum moisture

Sample

After 96 hour of saturation

Undisturbed

Rut

Undisturbed

Rut

Moisture content, %

18.6

22.7

21.8

26.3

Bulk density, kg m-3

1600

1480

1602

1480

Porosity, %

40.1

44.6

40.0

44.6

Saturation, %

74.3

75.4

87.2

87.5

2.54 mm

8.9/63

5.7/40

4.3/30

5.4/38

5.08 mm

9.5/107

6.0/64

3.8/40

5.3/56

CBR depth ratio %/ Ncm-2

211

M. Allman et al.

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205â&#x20AC;&#x201C;215)

Fig. 3 Percentage of individual forest stand damage classes (classification according to Weise 2002) resistance of the undisturbed stand sample adjusted to optimum moisture was 63 N cm-2 (CBR index 8.9%) at a depth of approx. 2.54 mm, and 101 N cm-2 (CBR 9.5%) at 5.08 mm. The penetration resistance of the soil sample after saturation decreased to 30 N cm-2 (CBR 4.3%) at a depth of 2.54 mm, and 40 N cm-2 (CBR 3.8%) at a depth of approx. 5.08 mm. Both optimum and saturated soil sample resistance from the rut were lower than the soil from the undisturbed stand.

The percentages of forest stand damage classes based on measured rut depths are presented in Fig. 3.

Significant differences in damage intensity were recorded in both stands. We determined that most of the ruts in stand 588 were in the moderate damage class. Only 8% of ruts with severe damage were recorded, and there were no ruts with very severe damage. In the second stand, 27% of ruts with severe and 18% with very severe damage classes were recorded, respectively. These results were affected by higher skidded volume in the second forest stand, and additionally the skidding process was conducted during the period with intensive rainfall. These results confirm the importance of carefully planning the skidding process with regard to climatic conditions and the requirement of determining the limits of soil damage.

4. Discussion The bulk densities of fresh soils showed significant differences between the undisturbed forest stand and

212

rut samples. In stand 588, the average bulk densities ranged from 1.16 g cm-3 (undisturbed stand) to 1.56 g cm-3 (rut). In stand 574, it was approx. 1.43 g cm-3 in the stand and approx. 1.79 g cm-3 in the rut. In both cases, a significant increase of soil bulk density was recorded as a result of soil compaction caused by skidder traffic, and the estimated soil bulk densities from ruts exceeded the critical values (Table 3) in both stands. Kindernay (2010) conducted research on cambisols in various parts of Slovakia after felling and timber transport made by a harvester and a forwarder. He determined the bulk density of rut soil at approx. 0.96 g cm-3, which is equal to our soil bulk density from the natural forest stand. Rab (1992) states the bulk density of rut soil is approx. 1.12 g cm-3, and Anderson et al. (1992) determined it to be 1.10 g cm-3. In our case, the upper soil layers were compacted to a higher level.

The estimation of bulk density and its comparison with critical values (Table 3) is suitable for estimating soil damage levels after skidding. However, it is not applicable as a criterion for predicting future damage caused by machine operation in given conditions.

The measurements of the soil CO2 concentration confirmed a significant increase in ruts after skidding compared to undisturbed forest stand soil. In stand 588, the CO2 concentration was approx. 0.3% in undisturbed soil and 0.57% in the rut. In stand 574, it was approx. 0.56% and 1.73% in the rut. These results are similar to measurements conducted by Gebauer et al.

(2012). They measured the soil CO2 concentrations in forest stands after timber extraction and skidding with harvester technology. Their results show that the critical value (0.6%) was significantly exceeded in almost all cases after the passage of harvesters and forwarders, and in some cases the value was exceeded several times (e.g. 1.2% and 3.4% CO2 in a harvester rut as opposed to 0.4% and 0.5% CO2 on surfaces unaffected by harvesters). The lower values of CO2 content in stand 588 were probably affected by the dry conditions during the measurements. The values from stand 574 were higher because they were taken during the wet season.

Measurement of the soil CO2 concentration is also applicable as a method for assessing soil damage after mechanical operation, but not for determining whether the machine will cause unacceptable soil damage.

This method requires a special measuring device, and the CO2 content in the soil varies during the day and during the year (Hirano et al. 2003).

In our case, the soil was the most susceptible to compaction when its moisture content was, according to the Proctor standard test, between 24.3â&#x20AC;&#x201C;31.0% in stand 588, and between 22.0â&#x20AC;&#x201C;21.8% in stand 574.

Croat. j. for. eng. 36(2015)2

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205–215)

After reaching the plastic limit, the soil changes its physical properties from the solid to plastic viscous phase, which has minimal bearing capacity even with the use of low pressure tyres, so skidder traffic on forest soil will create ruts on the skidding line. In our case, the moisture content, where the soil reached the plastic limit, was 31–33% for silt loam soil in stand 588 and 22–29% for silt loam soil from stand 574.

The actual soil moisture content in stand 588 was approx. 20.4 to 20.9%. The soil moisture content in this stand did not exceed the critical value. The moisture content of the soil from stand 574 was significantly higher than in stand 588. The critical value was exceeded in this case. The results of damage measurement (Fig. 3) show that, in the case of forest stand where the plastic limit was not exceeded, the damage is significantly lower than in stand 574, where the soil moisture exceeded the limit value. This difference was caused by heavy rainfall during skidding in stand 574.

The plastic limit of soil may be used in forestry practice as a simple criterion for the operation of forestry machinery in the skidding process. If the soil moisture content exceeds this limit, forest machines should cease operation until the moisture decreases below the set limit. The practical assessment of soil moisture content will require a moisture probe for measuring the actual moisture. For an easy and fast decision in the field, it is possible to use the »thread rolling method« (Persson 2013, Lüscher et al. 2009).

For this test, it is necessary to take a small sample of the soil and try to roll it into a thin thread. The thinner and longer the thread, the higher the soil moisture capacity. If it is possible to roll the thread to a diameter of 3 mm before it starts to break into shorter pieces, the soil reached the plasticity limit (Klobouček et al. 1979).

5. Conclusion It is impossible to choose an ideal logging technology that causes no damage to the forest soil and stand because of the high variability of environmental conditions in our forests. Despite this, it is the duty of the forest manager to choose optimal technology which causes minimal forest stand and soil damage. Exact and easily measurable limits for forest machinery will help in the decision-making process for given conditions. The limits will help to stop the operation of machinery under conditions where they would probably cause irreparable damage to the forest stand or soil and plan their operation during seasons with minimal rainfall.

Croat. j. for. eng. 36(2015)2

M. Allman et al.

Soil vulnerability depends on its texture, moisture content, the machine weight and its pressure on the soil surface. Soil moisture is one of the most important factors and is also easy to measure. We used the soil moisture content that changes the soil from the solid to plastic phase (plasticity limit) as the critical value for skidder operation. In our case, the critical values were approx. 31% in stand 588 and 30% in stand 574.

If the soil moisture reaches its critical value, it is necessary to immediately stop the operation of forest machinery until the moisture content drops below the plasticity limit. Determination of such limits for various soil types (textures) enables easy determination of the suitability of conditions for the operation of machinery using a simple moisture probe.

Acknowledgement The research was financed by VEGA grant 1/0323/011 »Analysis of forest technics chassis impact on soil surfaces and maximum damage limits«.

6. References Arshad, M.A., Coen, G.M., 1992: Characterization of soil quality: physical and chemical criteria. American Journal of Alternative Agriculture 7(1–2): 5–12.

Anderson, H., Boddington, D., Van Rees, H., 1992: The longterm effects of sawlog-only harvesting on some soil physical and chemical properties in East Gippsland. Department of Conservation and Environment, Victoria, Australia, 1–29.

Becker, G., Hofmann, R., Gross, M., 1986: Boden- und Wurzelschäden durch Befahrung von Waldbeständen. Der Forstund Holzwirt 41(14): 367–370.

Benecke, P., 1982: Auswirkungen des Befahrens mit schweren Fahrzeugen auf bodenphysikalische Kennwerte. Mitt. d. Dt. Bodenkunkdl. Gesellsch. 34: 181–185.

Bredberg, C., Wästerlund, J., 1983: Wurzel- und Bodenschäden durch Fahrzeuge. Forstwissenschaftliches Centralblatt 102(1): 86–98.

Ebel, A., 2006: Druckverteilung auf Kontaktflächen unter Forstreifen. Dissertationsschrift an der Fakultät für Forstwissenschaften und Waldökologie der Georg August-Universität Göttingen, 1–134.

Forbrig, A., 2000: Konzeption und Anwendung eines Informationssystems über Forstmaschinen auf der Grundlage von Maschinenbuchführung, Leistungsnachweisen und technischen Daten. KWF-Bericht Nr. 29, zugl. Dissertation der TU München, 1–213.

Gebauer, R., Neruda, J., Ulrich, R., Martinková, M., 2012: Soil Compaction – Impact of Harvesters’ and Forwarders’ Passages on Plant Growth, Sustainable Forest Management – Current Research, Dr. Julio J. Diez (Ed.), ISBN: 978-953-510621-0, InTech.

213

M. Allman et al.

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205–215)

Güldner, O., 2002: Untersuchungen zu Bodenveränderungen durch die Holzernte in Sachsen und Entwicklung eines Konzepts zur ökologisch verträglichen Feinerschliessung von Waldbeständen. In: Tagungsbericht der Sektion Forsttechnik des Verbandes Deutscher Forstlicher Versuchsanstalten, Sopron, Ungarn, 1–10.

McMahon, S.D., 1995: Accuracy of two ground survey methods for assessing site disturbance. J. For. Eng. 6(2): 27–34.

Hamberger, J., 2002: GPS als Mittel zum umweltschonenden Maschineneinsatz: Navigation von Forstmaschinen und Dokumentation ihrer Fahrbewegungen. Forstliche Forschungsberichte München, Nr. 188, zugl. Dissertation der TU München, 1–192.

Rab, M.A., 1992: Impact of timber harvesting on soil disturbance and compaction with reference to residual log harvesting in East Gippsland, Victoria-a review. VSP Tech. Rep. No. 13. Native Forest Research, Department of Conservation and Environment, Victoria, Australia, 18 p.

Hildebrand, E.E., 1981: Befahrung und Bodenverdichtung unter dem Aspekt der Bodenfunktion als Waldstandort.

Mitt. d. Dt. Bodenkdl. Gesellsch. 32: 51–85.

SHMU, 2012: http://www.shmu.sk/en/?page=1610&id= (Last viewed 13. 8. 2013)

Hildebrand, E.E., Wiebel, M., 1982: Der Einfluss von Befahrung auf die Qualität von Forststandorten. AFJZ 153: 24–34.

Hirano, T., Kim, H., Tanaka, Y., 2003: Long‐term half‐hourly measurement of soil CO2 concentration and soil respiration in a temperate deciduous forest. Journal of Geophysical Research: Atmospheres (1984–2012) 108(D20): DOI: 10.1029/2003JD003766.

Hofmann, R., Becker, G., 1990: Bodenschäden im Wald durch den Einsatz von Forstmaschinen. AFZ-Der Wald 20: 478–480.

Kindernay, D., 2010: Stanovenie limitov ťažbovo – dopravnej erózie a poškodenia lesa na vybraných stanovištných podmienkach, Technická univerzita Zvolen, 1–141, LF-111162078.

Klobouček, B., Kaun, M., Luxemburk, F., Procházka, M., Stahl, J., 1979: Silniční laboratoř, SNTL/ALFA Praha, 468 p.

Löffler, H., 1983: Risiken moderner forstlicher Arbeitsverfahren für Boden und Bestand. FwCbl. 102(1): 65–67.

Löffler, H., 1986: Bodenschäden bei der Holzernte - Ursachen, Folgen, Vorbeugung. HZBL 149: 2190–2192.

Lüscher, P., Frey, B., Sciacca, S., Frutig, F., 2009: Physikalischer Bodenschutz im Wald- Datengrundlagen und Umsetzungsstrategie. In: Böden - eine endliche Ressource, September, Bonn.

Poršinsky, T., Sraka, M., Stankić, I., 2006: Comparison of two approaches to soil strength classifications. Croatian Journal of Forest Engineering 27(1):17–26.

Scheer, Ľ., 2010: Biometria, Vysokoškolská učebnica TU Zvolen, 334 p.

Schürger, J., 2012: Zhodnotenie negatívneho vplyvu traktorových technológií na pôdu v lanovkových terénoch.

Dizertačná práca, Zvolen, 1–198.

Skoupý, A., 2011: Multikriteriálni hodnocení technologií pro soustřeďování dříví, Lesnická práce, 211 p.

Stanovský, M., Schűrger, J., Jankovský, M., Allman, M., Messingerová, V., Hnilica, R., Kučera, M., 2012: Znečistenie pôdy olejmi a možnosti čistenia pôdy po znečistení olejmi, Použitie ekologicky vhodných médií v hydraulických a mazacích systémoch lesných strojov, TU Zvolen 76 p.

STN 72 1002 Klasifikácia zemín pre dopravné stavby, Úrad pre normalizáciu, metrológiu a skúšobníctvo SR 1995.

Ulrich, R., 2002: Použití harvestorové technologie v probírkach, MZLU Brno, 98 p.

Wästerlund, I., 1989: Strength components in the forest floor restricting maximum tolerable machine forces. Journal of Terramechanics 26(2): 177–182.

Wehner, T., 2002: Technikkonzepte und ihre Auswirkungen auf den Boden. Schriftenreihe Freiburger Forstliche Forschung Band 18: 97–107.

Lukáč, T., 2005: Viacoperačné stroje v lesnom hospodárstve.

Zvolen, Technická univerzita, 137 p.

Weise, G., 2002: Evaluation scheme for soil compaction by large machinery in the forest. Forsttechnische Informationen 1(2): 10–12.

Majnounian, B., Jourgholami, M., 2013: Effects of Rubber Tired Cable Skidder on Soil Compaction in Hyrcanian Forest. Croatian Journal of Forest Engineering 34(1):123–135.

Zander, J., 1988: Experimentelle Untersuchungen über den Einfluss des Befahrens auf die Oberbodenverdichtung.

FwCbl. 107(1): 112–122.

Matthies, D., Weixler, H., Hess, U., 1995: Befahrungsbedingte Strukturveränderungen von Waldböden. AFZ-Der Wald 22: 1218–1221.

Ziesak, M., 2004: Entwicklung eines Informationssystems zum bodenschonenden Forstmaschineneinsatz. Dissertation der TU München, 1–145.

214

Croat. j. for. eng. 36(2015)2

Damage Caused by Wheeled Skidders on Cambisols of Central Europe (205–215)

M. Allman et al.

Authors’ address:

Received: December 12, 2013 Accepted: October 16, 2014 Croat. j. for. eng. 36(2015)2

Michal Allman, MSc. e-mail: allman.michal@gmail.com Michal Ferenčík, PhD.* e-mail: ferencik@tuzvo.sk Martin Jankovský, MSc. e-mail: jankovskyma@gmail.com Miroslav Stanovský,MSc. e-mail: stanovsky@tuzvo.sk Prof.Valéria Messingerová, PhD. e-mail: messingerova@tuzvo.sk Department of Forest Harvesting, Logistics and Ameliorations Technical University in Zvolen, Faculty of Forestry T. G. Masaryka 24 960 53 Zvolen SLOVAKIA * Corresponding author

215

Original scientific paper

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C Skidder in Northern Iran Ahmad Solgi, Ramin Naghdi, Petros A. Tsioras, Mehrdad Nikooy Abstract Skidding by means of heavy forestry machinery can affect soil physical properties. We assessed the effects of ground based skidding on soil bulk density and total porosity under the Iranian mountainous forest conditions. Treatments included a combination of four levels of traffic intensity (1, 3, 6, and 15 passes) of a Timberjack 450C rubber skidder and two levels of slope (<20% and >20%). The bulk density was highest in samples taken in the wheel tracks and between them, and decreased towards both ends of the track (0.5 to 4 m). The results showed that bulk density increased with traffic frequency, while total porosity decreased. Average soil bulk density ranged from 0.96 g cm-3 (after one machine pass and slope <20%) up to 1.41 g cmâ&#x20AC;&#x201C;3 (after 15 machine passes and slope >20%) on the skid trail, while the respective value was 0.7 g cmâ&#x20AC;&#x201C;3 for the undisturbed area. On compacted soil, total porosity at the 0â&#x20AC;&#x201C;10 cm depth decreased by 37% compared with non-compacted soil. The results showed that slope steepness had a strong effect on the soil disturbance, with the critical value for bulk density occurring after 15 machine passes at slope <20% and six machines passes at slope >20%. The impacts of soil compaction could be evidenced in a distance of up to 2 m from the end of the skidding trail. The latter finding suggests that special interest in the form of managerial measures should be taken during the skidding operations in an effort to minimize the adverse effects of ground based skidding on the physical properties of the soil.

Keywords: bulk density, slope, soil compaction, total porosity

1. Introduction Wood extraction by means of forest machinery has replaced animal skidding in many parts of the world.

This change has resulted in higher productivity rates and efficiency of the forest operations, but at the same time, has brought up the problem of soil disturbance.

One of the most commonly witnessed forms of soil disturbance is soil compaction. Soil compaction is a typical process that may appear as a result of inappropriate use of heavy forest machinery (Ampoorter et al. 2010, Greacen and Sands 1980).

Soil compaction refers to the compression of pores, which leads to decreased porosity (Gayoso and Iroume 1991) and pore continuity (Berli et al. 2004), increased bulk density (Eliasson 2005, Greacen and Sands 1980, Solgi et al. 2013), increased soil strength Croat. j. for. eng. 36(2015)2

(Horn et al. 2004), decreased gas exchange rates between soil and atmosphere (Shestak and Busse 2005) and lower water infiltration (Dickerson 1976), which in turn leads to increased runoff. Furthermore, when air filled porosity falls below 10% of the total soil volume, microbial activity can be severely limited in most soils (Brady and Weil 2003). Soil compaction, with its multiple adverse impacts to the management and long term productivity of forested areas, is a major problem that should be properly addressed (Matangaran and Kobayashi 1999).

A large number of factors influence the extent and severity of soil compaction. These factors include site and soil properties, such as soil texture, the magnitude and nature of compressive forces, skid trail conditions, forest stand characteristics, harvesting system, and training, experience and expertise of equipment op-

217

A. Solgi et al.

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217–225)

erators (Demir et al. 2007). Traffic intensity (number of machine passes) plays an important role in soil compaction because deformations can increase with the number of passes and may lead to excessive soil disturbance. Also, slope steepness has been associated with a strong effect on soil disturbance during timber harvesting (Krag et al. 1986). Najafi et al. (2009) showed that disturbance increased in both magnitude (extent) and depth with an increasing slope.

Effective management of machine mobility, the control of site disturbance, and moderation of potential soil damage due to wood harvesting and extraction machinery traffic requires analysis of the effects of soil-machine interaction. The interaction should take into account the influence of machine variables on a range of forest terrain that may be encountered (Nugent et al. 2003).

The objective of this study was to determine the impacts of ground skidding on soil disturbance at different levels of trail slope and traffic intensity, with regard to a) soil bulk density and b) soil porosity in the wheel tracks, between them and in an area 4 m wide on each side of the skid trail. The results are discussed and threshold levels for machine traffic intensity and skid trail slope are proposed.

2. Materials and methods 2.1 Site description The study was conducted during the period August – September 2013 in compartment 41 of the third district of Nav-Asalem forest, Guilan Province, northern Iran (Fig. 1), with an area of 62 ha and a slope of 0–50%. The general location of the study site is between 37°37'and 37°61'N latitude and 48°39' and 48°44'E longitude. The area is covered by Fagus orientalis and Carpinus betulus stands. Canopy cover has been estimated to be 85%, average tree diameter was 32.53 cm, average tree height was 21.76 m and stand density has been measured to be 180 trees/ha. Elevation is approximately 1400 m above sea level with a northern aspect. The average annual rainfall in the area amounts to 1200 mm, with the lowest monthly average precipitation value of 25 mm in August and the highest of 120 mm in October. Mean annual temperature is 15 °C, with the lowest values in February.

At the time of skidding, weather conditions were wet with the soil having an average gravimetric moisture content of 28%. Soil texture along the trail was determined to be clay loam after analysis with the Bouyoucos hydrometer method (Kalra and Maynard, 1991).

The soil had not been driven on before the experiment.

218

Fig. 1 Compartment 41 of the third district of Nav-Asalem forest in northern Iran (study area) Single tree selection method was used in the study area. All logs (of various dimensions) were extracted from stump area to road side landing by a ground based skidding system. The machine used was the 4WD Timberjack 450C rubber tired skidder, weighing 10.3 ton without load (axle weight proportion 55% on the front to 45% on the rear axle).The skidder was equipped with the engine model 6BTA5.9 (engine power of 177 PS) and was fitted with tires of 24.5–32 inflated to 220 kPa.

2.2 Experimental design and data collection A skid trail, 4 m wide and 800 m long with upslope skidding direction, was selected for the experiments.

The skid trail passed through the stand in an east west direction and has been used recently. The longitudinal profile showed that the slope of the skid trail ranged from 0 to 36%. In this study, the impacts of skidding on the skid trail surface soil layer (0–10 cm depth) have been examined by a) measuring dry bulk density and total porosity, at different levels of slope and traffic intensity and b) comparing these results with the respective values in samples taken from undisturbed area. Our experimental design included two slope classes (<20%, and >20%) and four levels of traffic intensity (1, 3, 6, and 15 passes). Each treatment was replicated three times.

Twenty four plots, 10 m long and 12 m wide, were delineated randomly prior to skidding. Each plot was divided in three parallel transects with similar dimenCroat. j. for. eng. 36(2015)2

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217–225)

A. Solgi et al.

2.65 particle density, g cm–3; VC volume of the soil cores, 196.25 cm3; Soil bulk density was calculated as Eq. (2) (Tan et al. 2005): Db =

Wd VC

(1)

Where: Wd weight of the dry soil, g.

2.3 Statistical analysis Fig. 2 Study layout sions (10 m long and 4 m wide), with the central transect characterized as the skid trail area. Buffer zones of at least 5 m length were created between plots in order to avoid interactions. In each plot, samples were taken along four randomized lines perpendicular to the direction of travel, leaving 2 m buffer zone between them to avoid interactions. More specifically, soil samples from the 0–10 cm depth were collected at three different points of each line, one on the left track (LT), one between tracks (BT) and one on the right track (RT). Furthermore, 8 soil sample points were taken on each side of the lines at 0.5 m intervals, extending from the skid trail area to undisturbed soil.

For control purposes, soil samples were taken from the undisturbed area, where there was no skidding impact, at least 50–60 m away from the skid trail (at least two – tree lengths away to reduce possible impacts) (Fig. 2).

The soil samples were collected with the help of a soil hammer and cylinders (diameter 5 cm, length 10 cm). All samples were put in polyethylene bags and labelled. Then the samples were brought to the laboratory, where they were promptly weighed. The samples were oven dried at the temperature of 105 °C for 24 h and weighed for a second time. The formulas below were used to determine bulk density and total soil porosity: Total porosity was calculated as Eq. (1) (Fernández et al. 2002, Ezzati et al. 2012): Db 2.65 AP = × 100 VC 1

Where: AP total apparent porosity, %; Db soil bulk density, g cm–3; Croat. j. for. eng. 36(2015)2

The data were analyzed using two-way ANOVAs in the SPSS 11.5 software. Mean values of physical soil properties at each plot were compared to those in undisturbed (untrafficked) areas using Duncan’s multiple range test (Zar 1999). One-way ANOVA (significance test criterion P≤0.05) was used to compare the physical soil properties in four traffic intensities (main effects) with those in undisturbed areas. Paired t-tests were used to analyze soil properties data in two slope gradients at an alpha level of 0.05.

3. Results 3.1 Soil bulk density Soil bulk density was measured as 0.7 g cm–3 in the general harvesting area (undisturbed area). On the skid trail, soil bulk density increased both with traffic frequency (p<0.05) and skid trail slope (p<0.05), but the interaction effects of traffic frequency × skid trail slope were found not significant (Table 1). Compared to the undisturbed soil condition, average soil bulk density at the 0–10 cm depth increased by 77% in compacted plots.

Table 1 Analysis of variance (P values) of the effects of number of passes and slope class on dry bulk density and total porosit Source of Variance Dry Bulk Density

(1) Total Soil Porosity

d.f. P value*

Number of passes

3

0.001

Slope class

1

0.029

Number of passes × slope class

3

0.137

Number of passes

3

0.001

Slope class

1

0.014

Number of passes × slope class

3

0.382

*P values less than 0.05 are given in bold

219

A. Solgi et al.

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217–225)

Fig. 3 Soil bulk density under the main track (LT and RT), log track (BT) and at various distances from the track (a: slope <20%; b: slope >20%) Table 2 Effect of slope and traffic intensity on bulk density (g cm–3) Slope <20% >20%

Cycles 0*

1 a

3 a

0.69

0.96

a

1.09

b

0.71

6 a

1.39a

b

1.41a

1.23

b

1.13

15 a

1.25

1.40

*0 passes denotes undisturbed soil

Table 3 Dry bulk density increase (%) per slope class compared to the previous traffic intensity level Slope class

Traffic intensity 1

3

6

15

<20%

37.14

13.54

12.84

13.00

>20%

61.42

10.62

11.68

0.71

Average soil bulk density values ranged from 0.96 g cm–3 (one pass and slope of <20%) to 1.41g cm–3 (15 passes and slope of >20%) on the skid trail compared to 0.7 g cm–3 in the undisturbed area. The results showed that the dry bulk density was affected by skid trail slope (Table 2).

Average dry bulk density of the samples with one pass and slope of >20% (1.13 g cm–3) was higher than of those after 3 passes and slope of <20% (1.09 g cm–3).

Within all traffic frequencies, dry bulk density increased considerably faster with an increase in slope from <20% to >20% (Table 2).

220

The results showed that most of the potential impact occurred after the initial passes. As it can be noticed in Table 3, substantial increases in bulk density (61.42%) for skid trails at steep terrain appear right after the first skidder pass.

The degree of bulk density of the soil just under the main track (right and left track) or log track (between tracks) differed from those at various distances from the track. In each plot, the highest bulk density was observed in the sample points along the main track.

As we move away from the main track, bulk density decreased (Fig. 3).

Bulk density increased at different rates at the sampling points »outside« the skid trail area; however, the amount of increase varied between the two slope classes and the distance from the skid trail (Fig. 4). Also, as traffic intensity increased, the difference between bulk density values on the skid trail, as well as between the skid trail and the adjacent sampling points decreased.

This finding suggests that the higher the skid trail usage, the more area adjacent to it is compacted.

3.2 Total porosity Total porosity changes were influenced significantly by the number of skidder passes (p<0.05) and the slope (p<0.05); however, the interaction between numbers of skidder passes and slope was not significant (p>0.05) (Table 1). Total porosity on the skid trail was considerably lower than the total porosity in the undisturbed area (Table 4).

Croat. j. for. eng. 36(2015)2

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217â&#x20AC;&#x201C;225)

A. Solgi et al.

Fig. 4 Effect of different numbers of skidder passes on soil bulk density under the main track (LT and RT), log track (BT) and at various distances from the end of the track (a: one pass; b: three passes; c: six passes; d: fifteen passes) Table 4 Effect of skidder passes on total porosity (%) Slope

Number of passes 0*

1 a

3 b

6

15

<20%

74.62

62.82

56.82

52.67

49.18e

>20%

74.32a

56.93b

53.12c

49.11d

46.67e

*0 passes denote undisturbed soil

Croat. j. for. eng. 36(2015)2

c

d

On compacted soil, total porosity at the 0â&#x20AC;&#x201C;10 cm depth decreased by 37% compared to non-compacted soil. The average total porosity in the undisturbed area is 74% and on the skid trail 53%. The total porosity was significantly lower on the plots with a slope of >20% than those of <20% in all traffic treatments.

In each treatment, as we move away from the main track, the total porosity increased. The lowest total po-

221

A. Solgi et al.

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217â&#x20AC;&#x201C;225)

Fig. 5 Soil total porosity under the main track (LT and RT), log track (BT) and at various distances from the track rosity value was observed in the soil just under the main track, while its highest value was measured in the undisturbed area (Fig. 5).

4. Discussion Soil compaction was affected by trail slope, with dry bulk density increasing faster at higher slope levels. This may be a consequence of the difficulties that the skidder found when logging in steep terrains. Under these conditions, the machine slipped continuously and remained in a given place for a longer period of time, puddling and dragging the soil (Gayoso and Iroume 1991). When a skidder passes slower on a steep slope, the top soil is obviously vibrated more and consequently gets more disturbances compared to flat terrain. In the case of uphill skidding, higher soil compaction can be explained by the higher load of the skidder rear axle (Najafi et al. 2009). The increase of bulk density and decrease of total porosity in the steep slope trail and at various distances from it may be associated with the lower speed of skidder on steeper slope.

Many previous studies indicated that most of the potential impact occurred after the initial passes (Lacey and Ryan 2000, Startsev and McNabb 2000). As it can be noticed in Table 3, strong increases in bulk density (61.42%) for skid trails already appear right after the first skidder pass at a slope higher than 20%. Such

222

high bulk density changes for this traffic intensity are not uncommon, and they have been reported by Williamson and Neilsen (2000). At lower inclination than 20%, three skidding cycles increased bulk density by 55.7%, which is in line with Ampoorter et al. (2007), who found that bulk density increases more gradually with 50% of the total impact occurring after 3 passes. Our results contradict the findings of other studies, which report that significant changes in soil compaction can occur after six machine passes during skidding operations (Froehlich 1978, Jamshidi et al.

2008). However, such differences may be attributed to a number of reasons, including different soil attributes, different machine types and dimensions, among others (Bustos and Egan 2011, Rab et al. 2005, Williamson and Neilsen 2000).

At lower slope levels, dry bulk density reached the critical value after 15 passes, compared to only 6 passes needed at slope of >20%. There was no significant difference between treatments of 6 passes and slope of >20% and those of 15 passes and slope of <20% from a bulk density point of view. Given that reaching the critical value has an adverse effect on site productivity, a number of managerial measures could allow for increased traffic intensity at reduced soil compaction. In this context, changes in skidding equipment should be considered, when they are not constrained by high purchase costs, increased operational costs, type of silvicultural prescription and potential residual stand damage (Bustos and Egan 2011, Bustos et al. 2010) There were significant (pâ&#x2030;¤0.01) differences in total porosity between treatments with the slope of <20% and those with more than 20%, which is in line with Solgi and Najafi (2014). The major differences in pore volume between the different slope treatments could probably be explained by more soil compaction on the steep trail, indicating the effect of slope on soil porosity during ground skidding. No significant differences were found in porosity between the undisturbed area and various points of the track when skidder passed once, but there was a difference when traffic increased to three passes (at slope of >20%) and more than six passes (at slope of <20%).

According to our findings, soil compaction decreased as the distance from skid trail increased. These results agree with those quoted by Matangaran and Kobayashi (1999). The extent of change in mean bulk density was affected by traffic intensity and slope and was found to be statistically significant for samples taken from points up to 2 m from the end of the trail.

Under realistic conditions, where skidder operators do not necessarily drive on the skid track and the number of passes can exceed 15, it is easy to conclude that Croat. j. for. eng. 36(2015)2

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217–225)

soil compaction extends to the adjacent area of the skid trail. For this reason, it is necessary to design skidding roads in a way that minimizes logging impact (Matangaran and Kobayashi 1999), which can be accompanied by reduced skidding cost and better spatial allocation of the respective equipment (Najafi et al. 2008).

Special attention should be given to the weather conditions suggesting that machine traffic should be reduced to dry and favourable weather conditions (Hittenbeck 2013). In the case of existing skid trails, restricting skidding traffic to a smaller portion of the harvest site (Zenner et al. 2007) designated as permanent skidding tracks, might be an alternative, when applicable. Special interest should be given to the human factor. Tiltable seats and cabins can reduce the operator stress at higher inclinations (Hittenbeck 2013), with a possible positive effect on work safety, as well as on a more considerate driving style of the machine operator under adverse conditions. Finally, the forest crews should be aware of the adverse impacts of soil compaction in forest stands and participate in training seminars, in an effort to minimize the environmental impacts (Tsioras 2010).

5. Conclusion This study was conducted with the overall objective to assess the impacts of ground skidding on soil disturbance at different levels of trail slope and traffic intensity. Dry bulk density increase by 61.42% was measured after the first skidder pass at slope >20%, while it took more than three passes at slope <20% to reach this value. The critical value for bulk density was also achieved after 6 machine passes for slope >20% compared to 15 needed at slope <20%. Bulk density increase extended for the respective slope and traffic intensity up to an area of 2 m from the end of each wheel track. Total porosity decrease ranged from 15.5% for the low traffic (1 pass and slope of <20%) up to 37.27% for the severe (15 passes and slope of >20%) treatments.

The impacts of skidding operations on soil disturbance can be ameliorated with a number of managerial and technical measures. Changes in equipment and better forest road design, which takes into consideration the, above mentioned, soil responses to compaction and restriction of traffic in heavily trafficked parts of the forest compartments, can mitigate the effects of ground based skidding operations. However, all these changes will have suboptimal results, unless special interest is given to the forest operation crews.

For this reason, training seminars on the proper use of forest machinery should be introduced, which will Croat. j. for. eng. 36(2015)2

A. Solgi et al.

benefit the forest enterprises not only in reducing logging impact but also in various other levels, such as improved operational efficiency, reduced operational costs and higher safety during forest work.

Acknowledgements This paper has been written based on the results of a research entitled »Assessing the effects of various systems of wood extraction from forests on soil physical, chemical and micromorphological properties and regeneration in northern forest of Iran«, which was sponsored by Iran National Science Foundation – Depute of Science and Technology – Presidential Office.

The authors are grateful to the Iran National Science Foundation.

6. References Ampoorter, E., Goris, R., Cornelis, W.M., Verheyen, K., 2007: Impact of mechanized logging on compaction status of sandy forest soils. Forest Ecology and Management 241(1–3): 162–174.

Ampoorter, E., Van Nevel, L., De Vos, B., Hermy, M., Verheyen, K., 2010: Assessing the effects of initial soil characteristics, machine mass and traffic intensity on forest soil compaction. Forest Ecology and Management 260(10): 1664–1676.

Berli, M., Kulli, B., Attinger, W., Keller, M., Leuenberger, J., Flühler, H., Springman, S.M., Schulin, R., 2004: Compaction of agricultural and forest subsoils by tracked heavy construction machinery. Soil and Tillage Research 75(1): 37–52.

Brady, N.C., Weil, R.R., 2003: Elements of Nature and Properites of Soils, 2nd ed. Prentice Hall, 1031 p.

Bustos, O., Egan, A., 2011: A comparison of soil compaction associated with four ground-based harvesting systems.

Northern Journal of Applied Forestry 28(4): 194–198.

Bustos, O., Egan, A., Hedstrom, W., 2010: A comparison of residual stand damage along yarding trails in a group selection harvest using four different yarding methods. Northern Journal of Applied Forestry 27(2): 56–61.

Demir, M., Makineci, E., Yilmaz, E., 2007: Investigation of timber harvesting impacts on herbaceous cover, forest floor and surface soil properties on skid road in an oak (Quercus petrea L.) stand. Building and Environment 42(3): 1194–1199.

Dickerson, B.P., 1976: Soil compaction after tree-length skidding in northern Mississippi. Soil Science Society of American Journal 40(6): 965–966.

Eliasson, L., 2005: Effects of forwarder tyre pressure on rut formation and soil compaction. Silva Fennica 39(4): 549–557.

Ezzati, S., Najafi, A., Rab, M.A., Zenner, E.K., 2012: Recovery of soil bulk density, porosity and rutting from ground skidding over a 20-year period after timber harvesting in Iran.

Silva Fennica 46(4): 521–538.

223

A. Solgi et al.

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217–225)

Fernández, R.A., Mac Donagh, P.M., Lupi, A.M., Rodolfo, M., Pablo, C., 2002: Relations between soil compaction and plantation growth of a 8 years old Lobolly pine second rotation, in Misiones, Argentine. Written for presentation at the 2002 ASAE annual international meeting/CIGR XVth world congress sponsored by ASAE and CIGR Hyatt Regency Chicago Chicago, Illinois, USA. Paper Number: 025012.

Froehlich, H.A., 1978: Soil compaction from low groundpressure, torsion-suspension logging vehicles on three forest soils. Forest Research Laboratory, Oregon State University, Corvallis, p. 12.

Gayoso, J., Iroume, A., 1991: Compaction and soil disturbances from logging in southern Chile. Annales des Sciences Forestieres 48(1): 63–71.

Greacen, E.L., Sands, R., 1980: Compaction of forest soils. A review. Australian Journal of Soil Research 18(2): 163–189.

Hittenbeck, J., 2013: Estimation of trafficable grades from traction performance of a forwarder. Croatian Journal of Forest Engineering 34(1): 71–81.

Horn, R., Vossbrink, J., Becker, S., 2004. Modern forestry vehicles and their impacts on soil physical properties. Soil and Tillage Research 79(2): 207–219.

Jamshidi, R., Jaeger, D., Raafatnia, N., Tabari, M., 2008: Influence of two ground-based skidding systems on soil compaction under different slope and gradient conditions. International Journal of Forest Engineering 19(1): 9–16.

Krag, R., Higginbotham, K., Rothwell, R., 1986: Logging and soil disturbance in southeast British Columbia. Canadian Journal of Forest Research 16(6): 1345–1354.

Lacey, S.T., Ryan, P.J., 2000: Cumulative management impacts on soil physical properties and early growth of Pinus radiata. Forest Ecology and Management 138(1–3): 321–333.

Matangaran, J.R., Kobayashi, H., 1999: The effect of tractor logging on forest soil compaction and growth of Shorea selanica seedlings in Indonesia. Journal of Forest Research 4(1): 13–15.

Najafi, A., Sobhani, H., Saeed, A., Makhdom, M., Mohajer, M.M., 2008: Planning and assessment of alternative forest road and skidding networks. Croatian Journal of Forest Engineering 29(1): 63–73.

Najafi, A., Solgi, A., Sadeghi, S.H., 2009: Soil disturbance following four wheel rubber skidder logging on the steep trail

224

in the north mountainous forest of Iran. Soil and Tillage Research 103(1): 165–169.

Nugent, C., Kanali, C., Owende, P.M.O., Nieuwenhuis, M., Ward, S., 2003: Characteristic site disturbance due to harvesting and extraction machinery traffic on sensitive forest sites with peat soils. Forest Ecology and Management 180(1–3): 85–98.

Rab, A., Bradshaw, J., Campell, R., Murphy, S., 2005: Review of factors affecting disturbance, compaction and trafficability of soils with particular reference to timber harvesting in the forests of south west western Australia. Sustainable Forest Management Series, 145 p.

Shestak, C.J., Busse, M.D., 2005: Compaction alters physical but not biological indices of soil health. Soil Science Society of American Journal 69(1): 236–246.

Solgi, A., Najafi, A., 2014: The impacts of ground based logging equipment on forest soil. Journal of Forest Science 60(1): 28–34.

Solgi, A., Najafi, A., Sam Daliri, H., 2013: Assessment of crawler tractor effects on soil surface properties. Caspian Journal of Environmental Sciences 11(2): 185–193.

Startsev, A.D., McNabb, D.H., 2000: Effects of skidding on forest soil infiltration in west central Alberta. Canadian Journal of Soil Science 80(4): 617–624.

Tan, X., Scott, X.C., Kabzems, R., 2005: Effects of soil compaction and forest floor removal on soil microbial properties and N transformations in a boreal forest long term soil productivity study. Forest Ecology and Management 217(2–3): 158–170.

Tsioras, P.A., 2010: Perspectives of the forest workers in Greece. iForest 3: 118–123.

Williamson, J.R., Neilsen, W.A., 2000: The influence of forest site on rate and extent of soil compaction and profile disturbance of skid trails during ground-based harvesting. Canadian Journal of Forest Research 30(8): 1196–1205.

Zar, J.H., 1999: Biostatistical analysis, 4th edition, Prentice Hall, Upper Saddle River, NJ, USA. 662 p. plus appendices.

Zenner, E.K., Fauskee, J.T., Berger, A.L., Puettmann, K.J., 2007: Impacts of skidding traffic intensity on soil disturbance, soil recovery, and aspen regeneration in North Central Minnesota. Northern Journal of Applied Forestry 24(3): 177–183.

Croat. j. for. eng. 36(2015)2

Soil Compaction and Porosity Changes Caused During the Operation of Timberjack 450C ... (217â&#x20AC;&#x201C;225)

A. Solgi et al.

Authorsâ&#x20AC;&#x2122; address: Ahmad Solgi, MsC.* e-mail: solgi_ahmad231@yahoo.com Assoc. prof. Ramin Naghdi, PhD. e-mail: rnaghdi@guilan.ac.ir Assist. Prof. Mehrdad Nikooy, PhD. e-mail: nikooy@guilan.ac.ir University of Guilan Faculty of Natural Resources Department of Forestry 1144 Somehsara, Guilan Province IRAN

Received: June 04, 2014 Accepted: March 12, 2015 Croat. j. for. eng. 36(2015)2

Petros A. Tsioras, PhD. E-mail: ptsioras@for.auth.gr Aristotle University of Thessaloniki Department of Harvesting and Technology of Forest Products Lab of Forest Utilization 54124 Thessaloniki GREECE * Corresponding author

225

Original scientific paper

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads Eric R. Labelle, Dirk Jaeger, Benjamin J. Poltorak Abstract In cut-to-length mechanized forest harvest operations, trees are cut, delimbed, and bucked to standard lengths directly in the harvest block. This in-stand processing, generates harvesting residue composed of tree limbs, tops, and foliage, which is frequently placed on machine operating trails to prolong trail trafficability and protect forest soils against heavy loadings. These so-called brush mats vary both in quantity and quality based on harvested wood and stand characteristics. The objectives of this study were to determine, quantify, and compare the load distributing capabilities of hardwood and softwood brush mats of different amounts (10, 20, 30, and 40 kg m-2) compared to no brush (0 kg m-2). This was done by laboratory tests analyzing the difference in strain recorded below brush mats at small scale when exposed to single and repetitive loadings. Brush mats (approx. 37 cm x 37 cm in area) were placed inside a test structure including a top open box with the bottom filled with a 15 cm thick layer of sand, below which strain gauges were installed. The entire test structure was positioned on a load frame programmed to lower a loading disk directly over the brush mat, thereby applying increasing loads up to 10 kN on the mat. Results suggest that for specific brush amounts and loadings, softwood brush showed a slightly better capacity to laterally distribute exerted loads than hardwood brush, especially at brush amounts of 10 and 20 kg m-2. At higher brush amounts, the differences of recorded loadings (strains) between the tested softwood and hardwood brush were reduced and at 40 kg m-2 hardwood brush contributed to a lower response of the strain gauges than softwood brush when subjected to 5 and 10 kN loadings.

Keywords: brush mat, brush compressibility, strain, load distribution, soil protection, forest operations

1. Introduction During mechanized forest operations, forest machinery is operated off road directly on the ground, thus exerting high surface contact pressures on the unbound surface of forest stands. These types of operations may have adverse effects on forest soil productivity and stand vitality and growth. As an alternative, operating heavy machines on a layer of harvesting residue (tree limbs, tops, and foliage) created during the processing phase and placed on the machine operating trail as a brush mat can reduce these negative impacts and attenuate high surface contact pressures by laterally distributing machine loads over a greater area (Labelle and Croat. j. for. eng. 36(2015)2

Jaeger 2012). In particular, machine operating trails covered with a brush layer showed reduced rutting and less pronounced decreases in porosity and hydraulic conductivity (Jakobsen and Moore 1981, McMahon and Evanson 1994, McDonald and Seixas 1997, Han et al.

2006, Eliasson and W채sterlund 2007, Gerasimov and Katarov 2010, Poltorak 2011). Unlike machine related parameters used to minimize soil disturbance (reduced payload, flotation tires, steel flexible tracks, etc.), brush is most often available on harvesting sites and its use does not negatively affect machine productivity. However, since branches used to create brush mats are obtained from processed trees, the quantity and quality of material available as a protective layer is dependent on

227

E. R. Labelle et al.

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227â&#x20AC;&#x201C;242)

the particular forest stand as well as on the type and intensity of silvicultural treatment applied and the characteristics of harvested wood.

With an increase in partial harvest frequency, mainly through commercial thinning and shelterwood silvicultural treatments, the amount of brush available from harvested trees for soil protection is much less than in clearfelling operations. Since smaller sized harvesting machinery is commonly used in thinning operations, one could assume that thinner brush mats of the fewer available brush would be sufficient for soil protection. However, despite employing lighter and narrower harvesters during partial harvests, forwarders, used to transport wood from the felling site to roadside landings, are usually the same as those used in clearfelling operations. Such forwarders, with loaded mass of 15 to 40 metric tons exert high static ground pressures (70 to 180 kPa, Kozlowski 1999) and often require a thick brush mat to effectively reduce peak loads. Therefore, it becomes important to predominantly apply available brush to those sections of the operating trails, which are most susceptible to soil disturbance by machine traffic (i.e. terrain depressions with high soil moisture, high traffic areas, machine operating trails with acute intersection angles, etc.).

While brush quantity proved to be a significant factor for determining the capability of brush mats to effectively reduce peak loadings by laterally distributing applied loads (Labelle and Jaeger 2012), the impact of tree species on load distributing pattern of brush mats of identical quantity needs more investigation. Branches of different species (e.g., hardwood and softwood) have diverse physical and, in particular, mechanical properties, which might impact their ability to reduce surface contact pressures once they are used to form a brush mat. To determine and quantify the ability of hardwood and softwood brush mats in distributing applied loads, the following research objectives were studied.

Analyze the lateral load transfer capability of hardwood and softwood small scale brush mats of 20 kg m-2 exposed to repetitive loadings.

In this study, strain below brush mats is defined as the deformation of steel channels (below which strain gauges were installed) relative to a reference length.

These steel channels were located below a 15 cm layer of sand above which different brush mat amounts of different species were subjected to vertical loadings.

1.1 Research objectives

Hardwood and softwood brush mats of varying quantities (10, 20, 30, and 40 kg m-2) were each replicated once for a total of two test series per brush amount. Once placed inside the test structure (additional details in section 2.3), the mats were exposed to increasing loadings up to a maximum of 10 kN. The vertically downward transferred loadings were recorded below the mats by strain gauges installed on steel channels at the bottom of the test structure and covered with a layer of sand upon which the varying brush mats were positioned. The purpose of adding a sand layer inside the test structure was to allow for a consistent flexible medium, below which load distribution capabilities of various small scale brush mats

Determine and quantify the capability of hardwood and softwood small scale brush mats of varying quantities to reduce vertically downward transferred loadings when exposed to single loadings compared to tests without applying brush mats.

Analyze the lateral load transfer capability of various hardwood and softwood small scale brush mats (similar conditions as outlined in number i).

Quantify the capability of hardwood and softwood small scale brush mats of 20 kg m-2 to reduce vertically downward transferred loadings compared to applying no brush when exposed to repetitive loadings.

228

2. Methodology 2.1 Brush mat characteristics To determine the relative competence of brush in distributing applied punctual loadings when placed over a layer of sand, we constructed hardwood and softwood small scale brush mats from green (fresh) branches collected from living trees during a timber harvesting operation within a natural forest stand in Fredericton, New Brunswick, Canada in summer (August). Hardwood mats were composed of yellow birch (Betula alleghaniensis Britton) branches, while softwood brush mats were formed from balsam fir (Abies balsamea (L.) Mill.). Yellow birch and balsam fir were chosen because of their wide natural distribution range and high frequency in forest stands throughout eastern Canada. Even though branch diameter was not individually tallied during this study, hardwood and softwood branches were limited to a diameter of 3 cm at the large end.

To determine brush water content, sub-samples from branches used for each brush mat were oven dried at 105 Ë&#x161;C until constant mass was achieved. Quantifying branch water content was of interest due to potential implications on brush compressibility and associated ability of laterally distributing applied loads.

2.2 General description of test scenarios

Croat. j. for. eng. 36(2015)2

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

E. R. Labelle et al.

Fig. 1 A) Load frame with custom built load test structure, B) Three strain gauges installed below steel channels (channels were turned upside down to show strain gauges), C) Horizontal crosshead of load frame with load cell and steel circular loading disk resting on a softwood brush mat could be equally compared amongst each other and with the no brush test scenario, where the loading was applied directly to the sand layer without any brush cover. The loading was generally performed in two steps: first, a consolidation loading of 10 kN was applied to allow for settling of branches for improved mat performance. After the load was completely released, the main-test loading followed exposing the mats again to a maximum load of 10 kN. This combination of consolidation and main-test loading constituted one test series. In total, two test series T1 and T2 were performed for all brush mat quantities. During the main-test loading of the mats, the vertically transferred loadings below the mats and the sand layer were recorded to assess the mats’ abilities to reduce peak loadings by transferring loadings into side areas.

Besides testing brush mats, we also recorded the response of strain gauges when no brush mat was applied and the loads were thus exerted directly to the sand. As for the brush tests, in total, two test series were also performed directly over sand, each of which constituted one consolidation and main-test loading.

To avoid any consolidation of the sand due to multiple loadings during the different brush test scenarios, the sand was loosened with a spade, removed from the load box after each test, and refilled with special attention given to re-using the same mass of sand, thereby assuring a relatively constant density.

2.3 Compression and strain instrumentation To reduce variability associated with in-field testing, we assessed the ability of hardwood and softwood brush mats of different brush amounts in transferring applied loadings in a more controlled laboratory test Croat. j. for. eng. 36(2015)2

environment. To allow testing of different brush amounts, a test structure composed of a top open load box at its base (inside dimensions 36.7 cm long, 36.8 cm wide and 20.0 cm high) and a two sided brush support frame (90 cm high) was designed and constructed from structural lumber (Fig. 1 A). This test structure was designed with three main intentions: Þ confine branches to a defined area during testing, Þ allow a 15 cm thick layer of sand to be placed and contained in the load box of the test structure, Þ be small enough to fit inside the load frame of a universal testing machine.

The load box section of the test structure was filled with 15 cm of sand prior to any test. At the bottom of this load box, located below the sand, received loadings were measured by three (350 ohms) general purpose strain gauges installed in the middle of the downward side of three separately set up and independently operating steel channels (36.6 cm long x 2.5 cm wide x 0.6 cm thick; Fig. 1 B and Fig. 2). One steel channel and corresponding strain gauge were positioned in the middle of the test structure directly below a loading disk, and the other two steel channels and associated strain gauges were offset 16.2 cm adjacent to each side of the centre gauge to detect any lateral load distribution in relation to the middle gauge. All three strain gauges were connected in a three wire (excitation, ground, and nominal gauge resistance 350 ohms) quarter bridge circuit to a strain indicator and recorder. This three wire connection type offered the benefits of intrinsic bridge balance, automatic compensation for the lead wire temperature change on bridge balance, and increased mea-

229

E. R. Labelle et al.

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227â&#x20AC;&#x201C;242)

Fig. 2 Top and side view schematic of base plate, steel channels, and strain gauges installed at the base of the load box

surement sensitivity compared to a two wire configuration (Micro Measurements 2011). To assure unbiased measurements, all three strain gauges were independently subjected to an identical vertical load

230

and their responses in micro strains were monitored.

The strain indicator and recorder was set to record data at a rate of one reading per second for each of the three channels used.

Croat. j. for. eng. 36(2015)2

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227â&#x20AC;&#x201C;242)

To assess the load distributing capability of different brush mats, varying brush quantities of hardwood and softwood were placed in the test structure on top of the sand filled load box. To compress brush, a 30 kN load capacity testing load frame was used. The load frame consisted of two vertical side columns allowing a horizontal crosshead to travel up to 1122 mm on the vertical axis (InstronÂŽ 2012). The load frame measured applied loads with a 30 kN capacity compression load cell directly mounted to the horizontal crosshead. Fastened below this load cell was a steel circular loading disk measuring 15.4 cm in diameter (area of 186.3 cm2) and 1.9 cm in thickness used to exert loads from the crosshead to the brush mat (Fig. 1 C). During the loading process, the load frame software recorded displacement of the crosshead (mm) and associated load (kN) measured by the load cell.

Using the load frame, loads of up to 10 kN (equivalent to surface contact pressure of 537 kPa below loading disk at even load distribution) were exerted on top of brush mats placed inside the test structure with a constant 50 mm min-1 downward movement of the horizontal crosshead. We compared strain gauge responses when loading hardwood and softwood brush mats of varying quantities to the strain responses recorded when exerting the loadings directly over bare sand in the load box (without any brush cover).

2.4 Testing procedure Before testing, the load box section of the load test structure was evenly filled with 15 cm (33 kg) of air dry sand. Due to size limitations within the load test structure and to reduce sidewall friction in the load box, all branches (hardwood and softwood) had to be Table 1 Brush water content by composition and amount for each of the two test series Composition

Brush amount, -2

Brush water content, % green mass

kg m

Test series 1

Test series 2

Hardwood

10

52.4

54.4

Hardwood

20

51.2

52.6

Hardwood

30

55.0

52.2

Hardwood

40

55.2

51.1

Softwood

10

49.1

47.4

Softwood

20

50.3

48.3

Softwood

30

50.1

47.8

Softwood

40

50.6

52.2

Croat. j. for. eng. 36(2015)2

E. R. Labelle et al.

trimmed to a 35 cm length before testing. Once a specific brush mat target amount was reached by using a scale (10, 20, 30, or 40 kg m-2 corresponding to 1.35, 2.70, 4.05, and 5.41 kg of brush per test series, respectively), a brush mat was created within the load test structure by placing the branches over the sand perpendicular to the orientation of the strain gauges and measuring its initial loose (no load) thickness. We then zeroed the strain gauges and the load frame (load cell and positioning of horizontal crosshead) and programmed the software to lower the horizontal crosshead at a constant speed of 50 mm min-1.

First, the consolidation loading was performed to allow the brush mat to settle. With the movement of the crosshead both the load frame and strain recording systems started recording data (one reading per second) at the exact same time. The consolidation loading was paused when a load of 0.09 kN was reached (pre-loading state) to allow for standardized measuring of the thickness of the brush mat (summary of test procedure; Fig. 3). The load of 0.09 kN (surface contact pressure of 4.8 kPa below loading disk) on top of the brush ensured similar brush mat thickness measurement conditions as those applied by Labelle and Jaeger (2012) during laboratory testing of forwarder traffic over different brush mat amounts. By assuring similar measurement conditions, we could compare mat thicknesses of these lab scale tests to thicknesses of full scale brush mats. The thickness of the brush mat was measured as the vertical distance between the downward side of the loading disk and the top of the sand layer. Afterwards, the consolidation loading was resumed by programming the load testing machine to further increase the loading by downward movement of the crosshead and loading disk at a constant velocity until a terminal load of 10 kN was exerted to the brush mat covering the sand layer.

Once this load was reached, we re-measured the thickness of the brush mat and the load was released by upward movement of the horizontal crosshead until the loading disk was out of contact with the brush layer, thus enabling it to rebound freely.

After completion of the consolidation loading, the main-test loading was applied to determine brush mat compressibility and corresponding strain gauge responses. In this context, we define brush mat compressibility as the length of crosshead and loading disk vertical travel from the pre-load stage until the target load was reached. The main-test loading event was identical to the consolidation loading event, as it included a pre-loading phase until a load of 0.09 kN was reached to allow again for standardized determination of brush layer thickness and then application of the full load (10 kN) with measurements of corresponding brush layer thickness.

231

E. R. Labelle et al.

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227â&#x20AC;&#x201C;242)

Fig. 3 General testing sequence (example of 10 kg m-2 brush mat) performed at the load test structure. The steps listed were also performed for 20, 30, and 40 kg m-2 brush amounts for hardwood and softwood brush mats. In addition, 20 kg m-2 brush mats were subjected to a third, fourth and fifth loading each consisting of the same applied loads as described above

Lastly, brush mats containing 20 kg m-2 of hardwood or softwood brush were not only exposed to the consolidation and main-test loadings but to additional three loadings to determine the load diverting/distributing behaviour of brush mats under repetitive loadings.

Once testing of a specific brush mat was completed (consolidation test and main-test), all brush was removed from the test structure and discarded. Sand

232

was also removed from the test box and for the next test series the load box was refilled with the same amount of sand and fresh, uncompressed branches were used for subsequent tests.

In addition to testing hardwood and softwood brush mats, we also performed two separate test series (both including consolidation and main-test loadings as for any brush mat) directly on top of the 15 cm sand layer to obtain strain responses under a no brush scenario.

Croat. j. for. eng. 36(2015)2

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227â&#x20AC;&#x201C;242)

2.5 Data analysis To compare the relative competence of hardwood and softwood brush mats in distributing applied loads, we determined strain gauge responses at three target loads (1, 5, and 10 kN which correspond to surface contact pressures of 54, 268, and 537 kPa below the loading disk). However, as each loading ranging from 0.09 to 10 kN was controlled by the rate of the crosshead vertical displacement (50 mm min-1) rather than by an actual rate of loading (e.g. 1 kN min-1), exact strain responses at the three target loads had to be obtained by linear interpolation between the two nearest strain responses (one below and one above the target load).

According to the research objectives of this study, we analyzed the recorded data in different respects: To assess the capability of a mat to reduce vertically downward transferred loads, the strain responses of the middle gauge was compared for hardwood and softwood brush mats of varying quantities at the three target loads and also for the no brush scenario.

To assess the capability of a mat to transfer loads laterally, we averaged the responses of both side gauges per test series (in total four strain responses of the two side gauges during two test series for each testing scenario) at the target loads mentioned above, and compared them to the average response recorded below the middle gauge for both test series.

Both analysis steps were applied to single loadings and repetitive loadings.

2.5.1 Statistical analyses Statistical analyses were performed with the Minitab 17 statistical package. To assess the competence of brush mats at distributing applied loads below the mats, analysis of variances were made and differences between means were tested post hoc using Tukey pairwise comparisons. Response variables used were strain on middle gauge and strain on side gauges while using the amount of brush as the term for comparisons. A significance level of 5% was used throughout all statistical analyses.

3. Results 3.1 Brush mat characteristics Branches collected in late August 2010 from the same stand and tested within a two day period had a water content (percent of green mass) varying between 49.1 and 50.6% for softwood test series 1 and between 47.4 and 52.2% for softwood test series 2 (Table 1).

Hardwood brush mats had generally slightly higher Croat. j. for. eng. 36(2015)2

E. R. Labelle et al.

water content ranging from 51.2 to 55.2% during test series 1 and between 51.1 and 54.4% during test series 2. When combining all brush mats from the same species, softwood branches had lower average water content by 3.5% compared to hardwood branches.

Aside from determining brush water content, we also measured brush mat thickness during different loading stages to determine if the various quantities of hardwood and softwood brush showed similar responses with respect to compression when exposed to identical loading. When combining results from both test series during the no-load stage (0 kN load), 30 and 40 kg m-2 brush mats of hardwood composition were each slightly thicker than softwood brush mats of the same amounts (Fig. 4 A). On average, hardwood brush mat thickness increased from 24 to 70 cm and softwood brush mat thickness from 24 to 62 cm when brush amounts were quadrupled from 10 to 40 kg m-2 brush amounts.

The application of the pre-load of 0.09 kN considerably reduced brush thickness to 11 and 42 cm and to 10 and 37 cm for 10 and 40 kg m-2 amounts of hardwood and softwood composition, respectively (Fig. 4 A). This translates to a compaction to 45.8 and 60.0% of the initial no-load thickness of the 10 and 40 kg m-2 hardwood brush amounts, respectively. Similar results were obtained when assessing softwood brush, where a compaction to 41.7 and 59.7% of the initial thickness was found for the 10 and 40 kg m-2 brush mats, respectively, when the pre-load stage was reached.

The difference in thickness was less apparent once the full-load of 10 kN was applied. Regardless of brush amounts tested during the consolidation loading event, thickness under full-load was below 10 cm for both hardwood and softwood brush mats. Also, when combining brush amounts (10, 20, 30, and 40 kg m-2), all hardwood mats were compacted between no-load and full-load state from 14.2 up to 6.2% of initial thicknesses, while the softwood mats were compacted from 16.0 up to 7.9% of initial thicknesses. This indicated that despite differences in starting thickness, brush mats of hardwood and softwood composition presented similar compressibility behaviours throughout the three loading stages.

Aside from the anticipated reduced thickness due to the consolidation loading event, the same general trends as described above were also apparent for the main-test loading performed over the same brush mats (Fig. 4 B). However, once under full-load of 10 kN, brush mat thickness was only 4.0 and 5.3% lower during the main-test loading event compared to consolidation loading event for hardwood and softwood mats (all amounts combined), respectively.

233

E. R. Labelle et al.

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227â&#x20AC;&#x201C;242)

Fig. 4 Hardwood and softwood brush mat thickness of different brush amounts (10, 20, 30, and 40 kg m-2) for test series 1 and 2 at different loading stages (no-load, pre-load, and full-load) for A) consolidation loading events and B) main-test loading events

Following consolidation and main-test loading events from each test series, sand was loosened with a spade and removed from the load box. Afterwards the load box was refilled using the same mass of sand.

Average sand density, measured from four samples was 1.57 g cm-3 (1.58, 1.56, 1.55, and 1.58 g cm-3) at 1.6 % gravimetric moisture content.

3.2 Difference between consolidation loading and main-test loading For analyzing the performances of brush mats, it was important to identify a loading scenario (e.g., consolidation or main-test loading) most representative for the matsâ&#x20AC;&#x2122; behaviour. Therefore, we examined the differences between strain responses recorded during the consolidation and main-test loading events of the two test series T1 and T2 at mats of identical brush amounts. The analysis was based on the response of the middle gauge directly located below the loading disk recorded during both test series. Combining strain gauge responses from two tests series for each brush amount would yield an average loading curve not necessarily representative of what is occurring during each test, since each brush mat may have compressed at different rates. After initial analyses in section 3.1, very similar trends were noticed for both test

234

series. However, in an attempt to avoid any shadowing since the entire strain and load curves are of interest, only results obtained during test series 2 are presented in Fig. 5 and 6.

When exposed to the 10 kN load, strain recorded by the middle strain gauge during the main-test loading event was on average 22.8% lower (all brush amounts combined) compared to strain measured during the consolidation loading event (Fig. 5 Aâ&#x20AC;&#x201C;D). The difference in response of the middle gauge between consolidation and main-test loadings generally increased as brush amounts increased from 10 to 40 kg m-2 and was more prevalent for hardwood than for softwood brush mats. In fact, in relation to consolidation loading, brush mats of 40 kg m-2 showed a strain reduction of 54.8% for hardwood and 29.1% for softwood during the main-test loading event at the full 10 kN load. Furthermore, the difference between consolidation and main-test loading curves could be observed at a much lower applied load, as brush amount increased from 10 to 40 kg m-2 for both hardwood and softwood mats.

To reduce bias associated with combining both loading events, we used strain gauge responses from the main-test loading event and not from the consolidation loading for further analyses. In addition, off road traffic of forest machinery constitutes more than a Croat. j. for. eng. 36(2015)2

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

E. R. Labelle et al.

Fig. 5 Response of middle strain gauge to consolidation loading and main-test loading for A) 10 kg m-2, B) 20 kg m-2, C) 30 kg m-2, and D) 40 kg m-2 brush amounts for hardwood and softwood mats obtained from test series 2 single loading event (impact of a single wheel), which also justifies the use of the results of the second loading event (main test).

3.3 Effect of brush mat amount and composition on strain

test series 1 and 2 (in micro strains, με) for bare sand as well as for four brush quantities. Setting the recorded strain below bare sand at 100%, the table also shows the percent reduction of strain for the four tested brush quantities during the three target loadings of the maintest loading event.

Table 2 shows the mean responses of the middle gauge located vertically below the loading disk during

Strain response recorded directly below bare sand increased from 337.0 με at 1 kN to 864.1 με at 5 kN and

Croat. j. for. eng. 36(2015)2

235

E. R. Labelle et al.

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

Table 2 Mean response of middle strain gauge of both test series (N=2) in micro strains (me) and in percent reduction compared to the response of middle gauge under bare sand to three loads (1, 5, and 10 kN) when applying no brush (0 kg m-2) or four brush quantities of hardwood or softwood after one preliminary 10 kN consolidation loading. Different lower case letters indicate a statistical difference at alpha 0.05 for a specific loading based on Tukey pairwise comparisons Sand

Loading, kN

1

5

10

Hardwood

0 kg m-2

10 kg m-2

20 kg m-2

30 kg m-2

40 kg m-2

me

337.0

155.6 a

123.0 ab

80.5 bc

61.9 c

%

100.0

53.8

63.5

76.1

81.6

me

864.1

385.0 a

296.0 ab

224.9 b

177.0 b

%

100.0

55.4

65.7

74.0

79.5

me

1310.0

465.5 a

362.0 ab

276.0 bc

222.0 c

%

100.0

64.5

72.4

78.8

83.1

Sand

Softwood -2

1

5

10

-2

-2

0 kg m

10 kg m

20 kg m

30 kg m-2

40 kg m-2

me

337.0

125.8 a

112.7 ab

75.1 bc

60.1 c

%

100.0

62.7

66.6

77.7

82.2

me

864.1

292.8 a

261.8 ab

214.5 bc

186.4 c

%

100.0

66.1

69.7

75.2

78.4

me

1310.0

369.5 a

320.5 ab

266.0 bc

233.0 c

%

100.0

71.8

75.5

79.7

82.2

up to 1310.0 με at the maximum loading of 10 kN (Table 2). When adding branches on top of the sand, it became obvious that all tested brush mats contributed significantly to reducing strain on the middle gauge compared to tests using bare sand (0 kg brush m-2).

Furthermore, the more brush was applied, the higher was the reduction of the middle strain response at all target loadings. At the 5 kN loading, hardwood brush mats reduced strain from 864.1 με to 385.0 με and 177.0 με, which translates to a 55.4 and 79.5% reduction of the strain recorded under bare sand for 10 and 40 kg m-2 brush mats, respectively. One-way ANOVA showed a statistical difference (p=0.008) between strain recorded by the middle gauge and brush amount. Based on Tukey pairwise comparisons, statistical differences existed between the strain recorded below 10 and 30 kg m-2 as well as 10 and 40 kg m-2 hardwood mats (Table 2).

40 kg m-2 brush mats, respectively, compared to the response recorded directly under bare sand. Likewise to hardwoods, one-way ANOVA showed statistical difference (p=0.009) between strain recorded by the middle gauge and softwood brush amount.

Amongst tested brush amounts, means of strain on middle gauge were statistically different between 10 and 30 kg m-2, 10 and 40 kg m-2, as well as 20 and 40 kg m-2 (Table 2).

At the same target loading of 5 kN, softwood brush mats reduced the strain response over bare sand from 864.1 με to 292.8 με with a 10 kg m-2 brush mat down to 186.4 με under the 40 kg m-2 mat. This translates to a reduction of 66.1 and 78.4% of the strain for 10 and

A second analysis compared the mean responses of the two side gauges (again as means from both test series T1 and T2) to the mean response of the middle gauge to give additional evidence of the ability of the brush mats to transfer the exerted loading to side ar-

236

At same brush amounts and loadings, softwood showed a slightly better ability to distribute applied loads than hardwood, especially at the most practically implementable brush amounts of 10 and 20 kg m-2. At higher brush amounts, the differences between the tested softwood and hardwood brush diminished and at 40 kg m-2 hardwood brush contributed to a lower response of the middle gauge by approximately 5% at 5 and 10 kN loadings compared to softwood mats.

Croat. j. for. eng. 36(2015)2

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

eas. We expected brush mats with a high load distributing capability to have relatively high responses of the side gauges compared to the middle gauge. Thus, lateral load transfer would be indicated by rather high responses of the side gauges. For this analysis, we compared the absolute strain responses of the side gauges and, in addition, we expressed these responses in percent of the middle gauge response.

In general, average side gauge response decreased with an increase in brush amount. This tendency was anticipated since lower total strain was recorded with an increase in brush. Average side gauge strain record­ ed at the 5 kN load decreased from 83.5 με to 59.5 με (28.7% reduction) between the 10 and 40 kg m-2 hardwood brush mats, respectively, and from 109.8 με to 75.5 με (31.3%) for softwood brush mats of the same amounts (Table 3). When considering the effect of the full 10 kN load, almost identical results were found, with 28.5 and 31.6% reductions of side gauge response between 10 and 40 kg m-2 hardwood and softwood mats, respectively.

The average response of the two side gauges in percent of the related middle gauge increased slightly

E. R. Labelle et al.

with an increase in brush amount for both hardwood and softwood composition (Table 3). The only exceptions were observed at the 1 kN loading for both hardwood and softwood brush mats. For softwood, the percent strains were quite constant for any amount of brush, increasing from about 30% to approximately 40% as the load increased from 1 to 10 kN (Table 3).

The percent strain for hardwood brush mats was higher with an increase in brush amount and load. Softwood brush mat showed higher absolute and relative responses of side gauges when related to the response of middle gauge, suggesting higher capability to transfer loads laterally compared to hardwood brush mats.

However, since our benchmark used for comparison »average response of middle gauge« is also varying with brush amount, absolute strain readings of the side gauge were also considered.

3.4 Relationship between load, strain, and brush mat compressibility When combining all three measured parameters: applied load (through displacement of the crosshead), brush mat compressibility, and strain on middle gauge,

Table 3 Mean response of side gauges of both test series (N=4) to received loadings in micro strains (me) and in percent of the loading of the corresponding middle strain gauge for the different test scenarios of hardwood or softwood mats and bare sand after one preliminary 10 kN consolidation loading. Different lower case letters indicate a statistical difference at alpha 0.05 for a specific loading based on Tukey pairwise comparisons Sand

Loading, kN

1

5

10

Hardwood

0 kg m-2

10 kg m-2

20 kg m-2

30 kg m-2

40 kg m-2

me

17.2

28.0 a

25.5 a

19.9 ab

13.9 b

%

5.1

18.0

20.7

24.7

22.4

me

53.9

83.5 a

79.5 a

70.3 ab

59.5 b

%

6.2

21.7

26.8

31.3

33.6

me

51.8

109.5 a

103.3 ab

87.5 ab

78.3 b

%

4.0

23.5

28.5

31.7

35.2

Sand

Softwood -2

1

5

10

0 kg m

10 kg m

20 kg m

30 kg m-2

40 kg m-2

me

17.2

39.3 a

34.4 ab

24.2 bc

16.9 c

%

5.1

31.2

30.6

32.2

28.2

me

53.9

109.8 a

100.2 a

85.2 a

75.5 a

%

6.2

37.5

38.3

39.7

40.5

me

51.8

147.0 a

130.0 a

109.5 a

100.5 a

%

4.0

39.8

40.6

41.2

43.1

Croat. j. for. eng. 36(2015)2

-2

-2

237

E. R. Labelle et al.

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

Fig. 6 Relationship between strain response of middle gauge, load, and brush compressibility recorded during main-test loadings of test series 2 on different brush amounts of hardwood (A, C, E, and G) and softwood (B, D, F, and H) composition along with a hypothetical brush mat (I) showing ideal performance with respect to brush compressibility and vertically transferred loads

it became apparent that softwood brush mats were more efficient at distributing applied loads through lower strain on middle gauge compared to hardwood brush mats throughout all test scenarios (Fig. 6). For most cases, we also noticed that hardwood brush mats offered slightly higher compressibility (demonstrated by a longer vertical travel of the crosshead) for the same applied load than softwood brush mats. Regardless of brush composition, we detected an increased compressibility at higher brush amounts (30 and 40 kg m-2) especially at loadings of 5 kN and higher (Fig. 6 E–H), while mats with low brush amounts (10 and 20 kg m-2) seemed to compress more slowly (Figures 6 A–D). Also, middle strain gauge response increased at a slower rate with an increase in brush mat amount for both hardwood and softwood mats.

To fully understand the interactions of all three parameters, a hypothetical (ideal) brush mat response to applied load was created (Fig. 6 I). In this idealized

238

scenario, strain recorded from the middle gauge would increase at a slow and steady rate with increasing load. This would offer better protection against heavy machine loadings, since a higher percentage of the applied load would be transferred to side areas. A firmer and less compressible brush mat should be able to distribute applied loadings more efficiently compared to a looser brush mat offering higher compressibility. For this reason, the idealized scenario also demonstrates relatively low compressibility of the brush mat even at the maximum loading of 10 kN.

3.5 Effect of repetitive loadings on strain Finally, we exposed hardwood and softwood 20 kg m-2 brush mats to five repetitive loadings to determine how this would affect their capability to distribute the loads exerted on top of the mats in ver­ tical and lateral directions. The highest decrease of middle gauge response was observed between the first Croat. j. for. eng. 36(2015)2

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

E. R. Labelle et al.

Table 4 Response of middle gauge (in micro strains; me and percent change (+/-) compared to previous loading event) to repetitive loadings (one to five loadings) of 20 kg m-2 hardwood and softwood brush mats. The repetitive loadings were not replicated

Loading kN

1

5

10

Hardwood

Softwood

Repetitive loadings

Repetitive loadings

1

2

3

4

5

1

2

3

4

5

me

146.0

123.1

126.1

123.9

127.8

147.7

122.5

127.0

127.2

126.3

%

0.0

–15.7

+2.4

–1.7

+3.2

0.0

–17.1

+3.7

+0.2

–0.7

me

330.2

297.8

298.8

296.6

299.6

329.4

280.7

281.8

277.8

275.8

%

0.0

–9.8

+0.3

–0.7

+1.0

0.0

–14.8

+0.4

–1.4

–5.2

me

410.0

363.0

359.0

357.0

352.0

392.0

343.0

337.0

332.0

328.0

%

0.0

–11.5

–1.1

–0.6

–1.4

0.0

–12.5

–1.7

–1.5

–1.2

two loadings for both hardwood and softwood mats at all three applied loads and ranged from 9.8 to 17.1% (Table 4). From there, percent changes decreased during the third to fifth loading to a maximum of 3.7% moving undirected around the strain responses of the main-test loading, somehow indicating a rather stable mat performance with respect to reduced vertically transferred loads. Even more, slightly decreasing response values of the middle gauge from the first to fifth loading events for the 10 kN load indicated enhanced performance of the mats as continued loading consolidated the mats. In fact, during the fifth loading an average decrease of strain of 3.7 and 4.2% compared to strain recorded during the main-test loading was apparent when comparing middle gauge responses at full compression load of 10 kN for hardwood and softwood, respectively.

4. Discussion 4.1 Branch properties and brush mat compressibility During this study, branch diameter was limited to 3 cm (at the large end) for both hardwood and softwood smallscale brush mats. Rational for this diameter limit is based on results obtained from Labelle and Jaeger (2012) and an associated MScFE project (Poltorak 2011; conducted from on-going in stand forest operations in New Brunswick), both indicating the most frequently tallied branch diameter in the creation of brush mats was within the 1–3 cm category. Likewise, McMahon and Evanson (1994) reported an average branch diameter of 3.3 cm in unconfined field brush mat tests. In general, pre-load thickness of the small scale brush mats was half of the pre-load Croat. j. for. eng. 36(2015)2

thickness of full scale brush mats of the same quantities as reported by Labelle and Jaeger (2012). The lower pre-load thickness of the small scale brush mats as compared to those used in full scale tests was anticipated since the latter permitted the use of tree tops and larger branch diameter (up to 7 cm) and length (up to 5 m) for its composition. In addition, all branches used in the current study were trimmed to a length of 35 cm to fit within the test structure. This was considered necessary because of the relatively small area used for supporting and testing brush mats (37 x 37 cm or approx. a 1:7 scale in comparison to one square meter). We believe that allowing larger diameter branches in such a confined space could have caused bias and yield erroneous results through increased friction between large diameter branches and side walls of the test structure. However, since our tested brush mats were composed of smaller branches than full scale mats, the test results may be influenced by the higher number of branch intersections causing increased internal friction adding to lateral load transfer. On the other hand, the thicker and longer branches used in the full scale mats may likely contribute to increased lateral load transfer, too, compared to the test results of this study. Further research is needed to gain more insight into load diverting pattern of full scale brush mats.

Average branch water content of 51% (in relation to green mass) for all brush mats is similar to water contents of 45–50% for Sitka spruce (Picea sitchensis Bong. Carr.) and 48% for mixed wood brush mats report­ed by Dibdiakova (2011) and Poltorak (2011), respectively. While an increase in branch water content could increase flexibility and potentially lower lateral load distributing capabilities compared to stiffer

239

E. R. Labelle et al.

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227â&#x20AC;&#x201C;242)

branches of lower water content, it could also increase compressibility and internal friction. However, due to the low variation of brush water content, we were not able to further analyse the potential effect of brush water on load distributing capabilities.

Despite major differences in compressibility between loading stages (no-load, pre-load, and full-load), hardwood and softwood brush demonstrated similar compressibility for respective loadings stages. Perhaps the confined test structure influenced the compressibility of the small scale brush mats by increasing friction between branches and the side walls of the test structure. During an unconfined test, branches under loading would have the possibility of moving in both lateral and longitudinal directions, and we would anticipate hardwood brush mats to be less compressible under loading compared to softwood brush mats of the same amount because of fewer and in average more coarse branches per mass unit.

4.2 Applied loads Applied loads of 1, 5, and 10 kN translated to 54, 268, and 537 kPa of surface contact pressure underneath the steel circular loading disk. Static nominal ground pressure exerted by loaded forwarders usually range between 70 and 180 kPa (Kozlowski 1999).

However, these static nominal ground pressures assume a penetration into the soil equal to 15% of the wheel diameter, thus greatly increasing the contact area and reducing actual peak pressures that would be observed on harder surfaces. Labelle and Jaeger (2012) reported peak dynamic surface contact pressures up to 380 kPa below a 30 metric ton Timbco eight wheel forwarder when operated directly over a steel rigid surface. These peak dynamic surface contact pressures were in relation to the size of a loading plate (30.5 cm x 30.5 cm) and were likely higher directly underneath tire threads. In addition, many forwarders are equipped with 600 or 650 mm wide tires as opposed to the 710 mm wide tires installed on the Timbco, which would also contribute to higher ground pressures for the same loading. Considering these factors, the pressures applied to the smallscale brush mats during this study are within a realistic range.

4.3 Effect of brush mat amount and composition on strain In general, an increase in brush amount corresponded to increased load distribution through higher strain readings at side strain gauges in relation to the gauge directly located below the loading disk. At a fixed load, higher brush amounts seemed to behave with more rigidity than thinner brush mats, thereby

240

allowing applied loading to be distributed more efficiently away from the loading disk onto side gauges.

Increasing the amount of branches placed inside the test structure also placed further distance between the loading disk and strain gauges located below the sand layer and brush mat, which in itself could influence stress propagation within the brush mat and corresponding sand layer. Assuming that Boussinesqâ&#x20AC;&#x2122;s (1885) stress propagation theory within a soil profile would also apply to the combination of a brush mat over a layer of sand, it is possible that with higher brush amounts, more stress was distributed to the side walls of the test structure, therefore reducing total vertical stress being transferred to the middle strain gauge. Although considerable differences exist between a soil layer and a brush mat, it is conceivable that the stress propagation formula would apply to a certain extent to how branches are capable of distributing applied loads.

Softwood brush mats were slightly more suitable at distributing applied loads laterally than hardwood brush mats. This can be attributed in part to a higher degree of small diameter branches in softwood mats compared to hardwood mats, which increases the number of contact and interaction points of the branches and with this the overall internal friction and overall brush mat stiffness. Hardwood branches have higher wood density and fewer secondary branches compared to softwood, which might impact their effectiveness to laterally distribute loadings.

4.4 Effect of repetitive loadings on strain Once placed on machine operating trails, brush mats can be exposed anywhere from two to five loadings per machine pass over depending on the number and configuration of axles. Assessing strain gauge responses below brush mats receiving a single loading (consolidation loading) indicated in average 22.8% higher strain at the middle gauge compared to strain recorded during the main-test loading event at the full 10 kN load. Strain on side gauges indicated similar responses to a third, fourth, and fifth loading. Han et al.

(2006) reported that brush mats (7.5 kg m-2 and 15 kg m-2) were effective at minimizing compactive energy of a loaded eight wheel drive Valmet 890.1 forwarder (31,434 kg) equipped with bogie wheel tracks for only the first two to three passes, after which the mat deteriorated and ceased to be beneficial. Labelle and Jaeger (2012) showed a benefit of using 20 to 30 kg m-2 brush mats in reducing peak loads up to the maximum traffic frequency tested of 12 loaded passes, despite showing a significant peak load increase following the first two passes.

Croat. j. for. eng. 36(2015)2

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

E. R. Labelle et al.

5. Conclusion

6. References

The use of brush as a covering layer placed on machine operating trails during cut-to-length forest operations can lower peak pressures by transferring applied loads to a greater area. However, since brush mat quality and quantity is dependent on stand characteristics and silvicultural treatment, this study attempted to quantify the difference in loading resistance recorded below hardwood and softwood small scale brush mats of different amounts. In particular, this study analyzed the performance of hardwood and softwood brush mats of varying quantity with respect to vertical and lateral load transfer. Results indicated that softwood brush mats offered slightly better protection against vertical load transfer and, in addition, showed a better performance in lateral load transfer compared to hardwood brush mats. Further research on larger scale brush mats should be performed before the effects of hardwood and softwood brush mats on load distribution can be extrapolated to actual in-stand forest operations.

Boussinesq, J., 1885: Application des potentiels à l’étude de l’équilibre et du mouvement des solides élastique. GauthierVillais, Paris, 721 p.

Acknowledgements Financial assistance was provided by the Natural Sciences and Engineering Research Council of Canada, FPInnovations, and the University of New Brunswick. The authors are grateful for manuscript revisions provided by Dr. Robert J. Rogers from the Mechanical Engineering Department at the University of New Brunswick and Mr. Mark Partington from FPInnovations. Technical assistance obtained from Mr. Scott Fairbairn and Mr. Dean McCarthy was also appreciated. Statistical consultations were generously provided by Dr. William Knight from the Mathematics Department at the University of New Brunswick and Dr. Marcus Lingenfelder from the Chair of Forest Operations at the University of Freiburg.

Croat. j. for. eng. 36(2015)2

Dibdiakova, J., 2011: Inherent properties in branches and stems of Norway spruce and Sitka spruce Skog & Landskap. 4 p.

Eliasson, L., Wästerlund, I., 2007: Effects of slash reinforcement of strip roads on rutting and soil compaction on a moist fine-grained soil. For. Ecol. Manage. 252(1): 118–123.

Gerasimov, J., Katarov, V., 2010: Effect of bogie track and slash reinforcement on sinkage and soil compaction in soft terrains. Croatian J. For. Eng. 31(1): 35–48.

Han, H.S., Page-Dumroese, D., Han, S.K., Tirocke, J., 2006: Effects of slash, machine passes, and soil moisture on penetration resistance in a cut-to-length harvesting. Int. J. For. Eng. 17(2): 11–24.

Instron®, 2012: 3360 series dual column tabletop universal testing systems. Address viewed online on January 10, Online URL: http://www.instron.us/wa/product/3300-DualColumn-Testing-Systems.aspx.

Jakobsen, B.F., Moore, G.A., 1981: Effects of two types of skidders and of a slash cover on soil compaction by logging of mountain ash. Australian J. For. Res. 11: 247–255.

Kozlowski, T.T., 1999: Soil compaction and growth of woody plants. Scandinavian J. For. Res. 14(6): 596–619.

Labelle, E.R., Jaeger, D., 2012: Quantifying the use of brush mats in reducing forwarder peak loads and surface contact pressures. Croatian J. For. Eng. 33(2): 249–274.

McDonald, T.P., Seixas, F., 1997: Effect of slash on forwarder soil compaction. Int. J. For. Eng. 8(2): 15–26.

McMahon, S., Evanson, T., 1994: The effect of slash cover in reducing soil compaction resulting from vehicle passage.

LIRO report Rotorua, NZ. 19(1): 1–8.

Micro Measurements, 2011. The three-wire quarter bridge circuit. Application Note TT-612. 3 p.

Poltorak, B.J., 2011: Mitigating soil disturbance in forest operations. Master of Science in Forest Engineering Thesis.

University of New Brunswick. 158 p.

241

E. R. Labelle et al.

Received: May 31, 2015 Accepted: July 28, 2015

242

Assessing the Ability of Hardwood and Softwood Brush Mats to Distribute Applied Loads (227–242)

Authors’ address: Assist. Prof. Eric R. Labelle, PhD.* e-mail: eric.labelle@tum.de Technische Universität München Assistant Professorship of Forest Operations Hans-Carl-von-Carlowitz-Platz 2 D-85354, Freising GERMANY Prof. Dirk Jaeger, PhD. e-mail: dirk.jaeger@foresteng.uni-freiburg.de Albert-Ludwigs-Universität Freiburg Institute of Forest Sciences Werthmannstrasse 6 D-79098, Freiburg GERMANY and University of New Brunswick Faculty of Forestry and Environmental Management 28 Dineen Drive, Fredericton, NB E3B 5A3 P.O. Box 4400 CANADA Benjamin J. Poltorak, MSc. e-mail: ben.poltorak@earthmaster.ab.ca Earthmaster Environmental Strategies Ltd. 103-4813 4th Avenue Edson, Alberta CANADA * Corresponding author Croat. j. for. eng. 36(2015)2

Original scientific paper

Nutrient Concentration on Skid Trails under Brush-Mats – Is a Redistribution of Nutrients Possible?

Herbert Borchert, Christian Huber, Axel Göttlein, Johann Kremer Abstract In mechanized timber harvesting, it is common practice to build brush mats from logging residues on skid trails. Protective effects of brush mats against soil compaction are documented by several studies. On the other hand, a large quantity of nutrients is concentrated on the skid trail. Fully mechanized harvesting has been criticized frequently for this reallocation of nutrients. Is there really a risk of nutrient leaching below skid trails or imbalances? Are the nutrients redistributed through nutrient uptake by roots of adjacent trees? Effects of fully mechanized thinning on soil, water and nutrient balance were examined in a seventy years old spruce stand on a nutrient poor site in Bavaria. Sections of the trails were covered with brush mats, while other sections remained uncovered. For five replications, soil physical properties, soil chemistry, matter and water balances and the density of fine roots were measured in the middle of the trail, under the tire tracks, at the transition of trail and stand and inside the stand over a period of two years. Logging operation caused soil compaction. The macropore volume decreased and both hydraulic conductivity and air permeability were severely reduced. The nutrients were largely kept in the forest ecosystem. Results of the soil moisture monitoring indicate that, within the sections covered by a brush mat, tree roots extracted water from the soil between the tracks. Without cover, the trees scarcely extracted water from this area. Hence, building a brush mat can facilitate water availability and thus enable redistribution of nutrients.

Keywords: soil compaction, brush mat, nutrient balance

1. Introduction When opening up forest stands with skid trails, traffic of heavy machinery is concentrated on these lanes. Severe changes of the structure and aeration may occur below tracks. In order to reduce such changes, skid trails are often covered with logging residues. Thus the pressure on the soil surface is supposed to be reduced. Many studies have proved the protective effect of brush mats. Schäfer and Sohns (1993) detected a decreasing penetration resistance according to the thickness of a brush mat. Also Becker et al. (1989) found a positive effect of the brush mat on the penetration resistance. Jacke et al. (2008) proved experimentally a wider distribution of the load below a brush mat. Eliasson (2005), Labelle and Jaeger (2012) and Jaeger et al. (2012) determined reduced soil compaction if the track was covered with a brush mat.

Croat. j. for. eng. 36(2015)2

However, if slash is displaced from inside the stand to the skid trail, lot of nutrients will also be relocated.

The slash comes mainly from the top of the tree, where the nutrient concentration is high. According to Weis and Göttlein (2012), the concentration of many nutrients is e.g. in needles up to twenty times of that in wood. It is not clear to which extent and how quickly tree roots can grow into the compacted area and collect the deposited nutrients. In the case of poor nutrient supply, spruce can have an extensive root system (Schmidt-Vogt 1991). However, on skid trails soil compaction may hamper rooting (Kremer 1998, Gaertig et al. 2001, Eppinger et al. 2002, Schäffer 2005). There is little knowledge on the dynamics of slash decomposition. Due to the high nutrient concentration, the rate of mineralization is higher than that of litter (Lundmark-Thelin and Johansson 1997). An accumulation of

243

H. Borchert et al.

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

slash and deadwood (e.g. after a clear cut or a dieback caused by bark beetles) can cause nitrate leaching, if the nutrient uptake by plants is insufficient (Emmet et al. 1991, Weis et al. 2001, Blumfield and Xu 2003, Huber et al. 2004, Huber 2005, Weis et al. 2006). Also, other nutrients deposited with the brush mat can be leached. Potassium and phosphor in needles and twigs are mineralised quickly (Ouro et al. 2001, Huber et al. 2004, Thiffault et al. 2006, Palviainen et al. 2005).

According to a study carried out in Finland, 49% of P and 90% of K were set free from needles, twigs and fine roots three years after a clear cut (Palviainen 2005).

Moreover K is very mobile in the soil. So the risk of leaching is high, if there is no plant uptake (Olsson 1999, Huber et al. 2004, Huber 2005). P can cause an eutrophication of lakes and rivers if there is surface runoff (Palviainen et al. 2004). It is not clear how far these processes also apply to brush mats, because the insolation is much less than in the cited case studies.

Insolation is a main driver for mineralisation (Huber et al. 2004, Palviainen 2005).

Due to rising energy prices, slash is intensively extracted and then chipped and used for energy purposes. The use as brush mat and processing for wood chips compete with each other. Von Teuffel (2012) claims that, if biomass is deposited on skid trails, nutrients are permanently out of reach of the trees. For this reason, he suggests the use of biomass for energy purposes and the return of the extracted nutrients with treated wood ash.

Practitioners of forestry are not really sure how to deal with the slash. Therefore, the Centre for Forest Nutrition and Water Resources, the Chair of Forest Work Science and Applied Informatics, both Technical University of Munich and the Bavarian State Institute of Forestry conducted a study on the dislocation of biomass and nutrients in harvesting. The focus of the study was: Does nutrient leaching occur after mineralization of slash deposited on skid trails or is a reallocation possible by taking up nutrients by roots of adjacent trees?

2. Materials and methods A stand about 70 years old in the community of Eslarn in the region of Upper Palatinate was thinned at the end of March 2007. Skid trails were cut by a harvester, equidistant at 30 meters. Thinning was carried out in a combined method of fully mechanized cutting by harvester and felling of trees that were out of crane reach by forest workers. The harvester was a Valmet 901.3 6WD, weight of about 13 t, inflation pressures in front tires between 1.6 and 2.7 bar and in rear 1.6 bar.

The timber was logged by a Timberjack 1110 8WD for-

244

warder, bare weight of about 15 t and a load capacity of about 11 t. The inflation pressure in the front tires was between 2.0 and 3.4 and in the rear 2.5 to 4.2 bars. During logging, the soil water content was about 30 Vol.%.

The stand stocks on a brown soil developed from gneiss. The mineral soil of sandy loam or loamy sand is covered by an organic layer of moder and has a skeleton fraction up to 14%. The base saturation between 6–11% indicates a very poor soil. With a mean temperature of 6.5 °C the climate is cool compared to the Bavarian average. The yearly precipitation of 800 mm (1971–2000) is slightly below average. In spite of poor nutrient stocks in the soil, the stand grows vigorously.

The diameter in breast height (dbh) of the stem of mean basal area was 23.7 cm before thinning, top height 30.7 m, stand density 1162 trees and the growing stock 670 m³ standing gross volume. 335 trees/ha or 132 m³/ha were cut during the first thinning. With four replications, skid trail sections were covered with a brush mat and others were kept free of slash.

2.1 Measuring the brush mat At four sections covered with brush mat, the height profile was measured over a width of 4 m at three spots before the first crossing and after logging again. From each replication in the area of the tire tracks, samples of the brush matt were taken in a square of 1 m². The samples structure and weight were determined.

2.2 Soil physical investigations Four replications of soil samples were taken at three spots in the tracks and in the adjacent stand (reference) immediately after thinning, and two years later again.

Core samples were taken in four depths (5–10 cm; 15–20 cm; 25–30 cm; 35–40 cm). Bulk density, pore volume, pore size distribution, air permeability and hydraulic conductivity were measured in the laboratory.

In addition, one soil column was taken from a track of each replication, covered with a net, scanned in a computer tomography and replaced precisely at the same place. Two years later, the same samples were taken, scanned again, and the roots which grew into the columns in the meantime were counted.

2.3 Chemical and hydrological investigations The chemical and hydrological investigations were conducted in each case in the middle of the skid trail, in the tire track, in the transition zone to the adjacent trees and inside the stand. These four locations form a transect. At one section of the skid trails covered with a brush mat and one section without a brush mat, two transects were established. Thus five replications could be analyzed, one more than in case of soil physics.

Croat. j. for. eng. 36(2015)2

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

H. Borchert et al.

2.4 Element input from mineralization

2.10 Forest growth

Immediately after thinning, samples of needles and twigs were collected and wrapped in bags of nylon. At each of the five replications and four sampling locations litterbags were placed on the ground. During the following two years at different times litterbags were collected, the losses in mass and the nutrient mineralization were determined after HNO3 pressure digestion.

Changes of the lightning conditions by thinning, and changes of the water and nutrient supply can affect forest growth. The growth of the stand was investigated on two plots. One plot was established on both sides of a skid trail section covered with a brush mat, the other along a section without a brush mat. On both plots (30×25.5 m and 25×25 m) the dbh was measured with a circumference measuring tape before thinning and after each following four growing seasons. All trees were numbered and the coordinates of the stem base recorded.

2.5 Element input from precipitation At all sampling locations and in the open field, precipitation collectors were established. The samples were collected monthly. The element concentration was determined from composite samples.

2.6 Changes in soil chemistry Two years after thinning, soil samples were taken from all five replications at all sampling locations. In the laboratory, the pH value, the C and N content, the cation exchange capacity (CEC) and the exchangeable cations in a NH4CL extract were measured separately for the organic layer and the soil depth of 0–10 cm, 10–20 cm and 20–40 cm.

2.7 Nutritional state of the trees During thinning, from ten trees at each replication, needles aged 1–3 were collected from the 7th whorl.

Separate samples from inside the stand and from trees standing adjacent to the skid trail were collected.

Three years later, needles were collected again. The element contents of C, N, P, K, Ca, Mg, Al, Fe, Mn, Na, Cu and Zn were determined from composite samples.

3. Results of the study 3.1 Brush mat Fig. 1 shows one brush mat profile, typical for all measured profiles. The uneven distribution of the brush mat is noticeable. The maximum height is in the middle of the skid trail. An even distribution of the slash would have been preferable. Prior to machine traffic, the mean thickness of the brush mat was on average 37 cm in the area of the left track and 24 cm on the right side with variation coefficients of 47% and 64%. After compaction, only 9 cm remained on average on the left side and 8 cm on the right. Thus the mat was compacted to a quarter or a third of its former height.

The weight of the brush mat ranged from 14 to 48 kg/m² with an average of 27 kg/m².

2.8 Chemistry of seepage water At each sampling location, a pore water sampler was installed in a depth of 40 cm. Water samples were collected in intervals of one to two months between May 2007 and the end of 2009. The pH value, HCO3-, dissolved organic carbon, total nitrogen, NH4+, NO3-, H2PO4-, SO42-, Cl-, Al, Fe, Mn, Ca, Mg, K und Na of the pore water was determined. The water fluxes were calculated by means of a water balance model. Some meteorological input data were measured directly, some were taken from a forest climate station 25 km away.

The element losses were calculated from the substance concentration and the modeled water fluxes.

2.9 Soil water content At all sampling points, ECH2O soil moisture sensors were installed detecting the water content in the upper 20 cm of mineral soil hourly.

Croat. j. for. eng. 36(2015)2

Fig. 1 The profile of a brush mat before compaction by forest machinery and after compaction

245

H. Borchert et al.

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

Table 1 Comparison of soil parameters between different locations after logging in 2007 Parameter

Bulk density, g/cm3

Porosity,%

Large macropores, 50 mm

Small macopores, 10–<50 mm

Residual pores, <10 mm

Air permeability of macropores (ka), mm2

Hydraulic conductivity (Ksat), m/d

Depth, cm

Inside stand

Track without Track with brush mat brush mat

5–10

1.05 a

1.44 b

1.37 c

15–20

1.33 a

1.53 b

1.44 b

25–30

1.3 a

1.51 b

1.39 a

35–40

1.39

1.51

1.47

5–10

57% a

40% b

47% c

15–20

48% a

37% b

46% a

25–30

49% a

38% b

47% a

35–40

45% a

38% b

44% a

5–10

16% a

4% b

5% b

15–20

12% a

4% b

7% b

25–30

15% a

5% b

10% b

35–40

11% a

8% ab

8% b

5–10

13% a

6% b

6% b

15–20

10% a

5% b

7% c

25–30

8% a

6% b

7% b

35–40

8%

6%

7%

5–10

28% a

30% a

36% b

15–20

27% a

27% a

32% b

25–30

25% a

24% a

30% b

35–40

26% a

23% a

29% b

5–10

441 a

39 b

110 c

15–20

224

76

105

25–30

304 a

137 b

191 ab

35–40

216

114

109

5–10

12.4

5.4

0.4

15–20

4.8

2.8

0.6

25–30

4.2

1.3

1.1

35–40

1.5

11.0

0.6

a, b, c – data indicate significant differences between locations at the level of 95% as a result of a t test

3.2 Soil physics In Table 1, the data of the samples collected inside the stand represent the reference. The soil bulk density increased significantly under the track without a

246

brush mat down to a depth of 30 cm and under the brush covered track to a depth of 20 cm. Down to a depth of 10 cm without a brush layer, the increase was significantly higher than with cover. The porosity decreased significantly in the track without brush mat down to the depth of 40 cm. With a brush mat, the porosity decreased only to a depth of 10 cm. In both tracks, the volume of the macropores was reduced to great depth. Here only little differences between both tracks were observed. The volume of the residual pores in the track with a brush mat increased compared to the reference. Apparently, there was a movement of macropores to residual pores. The volume of the residual pores did not shift significantly below the track without a brush mat. The results show that the traffic of forest machinery caused a soil compaction for both tracks covered with a mat and bare ground. Without a brush mat, the degree of compaction was at a higher level.

The air permeability was significantly reduced in the upper soil layer. The reduction was higher in the track without a brush mat compared to the covered track. In greater depth, only the track without a brush mat had significantly lower air permeability. In spite of the reduction, the air permeability is still slow to medium even in the track without a brush mat in the upper soil layer, according to the Bruggenwert (1966) classification.

The data regarding the hydraulic conductivity were not normally distributed. There were many low and few high values. Thus a t test was not possible in this case. The Mann-Whitney U test was performed instead. This test shows significant differences between the reference and both tracks. In contrast, the samples of both the track with and without a brush mat can belong to the same population. There were significant differences only in the depth of 35–40 cm.

Sampling was repeated two years later. Then a K value of 0.7 m/d was measured in the track without a brush mat in the same depth and no significant differences to the track with a brush mat.

The functionality of the soil has been limited to some degree due to compaction.

Two years after compaction, the bulk density did not change significantly neither in the tracks with a brush mat nor the tracks without it. Only in the second depth below the track without a brush mat, the average of 1.61 g/cm³ was significantly higher. The porosity was significantly higher in the upper soil layer below the track without a brush mat (5%) and below the track with cover (6%) than immediately after logging. Below the track without a brush mat, the porosity rose significantly at a depth of 25–30 cm and 35–40 cm.

Croat. j. for. eng. 36(2015)2

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

H. Borchert et al.

the litterbags placed in the middle of the trails contained significantly more Ca at the end than those inside the stand.

3.4 Element input from precipitation Element input from precipitation was larger in the middle of the skid trail compared to the open field (Table 2). Thus it is also affected by the surrounding forest stand. The N input was slightly above the Bavarian average. The sulfur input was comparatively low.

The deposition of Mg and K was average, and of Ca very low.

Table 2 Average throughfall deposition in 2007 to 2009 Measuring point

Fig. 2 The rate of mineralization of dry matter and different elements in litterbags filled with needles deposited inside the stand The air permeability did not change significantly during the two years. The hydraulic conductivity only changed significantly in the depth of 35–40 cm below the track without a brush mat and was then at the same low level as in the track with a brush mat.

3.3 Element input from mineralization After about two years, 36% of the needles (Fig. 2) and about 60% of the twigs (dry weight) remained in the litterbags. The mineralization of K was remarkably fast in both cases, whereas N decomposed only slightly (needles) or was even accumulated in the twigs.

Sulfur was also mineralized only slightly. The needles delivered phosphor and calcium only slowly, too.

The mineralization of Mg was similar to that of dry matter. The needles were mineralized faster inside the stand and at the transition of trail and stand than on the operating trail. Significantly more dry matter remained in the litterbags placed in the middle of the skid trail and on the tracks (95% level). In case of the twigs, only marginal differences in mineralization were observed referring to the location. Remarkable differences occurred in mineralization of N from needles. N has been mineralized inside the stand and at the transition to the trail, whereas N was accumulated in needles placed on the skid trail. These differences were significant as well. The location did not affect the mineralization of K, Ca, Mg and P significantly. Only Croat. j. for. eng. 36(2015)2

H2O

N

S

Ca –1

mm

Mg

K

–1

kg ha y

Open field

932

10.9

3.7

1.8

0.7

3.0

Inside stand

633

20.7

5.6

4.1

1.2

15.2

Edge of skid trail

647

21.0

5.7

4.3

1.4

15.6

Middle of skid trail

765

15.2

4.5

3.2

1.0

10.2

3.5 Changes in soil chemistry The potassium stocks were highest in all depths in the middle of the skid trail covered with a brush mat (Table 3). In some cases, they were significantly higher compared to the reference, in others compared to the skid trail without a brush mat. This is in good agreement with the high mineralization rates of K in needles and twigs. Also, the Ca and Mg stocks were highest in the organic and upper soil layer on skid trails with a brush mat. However, the differences were only significant in case of Ca in the upper soil layer. The Ca stock was significantly lower at a depth of 20–40 cm below the trail with a brush mat compared to those without a mat. The N stocks were significantly lower in depths of 10–20 and 20–40 cm on skid trails with a brush mat than those without a mat. The N stock inside the stand was significantly lower in a depth 20–40 cm than at the skid trail without a brush mat.

3.6 Nutrition state of trees At the beginning, the element contents of needles were low in case of N, P, Ca and Mg when compared to other regions of Bavaria (Weis and Göttlein 2012).

They were very low in case of Fe and slightly above average for Mn. Three years after thinning, there were hardly any differences between treatments and locations. Three-year needles from trees along skid trails

247

H. Borchert et al.

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

Table 3 Total C and N pool and the stock of different elements at the cation exchanger in soil samples from inside the stand and in the middle of the skid trail measured three years after thinning Depth

Organic layer

0–10 cm

10–20 cm

20–40 cm

C

Location

N

K

Ca

Mg

–1

Al

Mn

Fe

CEC

–1

t ha

kmol ha

Inside stand

62.5

2.5

0.9 ab

4.6

1.6

2.7

0.9

1.2

17.4

Skid trail with mat

72.8

3.0

1.3 a

7.2

2.8

3.9

1.6

1.5

23.7

Skid trail without mat

58.6

2.3

0.6 b

4.0

1.8

3.2

0.9

1.4

17.1

Inside stand

30.5

1.7

0.6 a

0.8 a

1.1

27.1

0.6

6.9

43.6

Skid trail with mat

32.9

1.8

1.2 b

2.2 b

2.0

30.9

0.7

6.1

49.7

Skid trail without mat

35.2

2.1

0.6 ab

1.4 ab

1.6

30.4

0.9

9.8

52.0

Inside stand

14.8 ab

1.5 a

0.7 ab

0.8

1.0

22.7

1.3

0.4

32.3

Skid trail with mat

15.5 a

1.2 b

0.9 a

0.9

1.0

25.1

0.8

0.6

34.5

Skid trail without mat

11.3 b

1.5 ab

0.5 b

0.9

0.9

21.5

1.3

0.6

30.1

Inside stand

13.1

2.7 ac

1.4 ab

1.6 ab

1.7

34.4

2.1

0.2

51.6

Skid trail with mat

13.0

2.0 b

1.9 a

1.5 a

1.9

49.3

2.0

0.3

68.5

Skid trail without mat

11.3

3.1 c

1.0 b

2.2 b

1.8

37.8

2.2

0.4

54.8

a, b, c – data indicate significant differences between locations at the level of 95% as a result of a t test

with a brush mat had significantly lower concentrations of N and Mg compared to trees from inside the stand.

area affected is small. The nutrients transferred to the skid trails remained largely in the ecosystem except for one case.

3.7 Chemistry of seepage water The nitrate concentration of seepage water was very low during the investigation period, both inside the stand and below the skid trails. A very high nitrate concentration was measured only on one of five skid trails covered with a brush mat. In this case, the concentration of Ca Mg and K was high as well. Without this outlier, the concentration of N, Ca, Mg and K did not differ significantly between skid trails with and without a brush mat.

3.8 Element losses with seepage water The water fluxes increased on skid trails because of reduced interception (Table 4). The soil in Eslarn is not nitrogen saturated. Thus, the nitrogen losses were very low inside the stand. On skid trails with a brush mat the nitrogen losses (mainly nitrate) and the losses of other nutrients were very high in one case. In the other four cases, the nutrient losses were similar to those of skid trails without a brush mat, but slightly higher than inside the stand. These additional losses are marginal referring to the whole stand because the

248

Table 4 Material flow with the percolating water in a soil depth of 40 cm on average of 2007 to 2009 Measuring point

H2O

N

Ca

Mg

K

mm

kg ha–1 y–1

301 0.7

2.5

2.2

4.9

Skid trail with brush mat, without outlier 487 1.4

4.2

5.4

5.8

Inside stand

Skid trail with brush mat, only outlier

487 76.8 9.4 12.5 15.0

Skid trail without brush mat

487 1.4

3.5

4.2

6.8

3.9 Development of soil water content The curve of the water content of soils on skid trails with a brush mat and inside the stand is conspicuously similar. After rainfall, the water content rises considerably. The maxima do not differ between the treatments, because the porosity in the middle of the skid trails was not affected. After rainfall, the water Croat. j. for. eng. 36(2015)2

Nutrient Concentration on Skid Trails under Brush-Mats ... (243â&#x20AC;&#x201C;252)

H. Borchert et al.

Fig. 3 The development of the volumetric water content inside the stand (reference), in the middle of the skid trail with and without a brush mat content decreases again rapidly due to the drainage of macropores. Afterwards, the water content decreases further inside the stand and on skid trails covered with a brush mat during the growing season. In contrast, on skid trails without a brush mat, the water content hardly decreases.

The bars below Fig. 3 show periods with significant (Sig.) and with no significant (N. sig.) differences and periods not ratable (N. rat.). The upper bars show differences between both skid trails, the bars in the middle between the skid trail with brush mat and the reference and that below differences between the skid trail without a mat and the reference.

The differences were significant for a long period particularly during the first growing season. In the second year, some instruments failed during summer, thus significance could not be calculated in July and August. Afterwards, the differences were significant till November. In the third year, the time period ended already at the end of July for the skid Croat. j. for. eng. 36(2015)2

Fig. 4 The rooting of soil columns below tire tracks covered with a brush mat (right) and without a mat (left) two years after passing of forest machinery trails without a mat. During winter, there are periods when the soil is dryer on skid trails without a brush mat. These periods coincide with frost periods. The

249

H. Borchert et al.

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

soils were frozen on skid trails without a brush mat, and for this reason the instruments recorded wrongly dryer soils.

3.10 Rooting Two years after logging, five to six times more roots grew in soil columns of tracks covered with a brush mat than in the soil below bare tracks (Fig. 4).

3.11 Forest growth After thinning, an improved lightning can particularly be expected for trees along new skid trails. All trees with a distance up to 4 m from the middle of the skid trail were defined as edge trees. Neither a t test nor a Mann-Whitney U test showed significant differences of the average dbh between both plots and between edge trees and the other trees before thinning. The MannWhitney U test was performed because the dbh distribution of the stand was not normal. Even four growing seasons later no significant differences of the dbh were found between the plots. The mean diameter increment of edge trees of 0.47 cm /year at the plot along the skid trail with a brush mat was significantly higher than that of the other trees (0.31 cm/year). The edge trees on the plot without a brush mat did not differ significantly from the others. It could be possible that the edge trees along the trail with a brush mat had an improved competitive situation after thinning. Then the increased growth could be due to a better lightning. Thus a competition index C was calculated for each tree. The index is the sum of the quotient of dbh difference and distance of the next three neighbor trees: C1 =

dbh1 − dbhN1 dbh1 − dbhN2 dbh1 − dbhN3 + + (1) distanceN1 distanceN2 distanceN3

Where: N1, N2 and N3 are the next neighbor trees.

No significant differences could be found between the edge trees and the other trees on both plots, neither immediately after thinning nor at the end.

4. Discussion As the results show, CTL harvesting can cause an accumulation of nutrients in the area of skid trails in case the slash is deposited on the trail. The nutrients are decomposed at different pace. Potassium was mineralized very quickly. This is in accordance with other studies mentioned before. Hence, the potassium content was elevated considerably in soils below skid trails covered with brush. Phosphor was mineralized rather slowly in difference to the results of Palviainen

250

(2005). The nutrient losses by leaching were low except for one case. The high nutrient losses on one skid trail must not have been caused by depositing the slash there. The nutrient reallocated to the skid trails remained mostly within the ecosystem.

The brush mats were rather thick. Almost everywhere they had more than 15 kg/m², and because of that they should have a considerable protective impact according to Jacke et al. (2008). On average they were heavier than 20 kg/m², a weight causing a significant reduction of measurements exceeding the 80% standard Proctor density threshold as Jaeger et al. (2012) observed. However, the soil was compacted below the brush mat. As shown by the bulk density and porosity, the compaction was not as severe as without a brush mat. The functionality of the soil regarding aeration and drainage was reduced but still existed.

The level of water content in Fig. 3 seems to be high compared to porosity (Table 1). After rapid drainage following rainfall, the water content is about 43% during periods of wet soils. The porosity less the large macropores is about 39% in the upper 20 cm. Probably Fig. 3 overvalues the water content up to 10%. However, important are the differences between the curves.

The increased moisture in the soil on the skid trail not covered with a brush mat might be partly due to higher infiltration, because there is less interception. However, it can also indicate a hindered water uptake of the tree roots. Although the soil was not deformed seriously, the roots of the trees adjacent to the skid trail without a brush mat apparently were crushed or sheared. At least for three years, the tree roots recolonized only a little the space between the tire tracks. In case of skid trails with a brush mat, the trees could still take water. Thus nutrients from the skid trail could be transferred back inside the stand. As the nutrients mineralize in parts slowly, the relocation will be slow as well. Generally, relocation is possible from the investigated skid trails with a brush mat.

Dietrich (2011) observed a correlation between soil moisture and the height of a brush mat on skid trails as well. He assumed a mulching impact of the slash and therefore expected higher water content below a brush mat. The measurement showed the opposite: The higher the brush mat, the dryer was the soil.

Apparently, the trees adjacent to skid trails without a brush mat could not benefit from the better lightning after thinning. Maybe their growth was limited by nutrient or water supply, because soil compaction below tracks hindered rooting. As forest growth was observed only at one skid trail with and one without a brush mat, the results are valid only for these plots.

The observed growth reactions are plausible. We Croat. j. for. eng. 36(2015)2

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

should not expect this reaction each time. Increment differences could be expected only if nutrient or water supply limits tree growth.

A growth surplus of the trees growing along skid trails with a brush mat could explain the lowered N and Mg concentrations in the needles. Here, a dilution may have occurred. An increased growth could also explain the lower nitrogen content in the soil covered with a brush mat. Possibly the nitrogen uptake from the soil increased. Concurrently nitrogen was mineralized very slowly, so that at least temporarily a depletion of nitrogen below skid trails with a brush mat could occur. Furthermore, a depletion of nitrogen could have been caused by a nitrogen blockade. Microbial decomposition of the slash requires finally much nitrogen.

5. Conclusion Soil deformations by machinery traffic are largely irreparable, at least in the foreseeable future. As the investigation shows, rooting of the soil can already be hindered by soil compaction even if no rutting occurs.

Thus the uptake of water and nutrient by trees is hampered. This can have an impact on tree growth, if water or nutrients are in short supply. Particularly on dry and nutrient poor sites, the skid trails should be treated with care so that the soil can serve as a reservoir of water and nutrients. If forest machinery drives off road, the soil can be protected (1) by applying low ground pressure, (2) choosing periods of dry conditions or (3) covering the soil with a brush mat of sufficient thickness. Hence, we suggest continuing to lay a brush mat.

The statement of von Teuffel (2012), according to which the biomass deposited on the skid trail means that nutrients are permanently out of reach of the trees, is not true. If the machinery operates carefully, a relocation of nutrients remains possible.

The situation becomes critical if soil deformation hinders rooting in spite of a brush mat coverage. In this case the accumulated nutrients are scarcely available for the trees. Unfortunately, hampered rooting is not visible by means of the track shape. Methods should be developed for measuring the rooting on skid trails easily. At least returning of the slash from skid trails by machinery should be considered in case of hampered rooting on poor sites.

6. References Becker, G., Hofmann, R., Roeder, A., Eisenbarth, E., Hanewinkel, M., 1989: Bodenschäden durch Forstmaschinen auf Tonstandorten? – Entstehung, Messung, Begrenzung. Forst und Holz 10: 507–512.

Croat. j. for. eng. 36(2015)2

H. Borchert et al.

Blumfield, T.J., Xu, Z.H., 2003: Impact of harvest residue on soil mineral dynamics following clearfall harvesting of a hoop pine plantation in subtropical Australia. Forest Ecology and Management 179(1–3): 55–67.

Bruggenwert, M.G.M., Janse, A.R.P., Koenigs, F.F.R., 1966: Algemene Bodemkunde. Handleiding kandidaatspraktikum Landbouwhogeschool, Wageningen, 162 p.

Dietrich, K., 2011: Auswirkungen einer Reisigauflage auf die Bodenfeuchte einer Rückegasse. Masterarbeit an der Fakultät Forst, Geo und Hydrowissenschaften der Technischen Universität Dresden, 82 p.

Eliasson, L., 2005: Effects of forwarder tyre pressure on rut formation and soil compaction. Silva Fennica 39(4): 549–557.

Emmett, B.A., Anderson, J.M., Hornung, M., 1991: The controls on dissolved nitrogen losses following two intensities of harvesting in a Sitka spruce forest (N. Wales). Forest Ecology and Management 41(1–2): 65–80.

Eppinger, M., Schack-Kirchner, H., Hildebrand, E., 2002: Rückegassen als Wurzelraum? AFZ/Der Wald 10: 489–491.

Gaertig, T.H., Schack-Kirchner, H., Hildebrand, E.E., 2001: Steuert Gasdurchlässigkeit im Boden Feinstwurzeldichte und Vitalität der Eiche? AFZ/Der Wald 25: 1344–1347.

Huber, C., Baumgarten, M., Göttlein, A., Rotter, V., 2004: Nitrogen turnover and nitrate leaching after bark beetle attack in Mountainous Spruce Stands of the Bavarian Forest National Park. In: Wieder, R.K., Novak, M. and Vile, A. (Ed.): Biogeochemical Investigations of Terrestrial, Freshwater and Wetland Ecosystems Across the Globe. Water, Air, and Soil Pollution: Focus 4, 391–414.

Huber, C., 2005: Long lasting nitrate leaching after bark beetle attack in the highlands of the Bavarian Forest National Park. Journal of Environmental Quality 34(5): 1772– 1779.

Jacke, H., Sengpiel, A., Brokmeier, H., 2008: Zur Druckverteilung unter Reisigmatten. Forst u. Technik 10: 22–27.

Jaeger, D., Labelle, E.R., Poltorak, B.J., 2012: Soil disturbances by off-road traffic of forwarders, magnitude, persistence and mitigation. Proceedings 45th FORMEC conference, October 8–12, Dubrovnik, Croatia.

Kremer, J., 1998: Befahrungsbedingte Strukturveränderungen von Waldböden und ihre Auswirkungen auf das Wachstum von Fichten, Kiefern und Buchen auf ausgewählten Standorten. GCA-Verlag (Forschen und Wissen Forstwirtschaft), Herdecke, 177 p.

Labelle, E.R., Jaeger, D., 2012: Quantifying the use of brush mats in reducing forwarder peak loads and surface contact pressure. Croatian Journal of Forest Engineering 33(2): 249– 274.

Lundmark-Thelin, A., Johansson, M.B., 1997: Influence of mechanical site preparation decomposition and nutrient dynamics of Norway spruce (Picea abies (L.) Karst.) needle litter and slash needles. Forest Ecology and Management 96(1–2): 101–110.

251

H. Borchert et al.

Nutrient Concentration on Skid Trails under Brush-Mats ... (243–252)

Olsson, A., 1999: Effects of biomass removal in thinnings and compensatory fertilization on exchangeable base cation pools in acid forest soils. Forest Ecology and Management 122(1): 29–39.

Ouro, G., Perez-Battalon, P., Merino, A., 2001: Effects of silvicultural practices on nutrient status in a Pinus radiata plantation: Nutrient export by tree removal and nutrient dynamics in decomposing logging residues. Ann. For. Sci. 58: 411–422.

Palviainen, M., 2005: Logging residues and ground vegetation in nutrient dynamics of a clear-cut boreal forest. Dissertationes Forestales 12. Faculty of Forestry, University of Joensuu, Finland, 38 p.

Palviainen, M., Finér, L., Mannerkoski, H., Piirainen, S., Starr, M., 2005: Changes in the above and below ground biomass and nutrient pools of ground vegetation after clear cutting of a mixed boreal forest. Plant and Soil 275(1–2): 157–167.

Schäfer, T., Sohns, D., 1993: Minderung der Bodenverdichtung durch eine Reisigauflage. AFZ/Der Wald 9: 452–455.

Schäffer, J., 2005: Bodenverformung und Wurzelraum.

Teuffel et. al. (Hrsg.) Waldumbau für eine zukunftsorientierte Waldwirtschaft, Springer, 345–361.

Schmidt-Vogt H., 1991: Die Fichte, Band II/3. Paul Parey, Hamburg-Berlin, 781 p.

Thiffault, E., Paré, D., Bélanger, N., Munson, A., Marquis, F., 2006: Harvesting intensity at clear felling in the boreal forest: Impact on soil and foliar nutrient status. Soil Science Society of America Journal 70(2): 691–701.

Von Teuffel, K., 2012: Nachhaltigkeit und Holznutzung.

AFZ-Der Wald 4: 8–9.

Weis, W., Huber, C., Göttlein, A., 2001: Regeneration of mature Norway spruce stands. The impact of clear cutting and selective cutting on seepage water quality and soil fertility.

In: Optimizing Nitrogen Management in Food and Energy Production and Environmental Protection. The Scientific World Journal 1: 493–499.

Weis, W., Rotter, V., Göttlein, A., 2006: Water and element fluxes during the regeneration of Norway spruce with European beech: Effects of shelterwood cut and clear cut. Forest Ecology and Management 224(3): 304–317.

Weis, W., Göttlein, A., 2012: Nährstoffnachhaltige Biomassenutzung. LWF aktuell 90: 44–47.

Authors’ address: Herbert Borchert, PhD.* e-mail: Herbert.Borchert@lwf.bayern.de Bavarian State Institute of Forestry Hans Carl von Carlowitz Platz 1 GERMANY Prof. Christian Huber, PhD. e-mail: christian.huber@hswt.de University of Applied Science Am Hofgarten 4 GERMANY Prof. Axel Göttlein, PhD. e-mail: goettlein@forst.tu-muenchen.de Forest Nutrition and Water Resources Technical University of Munich Hans Carl von Carlowitz Platz 2 GERMANY

Received: April 30, 2014 Accepted: January 14, 2015

252

Johann Kremer, PhD. e-mail: kremer@wzw.tum.de Forest Work Science and Applied Informatics Technical University of Munich Hans Carl von Carlowitz Platz 2 85354 Freising GERMANY * Corresponding author Croat. j. for. eng. 36(2015)2

Original scientific paper

The Effects of Mechanised Log Skidding on some Components of Farm Tractors in Ondo State, Nigeria Richard Omoyele Akinbamowo, Basil Oluwafemi Akinnuli Abstract An investigation was made into the effects of mechanised logging and log skidding on some components of farm tractors used under the Tractor Hiring Units (THU) of the Ministry of Agriculture in Ondo State, Nigeria. Thirty tractors of various makes and models were examined after a period under logging operations based on a template of eight key criteria and thirty four sub criteria related to vital tractor systems and components. Data collected were analysed to determine the percentage of the components that are good, damaged but repairable or damaged or missing parts. The results show that the highest occurrence of failure are with worn out tyres (80%), followed by broken rear work lights, headlamps and road warning lights (60%) and broken lateral stabilizers (53.3%). Similarly, the least damaged components of tractors under the study were the tyre rims (66.7% good), wheel hubs and axle arms (56.7%) and hydraulic linkage arms and forks as well as brake pedal travel and release (46.7%).

Keywords: tractors, damage, serviceable, faults, forestry

1. Introduction Sustaining the natural productivity of forests during harvesting operations is a major concern for policy makers in the forest zone of Nigeria and across the world. Apart from the known contributions of deforestation to global warming and climate change, this will guarantee that remaining trees, wildlife and soil nutrients are retained in their original position and proportion. With the advent of modernisation, forestry extraction activities have adopted mechanisation to remove the drudgery associated with the process.

The adverse effects of machines on the forest eco system and residual stand have been documented by several researchers including Ezebilo (2004), Adekunle and Olagoke (2010) and FAO (2011), however challenges of the failure of tractors that are primarily designed and used for arable crop farming and now converted to use in forestry mechanisation without necessary adaptation have not been adequately studied. As an alternative, horse logging is generally considered more expensive and less productive than tractor logging (McNamara and Kaufman 1985).

Croat. j. for. eng. 36(2015)2

A major incentive for tractor use in log extraction is that hiring rates for forestry activities are higher than for crop husbandry purposes. The cost of acquiring a medium range agricultural 4 WD tractor with forestry specifications and protection is quite high. In addition, tractor owners take decision to deploy their tractors for log extraction to cure the perennial problem of underutilisation during the dry season to shore up the annual use of tractors, which have been found to be at a deplorable range of 678.92 to 534.4 hours by Oluka (2000).

The consequence of this is that failures may occur more frequently, more intensly and in unusual places on these tractors. Upon more investigation, it is suspected that the increased cost of hiring may not account for the repair costs and reduction in service years.

According to FAO (2011) estimates, Nigeria has 693,000 ha of forest plantations. Ondo state has a total of 3076.16 square kilometres of forest reserves, of which 92.2% are high forests, and 6.3% is savannah, while 1.5% is mangrove swamp. About 1391.63 (45.3%)

253

R.O. Akinbamowo and B.O. Akinnuli The Effects of Mechanised Log Skidding on some Components of Farm ... (253–258)

Table 1 Template for survey of tractor conditions Make/model of tractor

Registration No

Engine No

Chasis/serial No

Location

Criteria

2 1

Transmission system

4

Worn clutch pressure plate

3

Brakes and final drive

Clutch

Damaged release bearings

7

Loss of power assistance Steering Connection loose or damaged

9

Low oil level Damaged steering shaft assembly, Alignment drag links and tie rod assembly

11

Oil leak from steering ram

12

Excessive brake pedal travel

13

Brake grab and drag

14

Brakes

Hydraulic system

Brakes fail to release

16

Brakes inefficient

17

Damaged a) Top link b) Lifting arm (L&R)

19 20

c) Lifting forks (L&R) Linkages

21

Drawbar

23 24

e) Broken lateral stabilizers

Drawbar

7

Rims/tyres

General/ physical condition

26 27

Broken drawbar bracket Missing hooks and pins Broken rim

25 6

d) Lower links (L&R)

Lift fails to operate/erratic during lowering/raising

22 5

Braking is unbalanced

15

18 4

Clutch slip

6

10

Traction Devices

Incorrect tyre pressure Damaged wheel hubs and axle arms

29

Damaged front grille, hood or side panels

31

General

Damaged instruments panel ROPS and FOPS*

32

Broken rear work lights Lighting Broken headlamps and road warning 33 Equipment lights

8 Accessories 34

Wheel spanner, jack and repair tools

* ROPS – Roll Over Protection Structure; FOPS – Falling Object Protection Structure

254

2. Material and methods A small sample of 30 tractors was studied to collect data on tractor failures after the tractor has been in use for skidding for a period of three months. Physical inspection of tractor condition was conducted and documented by experienced agricultural mechanics in a conditional survey. Data was elicited on vital tractor systems using a template of eight key criteria and thirty four sub criteria shown in Table 1. Selected tractors have previously been used for similar operation and they were reported to be in good condition before commencement.

Results were scored in code of Good (1), Faulty and Serviceable (2), Defective or missing (3) for easy analysis. Analysis was done with SPSS statistical package.

Worn out tyres

28

30

Score

Oil leak from gearbox Difficulty in engaging/disengaging gear

8 Steering and alignment

Gear

3

5

2

Subcriteria Gear noisy in neutral and in gear

1

of these forest reserves have been encroached (DAS 2009). Similarly, 86,082 and 87,370 trees were harvested in 2007 and 2008, respectively.

Tree harvesting methods are often classified either as »Low Level« that is chainsaw felling and tree processing–line skidding, »Medium Level« that is chainsaw felling, line skidding and tree processing at landing or »High Level« which involves harvester felling and tree processing – forwarding logs to landing. Logging of timber in Nigeria is often done selectively, first selecting the mature trees. During the harvesting, the trees are felled, delimbed and topped, then skidded to the landing with tractors fitted with a winch (optional), cable or chains. In certain cases, after buckling, trees may be sawn in situ, loaded into modified tractor trailers for haulage.

Although no research has been carried out in the area of study on the failure rate of these tractor components and systems, this paper seeks to highlight the location of common failures and repair costs.

3. Results and discussion The distribution of tractors by location in Table 2 shows that the highest percentage (36.7%) of the sampled tractors were from Owo zone, 30% of the tractors were from Ikare zone, 16.7% were from Okitipupa and Akure zones.

Fig. 1 shows the distribution of tractors by their model. TAK Tractor with 13 units forms the majority of the sampled tractors. 6 of the tractors were Massey Ferguson model, while 5 were Swaraj model. 2, 3, and 1% were Mahindra, New Holland and Steyr Ursus models, respectively.

Croat. j. for. eng. 36(2015)2

The Effects of Mechanised Log Skidding on some Components of Farm ... (253–258) R.O. Akinbamowo and B.O. Akinnuli

Table 2 Distribution of respondents by location Zones

Frequency

Percent

Akure

5

16.7

Ikare

9

30.0

Okitipupa

5

16.7

Owo

11

36.7

Total

30

100.0

Source: Field Survey 2014

Fig. 1 Distribution of tractors by models

3.1 Brake condition Table 3 shows that half of the sampled tractors had serviceable problems pertaining to excessive brake pedal travel, 46.7% of them were in good condition not

having this problem, whereas the status of only few tractors (3.3%), as regards the challenge, was not provided. Results further show that majority (53.3%) of the tractors had service problems related to brake grab

Table 3 Percentage of occurrence of tractor faults Criteria

Brakes

Good, %

Serviceable, %

Bad, %

No response, %

Excessive brake pedal travel

46.7

50.0

–

3.3

Brake grab and drag

43.3

53.3

–

3.3

Braking is unbalanced

30.0

63.3

–

6.7

Brakes fail to release

46.7

36.7

–

16.7

Brakes inefficient

26.7

63.3

–

10.0

Damaged a) Toplink

33.3

13.3

46.7

6.7

b) Lifting arm (L&R)

46.7

40.0

13.3

–

c) Lifting forks (L&R)

46.7

26.7

26.7

–

d) Lower links (L&R)

40.0

13.3

40.0

6.7

e) Broken lateral stabilizers

16.7

30

53.3

–

Lift fails to operate/erratic during lowering/raising

26.7

20.0

13.3

40.0

Broken drawbar bracket

43.3

13.3

40.0

3.3

Missing hooks and pins

13.3

36.7

46.7

3.3

Broken rim

66.7

13.3

16.7

3.3

Worn out tyres

3.3

13.3

80.0

3.3

Incorrect tyre pressure

16.7

43.3

16.7

23.3

Damaged wheel hubs and axle arms

56.7

20.0

16.7

6.7

Broken rear work lights

3.3

30.0

60.0

6.7

Broken headlamps and road warning lights

6.7

46.7

40.0

6.7

Damaged front grille, hood or side panels

30.0

50.0

10.0

10.0

Damaged instruments panel

20

43.3

10

26.7

ROPS and FOPS

6.7

50

10

33.3

–

30

3.3

66.7

Loss of power assistance

26.7

46.7

6.7

3.3

Connection loose or damaged

43.3

43.3

6.7

6.7

Low oil level

26.7

56.7

3.3

13.3

Linkages

Drawbar

Traction devices

Lighting equipment

General Accessories Steering

Wheel spanner, jack and repair tools

Croat. j. for. eng. 36(2015)2

255

R.O. Akinbamowo and B.O. Akinnuli The Effects of Mechanised Log Skidding on some Components of Farm ... (253â&#x20AC;&#x201C;258)

and drag, 43.3% of them were in good condition, while there was no response for 3.3%. Most (63.3%) of the tractors had problems of unbalanced braking, some (30%) of them were in good condition, while the status of few (6.7%) of the tractors was not known. Pertaining to the sub criteria of brakes failing to release, it is obvious from Table 3 that majority (46.7%) of the tractors do not encounter the problem. 36.7% of them had this problem, while the status of some (16.7%) of them was not known. With respect to inefficient brakes, 26.7% were in good condition, 63.3% had problems, while the status of 10% of them concerning this challenge was not known. Summarily, the brake system of most tractors sampled with respect to all signs of brake performance, were high in percentage with respect to faulty but serviceable status. From this, it could be said that brake system of most tractors used in the study is in poor condition.

3.2 Hydraulic linkage Results show that Top-link of 46.7% of the tractors was damaged or missing, 33.3% were in good condition, 13.3% were faulty but serviceable. However, the status of Top-link of 6.7% of the tractors was not known.

46.7% of the tractors had their lifting arm (Left or Right) in good condition, though 40% of the lifting arms (L&R) were faulty but serviceable. 13.3% of the lifting arms (L&R) were critically damaged. Similarly, the lifting forks (L&R) of 46.7% of the tractors were in good condition. The status of 26.7% of them was faulty but repairable, with another 26.7% critically bad. Results further show that 40% of the tractors had their lowering link (L&R) in good and bad condition, respectively, 13.3% of the lowering links were faulty but serviceable, while there was no response to indicate the status of 6.7% of them. Analysis also reveals that broken lateral stabilizers are common with a majority (53.3%) of the tractors, 30% were faulty but serviceable, while few (16.7%) were in good condition. As to whether lift fails to operate/erratic during lowering/raising, most (40%) of the respondents gave no comment. This might probably be because tractors were not examined with mounted implement during data collection. However, 20% of response indicated faulty but serviceable conditions, while 13.3% were in critical condition and only 26.7% of them in good condition. Based on this analysis, it is apparent that linkages conditions of most tractors in the study area are moderately functional.

3.3 Drawbars Analysis of the condition of tractor drawbars shows that the drawbar bracket of 43.3% of the tractors was in good condition, 40% of those brackets were

256

Fig. 2 Damaged drawbar bracket of Massey Ferguson 375 tractor critically bad (Fig. 2), while 13.3% were faulty but serviceable. The status of few (3.3%) of them were not known. This result might be the outcome of improvisation of winches, converting drawbar brackets to hitch point as commonly practised by operators during log skidding. With respect to missing hooks and pins, majority (46.7%) of the tractors reported them missing, 36.7% were partly missing, while few (13.3%) were in good condition. The status of only 3.3% of them was not known. Thus, the drawbar status of most tractors is averagely poor.

3.4 Traction devices The results show that 66.7% of the tractors did not have problems of broken rim, but 13.3% of them had a minor problem in this area, while 16.7% had critically bad rims. The status of 3.3% of them was not known. Results also revealed that most (80%) of the tractors in the study encountered problems of bad, worn out tyres. 13.3% of them had partly worn out tyres, while only 3.3% of them had their tyres in good condition. There was no response to indicate the status of 3.3% of them. It was found that most (43.3%) of the sampled tractors suffer from incorrect tyre pressure of 30 psi. This might be attributable to the fact that most of the tractors were not serviceable during data collection. Only a few (16.7%) of the tractors in the study were in good condition and free from this challenge, some (23.3%) of the respondents did not unveil the status of this component probably because this problem can easily be remedied. In the same vein, the wheel, hub and axles arms of majority (56.7%) of the tractors were found to be in good condition, while 20% were faulty but serviceable. However, 16.7% of them were critically bad. The status of about 6.7% was not known.

Croat. j. for. eng. 36(2015)2

The Effects of Mechanised Log Skidding on some Components of Farm ... (253â&#x20AC;&#x201C;258) R.O. Akinbamowo and B.O. Akinnuli

3.5 General physical condition and accessories The results further show that front grille, hood or side panels of majority (50%) of the tractors were faulty but repairable, 30% of the tractors had these components in good condition, while about 10% of them had the component in critically bad condition.

There was no response for the remaining 10% regarding the condition of this component. Also, the instrument panels of 43.3% of the sampled tractors were faulty but serviceable, 20% of them had the component in good condition, while the remaining 10% had it in bad condition. The status of this component in some 26.7% of the tractors was unknown. Furthermore, Roll Over Protection Structure (ROPS) and Falling Object Protection Structures (FOPS) in half of the tractors were not available. 10% of them had the components in bad condition and only 6.7% of the tractors had it in good condition when available.

There was no response to indicate the status of the components in about 23.3% of the tractors. Here, the inadequacy of response might indicate the lack of comprehension of what constitutes ROPS and FOPS by respondents or the general lack of ROPS on many tractors in the state. Wheel spanner, jack and repair tools of about 30% of the tractors were faulty but serviceable, none of them had the components in good condition, while only 3.3% of them had it in critically bad condition. The majority (66.7%) of the respondents did not indicate the status of these components.

Summarily, with respect to all components, it could

be said that most of the tractorsâ&#x20AC;&#x2122; accessories are faulty but repairable.

Furthermore, the analysis of lightning equipment in Table 3 shows that the rear work light of most (60%) of the tractors was broken and in bad condition, 30% of them had faulty but repairable rear work light, while only 3.3% of them had good rear work lights.

6.7% of the respondents did not indicate the status of their tractors rear work light. Considering the status of headlamps and road warning lights, majority (46.7%) of the tractors had faulty but repairable headlamps and road warning lights, 40% of them had bad headlamps and road warning lights, while few (6.7%) of them had good headlamps and road warning lights.

The status of headlamps and road warning was unknown in about 6.7% of the tractors. Based on this result, it could be said that lightning status of most of the tractors in the study area is poor, which is understandable in view of the high possibility of frequent contacts with trees, shrubs and other vegetal disturbance in the work environment.

Assessment of the steering condition with respect to loss of power assistance indicated that 26.7% of the tractors were in good condition, 46.7% of the tractors had minor and rectifiable faults. Few (6.7%) of them were in bad condition, while 3.3% of the respondents did not indicate whether their tractors had challenges with power assistance or not. Results further show that majority (43.3%) of the tractors had no loose or damaged connection, the same percentage were

Fig. 3 Analysis of tractor component failure on good/bad basis Croat. j. for. eng. 36(2015)2

257

R.O. Akinbamowo and B.O. Akinnuli The Effects of Mechanised Log Skidding on some Components of Farm ... (253–258)

faulty but serviceable, while 6.7% had steering hoses that needed replacement. There was no assessment for 6.7%. 56.7% of the tractors had serviceable problems of low steering oil level, 26.7% were in good condition, while very few (3.3%) were in bad condition meaning that they had critical low oil level.

There was no response for 13.3% of them. Analysis also revealed that 40% of the sampled tractors had repairable faults of shaft assembly, drag link and tie rod. 26.7% of them were in good condition, while 10% needed a replacement of these components.

There was no response for 23.3% of tractors examined. Also, 43.3% of the tractors had oil leaking from steering ram, 23.3% of them were without this problem, while the status of 33.3% of the tractors was unknown. Based on these results and with respect to all the components related to steering system, it could be said that the status of majority of tractors in the study area is slightly below the average.

Analysing on the basis of mean percentages of major criteria (components and symptoms) and aggregating to strictly good and bad scores on tractor basis and ignoring undisclosed responses, the results (Fig. 3) indicated that the braking unit had a good/bad ratio of 0.72, the linkages 1.47, drawbar 1.13, traction devices 1.38, while the good/bad ratio of the lighting, general physical condition and steering was 0.13, 0.40 and 0.66, respectively.

other machines to enhance productivity. The current study has indicated that the use of conventional tractors without adaptation to forestry use is largely harmful to the machine, although the examiners could be subjective, and the quantity and uniformity of the assessment during the study were not always sufficient.

Further study on the cost of these failures might be required to assure tractor owners that using tractors this way is not the case of being »penny wise and pound foolish«.

4. Conclusion

McNamara, D., Kaufman, L.A., 1985: Can Horses compete with Tractors? Forestry note No 95, State of California The Resources Agency/Department of Forestry. Sacramento, CA 95814, 7 p.

The key to sustainable forest harvesting is to apply the best knowledge available in six critical areas: harvest planning, forest roads, felling, extraction, long distance transport and post-harvest assessment. Most of these operations include the use of tractors and

5. References Adekunle, V.A.J., Olagoke, A.O., 2010: The impacts of timber harvesting on residual trees and seedlings in a tropical rain forest ecosystem, South Western Nigeria. International Journal of Biodiversity Science, Ecosystem Services & Management 6(3): 131–138.

DAS, 2009: Digest of Agricultural Statistics Published by Research and statistics department Ministry of Economic Planning and Budget, Akure, 65 p.

Dolan, J.A., 2004: Motor vehicle technology and practical work. Heineman educational books, Oxford, 527 p.

Ezebilo, E.E., 2004: Threats to Sustainable forestry Development in Oyo State, Nigeria. Master thesis, Department of Southern Swedish Forest Research Centre Alnarp, Swedish University of Agricultural Sciences September, 42 p.

FAO, 2011: Estimates. www.fao.prg.forestry/harvesting Greacen, E.L., Sands, R., 1980: Compaction of forest soils: A review. Australian Journal of Soil Research 18(2): 163–189.

Oluka, S.I., 2000: Costs of tractor ownership under different management systems in Nigeria. Nigerian Journal of Technology 19(1): 15–28.

Authors’ address: Richard Omoyele Akinbamowo, PhD.* e-mail: akinba@yahoo.com Ministry of Agriculture Alagbaka, Akure Ondo State NIGERIA

Received: July 28, 2014 Accepted: March 15, 2015

258

Basil Oluwafemi Akinnuli, PhD. Federal University of Technology P.M.B. 704, Akure Ondo State NIGERIA * Corresponding author Croat. j. for. eng. 36(2015)2

Original scientific paper

Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico (a Case Study) Ciro Hernández-Díaz, Jesús Soto-Cervantes, Javier Corral-Rivas, Eusebio Montiel-Antuna, Ramón Alvarado, Rodolfo Goche-Télles Abstract The impacts of forest roads on soil were studied in a timber harvesting area of 92 ha in the municipality of San Dimas, state of Durango, Mexico. The area included 3127 m of main roads, 2907 m of secondary roads and 2979 m of tertiary roads. The timber was harvested at the beginning of 2012. After logging, soil loss by run-off during the rainy season was assessed along the truck ruts. This variable was correlated with the width and the longitudinal and transverse slopes of the road. The overall average road density (98 m/ha) indicated an average external yarding distance of 102 m. This is a short distance considering that timber was drawn with a jammer, which can pull logs over a distance of 300 m or more. Run-off in the rainy season decreased the ground level by between 38 and 58 mm along the truck ruts, and the soil loss was different in each type of road. The findings have led us to propose the elimination of some tertiary roads, to reduce the total road density to 78 m/ha. This is more than sufficient for logging, especially if the jammer capacity is improved, e.g., by applying the highlead system or the aerial yarding system with jammer. We estimate that soil loss would be reduced by 20% with the proposed changes to the road network. Additionally, the new road network would enable almost 20% of the area now occupied by roads to be reforested.

Keywords: yarding systems, road density, run-off, erosion

1. Introduction Rural roads are essential for the social and economic development of communities in mountainous or semi desert locations, where access to basic health services and education is very complicated (Mills 1997). These roads enable people to study, enjoy and live within wild areas and forests, and they facilitate leverage of resources and ecosystem services (Keller and Sherar 2004, Tolosana et al. 2000, Bruce et al.

2011, Deegen et al. 2011).

Forest roads also have major environmental impacts. Building forest roads involves removal of vegetation and soil, thus favoring run-off, pollution of streams, and the risk of erosion (Keller and Sherar 2004, Naghdi 2004). The ecosystem also becomes fragmented and weakened, wildlife habitats are altered (Gucinski 2012), and anthropogenic activities Croat. j. for. eng. 36(2015)2

that further increase the impacts on ecosystem impacts are generally favored (Dykstra and Heinrich 1996, Keller and Sherar 2004, FAO 2008).

The impacts may be minimal if the roads are designed and managed carefully, with the aim of protecting nature (Demir 2007). Therefore, the available equipment, machinery and technology should be optimally used to harvest wood products in the cutting areas, in order to build and maintain only those roads that are actually required (Hernández 1993).

Building forest roads is expensive, particularly in sites with steep topography and the presence of rock.

Investment is only made in constructing forest roads when the cost can be justified. However, the justification is usually based solely on financial issues or short term social interest, and sufficient attention is not given to the potential environmental impacts, which may

259

C. Hernández-Díaz et al. Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267)

Fig. 1 Jammer with one reel, which is the type commonly used in timber harvesting operations in Mexico reverse the balance of the initially perceived benefits in the medium and long term (López 2009).

In Mexico, timber harvesting requires efficient planning (López 2009) and also improved planning and design of forest roads (Daniluk 2002). In general, timber harvesting operations should not be analyzed separately, but following comprehensive, integrated plans (Hernández 1993). In addition, logging roads should be designed and laid out by competent engineers, who are aware of the need to minimize soil disturbance, to establish proper drainage systems and avoid crossing watercourses (Dykstra and Heinrich 1996, Álvarez et al. 2010, Tenerife 2011, Naghdi 2004).

One of the main parameters for evaluating a road network is its density, which indicates the length of road per unit area (Keller and Sherar 2004). In Europe, the average density of forest roads for ground skidding is about 25 m per hectare of forest (Dykstra and Heinrich 1996).

The machine most commonly used for timber harvesting in Mexico is the jammer »motogrúa« (Men-

260

doza 1997) (Fig. 1). This is a timber yarding machine that is usually equipped with just one cable winch, used to directly pull one log at a time (direct pulling system). The average external yarding distance with this system is almost never longer than 150 m, which implies an average road density >65 m/ha.

Use of a two drum jammer would enable application of the highlead yarding system and the skyline yarding system with jammer. The latter has already been successfully tested in Mexico, showing that it is possible to yard logs from distances of 400 m or more (Hernández et al. 2002), whereby the density of roads required could be reduced to around 25 m/ha, as in Europe.

The skyline yarding system with jammer can also be applied in areas where the highlead system is appropriate; it requires minimal additional investment and training and would reduce the density of roads needed in various harvesting areas by more than half (Hernández et al. 2002).

Owners and managers of timber harvesting companies should train operators and make maximum Croat. j. for. eng. 36(2015)2

Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267) C. Hernández-Díaz et al.

use of the available technology and capacity and reaching range of equipment and machines for log yarding (Dykstra and Heinrich 1997). This would reduce the need for roads, reducing environmental impacts and also increasing the area available for the production of trees and biodiversity conservation.

The objectives of this study were to evaluate some impacts on soil of the existing network of roads in a cutting area that was harvested in 2012 and to compare these with the impacts associated with the proposed road network.

2. Materials and methods 2.1 Location The study was conducted in a property known as La Trinidad (Lots Four and Five of Fraction ll), where

timber was harvested in 2012 (Fig. 2). The site is located between latitude 24°18’06” and 24°17’35” North and longitude 105°49’00” and 105°47’28” West, in the municipality of San Dimas, Durango, Mexico. The elevation in the area ranges from 2200 to 2800 meters above sea level.

2.2 Methods With the aid of a GPS navigator, the coordinates of the study area were plotted on a map, and the area in which timber harvesting had recently been carried out was calculated. Before the rainy season in 2012, the study area was inspected to establish the location of the roads and trails, to measure them and to plot them on a map of known scale. The contour lines were then added to the map (equidistance, 20 m) along with the slope and the current yarding distance, and also the

Fig. 2 Maps showing the location of the study area Croat. j. for. eng. 36(2015)2

261

C. Hernández-Díaz et al. Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267)

proposed distance for harvesting the logs, assuming that the yarding equipment was used to full capacity for the type of terrain.

In the roads under study, sections were marked every 50 m as possible sites for sampling points for placing stakes to enable measurement of erosion. Thirty-one (31) sampling points (50% of the possible points) were positioned on the main road, which is 3127 m long. The positions of all sampling points along the road were randomly selected, using the following system. Strips of paper were marked with numbers from 1 to 62 and placed in a bag. Paper strips (31) were removed from the bag at random, and the number on each paper was multiplied by 50 to establish the distance, in meters, from the initial reference point of intersection of the main road, for the location of each sample.

The same procedure was followed for selecting the position of sampling sites in secondary and tertiary roads. In Mexico, tertiary roads (brechas) are the simplest and cheapest but also the lowest quality roads used for timber harvesting; these roads do not have bridges, culverts, ditches or bed liners.

At each sampling point, a wooden stake was stuck as far as possible into the ground in the middle of the rolled tracks: it was assumed that vehicles would no longer pass along the road because timber harvesting was completed in that area. The stakes were used to mark and compare the level of the ground at the beginning and the end of the rainy season, to enable estimation of the amount of soil lost by run-off along the truck tracks. The width, gradient and cross slope of the road were also measured at each sampling point.

These data were used to estimate the area devoid of vegetation for each type of road and also to evaluate the possible effect of the slope on the magnitude of soil loss due to run-off. Soil loss per area equivalent to one hectare covered by truck tracks in each type of road, was estimated as follows: (1) SL = ( lb − la ) ×10.000

Where: SL soil loss in the truck tracks, m3/ha; (lb – la) difference in soil level measured before and after the rainy season, m.

Density and average external yarding distance were also estimated (assuming that yarding with jammer is always done from the bottom upwards in the direction of the slope):

262

RL

RD = A

(2)

AEYD =

10.000

RD

(3)

Where: RD road density, m/ha; RL total length of the road of interest in the cutting area, m; A cutting area, ha; AEYD average external yarding distance, m.

The exposed surface without vegetation per hectare was calculated as follows:

EA = RD × Aw

(4)

Where: EA area exposed by the type of road, m2/ha; RD density of the same kind of road, m/ha; Aw average width of the road, m.

2.3 Statistical analysis Nonparametric tests were used, as the data did not meet the basic conditions required for application of regression analysis. The Spearman’s correlation coefficient (rs), which indicates the degree of relationship that may exist between two variables, was calculated and expressed as follows (Mora 2008):

 n  rs = 1 −  6∑di 2  / n n2 − 1  1 

((

))

(5)

Where: rs magnitude of the correlation between the two variables, which can be direct (+) or indirect (–) and can acquire absolute values between zero and unity; di2 squared differences between each observation range of the two variables, when the observations are ordered with respect to each variable; n number of observations in the sample.

Furthermore, the non-parametric Kruskal Wallis test, also known as the H Test (Acuña 2013), was used to determine whether the magnitude of soil loss by runoff was the same for the three types of roads. This test enables comparison of independent samples and does not require that the data conform to a normal or any other distribution; it is based on comparison of the medians of the samples using the H statistic, expressed as follows:

 R12 R2  R2 12 H= + ∑ 2 + ∑ 3  − 3 ( N + 1) (6)

∑ n2 n3  N ( N + 1)  n1 Croat. j. for. eng. 36(2015)2

Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267) C. Hernández-Díaz et al.

Where: H test statistic; R1, R2, R3 ranges for each observation of each sample, obtained when considering together all observations in the three samples compared; n1, n2, n3 the size (number of observations) of each sample; N the sum of all the observations in the three samples.

3. Results and discussion The logging area under study covers 92 ha and includes 3127 m of main roads, 2907 m of secondary roads and 2979 m of tertiary roads. The average road density is 98 m/ha, which is equivalent to an average external yarding distance of 102 m from the road. This is considered a very short distance given the capacity and the potential reach of the jammer.

3.1 Yarding distance and current density of roads In some parts of the cutting area, the roads were separated by distances of more than 250 m and sometimes even more than 300 m. In these sites, it was not possible to pull up the logs with the jammer using the traditional system of »direct pull« with a single cable, as the maximum average range with this system is 150 m.

Comparison of the current road density (98 m/ha) with the technically feasible density for jammer, suggested as 33 m/ha or less (Dykstra and Heinrich 1996, Hernández et al. 2002) shows that the road density in the cutting area exceeds the cited parameter by three times. This may lead to negative environmental impacts such as loss of vegetation, reduced water infiltration and increased soil compaction, runoff, habitat fragmentation and pollution of streams and water bodies (Akbarimehr and Naghdi 2012, Grace and Clinton 2007).

3.2 Soil loss by run-off from the road The average values of parameters measured in random samples from each type of road are shown in Table 1. The decrease in the ground level corresponds to the difference in level at the beginning and the end of the rainy season in 2012. The width of each road was measured considering the area devoid of vegetation due to the passage of the road. This was measured from the start of the ravine on one side of the road to the area where soil removal was noticeable on the other side of the road. We noted that the road was wider in many areas due to truck tracks. The longitudinal and transverse slopes were measured from the center of the road at each sampling point.

Croat. j. for. eng. 36(2015)2

Table 1 Estimated average values of each variable, in relation to types of road Variable

Main roads Secondary roads Tertiary roads (n1 = 13) (n2 = 24) (n3 = 10)

Reduction of ground level, cm

5.8

4.1

3.8

Soil loss, m3/ha

580

410

380

Road width, m

4.5

5.0

4.5

Track total width, m

1.6

1.4

1.2

Longitudinal slope of road, %

12.9

10.5

6.2

Hillside slope, %

22.5

22.5

11.1

According to data recorded at the weather station closest to the study area (located at a distance of 40 km, La Victoria, San Dimas, Durango), the precipitation level between June and October 2012 was 620.4 mm.

Run-off and sediment delivery to water streams have been shown to be proportional to the amount and distribution of annual precipitation (Kahklen 2001), although confirmation of this is beyond the scope of this study.

3.3 Statistical analysis The values of the Spearman’s correlation coefficient (rs) for the relationship between the soil loss in each road type and the variables evaluated are shown in Table 2. When the correlations were significant (e.g. hillside slope on main roads [rs = 0.83]), it was concluded that the observed effects are not due to chance and that these variables are possible causes of soil loss.

Analysis of the data classified by road type revealed significant correlations in only three cases, particularly in secondary roads, for which slightly more Table 2 Spearman’s coefficients (rs) for correlations between soil loss and the variables evaluated. The values shown in bold type indicate significant correlations (p<0.05) Road type

Road width, m

Hillside slope, %

Road longitudinal slope, %

ns

0.83

ns

0.48

ns

0.66

ns

ns

ns

0.69

0.63

0.50

Main road (n1=13) Secondary road (n2=24) Tertiary road (n3=10) All roads (N=47) ns – not significant (p<0.05).

263

C. Hernández-Díaz et al. Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267)

observations (24) were made (Table 2). In relation to the tertiary roads, for which fewer observations were made, no significant relationship between soil loss and any of the explanatory variables was detected. However, when the data were analyzed together for all the three types of roads (last row), the three coefficients were found to be moderately significant. The relationship between soil loss and each of the three variables considered may have been clearer if larger samples had been available for each type of road.

The proposed closure of some tertiary roads relies on the assumption that by leaving only the main and secondary roads, the yarding distance would not be greater than 300 m. This distance is achievable with the jammer if used with the highlead system, especially if it is supported with other methods like using animals (horses or oxen) to pre pull the logs toward the jammer lane in moderately sloping areas.

The results of the Kruskal Wallis test indicate significant differences in soil loss between the three types of roads analyzed (p<0.05), as the calculated H value (375,792) was much higher than the tabulated H value (6221). This is consistent with the results of the Spearman test. It also confirms that the analyzed variables have an effect on the magnitude of soil loss on forest roads and that the magnitude differs depending on the characteristics of the road.

Another option is to apply the skyline yarding system with jammer, mentioned in the introduction. In this system, installation of the jammer in each workstation is slightly more complicated and time consuming than with the direct pull system. However, given that it extends the yarding distance by two or three times, a much larger area is drained from each road with the skyline system, which involves extraction of a larger volume of wood at each station. The use of this system also reduces the necessary effort by workers and risks involved. Since fewer roads are required, larger areas will be covered by vegetation, including timber trees, and runoff and the overall environmental impact will be reduced (Hernández and Alcázar 2003).

3.5 Proposals regarding yarding distance and road density

3.6 Soil loss associated with the proposed road network

The primary and secondary roads are planned to reach places outside the cutting area, and be used also for other purposes besides harvesting timber; therefore, proposals for amending these roads are beyond the scope of this study. However, on the basis of the results and the observed density of roads inside the study area, we suggest closing 1827 m of tertiary roads inside the cutting area, which would reduce road density from 98 m/ha to 78 m/ha. These roads represent 20% of the existing roads and their closure would significantly reduce the number and magnitude of environmental impacts.

Table 3 shows the estimated soil loss for each road type in the study area. The proposal involves closure of only some of the tertiary roads, and therefore the estimated soil loss is only reduced in this type of road.

3.4 Comparison of soil loss in relation to type of road

With the current length of tertiary roads (2979 m), soil loss is approximately 135.8 m3, and it is estimated that reduction of this type of road to only 1152 m would reduce soil loss to 52.5 m3, which indicates a reduction of 61.3% in soil loss in the rainy season associated with this type of road. It is also estimated that if only the primary and secondary roads were left, and by closing some of the tertiary roads, the total soil loss would be

Table 3 Estimated changes in soil parameters associated with the proposed changes to the road network in the study area Road type

Road length, m

Track width, m

Reduction of ground level, cm

* Total soil lost within both tracks, m3

** Estimated soil loss due to tracks, m3/ha

Main road

3127

1.6

5.8

290.2

580.0

Secondary road

2907

1.4

4.1

166.9

410.0

Tertiary road

1152

1.2

3.8

52.5

380.0

–

1.5

4.8

–

479.2

7186

–

–

509.6

–

Average Total

* Total amount of soil actually lost within both tracks along each road type ** Estimated soil loss per hectare of track area (not forest area)

264

Croat. j. for. eng. 36(2015)2

Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267) C. Hernández-Díaz et al.

Table 4 Area devoid of vegetation with the current road network and with the proposed network Current road network Road type

Proposed road network

Length,

Road width,

Area devoid of vegetation,

Length,

Road width,

Area devoid of vegetation,

m

m

m2

m

m

m2

Main road

3127

4.5

14,071

3127

4.5

14,071

Secondary road

2907

5.0

14,535

2907

5.0

14,535

Tertiary road

2979

4.5

13,405

1152

4.5

5184

Total

9013

14

42,011

7186

14

33,790

509.6 m3 rather than the 592.9 m3 currently lost in the rainy season; in other words, the global soil loss in the truck tracks would be reduced by more than 14%, due to the reduction in the density of forest roads.

3.7 Bare soil area due to roads Table 4 shows the area devoid of vegetation due to the presence of roads in the current network and the corresponding area after closure of the roads and reforestation of the sites. An area of 8221 m2 could be returned to forest production, which represents a reduction of 19.6% of area that is devoid of vegetation due to the presence of roads. This is important in relation to both the increased timber production and also the reduction in environmental impacts associated with deforestation.

The use of two drum jammers (one drum for the mainline and the other for the haulback line), like the ones used in Durango before 1975, yields a substantial improvement enabling a reduction in the area occupied by tertiary roads.

This machine could be used to apply the skyline yarding system with jammer and to harvest the logs from a distance of 400 m or more. It is also possible to pull logs from an average distance of 300 m (equivalent to an average road density of 33 m/ha) by only applying the highlead yarding system, which was used about 100 years ago when the jammer was initially brought to Mexico. Any of these two systems, with this road density, would enable harvesting all of the trees in that cutting area, which could not be reached with the current yarding system (direct pulling) given the current road network.

We recommend that future studies should use more sampling points for measuring the variables analyzed in this paper. Other independent variables that could contribute to explaining the magnitude of soil loss on forest roads should also be studied: a) degree of soil Croat. j. for. eng. 36(2015)2

compaction, b) presence of ditches, culverts and other roadworks, c) type of soil covering the road bed, d) intensity and frequency of rainfall, and e) the quantity and weight of trucks using the road, among others.

Medium and long-term studies should also be conducted to evaluate the effect of the intensity of annual precipitation. Finally, observations should also be made in several different timber harvesting areas, to enable comparisons and possible extrapolation of the results.

4. Conclusions The width and the longitudinal and transverse slopes of the roads affect soil loss by run-off via truck ruts. Although the unitary loss is different for each type of road, the overall loss increases with the number and length of roads in forestry areas. Here, we propose closing of some tertiary roads to decrease the total density of roads in the cutting area by 20%, with expected proportional reductions in environmental impacts. Moreover, almost 20% of the area currently occupied by these roads could be returned to timber production.

5. References Acuña, E., 2013: Pruebas no paramétricas. University of Puerto Rico. In: http://math.uprm/edu/~edgar 23 p.

Akbarimehr, M., Naghdi, R., 2012: Reducing erosion from forest roads and skid trails by management practices. Journal of Forest Science 58(4): 165–169. Álvarez, D., Betancourt, Y., Rodríguez, J.F., Pastor, J.F.B., Vallalba, M.J., Alaejos, J., Prades, C., Álvarez, E., Candano, F., 2010: Aprovechamiento forestal. University of Pinar del Rio, Cuba; University of Huelva, Spain; University of Cordoba, Spain. 147 p.

265

C. Hernández-Díaz et al. Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267) Bruce, J.C., Han, H.S., Akay, A.E., Chung, W., 2011: ACCEL: Spreadsheet-Based Cost Estimation for Forest Road Construction. Western Journal of Applied Forestry 26(4): 189–197.

Daniluk, G., 2002: Código de cosecha forestal uruguayo. Departamento Forestal. Facultad de Agronomía. World Forestry Institute. Montevideo. Uruguay. 89 p.

Deegen, P., Hostettler, M., Navarro, G.A., 2011: The Faustmann model as a model for a forestry of prices. European Journal of Forest Research 130(3): 353–368.

Demir, M., 2007: Impacts, management and functional planning criterion of forest road network system in Turkey.

Transportation Research Part A 41(1): 56–68.

Dykstra, D.P., Heinrich, R., 1996: Código modelo de prácticas de aprovechamiento forestal de la FAO. 86 p.

Dykstra, D.P., Heinrich, R., 1997: Forest harvesting and transport: Old problems, new solutions. Proceedings of the XI World Forestry Congress 13–22 October. Antalya, Turkey: 171–186.

FAO, 2008: Promoting responsible forest harvesting practices: www.fao.org/forestry/ harvesting/11835/ en: (Accessed April 14th 2012).

Grace, J.M. III., Clinton, B.D., 2007: Protecting soil and water in forest road management. American Society of Agricultural and Biological Engineers 50(5): 1579–1584.

Gucinski, H., Furniss, M.J., Ziemer, R.R. Brookes, M.H., (Editors) 2001: Forest roads: a synthesis of scientific information. Gen. Tech. Rep. PNW–GTR-509. Portland, OR: US Department of Agriculture, Forest Service, Pacific Northwest Research Station 103 p.

Hernández, D.J.C., 1993: Obtenga mejores rendimientos en la extracción de madera, combinando el derribo y el arrime de trocería. Tema didáctico Núm. 1. SARH–INIFAP– CIRNOC; Campo Experimental Valle del Guadiana. Durango, Mexico, 28 p.

Hernández, D.J.C., Alcázar, V.C., Unzueta, A.E., Sánchez, Q.A., 2002: Arrime de trocería combinando los sistemas de motogrúa y cable aéreo. Folleto científico Núm. 1. University Juarez of the State of Durango, Mexico, 38 p.

266

Hernández, D.J.C., Alcázar, V.C., 2003: Sistema combinado para el arrime de trocería, con motogrúa y cable aéreo. Folleto técnico Núm. 1. University Juarez of the State of Durango, Mexico, 36 p.

Kahklen, K., 2001: A method for measuring sediment production from forest roads. United States Department of Agriculture Forest Service Pacific Northwest Research Station Research Note PNW-RN-529.

Keller, G., Sherar, J., 2004: Ingeniería de caminos rurales; Guía de campo para las mejores prácticas de administración de caminos rurales. US Agency for International Development (USAID), in cooperation with USDA, Forest Service, International Programs and with Conservation Management Institute, Virginia Polytechnic Institute and State University, USA, 181 p.

López, S.P.M., 2009: Caracterización del abastecimiento forestal en el ejido El Brillante, P.N., Dgo. Tesis de licenciatura; Facultad de Ciencias Forestales; Universidad Juárez del Estado de Durango, Mexico, 89 p.

Mendoza, B.M.A., 1997: Rendimiento de un sistema de cable en relación a la intensidad del régimen silvícola. Madera y Bosques 3(1): 13–32.

Mills, K.A., 1997: Forest roads, drainage, and sediment delivery in the Kilchis River watershed. Oregon Department of Forestry, Tillamook District.

Mora, V.M., 2008: Fórmulas y tablas estadísticas. 1a. ed., 1a. reimpr.; Editorial UCR, San José C. R. 88 p.

Naghdi, R., 2004: Comparative study of tree length and cut to length logging method. PhD thesis College of Natural Resources, University of Tehran Iran, 320 p.

Tenerife-Área de Medio Ambiente, 2011: Circulación por pistas forestales con vehículos de motor. Consejero Insular del Área de Medio Ambiente; Resolución del del 14 de abril de 2011 que regula el tránsito motorizado por pistas forestales. España; 9 p. https://www.tenerife.es/.

Tolosana, E.E., González, G.V.M., Vignote, P.S., 2000: El Aprovechamiento maderero. Mundi-prensa: 419–438.

Croat. j. for. eng. 36(2015)2

Impacts of Forest Roads on Soil in a Timber Harvesting Area in Northwestern Mexico... (259–267) C. Hernández-Díaz et al.

Authors’ address – Adresa autorâ:

Received: July 08, 2014 Accepted: April 12, 2015 Croat. j. for. eng. 36(2015)2

Prof. Ciro Hernández-Díaz, PhD.* e-mail: jciroh@ujed.mx Prof. Javier Corral-Rivas, PhD. e-mail: jcorral@ujed.mx Ramón Alvarado, MSc. e-mail: alvonty@yahoo.com Universidad Juárez del Estado de Durango Institute of Silviculture and Wood Science Fracc. Ciudad Universitaria, Blvd. Guadiana 501 Durango 34120 MÉXICO Jesús Soto-Cervantes, mag. ing. silv. e-mail: jesus9003@hotmail.com Eusebio Montiel-Antuna, M.A. e-mail: e.montiel@ujed.mx Rodolfo Goche-Télles, PhD. e-mail: jgoche@ujed.mx Universidad Juárez del Estado de Durango Faculty of Forest Sciences Fracc. Ciudad Universitaria, Blvd. Durangoy Río Papaloapan 34120 MÉXICO * Corresponding author

267

Original scientific paper

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance of an Experimental Road Pavement József Péterfalvi, Péter Primusz, Gergely Markó, Balázs Kisfaludi, Miklós Kosztka Abstract Forest roads should be constructed to provide economic wood transport routes while causing minimal environmental impact. Therefore, the extended use of local materials (soil, stone) is essential. As cohesive soils cannot be drained by gravity and saturated cohesive soils have low bearing capacity, their use as a building material raises problems. This issue can be solved by lime stabilizing the soil. An experimental road was constructed to evaluate the effect of lime stabilized cohesive soil on the pavements built on top of it. Nine pavement versions were built on three different thickness (15, 25 and 35 cm) of lime stabilized soil. A traditional pavement without lime stabilization was also built for comparison. The bearing capacity of the stabilized layers and the finished pavements were calculated. The long term performance of the pavements was tested by measuring the effect of artificial traffic on their bearing capacity. Results showed that the bearing capacity modulus of the lime stabilization was around 500 MPa. 25–35 cm of lime stabilization under the pavements was necessary for good long term performance. 35 cm thickness of the stabilized local soil was enough to withstand the applied traffic without serious damage. Therefore, lime treated cohesive soil can be recommended as a subgrade layer in forestry conditions.

Keywords: forest opening up, lime-stabilization, road test, bearing capacity

1. Introduction Stabilized local soil can usually be applied as the subgrade course of a road pavement. Cohesive soils can be stabilized by lime. Between 1960 and 1970 approximately 53 km of forest roads were built on lime stabilized subgrade in South-Western Hungary. The economic and technological conditions of the following period were not conductive to the spreading of the lime stabilization method. As a result the stabilization experiments were neglected. Nowadays, the tightening economic situation, the increasing ecological requirements and the appearance of modern rotary mixers and binding material feeders have led to the rediscovery of stabilizing methods. For reasons of environmental protection, the use of local soil, advantageous as input of external material (e.g. crushed stone) onto the area, can be reduced. Lime as an adhesive is present in nature and the quantity of application is not significant.

Croat. j. for. eng. 36(2015)2

Current public road regulations in Hungary do not consider stabilized local soil as a bearing layer in the pavement design process. Stabilizations are considered only as soil improvement methods. It was hypothesized that stabilized local soil can act as a standalone pavement type for low volume roads. To evaluate the relevance of this hypothesis an experimental road was built. The first aim of the experimental road was to determine the bearing capacity of lime stabilized layers that can be taken into account in the design process of forest road pavements. The study of traffic resistance of pavements built on lime stabilized local soil was the second aim of the experimental road. By constructing one of the test sections with only lime stabilized soil, we have the opportunity to examine the possibilities of this method in the improvement of trafficability of dirt roads. Final purpose of the experimental road is to prove that the actual thickness of crushed stone layer in a pavement, built onto a lime stabilised local soil is reducible. If the vol-

269

J. Péterfalvi et al.

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

ume of crushed stone could be reduced, not only material costs would be lower, but transportation costs could be decreased as well.

1.1 Soil treatment with lime Lime has been used for a long time to improve mechanical properties of cohesive soils. By absorbing water, lime transforms to lime hydrate, while producing heat. If lime hydrate is mixed into the soil, the Ca++ ions bind to the surface of clay particles and they remove water and other ions from there. As a result, the plasticity index of the soil decreases significantly, and the soil becomes granular (Szendefy 2013). If enough lime is added, the pH increases so much that clay particles start to disintegrate. The resulting aluminium and silicon ions react with the calcium ions and together they form hydrates. These hydrates form a network that further increases the bearing capacity of the soil (NLA 2004). According to experiences, moist cohesive soils can be dried by the addition of 1–3 mass percentage of lime. In addition to the drying effect, the result of a 3–5 mass percentage lime treatment is significant increase in the bearing capacity, the shear strength and the optimum moisture content of the soil.

The quantity of the lime is determined by the properties of the soil to be treated and the lime itself. The current practice of the quantity determination is to conduct laboratory tests (Tárczy 2007).

Lime stabilized soils can be produced on-site or at a mixing plant. In case of on-site mixing, lime can be mixed right into the soil or soil can be excavated and mixed with lime. The first step of on-site mixing is to spread dry quicklime or dry hydrated lime on the surface of the soil to be treated in the predetermined amount. Application of quicklime is more favorable due to its higher lime content and faster reactions. The produced heat is also higher and, therefore, the construction period may be extended. The downside of quicklime is that it requires more water and it is more difficult to mix with soils than hydrated lime.

Lime application can be done by hand or by machines. Mechanical spreading is more efficient. The amount of applied lime can be regulated by the feeder opening and the speed of the feeder. Mixing can be done by a grader or by rotary mixing. If its capacity is appropriate, the use of a rotary mixer is more favorable since it achieves better mixing quality in less time. Water has to be added to the soil as needed during the mixing procedure. This may be done by water sprayer trucks or a rotary mixer with water supply capability.

Mixing can generally be carried out in one stage. In case of soils with high clay content, it is appropriate to mix in two stages. 24–72 hours should pass between the two

270

mixing stages. After the first stage, the subgrade has to be sealed by light rolling. In order to achieve appropriate bearing capacity, the stabilized soil needs compaction. Compaction can be performed several ways. The common practice is to use a sheep foot roller or a vibrating steel wheel roller followed by a rubber tired roller.

Compaction should be commenced shortly after mixing though a delay of four days is usually acceptable in humid conditions (Little 1995, Tárczy 2007).

1.2 Full scale road tests In order to determine the lime quantity and quality suitable for a given soil, laboratory tests should be conducted. Based on these tests, different soil and lime types can be compared. Though the best way to study the behavior of various pavements and bases on different soils is to study them on real life roads. Since the deterioration of roads is a slow process, full scale accelerated pavement testing has gained ground. For these tests, different pavement types are actually built and tests are conducted on them. The aim of the tests is usually to determine the reaction of different pavement types to a given amount of traffic load (Metcalf 1996).

The most widely known tests of this kind were the AASHO (American Association of State Highway Officials) Road Tests. These experiments were carried out in the USA between 1956 and 1962. The aim of these tests was to develop a methodology for highway pavement design. For this reason, 470 types of pavements were built on weak soil. Pavements consisted of sandy gravel, crushed stone and asphalt concrete courses.

The experimental road section was loaded with artificial traffic for two years. The condition of the pavements was evaluated regularly. Evaluation was carried out by visual deterioration assessment and the measurement of longitudinal road profile and central deflection. The most important result of the experiment was the determination of the connection between design parameters of pavements (equivalent thickness), axle load and configuration as well as the number of loads. The resulting functions proved to be useful in case of different types of soils and pavements (Highway Research Board 1961).

The experimental road »SERUL« (Laval University Road Experimental Site) was built in Canada specifically to study forest road pavements in cold conditions. Several pavements were built on the local soil.

The study of the effect of different soil types is made possible by a three meters deep concrete trough the experiments on this site focused on the effect of thawing frosting and of the width of tires (LeBel et al. 2000).

Behak conducted experiments to study the effects of lime stabilization in 2011. Two test sections were Croat. j. for. eng. 36(2015)2

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

built with lime applied in 3% and 5%, respectively.

Central deflection was measured and visual deterioration assessment was carried out before and after the passing of real but known amount of traffic. Central deflection was changed from 2.44 mm to 0.77 mm in 4 months after construction (Behak 2011).

In Hungary an experimental road was built to study pavements typical for agricultural roads. 72 types of pavements were built on sand. Pavements consisted of asphalt and cement concrete surfacing, mechanical, cement and blast furnace slag stabilizations and/or crushed stone base courses. The pavements were loaded with a total of 11 500 ESAL (Equivalent Standard Axle Load). After evaluating the initial state of the pavements, the effect of five loading periods was measured. Meteorological data (daily min. and max. temperature, precipitation, soil temperature), performance (central deflection, deflection bowl) and traffic ability (longitudinal profile, rutting, cracking, potholing and other surface failures) properties were determined. Based on these measurements, it was concluded that the 11 500 ESAL resulted in complete failure only on those sections where construction deficiency occurred. Clear decrease in the bearing capacity was evincible by central deflection measurements only after the last loading period. It was proven that the AASHO design formulas can be used in case of low volume roads, though they lead to slight overdesign. It was also concluded that the various stabilization methods are suitable for agricultural and forest roads (Kosztka 1989).

2. Materials and Methods A test road was built in 2006 to examine the behavior of pavements built on lime stabilized local cohesive soil. Experiments were conducted to determine the bearing capacity of the lime stabilized layers. Artificial traffic loaded the road between 2007 and 2008 and bearing capacity was monitored to evaluate the performance of the road. Since then, real life forestry traffic has been loading the road and control measurements are going to be done to evaluate the long term performance of the pavements.

J. Péterfalvi et al.

Þ liquid limit, 44.6%; Þ plastic limit, 22.4%; Þ plasticity index, 22.2%; Þ flow index, 18.4%; Þ maximum dry density, 1.82 g/cm3; Þ optimum water-content for compaction, 15%.

Precipitation conditions were analyzed using the data of the meteorological station of Bánokszentgyörgy. There were 94 precipitation days; for forest road building it means only a low number of days for constructing the subgrade. The situation is especially disadvantageous as there is a significant quantity of rainfall (694 mm/year), which slows and complicates drying. This problem is increasing with clay subgrade that cannot be desiccated by gravity. Saturated subgrades permanently lose their bearing capacity and, therefore, they cannot support pavement courses.

Among temperature conditions in the tested area, frost and thaw and their periodic alternation are of essential importance for road construction works. Frequent thaw frost effect can damage the pavement and thus can open a way to water into the pavement. Frost-thaw conditions can be characterized by these data: Þ winter temperature, +3,7 °C; Þ January temperature, –0,4 °C; Þ number of frost days, 100–110 days.

2.2 Designed pavements Different pavement versions were designed with about the same design bearing capacity. We used the AASHO Pavement Thickness Design Guide for it. This equivalent thickness was 30 ecm. Road sections with the same lime stabilization layer thicknesses were placed next to each other. The 15, 25 and 35 cm lime stabilization layer was built using local soil in the first 360 meters. Nine different pavements were built with asphalt concrete and crushed stone surfaces in different thickness, each 40 m in length. Three different pavement courses were designed on the lime stabilization layer: Þ well graded crushed dolomite course; Þ hot asphalt base course; Þ Finnish asphalt course.

2.1 Test site The 580 m long experimental road is located at the Forestry of Bánokszentgyörgy in Zala County, which belongs to the »Zalaerdő Forestry Company«.

The subgrade of the experimental road was constructed using the medium clay soil in place. The most important soil physical properties are as follows: Croat. j. for. eng. 36(2015)2

Fig. 1 Cross section template

271

J. Péterfalvi et al.

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

Fig. 2 Pavement versions The most important dimensions of the cross-section can be seen in Fig. 1. A traditional pavement, without lime stabilization, was built in the last 220 m as a control section. The control section was composed of sandy gravel and crushed stone layers built on the subgrade. To better understand the bearing capacity of the lime stabilization layer, the 5th pavement version was built with just 35 cm lime stabilization layer and with 2 cm of fine crushed stone. Fig. 2 shows the constructed pavements.

2.3 Lime stabilization bearing layer construction Constructing the lime stabilization layer was done as follows: Þp umping powdered quicklime into the binding material feeder; Þb ringing the binding material feeder to the starting point of construction; Þ c alibrating the binding material feeder; Þ c hecking the binding material feeder; Þ s preading the binding material; Þm ixing with a soil milling machine; Þ c ompacting with the vibration roller.

The binding material feeder spread the powdered quicklime on the whole surface of the subgrade in two passes. The spreading did not overlap; therefore, the spreading was in accordance with the plan. The soil milling machine performed the mixing in two courses with overlapping (Fig. 3). During the mixing, the machine set the thickness of the milling automatically.

272

Fig. 3 Soil milling machine The production line finished the lime stabilization with three different thicknesses and a width of 3.50 m in three hours.

2.4 Mechanical model of pavements The layers stabilized with lime, and the mechanical parameters of granular layers without binding material within the multi layered pavements, can be represented within the elastic pavement structure model shown in Fig. 4. At the bottom, the model of the completed subgrade corresponds to the calculation of single layer systems as elastic, homogeneous infinite half space; that is, our topic will be a mass of unlayered soil, bounded with Croat. j. for. eng. 36(2015)2

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

J. Péterfalvi et al.

2.4.1 Correspondences of infinite homogeneous half space In the applied model, the ground/subgrade is loaded with a p weight acting on a 2r=30 cm diameter rigid/flexible plate. The estimated calculation of shape changes caused by weight, regarding the elastic homogeneous half space, was worked out by Boussinesq in 1885. The deflection formula is the following (Yoder and Witczak 1975):

p E = c × 1 − m2 × r

(1) d Where: E elastic modulus of half space material, MPa; c Boussinesq plate factor (c=p/2 stiff and c=2,0 elastic); r radius of applied plate, mm; p greatest pressure applied, MPa; d vertical excursion under the plate, mm; μ Poisson ratio (–).

(

Fig. 4 Model of flexible pavements

)

In the formula above, μ stands for the Poisson ratio, the well known material parameter in elasticity studies. The value of this may vary between 0≤μ≤0.5. If the soil is considered as incompressible liquid (cohesive soil (clay)), it can be μ=0.5. In this way, the subgrade can be calculated as homogeneous elastic half space.

2.4.2 Solution for the two layer system

Fig. 5 Diagram by Burmister (1945) for two layer systems

The exact solution for the two layer system from mathematical mechanical point of view was given by Burmister in 1945. Later, he generalized his method for n layers, between 1954 and 1956. He also presented a diagram (Fig. 5) for the solution of the two layer system – to prevent difficult calculations. To calculate the centre bending, he applied the formula used for single layer systems (Burmister 1945):

horizontal plane, its dimensions are infinite horizontally and in depth. There are layers of H2 thickness stabilized with lime over the subgrade along the experimental road sections; over them, there are granular (i.e. macadam) base layers of H3 thickness without bonding.

Over all these, there are asphalt concrete layers of H4 thickness altogether. The layers are described by the E1, E2, E3, E4 moduli and the μ1, μ2, μ3, μ4 Poisson’s ratio numbers (Huang 2003). The load is a ca. 2r =30 cm diameter circle shaped, steady load coming from a 50 kN weight, which nearly equals the tire pressure of big trucks p=0.7 MPa. More details about analytical sizing of elastic pavements can be read in the work of Bocz et al. (2009).

Croat. j. for. eng. 36(2015)2

(

d = 2 × 1 − m2

) pE× r × F

(2)

1

In the formula, the modulus of the lower layer (the half space) is taken into account, which is then multiplied with an F deflection factor, defined on the basis of the known h/r and E2/E1 rates. The deflection factor equals the fraction,

F=

E1

Ee

(3)

Where: Ee equivalent surface modulus (Nemesdy 1985).

It is important that the surface modulus is not a layer modulus, but an average parameter typical of all

273

J. Péterfalvi et al.

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

the layers together coming from the measurements done on the surface of the multilayer systems (Papagiannakis and Masad 2008).

2.4.3 Defining the modulus of granular layers The modulus of the layers consisting of round grains without bonding material depends on the modulus of the layer below them. One of the oldest solutions is applied in the Shell sizing manual (Claussen et al. 1977):

E2 = E1 × 0.2 × H 20 ,45

(4)

Where: H2 thickness of the macadam or granular layer, mm; E1 modulus of the bottom layer, MPa. The defect of this correspondence is that it does not make any quality difference between the crushed stone and the round grain courses (Nemesdy 1991).

This defect is then covered by the significant work of Barker et al. (1977). They studied the thicker granular base layers within the elastic pavements, making differences between the macadam and the mechanical stabilization. It can be applied to crushed stone foundations with the following formula (Barker et al. 1977):

(

E2 = E1 × 1 + 10.52 × log ( H 2 ) − 2.10 × log ( E1 ) × log ( H 2 )

)

(5)

To gravel foundations, mechanical stabilization:

(

E2 = E1 × 1 + 7.18 × log ( H 2 ) − 1.56 × log ( E1 ) × log ( H 2 )

)

(6)

The modulus of the sandy gravel layers can also be estimated with this correspondence. When applying these correspondences, it is important to know that the E moduli are given in psi (pounds per square inch), the H layer thicknesses are given in inch in the original study (1 psi=0.006894 MPa and 1 inch=2.54 cm).

2.5 Static and dynamic load bearing capacity measurement The multilayer mechanical model of the elastic pavements requires each layer to be taken into account with the thickness and modulus values typical of them. The load bearing capacity modulus of the subgrade and the pavement courses can be defined with the help of static and dynamic field measuring tools.

The static modulus can be measured on the completed subgrade, as the load bearing capacity modulus of the bottom layer is considered as an elastic, homogeneous and isotropic half space. The plate bearing test is a test in which a load is applied in increments to the

274

soil using a circular loading plate and a loading device, released in decrements and the entire process is repeated. During the experimental process, a 30 cm diameter elastic plate is loaded with e.g. a hydraulic jack and truck balance. The load is applied in increments, waiting for the consolidation time, while the deformation caused by the load is being measured. Reaching the p=0.40 MPa pressure, the plate is unloaded, then the process is performed again. The pressure deformation curves can now be plotted using the data of the two load processes. The static modulus of the subgrade is calculated on the basis of the second load plate process (this is why number 2 is shown in the subscript) using the no. (1) deflection formula. Seeing the geometric dimensions of the plate applied at the E2 static experiment, a ca. 75 cm thick upper layer can be studied (v ≤ 2.5 ´D » 75 cm). Doing the calculations, rigid plate model and material dependent Poisson ratio (clay μ=0.5, limed soil μ=0.3) were taken into account. The disadvantage of the static experiments is that they are very time consuming and not properly modeling the moving wheel stress of trucks. These disadvantages are intended to eliminate using the dynamic tools.

The dynamic weight measurement method records the shape change caused by an 18±2 milliseconds load of a ca. 10 kg weight dropped from 70–75 cm height.

The weight is the same as the weight applied in the static measurement. The experiment models the material behavior typical of dynamic weight, since consolidation cannot take place in such a short period of time.

The dynamic load bearing capacity (Ed) method is capable to study base course or subgrade with at most 63 mm biggest grains, and thick at most twice the plate diameter (30 cm) (Subert 2005). In the case of the B&C (Bearing Capacity&Compaction Rate Tester) small plate light drop weight device, the dynamic effect is the 16 cm distance covered by a truck driving at ca.

35 km/h. The dynamic load bearing capacity modulus of the subgrade was defined at five points on each experimental road section. Four measurements were performed in the line of the expected tracks (two of them at the beginning, and two at the end of the section), and one measurement was made in the middle of the section. At the place of the measurement, preload was performed with three drops, then three measuring drops. The average values of each measuring points were defined with the help of the three measuring series (in triangle formation). This means 5 ´3=15 measurements per experimental sections. During the calculations, rigid plate model and variable Poisson ratio were applied: the expected value was μ=0.5 in the case of clay, and μ=0.3 in the case of soil mixed with lime.

The resulted modulus values were given as integers.

Croat. j. for. eng. 36(2015)2

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

J. Péterfalvi et al.

Fig. 6 Characteristic vehicle of artificial traffic

Fig. 7 Wheel load scaling

2.6 Measuring the central deflection with Benkelman beam

MAN, Fig. 6). Axle configuration, and loaded and unloaded axle loads were registered for each truck type.

Measurements of axle load were carried out by wheel load scales (Fig. 7).

The bearing capacity of a pavement can be characterized by the rate of elastic deformation caused by load. The method based on the measurement of central deflection is widely used. It generally provides good and reliable results. The deflection is measured with a Benkelman beam of 2:1 measurement probe support beam ratio. The tip of the measurement probe is placed under the dual tire of a loaded truck, where the maximum deflection is expected. Central maximum deflection is the elastic deformation measured with the beam converted to 50 kN wheel load. It is assumed that the conversion is linear. Deflection measurement was developed specifically for asphalt pavements, however we used it to characterize the unsurfaced lime stabilized course as well for information purposes. According to the experiences gathered so far, the lime treated layers work together sufficiently and they act as a flexible base course. In case of the crushed stone pavement types, Benkelman deflection measurements were not carried out.

The measurement of central maximum deflection was performed before and after each loading period.

Therefore, the changes caused by the traffic load became comparable to the initial state. Measurements were carried out on the left and right wheel track in every five meters. This density of data proved to be enough for statistical analysis.

2.7 Artificial traffic The reactions of different types of pavements to loads were studied by the application of artificial traffic. Traffic was generated by trucks characteristic for the Hungarian forestry companies (IFA, KAMAZ, and Croat. j. for. eng. 36(2015)2

The precision of the measurements was ±50 kg. The experimental road was loaded intensively. The number of passing trucks was registered. Different axle loads were converted to 100 kN ESAL by the use of equivalent axle load factors. Based on the weight of timber to be transported and the type of trucks used, traffic (T100) could be calculated. The number of ESALs on the experimental road is shown in Fig. 8. During the test period, 2100 ESAL artificial traffic was applied.

Since loading was not continuous, the intensity of traffic was high during the loading periods.

3. Results and discussion 3.1 Bearing capacity modulus of the subgrade As to all ten experimental sections static E2 values were tried to be defined. The measurements were made by turns at the centre of every section, and left and right from it, in the line of the expected tracks. Unfortunately, on the wet subgrade the measurement could be successfully performed at only two sections. Though these results agreed quite well with the measurements performed in advance, an average of 10 MPa bearing capacity value was defined on the subgrade.

3.2 Evaluation of the static and dynamic load bearing capacity measurements After building the lime stabilization, the increase of bearing capacity was studied with two types of model effect (static, dynamic). The data series of the

275

J. Péterfalvi et al.

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

Fig. 8 Artificial traffic on the experimental section Table 1 Values of the moduli determined by plate bearing tests Ver.

Thickness of lime stabilization, cm

01

Earthwork

Lime, 0 h

Lime, 24 h

Lime, 48 h

25

–

–

34

45

02

25

–

–

53

65

03

25

–

–

53

53

04

35

–

–

53

69

05

35

9

–

50

54

06

35

12

–

41

50

07

15

–

19

27

38

08

15

–

36

39

43

09

15

–

–

37

45

10

None

10

–

–

–

MEAN

10

28

43

51

two types of measurements were intended to define the static modulus necessary for planning, since the current technical regulation contains specifications about this. Apart from the lime stabilization layers, the load bearing capacity changes of the pavements built upon the lime stabilization layers were also studied.

Static (E2) and dynamic (Ed) load bearing capacity values were defined as mentioned above. In pavement versions No. 2 and 7 (Asphalt concrete (AC 22) bind-

276

Bearing capacity E2, MPa

ing course with large shear strength), the B&C light drop weight device did not show any excursions, so it was not possible to perform the measurement. During the calculations, the Poisson ratio was μ=0.5. The measurement results are shown in Table 1 and 2.

According to the professional literature, the dynamic measurement modulus does not correspond to the static measurement load bearing capacity modulus nor to the soil, so a general correspondence cannot be Croat. j. for. eng. 36(2015)2

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

J. Péterfalvi et al.

Table 2 Values of the moduli determined by dynamic bearing tests Ver.

Thickness of lime stabilization, cm

01

Bearing capacity Ed, MPa Subgrade

Lime, 0 h

Lime, 24 h

Lime, 48 h

Lime, 120 h

25

11

24

34

39

50

02

25

12

41

55

60

66

03

25

15

36

55

62

74

04

35

12

52

68

72

74

05

35

10

24

47

52

54

06

35

9

24

40

45

50

07

15

8

26

35

43

55

08

15

10

32

46

51

65

09

15

11

22

33

36

42

10

None

10

–

–

–

–

MEAN

11

31

46

51

59

given (Subert 2005). This topic is detailed in Tompai’s essay (2008). Parallel measurements are needed to define the threshold limit values typical of the type of material and the conditions on every type of soil, granular pavement layer, in every case.

The measurements performed on the lime stabilization showed that the load bearing capacity values

defined with different model effects are nearly the same. Fig. 9 clearly shows that both the 24 hours and the 48 hours static (E2), and the average dynamic modulus (Ed) values show similar results. Therefore, the load bearing capacity values defined with dynamic model effect can be considered equal to the static E2 needed for planning.

The increase of load bearing capacity of the lime stabilization was followed up through five days with the B&C device. The measurement results are summarized in Fig. 10. According to the measurements, the lime stabilization layer thickness and the increase of load bearing capacity do not correspond. The soil lime reaction is a quite complex process, and there are several factors that affect the rate of load bearing capacity that are not yet revealed. One of the most probable cause is the high heterogeneity of the soil water content, which can have a significant effect on the reaction of the soil lime mixture. In spite of all this, it can clearly be seen that the highest increase of bearing capacity was produced in the 25 cm and 35 cm thick lime stabilization experimental sections. During building, the lime stabilization was followed by the spreading and compaction of the crushed stone course, on the surface of which, the load bearing capacity was also measured. The studies performed on the surface of the crushed stone courses showed that the dynamic moduli are ca. twice as much as the static ones.

Fig. 9 Values of dynamic and static bearing moduli measured on the same sites Croat. j. for. eng. 36(2015)2

The static and dynamic load bearing capacity moduli measured parallel on the surface of the completed

277

J. PĂŠterfalvi et al.

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269â&#x20AC;&#x201C;282)

pavements do not show correspondence with each other at all. The main cause may be that, while the static load bearing capacity measurement can study a 75 cm thick layer, the detection range of the light drop weight measurement is only 30 cm. Therefore, the dynamic measurement is not suitable to study the pavements built of different layers, since it cannot detect the favorable and unfavorable features of the subgrade.

3.3 Bearing capacity modulus of lime stabilization layers The results of the demonstrated static and dynamic load bearing capacity measurements show that the modulus of the lime treated soil layers can be deduced.

The built subgrade and the lime stabilization together make a two layer system. Since the load bearing capacity modulus of the cohesive clay subgrade is known (10 MPa), and also the H thickness of the lime treated layers and the (Ee) value of the surface moduli measured on their surface, the F deflection factor can be calculated using correspondence No. (3). According to the theory of Burmister (1945), the modulus of the H thickness lime treated layer can be defined knowing these parameters (Fig. 5). After the calculations, the modulus of the lime stabilization layers resulted as Elime=500 MPa, which is almost the same as the modulus of a well compacted continuous grain distribution

macadam layer. According to this result, the lime treated soil layers can be counted as pavement layers, further on.

It is also possible to originate this two layer system in a single layer system, which behaves similarly to the original one, under the same conditions. According to the current public road design regulations, the lime stabilization is counted as improved subgrade, and not as a part of the pavement. In this case, the modulus of the improved subgrade is the equivalent surface modulus (Ee) defined on the surface of the two layer system. According to the field measurements, this value was defined to three thickness groups: Hlime=15 cm Eis.=40 MPa Hlime=25 cm Eis.=50 MPa Hlime=35 cm Eis.=60 MPa Where: Hlime thickness of the lime-treated layer; Eis. modulus of the improved subgrade.

3.4 Load bearing capacity modulus of granular layers The moduli of the granular layers were estimated with the correspondences by Barker et al. (1977), described in point 2.5.3. The lime stabilization layers, as improved subgrade, were introduced into the model.

Fig. 10 Increase of bearing capacity of lime stabilized layers as a function of time

278

Croat. j. for. eng. 36(2015)2

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269â&#x20AC;&#x201C;282)

J. PĂŠterfalvi et al.

Table 3 E moduli of granular layers by pavement versions Ver.

Subgrade

Granular layers

E, MPa

H, cm

E, MPa

1

50

25

170

2

50

15

120

3

50

15

120

4

60

0

0

5

60

2

0

6

60

15

160

7

40

15

120

8

40

15

120

9

40

30

150

10

10

45

60

The results of the calculations are summarized in Table 3. It can clearly be seen that, in the case of the control section, the moduli of the granular layers were the lowest because of inappropriate compaction of the subgrade. At ver. 9 built with the same total thickness (15 cm lime stabilization), the crushed stone layer modulus was 150 MPa, which is more than twice as much as the traditional solution (15 cm sandy gravel).

Where the thickness of the lime stabilization is 35 cm, the load bearing capacity modulus of the improved subgrade equals the load bearing capacity of the entire pavement control section.

more or less deformed due to the artificial traffic. The main goal of the Benkelman beam measurements was to determine the deterioration curve of pavements.

The central deflection changing was linear compared to the traffic. As a result of one year resting, the pavements of all versions regenerated. The regeneration is clearly visible on the central deflection data (Fig. 11).

On pavements with 25 cm and 35 cm thick stabilized courses, deflection values are similar, while higher deflection were measured on the ones with 15 cm stabilized courses.

3.6 The effect of traffic on experimental road

A certain part of the evolved deformations does not correspond to the size of the traffic, but was caused by technological problems. The deterioration of the pavement versions started in 2008. The pavement of the control section (ver. 10) was completely deteriorated on the whole length, the 15 cm thick sandy gravel and the 30 cm thick crushed stone layer above it entirely mixed with each other. This huge deformation was already experienced at T100=700 ESAL. The reason was that the sandy gravel and the crushed stone courses were not properly compacted. As an impact of the traffic, under the improperly compacted pavement, the soil failed. Beside the control section, there were three pavement versions where huge deformations evolved not connected to the size of traffic.

According to the deflection measurements, the followings can be stated about the sections after the evaluation of the measurement results and fieldwork.

Generally, the experimental sections performed well under traffic load. The experimental pavements got

On the first Finnish asphalt section (ver. 3), a big rut evolved along the right side track, which is probably just a local failure. This was confirmed by the statistic process of the deflection measurements.

The failure was also predicted during the graphical

3.5 Cost of pavements The estimated bearing capacities of pavement ver.

6 and traditional pavement (the control section) are similar. Therefore, their construction costs could be compared. Both pavement types were built with the same width, therefore their construction costs were calculated per linear meter. The costs are shown in Table 4 on the price level of 2006. Based on this data, it can be stated that, under the given circumstances, the pavement with lime-stabilization was half as expensive as the traditional one.

Croat. j. for. eng. 36(2015)2

279

J. Péterfalvi et al.

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

Table 4 Cost of traditional pavement and a pavement with lime stabilization Traditional pavement

Pavement with lime stabilization

Course thickness

Material

Cost, €/lm

Course thickness

Material

Cost, €/lm

15

Sandy gravel

11.9

35

Lime stabilization

9.9

35

Crushed stone

27.7

15

Crushed stone

11.9

50

Total

39.6

50

Total

21.8

Fig. 11 Deterioration curves of experimental pavements based on central deflection data process of the deflection measurements, and later, the failure evolved as a result of the experimental traffic. On the second AC 22 (ver. 7, Fig. 12) section and on the ver. 8 Finnish asphalt section (Fig. 13), there were big and long deformations along the left side track.

The reason is probably a technological problem, again. A part of the crushed stone layer under the asphalt layer, because of the improper side support, was

pushed sidelong, and therefore, the asphalt layer got damaged under the large bending strain. The left roadside is narrower than the right one, and the water running away from the asphalt was leaking in at the roadside in greater amount. Exploring the pavement, it became clear that the lime stabilization layer also »broke« under the experimental traffic. The macadam pavements built on lime stabilization were all in good condition after the traffic (Fig. 14).

Fig. 12 Left rutting on AC 22 (Ver. 7)

Fig. 13 Left rutting on Finnish asphalt (Ver. 8)

280

Croat. j. for. eng. 36(2015)2

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

J. Péterfalvi et al.

transportation and construction costs as well as environmental impacts can be reduced. In order to get satisfactory performance from the pavements built on cohesive soil, 25–35 cm of lime stabilized subgrade should be designed to support them. For low traffic roads, 35 cm of stabilized local soil with a thin (15 cm) crushed stone surfacing could be enough to provide trafficability. 25–35 cm thickness can be constructed in one stage of mixing if the constructing machine is appropriate. For mixing, it is necessary to apply a special rotary mixer machine, or a mixing adapter attached to an agricultural tractor with 150 kW output power.

Fig. 14 Macadam pavement built on lime stabilization (Ver. 6)

Based on the experiences gained from the road tests, a 6 km long forest road was built with stabilized local cohesive soil as subgrade. It was confirmed that pavements built on stabilized cohesive soil are durable only if appropriate drainage is provided.

Acknowledgements

Fig. 15 Unsurfaced lime stabilization (Ver. 5) The unsurfaced lime stabilization (ver. 5, Fig. 15) endured the traffic very well so far; the tires erased a 1 cm thick layer from the surface, therefore, a certain depth of rut evolved, though far from the level it could be objected. Bigger cracks could be observed on the surface, as recorded on photos.

4. Conclusions A test road was built to examine the effects of lime stabilized soil as subgrade. As expected, the applied lime increased the bearing capacity of the original cohesive soil. When designing a pavement with lime stabilized medium clay subgrade, it can be taken into account with a 500 MPa layer modulus. Doing so, the required bearing capacity can be reached with smaller amount of additional building materials. Therefore, Croat. j. for. eng. 36(2015)2

The experimental road was built as collaboration between the Regional Knowledge Centre of Forest and Wood Utilization (ERFARET), the Institute of Geomatics and Civil Engineering of the University of West Hungary, the Carmeuse Hungary Ltd. and the Zalaerdő Forestry Company. Research work of Péter Primusz was supported by the European Union and the State of Hungary, co-financed by the European Social Fund in the framework of TÁMOP-4.2.4.A/2-11/1-2012-0001 »National Excellence Program«. Previous research for this research project: Péter Pázmány programme (RET03/2004), 1.3. Technical development of forest management, Development of forest opening up networks.

The project completion date: 2006–2008.

5. References Barker, W.R., Brabston, W.N., Chou, Y.T., 1977: A General System for the Structural Design of Flexible Pavements. Proceedings of the Fourth International Conference on the Structural Design of Asphalt Pavements, Ann Arbor: 209–248.

Behak, L., 2011: Performance of full-scale test section of lowvolume road with reinforcing base layer of soil-lime. Transportation Research Record, Journal of the Transportation Research Board 2204: 158–164.

Bocz, P., Devecseri, G., Fi, I., És Pethő, L., 2009: Pályaszerkezetek analitikus méretezése. Közlekedésépítési szemle 59(5): 8–22.

Burmister, D.M., 1945: The General Theory of Stresses and Displacements in Layered Soil Systems. Journal of Applied Physics 16(2): 89–94.

Claussen, A.I.M., Edwards, J.M., Sommer, P., Ugé, P., 1977: Asphalt Pavement Design. The Shell Method. Proceedings of

281

J. Péterfalvi et al.

Evaluation of the Effect of Lime-Stabilized Subgrade on the Performance ... (269–282)

the Fourth International Conference on the Structural Design of Asphalt Pavements, Vol. I, Ann Arbor: 39–74.

Highway Research Board 1961: The AASHO Road Test, Highway Research Board Special Report 61A, National Academy of Sciences: Washington, D.C. 56 p.

Huang, Y.H., 2003: Pavement Analysis and Design, Second Edition, Prentice Hall, ISBN-13: 9780131424739, 792 p.

Kosztka, M., 1989: A Makk-pusztai kísérleti úton végzett megfigyelések a vékony útpálya-szerkezetek tönkremenetelének folyamatáról. Erdészeti és Faipari Tudományos Közlemények 2: 25–36.

LeBel, L., Doré, G., Provencher, Y., 2000: Laval University’s full-scale experimental site for construction and maintenance of forest roads. Proceedings of the COFE-CWF Conference.

11–14 September, Kelowna, British Columbia, Canada.

Little, D.N., 1995: Handbook for stabilization of pavement subgrades and base courses with lime. Lime Association of Texas, USA. Metcalf, J.B., 1996: NCHRP Synthesis of highway practice: Application of full-scale accelerated pavement testing. TRB, National Research Council, Washington D.C., USA. National Lime Association (NLA) 2004: Lime-treated soil construction manual. National Lime Association, USA.

Nemesdy, E., 1985: Útpályaszerkezetek méretezésének és anyagállandó-vizsgálatainak mechanikai alapjai. Kutatási részjelentés I., BME Útépítési Tanszék, Budapest.

Nemesdy, E., 1991: A zúzottkőalapok és kavicsalapok szerepe és hatékonysága az új út-pályaszerkezetekben. Közlekedésépítés- és Mélyépítéstudományi Szemle 41(7): 241–253.

Papagiannakis, A.T., Masad, E.A., 2008: Pavement Design and Materials, Wiley&Sons, Hoboken NJ, ISBN-10: 0471214612, ISBN-13: 978-0471214618, 552 p.

Subert, I., 2005: A dinamikus tömörség- és teherbírásmérés újabb paraméterei és a modulusok átszámíthatósági kérdése.

Közúti és mélyépítési szemle 55(1): 5.

Szendefy, J., 2013: Impact of the soil-stabilization with lime.

Proc. of the 18th International Conf. of ISSMGE Paris: 2061– 2064.

Tárczy, L., 2007: Meszes talajkezelés. Közúti és mélyépítési szemle (2): 26–28.

Tompai, Z., 2008: Földművek és kötőanyag nélküli alaprétegek teherbírásának és tömörsé-gének ellenőrzése könnyű ejtősúlyos módszerekkel. Budapesti Műszaki és Gazdaságtudományi Egyetem, Építőmérnöki Kar, Ph.D. theses, Budapest, 162 p.

Yoder, E.J., Witczak, M.W., 1975: Principles of Pavement Design, Second Edition, John Wiley & Sons, Inc. ISBN: 9780471977803, 711 p.

Authors’ address – Adresa autorâ:

Received: April 30, 2014 Accepted: October 06, 2014

282

Assoc. prof. József Péterfalvi, PhD.* e-mail: peterfalvi.jozsef@emk.nyme.hu Assist. prof. Péter Primusz, PhD. e-mail: primusz.peter@emk.nyme.hu Assoc. prof. Gergely Markó, PhD. e-mail: gergely.marko@gmail.com Balázs Kisfaludi, MsC. e-mail: kisfaludi.balazs@emk.nyme.hu Prof. emerit. Miklós Kosztka, CSc. e-mail: kosztka.miklos@emk.nyme.hu University of West Hungary Faculty of Forestry Department of Forest Opening Up Institute of Geomatics and Civil Engineering Bajcsy-Zsilinszky u. 4.

Sopron, 9400 HUNGARY * Corresponding author Croat. j. for. eng. 36(2015)2

Original scientific paper

Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood Johanna Routa, Marja Kolström, Johanna Ruotsalainen, Lauri Sikanen Abstract Moisture is the most important factor influencing the quality and calorific value of fuel wood.

Drying models for estimating the optimal storage time based on average moisture change in fuel wood stacks stored outdoors have been developed for different stem wood piles. Models are an easy option for making an estimate of the moisture content of an energy wood pile if compared with sampling and measuring the moisture of samples. In this study, stem wood models were validated against data from forest companies. Fourteen reference piles of covered pine stem wood and 8 piles of uncovered pine stem wood were studied. The results of the validation are promising. The difference between the measured and modelled moisture was on average only 0.3% with covered piles and 2.5% with uncovered piles. The models presented can be implemented in every location in Finland, because the Finnish Meteorological Institute has a database for interpolated meteorological observations covering the whole country in a 10 km x 10 km grid. For international use, model parameters need to be estimated case by case, but it should also be possible to implement the approach itself worldwide.

Keywords: energy wood, quality, storing, natural drying, model validation

1. Introduction Thinning is a harvesting method mostly used in Europe and within plantations all over the world.

Thinning wood is a typical raw material for the pulping industry as well as for energy and biorefining in the Nordic countries. Especially in Finland, thinning wood from young stands has been increasingly used for energy. In 2010–2013 it was the major source of forest chips for energy (Torvelainen et al. 2014).

Increased use of forest biomass for energy and rising transportation costs are forcing biomass suppliers towards better moisture content (MC) management in the supply chain. Biomass fuel quality is often defined by the calorific value, and lower moisture content results in increasing calorific value (Hartmann and Kaltschmidt 2001, Stokes et al. 1987).

Natural drying is used to reduce the moisture content of energy wood after cutting and during storage.

Croat. j. for. eng. 36(2015)2

Storing time at the roadside depends on the need for energy wood. The supply of energy wood operates year round, but the demand is notable from October to March (Andersson et al. 2002).

After tree cutting, wood starts to react with the surrounding microclimate (Routa et al. 2015). In Nordic conditions, the moisture content of wood drops rapidly in the spring. In late August and September, evaporation usually decreases, and the moisture content of the wood increases. In some cases, it can even be higher than the »green« moisture right after cutting. Maximizing natural drying and minimizing re-moistening are key elements in the quality management of energy wood (Routa et al.

2012). The timing of the operations in relation to the seasons is crucial in order to maximize the quality and monetary value of the energy wood.

In natural drying, the weather conditions are a very important part of the drying process. The most

283

J. Routa et al. Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291)

important parameters are evaporation, precipitation, humidity, temperature, solar radiation and wind conditions (Routa et al. 2015). In addition, the material, size and shape of the energy wood pile and the location of the pile, also affect the drying process.

The latest research methodology for monitoring moisture change has been the constant weighing of piles in racks built on load cells (Erber et al. 2012, 2014).This methodology allows moisture changes to be monitored in much more detail than previous sampling methods. The method also gives the moisture of the whole pile, which is challenging in determening the use of sampling methods (Röser et al. 2011).

Measurements can be taken automatically and as often as needed. This also enables an exact investigation of the effect of weather on energy wood storage and its moisture content.

In the 1980s, the first ideas of using models to predict the moisture content of wood were presented.

Stokes et al. (1987) published their models for soft and hardwoods in south eastern USA. Also, for example, Liang et al. (1996), Gigler et al. (2000), Filbakk et al.

(2011), Murphy et al. (2012), Erber et al. (2012) and Dong-Wook and Murphy (2013) have developed different drying models for different species. All approaches to fuel wood moisture content modelling have one common target variable: moisture contents or rather the alteration in moisture content during a specific period. The alteration can be explained by a large variety of explanatory variables, such as meteorological variables, parameters of storing, material type and duration of storage. Today’s practice is to measure the moisture content of wood chips when they arrive at the heating plant, but for efficient planning of operations the information is available too late. With the prediction models, it is possible to have an estimation of the moisture content of fuel in advance and plan supply operations so that the fuel is transported to the heating plant in a timely manner.

The objective of this study was to develop a model to forecast the energy wood moisture content and validate the model. Changes in moisture content were linked to weather conditions and microclimate.

The model should be easy to apply to the planning systems in operational stage and that is why the model should be quite simple and quick to calculate.

To verify fuel wood drying models, reference piles are a good option. Samples must be taken from the piles, which should consist of similar material as regards assortment and tree species. The models developed were validated against real life data from forest companies.

284

2. Material and methods 2.1 Experimental design At Mekrijärvi Research Station of the University of Eastern Finland (62°46’N, 30°59’E), two drying racks for round wood with continuous measuring systems were built for research purposes in March, 2012 (Fig. 1). The purpose of the racks is to simulate energy wood storage at the roadside in the forest after cuttings. The racks are similar to those used on timber trucks to carry logs, and their size is 2.5 m in width, 2.6 m in length and 2.8 m in height. As the piles in the racks are quite small compared with real storage in the field, there are cover papers at the bottom and sides of the racks to avoid them drying too quickly. The small diameter energy wood stems of pine were piled up by a machine into the metal racks at the end of March 2012. In the system, four load cells continuously measure the weight of the pile in the rack. These four cells are connected to a junction box, which is connected onwards to a weighing transmitter. Data of weight are stored in a file that can be utilized for data management.

At the Mekrijärvi Research Station, there is a wellequipped meteorological station operated by the Finnish Meteorological Institute (FMI), which provides data on relative air humidity (%), air temperature (°C), wind speed (m/s), wind direction (°), solar radiation (W/m2) and rainfall (mm), air pressure (hPa), ground temperature (°C), rainfall intensity (mm/h), visible distance (m), height of clouds (m) and snow depth (cm).

The meteorological data is collected by a data logger.

The weather data can also be obtained from grid data.

The FMI provides gridded weather data for the whole of Finland. This data set consists of weather observations (e.g. temperature, humidity, precipitation), which have been interpolated to a 10 km x 10 km grid using the Kriging interpolation method (Venäläinen and Heikinheimo 2002).

Fig. 1 Drying racks with small diameter stems at Mekrijärvi ­ Research Station Croat. j. for. eng. 36(2015)2

Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291) J. Routa et al.

The mean annual precipitation in this area is 668 mm, and the mean annual temperature 2.1 °C. The mean snow depth in the Mekrijärvi area is approx.

45–65 cm in the winter months. In the winter of 2012–2013, snow depth was close to the average and the permanent snow cover period was typical. The drying season (from April to October 2012) was unfavourable for wood drying. The mean temperature (9.5 °C) for these seven months was somewhat lower than the long term average mean temperature (9.8 °C) (Drebs et al. 2002). Most of the precipitation occurred in July (163.8 mm) and June (104 mm), a total of 605.3 mm during the investigation period (Fig. 2), which is almost 50% more than the long term average of 439 mm. In the latter part of the drying season (from April to the beginning of June 2013), the precipitation was in total 156.3 mm. The long term averages (1971–2000) were taken from the nearby station at Ilomantsi Kirkonkylä, because there were no data from the Mekrijärvi station, which was founded in 1999. The Ilomantsi Kirkonkylä station is located only 11.6 km from Mekrijärvi, and therefore represents the same climate conditions.

The moisture content is determined based on weight changes in the energy wood storage pile. When the weight of the pile decreases, the moisture content of the material decreases, and when the weight increases, the moisture content of the material increases. The weight of snow on the pile is a challenge for moisture content estimates, because the weight of snow does not indicate changes in the moisture content of the material.

2.2 Study material and sampling method Small size stems were cut at the end of October, 2011. They were piled onto a roadside storage in the field right after cutting. From this storage, the stems were transported to the racks at the research station at the end of March 2012. So the material was in storage for five months under winter conditions in the field.

When the stems were piled onto the racks, six sample stems from each pile were randomly selected.

Five of the sample stems were pine and one of them birch. Three sample discs were taken from each sample stem; one from the bottom of the stem, one from the middle of the stem and one from the top of the stem.

All of these sample discs were taken because variation of the moisture content within the stem could be significant (Kärkkäinen 2003). In total 18 discs presented one pile in order to determine moisture content at the beginning of the storage period under study. The moisture content (wet basis) was determined using the oven dry method (EN 14774-2). Sampling was carried out in accordance with the solid biofuel standard EN 14778.

The weight changes in these two piles were similar during the summer of 2012. In early September, the other pile was covered with a cover paper manufactured

by Walki. The width of the paper was 4 metres. The paper is developed for this purpose, and it should keep the rain and snow away from the pile. The paper can be chipped with energy wood and combusted at a heating plant.

After the storage period, it was assumed that the moisture content varies within the pile. When the piles were unloaded, the material from each pile was chipped using a big drum chipper. The samples for the moisture content analysis were taken from the chips and they were taken from the top, the middle and the bottom of the pile. The fourth sample was a mix of the previous three samples. All the samples were analysed using the oven dry method. At the end of the storage period, we had four measurements of the moisture content per pile.

2.3 Validation data

Fig. 2 Precipitation (mm) during the effective drying period at ­Mekrijärvi Research Station in 2012 Croat. j. for. eng. 36(2015)2

The validation data for covered small diameter pine stem wood has been collected in central Finland during 2010–2011. The sampled piles were selected so as to represent the average energy wood storages in Finland.

The materials of the piles were typical of first thinning.

Most of the stands were harvested as an integrated energy and pulpwood harvesting, where all the pulp wood diameter wood (diameter >6 cm) was taken as pulp wood, and the rest of them were collected for

285

J. Routa et al. Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291)

Fig. 3 Moisture samples taken from chip piles energy wood. All the storage piles were covered with the Walki cover paper. The size of the roadside storages varied from 17 m3 to 295 m3. The energy wood was driven to the Äänekoski power plant and chipped there. The moisture samples were taken from piled chips; 6–8 samples were taken with ladle sampling to a big plastic tub. All the samples were spilled onto a table, where they were mixed and then the moisture samples were collected from nine points by hand to a duplicate plastic bag (5 litres). The plastic bags were delivered immediately to the laboratory, where the moisture content was measured using the oven dry method (EN 14774-2). Sampling was carried out in accordance with the solid biofuel standard EN 14778.

Uncovered pine stem wood was delivered by the Tornator Company. The stems were from cuttings made 2–21 months before. Eight test piles were chipped at the Fortum power plant on 10th of November 2014. The moisture samples were taken from the chip piles. Five samples were taken with ladle sampling to a big plastic tub, and all samples were spilled onto a table, where chips were divided into four parts (Fig. 3). One part was put into a duplicate plastic bag (5 litres). Plastic bags were delivered immediately to the laboratory, where the moisture content was measured using the oven dry method. Sampling was carried out in accordance with the solid biofuel standard EN 14774.

2.4 Data analysis At first, data from continuous measurements was prepared for the analysis. The running mean of the weight of the piles (average of ten previous measurements), the moisture content and the daily moisture change for each day were calculated. The data from 1st of April to the end of October was used, and the winter months were excluded. The weather parameters were interpolated to the grid, and then the evaporation was calculated using the PenmanMonteith equation (Monteith 1981) by the Finnish

286

Meteorological Institute. The interpolation method is explained in detail in Venäläinen and Heikinheimo (2002), except that the precipitation is obtained mainly from the weather radar network and the radiation parameters are nowadays taken from a weather model because of the lack of radiation measurements and synoptic cloud observations. Net evaporation (mm) was calculated by subtracting precipitation from the reference evaporation.

Different modelling approaches were tested; the linear regression model, multiple linear regression model and non-linear model. Temperature, precipitation, evaporation, wind speed and humidity were used as determining variables. Also, net evaporation was tested. Net evaporation means the difference between evaporation and precipitation. In fact, this variable contains all the most important weather parameters that affect energy wood drying.

The target variable is the moisture content alteration per day in % on a wet basis (DMC=daily moisture change). The analyses were performed with IBM SPSS Statistics version 20.

A Mann-Whitney test was used to compare the difference between measured and modelled moisture contents with IBM SPSS Statistics version 20, using the critical level at p<0.05. The Mann-Whitney test is considered to be one of the most powerful nonparametric tests especially testing differences in the location of the distribution (Ranta et al. 1992).

3. Results 3.1 Results of modelling Stand models and roadside storage models for small diameter stem wood were developed using three different approaches: linear regression, multiple linear regression and non-linear regression. In statistics, the non-linear regression model has the best statistical values (Table 1). When these models were applied to validation data of covered pine piles, it was Table 1 Statistical details of different models Test values

Linear

Multiple linear

Non-linear

regression model regression model regression model

F

784.7

171.3

355.5

p

0.000

0.000

0.000

2

R

0.705

0.726

0.766

Standard error

0.17

0.17

0.15

Croat. j. for. eng. 36(2015)2

Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291) J. Routa et al.

Fig. 4 Difference between measured and modelled moisture with different model types, pine stem wood, covered

Fig. 5 Measured and modelled moisture content of 14 different covered energy wood piles

Table 2 Drying models for covered and uncovered stem wood stored on roadside

Using the models starts with determining the moisture content of fresh wood. For that reason, average moisture of fresh wood, depending on the cutting month, is presented in Table 3 (Hakkila 1962, Gislerud 1974, Kärkkäinen 1976, Nurmi 1999, Hillebrand and Nurmi 2001, 2007, Nurmi and Lehtimäki 2011 and Routa, unpublished data). After cutting, the stems are stored and the model can be applied to estimate the daily change of the moisture content, and with that estimate the current moisture content of the wood material within the pile.

Roadside storage models DMC=Coef*(evaporation–precipitation)+const.

Moisture content (i)=moisture content(i–1)–DMC Model

Coef.

Const.

R2

SE

Pine birch mix, covered

0.062

0.051

0.70

0.2

Pine birch mix, uncovered

0.062

0.039

0.64

0.2

3.2 Model validation

found that the linear regression model gives the most reliable results (Fig. 4).

The moisture content estimation is made by the model acquired from the rack experiment. The result is compared to the moisture content of the reference pile.

Linear regression models were chosen because they appeared to be most functional, and the structure of the model was simple and understandable. For model form, the simplest regression model was chosen with one determining variable, net evaporation (Table 2).

3.2.1 Covered stem wood model The validation results against covered stem wood model are shown in Table 4 and Fig. 5. The difference between the measured and the modelled moisture content varied from 0.4 to 5.95% in 14 different piles.

Table 3 Moisture of fresh stem wood depending on the cutting month in Finland Moisture content of fresh stem wood, monthly, % Species

Jan

Feb

March

April

May

June

July

Aug

Sep

Oct

Nov

Dec

Pine

57

57

57

56

56

55

55

57

57

57

57

57

Birch

45

45

45

46

48

42

42

42

42

44

45

45

Croat. j. for. eng. 36(2015)2

287

J. Routa et al. Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291)

On average, the difference was only 0.3%. A statistical test was carried out (Mann-Whitney test) and the difference between the measured and modelled value was not statistically significant (p=0.7) 3.2.2 Uncovered stem wood model The validation results against the uncovered stem wood model are shown in Table 5 and Fig. 6. The difference between the measured and the modelled moisture content varied from 1 to 14% in 8 different piles. On average, the difference was 2.5%. The difference between the measured and the modelled value was not statistically significant (p=0.6). In this experiment, the age of the energy wood piles varied from 2–21 months. It can be seen that the oldest piles, which had been stored during the winter, have the highest difference between the modelled and the measured moisture content. The moisture content of energy wood increases during the winter, when the evaporation is really low, and in springtime melted snow increases the moisture of the pile especially in storage piles without a cover. It is a very site-specific situation, and it is hard to model how much the moisture increases, but the average amount of 5% has been used.

4. Discussion In this study, a forecast model for the moisture content of small size stems in roadside storage in both

Fig. 6 Measured and modelled moisture content of 8 different uncovered energy wood piles

cases, uncovered and covered piles, have been developed. With the detailed experimental data, the non-linear regression model produced the best statistical values. However, when the models were

Table 4 Measured and modelled moisture content, difference, % and difference in % units of 14 different covered energy wood piles Covered stem wood model Pile

Measured

Modelled

Difference between measured

Difference between measured

number

moisture

moisture

and modelled moisture, %

and modelled moisture, units

1

44.68

45.04

–0.36

0.36

2

46.85

43.90

2.94

2.94

3

45.50

48.94

–3.44

3.44

4

53.12

54.28

–1.16

1.16

5

62.07

61.82

0.25

0.25

6

36.60

39.75

–3.16

3.16

7

54.28

50.27

4.01

4.01

8

55.77

55.15

0.62

0.62

9

55.95

55.13

0.82

0.82

10

60.43

57.90

2.54

2.54

11

60.78

60.07

0.70

0.70

12

26.40

32.35

–5.95

5.95

13

47.7

42.64

5.06

5.06

14

50.3

48.76

1.54

1.54

Average

50.03

49.71

0.32

2.33

288

Croat. j. for. eng. 36(2015)2

Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291) J. Routa et al.

validated with the imprecise data from real life, the linear regression model gave nearly similar estimations for the moisture content as did observations from the field. For practical use, the linear-regression model was selected, a factor of which is net evaporation.

The models presented can be implemented at every location in Finland, because the Finnish Meteorological Institute has a database for interpolated meteorological observations covering the whole country in a 10 km x 10 km grid. For international use, model parameters needs to be estimated case by case, but it should also be possible to implement the approach worldwide.

Using accurate weather observations for modelling moisture changes in a single place would be preferred, especially from a scientific perspective. However, nowadays the weather observation network is relatively sparse in most countries to represent different areas well. For example, in Finland radiation measurements, which are needed to calculate the evaporation, are only made at a few stations.

Therefore, using gridded data, despite its limitations, is the best option when intended for wide use for the forecast models. With this application, the moisture models, which now use only weather observations and present history, could in the future be relatively easily connected to numerical weather forecast models.

The initial moisture content of wood is important for the accuracy of the estimation. If initial moisture is not measured, there is a risk that it differs from the average table value given in Table 3. The difference

will then remain through the storing process and can cause imprecise information of the moisture content.

Winter is a challenging period for the estimation procedure because of the ice and snow. It is difficult to estimate how big a proportion of the snow, for example, ends up in the heating plant and then increases the final moisture value of the pile. In the validation data of this study, there were two piles uncovered (piles 5 and 6), which were stored over the winter, and the difference of the moisture content between the measured and the modelled value is large, i.e. 11% and 14%. This might be due to the snow in the winter season, which has considerably increased the moisture content of the pile.

Data for the forecast models originates from automated monitoring in the spring, summer and autumn, so the daily moisture alteration during winter cannot be estimated by those models. Therefore, this application is recommended to be used from April to October in Finland. It can be assumed that the moisture content of fuel wood increases in the springtime when melted snow penetrates the stacks. When energy wood storages are stored at the roadside over the winter, the calculation has to be stopped during the period from 1 November to 31 March. The calculation starts again on 1 April with moisture content that has been achieved with the model by 31 October. If the storage is uncovered, the moisture content of storage should be increased during the winter period by approximately 5% units.

Measuring the moisture content in different phases of supply chain is challenging. Exact moisture

Table 5 Measured and modelled moisture content, difference, % and difference in % units of 8 different uncovered energy wood piles Uncovered stem wood model Pile

Age

Measured

Modelled

Difference between measured

Difference between measured

number

(months from logging)

moisture

moisture

and modelled moisture, %

and modelled moisture, units

1

2

55.83

57.25

–1.42

1.42

2

4

46.97

53.51

–6.54

6.54

3

7

41.58

42.56

–0.98

0.98

4

9

45.66

48.09

–2.43

2.43

5

12

56.00

41.93

14.07

14.07

6

15

47.20

36.24

10.96

10.96

7

17

40.74

39.23

1.51

1.51

8

21

37.02

32.53

4.49

4.49

46.38

43.92

2.46

5.30

Average Croat. j. for. eng. 36(2015)2

289

J. Routa et al. Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291)

monitoring based on the weight changes of racks, used in this study, is possible only for a limited number of cases. Quality of the real life observations can easily be questioned, because the variation of observations in one truck load is remarkable, and sampling is seldom done according to good scientific principles.

The practitioners of the forest energy business have stated that their requirement of the moisture estimate accuracy for enterprise resource planning purposes would be ±5% of the moisture content. In this study, 77% of observations meet this limit.

5. Conclusions Modelling is an easy option to make an estimate of the moisture content of an energy wood pile if compared with sampling and measuring the moisture of samples. Models are also a considerably more reliable method for allocation and prioritisation of piles than the »educated guesses« used earlier. In practice, piles are often kept in storage too long »just to be sure« that they are dry enough. This increases storages levels and due to that, the capital costs of supply. In addition, dry matter losses increase due to too long storage times.

Some forest companies have already started to use models as a part of their Enterprise Resource Planning (ERP) systems, and the feedback has been encouraging; models work well enough to give added value. A need for further development is still recognized, especially concerning the varying weather conditions in autumn.

Some fuel chip reception stations on heating plants are already using automated continuous moisture metering. If the chain of custody is proof, this information can be used effectively for future development of models.

Acknowledgements The work was supported by the European Regional Development Fund, the Finnish Funding Agency for Technology and Innovation (TEKES) (Laava project), the Sustainable Bioenergy Solutions for Tomorrow (BEST) research program coordinated by FIBIC Ltd. and CLEEN Ltd., and the European Union Seventh Framework Programme, INFRES – project [grant number 311881, 2012–2015]. We also thank Tornator Ltd., Stora Enso Ltd., Metsä Group Ltd. UPM Ltd., JLTuote Ltd. and Fortum Ltd, and the intellectual support of B.Sc. Janne Immonen and B.Sc. Mika Auvinen is gratefully acknowledged. Special thanks go to the research team of Mekrijärvi Research Station for implementing the drying rack experiments.

290

6. References Andersson, G., Asikainen, A., Björheden, R., Hall, P.W., Hudson, J.B., Jirjis, R., Mead, D.J., Nurmi, J., Weetman, G.F., 2002: Production of forest energy. In: Bioenergy from Sustainable Forestry. Guiding Principles and Practice. Edited by Richardson J., Björheden R., Hakkila P., Lowe A.T., Smith C.T. Kluwer Academic Publishers, The Netherlands, 344 p.

Dong-Wook, K., Murphy, G.E., 2013: Forecasting air drying rates of small Douglas-fir and hybrid poplar stacked logs in Oregon, USA. International Journal of Forest Engineering 24(2): 137–147.

Drebs, A., Nordlund, A., Karlsson, P., Helminen, J., Rissanen, P., 2002: Tilastoja Suomen ilmastosta 1971–2000. Ilmastotilastoja Suomesta: 1, Ilmatieteen laitos, Helsinki, 100 p.

Erber, G., Kanzian, C., Stampfer, K., 2012: Predicting moisture content in a pine logwood pile for energy purposes.

Silva Fennica 46(4): 555–567.

Erber, G., Routa, J., Kolström, M., Kanzian, C., Sikanen, L., Stampfer, K., 2014: Comparing two different approaches in modelling small diameter energy wood drying in logwood piles. Croatian Journal of Forest Engineering 35(1): 15–22.

Filbakk, T., Hoibo, O., Dibdiakova, J., Nurmi, J., 2011: Modelling moisture content and dry matter loss during storage of logging residues for energy. Scandinavian Journal of Forest Research 26(3): 267–277.

Gigler, J., van Loon, W.K.pP, Seres, I., Meerdink, G., Coumans, W.J., 2000: PH – Postharvest Technology: Drying characteristics of willow chips and stems. Journal of Agricultural Engineering Research 77(4): 391–400.

Gislerud, O., 1974: Heltreutnyttelse. II. Biomasse og biomasseegenskaper hos tynningsvirke av gran, furu, bjork og or.

Norsk Institutt for skogforskning, Rapport 6: 1–59.

Hakkila, P., 1962: Polttohakepuun kuivuminen metsässä.

Communicationes Instituti Forestalis Fenniaie 54(4): 1–82.

Hartmann, H., Kaltschmitt, M., 2001: Energie aus Biomasse: Grundlagen, Techniken und Verfahren, Springer Verlag, Berlin–Heidelberg–New York, 770 p.

Hillebrand, K., Nurmi, J., 2001: Hakkuutähteen laadunhallinta. VTT Energia. Raportteja 2/2001, 51 p.

Kärkkäinen, M., 1976: Puun ja kuoren tiheys ja kosteus sekä kuoren osuus koivun, kuusen ja männyn oksissa. Silva Fennica 10(3): 212–236.

Kärkkäinen, M., 2003: Puutieteen perusteet. Metsälehtikustannus, Hämeenlinna, 451 p.

Liang, T., Khan, M.A., Meng, Q., 1996: Spatial and temporal effects in drying biomass for energy. Biomass and Bioenergy 10(5): 353–360.

Monteith, J.L., 1981: Evaporation and surface temperature.

Quarterly Journal of the Royal Meteorological Society 107(451): 1–27.

Croat. j. for. eng. 36(2015)2

Validation of Prediction Models for Estimating the Moisture Content of Small Diameter Stem Wood (283–291) J. Routa et al.

Murphy, G., Kent, T., Kofman, P.D., 2012: Modeling air drying of sitka spruce (Picea sitchensis) biomass in off forest storage yards in Ireland. Forest Products Journal 62(6): 443–449.

logistic management. In: Special issue. Abstracts for international conferences organized by LSFRI Silava in cooperation with SNS and IUFRO. Mezzinatne 25: 33–35.

Nurmi, J., 1999: The storage of logging residue for fuel. Biomass and Bioenergy 17(1): 41–47.

Routa, J., Kolström, M., Ruotsalainen, J., Sikanen, L., 2015: Precision Measurement of Forest Harvesting Residue Moisture Change and Dry Matter Losses by Constant Weight Monitoring. International Journal of Forest Engineering 26(1): 71–83.

Nurmi, J., Hillebrand, K., 2007: The characteristics of wholetree fuel stocks from silvicultural cleanings and thinnings.

Biomass and Bioenergy 31(6): 381–392.

Nurmi, J., Lehtimäki, J., 2011: Debarking and drying of downy birch (Betula pubescens) and Scots pine (Pinus sylvestris) fuelwood in conjunction with multi-tree harvesting. Biomass and Bioenergy 35(8): 3376–3382.

Stokes, B.J., Watson, W.F., Miller, D.E., 1987: Transpirational drying of energywood, ASAE paper No. 87‐1530, St. Joseph, MI: American Society of Agricultural Engineers, 13 p.

Ranta, E., Rita, H., Kouki, J., 1992: Biometria. Yliopistopaino, Helsinki, 569 p.

Routa, J., Kolström, M., Ruotsalainen, J., Sikanen, L., 2012: Forecasting moisture changes of energy wood as a part of

Röser, D., Mola-Yudego, B., Sikanen, L., Prinz, R., Gritten, D., Emer, B., Väätäinen, K., Erkkilä, A., 2011: Natural drying treatments during seasonal storage of wood for bioenergy in different European locations. Biomass and Bioenergy 35(10): 4238–4247.

Torvelainen, J., Ylitalo, E., Nouro, P., 2014: Puun energiakäyttö Suomessa. Metsätilastotiedote 31/2014. Suomen virallinen tilasto. Metsäntutkimuslaitos, 7 p.

Venäläinen, A., Heikinheimo, M., 2002: Meteorological data for agricultural applications. Physics and Chemistry of the Earth 27(23): 1045–1050.

Authors’ address: Johanna Routa, PhD.* e-mail: johanna.routa@luke.fi Lauri Sikanen, PhD. e-mail: lauri.sikanen@luke.fi Natural Resources Institute Finland, Luke P.O. Box 68, FI-80101 Joensuu FINLAND Marja Kolström, PhD. e-mail: marja.kolstrom@uef.fi University of Eastern Finland, School of Forest Sciences P.O. Box 111, FI-80101 Joensuu FINLAND

Received: March 13, 2015 Accepted: May 6, 2015 Croat. j. for. eng. 36(2015)2

Johanna Ruotsalainen, MSc. e-mail: johanna.ruotsalainen@fmi.fi Finnish Meteorological Institute, Aviation and Military Services Yliopistonranta 1 F, 70210 Kuopio FINLAND * Corresponding author

291

Original scientific paper

Are Forest Functions a Useful Tool for Multi-objective Forest Management Planning? Experiences from Slovenia Tina Simončič, Andrej Bončina Abstract The concept of forest functions evolved in Central Europe as an important tool in the practice of multi-objective forest management. It is based on designating forest function areas that are relatively more important for the selected services. Recent practice has raised a number of concerns regarding the suitability and effectiveness of the concept of forest functions in satisfying increasing social demands on forests. This paper presents the main results of a survey of forest functions in Slovenia as seen by forestry experts (n=162). There was broad agreement among respondents that there are too many forest function types, and that at most two levels of importance should be applied. Principal component analysis identified four main purposes for designating forest function areas: harmonisation of forest uses, identification of conflict areas, and argumentation for land use planning; setting management priorities and strategies such as limitations for harvesting and skidding; providing a framework for financial subsidies for adjusted forest management; guiding forest road planning and construction.

Respondents identified designation of forest function areas in both public and private forests, and their high importance for land use planning as the major strengths of the concept. Major weaknesses were an insufficient monitoring and planning system, and complicated forest function mapping. It seems that forest functions have remained an important tool in the practice of multi-objective forest management. However, improved planning methods, increased public participation and greater integration of forest functions in forest policy are needed.

Keywords: multiple forest use, integration model, concept of forest functions, services, forestry experts, survey

1. Introduction In Central Europe, the integration model of multiobjective forest management prevails. This management approach considers all forest functions at the same place and time, although their importance can differ (Borchers 2010). The pillar of the integration model is the »concept of forest functions«, which is based on the designation of areas with important forest functions (hereafter forest function areas) that are of relatively higher importance for the selected forest services (functions) than the surrounding forest area (Blum et al. 1996). The concept was developed in the 1950s by Dietrich (1953), who defined a forest function Croat. j. for. eng. 36(2015)2

as a social demand imposed on forests. Most of the variants and definitions that followed relied on Dietrich’s work (e.g. Rupf 1960, Hasel 1971). Multifunctional forest management was developed due to increasing demands for environmental services (e.g.

Mantel 1990). It first came into use through the wake water paradigm, which is based on the assumption that management for sustainable timber production ensures ecological and social functions at the same time (Glück 1982). Later, »forest function mapping« was integrated into multifunctional forest management (Riegert and Bader 2010). The concept of forest functions was gradually affirmed in the practical forestry of Central European countries (especially in

293

T. Simončič and A. Bončina

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

Switzerland, Germany, Austria and Slovenia) in the 1980s and 1990s (Volk 1987, Anko 1995) and has remained an important tool in multi-objective forest management.

Three groups of forest functions are commonly defined by forestry legislation: production, ecological (or also protective) and social functions (e.g. Forst Act 1975, ZG 1993). The production function refers to the use of timber and other wood and non-wood products. Ecological functions include protection against natural hazards; the protection of soil, water and climate; and the conservation of natural habitats and biological diversity. Social functions are mainly connected to recreation and other cultural and educational values, and the protection of natural and cultural heritage. Detailed classifications of forest functions differ significantly among Central European countries (Simončič et al. 2013). For example, in Germany approx. 20 forest function types are classified, although the number may differ among federal states (e.g. Volk and Schirmer 2003). In Austria and Switzerland, the classification systems are simpler. In Austria, protective, protection, social and welfare functions are distinguished (Fürst and Schaeffer 2000), whereas in Switzerland, protective, protection, social and nature conservation functions are commonly defined (BUWAL 1996). Forest development plans (Ger. Waldentwicklungspläne) are the main tools for designating forest function areas and for prescribing management guidelines to promote the selected functions.

The concept of forest function areas has contributed greatly in emphasizing the public importance of forests (Bachmann 2005, Bürger-Arndt 2012) and mitigating conflicts between forest uses (Hanewinkel 2011). In addition, forest function areas have become influential in spatial planning as an important argument for environmental impact assessment in forest areas (e.g. Berger and Ray 2004, Schulzke and Stoll 2008). They have also led to better communication between forestry practitioners and stakeholders (Krott 1985). Nevertheless, a number of concerns have been raised regarding the suitability and effectiveness of the concept of forest functions in practicing multi-objective forest management. Applying fine scale mapping, overlapping and ranking of forest function areas has often failed to meet the diverse demands on forests, mainly due to poorly defined management measures associated with the forest function areas (Weiss et al.

2002), the lack of financial support for adjusted forest management (Buttoud 2002) or limited options for the participation of forest owners and public in the designation process (Ruppert-Winkel and Winkel 2009). In addition, the concept has often been criticized for be-

294

ing too general and prescriptive (e.g. Krott 1985). Another point of concern is that the discourse has not considered an effective reward system for social services provided by forest enterprises (Pistorius et al.

2012). However, there are significant differences among CE countries in how the concept has been applied (Simončič et al. 2013).

In Slovenia, forest functions have been used in forest management planning for nearly three decades.

However, with the exception of recent research (e.g.

Bončina and Matijašić 2010, Planinšek and Pirnat 2012, Simončič and Bončina 2012, Mavsar et al. 2013, Simončič et al. 2013, 2015), they have not been a popular topic of interest among scientists. Accumulated experience in the implementation of the concept during the last decades and new regulations regarding multi-objective forest management underscore the need to evaluate the effectiveness of forest functions as a tool in the practice of multi-objective forest management. We used a survey among forestry experts in Slovenia to explore: Þ t heir perceptions on the designation of forest function areas, including the importance of forest function areas in practicing multi-objective forest management; Þw hether these perceptions differ among different groups of forestry experts.

2. The concept of forest functions in Slovenia In Slovenia, wood and non-wood forest functions gained equal importance with the enforcement of the Forestry Act in 1993 (ZG 1993). In the last three decades, the classification of forest function types has been developed (Anko 1995), and detailed criteria and procedures for designation of forest function areas have been elaborated (Pravilnik 1998, 2010). The forestry act classifies three main groups of forest functions (social, ecological and economic) and further defines 17 forest function types (Table 1).

Forest function areas are designated in the regional forest plans, which are the strategic plans made at the level of forest management regions (14 in Slovenia). Regional forest plans are aimed at defining objectives, priorities and controlling mechanisms for ensuring public interests and management of the forest.

They are approved by the government. In addition, forest function areas are supplemented in the forest management unit plans, in which operational and frame planning is combined (Bončina 2001). Forest function areas are updated every 10 years in the frameCroat. j. for. eng. 36(2015)2

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

work of regional forest plan revisions. This is a multistep process consisting of: Þ c ollecting information about forest functions from various institutions (e.g. water protection zones, Natura 2000 sites, hiking trails, natural hazard potential); Þ c hecking and harmonizing information about forest functions with forest management unit plans; ÞG IS analyses and preparation of forest function maps; Þ s etting management guidelines associated with the forest function areas; Þh armonizing the maps and associated management prescriptions with other institutions, the public and forest owners.

Forest function mapping in Slovenia is partly similar to the methodology used in Germany and Austria.

The forest function map is elaborated on a 1:25,000 scale. The minimum mapping area has the same threshold as for the designation of forest area, which is 0.25 ha. To avoid multiple overlapping that often occurs between 17 types of functions, a synthesis map of the four main categories of forest functions is produced in the regional forest plan, although the database enables the presentation of individual functions on any spatial level (Fig. 1). The importance of each function is ranked according to three levels: Þ fi rst level – function determines management regime; Þ s econd level – function influences management regime; Þ t hird level – function has no significant influence on management regime.

Each forest area is designated with a function; if no function is explicitly important, wood production is automatically ranked as primary (first or second level of importance). Due to overlapping, the sum of forest function areas is greater than the surface of the forest area (Fig. 2).

In private forests, financial support is available if additional measures are needed when there are tradeoffs between owners’ objectives and public demands.

The main benefits available for private owners for providing non-timber functions are the right to full or partial financial support of silvicultural and protective measures. The amount of subsidies partly depends on the importance of social and ecological forest functions. In the case of the first or second level of importance, the basic amount of subsidies available for management is increased by 20% and 10%, respectively.

Croat. j. for. eng. 36(2015)2

T. Simončič and A. Bončina

Table 1 Distribution of forest function areas in Slovenia according to the first and the second level of importance (source: SFS 2012).

Total forest area amounts to 1.2 million hectares Percentage of the whole forest area

Function

First level, %

Second level, %

Protection

15.4

24.9

Hydrologic

5.1

44.6

Habitat protection

5.0

59.6

Climatic

2.9

3.5

Protective

2.2

0.4

Hygienic-health

2.3

6.0

Recreational

2.4

5.0

Touristic

2.5

2.4

Educational

0.6

0.4

Research

0.8

0.0

Protection of natural heritage

3.0

14.6

Protection of cultural heritage

0.4

13.3

Aesthetic

2.8

7.0

Defence

1.1

1.3

Timber production

59.6

24.4

Non-wood products

1.4

20.1

Game management

2.8

0.0

3. Methods 3.1 Survey methodology A web based questionnaire (implemented with SurveyMonkey; www.surveymonkey.com) was conducted during February and June 2013 among different groups of forestry experts (Table 2). The questionnaire was first pilot tested through face-to-face interviews with the scientists of the study and further refined. Before data collection, it was additionally tested on six representatives (two local foresters, two scientists, and two planners). The survey lasted 25 minutes on average. Invitations to respond to the questionnaire were distributed by email. Each questionnaire was enclosed with a cover letter identifying the general purpose of the study and key contact person.

The questions were conducted based on our previous research (e.g. Simončič and Bončina 2012), a literature review, analyses of existing legal documents, personal discussions and interviews with forest plan-

295

T. Simončič and A. Bončina

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

ners and local foresters, and consultations with on-theground practitioners. For the purpose of the paper, only one part of the questionnaire is presented. The questionnaire contained structured questions. The socio demographic characteristics included information about the respondents’ sex, age, education, work location and work position. The questions about types and ranking of forest functions were the multiple response type. Before the interviews, we prepared a list of 16 purposes that we hypothesized forestry experts might consider as the main reasons for designating forest function areas. The respondents were then asked to express the degree of importance of forest function areas to the pre listed purposes with a grading scale.

The grading scale was a five point ordinal Likert type scale (Likert 1932): Þ ( 1) not at all important; Þ ( 2) rather unimportant; Þ ( 3) not important and not unimportant; Þ ( 4) rather important; Þ ( 5) very important.

The questions consisted of individual Likert items.

For a general evaluation of the concept of forest functions, we prepared a list of 17 statements associated with the designation of forest function areas and subsequent management. Answers to each question were given as a reflection of choices from the strongest agreement (1) to the strongest disagreement (5). We used affirmative and negative statements to encourage respondents to carefully consider each statement and to decrease automatic responses. We then applied cross dating to get parallel statements and to be able to perform statistical tests.

3.2 Respondent profile The survey population consisted of forestry experts from three institutions. A total of 162 responses were analyzed out of approximately 800 people, representing about 25% of the population. The respondents were then classified into three main groups according to their work positions. For the total sample, scientists represented 30% and practitioners (local foresters and planners) about 22% of the population. The average age of

Fig. 1 Map of selected forest function areas at the national level with the first and second level of importance (source: SFS 2014). Protection refers to indirect protection; protective means direct protection of objects; production refers to the timber production

296

Croat. j. for. eng. 36(2015)2

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

T. Simončič and A. Bončina

Fig. 2 Section from forest function map at the landscape spatial scale. Only protection and recreational functions of first level of importance are shown

Table 2 Respondents included in the survey Organization*

Working position

Number of responses, n

Local

SFS

District forester

71

foresters

SFS

Head of local unit

24

SFS

Forest planner at local or regional unit

29

SFS

Other employee of regional unit

19

Group

Planners

SFS

Scientists

Employee of central unit

4

BF

Researcher

14

SFI

Researcher

1

* SFS – Slovenia Forest Service; SFI – Slovenian Forestry Institute; BF – Biotechnical Faculty, Department of Forestry

Croat. j. for. eng. 36(2015)2

the interviewees was 45 years. Men (88%) prevailed in the survey. The majority of interviewees had university education (43%), followed by higher professional school (37%), a master’s or PhD degree (16%) and high school (4%). The respondents mainly work in the forest or forested landscape (74%), followed by agriculture (14%) and the suburban and urban landscape (12%).

3.3 Statistical data analysis The results were analyzed using Excel and SPSS (IBM 2011). Mean, standard deviation and frequency distribution were used as the basic statistics in the data analysis. The differences between different groups of forestry experts were tested using the χ2 test. Due to the sample size, the Likert grades were joined into the following categories: Þ s trongly disagree and disagree; Þn eutral; Þ a gree and strongly agree.

297

T. Simončič and A. Bončina

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

Table 3 Respondent opinions on the number of forest functions The Forestry Act and planning regulations define 17 forest functions.

Local foresters, %

Planners, %

Scientists, %

All, %

Number of forest functions is adequate

49.5

21.2

13.3

37.0

Number of forest functions is too high

44.2

78.8

86.7

59.3

Number of forest functions is too low

0.0

0.0

0.0

0.0

Undecided

6.3

0.0

0.0

3.7

What is your opinion on the number of forest functions?

4. Results

The number of responses allowed us to only test differences between local foresters and planners, whereas differences with scientists were analyzed by comparing the frequency distribution of responses.

4.1 Number and types of forest functions

The factors influencing the perceived importance and general evaluation of the concept were analyzed by bivariate Spearman correlation coefficient (r) between the respondents’ socio-demographic variables and their opinions, which is commonly used to analyze Likert scale data (Norman 2010). In our case, we compared independent categorical variables such as gender, age, working position, etc. with dependent variables consisting of ordinal data (Likert grades).

Table 4 Respondent opinions on the types of forest functions

We applied principal component analysis (PCA; Hill and Lewicki 2007) in SPSS to identify the major categories of importance of forest function areas from the list of 16 statements. PCA is a type of exploratory factor analysis that explains the maximum amount of common variance in a correlation matrix using the smallest number of explanatory factors (Field 2000). We chose this approach because the correlation analysis found a degree of interdependence of the data, estimated by Pearson correlation coefficient, at 0.05 and 0.01 significance levels. The reliability of the PCA was evaluated using the KaiserMeyer-Olkin measure of sampling adequacy (KMO).

KMO greater than 0.7 is considered as an acceptable reliability coefficient. Also, we applied Bartlett’s test of sphericity to check the suitability of our data for data reduction. The significant value for this analysis (P=0.00) led us to reject the null hypothesis and conclude that there are correlations in the data set that are appropriate for factor analysis. Based on the Kaiser criterion, only components with an eigenvalue greater than one were considered. Thus, the first four principal components (PCs) were extracted (controlling for 68.7% of the variance) and subsequently rotated with varimax rotation to increase their interpretability.

298

The majority (59.3%) of the survey respondents indicated that the number of forest functions is too high (Table 3). We found a statistically significant difference between different groups of forestry experts

Which (if any) forest functions would you no longer designate?

Local foresters, %

Planners, Scientists, % %

All, %

Defence

28.4

42.3

46.7

34.6

Hygienic-health

23.2

38.5

13.3

27.2

Touristic

12.6

38.5

13.3

21.0

Climatic

12.6

30.8

20.0

19.1

Aesthetic

8.4

32.7

26.7

17.9

Educational

8.4

26.9

0.0

13.6

Protective

10.5

11.5

0.0

9.9

Non-wood products

8.4

7.7

6.7

8.0

Research

5.3

9.6

0.0

6.2

Protection of cultural heritage

4.2

11.5

0.0

6.2

Protection of natural heritage

2.1

11.5

0.0

4.9

Recreational

1.1

3.8

0.0

1.9

Hydrologic

0.0

3.8

0.0

1.2

Wood production

1.1

1.9

0.0

1.2

Protection

0.0

0.0

6.7

0.6

Game management

0.0

1.9

0.0

0.6

Habitat protection

0.0

0.0

0.0

0.0

Croat. j. for. eng. 36(2015)2

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

(P<0.001). The frequency distribution of the responses showed that the number of forest functions is adequate for about half of local foresters, whereas there is broad agreement among planners and scientists that there are too many types of forest functions.

Most respondents would no longer designate areas with the defence, hygienic and health, touristic, climate and aesthetic functions (Table 4). There is a statistically significant difference between different groups of forestry experts regarding the touristic (P=0.001), educational (P=0.002) and aesthetic functions (P=0.001). The frequency distribution of responses shows that a higher share of planners compared to the other two groups would no longer designate touristic, educational and aesthetic functions, the latter also being the case for scientists.

We asked the respondents if they would combine any forest functions. The most common combinations of functions were the following: recreational and touristic (58.6%), protection and protective (38.3%), educational and research (38.3%), protection of cultural and natural heritage (32.7%), and climatic and hygienic– health (27.8%). We found statistically significant differences among forestry experts in combining climatic (P=0.000), recreational (P=0.005) and educational functions (P=0.007). Most (86.3%) local foresters would not combine the climatic function with any of the other functions, whereas about half of scientists and planners would combine the climatic function with other functions. About half (51.6%) of local foresters would not combine the recreational function with other functions, whereas the majority of planners (75.0%) and scientists (60.0%) would combine the recreational function with other functions.

T. Simončič and A. Bončina

quency distribution of the responses points to the largest differences among local foresters and the other two groups, with local foresters being less critical of the current ranking system.

4.3 Perceived importance of forest function areas The lowest importance of forest function areas was given to the following purposes: financial subsidies for management restrictions, financing additional works, planning silviculture and protection works and selection of trees to be cut (Table 6). The highest importance was given to arguments against deforestation of forestland, basis for environmental impact assessment and influence on forest road construction. A higher share of planners (69.2%) compared to local foresters (51.6%) pointed to the importance of forest function areas for environmental impact assessment (P=0.015), whereas a higher share of local foresters (60.0% and 74.7%, respectively) compared to planners (34.6% and 53.8%, respectively) pointed to the importance of forest road planning (P=0.009) and the implementation of harvesting and skidding (P=0.034).

PCA analysis revealed four major categories of importance among the 16 designation purposes, which explained 68.7% of the variability in decision making (Table 7). The highest importance of designating forest function areas (PC 1), accounting for 23.2% of the total variability, was for planning forestland use and broader land use planning. PC 1 had the highest loadings of factors (six factors with factor loadings higher than 0.70). The second category (PC 2) represented the importance of forest function areas for planning and implementing management measures and explained 21.5% of the variance. We identified a third PC as the importance of financial subsidies. It additionally explained 14.9% of the variability. PC 4, which describes the importance for forest road construction, additionally explained 9.1%.

Respondent's age and forest management region had no significant correlations with perceived importance of forest function areas, whereas working posi-

4.2. Ranking of importance of forest functions The majority (58.7%) of respondents would change the current ranking system and most would apply the first and second level of importance (Table 5). We found statistically significant differences among different groups of forestry experts (P=0.001). The fre-

Table 5 Respondent opinions on ranking the importance of forest functions Which levels of importance would you use?

Local foresters, %

Planners, %

Scientists, %

All, %

Current system of three levels of importance

53.7

19.2

20.0

39.5

First and second level of importance

27.4

46.2

33.3

34.0

First level of importance

7.4

23.1

20.0

13.6

First level of importance or second where the areas overlap

9.5

9.6

26.7

11.1

Undecided

2.1

1.9

0.0

1.9

Croat. j. for. eng. 36(2015)2

299

T. Simončič and A. Bončina

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

Table 6 Respondent perceptions of the importance of forest function areas (the frequency distribution of the responses in %) Statement

Likert scale* 1

2

3

4

5

Avg.±st. dev.

Assessment of deforestation of forestland

0.6

1.9

12.3

45.1

40.1

4.22±0.78

Environmental impact assessment

0.6

3.7

15.4

39.5

40.7

4.16±0.86

/

3.7

19.1

39.5

37.7

4.11±0.84

Planning road construction

0.6

2.5

22.2

43.2

31.5

4.02±0.83

Participation in elaboration of land use plans

1.2

6.8

24.7

47.5

19.8

3.78±0.88

Identification of conflict areas

2.5

5.6

25.9

44.4

21.6

3.77±0.93

Harvesting and skidding implementation

0.6

11.1

22.2

48.1

17.9

3.72±0.91

Harmonization of multiple forestland uses

1.2

6.2

34.6

40.7

17.3

3.67±0.88

Participation with forestland users

2.5

10.5

34.0

40.1

13.0

3.51±0.93

Maximum allowable cut

1.9

9.9

39.5

34.6

14.2

3.49±0.92

Subsidies for silviculture works

2.5

15.4

32.1

34.6

15.4

3.45±1.01

Planning additional works

4.9

9.9

38.9

34.6

11.7

3.38±0.99

/

19.8

38.3

32.7

9.3

3.31±0.89

Planning silviculture and protection works

1.9

18.5

36.4

36.4

6.8

3.28±0.91

Financing additional works

5.6

19.8

32.1

30.9

11.7

3.23±1.07

Financial subsidies for management restrictions

10.5

21.0

27.8

25.9

14.8

3.14±1.21

Forest road construction

Selection of trees to be cut

* 1 – unimportant; 2 – rather unimportant; 3 – not important and not unimportant; 4 – rather important; 5 – very important

tion had the strongest. Local foresters and local planners acknowledge forest function areas as more important for the selection of trees to be cut (r=–0.21, P<0.01), maximum allowable cut (r=–0.29, P<0.01) and harvesting and skidding implementation (r=–0.23, P<0.01), whereas higher officials and scientists find forest function areas more important for identification of conflict areas (r=0.17, P<0.05), harmonization of multiple forestland uses (r=0.22, P<0.01), environmental impact assessment (r=0.20, P<0.05) and assessment of deforestation of forestland (r=0.18, P<0.05). Men find forest function areas more important for the selection of trees to be cut (r=–0.17, P<0.05) and maximum allowable cut (r=–0.19, P<0.05), whereas women perceive environmental impact assessment as more important (r=0.18, P<0.05), although this may be related to the higher share of women among forest planners and scientists compared to the share of women among local foresters.

4.4 General evaluation of the concept of forest functions

(p[rating<3]>0.50): the lack of financial instruments, complicated forest function mapping, poor monitoring of the effectiveness of management measures and insufficient participation of stakeholders, especially forest owners in the designation process (Table 8). The main advantages of the concept (p [rating>3]>0.50) were designation of forest function areas in public and private forests, ranking of the importance of functions and usefulness of forest function maps for spatial planning. Five statements showed statistically significant differences among forestry experts.

The frequency distribution of responses indicated that planners are more critical of forest function maps (p [rating>3]=0.35) compared to local foresters (p [rating>3]=0.13) and of the system of financial instruments (planners p [rating>3]=0.885; local foresters p [rating>3]=0.632). Significant differences were also found regarding ownership focus. For example, 1.9% of planners support the designation of forest functions only in agreement with the owners, whereas the proportion of local foresters is higher in this regard (16.8%).

Respondent opinions point to the following greatest weaknesses of the concept of forest functions

The strongest correlations were found between the general evaluation of the concept and respondent

300

Croat. j. for. eng. 36(2015)2

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

Table 7 Factor loadings in the PCA analysis of respondent perceptions of the importance of forest function areas (N=162, KMO=0.841) Importance

Categories of importance* PC1

PC2

PC3

PC4

Harmonization of multiple forestland uses

0.82

–

–

–

Environmental impact assessment

0.76

–

–

0.43

Participation in elaboration of land use plans 0.76

–

–

–

Identification of conflict areas

0.75

–

–

–

Participation with forestland users

0.73 0.31

–

–

Assessment of deforestation of forestland

0.71

–

–

0.47

Selection of trees to be cut

–

0.85

–

–

Maximum allowable cut

–

0.80

–

–

Planning silviculture and protection works

–

0.78

–

–

Harvesting and skidding implementation

–

0.72

–

–

Financing additional works

–

–

0.85

–

Financial subsidies for management restrictions

–

–

0.80

–

Subsidies for silviculture works

–

–

0.74 0.32

Planning additional works

0.31 0.35 0.55

–

Forest road construction

–

0.54

–

0.62

Planning road construction

–

0.58

–

0.62

Extraction Method: PCA with varimax rotation and Kaiser normalization.

Bolded loading indicates a value greater than 0.50, loadings below 0.25 are not shown. *Main principal components (PC): PC1 – planning forest land use and broader land use planning; PC2 – planning and implementing management measures; PC3 – financial subsidies; PC4 – road construction.

working position. Negative correlations point to the conclusion that local foresters and forest planners at local and regional units are more critical of unclear forest function maps (r=–0.18, P<0.01), designation of forest functions areas without owner agreement (r=–0.24, P<0.01) or in private forests in general (r=–0.26, P<0.01), whereas higher officials and scientists are more critical of the system of financial instruments (r=0.16, P<0.05) and monitoring of management measures (r=0.33, P<0.01). Men tend to be more critical of financial instruments (r=–0.20, P<0.01) and the monitoring system (r=–0.17, P<0.01) than women, whereas women are more critical of forCroat. j. for. eng. 36(2015)2

T. Simončič and A. Bončina

est function maps (r=–0.18, P<0.01) and the complicated description of forest functions in management plans (r=–0.17, P<0.01).

5. Discussion Our study addressed several topics regarding the concept of forest functions in Slovenia. The first was the classification system (i.e. number, types and ranking of forest functions). There was broad agreement among respondents (although less for local foresters) that the current number of forest functions is too high.

The respondents would either combine many of the existing forest functions, or would not designate some of them. One of the reasons for such a response could be that some forest function types are designated for similar reasons (e.g. recreational and touristic functions) or with regard to rather vague designation criteria (e.g. hygienic–health function). Other CE countries, such as federal states in Germany (e.g. Gross 2007) or in the eastern part of Central Europe, even have more detailed classification of forest function types (Simončič et al. 2013), whereas Austria and Switzerland classify only four to five main functions (BUWAL 1996, Fürst and Schaeffer 2000). The latter approach seems to be more appropriate for forest management given that differentiating and mapping a high number of functions is not practical for collaboration with stakeholders or for implementing forest management (Bončina et al. 2014). In addition, some of the existing forest functions (e.g. climatic function) are not dependent on forest management and can be provided without spatial designations.

Most of the respondents in our survey would apply only the first and second level of importance. The current ranking system of the importance of forest functions used in Slovenia is similar to the Austrian system, which applies four ranks (WEP 2006). In Germany, only recreational (two levels according to the intensity of recreation) and hydrological functions (two levels according to water regulations) are commonly ranked (Waldfunktionen Kartierung 2004). In Switzerland, most cantons apply one level – the priority function (Ger. Vorrangfunktion, Kantonale Waldplanung 2007), and some also a second level – the secondary function (Ger. Nebenfunktion). Forest functions are ranked between each other, which differs from the Slovenian approach, where multiple functions can have the first level of importance in the same forest area. The approach used in Switzerland clearly defines priorities between functions, which is important for prescribing management regimes, since management regimes associated with each function might not be completely compatible.

301

T. Simončič and A. Bončina

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

Respondents identified several reasons why forest function areas are an important tool in the practice of multi-objective forest management, from identifying conflict areas and setting management priorities to collaboration with stakeholders and argumentation in spatial planning. The diverse importance of forest functions should show in the designation process; the designation criteria should be simple enough to articulate various demands on forests, but also clear and transparent, especially if state funds are available for adjustments of forest management to support public services.

In such cases, the participation of forest owners and other relevant stakeholders becomes even more important. Good examples are protection forests in Switzerland that are strongly supported by cantonal or even national budgets (Schmidt 2010).

Surprisingly, respondents placed the highest importance on the influence of forest function areas on spatial planning, which is probably connected to the dramatic land use changes during the last decade triggered by European Union subsidies for agricultural lands. Forest planners decide if small scale conversions from forest to agricultural lands are admissible, and in such cases forest function areas become important arguments against deforestation (Bončina and Matijašić 2010). The respondents assigned relatively low importance to forest function areas for implementing forest management, despite the mandate from the state that forest function areas of first level of importance should determine forest management regimes (ZG 1993). This could be connected to the lack of state funds to support adjusted management in both public and private forests, which is a weakness identified by foresters in this and other surveys (e.g. Bončina et al. 2014). In addition, many respondents criticized complicated forest function maps containing a large number of overlapping forest functions, which could be another reason for the relatively small management importance of forest function areas. Furthermore, large forest areas are ranked with the second level of importance, which has very little or even no influence on forest management regimes (Simončič and Bončina 2012). Experiences show that clear prioritization of forest function areas, which are not determined for the entire forest area but focused on areas with specific importance for multi-objective forest management, provides a much better basis for setting management measures, and at the same time significantly contributes to mitigating conflicts between forest uses (e.g.

Hanewinkel 2011).

Recently, the evolving concept of »ecosystem services« (EUSTAFOR and Patterson 2011) has been seen as a way forward to overcome some of the shortcom-

302

ings of the concept of forest functions (Bürger-Arndt 2013), as it improves communication with the public, evaluates non-monetary functions (services) and consequently establishes a reward system for those providing public services (Weiss et al. 2011). However, important conceptual differences between the two concepts exist (e.g. Pistorius et al. 2012) and should be considered when adopting the language of ecosystem services in the concept of forest functions.

We partly confirmed that forestry experts have different perceptions of the concept of forest functions.

Planners and scientists were more critical of classification and mapping compared to local foresters. This seems to be the result of the great deal of time planners need to spend in elaborating forest function maps. On the other hand, local foresters were more critical of the designation of forest function areas without the participation of private owners. Provision of public forest services may be more difficult to apply in private forests due to the divergent objectives of forest owners (Ficko and Bončina 2013) or the need to compensate for trade-offs between private and public demands (Cubbage et al. 2007), and local foresters directly involved with private owners may be much more aware of these issues.

6. Conclusion Forest functions remain an important tool in the practice of multi-objective forest management in Slovenia; they are the basis for presenting the public importance of forests, they play a large role in preventing deforestation of forestland, and are to some degree important for spatial differentiation of management measures and for financial support for providing public services. Improving the classification scheme and mapping of forest functions is a relevant task; however, changing the focus from »mapping« to management activities, which are necessary for providing the desired services, might be even more important. Nevertheless, the importance of forest function areas for multi-objective forest management will strongly depend on their overall integration into forest and environmental policy, especially the available financial support of the state.

Acknowledgments The authors wish to thank the Pahernik foundation for supporting the publishing of the results. We would also like to thank Christian Rosset, Juro Čavlović and Dragan Matijašić for comments on an early draft.

Croat. j. for. eng. 36(2015)2

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

7. References Anko, B., 1995: Funkcije gozda. Skripta, Biotehniška fakulteta Ljubljana, Oddelek za gozdarstvo in obnovljive gozde vire, 181 p.

Bachmann, P., 2005: Forstliche Planung I/III. Skript für die Lehrververanstaltungen »Gründzüge der Plannung«, »Forstliche Betriebsplannung« und »Waldentwicklungsplanung«. Professur Forsteinrichtung und Waldwachstum ETH Zürich, 346 p.

Berger, F., Rey, F., 2004: Mountain protection forests against natural hazards and risks: new French developments by integrating forests in risk zoning. Natural Hazards 33(3): 395–404.

Blum, A., Brandl, H., Oesten, G., Rätz, T., Schanz, H., Schmidt, S., Vogel, G., 1996: Wirkungen des Waldes und Leistungen der Forstwirtschaft. Allgemeine Forstzeitschrift 51(1): 22–26.

Bončina, A., 2001: Concept of sustainable forest management evaluation in forestry planning at the forest management unit level: some experiences, problems and suggestions from Slovenian forestry. In: Criteria and indicators for sustainable forest management at forest management unit level. Franc, A. (ed.). EFI Proceedings 38: 247–260.

T. Simončič and A. Bončina

EUSTAFOR, Patterson, T., 2011: Ecosystem Services in European State Forests. European State Forest Association, Brussels, 40 p.

Ficko, A., Bončina, A., 2013: Probabilistic typology of management decision making in private forest properties. Forest Policy and Economics 27: 34−43.

Forst Act, 1975: Forest act of the Republic of Austria. BGBl, Nr. 440/1975.

Field, A., 2000: Discovering statistics using SPSS for Windows 2000. Sage Publications, London, 496 p.

Fürst, W., Schaffer, H., 2000: Konzept des neuen Österreichischen Waldentwicklungsgesamtplanes »WEP-AustriaDigital«. FBVA-Berichte, Nr. 112.

Glück, P., 1982: Das Elend der Kielwassertheorie. Internationaler Holzmarkt 73(5): 15–18.

Gross, J., 2007: Waldfunktionen im Land Brandenburg. Landesforstanstalt, Volume 34 of Eberswalder forstliche Schriftenreihe, 47 p.

Hanewinkel, M., 2011: Multifunktionalität des Waldes. Forum für Wissen: 7–14.

Hasel, K., 1968: Die Zukunft der Deutschen Forstwirtschaft.

In: Jahrbuch des Deutschen Forstvereins. Landwirtschaftsverlag G.m.b.H., Hiltrup bei Münster (Westf.), 36–60.

Bončina, A., Matijašić, D., 2010: Gozdni prostor: načrtovanje, raba, nasprotja. Zbornik prispevkov, Biotehniška fakulteta Ljubljana, Oddelek za gozdarstvo in obnovljive gozdne vire, 65 p.

Hill, T., Lewicki, P., 2007: Statistics: Methods and Applications. StatSoft, Tulsa, OK, 832 p.

Bončina, A., Matijašić, D., Simončič, T., Bogovič, M., Devjak, T., Havliček, R., Klopčič, M., Pirnat, J., Pisek, R., Poljanec, A., 2014: Razvoj koncepta večnamenskega gospodarjenja z gozdovi: funkcije gozda, ekosistemske storitve in prednostna območja. Gozdarski vestnik 72(1): 44–46.

Kantonale Waldplanung Appenzell Innerrhoden, 2007: Waldfunktionen-kartiering, Bericht. 36 p.

Borchers, J., 2010: Segregation versus Multifunktionalität in der Forstwirtschaft. Forst und Holz 65(7/8): 44–49.

Likert, R., 1932: A Technique for the Measurement of Attitudes. New York: Archives of Psychology 140: 1–55.

Buttoud, G., 2002: Multipurpose management of mountain forests: which approaches? Forest Policy and Economics 4(2): 83–87.

Mantel, K., 1990: Wald und Forst in der Geschichte: ein Lehr und Handbuch. Hanover, Germany, Schaper, 511 p.

BUWAL, 1996: Fallbeispiele zur überbetrieblichen forstlichen Planung. Bundesamt für Umwelt, Wald und Landschaft, Bern, Switzerland, 77 p.

Bürger-Arndt, R., 2012: Kategorien, Indikatoren und Datenlage der Waldfunktionenkartierung. In: Bürger-Arndt, R., Ohse, B., Meyer, K., Höltermann, A. (eds.). Ökosystemdienstleistungen von Wäldern. Workshopbericht, BfN-Skripten 320, 51–55.

Bürger-Arndt, R., 2013: Waldfunktionen und Ökosystemleistungen im wissenschaftlichen Diskurs. In: Ring, I. (ed.). Der Nutzen von Ökonomie und Ökosystemleistungen für die Naturschutzpraxis. Workshop III, Wälder.-BfN-Skripten 334, 24–29.

Cubbage, F., Harou, P., Sillsa, E., 2007: Policy instruments to enhance multi-functional forest management. Forest Policy and Economics 9(7): 833−851.

Dieterich, V., 1953: Forstwirtschaftspolitik – Eine Einführung. Paul Parey, Hamburg und Berlin, 398 p.

Croat. j. for. eng. 36(2015)2

IBM Corp, 2011: IBM SPSS Statistics for Windows, Version 20.0. IBM Corp, Armonk, NY.

Krott, M., 1985: Zu den Waldfunktionen als Instrument der forstpolitischen Wissenschaft und Praxis. Centralblatt für das gesamte Forstwesen 102(1): 1–28.

Mavsar, R., Japelj, A., Kovac, M., 2013: Trade-offs between fire prevention and provision of ecosystem services in Slovenia. Forest Policy and Economics 29(0): 62–69.

Norman, G., 2010: Likert scales, levels of measurement and the »laws« of statistics. Advances in Health Science Education 15(5): 625–632.

Pistorius, T., Schaich, H., Winkel, G., Plieninger, T., Bieling, C., Konold, W., Volz, K.R., 2012: Lessons for REDDplus: A comparative analysis of the German discourse on forest functions and the global ecosystem services debate. Forest Policy and Economics 18: 4–12.

Planinšek, Š., Pirnat, J., 2012: Predlogi za izboljšanje sistema funkcij gozdov v Sloveniji. Gozdarski vestnik 70(5/6): 276–283.

Pravilnik o gozdnogospodarskih in gozdnogojitvenih načrtih, 1998: Ur. l. RS, št. 5/1998.

Pravilnik o načrtih za gospodarjenje z gozdovi in upravljanje z divjadjo, 2010: Ur. l. RS, št. 91/2010.

Riegert, C., Bader, A., 2010: German cultural history of forestry and forest functions since the early 19th century. In:

303

T. Simončič and A. Bončina

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

Cleveland, C. J. (ed.). Encyclopedia of Earth. Cleveland, Washington D.C. http://www.eoearth.org/article/German_ cultural_history_of_forestry_and_forest_functions_since_ the_early_19th_century.

Rupf, H., 1960: Wald und Mensch im Geschehen der Gegenwart. Allgemeine Forstzeitschrift 38(15): 545–552.

Ruppert-Winkel, C., Winkel, G., 2009: Hidden in the Woods?

Meaning, determining, and practicing of common welfare in the case of the German public forests. European Journal of Forest Research 130(3): 421–434.

Schmidt, R., 2010: Marketing of protection services – the example of Canton Bern. Schweizerische Zeitschrift für Forstwesen 161(9): 379–383.

Schulzke, R., Stoll, S., 2008: Forests and forestry in Hesse, Germany: Meeting the challenge of multipurpose forestry.

In: Correiro, M.M., Song, Y.C., Wu, J. (eds.). Ecology, planning and management of urban forests. Springer, 293–300.

SFS 2012: Forest inventory database. Slovenia Forest Service, Ljubljana, Slovenia.

SFS 2014: Forest inventory database. Slovenia Forest Service, Ljubljana, Slovenia.

Simončič, T., Bončina, A., 2012: Koncept prednostnih območij pri načrtovanju večnamenskega gospodarjenja z gozdovi.

Gozdarski vestnik 70(10): 415–428.

Simončič, T., Bončina, A., Rosset, C., Binder, F., De Meo, I., Čavlović, J., Gal, J., Matijašić, D., Schneider, J., Singer, F., Sitko, R., 2013: Importance of priority areas for multi-objective forest planning: a Central European perspective. International Forestry Review 15(4): 509–523.

Simončič, T., Spies, T.A., Deal, R.L., Bončina, A., 2015: A conceptual framework for characterizing forest areas with high societal values: Experiences from the Pacific Northwest of USA and Central Europe. Environmental Management 56(1): 127–143.

Volk, H., 1987: Umweltvorsorge durch Waldbiotopkartierung. Zur notwendigen Fortentwicklung der Waldfunktionenkartierung. Allgemeine Forstzeitschrift 42(22): 565– 568.

Volk, H., Schirmer, C., 2003: Leitfaden zur Kartierung der Schutz- und Erholungsfunktionen des Waldes (Waldfunktionskartierung) (WFK), 107 p.

Waldfunktionen Kartierung, 2004: Grundsätze und Verfahren zur Erfassung der besonderen Schutz und Erholungsfuntionen des Waldes im Freistaat Sachsen. Das Lebensministerium Freistaat Sachsen, Landesforstpräsidium.

WEP (Waldentwicklungsplan), 2006: Richtlinie über Inhalt und Ausgestaltung - Fassung 2006. Wien, Bundesministerium für Land und Forstwirtschaft, Umwelt und Wasserwirtschaft.

Weiss, G., Schönenberger, W., Weber, M., 2002: New silvicultural and planning paradigms for integrated mountain forest management. In: Presentation at OEFM Research Course on the formulation of integrated management plans for mountain forests, 30 June – 5 July, Bardoneccia, Italy.

Weiss, G., Ramcilovic, S., Mavsar, R., 2011: Financing mechanisms for forest ecosystem services in Europe and their implications for forest governance. Allgemeine Forst und Jagdzeitschrift 182(5/6): 61–69.

ZG, 1993: Zakon o gozdovih. Ur. l. RS, št. 30–1299/1993.

Authors’ address:

Received: October 21, 2014 Accepted: April 8, 2015

304

Tina Simončič* e-mail: tina.simoncic@bf.uni-lj.si Prof. Andrej Bončina, PhD. e-mail: andrej.boncina@bf.uni-lj.si University of Ljubljana Biotechnical Faculty Department of Forestry and Renewable Forest Resources Večna pot 83 1000 Ljubljana SLOVENIA * Corresponding author Croat. j. for. eng. 36(2015)2

Are Forest Functions a Useful Tool for Multi-objective Forest Management ... (293–305)

T. Simončič and A. Bončina

Appendix Table 8 General evaluation of the current concept of forest functions Statement

Likert scale* 1

2

3

4

5

The system of financial instruments for adjusted forest management on forest function areas is sufficient 29.6 42.0 24.7

3.7

0.0

2.02±0.83

0.003

Forest function map is too complicated due to a large number of forest functions**

27.8 37.7 25.9

8.6

0.0

2.15±0.93

0.000

Forest function map is clear due to overlapping of forest function areas

21.6 40.7 27.8

8.6

1.2

2.27±0.94

–

Monitoring of management measures supporting forest functions is sufficient

13.6 45.7 30.2

9.3

1.2

2.39±0.88

–

Participation of forest owners in the designation of forest function areas is not sufficient**

8.6

42.6 32.7 14.8

1.2

2.57±0.89

–

Stakeholders’ participation in the designation of forest function areas is sufficient

4.9

39.5 43.2 12.3

0.0

2.63±0.76

–

Forest function areas are uncritically adopted from other institutions (e.g. Natura 2000 sites)**

11.7 25.3 45.7 14.2

3.1

2.72±0.96

–

Descriptions of forest functions in forest plans are too extensive**

8.0

27.2 42.0 22.2

0.6

2.80±0.90

–

Descriptions of forest functions in forest plans are too general**

11.1 21.0 35.8 30.2

1.9

2.91±1.01

–

Forest function map is not useful for planning management measures**

6.8

13.6 47.5 29.0

3.1

2.92±0.91

–

Management measures on forest function areas are clearly defined in management plans

3.1

22.2 51.2 21.0

2.5

2.98±0.81

0.007

Information on forest function areas is not readily accessible to the public**

2.5

17.9 40.1 29.0 10.5

3.27±0.96

–

Forestry experts have enough/sufficient competences in designating forest function areas

3.1

13.6 37.7 35.2 10.5

3.36±0.95

–

Forest function map is useful for collaboration in spatial planning

1.2

8.6

38.3 45.7

6.2

3.47±0.79

–

The ranking levels of importance of forest functions are important for setting management priorities

1.2

8.6

33.3 48.8

8.0

3.54±0.81

–

Forest function areas should be designated only in agreement with forest owners**

1.2

9.3

24.7 37.7 27.2

3.80±0.98

0.018

Forest function areas should not be designated in private forests**

0.6

2.5

10.5 42.6 43.8

4.27±0.79

0.001

Avg.±st. dev. P-value

* 1 – complete dissatisfaction with the system; 5 – complete satisfaction with the system **Reverse coding applied

Croat. j. for. eng. 36(2015)2

305

Preliminary note

Combination of Artificial Neural Network with Multispectral Remote Sensing Data as Applied in Site Quality Evaluation in Inner Mongolia Fei Yan, Yinxi Gong, Zhongke Feng Abstract While abundant ground surface and site information is included in multispectral remote sensing data, traditional site quality evaluation system mainly uses artificial ground surface survey data. To construct an effective site quality evaluation system, this paper, with Wangyedian Forest Farm in Inner Mongolia as the object of study, has adopted an improved back propagation artificial neural network (BPANN) model based on a combination of multispectral remote sensing and surface survey data of the zone. With dahurian larch as an example, a neural network model based on a combination of remote sensing spectrum factor, site index and site factors has been constructed, which, applied in the site quality evaluation of sub compartments of the studied zone, has led to an optimized geographical position prediction model with an accuracy of 95.36%, and an increase of 9.83% as compared with neural network model based on traditional sub compartment survey data. The result indicates that multispectral remote sensing data is very suitable for forest site quality evaluation. Besides, the improved BP neural system features ideal accuracy of prediction, which testifies to the effectiveness and advantage of the methodology described in this paper.

Keywords: site quality evaluation, multispectral remote sensing, neural network

1. Introduction The growth and productivity of forest is closely related to the environmental conditions of the site and their quality, hence better understanding and management of its health requires accurate knowledge of the relationship between environmental factors and forest growth. Site refers to the total of external environmental conditions, which considerably influence the growth and development of the forest within a given space and is constituted of four categories, namely: climate conditions, topographical conditions, soil conditions and biological conditions (Zhang et al. 1992, Meng 1996, Enset al. 2013). Site quality evaluation refers to the judgment and prediction of the suitability for forestry or potential productivity of a forest site, hence quantifying the productive potential of the land (Skovsgaard and Vanclay 2008, Bruce 1981).

Croat. j. for. eng. 36(2015)2

The evaluation of site quality included two methods, direct evaluation method and indirect evaluation method. According to many researches, the most important and commonly used method is to use site index as an indicator to evaluated site quality (Zhang and You 1998). Site index was defined as the average height of dominant trees at the specific benchmark age in forest stand. However, using this indicator was extremely limited where the tree height could not be measured in non-forested land and of multiple tree species (Guo et al. 2012). Thus, the method of multivariate statistic mathematical modelling was adopted

by Carmean, Louwa, Curt et al. to retrieve the relationship between site index and site factors, using site factors for indirect evaluation of site index. This method was extensively used as it has provided an effective solution for the difficulty in uniform evaluation of for-

307

Fei Yan et al.

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

ested and non-forested land and of multiple tree species (Louwaand Scholes 2002, Carmean 1978, Curt et al. 2001, Farrelly 2011). However, many weaknesses in the design of analysis methods and the accuracy of prediction remain to be solved. For example, the use of stepwise regression analysis tends to incur biased estimation or ineffective prediction (Swenson 2005).

Main component analysis method may effectively simplify the data structure, but the accumulated contribution rate of the first two main component factors is usually below 70% (Huang et al. 2006); while quantification theory applied analysis method enables effective handling of discrete property factors but relies on enormous data accumulated through long term observation (Waring 2006).

The above problems are either directly or indirectly resulted from the non-linear complicated relationship between various site factors. Artificial neural networks (ANN), on the other hand, have attracted much attention for their unique properties including self-organization, self-adaptation, self-learning and distribution parallel processing (Mutluet et al. 2012, López 2001, Francl and Panigrahi 1997). Especially, BP (backward propagation) neural network model, a type of feed forward neural network constituted by non-linear transfer function neuron, enables effective prediction function by adopting a learning algorithm of error backward propagation (Mutlu et al. 2012). However, few studies have been conducted in the application of neural network in site quality evaluation up to now. Luan Zhaoping has applied BP neural network in studying the extent of influence by various site environmental factors on the height of wild Vaccinium uliginosum (Luan 2011).

Huang Jiarong et al. have chosen Pinus massoniana as an example in modelling the relationship between site factors and site index with the application of BP neural network for non-forested site quality prediction evaluation, with an average accuracy of 86.06%, higher than traditional multivariate regression model (Huang et al.

2006). All these site studies with BP neural network application have used traditional sub lot survey data, merely taking environmental factors of the site into consideration, without including bio vegetation factors, which most directly reflect site growth conditions into evaluation system, causing the prediction accuracy of the model to decrease. Meanwhile, most sub lot survey data are discrete non numerical data, causing convergence performance of model training and stability performance of prediction to deteriorate. Multispectrum remote sensing data include abundant ground surface bio vegetation information at low cost and with desirable availability, thus representing an improvement over the high demand for human and financial resourc-

308

es by traditional survey (Wu and Peng 2011, Niall et al.

2011). Ma Mingdong et al. have studied the correlation between remote sensing spectrum of plants and site index and established the site index inversion model of single vegetation index (Ma et al. 2006). However, the applicability of this method is constrained by the lack of universality of this estimation model semi-empirical formula, given the differences of surface and natural properties.

Therefore, for stable and effective forest site quality assessment, multispectral remote sensing data is introduced with 6 vegetation indexes closely related to forest productivity such as biological vegetation factor, which, combined with geographical topographical factor and soil factor, an improved BP neutral network with sensitiveness analysis and self-adapted lr decreasing gradient is applied in site quality evaluation.

Four plans are proposed based on different combinations of neutral network models and input data sets in predicting the site index. Accuracy and effectiveness of each plan is then analyzed and assessed for an optimized result, aiming at providing a more effective method for forest site quality assessment.

2. Materials and methods 2.1 Background information of the study site and data retrieval Wangyedian Experimental Forest Farm is located in the southwest of Kalaqin Qi in Inner Mongolia, Chifeng city at 118°09′–118°30′E, 41°21′–41°39′N with the central GPS coordinates 118.3825E, 41.5543N, between 500 and 1890 m above sea level with over 85% of its land classified as mountainous. The studied area is located in the mid temperate continental monsoon climate zone, with an average temperature of 6.2 ºC and 100–145 frost free days annually. Its soil types mainly include brown soil, cinnamon soil, meadow soil, black soil in mountainous area, among which brown soil covers most of the area. The annual hours of sunlight range between 2700 and 2900 hours, and main tree species include Dahurian larch, Pinus tabuliformis, Birch, Populus davidiana, Xylosma racemosum, etc. The name of the authority that issued the permit for each location is forestry bureau of Chifeng city in Inner Mongolia.

We had a formal contract to help them to carry out forest inventory in Wangyedian Farm Land and they gave us the authority to publish any research articles using the sub lot data. In addition, there were no specific permissions required for these locations or activities. The field studies did not involve endangered or protected species.

Croat. j. for. eng. 36(2015)2

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

Fei Yan et al.

According to the researches of site quality evaluation from Wu (2011), He (2012), Ma (2006) and Zhang (1998), TM remote sensing images were the traditional and best option for site quality study. Thus, the current research, with sub lot dominated by Dahurian larches as its object, uses TM images covering the area in April, 2010, with sun elevation angle of 64°, solar azimuth of 135°, with no enormous quantities of clouds or shadows guaranteeing desirable image quality. 1:10.000 topographical map is used for geometrical correction of image, with pixel root mean square error of 0.2, meeting the requirement for accuracy. Sub lot survey data, including site topographical information (altitude, aspect and slope) and soil information (soil type, soil thickness and humus layer thickness) with sub lot as unit of the studied area is also applied in the study.

ÞR VI (Ratio Vegetation Index)

2.2 Site information retrieval

This vegetation index reflects the absorption of solar radiation by forest vegetation in photosynthesis as well as information of vegetation growth including vegetation growth rate (He 2012).

Through gray scale ratio between near infrared band and red light band, this vegetation index expresses their difference in reflection rate. For TM image, the calculation formula of RVI is:

RVI = TM 4 / TM 3

(2)

RVI shows strong sensitivity towards green plants, while indicating significant correlation with forest parameters including biomass, leaf area index (LAI) and chlorophyll content, etc. (He 2012). ÞN DVI (Normalized Difference Vegetation Index) For TM image, the calculation formula of NDVI is:

(3)

The scope of the study presented in this paper is limited to Wangyedian Forest Farm, where site climate conditions show little fluctuation. Therefore, topography, soil and biology factors are dominating factors in ÞG vi and Bvi calculated on the basis of K–T variation influencing the forest site changes in the studied area.

Gvi = −0.284TM 1 − 0.243TM 2 − 0.543TM 3 + 0.724TM 4 Factors of these three categories are selected in this pa per, and site factors of the study area are retrieved with Gvi = −0.284TM 1 − 0.243TM 2 − 0.543 TM 3 + 0.724TM 4 + 0.084TM 5 − 0.180TM 7 (4) multispectral remote sensing data in combination with forest resource sub lot data, with sub lot defined as unit.

Bvi = 0.303TM 1 + 0.279TM 2 + 0.474TM 3 + 0.558TM 4 +

2.2.1 Spectrum information retrieval Bvi = 0.303TM + 0.279TM 2 + 0.474 TM 3 + 0.558TM 4 + 0.508TM 5 + 0.186TM 7 (5) As the band combination value 1of multispectral remote sensing data is closely related to the growth of Spectrum information of vegetation and soil is ground surface vegetation, six representative vegetaseparated through K–T transformation, hence acquirtion indexes, i.e. difference vegetation index (DVI), ing Gvi and Bvi components, which accurately reflect ratio vegetation index (RVI), normalized difference the variation in spectrum properties of forest vegetavegetation index (NDVI), green vegetation information and soil (Wu and Peng 2011). tion (Gvi), bright vegetation information (Bvi) and transformed soil adjusted vegetation index (TSAVI) ÞT SAVI (Transformed Soil Adjusted Vegetation Index) are selected to retrieve the bio vegetation factors in the For TM image, the calculation formula of TSAVI is: studied area.

ÞD VI (Difference Vegetation Index) This index indicates the difference between near visible light red band and infrared band values. For TM image, the calculation formula of DVI is:

DVI = TM 4 − TM 3

(1)

Where: TM4

near infrared band emission rate;

TM3

red light band reflection rate.

This index effectively reflects the soil background of forest vegetation as well as changes in vegetation coverage (He 2012).

Croat. j. for. eng. 36(2015)2

TSAVI = (TM 4 − TM 3 − 0.5 ) / (TM 4 + T

TSAVI = (TM 4 − TM 3 − 0.5 ) / (TM 4 + TM 3 + 0.5 ) (6)

By adding soil adjustment coefficient, this vegetation index has corrected the sensitivity of NDVI towards soil background, while explaining the characteristics of optical property variation of the background (He 2012).

2.2.2 Topography and soil information retrieval Topography and soil information used in this study is retrieved from forest sub lot survey data.

Eight properties, namely aspect, slope, elevation, soil type, soil thickness and humus layer thickness, are

309

Fei Yan et al.

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

Table 1 Sub lot Sub lot ID

Aspect

Slope

Elevation, m

Soil type

Humus layer thickness

Soil thickness

283

Southeast

Mid slope

1300

Brown soil

4

Medium

281

Southeast

Level

1300

Brown soil

1

Medium

178

Northeast

Mild slope

1100

Brown soil

9

Medium

212

West

Slope

1200

Brown soil

3

Slight

9

South

Mid slope

1000

Brown soil

6

Thick

496

Northwest

Slope

1500

Brown soil

9

Medium

524

North

Slope

1000

Brown soil

9

Medium

826

Southeast

Steep slope

1500

Brown soil

3

Medium

retrieved from forest resource sub lot survey data sheet to constitute the Dahurian larch growth site information table; several examples are illustrated in Table 1 while all these site factors of 826 sub lots in this forest farm are presented in supplement data file.

3.1 BP neural network In the light of the complicated non-linear relationship between multiple site factors and site quality, BP neural network (Back Propagation Neural Network), a multi-layer feed forward neural network capable of predicting based on multi scale data sources is adopted in this study.

Fig. 1 Back propagation network structure

As illustrated in Fig. 1, in multilayer feed forward network, the first layer is input layer, L layer is output layer, the mid layer is hidden layer. Suppose the neuron of layer l (l = 1, 2…, L) is the connection weight

coefficient of network connection. The cost function is supposed as:

value of the ist neuron in layer l is

Wij( l ) (i=1,2,…, nl ;

j=1,2,…, nl−1 ), then the connection relation of this network is discovered.

( )

1

2 1 nL d ti − xti( L ) ∑ 2 i=1

(

)

(9)

For output layer L:

∂E ∂x(l) ∂s(l) ∂E = (l)t × ti(l) × ti(l) = −(dti − xti(l) ) f '(Sti(l) )xti(t −1) = −d ti(L) xti(L −1) (l) ∂Wij ∂xti ∂Sti ∂Wti 1+ e ∂Et ∂xti(l) ∂sti(l) ∂E (10) = −(dti − xti(l) ) f '(Sti(l) )xti(t −1) = −d ti(L) xti(L −1) i=1,2,  , nl ; j=1,2,  , nl−1 ; l =1,2,  ,L ( l ) = (l) × (l) × ∂Wij ∂xti ∂Sti ∂Wti(l)

xi( l ) = f Si( l ) =

n l−1

(

− S(i ) l

(7)

)

Si = ∑Wij( l ) x (j l−1) x0( l−1) = qi( l ) , Wi0( l ) = −1 ( l)

j=0

(8)

Input and output samples of the given t group are set as

Et =

T

( 0)  xt( 0) =  xt1( 0) xt(20)  xt,n 0  , dt =   d t1 d t 2  d t,

(t= 1, 2…, T).

na



T

BP network is trained with this sample set, and this process enables the relation mapping between input and output through learning and adjustment of weight

310

( L−1)

( L)

Calculation dti is conducted, dti followed by the recurrence based on the above formula. The rest l dti( l ) and ∂Et / ∂Wij( ) may be derived by analogy. The ( l) expression dti where recurrence continues until the ( l) total is calculated, contains derivatives f ′ S ti . As

( )

( )

f – for s function is hypothesized, the derivative can be calculated as follows:

( )

(

(l)

)

xti( l ) = f S ti( l ) = 1/ 1 + e − Sti

(11)

Croat. j. for. eng. 36(2015)2

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

( ) = 1+ e

f ′ S ti

l) i

− S ti( ) l

( ) ( l)

= f S ti

(l)

e − Sti

( l)

( )

− S ti( ) l

( )

( )

(

( l)

)

( l)

(12)

On this basis the following network model is constructed:

Wij

changes in output is defined as sensitivity causal pa-

)

n ( l) = f S ti 1 − f S ti  = xti 1 − xrameter (SCP). Suppose input vector is [ xi ]1 , where ti   th

( l)

1 − f S ti( l )  = xti( l ) 1 − xti( l )  

( l)

(

Fei Yan et al.

( k + 1) = Wij ( k ) + jDij ( k + 1) ( l)

( l)

t

Dij( l ) = ∑dti( l ) xti( l−1)

(13) (14)

t =1

 n l+1  dti( l ) =  ∑dk( l+1)Wki( l+1)  xti( l+1) 1 + xti( l)  k =1 

dti( L ) = d ti − xti( L ) xti( L ) 1 − xti( L )

l = L, L − 1, ,1; i = 1, 2, , nl ; j = 1, 2,, nl−1

(

(

) (

)

)

(15) (16)

With the method of universal approximation, approximation of any non-linear mapping relation is enabled.

3.2 Improved BP neural network The improved neural network proposed in this paper is based on back propagation algorithm and, targeting at characteristics of the studied data sets, processing mechanisms of sensitivity analysis and self-adaptive lr decreasing gradient are introduced to filter out less influential factors, thus ensuring learning and convergence speed of the network.

As an excessive number of input data set shall result in excessive scale of neural network, sensitivity analysis algorithm is applied in this study for sorting the importance of properties of input data sets; therefore, by eliminating less important properties, a minimized and optimized data set is retrieved, enhancing the categorization accuracy and efficiency of neural network.

Model for Sensitivity Analysis (SA) can be expressed as follows: , where xi is the ith property value of the model, sensitivity algorithm is the algorithm that allows each property to change within feasible scope of value, where the extent of influence of these changes on model output value is assessed and regarded as the sensitivity coefficient of such property. In sensitivity analysis, the quantified value of the relation between changes in input and Croat. j. for. eng. 36(2015)2

xi is the i property of such input vector and n is the

total number of samples. Therefore, for the output result f(x) of BP neural network, SCP is defined as follows: N

(

)

SCP = ∑ f ( xi ) − f xi +  ij / N i =1

(17)

Where  ij is the variation value and disturbed value in jth component of xi .

Based on sensitivity analysis, the properties showing weakest sensitivity in diversified forestry factor are excluded while the other properties showing strong sensitivity are preserved so as to reduce the number of dimensions of the original input data set, hence improving the training accuracy and efficiency of BP neural network.

The study adopts self-adaptive lr gradient descent method to adjust the learning speed during network training, hence accelerating the training speed of network and avoiding the problem of minimum. The algorithm procedures of this method are as below: Step 1: check if the corrected value of weight reduced error function; Step 2: if so, increase learning speed at fixed step (in which case the learning speed is insufficient); Step 3: if not, decrease learning speed at fixed step (in which case the learning speed has been excessively adjusted); Step 4: if new error value is lower than previous error value, increase learning speed at fixed step.

This mechanism ensures stable learning of the network at maximum learning speed, while maintaining descent error. Once excessive learning rate is the case, learning speed automatically reduces to maintain stability of error descent. The following equation is the self-adaptive lr descent adjustment model:

 1.05lr ( k ) , SSE ( k + 1) < SSE ( k )  lr ( k + 1) = 0.7lr ( k ) , SSE ( k + 1) > 1.04 SSE ( k )  lr ( k ) , other 

(18)

Where: lr(k)

learning step;

SSE(k) momentum factor.

Based on the above optimized algorithm and various site factors, four schemes are adopted in this paper

311

Fei Yan et al.

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307â&#x20AC;&#x201C;319)

for site index prediction using BP neural network. In this study, all of the schemes implementation and data analysis were completed with MatlaB 2010 software: Scheme I: BP neural network + sub lot survey factors (XB factors, XBF): with topographical factors (elevation, aspect, slope) and soil factors (soil type, soil thickness, humus thickness) in traditional sub lot survey data as input data set, with site index selected as factor of output layer. Classic BP neural network is applied in prediction.

Scheme II: BP neural network + multispectral remote sensing bio factors (RS factors, RSF) + sub lot survey factors: Multispectral remote sensing bio factors (DVI, RVI, NDVI, Gvi, Bvi, TSAVI) are combined with topographical factors in traditional sub lot survey data (elevation, aspect, slope) and soifactors (soil type, soil thickness, humus layer thickness) as input data set, with status index selected as factor of output layer.

Classic BP neural network is applied in prediction.

Scheme III: Improved BP neural network + sub lot survey factors: Output and input data sets are identical with Scheme I, with improved BP neural network applied in site index prediction.

Scheme IV: Improved BP neural network + multispectral remote sensing bio factors + sub lot survey factors: Output and input data sets are identical with Scheme II, with improved BP neural network applied in status index prediction.

Fig. 2 Multivariate factor sensitivity analysis

4. Results and discussion 4.1 Sensitivity analysis Sensitivity analysis is made on the basis of sub lot survey factor data set (hereinafter referred to as sub lot data set) and multispectral remote sensing biofactor + sub lot survey factor data set (hereinafter referred to as multivariate data set), wherein sub lot data set consists of six factors (elevation, aspect, slope, soil type, soil thickness, humus layer thickness) and multivariate data set consists of 12 factors (DVI, RVI, NDVI, Gvi, Bvi, TSAVI, elevation, aspect, slope, soil type, soil thickness, humus layer thickness). 100 groups of data are selected as training samples for sensitivity analysis to conclude the sensitivity of each factor in the two data sets, as illustrated in Fig. 2 and Fig. 3.

Fig. 2 shows that the sequential order of sensitivity of corresponding factors is: aspect, soil thickness, elevation, soil humus layer thickness, slope, soil type.

With the 1st and 2nd insignificant factors, namely soil type and slope factors removed, a filtered sub lot data set consisting of four site factors, i.e. aspect, soil

312

Fig. 3 Multivariate factor sensitivity analysis

thickness, elevation and soil humus layer thickness, is constructed, where the low sensitivity of soil type is the result of relatively uniform distribution of soil type in the studied area.

Fig. 3 shows that the sequential order of sensitivity of corresponding factors is: aspect, soil thickness, NDVI, RVI, elevation, DVI, Gvi, slope, Bvi, soil humus layer thickness, TSAVI, soil type. Obviously, the variCroat. j. for. eng. 36(2015)2

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307â&#x20AC;&#x201C;319)

Fei Yan et al.

Fig. 4 BP neural network model of each scheme

Fig. 5 Curve of training MSE for each scheme Croat. j. for. eng. 36(2015)2

313

Fei Yan et al.

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307â&#x20AC;&#x201C;319)

Fig. 6 Fitting curves of prediction accuracy ous vegetation indexes acquired from multispectral remote sensing data show higher sensitivity.

With five factors: slope, Bvi, soil humus layer thickness, TSAVI and soil type, which show low sensitivity removed, a filtered multivariate factor data set is concluded, including aspect, soil thickness, NDVI, RVI, elevation DVI and Gvi. According to the results of sensitivity analysis, the input data sets of Scheme III and Scheme IV were determined. Four sub lot survey factors were selected for Scheme III and seven multispectral remote sensing bio

314

factors + sub lot survey factors were used for Scheme IV, separately (Fig. 4). The above sensitivity analysis mechanism has effectively removed factors with weak or no correlation, hence providing a solution to the redundant size and low efficiency of neural network, and improving the prediction efficiency and accuracy of neural network.

4.2 Parameter definition and modelling As the prediction of site index essentially boils down to function fitting, therefore, an artificial neural Croat. j. for. eng. 36(2015)2

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

Fei Yan et al.

Table 2 Predicted results of comparison of four sc Scheme

Datas

Model

MSE –3

Average accuracy, %

1

XBF

BP

3.31e

85.53

2

XBF+RSF

BP

8.27e–13

90.97

–10

88.71

3

XBF

Improved BP

9.14e

4

XBF+RSF

Improved BP

2.77e–14

95.36

–3

85.53

–13

90.97

1 2

XBF XBF+RSF

BP BP

3.31e 8.27e

MSE means square error

Fig. 7 Correlation analysis between predictive value and calculated value Croat. j. for. eng. 36(2015)2

315

Fei Yan et al.

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

network with 3 layer topology consisting of two hidden layers and one output layer is adopted in this study, where the constructed BP neural network model has selected 100 record entries from the above input data set as learning samples, while the number of neurons is adjusted based on the convergence of evaluation network training in order to determine the number of neurons of each hidden layer. To reproduce modelling results, random seeds are input in the neural network model of each scheme to determine the connection weight [1; 0; 0], connection bias [1; 1; 1] and layer connection weight [000; 100; 010] of the network.

With Msereg as performance function and Initlay as initialization function, BP neural network models are illustrated in Fig. 4 based on the four schemes.

The above illustration shows that the hidden layer and output layer in each layer of neural propagation function of each model are both Tansig neurons.

4.3 Training result analysis Training is conducted with network initial values set on the basis of the above parameters. Fig. 5 illustrates the error variation curves of the four schemes, from which the variation law between numbers of training and network output error of each scheme is determined.

It is seen that the training convergence speed of Scheme III and Scheme IV, where improved neural network is adopted, is obviously more desirable than Scheme I and Scheme II. Respective comparison of Scheme I and Scheme III, using the same category of input data set, shows that the optimal prediction accuracy of Scheme III is 9.14e–10, considerably higher than 3.31e–3 in Scheme I; while a comparison between Scheme II and Scheme IV shows that optimal prediction accuracy has increased from 8.27e–13 to 2.27e–14, proving the effectiveness of the improved neural network as proposed in this paper.

Fig. 6 and Fig. 7 illustrate the test results of 100 groups of data in the four schemes as well as correlation analysis of predictive value and calculated value.

Table 2 is based on a comparison of the results of the above neural network experiments.

Results of the four schemes are compared. From the perspective of input data set, the average prediction accuracy of Scheme II and Scheme IV has both exceeded 90%, significantly higher than Scheme I and Scheme III, indicating that the introduction of multispectral remote sensing biological factor shall significantly enhance the prediction accuracy of status index. Analyzed from the perspective of neural network application, the prediction accuracy of Scheme

316

III and Scheme VI is 88.71% and 95.36%, respectively, a significant improvement from 85.53% and 90.97%, respectively, of Scheme I and Scheme II. Meanwhile, the model convergence speed of Scheme III and Scheme VI is 34 and 40, respectively, notably more desirable than 38 and 77, respectively. of Scheme I and Scheme II, proving that improved BP neural network applied in this paper features better prediction accuracy and learning performance. With the optimized scheme determined as above, where multispectral remote sensing data is combined with sub lot survey data applied, the site index of Dahurian larch sub lot in Wangyedian Forest Farm in Inner Mongolia is predicted using ArcGIS 10.2 software.

Eventual prediction results obtained in this paper are compared with the test data and illustrated in Fig. 8.

The legend of Fig. 8 presents the site index, which means the average height of dominant tree larches at the specific benchmark age in forest stand. All the test data of site index was provided by Wangyedian Forest Farm.

5. Conclusions This paper studies the site quality assessment with multispectral remote sensing data combined with sub lot survey data of Dahurian larch in Wangyedian Natural Reserve in Inner Mongolia. Unlike the traditional method of using Richards growth function to build guiding curve model, BP neural network, which is capable of reflecting more complicated non-linear relationships, is used in the study to predict the site index.

Meanwhile, based on the characteristics of forest resources data, an improved BP neural network model is proposed to improve the prediction accuracy of status index and training speed of network. To obtain the most effective site quality evaluation system, site index prediction is conducted by combining different input data set and neural network models in this study, formulating four site quality evaluation schemes, whose prediction accuracy and performance are compared and analyzed.

The study shows that among the four schemes, the Dahurian larch site index prediction model determined by improved BP neural network with multispectral remote sensing data plus sub lot survey data applied features highest prediction accuracy of 95.36%. While the topographical and soil factors in traditional sub lot survey data and classic BP neural network was applied in prediction, the prediction accuracy was only up to 85.53%, which means an increase of 9.83% was achieved. Meanwhile, a comparison between improved BP neural network and classic BP neural netCroat. j. for. eng. 36(2015)2

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307â&#x20AC;&#x201C;319)

Fei Yan et al.

Fig. 8 Simulated results compared with test results; (a) is test result; (b) is prediction result

work indicates that the convergence speed of network training has effectively increased in improved BP neural network. The model convergence speed of Scheme III and Scheme VI is 34 and 40, respectively, notably more desirable than 38 and 77, respectively, of Scheme I and Scheme II. Results of the study indicate that multispectral remote sensing data is highly applicable in forest site quality evaluation, as it has expanded the quantity of information on site factors, while ensuring sufficient Croat. j. for. eng. 36(2015)2

time dimension with potentials for prediction over large areas, thus capable of providing effective evidence for forest site quality evaluation.

Further study shall develop a technical system of forest site quality evaluation fully based on remote sensing data, while using multispectral remote sensing data to retrieve soil and topographical factors of forests, hence reducing the cost for artificial sub lot survey, while increasing the extent and scope of forest site quality evaluation.

317

Fei Yan et al.

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

Acknowledgements Financial support for this study was provided by the Fundamental Research Funds for the Central Universities NO. BLX2014-10 and National Key Technology R&D Programme (No. 2012BAH34B01). We are grateful to the undergraduate students and staff of the Laboratory of Forest Management and »3S« technology, Beijing Forestry University, and for the great support of Wangyedian Forest Farm in Inner Mongolia.

6. Reference Bruce, F., 1981: The use of overlays in site quality mapping.

Canadian Journal of Forest Research 11(2): 362–370.

search Article. Agricultural and Forest Meteorology 107(4): 279–291.

Louwa, J.H., Scholes, M., 2002: Forest site classification and evaluation: a South African perspective. Forest Ecology and Management 171(1–2): 153–168.

Luan, Z.P., 2011: Study on growth site conditions of wild Benedict bilberry based on the BP neural network. Shandong Forestry Science and Technology 195(4): 11–16.

Ma, M.D., Jiang, H., Liu, S.R., 2006: The preliminary analysis of forest ecosystem site index using remote sensed data. Acta Ecologica Sinica 26(9): 2810–2816.

Meng, X.Y., 1996: Forest measurements. The second edition.

Beijing: China Forestry Publishing House 99–106.

Carmean, W.H., 1978: Site index curves for northern hardwood sinnorthern Wisconsin and upper Michigan. New York, Department of Agriculture 3–18.

Mutlu, A.C., Boyaci, I.H., Genis, H.E., 2006: Prediction of wheat quality parameters using near infra-red spectroscopy and artificial neural networks. European Food Research and Technology 233(2): 267–274.

Curt, T., Bouchaud, M., Agrech, G., 2001: Predicting site index of Douglas fir plantations from ecological variables in the Massif Central area of France. Forest ecology and management 149(1): 61–74.

Niall, F., Áine, N.D., Maarten, N., 2011: Site index of Sitka spruce (Picea sitchensis) in relation to different measures of site quality in Ireland. Canadian Journal of Forest Research 41(2): 265–278.

Ens, J., Farrell R.E., Bélanger, N., 2013: Effects of edaphic conditions on site quality for Salix purpurea »Hotel« plantations across a large climatic gradient in Canada. New Forests 44(6): 899–918.

Skovsgaard, J.P., Vanclay, J.K., 2008: Forest site productivity: a review of the evolution of dendrometric concepts for evenaged stands. Forestry 81(1): 13–31.

Farrelly, N., Dhubháin, Á.N., Nieuwenhuisb, M., 2011: Site index of Sitka spruce (Picea sitchensis) in relation to different measures of site quality in Ireland. Canadian Journal of Forest Research 41(2): 265–278.

Francl, L.J., Panigrahi, S., 1997: Artificial neural network models of wheat leaf wetness. Original Research Article.

Agricultural and Forest Meteorology 88(1–4): 57–65.

Guo, Y.R., Wu, B.G., Liu, Y., 2012: Research Progress of Site Quality Evaluation. World Forestry Research 25(5): 47–52.

Huang, J.R., Ma, T.X., Wang, Y.M., 2006: Forest site evaluation model studies on the basis of BP Neural Network. Journal of Mountain Agriculture and Biology 25(6): 479–483.

He, Z.H., Chen, X.X., Liang, H., 2012: Study of Remote Sensing Monitoring of Karst Basin Water-holding Dynamic Changing Based On Vegetation Indices Taking Guizhou Province as a Case. Territory&Natural Resources Study 4: 48–51.

López, G., Rubio, M.A., Martı́nez, M., Batlles, F.J., 2001: Estimation of hourly global photo synthetically active radiation using artificial neural network models Original Re-

318

Swenson, J.J., Waring, R.H., Fan, W., 2005: Predicting site index with a physiologically based growth model across Oregon. Canadian Journal of Forest Research 35(7): 1697– 1707.

Waring, R.H., Milner, K.S., Jolly, W.M., 2006: Assessment of site index and forest growth capacity across the Pacific and Inland Northwest U.S.A. with a MODIS satellite-derived vegetation index. Forest Ecology and Management 228(1–3): 285–291.

Wu, J., Peng, D.L., 2011: Advances in researches on hyperspectral remote sensing forestry. Spectroscopy and Spectral Analysis 31(9): 2305–2312.

Wu, J., Peng, D.L., 2011: Tree Species information extraction of farmland returned to forests based on improved support vector machine algorithm. Spectroscopy and Spectral Analysis 31(4): 1038–1041.

Zhang, W.R., Sheng, W.T., Jiang, Y.X., 1992: Classification of Forest Site System in China. Forest Research 5(3): 251–262.

Zhang, X.L., You, X.X., 1998: Application of 3S Technology to Forest Site Type Classification and Site Quality Evaluation in Beijing. Journal of Remote Sensing 2(4): 292–296.

Croat. j. for. eng. 36(2015)2

Combination of Artificial Neural Network with Multispectral Remote Sensing Data ... (307–319)

Fei Yan et al.

Authors’ address: Fei Yan, PhD. e-mail: yf-perfect@163.com Prof. Zhongke Feng, PhD.* e-mail: 15210026573@163.com Beijing Forestry University Department of GIS, RS and GPS No.35 Tsinghua East Road Haidian District, Beijing CHINA

Received: February 20, 2014 Accepted: December 31, 2014 Croat. j. for. eng. 36(2015)2

Yinxi Gong, PhD. e-mail: 119266664@qq.com GIS Research and Development Center The First Institute of photogrammetry and Remote Sensing SBSM East Youyi Road 334#, Xi’an City, Shaanxi Province CHINA *Corresponding author

319

Subject review

Expanding Ground-based Harvesting onto Steep Terrain: A Review Rien Visser, Karl Stampfer Abstract Timber harvesting on steep terrain has always been, and will remain, a challenge in terms of economic viability, safety and environmental performance. For almost a century motor-manual felling coupled with cable yarding has been the most appropriate harvesting system, but new technologies and innovations have led to machines and systems being developed that are modernising the way we operate on steep terrain. Specifically, they provide the opportunity for the mechanisation of operations with proven improvements in both safety and cost-effectiveness. The additional development of cable-assist machines is potentially making a real step-change by expanding the operating range onto very steep slopes. This paper reviews these developments, the main engineering considerations of how cable-assist works, as well as the advances being made in terms of how such equipment is integrated into harvesting systems.

The review also includes analyses of the operating guidelines that are either in place or being developed to help implement the systems.

Keywords: timber harvesting, steep terrain, cable assist, harvester, forwarder

1. Introduction Cable yarding remains the mainstay of steep terrain harvesting operations (Samset 1985, Studier and Binkley 1974, Heinimann et al. 2001). However, cable yarding remains both expensive (Dykstra 1975, Raymond 2012) as well as hazardous relative to groundbased harvesting operations (Slappendel et al. 1993, Klun and Medved 2007). The safety issues arise from the manual activities still common in cable logging, including the felling by chainsaw and using chokersetters in the extraction phase (Kirk and Sullman 2001).

A high level of risks to forest workers operating these systems provides both the need, but also the potential benefits, of mechanising the manual aspects of this system (Bell 2002, Montorselli et al. 2010).

Cable yarders can be used in many different configurations (Studier and Binkley 1974) with distinct advantages and disadvantages for each configuration (Harrill and Visser 2012). There are several options for the mechanisation and automation of existing cable yarding systems. For example, the advent of radiocontrolled chokers reduces the need for the operator, Croat. j. for. eng. 36(2015)2

and/or a separate ÂťpolemanÂŤ, to unhook the stems once on the landing (Stampfer et al. 2010). The use of video cameras increase operator visibility for increased productivity (Evanson 2013) and can also be used for advanced training (Parker 2010). The development of motorised grapple carriages can reduce, but not eliminate, the need for choker-setters (McFadzean and Visser 2013). On the landings, the opportunity to integrate processors has long eliminated the need for the use of skid-workers. Combining elements of ground-based systems into cable yarding has also shown to have benefits (Stampfer and SteinmĂźller 2004, Acuna et al. 2011). For example a study by Visser and Stampfer (1998) showed a 40% increase in cable extraction productivity when using mechanised versus chainsaw felling. Although the concept of carriages that can fell and extract are being investigated, no commercial success has yet been reported. As such, some work processes of cable yarding operations will remain manual and/or motor-manual.

Ground-based harvesting has benefited significantly from mechanisation and many options for fully mechanised systems are available (MacDonald 1999).

321

R. Visser and K. Stampfer

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321â&#x20AC;&#x201C;331)

Fig. 1 Safe operating range of ground-based harvesting machines related to terrain slope (%) and soil bearing capacity, as measured by California Bearing Ratio (after Heinimann 1995) The two most common options are the Cut-to-Length (CTL) and the Whole-Tree (WT) extraction systems.

CTL, first made popular in Scandinavia, uses a combination of a harvester to fell and process the tree into logs, followed by a forwarder that will accumulate the log piles and bring it to roadside. WT relies on a felling machine to put the trees down (and possibly prebunch them), a skidder to extract them to a landing, and a processing machine to buck the stems into logs.

Both mechanised systems have proven advantages in terms of safety (Bell 2002, Axelsson 1998, Laflamme and Cloutier 1988) and productivity (Raymond 2010).

While modern fully mechanised ground-based systems are a default option for safe and productive harvesting, they have always been limited by terrain factors such as slope (Olund 2001, Alam et al. 2013, Strandgard et al. 2014), soil strength and/or roughness (Amishev et al. 2009, Visser 2013). In the early 1970s, ground-based extraction machines had made considerable progress whereas mechanised felling and processing technology only just emerged on gentle terrain (Carson 1983). Feasibility limits were fixed for downhill skidding at a slope gradient of 50% for wheeled skidders and 60% for crawler tractors, depending on

322

surface roughness (FAO/ECE/ILO 1971). Practical experience later demonstrated that those limits had to be reduced in order to keep soil erosion within acceptable limits (Heinimann 1999). Reported slope limits of 30% and 40% for wheeled and tracked machines, respectively, were related primarily to machine traction and soil erosion, and these values have since been presented and propagated in many subsequent reports (e.g.

Arola et al. 1981) and guidelines (i.e. NZDOL 1999).

A study by Visser and Berkett (2015), recording the actual machine slope on 22 different operations (18 in New Zealand, 4 in Europe), showed that, under normal operating conditions, machines exceed established slope guidelines frequently and for extended periods of time. This, as well as anecdotal evidence from many countries, suggests that equipment advances have far outpaced operating guidelines.

Actual guidance on slope limits, based on either science or experience, is rare. Many guidelines refer to manufacturerâ&#x20AC;&#x2122;s specifications, yet few of the major forestry equipment manufacturers provide slope and/or operating limits for their purpose built machinery.

Komatsu has recently published operating guidelines that indicate a slope limit of 55% when using winch Croat. j. for. eng. 36(2015)2

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321–331)

assist (Komatsu 2015). Cavalli (2015) surmised that wheeled machines with chains or bands might have an upper limit of 45%, integral track machines up to 60%, and that tethered machines should be able to operate up to a range of 75 to 85% slope.

Work by European researchers expanded the considerations and understanding of steep terrain machinery and their constraints. Charts were developed to help indicate safe operating zones for ground-based harvesting machines related to terrain slope (%) and soil bearing capacity, as measured by California Bearing Ratio, or CBR (Fig. 1). This shows the need for low ground pressure machines (e.g. using high flotation tyres) for any soil less than 3% CBR, but with increasing CBR operating up to 50% slope is acceptable with any ground-based machine. Operating from 50% to 60% slope is a critical zone where purpose-built steep terrain harvesters are required, but operating above 60% is considered very critical and requires additional securing systems such as »cable-assist« or traction winch technology.

Slope is not the only factor that should be considered when assessing safe operations or system productivity on steep terrain (Strandgard et al 2014). Soil bearing capacity and the vehicle-terrain interface are also important (Horn et al. 2007), as is the operator skill factor (Heinimann 1999). Visser and Berkett (2015) demonstrate that ground roughness and stumps can have a larger impact on the slope of the actual machine than the underlying terrain slope. Operating machinery on steep terrain can also increase the impact on soil disturbance (Horn et al. 2007). It can be assumed that cable-assist system will reduce soil disturbance through reduced slippage of the wheel and/or tracks.

However, few studies (e.g. Wratschko 2006) have actually quantified the level of disturbance associated with cable-assist machinery on slopes where previously no machine had travelled.

R. Visser and K. Stampfer

et al. 1983). Based on such results, many new steep slope purpose built machines were equipped with self-levelling cab and boom. The greater the ability to self-level the cab and boom, the steeper the terrain that could be operated on (Peters 1991).

In the early 1980s these steep-terrain tracked machines were studied to assess the operational slope limits (Arola et al. 1981, Schiess et al. 1983). The results indicated that manual-mechanical machine controls are not suitable for difficult terrain conditions. The machines were operated on slopes up to 70%. More recently improved control systems, and swivel seats, 360 degree windows and/or rear facing cameras provide for greater visibility and improved operator performance. Modern self-levelling machines also redistribute the center of gravity uphill to improve overall stability (Fig. 2).

There have been some significant advances in equipment design in Europe with machines working successfully on slopes over 65% (33 degrees) not uncommon (Stampfer and Steinmüller 2001, Bombosch et al. 2003, Stampfer and Steinmüller 2004). Successful felling on steep terrain is mainly due to the machines having new under-carriage design. One of these concepts is demonstrated in the development of the Komatsu 911 X3M (initially developed as the Valmet 911 »Snake«) (Stampfer and Steinmüller 2001) that is fitted with four independently suspended »high-drive« tracks (Fig. 3). This improves traction and stability, as one track can negate an obstacle such as a tree stump, while the other three are still firmly flat on the ground.

Other developments include machines such as the Menzi-Muck A91 (www.menzimuck.com; Fig. 4) and the Kaiser S3 Spyder (www.kaiser.li). They have increased stability and maneuverability with wheels on

2. Machine improvements for steep terrain Improvements to allow machines to operate on steeper areas are two-fold: the need for increased stability of the base machine on the slope itself, as well improving the ergonomics for the operator. The latter has proven to be the easiest to resolve. With the goal of rationalising timber harvesting in steep terrain, early developments in North America included tracked machinery equipped with processors, fellerbuncher or harvester heads. Empirical studies showed a reduced productivity and operability when operating the machine without a self-levelling cab (Schiess Croat. j. for. eng. 36(2015)2

Fig. 2 Tigercat steep terrain feller-buncher shows that self-levelling cab feature is also used to redistribute the weight uphill for increased stability

323

R. Visser and K. Stampfer

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321–331)

Fig. 3 Komatsu 911.5 X3M, for example, can be fitted with four independently suspended »high-drive« tracks (photo from Komatsu Forest)

Fig. 4 Menzi Muck A91 as an example of a harvester with wheels on hydraulically actuated »arms« that can lower its centre of gravity to improve stability (photo from Raffaelle Cavalli)

hydraulically actuated »arms« (Hempill 1983). This allows the machine to rise over obstacles such as stumps, but more importantly it can lower itself on to the ground when felling larger timber in a difficult position.

Increasing the number of axles and providing independent suspension for those axles increases the number of contact points with the ground. The new purpose built Ponsse ScorpionKing provides four axles on two bogeys (Fig. 5). Another significant difference between the CTL type harvesters and the tracked felling machines is that a significant proportion of the weight is carried directly on the under-carriage. That is, only the weight of the cab, and/or the cab and the boom, need to self-level. In contrast, the WT felling machines carry the engine, pumps, fuel, hydraulics above the turntable raising the relative centre of gravity, as well as creating significant swing momentum when working. Another purpose built feature on the Ponsse ScorpionKing is the ability for the front and rear frames to tilt to the terrain, with the resulting pivot point for the cab being low. It also has an active stabilisation system based on detecting the direction and position of the crane, and then pressing the rear frame in the direction of work.

While points of contact, as well as distribution of weight on the ground affect the amount of traction a machine can gain, on tracked machines longer »cleats« can cut through a softer soil surface and also increase the level of traction. Wheeled machines can use chains (see also Fig. 3) or belts to successfully extend the operating range. The machine then develops the maximum amount of traction by ensuring the failure is between soil layers, and not between the tracks and the soil.

While these options invariably extend the operating range, they can be expensive, cumbersome, increase fuel usage and increase the level of soil disturbance.

While equipment modifications have increased the operating range, not all forest operations can employ such technology. Difficulty to access areas by machine is not the only limiting factor; larger diameter trees, and/or smaller crews that cannot support the high capital cost of such a machine within their system, can also limit mechanised options (Raymond 2008). As such, chainsaw felling is still common in some countries and is still the highest risk in terms of fatalities in the industry.

3. Cable assist system Fig. 5 Ponsse ScorpionKing (photo retrieved from Ponsse.com)

324

Cable assist machinery for forest operations have been commercially available in Europe since the 1990s (Sebulke 2011), with a number of different companies Croat. j. for. eng. 36(2015)2

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321â&#x20AC;&#x201C;331)

Fig. 6 Forestry machines can be stable on very steep slopes if the weight and centre of gravity are well managed (from McLean and Visser 2011) offering cable winch products that are either integrated onto the machine or a separate attachment (Sutherland 2012). Initially they were mainly used on forwarders (e.g. Bombosch et al. 2003, Wratschko 2006), but now there are numerous commercial options to extend that technology to harvesters (Sebulke 2011).

There is a limit with regard to the physical feasibility of operating machines on steep slopes (Hunter 1993). The loss of traction will prevent the machine from moving up and down the slope, but in terms of ÂťfailureÂŤ, the real safety concern is the risk of machine roll-over. The slope associated with static roll-over is relatively easy to calculate as shown in Fig. 6. If the machine Centre of Gravity (CoG) is on the uphill side of the Pivot Point (PP), then the machine will not roll.

Most forestry machines have relatively low CoG and are technically very stable in their intended direction of drive, both uphill and downhill. However, McLean

Fig. 7 A loaded forwarder typically has a high CoG, and if traversing across the slope or with the boom reaching out to the downhill side, it will be very unstable even on low slopes (from McLean and Visser 2011) Croat. j. for. eng. 36(2015)2

R. Visser and K. Stampfer

and Visser (2011) showed that the machines with a higher centre of gravity traversing across the slope, such as a loaded forwarder, can easily become statically unstable on very low slopes (Fig. 7). Machines with boom attachments, such as felling machines, can also become unstable if the boom is swung to the downhill side. The weight, but also the force from the momentum, can affect stability (Eger and Kiencke 2003).

While static roll-over slope limits are relatively easy to calculate, a dynamic factor will reduce the slope limit where a roll-over can occur. As such, loss of traction can become a significant factor and it can be deduced that most roll-over accidents result from an initial loss of traction. It results in an uncontrolled gain in momentum and if followed by hitting an object, such as a stump, or a change in terrain slope, can readily result in a roll-over. The basic physics with regard to a retaining traction on a slope is that the gravity force pulling the machine down (Wg) should not exceed the traction force (T) that the machine is able to develop on the ground. The benefit of cable-assist system is that the tension force provided by the cable (C) will add to the traction force (T) and thereby greatly increase the operating slope of the machine without it reaching its traction limit (Fig. 8).

In terms of calculations, Wg is simply the product of the downward force (W) by the sine of the angle of the slope. Wg will be 0 when on flat terrain, and in-

Fig. 8 Schematic diagram of forces of a cable assisted machine on a slope, where the cable assist force (C) is summed with the available traction force (T) to overcome the gravity down the slope (Wg) (from Visser 2013)

325

R. Visser and K. Stampfer

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321â&#x20AC;&#x201C;331)

crease up to the full weight of the machine when on a vertical slope, that is fully suspended by the rope. The traction force (T) is the product of the normal force of the machine on the ground (Wn) with the Coefficient of Friction (cof). As such, the normal force Wn is defined by the product of the machine weight (W) and the cosine of the slope and will decrease with increasing slope.

The cof is the relationship of the tractive force that can be developed between the tracks and the soil. This is both dependant on the machine (i.e. tracks can typically develope higher traction coefficients than tyres), as well as the inherent strength of the soil (in shear). It is a complex relationship but a typical range for cof is from 0.4 on wet, soft or weak soils, up to 1.0 for a tracked machine operating on dry firm soil. Fig. 9 shows the effect of slope angle on a range of traction coefficients from 0.4 to 1.0, overlaid on the gravity force using a 37 tonne machine, where the lines intersect represents the theoretical slope limit for the machine. For example, for a cof of 0.7, the maximum machine slope is 34 degrees. The graphic also shows the potential benefits of cable-assist system. The short black lines represent a cable tension of 10 tonnes. For

Fig. 9 Chart showing the relationship between Wg and the slope using the example of a 37 tonne machine, and the effect of differing traction coefficients and cable-assist tension on slope limit shown for different soil strength factors (cof). The intersection of the effect of gravity (Wg) and the different soil strength factors (cof) indicate the slope limit (shown in degrees). Adding a cable-assist system with 10 tonnes of tension, as illustrated by the black arrows, significantly increase the operating range. (from Visser 2013)

326

the previous example of cof of 0.7, it moves the traction limit from 34 to 48 degrees. Under all scenarios, it can be seen that a cable assist system with 10 tonnes will greatly increase the operating range.

4. Cable-assist design options While the cable-assist concept is simple enough, adding it to machines is not without its complications and has led to different design concept options. Consequently, two options have emerged. The most common is to mount the winch onto the chassis of the primary machine, whereby a number of manufacturers have preferred the Âťbolt-onÂŤ option to accommodate the opportunity to remove the unit when not required. The second uses a secondary machine to house, and provide power to the winch.

For the integrated system, adding a winch system to the machine itself adds both weight and increases power requirements (Fig. 10). In addition to the winch itself, modifications are required in terms of integrating the control system as well as some fairlead to ensure that the cable is not dragged over the ground or subjected to bending fatigue. In Europe, it is typical to connect the cable to suitable standing trees, whereas most countries in the Southern hemisphere prevent any anchoring to live trees, and instead stumps, deadmen or machines are used as anchors. An operational consideration is that the expected utilization of the cable-assist system is typically not high compared to conventional felling machines. The effort required to connect the system, and the subsequent movement limitations, means that an operator would not use the cable-assist option unless required.

Fig. 10 Schematic diagram of cable-assist machine configuration with winch integrated onto the machine Croat. j. for. eng. 36(2015)2

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321–331)

The other design option is a two-part system with the winch mounted on and powered by a second machine, typically either a bulldozer or excavator (Fig. 11). In addition to removing the weight and power requirements from the machine on the slope, this option provides flexibility in that the machine can be used in an ad hoc basis for multiple steep terrain machines. For example, first a felling machine is tethered, and subsequently a more standard excavator with grapple is used to pre-bunch or shovel the wood up or down the slope. The anchor machine is mobile and can readily be moved.

There are also other variations of both options. Two examples include an Austrian small purpose built mobile tracked winch system (T-Winch) with significantly lower power and fuel requirements compared to converting a conventional bulldozer or excavator (Ecoforest 2013), and Summit Manufacturing in Canada have built a cable assist unit that is mounted at the base of a yarding tower, that in turn is mounted on the boom of an excavator.

While cable-assist systems are becoming common, few studies have been published to sustain indications that such systems actually have productivity, cost-effectiveness, environmental or safety benefits. A number of operational studies indicate that it is very feasible to operate on slopes not achievable otherwise, and productivity levels are acceptable once the system is set up in place (Evanson and Amishev 2010, Evanson et al. 2013). However, no long term studies have been made to establish mechanical or operational delays, but given the more extensive and set up time and machine complexity they are expected to be higher.

With regard to safety, consideration of the actual ten-

R. Visser and K. Stampfer

Fig. 12 Chain used for last section to prevent wear and tear on cable (photo from Hamish Berkett) sion in the rope becomes critical if the machine is not stable on the terrain without the cable. Brief studies in both Austria and New Zealand have indicated that actual tensions regularly exceed the expected, and this is consistent with tension monitoring studies in cable logging (Hartsough 1993, Harrill and Visser 2014). Operators are also going through a steep learn-curve with regard to successful implementation, but digital onboard navigation tools are being developed to support operator decision making (Marshall 2012). Engineering modification are also possible: for example most New Zealand based operations prefer the use of a chain for the first 15–20 meters to minimize the wear and tear, or eliminate the risk of breaking the rope when felling or pulling trees across it (Fig. 12).

5. Rules and guidelines for cable-assist system

Fig. 11 Cable-assist machine configuration with winch on mobile type anchor Croat. j. for. eng. 36(2015)2

In terms of commonly available and referenced harvesting manuals (Liley 1983, OR-OSHA 2009, WorkSafeBC 2006, FITEC 2000, Safe Work Australia 2013), none have been updated to include cable-assist guidance. However, with the new developments in steep slope machinery, many safety organisations have revised references to slope limits. For example the latest »Safety and Health in Forestry Work« published by the International Labour Office (ILO 1998) stated: »mechanised harvesting should not be carried out in site conditions where the stability of the machine cannot be assured«. Equipment should not be operated on slopes exceeding the maximum gradient specified by the manufacturer or exceeding that which has been assessed as safe by a competent authority or

327

R. Visser and K. Stampfer

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321–331)

a competent person. Where the above specifications have not been made: (a) rubber-tyred skidders or forwarders should not be operated on a slope which exceeds 35%; (b) crawler tractors, feller-bunchers, excavator harvesters or similar machines should not be operated on a slope which exceeds 40%; (c) any other forestry equipment specifically designed for use on steep slopes should not be operated on a slope which exceeds 50%.

The Workers’ Compensation Board of British Columbia (WorkSafeBC 2006) has updated its Occupational Health and Safety Regulations and the stated slope limits are the same as those presented by the ILO (quoted above), except the word »should« is replaced by »must«. The regulations state that logging equipment must not be operated in a particular location or manner if its stability cannot be assured during that operation. Subject to this rule »… logging equipment may be operated beyond the maximum slope operating stability limits specified … if, (a) a qualified person conducts a risk assessment of that operation, and (b) written safe work practices acceptable to the Board are developed and implemented to ensure the equipment stability during operation.« The British Columbia Forest Safety Council developed a steep slope resource package to help manage safety of operations on slopes that exceed the BC guidelines (BCForestSafe 2011). Part 1 of this resource package is a Steep Slope Hazard Assessment Tool, a method to evaluate site-specific and machine-specific hazards and develop a plan to implement practices to mitigate machine stability risks. It recommends companies develop site specific slope management plans for their operations when exceeding slope limits.

In New Zealand, the revised Approved Code of Practice for Safety and Health in Forest Operations (MBIE 2012) contained a section for winch-assisted harvesting on steep slopes and references to specific slope limits have been removed. Specifically, it requires »All mobile plant using the assistance of a wire rope and/or winch shall be specifically designed, tested, demonstrated to be safe« and that »The tension on the wire rope shall be restricted to 33% of its breaking load at all times«. As such, they have aligned the safety requirements with that of cable yarding, where a factor of safety of 3 is common for most skyline applications.

Conversely, no European country has yet implemented specific cable-assist rules, although machine manufacturers have started to develop their own guidelines. Common to nearly all is the fundamental

328

principle that the machine must remain stable and have traction without the cable, and as such the cable is only a traction assist and no additional rules need be incorporated. Only a few manufacturers are providing slope limits, such as Komatsu noting a limit of 55%. In their operating manual, Ritter (2015) provides the following guidance for operation: ÞT he operation and maintenance of the winch may only be carried out by suitable, reliable person familiar with this work and over 18 years of age; ÞW orking alone is only allowed when wireless emergency communication exists; ÞB efore use, but at least every working day, check on your proper operating condition of the winch.

While such recommendations are direct, they do not address specific slope and/or stability limitations of cable-assist machinery. As cable-assist operations become more prevalent, it would be rational to develop standardised operating guidelines.

6. Conclusion Many developments have increased our ability to successfully harvest on steep terrain using groundbased equipment. Improvements have included additions such as self-levelling cabs for operator comfort, and more recently significant modifications of carrier bases to improve traction and stability. A possible major step-change has been the development of cableassist technology. Cable-assist system can significantly increase the ability to operate on steep slopes and avoid soil damaging slip, but the actual implementation, and understanding of its limitations, is in their infancy.

7. References Acuna, M., Skinnell, J., Evanson, T., Mitchell, R., 2011: Bunching with a Self-levelling Feller-Buncher on Steep Terrain for Efficient Yarder Extraction. Croat. j. for. eng. 32(2): 521–531.

Alam, M., Acuna, M., Brown, M., 2013: Self-Levelling FellerBuncher Productivity Based on Lidar-Derived Slope. Croat. j. for. eng. 34(2): 273–281.

Arola, R.A., Miyata, E.S., Sturos, J.A., Steinhilb, H.M., 1981: Felling and bunching small timber on steep slopes. U.S. Department of Agriculture Forest Service, North Central Forest Experiment Station, St. Paul, Minnesota Research Paper NC-203, 12 p.

Croat. j. for. eng. 36(2015)2

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321–331) Amishev, D., Evanson, T., Raymond, K., 2009: Felling and Bunching on Steep Terrain – A Review of the Literature.

Harvesting Technical Note HTN01-07, Future Forests Research Limited, Rotorua, New Zealand, 10 p.

Amishev, D., 2012: Mechanisation on Steep Slopes in New Zealand. Harvesting Technical Note HTN04-07, Future Forests Research Limited, Rotorua, New Zealand, 8 p.

Axelsson, S.Å., 1998: The mechanization of logging operations in Sweden and its effect on occupational safety and health. International Journal of Forest Engineering 9(2): 25–31.

BCForestSafe, 2011: Guideline: Steep Slope Resource Package. Supporting Guidance for Operating on Steep Slopes.

BC Forest Safety Council, Nanaimo, British Columbia, 26 p.

Bell, J.L., 2002: Changes in logging injury rates associated with use of feller-bunchers in West Virginia. J. Safety Res. 33(4): 463–471.

Bombosch, F., Sohns, D., Nollau, R., Kanzler, H., 2003: Are forest operations on steep terrain with wheel mounted forwarders without slippage possible? Austro 2003: High Tech Forest Operations for Mountainous Terrain. October 5–9, 2003, Schlaegl-Austria, 5 p.

Carson, W.W., 1983: Is New Zealand ready for steep country logging? New Zealand Journal of Forestry 28(1): 24–34.

Cavalli, R., 2015: Forest Operations in Steep Terrain. Presented at Conference CROJFE 2015 »Forest Engineering – Current Situation and Future Challenges«, March 18–20, 2015, Zagreb, Croatia (retrieved from www.crojfe2015.com/ home) Dykstra, D.P., 1975: Production rates and costs for cable, balloon, and helicopter yarding systems in old-growth Douglas-fir. Research Bulletin 18, Forest Research Laboratory, Oregon State University, 57 p.

Ecoforest, 2013: T-Winch: Mobile Traction Winch. (retrieved from www.ecoforst.at)

R. Visser and K. Stampfer

FAO/ECE/ILO, 1971: Symposium on forest operations in mountainous regions. Technical report, Joint FAO/ECE/ILO Committee on Forest Working Techniques and Training of Forest Workers, Krasnodar (USSR), August 31 to September 11, TIM/EFC/WP.1/1, 90 p.

FITEC, 2000: Best Practice Guidelines for Cable Logging.

Forest Industry Training and Education Council, Auckland, New Zealand, 138 p.

Harrill, H., Visser, R., 2012: Matching rigging configurations to harvesting conditions. Harvesting Technical Note HTN04-06, Future Forests Research Limited, Rotorua, New Zealand, 8 p.

Harrill, H., Visser, R., 2014: Skyline Tensions and Productivity of a Motorised Grapple Carriage. Report FFR-H015, Future Forests Research Limited, Rotorua, New Zealand.

Hartsough, B., 1993: Benefits of remote tension monitoring.

LIRO Report 18(23). Logging Industry Research Organisation, Rotorua, New Zealand, 13 p.

Heinimann, H.R., 1995: Mechanisierte Holzernte in Hanglagen. Wald und Holz 76(11): 32–36.

Heinimann, H.R., 1999: Ground-based harvesting technologies for steep slopes. In: Proceedings of the International Mountain Logging and 10th Pacific Northwest Skyline Symposium, Sessions and Chung (editors), March 28 – April 1, Corvallis, Oregon, USA, 1–19.

Heinimann, H.R., Stampfer, K., Loschek, J., Caminada, L., 2001: Perspectives on central European cable yarding systems in Proceedings of the international mountain logging and 11th Pacific Northwest skyline symposium, Schiess, P., and F. Krogstad (eds.). ������������������������������������ College of Forest Resources, University of Washington, Seattle, Washington.

Hempill, D., 1983: Spyder steep slope feller-buncher. Logging Industry Research Association, Rotorua, New Zealand.

Technical Release 5(2): 4 p.

Eger, R., Kiencke, U., 2003: Modeling of rollover sequences.

Control Engineering Practice 11(2): 209–216.

Horn, R., Vossbrink J., Peth, S., Becker, S., 2007: Impact of modern forest vehicles on soil physical properties. Forest Ecology and Management 248: 56–63.

Evanson, A.W., Amishev, D., 2010: A steep terrain excavator feller buncher. Harvesting Technical Note HTN03-02, Future Forests Research Limited, Rotorua, New Zealand, 8 p.

Hunter, A.G.M., 1993: A review of research into machine stability on slopes. Safety Science 16(3–4): 325–339.

Evanson, T., Brown, D., 2012: Improved Grapple Control using a Grapple Restraint. Harvesting Technical Note HTN10-04, Future Forests Research Limited, Rotorua, New Zealand, 7 p.

Evanson, T., 2013: Hauler Vision System: Testing of Cutover Camera. Harvesting Technical Note HTN05-03, Future Forests Research Limited, Rotorua, New Zealand, 6 p.

Evanson, T., Amishev, D., Parker, R., Harrill, H., 2013: An evaluation of a ClimbMAX Steep Slope Harvester in Maungataniwha Forest, Hawkes Bay. Harvesting Report H013, Future Forests Research Limited, Rotorua, New Zealand, 16 p.

Croat. j. for. eng. 36(2015)2

ILO, 1998: Safety and health in forestry work. Code of practice, Occupational safety, Occupational health, Forestry.

International Labour Office, Geneva 13.04.2, ISBN 92-2110826-0 Kirk, P.M., Sullman, M.J., 2001: Heart rate strain in cable hauler choker setters in New Zealand logging operations.

Applied ergonomics 32(4): 389–398.

Klun, J. Medved, M., 2007: Fatal accidents in forestry in some European countries. Croat. j. for. eng 28(1): 55–62.

Komatsu, 2015: Mehr Sicherheit: Grundlagen und Tipps fuer den sicheren und praxisnahe Einsatz Ihre Komatsu Forstmachine, 27 p. (www.komatsu.com)

329

R. Visser and K. Stampfer

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321–331)

Laflamme, L., Cloutier, E., 1988: Mechanization and risk of occupational accidents in the logging industry. Journal of Occupational Accidents 10(3): 191–198.

Raymond, K., 2010: Innovative Harvesting Solutions: A Step Change Harvesting Research Programme. New Zealand Journal of Forestry 55(3): 4–9.

Liley, B., 1983: Cable Logging Handbook. New Zealand Logging Industry Research Association, Rotorua, New Zealand, 147 p.

Raymond, K., 2012: Innovation to Increase Profitability of Steep Terrain Harvesting. Proceedings of NZ Institute of Forestry Conference, July 1–4, Christchurch, New Zealand, 15 p.

MacDonald, A.J., 1999: Harvesting systems and equipment in British Columbia. British Columbia Ministry of Forests: Forestry Division Services Branch, Victoria, British Columbia, 194 p.

Marshall, H., 2012: On-board Machine Stability Information System. Harvesting Technical Note HTN05-01, Future Forests Research Limited, Rotorua, New Zealand, 5 p.

McFadzean, S., Visser, R., 2013: Falcon Forestry Claw Grapple: Productivity and Ergonomics. Harvesting Technical Note HTN05-06, Future Forests Research Limited, Rotorua, New Zealand, 7 p.

McLean, D., Visser, R., 2011: Roll Over Limits of Forestry Machines on Steep Terrain Unpublished report. Forest Engineering, School of Forestry, University of Canterbury, New Zealand, 13 p.

MBIE, 2012: Approved Code of Practice for safety and health in forest operations. Ministry of Business, Innovation & Employment, Wellington, New Zealand.

Safe Work Australia, 2013: Guide to managing risks in cable logging. Safe Work Australia, Canberra, Australia, 62 p.

Samset, I., 1985: Winch and cable systems. Martinus Nijhoff/ Dr W. Junk Publishers, Dordrecht, Netherlands, 539 p.

Schiess, P., Schuh, D., Miyata, E.S., Mann, C.N., 1983: Concept Evaluation of a Walking Feller-Buncher – The Kaiser X5M Spider. Paper presented at the First Technical Conference on Timber Harvesting in the Central Rockies. Colorado State University, Ft. Collins, January 4–6, 50 p.

Sebulke, J., 2011: Holzernte mit Traktionswinden. Forst &Technik (3): 20–26.

Slappendel, C., Laird, I., Kawachi, I., Marshall S., Cryer C., 1993: Factors affecting work-related injury among forestry workers: A review. Journal of Safety Research 24(1): 19–32.

Stampfer, K., Leitner, T., Visser, R., 2010: Efficiency and Ergonomic Benefits of Using Radio Controlled Chokers in Cable Yarding. Croat. j. for. eng. 31(1): 1–9.

Montorselli, N.B., Lombardini, C., Magagnotti, N., Marchi, E., Neri, F., Picchi, G., Spinelli, R., 2010: Relating safety, productivity and company type for motor-manual logging operations in the Italian Alps. Accident Analysis & Prevention 42(6): 2013–2017.

Stampfer, K., Steinmüller, T., 2001: A new approach to derive a productivity model for the harvester Valmet 911 Snake. In: Proc. International Mountain Logging and 11th Pacific Northwest Skyline Symposium— A Forest Engineering Odyssey, December 10–12, Seattle, USA, CD ROM, 254–262.

NZDOL, 1999: Approved Code of practice for Safety and Health in Forest Operations. New Zealand Department of Labour Occupational Safety and Health Service, Wellington, 133 p.

Stampfer, K., Steinmüller T., 2004: Harvester und Seilgerät im Steilgelände – Valmet 911.1 X3 M und Syncrofalke. Institut für Forsttechnik, Department für Wald- und Bodenwissenschaften, Universität für Bodenkultur Wien, 24 p.

Olund, D., 2001: The future of cable logging. In The International Mountain Logging and 11th Pacific Northwest Skyline Symposium, Schiess, P., and F. Krogstad (eds.), College of Forest Resources, University of Washington, Seattle, Washington, 263–267.

Studier, D.D., Binkley, V.W., 1974: Cable logging systems.

Division of Timber Management, Forest Service, US Department of Agriculture, Portland, Oregon, 205 p.

OR-OSHA, 1993: Yarding and loading handbook. Oregon Occupational Safety and Health Division, Salem, Oregon, 184 p.

Parker, R.J., 2010: Using video clips for training – Breaking out. Harvesting Technical Note HTN03-03, Future Forests Research Limited, Rotorua, New Zealand, 6 p.

Peters, P.A., 1991: Mechanized felling on 40 to 100 percent slopes. ASAE – Paper No. 917545, presented at the winter meeting in Chicago, December 17–20, 9 p.

Raymond, K., 2008: Harvesting: Felling From manual to Mechanised to where? Proceedings of the Forest Industry Strategic Summit. Rotorua Convention Centre, July.

330

Strandgard, M., Alam, M., Mitchell, R., 2014: Impact of Slope on Productivity of a Self-levelling Processor. Croat. j. for. eng. 35(2): 193–200.

Sutherland, B., 2012: Review of tethered equipment for steep-slope operations. Internal Report IR-2012-08-20, FPInnovations, Vancouver, Canada, 11 p.

Visser, R., Stampfer, K., 1998: Cable extraction of harvester felled thinnings: An Austrian case study. Journal of Forest Engineering 9(1): 39–46.

Visser, R., 2013: Tension Monitoring of a Cable Assisted Machine. Harvesting Technical Note HTN05-11, Future Forests Research Limited, Rotorua, New Zealand, 5 p.

Visser, R., Berkett, H., 2015: Effect of terrain steepness on machine slope when harvesting. International Journal of Croat. j. for. eng. 36(2015)2

Expanding Ground-based Harvesting onto Steep Terrain: A Review (321–331) Forest Engineering 26(1): 1–9. DOI: 10.1080/14942119.

2015.1033211 WorkSafeBC, 2006: Cable yarding systems handbook.

Workers‘ Compensation Board of British Columbia, Vancouver, British Columbia, 204 p.

R. Visser and K. Stampfer

Wratschko, B., 2006: Einsatzmöglichkeiten von Seilforwardern. Masterarbeit, Institut für Forsttechnik, University of Natural Resources and Life Sciences, Vienna, 66 p.

Authors‘ address:

Received: July 10, 2015 Accepted: July 23, 2015 Croat. j. for. eng. 36(2015)2

Assoc. Prof. Rien Visser, PhD. e-mail: rien.visser@canterbury.ac.nz University of Canterbury College of Engineering School of Forestry Private Bag 4800, Christchurch NEW ZEALAND Prof. Karl Stampfer, PhD.* e-mail: karl.stampfer@boku.ac.at University of Natural Resources and Applied Life Sciences Vienna Department of Forest and Soil Sciences Institute of Forest Engineering Peter Jordan Straße 82 1190 Vienna AUSTRIA * Corresponding author

331

Vodič kroz eu fondove za šumarski sektor 26. svibnja 2015. godine Europska komisija je odobrila i usvojila Program ruralnog razvoja RH za razdoblje 2014.–2020. U izradi ovog Programa za sektor šumarstva sudjelovala je i radna skupina sastavljena od predstavnika svih relevantnih šumarskih institucija, uz koordinaciju nadležnog ministarstva.

Tim povodom, Komora je objavila Vodič kroz EU fondove za šumarski sektor (http://www.hkisdt.hr/ podaci/2015/ostalo/HKISDT_Vodic_kroz_EU_fondove.pdf), koji daje pregled i ostalih operativnih pro­ grama kao i Programa Unije, odnosno EU fondova koji su na raspolaganju šumarskom sektoru u novom programskom razdoblju EU-a 2014.–2020.

Autori Vodiča su Victoria Jane Primhak, konzultantica za EU fondove i Domagoj Troha, EU projekt menadžer. Autori će Vodič ažurirati slijedom eventualnih promjena u idućem razdoblju. Vodič je do­stu­ pan na hrvatskom i engleskom jeziku.

ISSN 1845-5719