Water Quality Research Journal of Canada sample issue

1

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Editorial Dr Ronald Gehr, Editor, Water Quality Research Journal of Canada, 2004–2013

As new Editors-in-Chief, before we divulge

IWAP journals has proved to be most advantageous to the

some of our plans for the future of the

readership which is now becoming more international.

journal, we want to pay tribute to Dr

Besides revising formats, improving turn-around time

Ronald Gehr, Professor at McGill Univer-

and other changes, one of the most significant long-term

sity, who took over the reins of the journal

contributions of Dr Gehr was to arrange to have all of the

in 2004. First, let us give a bit of history.

journal’s issues printed before the electronic dawn, archived

The Water Quality Research Journal of

at his home institution, McGill University, Montreal, and to

Canada (WQRJC) is the longest lived journal focusing on

make them available to the general public. The archived

water quality in Canada. The Canadian Association on

journals can be accessed free-of-charge through the

Water Quality (CAWQ) is the organization responsible for

CAWQ website (www.cawq.ca).

the WQRJC. The journal had humble beginnings in 1966

On behalf of the CAWQ board and all of its members,

as mimeographed copies of proceedings of the First

we pay homage to Dr Gehr for his extended efforts for the

Annual Conference on Water Pollution Research. Proceed-

journal.

ings of this annual Canadian conference were eventually

Now, let us shed some light on our plans for the future

printed and bound. In 1980 the journal was converted to

that we have elaborated in close communication with

an independent, peer-reviewed journal format with two

IWAP, which has published the journal now for about

issues per year, and was given the title Water Pollution

2 years.

Research Journal of Canada. By 1984 it progressed to a quar-

First of all, we plan to streamline the review process

terly publication. The journal changed its name to the

according to a new IWAP workflow so as to make it more effi-

current title in 1995. Throughout these metamorphoses,

cient and, especially, making its decision making faster. We

Environment Canada remained a staunch pillar supporting

know fast processing times have become a criterion by

the journal in various ways.

which potential authors select a journal. The leaner, more

When Dr Gehr became Editor-in-Chief, the journal was

international, Editorial Board will get more responsibility in

experiencing declining logistical and other support from the

the manuscript processing, in fact being in charge of it from

government and by 2008 the journal depended completely

the moment the paper is attributed to a particular editor

on the CAWQ. The government support of the early years

until his/her final decision on the manuscript. Personalized

provided a comfort zone and certainly a relaxed business

contact between editor and reviewer will ensure increased

model. Under his direction Dr Gehr nursed the journal

motivation of the reviewers and thus more timely reviews.

through many logistical changes as well as financial chal-

The size of the Editorial Board will be reduced to about 10

lenges to improve the journal’s stature, in addition to his

dedicated water professionals covering the variety of water

primary tasks of being the initial portal for all papers and

themes WQRJC is publishing on. In a first instance, the inter-

final arbiter in contentious situations. His dedication was

nationalization of the Editorial Board will occur by including

at the foundation of the journal’s success. Under his guid-

a number of US experts in addition to Canadian editors.

ance the journal eventually moved under the aegis of the

Second, we will carefully evaluate the journal’s Aims

International Water Association Publishing (IWAP) group

and Scope to provide a focus which will avoid major overlap

in 2011. Moving the journal into the suite of high quality

with key existing journals and therefore attract more high

doi: 10.2166/wqrjc.2013.111

2

Editorial

Water Quality Research Journal of Canada

|

49.1

|

2014

quality submissions. We will of course look at the journal’s

To end this Editorial, which has paid tribute to the work

record of contributions to the scientific literature and

that brought us where we are and indicated where we want

ensure that its stature in certain areas is maintained.

to go, we would like to invite you to submit your high quality

Third, we want our impact factor to increase as we fully

thought-provoking manuscripts to our journal, to propose

understand how important this has become in career

themes for special issues, to put the effort in writing up criti-

development. We will do this by ensuring that the best

cal reviews on contemporary topics and, last but not least, to

papers are attracted to the journal so that citations will

provide any suggestion that may cross your mind to make

follow. By carefully defining thematic issues, stimulating

our journal a better place to submit and read the latest scien-

submission of review papers, communicating convincingly

tific discoveries and practice-oriented innovations. We look

about the journal and adjusting its scope to ensure imminent

forward to working with and for you!

topics get the necessary visibility, we are pretty sure that the WQRJC’s impact factor will be increased substantially over the next 3 to 5 years. The autocatalytic effect of a journal’s

The new Editors-in-Chief

increase in impact on the quality of new submissions will

Peter A. Vanrolleghem

further enhance this anticipated progress.

Ronald L. Droste

3

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Editorial Drinking water: Assessing and managing risk

An adequate supply of safe drinking water is one of the key

designed to protect the health of the most vulnerable mem-

needs for a healthy life. Canadian drinking water supplies

bers of society, such as children and the elderly. The

are generally of excellent quality. However, water comes

guidelines set out the basic parameters that every water

into contact with many substances, all of which have an

system should strive to achieve in order to provide the clean-

impact on its quality. While many of these are harmless,

est, safest and most reliable drinking water possible. In

some may pose a health risk. The Canadian paradigm for

recent years, the theme that has preoccupied the water com-

understanding risk, the protection of drinking water and

munity has been risk: perception, assessment, management,

public health is continually evolving. As new information

communication, and all too often, controversy. This

emerges on topics such as chemical and microbiological

prevailing thread led Health Canada and the CDW to

threats to water quality, the potential impacts of climate

select the 2012 conference theme of ‘Risk Assessment and

change on water resources, and novel or improved analyti-

Management’.

cal and treatment technologies, this information brings with it challenges and opportunities.

The papers and posters presented at the 15th National Conference on Drinking Water addressed a wide range of

It is in this spirit that the Federal-Provincial-Territorial

topics, from detecting contaminants, to optimizing treat-

Committee on Drinking Water (CDW), in partnership with

ment processes, to risk assessment and management

the Canadian Water and Wastewater Association, convened

strategies. The range of topics reflected the multi-barrier

the 15th Canadian National Conference on Drinking Water

approach to safe drinking water that is promoted by the

in Kelowna, British Columbia in October 2012. The CDW

guidelines. This approach looks at each drinking water

has played a major role in drinking water quality in

system from the source all the way to the consumer’s tap

Canada since 1968, when the first edition of the Guidelines

to make sure all known and potential hazards are identified

for Canadian Drinking Water Quality was published by

and addressed so water remains free of contaminants. The

Health Canada. The CDW recognized the need for a

drinking water guidelines can be used as markers to make

national forum that would bring together scientists, regula-

sure the barriers are working and the treated drinking

tors and industry experts and focus on scientific data

water is safe. This Special Issue features papers carefully

relating to drinking water quality, on assessments of the

selected from the conference to reflect the source-to-tap con-

implications of these data for health and public policies

tinuum of the multi-barrier approach to drinking water

designed to protect the safe quality of the nation’s drinking

safety, supplemented by selected papers from other relevant

water supplies. The first National Conference was held in

sources.

Ottawa in 1984. Since that time, Canada’s drinking water

The province of Alberta is making strides towards a sys-

community has come together at this biennial event in

tematic risk assessment and management structure with the

each of the 10 provinces.

recent introduction of Drinking Water Safety Plans. As

The main responsibility of the CDW is to establish the

noted by the authors of this paper, the traditional regulatory

Guidelines for Canadian Drinking Water Quality. The

approach to maintaining the quality and safety of drinking

guidelines set out the maximum acceptable concentrations

water has largely been a reactive one. The work of Alberta

of microbiological, chemical and radiological contaminants

Environment and Sustainable Resource Development to

in drinking water. These drinking water guidelines are

develop a template for recording Drinking Water Safety

doi: 10.2166/wqrjc.2013.011

4

Editorial

Water Quality Research Journal of Canada

|

49.1

|

2014

Plans together with guidance notes to help complete them is

treatment units, such as RO and manganese greensand fil-

described in this paper. A drinking water safety plan is a

ters, to address the problem. The treatment performance

proactive method of assessing risk to drinking water quality,

of these systems was assessed by comparing raw and fin-

which better protects public health. The goal is to identify

ished water samples, with a focus on removal efficiency

key risks in a drinking water system as well as the interven-

and the effect of other water parameters. The study results

tions needed to bring them into control.

described here will help make informed decisions about

The multi-barrier approach includes an understanding

appropriate treatment.

of the contaminants entering a source of drinking water,

The storage and distribution of water poses its own

as illustrated by a study of the wastewater treatment systems

unique set of challenges. Hayes et al. in ‘Computational

that discharge into the Great Lakes basin and their effective-

modelling techniques in the optimization of corrosion

ness at removing chemicals of emerging concern. The Great

control for reducing lead in Canadian drinking water’

Lakes and their connecting channels form the largest fresh

describe the use of computational modelling techniques

surface water system on earth and are a source of drinking

to optimize corrosion control and reduce lead in drinking

water to millions of people. Chemicals of emerging concern

water. Modelling to optimize plumbosolvency control

may include pharmaceutical and personal care products,

was evaluated in Canadian and US contexts through

pesticides and others substances that are found in products

three case studies. In relation to regulatory compliance,

used daily in households, businesses, agriculture and indus-

supplementary orthophosphate dosing could be justified

try. The authors suggest that at least half of the 42 substances

in one water supply system but not in another. Compli-

examined in the study are likely to be removed by municipal

ance modelling illustrated differences in results that

wastewater treatment plants. The potential impact of these

were influenced by the stringency of the testing protocol

chemicals on ecosystem health and ultimately drinking

applied, the length of the lead service line and the

water supplies are poorly understood. Studies on anthropo-

copper premise pipe and by pipe diameters, as well as

genic inputs of chemicals are an important part of

flow characteristics (plug vs laminar). For either regulat-

quantifying exposure levels and generating science-based

ory compliance assessment or for the optimization of

information necessary to identify risks and inform risk

plumbosolvency control measures, routine sequential

management.

sampling from the same houses at a normalized flow

Water treatment is also a critical aspect of the multi-

will minimize these variable effects.

barrier approach, and the naturally occurring substances

It is clear from the papers presented in this Special Issue

in water can be as significant a treatment challenge as

that the water industry is continuously re-examining the

human inputs of contaminants. Thirunavukkarasu et al. in

ways in which it manages risk to drinking water, seeking

‘Performance of reverse osmosis and manganese greensand

deeper understanding and innovative solutions from the

plants in removing naturally occurring substances in drink-

source of the water, to how drinking water is treated,

ing water’ examine the performance of reverse osmosis

stored and distributed.

(RO) and manganese greensand plants in removing naturally occurring substances in drinking water. A number of

Stéphanie McFadyen

communities in Saskatchewan that depend on ground

Special Issue Editor

water as a source for drinking water have reported high

Water and Air Quality Bureau,

levels of naturally occurring substances, such as arsenic,

Health Canada,

uranium and selenium, in their raw water. Some of these

Ottawa, Ontario,

communities are installing new systems or retrofitting with

K1A OK9

5

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Implementation of Alberta’s drinking water safety plans D. C. Reid, K. Abramowski, A. Beier, A. Janzen, D. Lok, H. Mack, H. Radhakrishnan, M. M. Rahman, R. Schroth and R. Vatcher

ABSTRACT Traditionally, the regulatory approach to maintaining the quality and safety of drinking water has largely been a prescriptive one based on the ability of any given supply to meet standards set for a number of different chemical and biological parameters. There are a number of issues around the assumptions and the limitations of a sampling and analysis regime. The basis for such regimes is essentially reactive rather than proactive and, consequently, the cause of the concern may already have impacted consumers before any effective action can be taken. Environment and Sustainable Resource Development has developed a template for recording drinking water safety plans together with guidance notes to help complete them. The template has been developed in MS-Excel and has been designed in a straightforward step-wise manner with guidance on the completion of each sheet. It includes four main risk tables covering each main element of water supply which are prepopulated with commonly found ‘generic’ risks and these are carefully assessed before considering

D. C. Reid (corresponding author) K. Abramowski A. Beier A. Janzen D. Lok H. Mack H. Radhakrishnan M. M. Rahman R. Schroth R. Vatcher Alberta Environment and Sustainable Resource Development, Operations Division, Twin Atria Building, 4999–98 Avenue, Edmonton, T6B 2X3, Canada E-mail: donald.reid@gov.ab.ca

what action is required to deal with significant risks. Following completion of the risk tables, key risks are identified and the interventions required to bring them into control.

Key words

| implementation, legislation, water safety plans

INTRODUCTION Safe, secure supplies of drinking water are essential to all

been a prescriptive one based on the ability of any given

Albertans. Approximately three million Albertans, repre-

drinking water system to meet standards set for a number

senting more than 80% of the province’s population,

of different biological, chemical and physical parameters.

receive their drinking water from systems regulated by

In Canada, guidelines for drinking water quality are devel-

Environment and Sustainable Resource Development

oped and published by Health Canada through a Federal/

(ESRD). Alberta uses a multi-barrier approach to ensure

Provincial/Territorial

that safe drinking water is provided to all Albertans. This

These Guidelines for Canadian Drinking Water Quality

method is referred to as a ‘source to tap, multi-barrier

are adopted in Alberta and form the basis of the drinking

approach’. The term, ‘source to tap’ refers to the continuum

water quality standards used by the province.

Committee

on

Drinking

Water.

of environments that water passes through – from a water

The level of disease associated with contaminated drink-

body to the consumer’s drinking water tap and is a five-

ing water has fallen dramatically over the last 160 years or

pronged multi-barrier approach consisting of legislation,

so, virtually eliminating waterborne illness, but outbreaks

protection, drinking water systems, performance assurance

still can and do occur with the potential to have severe or

and knowledge. Detailed risk assessment or risk mitigation

devastating consequences for those involved (Rizak &

may or may not be part of these five prongs but is not a fun-

Hrudey ).

damental requirement for any of them.

These incidents highlight some of the limitations

Traditionally, the regulatory approach to maintaining

inherent in the ‘traditional’ regulatory approach. For

the quality and safety of drinking water has principally

example, there are a number of issues around the

doi: 10.2166/wqrjc.2013.063

6

D. C. Reid et al.

|

Implementation of Alberta’s drinking water safety plans

assumptions and the limitations of a sampling and analysis regime. Further, the basis for such sampling and analysis programmes is essentially reactive rather than being proactive and, consequently, the cause of the concern may

Water Quality Research Journal of Canada

|

49.1

|

2014

•

Assessing correctly what is required to be done in order

•

Determining how to obtain the necessary resources to

to reduce risks to an acceptable level. achieve this, how to prioritise and audit the tasks that

already have impacted consumers before any effective

have been identified, and how to deliver the actions

action can be taken. From this it is clear that the current gov-

within the required timescale.

ernance models being applied to assure the quality of drinking water received by consumers still has some limitations.

One

approach

designed

to

overcome

these

There are three other important considerations:

•

A DWSP cannot work in isolation so the operator must

limitations is the adoption and use of an approach termed

communicate and discuss their findings with the main

a ‘water safety plan’ or ‘drinking water safety plan (DWSP)’.

stakeholders and other relevant parties.

WHAT IS A DWSP?

• •

For the DWSP to work the actions that the operator has identified as necessary to mitigate the risks must be implemented.

Finally, the DWSP is a ‘living document’ and should not

The World Health Organization comments in the fourth edi-

just sit on the shelf as if to say ‘job done’; it should be

tion of their Guidelines for Drinking-water Quality (WHO

reviewed regularly and updated when necessary.

), ‘The most effective means of securing the safety of a drinking water supply is through the use of a comprehensive risk assessment and risk management approach that encom-

THE ALBERTA DWSP

passes all steps in the water supply from catchment to consumer.’ Binnie & Kimber () define a water safety

ESRD decided that developing a ‘template’ which would

plan as ‘a location-specific assessment of a water supply

have pre-populated generic risks for four key risk areas or

system, from the source, or sources, of the raw water through

‘nodes’ would be the most effective approach for introdu-

to the points of delivery, considering risks and hazards,

cing DWSPs to drinking water systems in the province.

means to address and monitor the hazards, and procedures

The template was built in MS-Excel and consists of a

for managing and operating the system under both normal

series of worksheets that together comprise the DWSP.

and exceptional circumstances’. This comprehensive risk

Functionality within the template is provided by various

assessment and risk management approach is termed a

macros which collate certain pieces of information into

‘water safety plan’ by the WHO and a ‘DWSP’ in Alberta.

summary sheets (such as risks which have been scored as high) or replicate data entries (e.g., name of the waterworks

DWSP – THE ALBERTA APPROACH

system).

The DWSP is intended to act as a single source for all of the relevant information about the water supply system. As

A DWSP framework was developed that built on pre-

such, the basic information needs to be entered for each of

existing regulatory requirements already being complied

the four elements of supply – source, treatment, network

with by operators.

and consumer. There are five sheets of this type in all –

The compilation of an effective DWSP is dependent on four principal processes:

•

Collecting and collating the best information about the

•

Analysing and understanding the risks that are present

water supply system.

core detail, source detail, treatment detail, network detail and customer detail.

Assessing the generic risks

and that in certain circumstances will threaten the

Each of the four risk sheets – source, treatment, network

safety of the supply’s customers.

and customer – have been populated with risks which are

7

D. C. Reid et al.

|

Implementation of Alberta’s drinking water safety plans

Water Quality Research Journal of Canada

|

49.1

|

2014

commonly encountered across the water industry in Alberta.

probability of something happening) and the cause (what

The information supplied provides a description of the risk,

chain of events leads to the hazard materialising).

the hazard that would result from it and the means by which it would happen (also known as ‘the causal chain’). In

The key risk sheet

addition to this, the comments box provides additional explanation or background.

The key risk sheet is essentially a list of all the key risks

In three of the sheets – source, treatment and network –

identified. The key risks are those risks with a risk score of

the risks have been grouped into sections to take account of

32 or more and these risks will have become coloured red

the different types of source, treatment process or component

automatically when the likelihood and consequence

part. If any of these groups do not apply to the system being

descriptions (and scores) are added. The key risks are the

considered, then a score of zero is entered, signifying that the

most significant risks and they all need to be considered

risk is not applicable. As each risk that is applicable to the

when determining what intervention is necessary to control

system is identified then the other fields on that risk line

each risk and reduce it to an acceptable level.

must be completed. These fields are as follows:

• • • •

Current monitoring – existing monitoring that is in place.

How risk is currently controlled – measures in place now

Completing the action plan

that are helping to control the risk. If not adequate then

The action plan is essentially the summary plan for the miti-

further intervention will be needed.

gation of all of the key risks though the provision of new, or

Assess if control is adequate – does the control mitigate

improved interventions (control measures). It is not antici-

the risk?

Do any standard procedures cover this – are there any standard operating procedures that help control this risk?

pated that all of the interventions will require capital investment. It may be possible to deal with the risk by changing or adding work procedures, or by providing additional

•

Likelihood and consequence – the probability of the risk

•

Likelihood and consequence scores – numerical values

•

for quantifying the risk score.

depending on the nature of the intervention, that not all

Risk score – calculated by multiplying the likelihood and

interventions can be delivered within a short timescale.

• •

occurring and its effects.

maintenance, or through increased or different monitoring regimes.

It was recognised that due to limited resources, and

consequence scores.

This means that in the short term there is no decrease in

Required intervention to prevent failure – what needs to

the vulnerability towards these risks.

be done to mitigate the risk if the score is 32 or greater.

It would be prudent, therefore, to consider what short-

Responsible party – person who is responsible for deliver-

term measures can be taken immediately to provide extra

ing the intervention.

protection. This might include things like increased visual monitoring, changes in alarm settings, additional or modified procedures that take account of the recognised risk.

Adding other site-specific risks

These measures are then entered in the short-term control box.

Once all of the generic risks have been completed the next

Next, the main interventions are considered. The

step is to assess whether the risks described cover all of

description should be an accurate account of what is

the risks in the system. When an additional risk is identified

intended to be done and how this will be achieved and, cru-

it is added to the bottom of whichever supply element risk

cially, who is responsible for delivering it. It may require

sheet it applies to. The process is the same as for the generic

funding, which should be indicated in the funding box and

risks with the additional requirements to describe the risk,

whether this has been approved, and then entered into the

the hazard and the cause of potential failure. It is important

two date boxes when it is expected that the work will

to have clarity about the difference between the risk (the

begin and when it is expected to be completed.

8

D. C. Reid et al.

|

Implementation of Alberta’s drinking water safety plans

Maintaining the plan

Water Quality Research Journal of Canada

|

49.1

|

2014

established departmental protocols the Code of Practice for Waterworks Systems Using High Quality Groundwater

The most common reason for updating the plan is likely to be

and the Code of Practice for a Waterworks System Consist-

to update the action plan following progress with an interven-

ing Solely of a Water Distribution System were revised and

tion. When an intervention is delivered it will reduce the

formally issued to affected facilities. Notices of change were

likelihood of that risk happening and the likelihood score

issued to waterworks systems operating under an approval,

should therefore be updated. This will in turn reduce the

and the standard approval template used to issue new or

risk score and indicate that this risk is now under control.

revised approvals was changed to reflect the DWSP require-

Other reasons to update the plan would include any

ment. The Standards and Guidelines for Municipal

changes made to the water supply system. These might be

Waterworks, Wastewater and Storm Drainage Systems

changes within the watershed, changes to part of the process

were also revised to take account of the new DWSP require-

in the water treatment works or changes to the network.

ments and these published in April 2012. In addition, ESRD

Should an incident or ‘near miss’ occur this would also

established a webpage where the DWSP template, support-

be an appropriate point to undertake a full review of the

ing documentation and training materials for self-directed

plan, in part to check whether the risk had been previously

training could be accessed (www.environment.alberta.ca/

identified and was thought to be under full control, or if not

apps/regulateddwq/dwsp.aspx) and a dedicated email

identified, why not. Some incidents or events may be rela-

account (aenv.dwsp@gov.ab.ca) to allow questions specifi-

tively easy to predict and may be seen as ‘accidents

cally related to DWSPs to be sent. All these elements were

waiting to happen’. Others, however, may be closer to an

completed within eight months and were achieved through

‘act of God’ where a very unusual set of circumstances has

the coordination and cooperation of many areas within

resulted in a catastrophic outcome. These sorts of risks are

ESRD. All drinking water systems regulated by ESRD

much more difficult to predict and require a fair degree of

have a requirement to complete and maintain a DWSP by

lateral thinking to anticipate. It is not surprising that they

December 31, 2013.

are sometimes missed.

ESRD will review the DWSP template after the initial completion date of December 31, 2013. Part of the ongoing commitment to the adoption of this policy direction is to

IMPLEMENTATION

develop and adopt a process of ‘continual improvement’ both to the content of the template but also to the mechan-

Implementation of the DWSP approach required a multi-

istic elements (such as the macros). Undoubtedly, there will

pronged approach. First, training workshops were devel-

be risks included in the present version of the template that

oped and delivered to operators and ESRD staff. Through

do not warrant being retained as ‘generic’ risks and, conver-

late 2011 and early 2012 a series of 15 workshops able to

sely, there may be system-specific risks that appear often

accommodate up to 35 participants were offered to all

enough to be adopted as generic risks in future template

municipal waterworks systems operators, and over 251 com-

revisions.

munities (with over 355 individual attendees) participated in these workshops. This training component continued beyond the provision of workshops with drinking water

CONCLUSIONS

operations specialists (who act as ‘circuit riders’ with geographically defined areas within the province) taking the

Alberta is the first province in Canada to introduce a regulat-

lead role in providing small-group or individual training sup-

ory requirement to have DWSPs for all regulated municipal

port to operators. The training elements were critical to the

drinking water systems. This requirement is thought to be

successful implementation of the DWSP.

the first such requirement in North America. As a leader

Changes to various legislative instruments and support-

in drinking water regulation and innovation for over 40

ing documents were also required to be made. Following

years, Environment and Sustainable Resource Development

9

D. C. Reid et al.

|

Implementation of Alberta’s drinking water safety plans

Water Quality Research Journal of Canada

|

49.1

|

2014

has eagerly embarked on the adoption, implementation and

the development and delivery of the training materials and

ongoing development of DWSPs as part of the department’s

workshops, Nigel Ayton, Rob Balfour, Stephen Sandilands,

ongoing commitment to Albertans to ensure they have a safe

Tommy Seggie, Scottish Water Horizons Ltd; Dr Lisa

and secure source of drinking water for their own health and

Barrott, Tim Darlow, MHW; Paul Yardley, Performance

the economic health of Alberta’s communities. The

Partnerships; and Ms Valerie Nelson for assistance with

approach developed in Alberta will be of interest to other

the preparation of this manuscript.

jurisdictions looking to adopt the DWSP concept where

The views expressed in this paper are those of the

issues of capacity constraints, geographic remoteness and

authors and do not necessarily reflect the policies of the

lack of resources may have been considered as significant

Government of Alberta or the Department of Environment

barriers to successful adoption. ESRD has demonstrated

and Sustainable Resource Development.

that it is possible to develop and implement DWSPs for all communities provided there is appropriate consideration given to the needs of the end-users and a commitment to ongoing partnerships with communities to support the implementation of the DWSP approach.

ACKNOWLEDGEMENTS The authors gratefully acknowledge the development of the DWSP template, Peter Dowswell, Water Safety Solutions;

REFERENCES Binnie, C. & Kimber, M.  Basic Water Treatment, 4th edn.

Thomas Telford, London.

Rizak, S. & Hrudey, S.  Strategic Water Quality Monitoring for Drinking Water Safety. The Cooperative Research Centre for Water Quality and Treatment. Research Report 27.

World Health Organization (WHO)  Guidelines for Drinkingwater Quality, 4th edn. World Health Organization, Geneva, Switzerland.

First received 10 December 2012; accepted in revised form 4 December 2013. Available online 17 December 2013

10

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Estimating nitrate loading from an intensively managed agricultural field to a shallow unconfined aquifer P. J. Kuipers, M. C. Ryan and B. J. Zebarth

ABSTRACT Nitrate loading from an intensively managed commercial red raspberry field to groundwater in the Abbotsford-Sumas Aquifer, British Columbia was estimated over a 1 yr period and compared with the nitrogen surplus calculated using a simple nitrogen budget. Nitrate loading was estimated as the product of recharge (estimated from climate data as total precipitation minus potential evapotranspiration (PET)) and monthly nitrate concentration measured at the water table. Most nitrate loading occurred when nitrate, accumulated in the root zone over the growing season, was leached following heavy autumn rainfall events. Elevated groundwater nitrate concentrations at the water table during the growing season when recharge was assumed to be negligible suggested that the nitrate loading was underestimated. The estimate of annual nitrate loading to the water table was high (174 kg N ha 1) suggesting that the tools currently available to growers to manage N in

P. J. Kuipers M. C. Ryan Department of Geoscience, University of Calgary, 2500 University Dr. NW, Calgary, Alberta, Canada T2N 1N4 B. J. Zebarth (corresponding author) Potato Research Centre, Agriculture and Agri-Food Canada, PO Box 20280, Fredericton, New Brunswick, Canada E3B 4Z7 E-mail: Bernie.Zebarth@agr.gc.ca

raspberry production are not adequate to protect groundwater quality. The calculated nitrogen surplus from the nitrogen budget (180 kg N ha 1) was similar to the measured nitrate loading suggesting that simple nitrogen budgets may be relatively effective indices of the risk of nitrate loading to groundwater.

Key words

| groundwater nitrate, nitrogen budget, red raspberry, Rubus idaeus

ABBREVIATIONS bgs below ground surface

increased mineral fertilizer and manure inputs (Rupert

EL

). Although there may be some dispute over the health

latent evapotranspiration

MW monitoring well PET potential evapotranspiration

risk associated with nitrate in water (Powlson et al. ), human health issues remain controversial (Ward et al. ) and aquatic ecosystem effects increasingly problematic (Vitousek et al. ; Galloway et al. ). Groundwater

INTRODUCTION

nitrate concentrations commonly exceed the drinking water guidelines for human health (10 and 11.3 mg

Nitrate is the most ubiquitous chemical contaminant in

NO3-N L 1 in the USA and EU, respectively), particularly

groundwater aquifers world-wide (Spalding & Exner )

in unconfined aquifers (Nolan & Stoner ; Rivett et al.

and groundwater nitrate contamination is increasingly

), which has important implications for management

being recognized as a long-lasting issue with broad environ-

of public water supplies.

mental and health implications (Puckett et al. ).

While non-point source groundwater nitrate contami-

Groundwater nitrate concentrations in many agricultural

nation can originate from various land uses, in many

regions have been increasing since about the 1950s in

cases nitrate contamination is associated with intensive

response to commercial production of mineral fertilizer

agricultural production (Power & Schepers ; Strebel

products, and have continued to increase over time with

et al. ; Puckett et al. ). Nitrate leaching occurs

doi: 10.2166/wqrjc.2013.136

11

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

when infiltrating water and elevated nitrate concen-

management practices with nitrate loading. Measurement

trations coexist in the soil (Meisinger & Delgado ).

of residual soil nitrate at the end of the growing season

Thus, the risk of nitrate leaching is enhanced by high min-

has been used as a proxy for the risk of nitrate leaching

eral N fertilizer inputs, high manure inputs, and cropping

from the soil root zone, and was used to assess the effects

systems with low efficiency of N utilization (Strebel et al.

of fertilizer and manure practices on the risk of nitrate

; Ruan & Schepers ). High manure inputs are fre-

leaching (Dean et al. ; Zebarth et al. ), however

quently associated with high livestock densities (Strebel

it is unclear how residual nitrate is quantitatively linked

et al. ) because it is not economical to transport

to nitrate loading to groundwater. As a result, the risk of

manure far from the animal operations (Eghball &

nitrate loading to the aquifer has been assessed primarily

Power ). The risk of nitrate contamination is

using N budget approaches (Zebarth et al. , ).

enhanced in sandy soils with limited water holding

The overall risk of nitrate contamination within a given

capacity, high water inputs as rainfall or irrigation (Strebel

field or geographic region may be assessed using a simple

et al. ; Power & Schepers ), and in permeable sur-

N budget approach (Meisinger et al. ), however it is

ficial aquifers which have limited ability to attenuate

unclear how reliable this approach is in providing quanti-

nitrate contamination (Rivett et al. ).

tative estimates of nitrate loading due to the large

The transboundary Abbotsford-Sumas Aquifer, located

uncertainties in some components of the N budgets (e.g.,

in British Columbia, Canada and Washington, USA, has

gaseous N losses through volatilization and denitrifica-

elevated nitrate concentrations over a wide portion of the

tion). These uncertainties in the nitrate loading associated

aquifer (Liebscher et al. ). This coarse sand and gravel

with specific management practices have made develop-

aquifer is highly vulnerable to nitrate contamination due

ment and adoption of practices to mitigate nitrate

to a thin and permeable surficial soil layer, high annual pre-

contamination challenging.

cipitation, permeable subsurface materials, and intensive 15

18

One approach to directly link changes in crop and soil

O isotopes, Wasse-

management practices with changes in groundwater nitrate

naar () concluded that nitrate in the aquifer originated

concentrations is through measurement of nitrate loss from

primarily from poultry manure. The elevated nitrate concen-

the root zone using a variety of techniques (Mulla & Strock

trations in the aquifer were attributed primarily to N inputs

). Nitrate loss from the root zone can be assessed using

as manure and mineral fertilizer in excess of crop N

soil sampling (Beaudoin et al. ), however this approach

removals in red raspberry production (Zebarth et al. ).

is not suitable for all soils and is not practical in coarse sand

Historic increases in groundwater nitrate concentrations

and gravel deposits. Nitrate loss can also be estimated using

have been attributed to intensification of agricultural

instrumented tile-drain plots (Milburn et al. ), however

production over the aquifer (Zebarth et al. ).

tile-drainage is not effective or common in these highly per-

agricultural production. Using

N and

A groundwater monitoring programme was initiated in

meable coarse sand and gravel deposits. Nitrate leaching

the early 1990s to monitor spatial and temporal changes in

can be measured directly using samplers such as the passive

groundwater nitrate concentrations (Hii et al. ). It is

capillary wick samplers ( Jabro et al. ) or equilibrium-

difficult, however, to relate groundwater nitrate concen-

tension lysimeters (Brye et al. ) or estimated from a com-

tration with specific land use practices due to the small

bination of nitrate concentration measured using suction

land holdings, large annual fluctuations in water table

lysimeters in combination with estimates of recharge rate

elevations and high groundwater flow rates. Simulation

(Vázquez et al. ). These approaches are generally effec-

modelling of nitrate transport through the vadose zone of

tive for comparison of management practices on a small plot

the Abbotsford-Sumas Aquifer has demonstrated that the

scale, but are limited in their ability to provide robust esti-

aquifer is highly vulnerable to nitrate loading from agricul-

mates of nitrate loading at a field scale due to the very

tural production (Chesnaux et al. ; Chesnaux & Allen

high inherent spatial variation in soil physical and biochemi-

), however limited field measurements for model

cal processes and the difficulty in quantifying preferential

calibration limit use of the simulation models to link

flow.

12

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

An alternative approach is to directly link changes in crop and soil management practices with changes in groundwater nitrate concentrations through monitoring nitrate at the water table. This approach provides better integration across time and space, and hence has the potential to provide a more robust estimate of nitrate loading. However, when making measurements at the water table, it can be more difficult to link nitrate loading to groundwater with the specific field or management area under study. This is particularly true in highly permeable

aquifers

where

groundwater

flow

may

be

dominantly horizontal and where land holdings are small (e.g., <1 ha). Stites & Kraft () quantified nitrate loading to groundwater in the Wisconsin central sand plain from a large (44 ha) agricultural field. Detailed depth profiles of nitrate and chloride concentrations below the water table were monitored. Annual spring applications of mineral fertilizer provided a chloride tracer that could be used to distinguish among years,

Figure 1

|

Site location (inset) and plan view showing the experimental field, the location of groundwater monitoring wells (MW1 and MW2), the time-weighted average groundwater flow direction (heavy line) and the range of variation in groundwater flow direction over the year of monitoring, and the surrounding land use.

Individual raspberry and blueberry fields are shaded grey, and one residential property was present. The white areas between fields are grassed laneways.

and allowed the mass of nitrate leached from the agricultural field within an individual year to be estimated. This

The aquifer is comprised of a succession of glacio-fluvial

approach was used to quantify the nitrate loading from a

sand and gravel interspersed with minor till and clayey silt

number of agricultural crops (Stites & Kraft ; Kraft &

lenses of unknown continuity collectively termed the

Stites ).

Sumas Drift (Armstrong et al. ; Easterbrook ). The

The objectives of this study were to estimate nitrate load-

aquifer is up to 70 m thick in some locations but has not

ing from an intensively managed commercial red raspberry

been fully characterized (Dakin ). The Sumas Drift is

field to groundwater in the Abbotsford-Sumas Aquifer over

underlain by low permeability glacio-marine and marine

a 1 yr period using calculated water surplus and monthly

sediments of the Fort Langley Formation (Halstead ).

groundwater nitrate concentrations measured at the water

The dominant soil types overlying the aquifer consist of

table, and to compare this estimate of nitrate loading with

up to 70 cm or more of silty aeolian deposits over sandy and

the nitrogen surplus calculated using a simple N budget

gravelly glacio-fluvial deposits (Luttmerding ). The soils

approach.

are generally well drained and are intensively managed. The dominant agricultural crop is red raspberry (Rubus idaeus L.) with significant areas in blueberry (Vaccinium corymbosum), forage grass (Dactylis glomerata L.) and pas-

METHODS

ture. A large number of poultry broiler, layer and turkey operations are located over the aquifer (Zebarth et al. ).

Hydrogeological setting

The aquifer is recharged primarily by direct precipiThe study site was located in the Abbotsford-Sumas Aquifer.

tation (Liebscher et al. ). Average annual precipitation

The aquifer covers an area of approximately 100 km2 in Brit-

(1971–2000) is 1,573 mm at the Abbotsford Airport, includ-

ish Columbia, Canada and a similar area in Washington

ing 64 cm as snow (Environment Canada ). About 70%

State, USA (Liebscher et al. ). The aquifer is located

of this precipitation occurs outside of the period of active

southwest of Abbotsford, British Columbia (49 30 N 122

crop growth in October to March. Horticultural crops are

170 W) (Figure 1).

commonly irrigated during the growing season. Average

W

W

13

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

annual temperature is 10.0 C (Environment Canada ).

W

Previous recharge estimates range from 65 to 70% of precipitation (Scibeck & Allen ), with an estimated 1 m of recharge for the Abbotsford Airport area under average precipitation conditions.

The water table varies from 40 m to < 2 m below ground surface, depending on the location and time of year (Kohut ; Wassenaar ; Environment Canada, unpublished data). The water table is also highly responsive to recharge, changing by as much as 3.5 m or more in elevation in a given year (Environment Canada, unpublished data).

Published estimates of hydraulic conductivity range from 3.8 × 10 6 to 2.8 × 10 2 m s 1 (Culhane ; Cox & Kahle ; Stasney ; Chesnaux et al. ), with a regional best estimate of 9.5 × 10 4 m s 1 (Cox & Kahle ). Porosity is generally assumed to be around 0.35 with resultant groundwater velocities in the order of 450 m yr 1 (Liebscher et al. ; Cox & Kahle ; Mitchell et al. ).

Study site and groundwater monitoring installations The study site was an 80 × 100 m field located within a portion of the Abbotsford-Sumas Aquifer which is known to have elevated groundwater nitrate concentrations (Liebscher et al. ; Hii et al. ) and which had relatively shallow water table depth to facilitate sampling (Figure 1). The study field was cropped to a red raspberry

Figure 2

|

Schematic diagram of bundle piezometer. The sampling ports are located 100 mm apart.

stand established in 2005 and surrounded by commercial red raspberry fields. Soils at the study site belong to the

using zip ties. Screening (210 micron Nytex®) covered the

Abbotsford soil series, which are defined as having 0.2–

ends of individual tubes and the centre stalks and was

0.5 m of medium textured aeolian deposit over gravelly out-

held in place with zip ties.

wash and are classified in the Canadian soil classification

The first bundle piezometers were installed in October

system as Orthic Humo-Ferric Podzols (Luttmerding ).

2006 when two wells (MW1A and MW1B, Figure 1) with

Field observations during monitoring well installation con-

36 tubes of 6.35 mm OD vinyl tubing each were installed

firmed that a 0.25–0.45 m deep surficial aeolian deposit

approximately 6 m apart with sample ports spanning differ-

was underlain by glacial sand and gravel at the study site.

ent depth intervals. Combining results from the two

Two installations of multi-level monitoring ‘bundle’

piezometer intervals provided a continuous 8.8 m sampling

piezometers (MacFarlane et al. ) were installed on the

profile terminating approximately 10 m below the ground

down-gradient edge of the study field (Figure 1). The

surface. A solid-stem auger was used to pre-drill the well

bundle piezometers consisted of a centre PVC stalk

holes, and was followed by a hollow stem auger used to

(12.7 mm diameter) surrounded by 36 or 72 smaller tubes

install the bundle piezometers. A sand pack was used to

which acted as dedicated sampling ports and terminated at

backfill the annulus to 1–1.5 m below ground surface (bgs)

100 mm incremental depths (Figure 2; MacFarlane et al.

at which point it was completed with a bentonite well seal

). The smaller tubes were secured to the centre stalk

and a cover was cemented in place at grade.

14

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

The second installation (MW2, Figure 1) occurred in

were collected using flow rates kept below 100 mL min 1 to

January 2007 using smaller 3.18 mm OD polyethylene

minimize mixing and encourage localized sample collec-

tubing allowing for a complete profile within one hole.

tion. The 20 mL samples were normally collected from

Instead of a sand pack, the non-cohesive in situ material

each working tube in a series with the exception of three

was permitted to collapse around the piezometer to the

sampling events during which sample frequency was

water table (MacFarlane et al. ; Ryan et al. ), at

reduced to every second port for all or a portion of the

which point it was backfilled using the native material

event due to time and weather constraints. Field doubles

obtained from drill cuttings to 1.52 m bgs. The installation

and blanks were collected on each sampling date. All

was sealed with bentonite and cement as described for the

samples were stored on ice prior to being refrigerated or

first installation.

frozen based on expected time between collection and analysis (frozen if 4 or more days expected).

In the laboratory, a 5 mL sub-sample was filtered through

Groundwater sampling

0.2 μm syringe filters and analysed for the concentration of Groundwater sampling was conducted on 12 dates in MW1

NO 3 using a Dionex ICS-1000 Ion Chromatograph with

from 2 November 2006 to 13 October 2007 and on 9 dates

A540 auto-sampler (APHA (American Public Health Associ-

in MW2 from 2 February 2007 to 13 October 2007. The

ation) ); Standard Method 4010-B). In-house standards,

times between sampling events ranged from 2 weeks to 2

which compare within 0.6% with commercial nitrate stan-

months, and were more frequent during the winter months

dards, were utilized for all ion chromatograph analysis.

when greater recharge was expected to occur.

Samples were analysed in a randomized order to reduce

Water table elevation was monitored in the centre stalks

errors associated with instrument drift. For quality control

of the bundle piezometers during each sampling event.

purposes, laboratory and field duplicates (one each per 10–

These water table elevations were used to confirm which

20 samples) were analysed where required. Where two or

sampling port was closest to the water table on each

more samples were analysed for a given sample event, an

sampling day. The irregular timing and relatively low fre-

average value was used in the data analysis. Average relative

quency of measurements made it difficult to use these

per cent difference for field and laboratory duplicates was less

water table elevations to estimate seasonal changes in

than 2.2% for nitrate (n ¼ 407, SD ¼ 7.3).

water table elevation. Consequently, water table elevations collected monthly from Environment Canada piezometers

Estimation of nitrate loading

(n ¼ 19) (Environment Canada, unpublished data) were used to generate water table elevation contour maps using

Nitrate loading in the current study could not be estimated

the Kriging algorithm within Surfer® as described by

as described by Stites & Kraft () because there were

Kuipers (). The contours in a 250 m radius of the

no seasonal patterns in chloride concentrations (Kuipers

study site were subsequently utilized to estimate study site

) that could be used to distinguish which year the nitrate

water table elevations, gradients and direction of flow. The

was leached from the root zone, and because of the small

network of Environment Canada piezometers span the aqui-

size of the study field. Consequently, nitrate loading to the

fer (Cox & Kahle ) and include six piezometers within

water table was estimated by multiplying estimated ground-

1 km of the study site.

water recharge with groundwater nitrate concentration

Groundwater samples were collected in the field using a peristaltic pump (Geotec II) with silicon tubing attached to

measured in the shallowest multi-level sampling port.

Nitrate loading was estimated for the 1 yr period from 1

volumes

November 2006 to 31 October 2007. The nitrate loading cal-

(5–750 mL depending on depth) were purged and all

culation was performed for 12 time intervals, centred on the

sample bottles were rinsed in the field three times using

12 sampling dates. Each time interval extended from

sample water prior to sample collection (Saunders ).

midway between the preceding sampling date and the cur-

Starting at the water table and working downward, samples

rent sampling date (or 1 November for the first sampling

each

bundle

piezometer

tube.

Three

well

15

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

date), to midway between the current sampling date and the

inputs were compared with outputs at the field level, similar

subsequent sampling date (or 31 October for the final

to previous studies (Zebarth et al. , ). The budget calcu-

sampling date).

lation assumed that the soil N processes were at equilibrium,

Groundwater recharge was estimated within each time

and consequently soil N cycling within the field (e.g., plant N

interval using a water balance method (Scanlon et al.

uptake, crop residue additions, soil N mineralization) could

). Precipitation data were obtained from an Environ-

be ignored. This approach has been used previously at regional

ment Canada climate station located at the Abbotsford

and field scales for cases where detailed information on N pro-

Airport. Using information from the same climate station,

cesses is not available (Meisinger et al. ).

daily potential evapotranspiration (PET) values were

Nitrogen inputs to the field included N from mineral fer-

derived by Agriculture Canada using Method 1 described

tilizers, manure, atmospheric deposition, and irrigation

in Baier & Robertson () to estimate latent evapo-

water. Total N input from manure was estimated based on

transpiration (EL) from simple weather observations,

measured N-content of the manure and the application

combined with the method described in Baier () to con-

rate reported by the land owner. Similarly, mineral fertilizer

vert values of EL to PET. Runoff was assumed to be

N applications were as reported by the land owner. Esti-

negligible based on the low site topography, well-drained

mates for atmospheric deposition were assigned the

soils, and the sandy nature of the underlying aquifer

average measured value reported in Belzer et al. () for

(Chesnaux et al. ). Irrigation contributions to recharge

a site at Abbotsford. The quantity of irrigation applied was

were assumed to be negligible due to high evaporation

not measured. A conservative estimate for irrigation N

during the summer months when irrigation is applied

inputs was made by assuming that the irrigation applied

(Zebarth et al. ). Thus, recharge within the time interval

was equal to the water deficit for the months of May, June,

was estimated as precipitation less PET, an approach used

July and August 2007 when PET exceeded precipitation.

previously to estimate recharge in this aquifer (Cox &

The water deficit was calculated as PET less precipitation

Kahle ; Chesnaux et al. ). In time intervals when

where PET was estimated as described above. Nitrate

PET exceeded precipitation, recharge was assumed to be

inputs as irrigation water were estimated by multiplying

zero. Note that PET and recharge were also calculated on

the assumed irrigation rate by the average nitrate concen-

a monthly basis such that monthly values during the moni-

tration of the irrigation water (7.5 mg N L 1) measured on

toring period could be compared with monthly values

three dates: 5 July, 5 August and 6 August 2007.

from historic climate data.

The measured nitrate concentration in the shallowest

Nitrogen outputs from the root zone included denitrification, volatilization, and crop removal from the field as

sampling port, averaged between MW1 and MW2 when

berries. Denitrification was estimated as 8% of the total N

both bundle piezometers were sampled, was assumed to

applied as mineral fertilizer or manure (Zebarth et al.

apply to the entire time interval. The quantity of recharge

). The N removed in harvested fruit was estimated at

(with units of m3 ha 1) was then multiplied by the nitrate

20 kg N ha 1 yr 1 (Kowalenko ; Dean et al. ). It

3

concentration (kg m ) to estimate the quantity of nitrate

was assumed that 20% of the total N in the applied poultry

loading to groundwater within each individual time interval

manure was lost by volatilization (Zebarth et al. ). It was

(kg ha 1). The quantity of nitrate loading was then summed

also assumed that a net increase in soil organic N storage

across time intervals to estimate nitrate loading over the 1 yr

occurred which was estimated as 20% of the total N in the

monitoring interval.

manure (Meisinger et al. ).

Nitrogen budget

RESULTS A nitrogen budget was calculated to estimate the annual surplus of N which may contribute to nitrate leaching. The

There were some seasonal changes in the direction of

budget calculated a mass balance for the root zone whereby

groundwater flow (Figure 1). This variation was likely due

16

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

in part to the influence of irrigation wells located in the general vicinity of the experimental site. However, the direction of groundwater flow was such that the groundwater samplers were located down-gradient of the study field over the entire monitoring period.

Precipitation

during

the

monitoring

period

was

1,938 mm (Table 1). This precipitation was 24% higher than the 30 yr (1978–2007) average of 1,560 mm. The increased precipitation primarily reflected 76 and 126% above-average precipitation in November 2006 and March 2007, respectively. Precipitation was 100 mm or more for each month from November 2006 to April 2007 (Figure 3(a)). Estimated PET during the monitoring period was 598 mm, which was slightly less than the 30 yr average of 628 mm. PET was highest from May to August (Figure 3(b)). Estimated recharge during the monitoring period (calculated on a monthly basis as precipitation less PET) was 1,582 mm, which is 41% higher than the 30 yr average of 1,122 mm (Table 1). Recharge was predicted to occur in every month except May to August 2007 (Figure 3(c)). The estimated water deficit during these 4 months was 241 mm, 27% higher than the 30 yr average of 190 mm (Table 1). Air temperature during the monitoring W

W

period (10.5 C) was similar to the 30 yr average (10.3 C).

Thus, climatic conditions during the monitoring period were generally drier than normal during the period of active crop growth but considerably wetter than normal than during the winter recharge period.

Figure 3

|

Estimation of recharge or water deficit for the 1 yr (November 2006 to October 2007) monitoring period calculated on a monthly basis from data obtained

Water table elevation at the start of the monitoring period in November 2006 increased over time until March

from the Abbotsford Airport climate station: (a) monthly total precipitation; (b) monthly potential evapotranspiration (PET); and (c) monthly estimated recharge or water deficit calculated as precipitation minus PET.

2007 (Figure 4(a)), which is consistent with high monthly Table 1

|

Comparison of climate parameters during the experimental monitoring period

precipitation from November 2006 until March 2007

(November 2006 to October 2007) in comparison with 30 yr (1978 to 2007) climate normals as measured at the Abbotsford Airport, calculated on a

(Figure 3(a)). Water table elevation subsequently declined

monthly basis

reaching a minimum value at the end of the monitoring period in October 2007, which is consistent with a calcu-

Parameter

Value W

Air temperature ( C) Precipitation (mm)

10.5

normal

102

10.3

lated water deficit for the period from May to August 2007. The maximum change in water table elevation over the monitoring period was 1.9 m.

124

1,560

There were also seasonal patterns in the magnitude and

598

95

628

depth distribution of nitrate concentrations below the

1,582

141

1,122

water table (Figure 5). Overall, nitrate concentrations

241

127

190

generally increased with depth, indicating that nitrate

Water deficita (mm) a

30 yr climate

% of 30 yr normal

1,938

PETa (mm) Estimated rechargea (mm)

Monitoring period

Calculation method described in text.

loadings to the aquifer are greater for fields up-gradient of

17

P. J. Kuipers et al.

|

Nitrate loading from an agricultural ďŹ eld

Water Quality Research Journal of Canada

Figure 5

|

|

49.1

|

2014

Depth distribution of nitrate concentration below the water table at four times during the 1 yr monitoring period. Note that the ground surface is at 49.3 masl.

For the simple N budget, N inputs were estimated to be 340 kg N ha 1 (Table 2). The largest N input was manure (solid poultry layer manure containing 5.4% total N, band applied on the crop rows and not incorporated), estimated at 287 kg total N ha 1. This rate of application was chosen based on the assumption that approximately one-third of the manure total N would be plant-available in the year of Figure 4

|

Calculation of nitrate loading during the 1 yr (November 2006 to October 2007)

Table 2

|

Calculated N balance for the experimental ďŹ eld during the experimental monitoring period (November 2006 to October 2007)

monitoring period calculated over 12 time intervals based on the timing of groundwater sampling: (a) water table elevation on each sampling date (ground surface is at 49.3 masl); (b) measured nitrate concentrations in the shallowest sampling port of MW1 and MW2 used for the calculation of nitrate loading; and (c) calculated nitrate loading for each time interval.

Budget componenta

Value (kg N ha 1)

N inputs Manure (total N)

287

the study site than for the study site. For depths close to

Mineral fertilizer N

26

the water table, distinctly different patterns of nitrate

Irrigation water N

18

concentration with depth were observed on different

Atmospheric deposition Total inputs

sampling dates. Nitrate concentration at the sampling port closest to the water table was highest in November

9 340

N outputs other than leaching

2006 (23.6 mg N L 1), averaged about 5 mg N L 1 for

Harvested fruit

20

January to June 2007, then increased to about 15 mg N L

Volatilization (manure)

57

from August to October 2007 (Figure 4(b)). Nitrate loading

DenitriďŹ cation

25

Increase in soil organic N storage

57

1

1

was estimated to be 174 kg N ha . The time periods of

Total outputs

greatest loading (Figure 4(c)) generally coincided with the times of maximum nitrate concentration at the water table (Figure 4(b)).

Surplus of N which has the potential to leach a

160 180

Assumptions used in developing the budget are described in the text.

18

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

application. The mineral fertilizer N rate (fertilizer analysis

obtain a robust estimate of nitrate loading given the high

[N-P2O5-K2O] of 8.5-15-30) was low at 26 kg N ha 1. Inter-

spatial variation in the crop and in soil N processes. It was

1

estingly, N inputs in irrigation water were 18 kg N ha ,

not possible in the current study to use the approach of

equivalent to more than half of the N applied as mineral

Stites & Kraft () because the depth distribution of chlor-

fertilizer. Atmospheric N deposition was estimated at

ide (Kuipers ) could not be used to identify which nitrate

9 kg N ha 1, however, there is a large degree of uncertainty

could be attributed to the current year. This was not unex-

in this estimate because there are few measured values of

pected because of the small size of the land holdings, and

atmospheric deposition in this region. It is important to

because high annual recharge makes identification of the

note that from the perspective of the producer, the

chloride associated with mineral fertilizer application pro-

supply of plant-available N was expected to be about

blematic. Efforts to apply a high rate of chloride at the

122 kg N ha 1 (i.e., mineral fertilizer N plus one-third of

field boundary to act as a marker failed when the applied

total N in manure) because there are currently no practical

tracer was not detected (Kuipers ).

means for accounting for soil N mineralization or atmospheric N deposition, and N in irrigation water is not considered.

The N outputs other than nitrate leaching were estimated 1

Consequently, nitrate loading in the current study was calculated from an estimate of recharge in combination with nitrate concentration in groundwater measured as

(Table 2). The largest outputs were for vol-

close to the water table as possible to represent the nitrate

atilization of manure N and increase in soil organic N

concentration in recent recharge. This approach has the

storage, each at 57 kg N ha 1. The N outputs for harvested

advantage of not requiring a tracer to distinguish what

fruit and denitrification were smaller at 20 and 25 kg N

year the nitrate was associated with and allows application

ha 1, respectively. It should be noted that with the exception

of the approach on relatively small land holdings. This

of the N in harvested fruit, there is a large degree of uncer-

approach does, however, make two important assumptions.

tainty with the estimated values of the N outputs.

First, it assumes that the recharge rate can be estimated with

at 160 kg N ha

1

(Table 2).

sufficient accuracy based on climate data. Although ground-

This surplus provides an approximation of the N which is

water recharge is an elusive parameter to estimate (Scanlon

available for nitrate leaching. This estimate is done at a

et al. ), our estimate agrees well with previous modelling

field scale, and assumes the system is at equilibrium, and

derived (Scibeck & Allen ) and water balance

therefore assumes that the internal N cycling within the

(Chesnaux et al. ) approaches. Second, it assumes that

field can be ignored. Consequently, this surplus is not a

the transit time for recharge from the root zone to the

predictor of nitrate leaching in any given year, but does pro-

water table is sufficiently rapid that the nitrate concentration

vide an index of the risk of nitrate leaching. By combining

at the water table can be associated with the recharge which

this

recharge of

occurred during the same time period. While this is not true

1,122 mm (Table 1), the resulting groundwater nitrate con-

of many aquifers, it is a reasonable assumption at this study

centration would be estimated to be about 16 mg N L 1.

site given the high annual recharge rate, relatively shallow

The estimated N surplus was 180 kg N ha

N

surplus

with

estimated

annual

The estimated N surplus of 180 kg N ha

1

was 3% higher

than the nitrate loading estimated from the 1 yr groundwater monitoring period of 174 kg N ha 1.

water table, and coarse sediments below the surface soil.

The seasonal patterns of groundwater elevation in this study are consistent with previous studies in the Abbotsford-Sumas Aquifer. Water table elevations in the Abbotsford-Sumas Aquifer are commonly at a maximum in

DISCUSSION

February or March and at a minimum in October to December (Liebscher et al. ; Zebarth et al. ). In

In this study, we estimated nitrate loading from an individ-

comparison, seasonal patterns of groundwater nitrate con-

ual commercial red raspberry field with known N

centrations have not previously been measured at the

management through monitoring of groundwater nitrate

water table. Nitrate concentrations in deeper wells do exhi-

concentrations. This approach was chosen in order to

bit seasonal patterns, and in some cases multi-year cycles in

19

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

concentrations, but it can be difficult to interpret these tem-

winter period in all years (Kowalenko ). The above-

poral patterns in groundwater nitrate concentration because

average recharge during the monitoring period occurred

the age and source of the nitrate measured is uncertain

during the time period when significant recharge already

(McArthur & Allen ).

occurs. Since essentially complete leaching occurs in any

The patterns of nitrate loss from the root zone predicted

case, the amount of nitrate leaching will be controlled

from this study are consistent with previous studies of soil N

not by precipitation, but rather by temperature, soil proper-

cycling in this region. Soil nitrate commonly accumulates in

ties and management practices. The air temperature during

the root zone during the growing season in response to

the monitoring period was near normal, and the manage-

spring application of manure and/or mineral fertilizer and

ment practices were consistent with previous years.

net mineralization of organic N in soil organic matter

Consequently increased recharge would be expected to

(Dean et al. ; Zebarth et al. ), followed by essentially

result in reduced groundwater nitrate concentrations due

complete loss of nitrate during the winter period in response

to greater dilution, but not necessarily to increase the quan-

to rainfall events (Kowalenko ). While there is evidence

tity of nitrate leached from the root zone.

of some downward movement of nitrate during the growing

The magnitude of the nitrate loading to groundwater

season (Zebarth et al. ), the magnitude of nitrate loss

during the monitoring period (174 kg N ha 1) was quite

from the root zone is not well known. Thus, nitrate loss is

high and would be expected to result in groundwater nitrate

expected to occur primarily with the onset of heavy rainfall

concentrations above the drinking water guideline. This

events in autumn, and the quantity of nitrate accumulated

occurred despite efforts by the grower to manage N carefully

in the root zone at the end of the growing season is used as

according to recommended practices (Hughes-Games &

an index of the risk of nitrate leaching (Zebarth et al. ).

Zebarth ). The high loading may in part reflect

The high nitrate concentration in October 2006 and the

that the N inputs from the perspective of the grower

high nitrate loading from October to December 2006 are con-

(122 kg N ha 1 from mineral fertilizer N plus one-third of

sistent with much of the nitrate leaching occurring in

total N in manure) represented less than 40% of the total

response to flushing of accumulated nitrate from the root

N inputs estimated using the N budget (340 kg N ha 1).

zone with the onset of heavy autumn rainfall events.

This suggests that the tools currently available to growers

It is interesting to note that despite considerable

to manage N in raspberry production are not adequate to

recharge during the winter period, nitrate concentrations

protect groundwater quality. The estimated N surplus for

at the water table remain close to 5 mg N L

1

from February

this study site is consistent with previous estimates of the

to June of 2007. This suggests that there may continue to be

N surplus in this region. For example, a regional N budget

significant losses of nitrate from the root zone, presumably

for Matsqui South, which is similar to the geographic

originating from soil N mineralization and atmospheric

extent of the Abbotsford Aquifer, was estimated at

deposition, over the winter period.

238 kg N ha 1 using 1991 census data (Zebarth et al. ).

In the current study, it was assumed that no recharge

It is interesting to note that the estimate of nitrate load-

occurs during May to August 2007 because PET exceeded

ing over the 1 yr monitoring period (174 kg N ha 1) is

total precipitation. However, soil nitrate concentration at

within 5% of the calculated N surplus from the N budget

the water table increased markedly in August 2007. This

(180 kg N ha 1). Given the large uncertainties in some com-

suggests that some recharge did occur during the growing

ponents

season in response to irrigation, and therefore that nitrate

deposition and gaseous N losses from volatilization and

loading to groundwater was somewhat underestimated.

of

the

N

budget,

particularly

atmospheric

denitrification, these are very similar estimates of nitrate

Precipitation during the 1 yr monitoring period was

loading. It should be noted that while the nitrate loading

considerably higher than the 30 yr climate normal. This

was estimated based on groundwater nitrate concentration

would result in above-average recharge. However, it may

over a specific 1 yr monitoring period, the N budget

have had a limited effect on nitrate loading. Essentially

approach assumes the system is at equilibrium and that

all nitrate is leached from the root zone during the

the N surplus reflect conditions over a multi-year period,

20

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Water Quality Research Journal of Canada

|

49.1

|

2014

whereas the actual N surplus may vary from year to year in

Water Network. The cooperation and assistance of

response to changes in management practices and environ-

the anonymous grower on whose land this study was

mental conditions. Although it would have been helpful to

performed, the BC Raspberry Industry Development

have measured nitrate loading over a longer monitoring

Council, and Mark Sweeney of the British Columbia

period, this result suggests that the simple N budgets may

Ministry of Agriculture are gratefully acknowledged.

be relatively effective indices of the risk of nitrate loading to groundwater in this aquifer which has shallow permeable soils and high annual recharge.

CONCLUSIONS This study estimated nitrate loading from an intensively managed commercial red raspberry field to groundwater in the Abbotsford-Sumas Aquifer, British Columbia over a 1 yr period and compared the result with the N surplus calculated using a simple N budget. Nitrate loading was estimated as the product of recharge (estimated from climate data as total precipitation minus PET) and nitrate concentration at the water table. Most nitrate loading occurred when nitrate, accumulated in the root zone over the growing season, was leached following heavy autumn rainfall events.

Elevated groundwater nitrate concentrations during the growing season when recharge was assumed to be negligible suggested that the nitrate loading was underestimated.

Nitrate concentrations at the water table remain close to 5 mg N L 1 from February to June suggesting that there may continue to be significant losses of nitrate from the root zone over the winter period, presumably originating from soil N mineralization and atmospheric deposition.

The estimate of nitrate loading was high (174 kg N ha 1) suggesting that the tools currently available to growers to manage N in raspberry production are not adequate to protect groundwater quality. The calculated nitrogen surplus from the nitrogen budget (180 kg N ha 1) was similar to the measured nitrate loading suggesting that simple nitrogen budgets may be relatively effective indices of the risk of nitrate loading to groundwater.

ACKNOWLEDGEMENTS Financial support was provided by the Natural Sciences and Engineering Research Council (NSERC) and the Canadian

REFERENCES American Public Health Association (APHA)  Standard Methods for the Examination of Water and Wastewater, 21st edn. Water Environment Federation, Alexandria, VA, USA. Armstrong, J. E., Crandell, D. R., Easterbrook, D. J. & Noble, J. B.  Late Pleistocene stratigraphy and chronology in south west British Columbia and north west Washington. Geol.

Soc. Am. Bull. 76, 321–330.

Baier, W.  Evaluation of latent evaporation estimates and their conversion to potential evaporation. Can. J. Plant Sci. 51, 255–266.

Baier, W. & Robertson, G. W.  Estimation of latent evaporation from simple weather observations. Can. J. Plant Sci. 45, 276–284.

Beaudoin, N., Saad, J. K., van Laethem, C., Machet, J. M., Maucorps, J. & Mary, B.  Nitrate leaching in intensive agriculture in northern France: Effect of farming practices, soils and crop rotations. Agric. Ecosyst. Environ. 111, 292–310.

Belzer, W., Evans, C. & Poon, A.  Atmospheric Nitrogen Concentrations in the Lower Fraser Valley. Environment Canada, Aquatic and Atmospheric Science Division, Vancouver, BC. DOE FRAP 1997:1–23.

Brye, K. R., Norman, J. M., Bundy, L. G. & Gower, S. T.  Nitrogen and carbon leaching in agroecosystems and their role in denitrification potential. J. Environ. Qual. 30, 58–70.

Chesnaux, R. & Allen, D. M.  Simulating nitrate leaching profiles in a highly permeable vadose zone. Environ. Model.

Assess. 13, 527–539.

Chesnaux, R., Allen, D. M. & Graham, G.  Assessment of the impact of nutrient management practices on nitrate contamination in the Abbotsford-Sumas aquifer. Environ.

Sci. Technol. 41, 7229–7234.

Cox, S. E. & Kahle, S. C.  Hydrogeology, Ground-water Quality, and Sources of Nitrate in Lowland Glacial Aquifers of Whatcom County, Washington, and British Columbia, Canada. USGS Water Investigations Rept 98–4195.

Culhane, T.  Whatcom County Hydraulic Continuity Investigation – Part 1, Critical Well/Stream Separation Distances for Minimizing Stream Depletion. Open File Technical Report 93–08, Washington State Department of Ecology, pp. 20.

Dakin, A.  Groundwater Resources of the Basins, Lowlands and Plains: Fraser Lowland. Groundwater Resources of

21

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

British Columbia. BC Environment, Victoria, BC, pp. 9.1–9.27.

Dean, D. M., Zebarth, B. J., Kowalenko, C. G., Paul, J. W. & Chipperfield, K.  Poultry manure effects on soil nitrogen processes and nitrogen accumulation in red raspberry. Can. J. Plant Sci. 80, 849–860.

Easterbrook, D. J.  Geologic map of western Whatcom County, Washington. U.S. Geological Survey Map I-854-B. Eghball, B. & Power, J. F.  Beef cattle feedlot management.

J. Soil Water Conserv. 49, 113–122.

Environment Canada  Canadian climate normals 1971–2000.

Accessed July 4, 2011 at http://climate.weatheroffice.gc.ca/ climate_normals/index_e.html.

Galloway, J. N., Dentener, F. J., Capone, D. G., Boyer, E. W., Howarth, R. W., Seitzinger, S. P., Asner, G. P., Cleveland, C. C., Green, P. A., Holland, E. A., Karl, D. M., Michaels, A. F., Porter, J. H., Townsend, A. R. & Vörösmarty, C. J.  Nitrogen cycles: past, present, and future. Biogeochemistry 70, 153–226.

Halstead, E. C.  Ground Water Supply - Fraser Lowland, British Columbia. National Hydrology Research Institute Paper No. 26. Environment Canada, Saskatoon, p. 80.

Hii, B., Liebscher, H., Mazalek, M. & Touminen, T.  Groundwater Quality and Flow in the Abbotsford Aquifer, British Columbia. Environmental Conservation Branch, Environment Canada, Vancouver, BC. Hughes-Games, G. & Zebarth, B. J.  Nitrogen Recommendations for Raspberries for South Coastal British Columbia. Resource Management Branch, BC Ministry of Agriculture and Food, Abbotsford, British Columbia (BC), Canada. Agdex 530/541. Revised February 1999.

Jabro, J. D., Kim, Y., Evans, R. G., Iversen, W. M. & Stevens, W. B.  Passive capillary sampler for measuring soil water drainage and flux in the vadose zone: design, performance, and enhancement. Appl. Eng. Agric. 24, 439–446.

Kohut, A. P.  Groundwater Supply Capability, Abbotsford Upland: British Columbia. Ministry of Environment, Victoria, BC, Open File Report.

Kowalenko, C. G.  The dynamics of inorganic nitrogen in a Fraser Valley soil with or without spring or fall ammonium nitrate applications. Can. J. Soil Sci. 67, 367–382.

Kowalenko, C. G.  Growing season dry matter and macroelement accumulations in Willamette red raspberry and related soil-extractable macroelement measurements.

Can. J. Plant Sci. 74, 565–571.

Kraft, G. J. & Stites, W.  Nitrate impacts on groundwater from irrigated-vegetable systems in a humid north-central US sand plain. Agric. Ecosyst. Environ. 100, 63–74.

Kuipers, P. J.  Estimating Nitrate Contributions from a Raspberry Field Managed under Beneficial Management Practices to Groundwater Nitrate in a Shallow Unconfined Aquifer. MSc Thesis, Dept of Civil Engineering, Univ.

Calgary, Calgary, AB, Canada.

Liebscher, H., Hii, B. & McNaughton, D.  Nitrates and Pesticides in the Abbotsford Aquifer, Southwestern British

Water Quality Research Journal of Canada

|

49.1

|

2014

Columbia. Inland Waters Directorate, Environment Canada, North Vancouver, BC. Luttmerding, H. A.  Soils of the Langley-Vancouver Map Area. BC Soil Survey Report No. 15.

MacFarlane, D. S., Cherry, J. A., Gillham, R. W. & Sudicky, E. A.  Migration of contaminants in groundwater at a landfill: a case study: 1. Groundwater flow and plume delineation. J. Hydrol. 63, 1–29.

McArthur, S. & Allen, D.  Abbotsford-Sumas Aquifer: Compilation of a Groundwater Chemistry Database with Analysis of Temporal Variations and Spatial Distributions of Nitrate Contamination. Report prepared for BC Ministry of Water, Land and Air Protection, Climate Change Branch.

Meisinger, J. J. & Delgado, J. A.  Principles for managing nitrogen leaching. J. Soil Water Conserv. 57, 485–498.

Meisinger, J., Calderon, F. & Jenkinson, D.  Soil nitrogen budgets. In: Nitrogen in Agricultural Systems (J. S. Schepers & W. R. Raun, eds). Agronomy Monograph 49. American Society of Agronomy, Crop Science Society of America, Soil Science Society of America, Madison, WI, pp. 505–562.

Milburn, P., Richards, J. E., Gartley, C., Pollock, T., O’Neill, H. & Bailey, H.  Nitrate leaching from systematically tiled potato fields in New Brunswick, Canada. J. Environ. Qual. 19, 448–454.

Mitchell, R. J., Babcock, R. S., Gelinas, S., Nanus, L. & Stasney, D. E.  Nitrate distributions and source identification in the Abbotsford-Sumas Aquifer, northwestern Washington State. J. Environ. Qual. 32, 789–800.

Mulla, D. J. & Strock, J. S.  Nitrogen transport processes in soil. In: Nitrogen in Agricultural Systems (J. S. Schepers & W. R. Raun, eds). Agronomy Monograph 49. American Society of Agronomy, Crop Science Society of America, Soil Science Society of America, Madison, WI, pp. 361–540.

Nolan, B. T. & Stoner, D. D.  Nutrients in groundwaters of the conterminous United States, 1992–1995. Environ. Sci. Technol. 34, 1156–1165.

Power, J. F. & Schepers, J. S.  Nitrate contamination of ground water in North America. Agric. Ecosyst. Environ. 26, 165–187.

Powlson, D. S., Addiscott, T. M., Benjamin, N., Cassman, K. G., de Kok, T. M., van Grinsven, H., L’hirondel, J.-L., Avery, A. A. & van Kessel, C.  When does nitrate become a risk for humans? J. Environ. Qual. 37, 291–295.

Puckett, L. J., Tesoriero, A. J. & Dubrovsky, N. M.  Nitrogen contamination of surficial aquifers – a growing legacy.

Environ. Sci. Technol. 45, 839–844.

Rivett, M. O., Buss, S. R., Morgan, P., Smith, J. W. N. & Bemment, C. D.  Nitrate attenuation in groundwater: a review of biogeochemical controlling processes. Water Res. 42, 4215–4232.

Ruan, W. R. & Schepers, J. S.  Nitrogen management for improved use efficiency. In: Nitrogen in Agricultural Systems (J. S. Schepers & W. R. Raun, eds). Agronomy Monograph 49. American Society of Agronomy, Crop Science Society of America, Soil Science Society of America, Madison, WI, pp. 675–693.

22

P. J. Kuipers et al.

|

Nitrate loading from an agricultural field

Rupert, M. G.  Decadal-scale changes of nitrate in groundwater of the United States, 1988–2004. J. Environ.

Qual. 37, S-240–S-248.

Ryan, M. C., MacQuarrie, K. T. B., Harman, J. & McLellan, J.  Field and modelling evidence for a ‘stagnant flow’ zone in the upper meter of sandy phreatic aquifers. J. Hydrol. 223, 223–240.

Saunders, L. L.  Manual of Field Hydrogeology. Prentice Hall, Upper Saddle River, NJ. Scanlon, B. R., Healy, R. W. & Cook, P. G.  Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeol. J. 10, 18–39.

Scibeck, J. & Allen, D. M.  Comparing modelled responses of two high-permeability, unconfined aquifers to predicted climate change. Glob. Planet. Change 50, 50–62.

Spalding, R. F. & Exner, W. E.  Occurrence of nitrate in groundwater – a review. J. Environ. Qual. 22, 392–402.

Stasney, D. V.  Hydrostratigraphy, Groundwater Flow and Nitrate Transport within the Abbotsford–Sumas Aquifer, Whatcom County, Washington. MS Thesis, Western Washington University, Bellingham, WA, USA. Stites, W. & Kraft, G.  Nitrate and chloride loading to groundwater from an irrigated north-central U.S. sand-plain vegetable field. J. Environ. Qual. 30, 1176–1184.

Strebel, O., Duynisveld, W. H. M. & Böttcher, J.  Nitrate pollution of groundwater in western Europe. Agric. Ecosyst.

Environ. 26, 189–214.

Vázquez, N., Pardo, A., Suso, M. L. & Quemada, M.  A methodology for measuring drainage and nitrate leaching in unevenly irrigated vegetable crops. Plant Soil 269, 297–308.

Water Quality Research Journal of Canada

|

49.1

|

2014

Vitousek, P. M., Aber, J. D., Howrath, R. W., Likens, G. E., Matson, P. A., Schindler, D. W., Schlesinger, W. H. & Tilman, D. G.  Human alteration of the global nitrogen cycle: sources and consequences. Ecol. Appl. 7, 737–750.

Ward, M. H., deKok, T. M., Levallois, P., Brender, J., Gulis, G., Nolan, B. T. & VanDerslice, J.  Workgroup report: drinking-water nitrate and health – recent findings and research needs. Environ. Health Persp. 113, 1607–1614.

Wassenaar, L. I.  Evaluation of the origin and fate of nitrate in the Abbotsford Aquifer using the isotopes of 15N and 18O in NO 3 . Appl. Geochem. 10, 391–405.

Zebarth, B. J., Bowen, P. A. & Toivonen, P. M. A.  Influence of nitrogen fertilization on broccoli yield, nitrogen accumulation and apparent fertilizer-nitrogen recovery. Can. J. Plant Sci. 75, 717–725.

Zebarth, B. J., Hii, B., Liebscher, H., Chipperfield, K., Paul, J. W., Grove, G. & Szeto, S. Y.  Agricultural land use practices and nitrate contamination in the Abbotsford Aquifer, BC, Canada. Agric. Ecosyst. Environ. 69, 99–112.

Zebarth, B. J., Kowalenko, C. G. & Harding, B.  Response of soil inorganic nitrogen content, and indices of crop yield, vigor and nitrogen status, to rate and source of nitrogen fertilizer in red raspberry fields. Commun. Soil Sci. Plant Anal. 38, 637–660.

Zebarth, B. J., Paul, J. W. & Van Kleeck, R.  The effect of nitrogen management in agricultural production on water and air quality: evaluation on a regional scale. Agric. Ecosyst.

Environ. 72, 35–52.

First received 21 June 2012; accepted in revised form 19 November 2012. Available online 27 August 2013

23

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Protecting our Great Lakes: assessing the effectiveness of wastewater treatments for the removal of chemicals of emerging concern Antonette Arvai, Gary Klecka, Saad Jasim, Henryk Melcer and Michael T. Laitta

ABSTRACT The Great Lakes and their connecting channels form the largest fresh surface water system on earth.

Over the past 10 years, focus on environmental monitoring has shifted to an array of recently discovered compounds known as ‘chemicals of emerging concern’ (CEC). These chemicals are found in products used daily in households, businesses, agriculture and industry, such as flame retardants, pharmaceuticals, personal care products, and pesticides. Wastewater treatment plants are among the important pathways by which CEC enter the Great Lakes, with concentrations highest in the vicinity of wastewater discharges. Treated sewage is often discharged into the nearshore waters, which also provide a source of drinking water to the public. In 2009–2011, the International Joint Commission addressed the need to assess the effectiveness of existing wastewater treatment technologies in the basin to remove CEC, as well as to gain insight on potential advanced technologies to improve their removal. This assessment encompassed three major activities, development of an inventory of municipal wastewater treatment plants that discharge in the basin; a survey of detailed operational data for selected wastewater facilities; and a comprehensive literature review and analysis of the effectiveness of various wastewater treatment technologies to remove chemicals of emerging concern.

Key words

| chemicals of emerging concern, Great Lakes, wastewater

Antonette Arvai Great Lakes Regional Office, International Joint Commission, Windsor, Ontario, Canada Gary Klecka The Dow Chemical Company, Midland, Michigan, USA Saad Jasim (corresponding author) SJ Environmental Consultants (Windsor) Inc., 2401-380 Pelissier Street, Windsor, ON N9A 6V7, Canada E-mail: sjenvcons@gmail.com Henryk Melcer Brown and Caldwell, Seattle, Washington, USA Michael T. Laitta US Section, International Joint Commission, Washington, DC, USA

INTRODUCTION Over the past 10 years, the emphasis on environmental moni-

compounds have regulations governing their release. Of con-

toring has shifted from the so-called legacy pollutants, such as

cern is the uncertainty of potential adverse effects on wildlife

polychlorinated biphenyls (PCBs), to a wide array of new

and humans due to chronic exposure to low concentrations

chemicals being discovered in the environment. The term

of these compounds. Some of these chemicals are accumulat-

‘chemicals of emerging concern’ has come to characterize

ing in sediments, birds, fish, and other aquatic life, as well as

the increasing awareness of the presence in the environment

in humans.

of many chemicals used by society, and the potential risk that

In October 2007, the International Joint Commission

they may pose to humans and ecosystems (Daughton ).

(Commission) established multi-Board priorities to be

Chemicals of emerging concern include new compounds

undertaken by its Great Lakes advisory groups during the

that have gained entry into the environment or those that

2007–2009 biennial reporting cycle. Within the context of

have been recently characterized due to increases in concen-

the Nearshore Framework Priority, the Chemicals of Emer-

trations in the environment or improvements in analytical

ging Concern Work Group was charged with reviewing the

techniques. In the United States and Canada, few of these

scientific and policy aspects related to identification, impact,

doi: 10.2166/wqrjc.2013.104

24

A. Arvai et al.

|

Chemicals of emerging concern removal in wastewater

Water Quality Research Journal of Canada

|

49.1

|

2014

and management of chemicals of emerging concern in the

examination of the performance of a subset of wastewater

Great Lakes. The Work Group found that discharges from

treatment plants in the Great Lakes basin. Further, literature

wastewater treatment plants are a significant source of con-

review and analysis of removal technologies was expected to

taminants to surface waters in the Great Lakes basin (IJC

provide insight into an array of potential wastewater treat-

), as concentrations of many chemicals of emerging

ment plant upgrades. The Work Group’s work plan

concern were generally highest in the discharge vicinity.

encompassed the following major activities:

Although pollution prevention is acknowledged to be the

•

develop an inventory of municipal wastewater treatment

•

conduct a detailed survey of operational data for selected

most effective and first barrier in protecting the environment and humans from exposure to chemicals of emerging concern (CEC), wastewater facilities also need to be proactive and prepared to address the presence of these chemicals in wastewater influents. The wastewater facilities therefore

•

become an important second barrier in protecting the

plants which discharge into the basin; facilities; and conduct a literature search and analysis of the effectiveness of various treatment technologies to remove chemicals of emerging concern.

environment from the discharge of CEC. As a result there is a need to assess and improve existing treatment technol-

The results of these activities served as a foundation for

ogies as large wastewater volumes are discharged without

an expert consultation. The resulting discussions and find-

receiving adequate treatment to remove chemicals.

ings

Another key finding was the necessity to assess the

from

the

consultation

formed

the

basis

of

recommendations provided to the Commission by the

impacts of chemicals of emerging concern on human and eco-

Work Group. These recommendations, listed at the end of

logical health in the basin. Human health and ecological

the paper, address reducing the discharge of CEC to the

health are interconnected. The health of ecological commu-

environment, with a particular focus on wastewater treat-

nities and populations may act as a sentinel for human

ment plants as a significant source of these contaminants.

health. Research to systematically determine the biological effects that may be occurring as a result of exposure to potentially toxic substances is limited. From an ecological perspective, analytical approaches for monitoring contaminants of possible concern can be supplemented with biological effects-based testing to understand contaminant effects at various levels of biological organization (sub-

METHODS Inventory of Canadian and US wastewater treatment facilities

cellular, cellular, organ system, individual, and population

In order to understand and comment on the effectiveness of

levels). Monitoring the effects of chemicals detected in the

municipal wastewater treatment systems which discharge

environment provides information that can bridge the gap

into the basin, an inventory of wastewater treatment systems

between chemical contamination and altered ecological status.

was developed for Canada (Ontario) and the United States.

This paper is based on the Chemicals of Emerging Concern Report, 2009–2011 Priority Cycle, prepared by the Chemicals of Emerging Concern Workgroup, International Joint Commission (IJC ). Under the 2009–2011 priority cycle, the Chemicals of Emerging Concern Work Group

The objectives of the project were to:

•

tabulate the total number of facilities which discharge

•

differentiate the facilities based on type of treatment

was charged to assess the performance of wastewater treatment plants in the Great Lakes basin with respect to the removal of chemicals of emerging concern. The Work Group was further instructed to provide the Commission

•

into the basin; (primary, lagoon, secondary/activated sludge, tertiary/ advanced); and analyze the distribution of facilities with respect to wastewater flow.

with a sampling of the information which might be derived

To develop the inventory, information was sought from

if a more extensive evaluation were undertaken, specifically

both the US and Canadian governments. The Work Group

25

A. Arvai et al.

|

Chemicals of emerging concern removal in wastewater

Water Quality Research Journal of Canada

|

49.1

|

2014

confirmed that the facilities indeed discharge into the geo-

representation, the facilities were selected to represent a

graphic/hydrogeological boundaries of the Great Lakes

range of the following.

basin, as defined by the Commission.

•

Ontario data were obtained from Environment Canada and the Ontario Ministry of the Environment, who had

Plant sizes – based on average flow, plants included: small facilities with 10

million gallons per day

(MGD) (38 million liters per day (MLD)); medium facili-

previously created a database of municipal plants and

ties 10–50 MGD (38–190 MLD); and large facilities

included pertinent information such as location, treatment

>50 MGD (190 MLD).

type, and average daily wastewater flow. With respect to the US data, the Environmental Protection Agency’s (US

•

EPA) Clean Watersheds Needs Survey (CWNS) and National Pollutant Discharge Elimination System permits were used to collect as much information as possible on municipal facilities. While the CWNS database identifies level of treatment, it is not specific to treatment technol-

•

Treatment technologies – included primary treatment, lagoon, fixed film, activated sludge, biological nutrient removal (BNR) and tertiary filtration.

Geographic location in the basin – plants were located on all five of the Great Lakes.

Basic data about the plants were collected including,

ogy, but rather defines treatment in terms of facility

location, flow, specific treatment processes and whether

performance. For US facilities, secondary treatment

chemicals were being used for phosphorus control. In

refers to the minimum level of treatment that must be

addition, data were collected on parameters that are used

achieved for discharges from all municipal wastewater

to control and evaluate plant operation, including solids

treatment facilities, and generally requires treatment that

and hydraulic residence times, wastewater flow, organic

will produce an effluent quality of 30 milligrams per liter

and nutrient concentrations, BOD5, and temperature.

of both 5-day biochemical oxygen demand (BOD5) and

Upon completion of the process of selecting plants and

total suspended solids (TSS). Further, secondary treatment

parameters of interest, facilities were contacted via tele-

must remove 85% of BOD5 and TSS from the influent

phone. Formal requests and templates for collecting the

wastewater, although lower percentage removals are auth-

necessary information were sent via e-mail and the US

orized in some circumstances. Advanced treatment is

Postal Service. For most parameters data were requested

more stringent than secondary, requiring the facility to

on a monthly basis (average, minimum, maximum) for a 3

achieve one or more of the following: BOD5 in the effluent

year period (2007–2009).

<20 mg/L, nitrogen removal, phosphorus removal, ammonia removal, metal removal, or specific synthetic organic

Review of the effectiveness of wastewater treatment

removal.

technologies

Additional work is currently underway to characterize the distribution of treatment operations (primary, lagoon,

In the absence of long-term measurements of the removal

activated sludge, tertiary) for the US facilities. The objective

efficiencies for CEC by municipal wastewater treatment

is to develop a harmonized database with information com-

plants within the Great Lakes basin, comparisons may be

parable to that of Ontario.

made of currently available data from published studies

Survey of operational data for selected facilities

various contaminants from wastewater treatment systems

that have reported influent and effluent concentrations of that practice similar technologies. A comprehensive literaIn order to provide a more detailed understanding of the

ture search was performed, constrained by publication

performance of municipal wastewater treatment systems, a

date (2000–2010) and treatment process – activated

survey of the operational data from a subset of wastewater

sludge, membrane bioreactors, sequencing batch reactors

treatment plants in the basin was undertaken. A total of

and lagoons. Chemicals selected for review were based

33 plants were solicited to participate in the survey, with

on the Review of Chemicals of Emerging Concern and

25 responding. Although not intended to be a statistical

Analysis of Environmental Exposures in the Great Lakes

26

A. Arvai et al.

|

Chemicals of emerging concern removal in wastewater

Water Quality Research Journal of Canada

|

49.1

|

2014

Basin, submitted to the International Joint Commission

and only eight facilities consist of primary treatment only.

(Klečka et al. ), which included more than 300 com-

With respect to total wastewater flow, greater than 95% of

pounds, and was the starting point for the search of

the municipal wastewater discharged into the basin receives

removal data regarding the performances of conventional

either secondary or tertiary (advanced) treatment. Approxi-

wastewater treatment processes. Additional criteria used

mately 1% of the plants do not practice secondary

to further screen documents included: (i) the document

treatment (primary and community septic tanks). Lagoons

should provide liquid removal (influent-to-effluent) of the

provide a low rate of secondary treatment and, while they

compounds of interest; and (ii) a description of the process

may constitute 37% of the facilities, they process only

or processes applied to treat the wastewater and the oper-

3.1% of the total flow.

ating conditions should be provided. Documents based

Of the 978 US facilities which discharge into the basin,

only on the occurrence of CECs were not further

311 and 563 achieve secondary and advanced treatment,

considered.

respectively (Table 2). Detailed information was not available in the survey database for 104 facilities, but likely the remaining facilities achieve at least secondary treat-

RESULTS AND DISCUSSION

ment performance, which is the minimum standard in the United States. Analysis of treatment performance with

Inventory of US and Canadian wastewater treatment

respect to total wastewater flow show that >95% of the

facilities

wastewater discharged into the basin meets the performance requirement for advanced treatment. Although 311

A total of 1,448 municipal wastewater treatment plants dis-

facilities meet the secondary treatment performance

charge 4.8 billion gallons (18 billion liters) per day of treated

requirement, these facilities receive only about 4% of the

effluent to the Great Lakes basin. Although there are differ-

total flow.

ences in how the two countries define such systems,

Many flows into the Great Lakes basin are not

collectively the combined group of secondary, advanced

accounted for in the above analysis, including wastewater

and tertiary plants treats 98% of the total wastewater flow

by-pass events, combined sewer overflows, industrial dis-

discharged to the Great Lakes basin.

charges, agricultural runoff, and the over one million

In Ontario, a total of 470 municipal wastewater treat-

private septic systems (IJC ) located in the basin. Also,

ment plants discharge into the basin (Table 1). Of these,

biosolids from wastewater treatment plants are often applied

212 and 68 are secondary/activated sludge and tertiary/

to agricultural land in the basin. This practice has the poten-

advanced treatment facilities, respectively. Note that smaller

tial to contaminate surface and ground water with organic,

communities are served by 175 lagoon treatment systems,

inorganic, and microbiological contaminants.

Table 1

|

Distribution of Ontario wastewater treatment plants in the Great Lakes basin

Facility type

Number of facilities

Percentage of total number of facilities

Total average daily flow MGD (MLD)

Percentage of total average daily flow

Primary

8

1.7%

25.4 (96.0)

1.7%

Community septics (all types)

7

1.5%

0.26 (1.0)

0.0%

Lagoons (all types)

175

37.2%

47 (178.0)

Secondary

212

45.2%

1331 (5038.1)

68

14.5%

120.7 (456.8)

470

100.0%

Tertiary Totals MLD ¼ million liters per day.

1549.4 (5769.1)

3.1% 87.3% 7.9% 100.0%

27

A. Arvai et al.

Table 2

|

|

Chemicals of emerging concern removal in wastewater

Water Quality Research Journal of Canada

|

49.1

|

2014

Distribution of US wastewater treatment plants in the Great Lakes basin

Facility type

Number of facilities

Percentage of total number of facilities

Total average daily flow MGD (MLD)

Percentage of total average daily flow

Secondary treatment

311

31.8%

135.9 (514.4)

4.2%

Advanced treatment

563

57.6%

3111.8 (11780)

95.8%

Unknowna

104

10.6%

n/a

n/a

Totals

978

100.0%

3247.7 (12294)

100.0%

MGD ¼ million gallons per day. a

Detailed information is not available.

Survey of operational data for selected facilities

biological nutrient (ammonia) removal are operated at higher solids residence times than those designed to

Of the 25 facilities that participated in the survey, 17 used an

remove organic (BOD) only and, as a result, BNR systems

activated sludge process, four used fixed film technologies,

remove various chemicals of emerging concern more effi-

two were lagoon based systems, and two were primary treat-

ciently than activated sludge systems operated at lower

ment plants. Wastewater flow rates for the facilities ranged

solids residence times. Consequently, the effectiveness of

from <10 to >50 MGD (<38 to >190 MLD).

ammonia removal provides a useful surrogate of the

The activated sludge facilities, which was the most

removal of biodegradable chemicals of emerging concern.

common secondary treatment process, used various modes

Three years of performance data from the 25 facilities

of operation, including plants designed for organic (BOD)

were reviewed. The activated sludge plants that reported

removal only and facilities designed to remove both organic

solids residence time data were divided into two groups:

and inorganic nutrients (e.g. ammonia). Chemical precipi-

residence times of <5 days and >5 days. Some facilities

tation of phosphorus was practiced at the majority of these

that operated at the lower residence times showed little to

plants. The second most common secondary treatment tech-

no change in ammonia from influent to effluent, indicating

nology used is based on biological fixed film technology;

low BNR (nitrification). These plants are most likely to

three plants used biological aerated filtration and one

show poorer removals for chemicals of emerging concern.

deployed the trickling filter/solids contact process. Fixed

In contrast, many of the activated sludge plants that

film processes are more advanced than the historically

were operated with higher solids residence times generally

more common trickling filter process. Of the remaining

showed effluent ammonia concentrations of <1.0 mg/L,

plants, two were lagoon-based and two employed only pri-

indicating almost complete nitrification and a high level of

mary treatment or activated carbon and chemical oxidation.

performance on a year-round basis. Also common to these

One critical operating parameter that has been corre-

plants was a greater hydraulic residence time, indicating

lated with removal efficiencies for many biodegradable

that the plants were conservatively designed with a year-

chemicals of emerging concern is the solids (biomass) resi-

round capability of nutrient removal. Given the correlation

dence time (Clara et al. a, b). Some chemicals are

previously discussed, removal efficiencies for biodegradable

difficult to biodegrade, and the microorganisms that have

chemicals of emerging concern by these facilities are likely

adapted to degrade them grow slowly. Because the solids

to be sustained throughout the year.

residence time is related to the microbial growth rate, it is

One of the plants evaluated operated their activated

a useful surrogate of the ability of the activated sludge pro-

sludge process in a membrane bioreactor mode. It produced

cess to retain the slower growing organisms and therefore

a very high quality effluent with <0.2 mg/L ammonia as well

to degrade the biodegradable chemicals of emerging con-

as very low effluent BOD5 and TSS concentrations. The high

cern. Further, activated sludge systems designed for

degree of nitrification is attributed to the higher level of

28

A. Arvai et al.

|

Chemicals of emerging concern removal in wastewater

Water Quality Research Journal of Canada

|

49.1

|

2014

control over solids residence times that can be achieved

concentrations, removal efficiencies and treatment pro-

with a membrane in place of a secondary clarifier. This

cesses. This was then analyzed from a weight-of-evidence

plant, too, will likely achieve a high level of removal

perspective. Much of the available information described

during summer and winter.

the removal of pharmaceuticals, personal care products, sur-

However, it is not necessary to invoke membrane bio-

factants, and hormones in the activated sludge process.

reactor technology to achieve high quality effluent. For

Removal efficiencies may be the result of biodegradation

example, one of the lagoon systems evaluated achieved the

or physical processes such as adsorption. Environmental

lowest effluent ammonia concentrations of all 25 plants

fate information for many chemicals of emerging concern

(<0.1 mg/L). The effluent BOD5 and suspended solids con-

is lacking, particularly with respect to information on biode-

centrations were similarly very low (2 and 6 mg/L,

gradability. In addition, the degradation products of these

respectively). This remarkable performance illustrates that

compounds are generally unknown and thus removal effi-

the plant has been designed conservatively and is well oper-

ciencies are typically based on removal of the parent

ated. However, lagoon technology cannot be deployed in

compound. This is a major impediment to the development

large population centers due to the large area required, but

of simulation models for predicting the removal efficiency

this plant does illustrate what can be achieved in small to

(extent and reliability) of wastewater treatment systems.

medium-sized communities.

Due to a lack of sufficient data to permit meaningful

Effective removal of the wide variety of chemicals of

statistical analysis, the analysis was limited to the treatment

emerging concern is dependent on both the nature of the

of 42 substances by activated sludge treatment facilities.

substance and the design and operation of the wastewater

Table 3 summarizes the results of analysis of the 42 sub-

treatment plant. Examination of the performance data col-

stances. They were grouped according to the likelihood

lected suggests that some plants are operated very well

that they would achieve a given removal efficiency during

while others are not. Facilities designed for primary treat-

wastewater treatment as a function of the frequency of

ment are unlikely to adequately reduce the concentrations

their observation in wastewaters.

of biodegradable chemicals of emerging concern. Some

As illustrated in Table 3, some substances such as DEET

very large cities are serviced by wastewater treatment

and testosterone were infrequently detected but demon-

plants designed for organic (BOD) removal only.

strated a high likelihood of removal. Other substances

In general, many of the facilities were operated at high

such as carbamazepine and diclofenac were frequently

solids residence times and are effective in both organic

detected but poorly removed. Only acetaminophen, caffeine

(BOD) and nutrient (ammonia) removal. Such facilities

and estriol occurred frequently and had a high probability of

are likely to show effective reductions of biodegradable

at least 75% removal efficiency. More than half of the 42

chemicals of emerging concern. In addition, conventional

substances fell into the middle, that is, they occurred at a

parameters (BOD, ammonia) are useful surrogates to

medium-to-high frequency and had a medium-to-high prob-

assess removal efficiencies for biodegradable chemicals of

ability of attaining 75% or better removal efficiency.

emerging concern. However, different indicators are

Results of the present study were compared to results of

required for substances that are poorly or non-biodegradable

recent investigations published by government agencies or

or that have a propensity to adsorb to biomass.

the Water Environment Research Foundation. For example,

Since this survey was not a statistical representation of

Stephenson & Oppenheimer () reported the fate of

the plants in the basin, extrapolation to other facilities in

pharmaceuticals and personal care products in municipal

the basin cannot be made.

wastewater treatment processes. The fate of 20 compounds was measured at eight municipal plants. The substances

Effectiveness of wastewater treatment technologies

were selected based on the frequency of occurrence in municipal wastewaters. All plants used a variation of the

From the literature review a database was created, which

activated sludge process, and six operated BNR. The results

included information on CEC influent and effluent

of this study are summarized in Table 4.

29

A. Arvai et al.

Table 3

|

|

Chemicals of emerging concern removal in wastewater

Water Quality Research Journal of Canada

|

49.1

|

2014

Summary of confidence level vs. removal efficiency for 42 chemicals of emerging concern by activated sludge systems

Confidence level (n ¼ # of records)

Low removal efficiency (<25% probability of 75%þ removal)

Medium removal efficiency (25–75% probability of 75%þ removal)

High removal efficiency (>75% probability of 75%þ removal)

Low (n < 9)

Atrazine Pyrene

Benzophenone Indomethacin Sulfamerazine

Musk ketone Di (2-ethylhexyl) adipate (DEHA) N,N-diethyl-toluamide (DEET) Testosterone

Medium (9 n 15)

Gemfibrozil Perfluorooctanoic acid (PFOA) Perfluorooctyl sulfonate (PFOS)

Di (2-ethylhexyl) phthalate (DEHP) Norfloxacin Ranitidine Roxithromycin Tetracycline

High (n > 15)

Carbamazepine Ciprofloxacin Clofibric acid Diclofenac Erythromycin Trimethoprim

Bezafibrate Bisphenol A Estrone (E1) 17α-Ethynyl estradiol (EE2) 17β-Estradiol (E2) Galaxolide Ibuprofen Ketoprofen Naproxen Nonylphenol Nonylphenol monoethoxylate (NP1EO) Nonylphenol diethoxylate (NP2EO) Octylphenol Sulfamethoxazole Tonalide Triclosan

Table 4

|

Acetaminophen Caffeine Estriol (E3)

Summary of removal efficiencies of pharmaceuticals and personal care products by activated sludge systems (Stephenson & Oppenheimer 2007)

Frequency of occurrence in samples

Poor removal (<25%)

Moderate removal (25–75%)

Good removal (>75%)

Infrequent (<25%)

Trichloroethyl phosphate (TCEP) Triphenyl phosphate

Octylphenol

Methyl-3-phenylproprionate

Intermediate (25–75%)

Butylated hydroxyanisole (BHA) N,N-diethyl-toluamide (DEET) Musk ketone

Ethyl-3-phenylproprionate

Frequent (>75%)

Galaxolide

Benzophenone Triclosan

Comparison of the results presented in Tables 3 and 4 indicates that removal efficiencies for many of the chemicals

Benzyl salicylate Butylbenzyl phthalate Caffeine Chloroxylenol Methylparaben Ibuprofen Octylmethoxycinnamate Oxybenzone 3-Phenylproprionate

unclear but may reflect different operating conditions among facilities.

common to both studies were similar while others were dia-

Drewes et al. () reported on the removal of eight

metrically opposed. The reason for this discrepancy is

endocrine disruptor compounds by seven municipal

30

A. Arvai et al.

|

Chemicals of emerging concern removal in wastewater

Water Quality Research Journal of Canada

treatment plants. All plants used a variant of the activated

|

49.1

|

2014

CONCLUSIONS AND RECOMMENDATIONS

sludge process and most operated BNR. Removal efficiencies for the steroid hormones ranged from 48 to 98%.

Much has been learned about the presence of chemicals of

Concentrations of bisphenol A, nonylphenol, and octylphe-

emerging concern in wastewaters during the past few years.

nol were reduced on average by 93, 61, and 80%,

However, the inability to answer questions is not surprising

respectively.

given the number of and range in molecular complexity of

Biodegradation

of

the

compounds

was

reported as the dominant removal mechanism.

the various compounds, combined with the spectrum of tech-

In a follow-up study, Drewes et al. () examined the contributions of household chemicals to sewage and their

nologies employed by municipal wastewater treatment plants and the range of operating conditions.

removal in municipal wastewater treatment systems. The

The results of the activities discussed in this report were

fate of 25 substances was measured at seven municipal

provided to attendees at an expert consultation held in

plants. Most of the facilities used activated sludge but

Romulus, Michigan on January 25–26, 2011. Based on the

one employed biological aerated filter technology. All of

results of the expert consultation, the Work Group made

the plants operated BNR. Removal efficiencies were vari-

the following recommendations to the Commission with

able for the compounds but generally exceeded 80%.

regard to reducing CEC discharged to the environment, par-

Notably, 2-phenoxyethanol, hydrocortisone, camphor,

ticularly from wastewater facilities.

propylparaben

and

isobutylparaben

achieved

98%

removal efficiency. Removal efficiencies for butylated hydroxytoluene,

butylated

hydroxyanisole,

1. Investigate the contribution of combined sewer/storm

DEET,

sewer overflows, by-passes, and industrial discharges to

3-indolebutyric acid and triclocarban ranged between 60

loadings of chemicals of emerging concern to the

and 70%.

Great Lakes in the next priority cycle. Combined

Taking into consideration that municipal wastewater treatment systems were not designed to remove chemicals

sewer/storm sewer overflows are potentially substantial contributors.

of emerging concern, results of the present study (IJC ),

2. Encourage primary treatment facilities to upgrade to sec-

as well as government and Water Environment Research

ondary treatment, with consideration also to advanced

Foundation reports, suggest that well operated, convention-

treatment technologies, and secondary plants to consider

ally designed plants are capable of achieving effective

adding BNR processes and optimizing processes to

reductions of a variety of substances.

improve removal of biodegradable chemicals of emerging

The weight of evidence suggests that at least half of the 42 substances examined in the present study are likely to be

concern. The utility of advanced oxidation as an advanced treatment process should be further explored.

removed in municipal wastewater treatment plants. An

3. Evaluate improvements to wastewater treatment systems

analysis as described above is limited in terms of reaching

in the context of sustainability and the triple bottom line.

definitive conclusions about the extent to which Great

Benefits of updating treatment technology needs to be

Lakes wastewater treatment facilities, as currently operated,

balanced with capital and operational costs, as well as

are able to remove various contaminants and to what extent

with environmental impacts such as increased power

and with what reliability.

requirements to operate equipment and emissions of

None of the analyses examined the impact of operating

greenhouse gases: CO2, NOx and SOx.

conditions on plant performance. Insufficient granularity

4. Develop a list of indicator compounds for use by facilities

and reproducibility in the various datasets preclude discern-

to assess the effectiveness of their treatment process to

ing the impact of operating conditions such as temperature,

remove chemicals of emerging concern. A list of criteria

loading rate, solids and hydraulic residence times. Likely

may be needed to define ‘indicator compounds’ for use

much of the variability in the reported data can be explained

as surrogates (currently being addressed by the Water

by differences in operating conditions at the various

Environment Research Foundation), that is, in addition

facilities.

to the conventional parameters of ammonia and BOD

31

A. Arvai et al.

|

Chemicals of emerging concern removal in wastewater

removal. This would require the development, standardization and validation of methodologies to analyze these wastewater contaminants. Further, in order for facilities to assess their effectiveness in removing CECs, there is a need to have laboratories available to perform such analysis. Therefore, the capacity of contract analytical laboratories must be increased.

5. Examine the role of biosolids from wastewater treatment facilities as a source of chemicals of emerging concern to the environment. Biosolids are often used as amendments to agricultural soils. Government agencies are currently addressing this issue.

6. Prepare a list of chemicals of emerging concern that are difficult to treat using current technologies and determine which require the development of a risk assessment/risk management strategy. Further, risk-based analysis of compounds needs to be conducted, especially in regard to human health.

7. Recommend biological effects monitoring of wastewater effluents. For example, bioassays of wastewater effluents could be used in combination with chemical analysis.

Compounds must be sorted with consideration to those that are highly consequential in small concentrations, that is, estrogens versus those that are not, that is, caffeine.

8. Increase public education regarding the use and disposal of pharmaceuticals and personal care products and their entrance into the environment and the wastewater treatment process. Public education also includes manufacturers in terms of promoting green chemistry.

ACKNOWLEDGEMENTS The study was supported by the International Joint Commission. The authors acknowledge the valuable contribution of the Chemicals of Emerging Concern Work Group for the preparation of the 2009–2011 Priority Cycle Report on Chemicals of Emerging Concern.

Water Quality Research Journal of Canada

|

49.1

|

2014

REFERENCES Clara, M., Strenn, B. & Gans, O. a Removal of selected pharmaceuticals, fragrances and endocrine disrupting compounds in a membrane bioreactor and conventional wastewater treatment plants. Water Res. 39 (19), 4797– 4807.

Clara, M., Kreuzinger, N., Strenn, B., Gans, O. & Kroiss, H. b The solids retention time – a suitable design parameter to evaluate the capacity of wastewater treatment plants to remove micropollutants. Water Res. 39 (1), 97–106.

Daughton, C. G.  Pharmaceuticals in the environment: overarching issues and overview. In: Pharmaceuticals and Personal Care Products in the Environment: Scientific and Regulatory Issues (C. G. Daughton & T. Jones-Lepp, eds).

Symposium Series 791, American Chemical Society, Washington, DC, pp. 2–38.

Drewes, J. E., Hemming, J. D. C., Schauer, J. J. & Sonzogni, W. C.  Removal of Endocrine Disrupting Compounds in Water Reclamation Processes. WERF Report 01-HHE-20T. Water Environment Research Foundation, Alexandria, VA. Drewes, J. E., Dickenson, E. & Snyder, S.  Contributions of Household Chemicals to Sewage and their Relevance to Municipal Wastewater Systems and the Environment. WERF Report 03-CTS-21UR. Water Environment Research Foundation, Alexandria, VA. IJC  Great Lakes Water Quality Agreement Priorities 2007–09 Series. Work Group Report on Great Lakes Chemicals of Emerging Concern, 2009. IJC, Special Publication 2009-01, Windsor, Ontario, Canada. Available from: www.ijc.org/en/ priorities/2009/reports/2009-chemicals.pdf.

IJC  Groundwater in the Great Lakes Basin: A Report of the Great Lakes Science Advisory Board to the International Joint Commission, 2010. IJC Publication, Windsor, Ontario, Canada. Available from: www.ijc.org/files/publications/E43. pdf.

IJC  Great Lakes Water Quality Agreement 2009–2011 Priority Cycle Report, International Joint Commission International Joint Commission, 2011. Available from: http://meeting.ijc. org/workgroups/cec.

Klečka, G., Persoon, C. & Currie, R.  Chemicals of emerging concern in the Great Lakes Basin: an analysis of environmental exposures. Rev. Environ. Contam. Toxicol.

207, 1–93.

Stephenson, R. & Oppenheimer, J.  Fate of Pharmaceuticals and Personal Care Products through Municipal Wastewater Treatment Processes. WERF Report 03-CTS-22UR. Water Environment Research Foundation, Alexandria, VA.

First received 14 January 2013; accepted in revised form 22 September 2013. Available online 7 November 2013

32

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Urban snow deposits versus snow cooling plants in northern Sweden: a quantitative analysis of snow melt pollutant releases Angela Lundberg, James Feiccabrino, Camilla Westerlund and Nadhir Al-Ansari

ABSTRACT High-velocity runoff from snow deposit transports suspended grain-attached contaminants while underground snow storages trapped these contaminants within the storage. The aim here is to quantify pollutant masses from an urban snow deposit and to investigate the conditions when pollutant control was increased by turning a snow deposit into a snow cooling plant with permeable underground snow storage. Pollutant masses in an urban snow deposit in northern Sweden were: Cu ¼ 67, Pb ¼ 17, Zn ¼ 160, P ¼ 170, SS ¼ 620,000, Cl ¼ 1,200, N ¼ 380 kg. A theoretical analysis showed that the fraction of surface runoff from a surface deposit largely depends on the hydraulic conductivity (K, m s 1) of the soil. For a melt rate of 30 mm, day 1, surface runoff would be about 97% for a soil with K ¼ 10 8, while nonexistent for K > 10 6. Similar soil conductivities are needed to ensure that all snow melt could be transported as groundwater from an underground storage. The largest pollution-control advantage with underground snow storage compared to a surface deposit would thus be that piping and filters for operation of the plant could be used to filter surface snow melt runoff before rejection.

Key words

| alternative energy, snow cooling plant, snow deposit, snow pollutants, urban snow

Angela Lundberg (corresponding author) James Feiccabrino Division of Geosciences and Environmental Engineering, Department of Civil, Environmental and Natural Resources Engineering, Luleå University of Technology, SE-971 87 Luleå, Sweden E-mail: Angela.Lundberg@ltu.se Camilla Westerlund Division of Architecture and Water, Department of Civil, Environmental and Natural Resources Engineering, Luleå University of Technology, SE-971 87 Luleå, Sweden Nadhir Al-Ansari Division of Mining and Geotechnical Engineering, Department of Civil, Environmental and Natural Resources Engineering, Luleå University of Technology, SE-971 87 Luleå, Sweden

INTRODUCTION Most northern towns in Sweden remove snow from roads to

depend on the flow velocity (Reinosdotter ). The impor-

increase traffic safety, and this polluted snow is usually

tance of protecting surface waters and groundwater aquifers

placed in a surface snow deposit. The major contaminants

from pollution has been highlighted by the European Union

in urban snow, metals, nutrients, salt, and hydrocarbons

(Directive //EC; Directive //EC). There has

(Viklander ) are hereafter collectively referred to as

therefore been a greater awareness of urban snow pollu-

snow pollutants. A large amount of the pollutants in the

tants. Increasing demands for cooling capacity during

snow is attached to particles that remain on the ground

summer have occurred in many northern countries, with

under the deposit after snow melt (Oberts ; Viklander

rising energy prices along with a public call for cleaner

; Oberts et al. ; Westerlund et al. ). However,

energy. Consequently, it is natural to consider the possibility

when the snow melt (and rain) rates are greater than the

of using deposited snow to meet these purposes, and a plant

infiltration capacity (IC) of the soil beneath the deposit, sur-

designed for cooling of a hospital in central Sweden has

face water from deposits may have the capacity to transport

proved to be a success (Larsson ). The snow melt there

dissolved and particle-bound pollutants to surface water

is collected and transported through pipes to a heat exchan-

recipients, such as lakes or streams (Viklander a). The

ger (Figure 1) chilling water circulating through the hospital

amount and size of suspended solids (SS) in surface runoff

building (Skogsberg & Nordell ).

doi: 10.2166/wqrjc.2013.042

33

A. Lundberg et al.

Figure 1

|

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

|

49.1

|

2014

The snow cooling plant with the liner-equipped snow storage with approximate water temperatures shown (modified from Skogsberg & Nordell (2001)).

The snow storage for this plant is located above the

Sweden. To assess the resultant pollution transport from

ground, covered with wood chips to reduce the rate of melt-

snow deposits of different designs is a huge task, requiring

ing and equipped with an impermeable liner to prevent

an enormous amount of data. The data include, accumu-

polluted snow melt from spreading (Skogsberg & Lundberg

lated pollution masses in the snow, melt rates, soil

). Snow cooling plants with other snow storage designs

parameters, topography of the location, deposit designs, geo-

and optional snow melt treatment can also be constructed.

chemistry of the soil and natural groundwater, original

The storage can be located under the surface of the

groundwater levels, and a combined sediment transport

ground and might be permeable or impermeable. The

and groundwater model adjusted for this type of problem.

snow melt might be filtered and/or treated before being

There is little information available for this study so an

recirculated through the snow storage or rejected. The

engineering approach with crude assumptions was applied.

investment costs for this plant, with the liner-equipped

The order of magnitude of the effects will be outlined only

storage, was high. A snow cooling plant having permeable

but, nevertheless, will provide guidance on how different

underground snow storage will work the same way apart

parameters influence the pollution transport from the two

from the fact that water can flow in or out of the storage

types of snow deposits. The detailed goals are as follows:

resulting in less control over dissolved pollutants. Turning an existing urban snow deposit into a snow cooling plant for a few buildings was a possible choice in Luleå, northern Sweden. Due to economic limitations a permeable under-

•

Estimate the pollutant load and dissolved matter for an

•

Analyze the factors that influence the surface flow frac-

existing surface snow deposit.

ground snow storage was designed, and an environmental

tion of snow melt from a surface deposit. For soil

impact assessment plan was requested by the city environ-

particles of different sizes, velocities required to set

mental community office, resulting in this study.

them in motion or deposition will be analyzed.

AIM AND SCOPE

•

For a planned snow cooling plant with permeable underground snow storage: (1) the order of magnitude of hydraulic conductivities, and gradients needed to ensure that there is no need for surface rejection of polluted

Comparison of pollutant pathways and possible environ-

water at the soil surface are to be determined; and (2)

mental effects for a surface urban snow deposit with a

the link of horizontal groundwater transport velocity to

snow cooling plant equipped with permeable underground

the soil porosity, hydraulic conductivity, and gradient

snow storage is dealt with for a typical city in northern

will be illustrated.

34

•

A. Lundberg et al.

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

|

49.1

|

2014

The conditions required for turning a surface snow

deposit located above the water table will percolate until

deposit into a snow cooling plant with permeable under-

the soil under the deposit approaches saturation, at which

ground snow storage that would reduce the total

point surface runoff will increase (Oberts et al. ). How-

pollutant load will be discussed, as well as the uncertain-

ever, for high melt rates and low infiltration capacities, the

ties in the estimates and the future for snow cooling

main snow melt pollutant pathway will be as surface

plants.

runoff where the mass transport of suspended material is a function of flow velocity, availability, and size distribution

Pollutant pathways

of the material (Westerlund & Viklander ). Low flow

Snow melt from a surface snow deposit might become sur-

velocities can carry larger particles and, consequently,

face runoff and/or infiltrate and then percolate to the

larger amounts of suspended solids. Rain on snow events

groundwater (Figure 2).

Infiltration rates for snow deposits vary from 0 to 100%

velocities carry only fine suspended solids, while high flow

will increase the surface water flow, allowing more suspended solids to leave the deposit (Oberts et al. ;

depending on soil properties (saturated hydraulic conduc-

Westerlund & Viklander ). A snow cooling plant with

tivity, initial water content, frost, etc.). Snow melt from a

permeable underground snow storage, having banks higher than the surrounding ground, will obstruct the surface water flow and thus hinder transport of suspended solids (Figure 3) but allow water transport through it. The water levels inside and outside the storage will control the flow direction through its sides (Figure 3(a)–(d)). The figure illustrates stationary conditions, but the flow pattern will change due to percolation raising the groundwater levels. If the storage is located in a flat area with a shallow water table, polluted water would be unable to leave the storage but

Figure 2

Figure 3

|

|

Pollutant pathways for a surface snow deposit located above a sloping

would increase the amount of polluted water requiring

groundwater table. The dashed lines illustrate how the groundwater table will rise as a result of enhanced infiltration and possibly also reach the soil surface.

water extraction from the storage, since fresh groundwater

White and black arrows denote clean and polluted water.

would be mixing with the polluted snow melt (Figure 3(a)).

Examples of different groundwater configurations in conjunction with a permeable underground snow storage. (a) Groundwater table above the storage water level; groundwater flowing into the pond. (b) Groundwater table below the storage water level but above the base of the storage. (c) Groundwater table below the bottom of the storage. (d) Sloping groundwater table intersecting the storage with some groundwater flowing through the sides. White arrows indicate clean water and black arrows polluted water.

35

A. Lundberg et al.

|

Pollutant pathways from snow deposits

METHODS

Water Quality Research Journal of Canada

|

49.1

|

2014

water table was well below the base of the storage; however, the exact depth could not be determined.

Site with existing snow deposit and proposed cooling plant

Contaminant masses: dissolved and particulate

The Luleå snow deposit is located on glacial till, typical for large areas in northern Sweden, having low porosity and hydraulic conductivity. The total porosity of glacial till is at most 17%, the average effective (water carrying) porosity is usually below 5%, and the hydraulic conductivity (K) is poor, between 10 7 and 10 9 m s 1 (Salonen et al. () cited by Mälkki ()). The distance from the present

The total pollutant mass (kg) of each pollutant i for the snow deposit was estimated by multiplying the average concentration of each pollutant CiAverage (kg, m 3) in the snow by the volume of the deposit VSD (m3): Pollutant massi ¼ CiAverage VSD

(1)

snow deposit to the nearest surface water is about 100 m and the slope of the soil surface to the water course is

Average urban snow pollutant concentrations were

about 1/100. The deposit storage volume is around

thus needed and the best source for these would be

200,000 m3 with a snow density of about 500 kg m 3 and

samples from the actual snow deposit. The snow in the

the conditions would be similar for the planned snow sto-

deposit was very packed, contained grit and sand and

rage close to the existing deposit. The area of the actual

was several meters thick so such samples were very diffi-

snow deposit varies with the season and is here assumed

cult to obtain, and did not appear to be available from

to have about the same volume and surface area

any study. However, pollutant concentrations (and pH) in

(≈40,000 m ) as the planned permeable underground snow

snow melt from road snow banks with both high and low

storage in order to make the conditions more comparable.

traffic loads, in Luleå, were accessible and could be used

The snow storage was designed to allow most, if not all, of

(Table 1).

2

the snow to melt in about 150 days (before the next winter

The concentrations of the different pollutants in the

season). It would be about 10 m deep (from the bottom to

snow deposit were estimated assuming that the city had

the top of banks) with a bottom around 6 m beneath the

50% high traffic and 50% low traffic streets so:

ground level, and banks built 4 m above the ground level and the snow reaching, on average, 5 m above the banks (Figure 4).

CiAverage ¼

Cihigh þ Cilow ρsnow 2:ρwater

(2)

When in operation, the water level should be maintained at 4 m above the bottom of the snow storage with a 2

Accordingly, the contents of SS, N, Cl, and suspended

water level area of ≈8,000 m . Ground-penetrating radar

and dissolved metals were determined. From the total

was used to determine the distance to the groundwater

mass of metals [copper (Cu), lead (Pb), zinc (Zn), phos-

table at the planned site, and it was concluded that the

phorus (P)], and their dissolved mass, the fraction of

Figure 4

|

Cross-section of proposed snow cooling plant with permeable underground snow storage: base width 40 m and side slopes of 1:4. The base area is 40 by 80 m (3,200 m2), the pond water surface area 72 by 112 m (8,064 m2), and the snow surface area 120 by 160 m (19,200 m2).

36

A. Lundberg et al.

Table 1

|

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

Measured pH, suspended solids (SS), nitrate (Ni), chloride (Cl), and metal and phosphorus concentrations (total and dissolved) for snow melt from snow samples (with standard deviations) taken from roads with high and low traffic load in Luleå from the central areas (Sources: Viklander 1997b; Reinosdotter & Viklander 2006)

Substance

Unit

pH

–

Low traffic load

1

High traffic load

7.3 ± 0.38

8.3 ± 0.36

|

49.1

|

2014

water equivalents the surface runoff QSURF (mm, day 1) can be estimated by: QSURF ¼ Melt IC; for Melt > IC and QSURF ¼ 0; for Melt IC

(4)

SS

mg L

4,471 ± 3,144

7,889 ± 6,744

N tot

mg L 1

3.8 ± 1.1

3.8 ± 1.1

mg L 1

setting IC equal to the saturated hydraulic conductivity K

Cl

4.4 ± 2.8

20 ± 15

of the soil. Expected fractions of average surface runoff for

Cu tot

μg L 1

310 ± 245

1,022 ± 1,089

the surface deposits were then determined for different

Cu dis

μg L 1

4.5 ± 2.45

7.0 ± 5.3

Pb tot

μg L 1

119 ± 87

217 ± 232

Incipient sediment motion requires velocities higher

Pb dis

μg L 1

0.3 ± 0.6

0.08 ± 0.03

than those to maintain the motion of the same particles.

Zn tot

μg L 1

931 ± 659

2,233 ± 2,308

For this purpose, Hjulström’s () diagram was used to

Zn dis

μg L

1

7.1 ± 6.7

1.5 ± 1.0

determine the order of magnitude for velocities required to

P tot

mg L 1

P dis

μg L

1

1.265 ± 0.6

2.039 ± 1.22

15 ± 17

6 ± 1.4

A rough estimate of IC of the soils can be achieved by

months.

erode (lift), and deposit soil particles of different sizes.

Conservative pollutants, i.e., pollutants that do not interact with the soil, travel with the same average velocity

particulate and dissolved contaminants was estimated.

vpollutant (m, day 1) as the groundwater:

Finally, the average pH-value was determined based on

i neff

the fractions of high and low traffic roads.

v pollutant ¼ K

Pollutant pathways

where i ( ) is the horizontal hydraulic gradient dh/dx and

Because suspended matter is transported by surface runoff,

(5)

neff ( ) is the effective porosity of the soil. The effective porosity of till soils was below 5% and likely variations in this

melt (and rain) rates and IC for surface deposits were

porosity have a minor influence on travel times compared

required to estimate the surface runoff. Since this melt is lar-

to the possible variations in K in the area, that range

gely governed by air temperature, it was determined using a degree-day method (also known as temperature index method) (Lundberg & Beyerel ). The Melt (mm day 1) is determined from the daily average air temperature T ( C) and a degree-day factor, DDF [mm, ( C day) 1] accordW

W

ing to: Melt ¼ DDF:T

between 10 6 and 10 9 m s 1. Soil surface slopes (≈i) varied from 0.001 to 0.1. In order to illustrate the effects of relevant parameters on travel times for conservative matters these times (t) were estimated for different combinations of K and i keeping neff ¼ 2.5% (assuming only advective transport and neglecting dispersivity) using the distance (dist) to the nearest water course ¼ 100 m:

(3) t ¼ dist=v pollutant

(6)

Even if this method was developed for natural snow, it has been used for snow deposits (Sundin ). Expected

The biggest advantage with underground snow storage

average melt was here calculated using DDF ¼ 4.0 [mm,

was that it cuts surface runoff; however, this only occurs if

( C day) 1] for May to July and DDF ¼ 2.0 for April and

the groundwater flux is large enough to transport all snow

August. Snow melts in excess of the IC can be assumed to

melt. Percolating snow melts from the storage will form a

become surface runoff when surface depression storage as

dome-shaped enlargement of the groundwater table under

for this application can be neglected. When expressed in

the storage and modify the natural groundwater flow

W

37

A. Lundberg et al.

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

pattern. A very rough indication of the size of this local influence on the groundwater table could be calculated by assuming that all infiltrated snow melt would remain

Table 2

|

Metal

|

49.1

|

2014

Estimated total and dissolved metal masses for the Luleå snow deposit

Total mass (kg)

67

Dissolved mass (g)

Dissolved fraction (%)

580

0.86

directly under the snow storage. If we divide the storage

Copper

mass (108 kg) with the projected storage area at the planned

Lead

17

19

0.11

operation water table (72 m × 112 m) and with the number

Zinc

160

430

0.27

of days the snow is expected to melt in (180 days) and the density of water (1,000 kg m 3) we get the daily infiltration rate of 0.08 m day 1. With an effective porosity of 2.5%

storage of 100,000 metric tons. Using a 50% high and 50% low traffic street ratio for snow contribution to the deposit

this would correspond to a daily rise in the water table by

results in a slightly basic pH of 7.8 (±0.37). The total mass

3 m. Even if this assumption is unrealistic, since the water

load of pollutants in the deposit indicates a high concen-

would spread fairly quickly over a much larger area, it still

tration of suspended solids (620 tons), 1,200 kg of chloride

gives an indication of the possible influence on the ground-

and 380 kg of nitrate. For phosphorus, the total mass was

water levels. An increased hydraulic gradient (i) in the

170 kg, out of which 0.64% was suspended. Of the metals,

original flow direction will be formed and a reversed gradient in the opposite direction will result due to this. Water will thus be flowing in all directions from the dome, so it is reasonable to assume that the groundwater flow cross-sectional area at the storage border will be the aquifer depth d (m) multiplied by perimeter (peri) of the water table in the storage when operating (368 m). For simplicity, if we assume that the natural groundwater flow at the site is negligible compared to the snow melt flow, we can estimate the required value of the product of the transmissivity (T ¼ K · d) and the hydraulic gradient i needed to transport all snow melt without having to reject any polluted snow melt at the soil surface. The total snow melt volume V (m3) is determined by:

zinc had the largest mass and copper the highest dissolved fraction (see Table 2).

Pollutant pathways Average daily melt rates calculated for the snow deposit using the DDF ¼ 4 for the summer months ranged from 52 to 64 (mm, day 1) (see Table 3) and melt rates for May and September using DDF ¼ 2 became 13 and 18 (mm, day 1), respectively.

The estimated fractions of surface runoff as a function of daily melt rates and infiltration capacities (assumed to be equal to K) are shown in Figure 5 which illustrates that for soils with K < 10 8 m s 1 almost all melt could be expected to become surface runoff for melt rates larger

V ¼ T i peri t

(7)

where t is the time (1 year in s). T· i can then be calculated

than just a few mm day 1. For soils with K ¼ 5 × 10 8 about 50% would become surface runoff for melt rates around 8 mm day 1 while there would not be any surface

by: T i¼

V peri t

Table 3

|

Monthly mean temperatures for Luleå with expected average daily monthly melt using the degree-day factors, DDF given below

(8)

May

Average daily air temperatures ( C) for Luleåa

6.7

Jun

Jul

Aug

13.1

16.1

14.2

Sep

8.9

W

RESULTS

DDF (mm, C 1 day 1) W

Contaminant masses: dissolved and particulate

Average melt (mm day 1)

The snow deposit mass, based on a snow density of 500 kg

Total melt (m water equivalent)

m 3 and a storage volume of 200,000 m3 resulted in a snow

a

2

4

4

4

2

13

52

64

57

18

0.41

1.57

2.00

1.70

Mean of average (30-year) mean high and mean low temperature values.

0.55

38

A. Lundberg et al.

|

Figure 5

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

|

49.1

|

2014

Estimated fractions of surface runoff versus melt rates for infiltrate capacities (K) equal to 1 × 10 9, 5 × 10 9, 1 × 10 8, 5 × 10 8, 1 × 10 7 and 5 × 10 7 (m, s 1), respectively, shown from the top.

runoff at all for soils with K ¼ 10 6 m s 1 even for melt (and

the storage while the hydraulic gradient will decrease. A gra-

1

dient of 0.5 just at the border of the storage seems realistic

rain) rates above 80 mm day .

To set sediments in motion by surface water, clay 1

due to the percolating snow melt’s rapid rise in the water

requires water velocities larger than about 1.50 (m s ),

table. Soil depths in the region seldom reach above 40 m

while silt, fine sands, and sand require surface water vel-

(soil depths at a planned drinking water source for the city

1

varied, e.g., from 0 to 40 m) so let us assume a hydraulic gra-

ocities around 0.20 (m s ) (Table 4).

For permeable underground snow storage, the product

dient of 0.5 and a soil depth of 20 m. With these values

of T and i (Equation (8)) restricts the annual amount of

inserted, then K has to be larger than about 10 6 m s 1 to

snow melt that could be transported (and filtered) by the

ensure that all snow melt can be transported through the

3

soil. With numerical values (V ¼ 100,000 m , t ¼ 3.2 × 107 s, and peri ¼ 368 m) inserted into the equation, then the product T· i has to be >9 × 10

6

1

soil.

The expected horizontal (conservative) groundwater

m s . The cross-sec-

transport time varies strongly with the hydraulic conduc-

tional area for the flow will increase with distance from

tivity and the hydraulic gradient (see Table 4). These parameters might, for a till, vary within large ranges (K

Table 4

|

from ≈10 6 to ≈10 9 m s 1 and i from ≈0.001 to ≈0.1), Water velocities required to set soil particles in motion and deposit them for different fine soil materials (Source: Hjulström 1935)

while the range of the effective porosity (neff ) is much smaller (1.25–5%). For a soil with neff ¼ 2.5% and a rather

Particle

Erosion velocity

Deposition velocity

Material

sizes (μm)

range (m s 1)

range (m s 1)

Fine sand

125–250

0.20–0.22

0.008–0.02

Very fine sand

63–125

≈0.20

0.005–0.008

Silt

4–63

0.20–1.50

<0.005

Clay

<0 4

>1.50

–

horizontal groundwater table (i ¼ 0.001) with low K (10 9 m s 1) the expected transport time for a 100 m distance would be close to 160,000 years while for a steep groundwater table (i ¼ 0.1) and high K (10 6 m s 1) the transport time would be less than half a year (Table 5). For the more likely combinations of K and i (K ¼ 10 7 to 10 8 m s 1

39

A. Lundberg et al.

Table 5

|

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

Groundwater transport times for a 100 m distance for different combinations of K, neff, and i. The top value is a reference velocity (time) for the area with typical

|

49.1

|

2014

Luleå and they are, therefore, only representative for that

combinations K, neff, and i. A and B are combinations giving short and long

location and that particular year. There is a strong corre-

transport times, respectively

lation between the snowfall and the amount of snow

Parameter combination

K (m s 1)

neff ( )

i ( )

Ref

1.0 × 10 7

0.025

0.010

A, min transport times

1.0 × 10 6

0.013

0.100

B, max transport times

1.0 × 10 9

0.050

0.001

158,549

Ref, but low K

1.0 × 10 8

0.025

0.010

793

Ref, but high K

1.0 × 10 6

0.025

0.010

7.9

Ref, but high i

1.0 × 10 7

0.025

0.100

7.9

Ref, but high neff

1.0 × 10 7

0.050

0.010

Time (years)

79 0.4

159

deposited in that year (Reinosdotter et al. ). However, even for the same amount of snowfall, the deposited snow amounts and contaminant concentrations will differ with time. Lower contaminant concentrations would be expected for a year with few, high-intensity snowfall events compared to a year with many low accumulation events since highintensity snowfall events force municipalities to remove the snow quickly (Viklander b). Snow handling practices

also

determine

which

anti-skid

material

(salt,

chemicals, or grit/sand) is used. Urban snow deposits from and i ¼ 0.01) the transport time would be 80–800 years. For

areas using salt de-icers instead of grit will have higher dis-

an effective porosity of 5%, the above transport times would

solved metal concentrations since salt-treated roads have

be twice as large.

less suspended solids, resulting in lower snow alkalinity and pH (Reinosdotter & Viklander ).

Melt rates for the surface snow deposit were estimated

DISCUSSION

using a degree-day method, normally used for natural snow; however, the technique has also been employed as a pilot

As stated earlier, an engineering approach with rather crude

urban snow deposit (Sundin ) and for a soot-dusted snow-

assumptions was employed in this study. In this section

pack (Lundberg & Beyerel ), both studies performed in

some of these simplifications, optional ways to perform the

Luleå. The DDF value (mm, day 1, C 1) used for the

estimates, a few attempts to determine whether the simplifi-

summer months in this study (4) was smaller than the

cations are likely to over- or underestimate the parameters,

values in the two other studies: (11) for the former and (6)

and some comparisons with other studies are discussed.

for the latter study. The pilot deposit in the first study was

Finally, the future for snow cooling is discussed, and we

small and high (about one-tenth of the surface area of this

explain why the plans for a snow cooling plant have not

study) with much larger melt from the sides, which explained

yet been realized.

the large melt. Also, the latter study with a soot-dusted snow-

W

The pollutant masses were estimated using snow

pack, reported larger DDF values than the ones used for this

samples taken from road banks but standard deviations of

study but a thick layer of grit and sand on the snow deposit sur-

the snow samples were of the same order as the pollutant

face, which insulates the snow and reduces the melt rate.

concentrations thus these estimates are rather crude. The

Reductions in melt rates by 35–85% due to the layer of

assumption of 50% high and low traffic loads also adds to

debris with a thickness from 0.1 to 0.4 m for glaciers were,

the uncertainty. An alternative way to estimate the pollutant

for example, reported by Skogsberg & Lundberg () so

masses could be to use regional snow melt runoff concen-

the DDF value used here does not seem unrealistic. The

tration values, which sometimes are available. However,

melt rates were assumed to be the same each day; however,

these concentrations also show large variations in the

for estimates of particulate transport, the average values are

metal content (dissolved and particulate) over the same

of minor importance since most of the suspended (and poss-

winter season with systematic differences between quality

ible bed load) transport will occur during extreme runoff

of snow melt runoff and snow (Reinosdotter & Viklander

events. Neglecting the variations in melt between different

). Therefore, it was difficult to justify the use of such

days and rain on snow events will underestimate the surface

values. The pollutant masses and concentrations determined

runoff on warm and/or rainy days. The average melt during

in this study were based on measurements from 1 year in

July was estimated to be 64 mm day 1. However, due to

40

A. Lundberg et al.

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

|

49.1

|

2014

variations in air temperature between the days, the maximum

The pollutants were assumed to be conservative so

melt rate could be expected to about be 100 mm. Then, if

adsorption was neglected in the estimates of transport

50 mm rainfall is added on such a day (hot day with following

times thus showing a worst-case scenario. However, most

intense thunderstorm), we end up with a combined melt and

pollutants interact with the soil at some level; not even inor-

rain contribution of 150 mm. For such an event, the surface

ganic chloride, which has often been used as a conservative

runoff fraction from a surface snow deposit would be large

tracer in hydrological studies (Christophersen & Neal ),

6

even for more permeable soils (≈64% for K ¼ 10

1

m, s ).

is any longer regarded as fully conservative (Öberg &

For the estimate of the hydraulic conductivity needed to

Sanden ). Therefore, the actual transport time will for

transport all snow melt from the underground snow storage

most cases be slower than the ones calculated here. For

the natural groundwater flux was neglected; soil depth was

further information regarding the effects of particle surface

assumed to be 20 m and the hydraulic gradient at the

potential, pH, and ionic strengths on the adsorption of

border of the storage to be 0.5 in all directions. The natural

metal ions in glacial till soil, readers are referred to, for

groundwater flow to the underground storage can be esti-

example, Al-Hamdan & Reddy (). For extremely low

mated as the upstream area multiplied by the natural

hydraulic conductivities and very high pollutant concen-

annual recharge (0.4 m) and the distance to the upstream

tration gradients, diffusion transport may also have to be

water divider was estimated to be 250 m so with the plant

considered. However, for groundwater velocities in glacial

width ¼ 72 m the resulting natural groundwater flow

till and the pollutant concentration gradients expected in

became about 18,000 m3 year 1. Consequently, the assump-

this case, the advection transport can be assumed dominant

tion of negligible natural flow (compared to the snow melt

allowing diffusion to be neglected, and dispersion does not

produced flow) could be justified. Hydraulic gradient of

influence the average transport velocity so the entire pollu-

6

about 0.15, a hydraulic conductivity of 10

1

m s , and a

tant mass is here assumed to move with piston flow.

soil depth of 20 m are required to transport the natural

The lined snow cooling plant in Sundsvall is still

flux. Then, it seems more likely that gradient in the flow

running successfully without failure after 11 years of oper-

direction would be about 0.65 (0.50 þ 0.15) in the original

ation, with recently increased capacity and functionality

flow direction and about 0.35 (0.50 0.15) in the opposite

(Larsson ). The water level in the plant has been

direction, not altering the total flow.

raised, the snow gun capacity, used to produce additional conditions were

snow when snow supply is limited, as well as the storage

assumed for the estimates of groundwater transport times.

volume has been doubled, and the snow dumping capacity

However, vertical conductivity is often lower than the hori-

tripled. Furthermore, the maximum power extraction

zontal conductivity for layered soils; and thin layers with

capacity has been doubled, the cooling energy more than

low conductivity will reduce the vertical flux considerably.

doubled, while the operation costs have been reduced by

Such effects were not taken into account. Soil under and

50%. Electric cooling has replaced many different appli-

around both the existing deposit and the proposed plant

cations resulting in a reduction in electricity consumption

was assumed to have typical conductivities for the area;

of more than 90%. Some purification of the polluted urban

however, for other sites local conditions need to be used.

melt water was achieved together with a 90% reduction in

The infiltration calculations for the surface deposit were sim-

CO2 emission (Larsson ). The plant received a Nordic

Homogenous and isotropic soil

plified by using a constant surface area and only vertical flux

environmental award and thereafter attracted international

was considered. In reality, water will move sideways due to

attention from Russia, Japan, Finland, and Norway (http://

differences in soil moisture contents. Then, the real infiltra-

sundsvall.lny.se/). The future for snow cooling was outlined

tion cross-sectional area will be larger than the ones used

by Larsson (), who stated that snow cooling is an extre-

here and large parts of the down-stream area of a surface

mely energy efficient solution which can be used in areas

deposit will likely act as infiltration areas when surface

with more than 1–2 months of snowfall a year. The tech-

water is flowing across it, reducing the runoff fraction reach-

nique is especially favorable when snow clearing is

ing the recipient.

needed. Snow can be produced by snow guns in places

41

A. Lundberg et al.

|

Pollutant pathways from snow deposits

Water Quality Research Journal of Canada

|

49.1

|

2014

where the temperature is less than 5 C for more than

to the effective porosity. This implies that depending on soil

200 h per winter season at a cost of about 2 EUR per

type and terrain characteristics, transport times for a 100 m

MWh cold. This means that the technique can in Europe

distance might vary from months to thousands of years.

W

be used, for example, in Scandinavia, the Alps, the Pyrenees,

Turning a surface snow deposit into a snow cooling

and the Carpathians. Apart from this snow plant, storage for

plant with a permeable underground snow storage would

cooling purposes has primarily been tested and applied in

reduce the total pollutant load if the hydraulic conductivity

Japan (GCT ; Hamada et al. ). The building of the

of the soil is low enough to produce surface runoff during

here-planned permeable plant has been postponed due to

events with high melt rates (and/or combined with high

too large investment costs (mainly the interiors of buildings).

rain rates), while it is still permeable enough to allow all snow melt from the permeable underground snow storage to be transported as groundwater. The hydraulic conduc-

CONCLUSIONS

tivities required to avoid surface runoff from both types of snow storages seemed to be of the same order of magnitude

In the snow deposit for Luleå city with roughly 46,000 inhabi-

(>10 6 m s 1). Thus from this perspective, the design

tants and about 105 ton deposited snow, around 620,000,

seemed not important; however, the snow cooling plant

1,200, and 380 kg of SS, Cl, and N, respectively, along with

with underground snow storage is still preferable since

67, 17, 160, and 170 kg of Cu, Pb, Zn, and P, respectively,

piping and filters for filtering the snow melt are needed for

were found. About 98% of the metals were in the particulate

the operation of the plant and can directly be used also for

phase, and normally remained on the existing surface snow

filtering the rejected water. The uncertainties in the esti-

deposit after melting. The hydraulic conductivity of the

mates of the conductivities are, however, large and more

underlying soil and the melt rate (sometimes combined

detailed studies are needed to confirm these preliminary

with rain rate) determined the fraction of polluted snow

findings. An additional advantage for the cooling plant is,

melt (and rain) that might become surface runoff from the

of course, the reduced need for electricity for cooling pur-

surface deposit spreading the particulate bound metals. For

poses. The demand for such plants can be expected to

1

rates of 30 mm day , the fraction of surface runoff would

increase in countries with snowfall for more than a few

be about 97% for a soil with K ¼ 10 8 m s 1, about 71% for

months per year, and the technique is particularly favorable

K ¼ 5 × 10

7

6

but nonexistent for a soil with K > 10

1

ms .

when snow clearing is required. For years with lower snow-

Rather high surface velocities are required to transport the

fall than normal years, snow guns can be used to produce

pollutants attached to the soil particles, silt, and sand.

additional snowfall provided that the air temperature is

For the underground snow storage, it was hard to deter-

less than 5 C for at least 200 h per winter.

W

mine the order of magnitude of hydraulic conductivities and gradients needed to ensure that all snow melt could be transported through the ground to avoid rejection of polluted

ACKNOWLEDGEMENT

water at the soil surface. However, with the assumptions that the soil depth was 20 m, the snow melt was moving

Dr Kjell Skogsberg, Snowpower (http://www.snowpower.

radially out from the snow storage, and the hydraulic gradi-

se/organisation_en.asp), is acknowledged for providing

ent at the snow storage border was 0.5, it was found that a

background knowledge on the planned permeable snow

hydraulic conductivity value of about 10 7 m s 1 was

cooling plant.

required to ensure that all snow melt could be transported through the soil.

The horizontal groundwater transport velocity from

REFERENCES

both surface snow deposit and from underground snow storages is directly proportional to the hydraulic conductivity and the hydraulic gradients and inversely proportional

Al-Hamdan, A. Z. & Reddy, K. R.  Adsorption of heavy metals in glacial till soil. Geotech. Geol. Eng. 24 (6), 1679–1693.

42

A. Lundberg et al.

|

Pollutant pathways from snow deposits

Christophersen, N. & Neal, C.  Linking hydrological, geochemical and soil processes on the catchment scale: an interplay between modeling and fieldwork. Water Resour.

Res. 26, 3077–3086.

Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for community action in the field of water policy. 22.12.2000.

Official J European Communities L 327/1.

Directive 2006/118/EC of the European Parliament and of the Council of 12 December 2006 on the protection of groundwater against pollution and deterioration. 27.12.2006.

Official J European Union L 372/19.

GCT  Good clean teach. The independent guide to ecotechnology.

Updated October 15, 2008. http://goodcleantech.pcmag.com/ news-and-events/280266-new-chitose-airport-in-japan-to-usesnow-for-30-percent-of-cooling-needs.

Hamada, Y., Nagata, T., Kubota, H., Ono, T. & Hashimoto, Y.  Study on a snow storage system in a renovated space. Renew.

Energy 41, 401–406. Available at: http://dx.doi.org/10.1016/ j.renene.2011.11.012 Hjulström, F.  Studies of the morphological activity of rivers as illustrated by the River Fyris. Bull. Geol. Inst. Univ. Uppsala 25, 221–527.

Larsson, P. E.  Best practice in tackling new and emerging challenges: Snow cooling as a Nordic solution. Rio þ 20 Reg Science Technol. Workshop for Europe. October 12–14, 2011, Helsinki, Finland.

Lundberg, A. & Beyerel, B.  Ash on snow – a tool to prevent flooding? Nordic Hydrol. 32 (3), 195–214.

Mälkki, E.  Groundwater flow conditions in the coastal bedrock area of the Gulf of Finland. Geol. Quart. 47 (3), 299–306. Öberg, G. & Sanden, P.  Retention of chloride in soil and cycling of organic matter-bound chlorine. Hydrol. Proc. 19, 2123–2136.

Oberts, G. L.  Influence of snow melt dynamics on stormwater runoff quality. Watershed Protect. Tech. 1 (2), 55–61.

Water Quality Research Journal of Canada

|

49.1

|

2014

Oberts, G. L., Marsalek, J. & Viklander, M.  Review of water quality impacts of winter operation of urban drainage. Water Qual. Res. 35 (4), 781–808.

Reinosdotter, K.  Sustainable Snow Handling. Thesis, Luleå University of Technology.

Reinosdotter, K. & Viklander, M.  Handling of urban snow with regard to snow quality. J. Environ. Eng. 132 (2), 271–278.

Reinosdotter, K. & Viklander, M.  Road salt influence on pollutant releases from melting urban snow. Water Qual.

Res. J. Can. 42 (3), 153–161.

Reinosdotter, K., Viklander, M. & Söderberg, H.  Development of Snow-handling Strategies in Swedish Municipalities – Potentials for Sustainable Solutions. Thesis Sustainable Snow Handling, Luleå University of Technology.

Salonen, V.-P., Eronen, M. & Saarnisto, M.  Applied Soil Geology (in Finnish). Kirja-Aurora, Turku.

Skogsberg, K. & Lundberg, A.  Wood chips as thermal insulation of snow. Cold Reg. Sci. Technol. 43 (3), 207–218.

Skogsberg, K. & Nordell, B.  The Sundsvall hospital snow storage. Cold Reg. Sci. Technol. 32, 63–70.

Sundin, E.  An approach to using the degree day method for urban snow deposit melt. Vatten 54 (2), 123–130.

Viklander, M.  Urban snow deposits – pathways of pollutants.

Sci. Total Environ. 189/190, 379–384.

Viklander, M. a Snow Quality in Urban Areas. Thesis, Luleå University of Technology.

Viklander, M. b Substances in urban snow. A comparison of the contamination of snow in different parts of the city of Lulea, Sweden. Water Air Soil Pollut. 114 (3–4), 377–439.

Westerlund, C. & Viklander, M.  Particles and associated metals in road runoff during snow melt and rainfall. Sci. Total Environ. 362, 143–156.

Westerlund, C., Viklander, M. & Bäckström, M.  Seasonal variations in road runoff quality in Luleå, Sweden. Water Sci. Technol. 48 (9), 93–101.

First received 15 August 2012; accepted in revised form 30 January 2013. Available online 27 August 2013

43

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Ecological benefit of the road salt code of practice Bruce W. Kilgour, Bahram Gharabaghi and Nandana Perera

ABSTRACT Despite an overall increase in total road salt used over the past 14 years (the data record in this manuscript), there has been a 26% reduction in the rate (normalized as tonnes of salt per cm of snow per km of road) of road salt application by the City of Toronto since that city implemented mitigations from the Road Salt Code of Practice. The ecological benefit of the reduced use of road salt was approximated by comparing the estimated 26% salt reduction to the distribution of chloride tolerances that has been recently published by the Canadian Council of Ministers of the Environment (i.e., CCME). Species sensitivity distributions predict that between 1 and 14% of taxa would benefit from a 26% reduction in chloride concentrations in surface waters. Assuming that a typical ‘healthy’ Canadian watercourse might support between 100 and 200 species of fish, invertebrates and plants, the Code of Practice might provide benefit to between 14 and 28 species. However, the net ecological benefit of implementing the Code may be undermined in rapidly urbanizing watersheds where road networks continue to expand at a rate of 3–5% per year and chloride loads to urban streams are steadily increasing.

Key words

Bruce W. Kilgour (corresponding author) Kilgour & Associates Ltd, 16, 2285C St. Laurent Boulevard, Ottawa, Ontario, K1G 4Z6, Canada E-mail: bkilgour@kilgourassociates.com Bahram Gharabaghi Nandana Perera School of Engineering, University of Guelph, Guelph, Ontario, N1G 2W1, Canada Nandana Perera Computational Hydraulics Int., 147 Wyndham St N., Guelph, Ontario, N1H 4E9, Canada

| chlorides, ecological benefits, maximum field distribution, road salt, species sensitivity distribution

INTRODUCTION Snow and ice conditions on the road system have a signifi-

Rutherford & Kefford ). Elevated concentrations of

cant impact on public safety, roadway capacity, travel time

chlorides in surface waters can cause changes in behavior

and economic costs (Keummel ). In Canada, control

(e.g., increased ‘drift’ of stream invertebrates; Crowther &

of snow and ice on road pavements and sidewalks is gener-

Hynes ), and increase mortality rates of aquatic organ-

ally achieved by a combination of de-icing with road salts

isms (Evans & Frick ; Benbow & Merritt ).

and mechanical plowing. Each year, approximately 5

Cladocerans (e.g., Ceriodaphnia) are considered particularly

million tonnes of road salts are used as de-icers on roadways

sensitive to chlorides with concentrations as low as about

in Canada (Environment Canada ). The City of Toronto,

450 mg/L causing harm to individuals during short-term

with about 5,500 km of dense road network (City of Toronto

exposures (Dowden & Bennet ; Elphick et al. ).

) relies on about 135,000 tonnes per year salt appli-

The larvae of some select rare and endangered Canadian

cation during the winter to provide safe transportation

freshwater mussels are also highly sensitive to chloride

surfaces for all road and sidewalk users in an efficient and

with concentrations as low as 113 mg/L causing harm to

affordable manner.

individuals in laboratory toxicity tests (Bringolf et al. ;

Both aquatic and terrestrial ecosystems can be adversely

Gillis et al. ; Gillis ).

affected by exposure to chloride concentrations associated

Chloride concentrations vary spatially in Canada.

with the typical use of road salts (USEPA (United States

Natural background freshwater chloride concentrations

Environmental Protection Agency) ; Novotny et al.

are generally in the 10–20 mg/L range, but stream chloride

; Environment Canada & Health Canada ;

concentrations in highly urbanized areas can be as high as

doi: 10.2166/wqrjc.2013.129

44

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

Water Quality Research Journal of Canada

10,000 mg/L. Loadings are highest in urban centers includ-

|

2014

of 860 mg/L which is to be applied to exposure durations of 4 h or less. The Canadian Council of Ministers of the

Concentrations of chlorides are generally increasing in

Environment (CCME) recently published a national guide-

ground (Kincaid & Findlay ; Mullaney et al. )

line for chloride (CCME ). CCME recommended that

Toronto

and

Montreal

(Mayer

al.

49.1

).

ing

et

|

and surface waters (Todd et al. ) in urbanized areas.

concentrations of 640 mg/L would protect most species

There may be a relationship between percent impervious-

(95%) during short-term acute exposures, while 120 mg/L

ness and chloride concentrations in streams, based on

would protect most species under longer-term chronic

work in the USA (Kaushal et al. ), and implying that

exposures. CCME () further recognized that some water-

areas with an imperviousness of >30–40% are likely to

sheds in southwestern Ontario contain species of Unionidae

have chloride concentrations in their surface waters of

(freshwater mussels; northern riffleshell – Epioblasma toru-

some 200–300 mg/L.

losa rangiana and the wavy-rayed lamp mussel – Lampsilis

Environment Canada classified road salt a toxic sub-

fasciola) that are not only highly sensitive to chloride

stance on the basis of extensive review of fate and effects

during their larval stages (concentration lethal to 10% of

(Evans & Frick ). That classification required that

larvae is ∼24 mg/L), but are rare and at risk of extirpation

Environment Canada consider regulatory instruments for

(COSEWIC (Committee on the Status of Endangered Wild-

mitigating the risks that were considered to be posed by

life in Canada) a, b), and that these watercourses may

the product. Environment Canada (EC) & Health Canada

require further protection greater than that afforded by the

(HC) () carried out additional research assessing various

general guidelines.

mitigation techniques (Marsalek ; Stone & Marsalek

The objective of this paper is to provide one estimate of

), and then developed its Code of Practice (EC ).

the ecological benefit of Environment Canada’s implemen-

The Code of Practice is an assemblage of best practice

tation of the Road Salt Code of Practice. The estimate here

guidelines for reducing the use of road salts in municipalities

is specific to the City of Toronto’s experiences. Reductions

that use large amounts of the product. Environment

in chloride loads to streams, and associated concentrations

Canada is encouraging municipalities that use more than

in surface waters, were modeled using road salt application

500 tonnes of road salt per year to implement the mitiga-

rates for the City of Toronto before and after implemen-

tions recommended in the Code. Municipalities that

tation of the Code of Practice. The ecological benefit was

implement the best practices (including reduced application

estimated by comparing the observed salt reduction (pre to

rates and more efficient timing of application) have been

post implementation of the Code) to the distribution of

able to reduce chloride concentrations in groundwater by

chloride tolerances (Posthuma et al. ) that has been

50% over a 3–4 year period (Bester et al. ; Stone et al.

recently published by the Canadian Council of Ministers

).

of the Environment (i.e., CCME).

The ecological benefit of implementing the Code is questionable, considering that our road networks are continuing to expand and chloride loads to watersheds are increasing.

METHODS

Estimating the ecological benefit of the Code of Practice requires an understanding of the relationship between

Normalized road salt loadings

exposure concentration (as well as frequency and duration) and ecological effect. Chloride tolerances determined from

Road salt application rates were computed for the Toronto

laboratory toxicity tests provide one line of evidence of the

area for the period 1996 to 2007. Salt application rates

potential effects that chloride concentrations can have in

were provided by the City of Toronto – Transportation Ser-

the

toxicity

vices and by the Ontario Ministry of Transportation. Road

thresholds for chloride including a longer-term chronic cri-

environment.

USEPA

()

developed

salt application rates within watersheds and catchments

terion of 230 mg/L which is to be applied to exposure

depend in part on road density and primarily on the

durations of 96 h or more, and a short-term acute criterion

weather. Higher road densities generally result in more

45

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

Water Quality Research Journal of Canada

|

49.1

|

2014

road salt being applied within a catchment. Greater

for all of the tens of thousands of aquatic organisms that

amounts of snow also generally require a greater amount

occur in surface water environments (creeks, streams,

of road salt application in a given year. Changes in surface

rivers, ponds, lakes) in Canada (Morton & Gale ). The

water concentrations pre-post the Road Salt Code of Prac-

SSDs, which are approximately normally distributed, how-

tice (i.e., before versus after 2004) could, therefore, have

ever, can generally be used to predict the percentage of

varied because of changes in road density (i.e., increase)

species that will be affected when exposed to chlorides for

or changes in snowfall. Road salt application rates were

short- or long-term periods. Both short- and long-term

thus standardized for road density and snowfall to better

SSDs were considered well fit using a log-Weibull model

understand the impact of the Code of Practice on road salt

with the following form (as per CCME ()):

application rates. Annual snowfall data for the Toronto area were obtained from the nearby Environment Canada weather station in North York. Digital road network data

x k y¼1 e λ

were obtained from the Ontario Ministry of Transportation.

The relationship between road salt application rate (tonnes

where, y is the percentage of species affected, x is the logar-

per year) and road density, as well as between road salt

ithm of the chloride concentration, and λ and k are

application rate and snowfall accumulation (cm per year)

constants that define the shape and form of the relationship.

were quantified annually for the years 1996 through 2007.

In the case of the short-term SSD, λ was 3.6268 and k was

Salt application rates were scaled to road density and snow-

10.9917; in the case of the long-term SSD, λ was 3.2119

fall accumulation, and ultimately expressed as kg salt/km

and k was 7.0473. The SSDs are reproduced in Figure 4.

road/cm snow. Total length of salt-applied roads within

The SSD models were used with the normalized

the watershed was considered for the normalization as all

reduction in road salt application rates to quantify a poten-

roads drain to the stream network within a couple of

tial ecological benefit to having implemented the Road

hours due to lower roughness in storm sewer systems. The

Salt Code of Practice. Here, the reduction in normalized

lag time in getting salty runoff to a stream is, therefore,

road salt application rates was taken as an estimate of the

much smaller than the time frame considered for short-

reduction in chlorides that could be anticipated, all other

term exposure for aquatic organisms (generally 24–48 h).

factors being considered, after implementation of best prac-

A simple t-test was used to test for differences in normalized

tices post the Road Salt Code of Practice. So for example, if

application rates between the period before (1996–2003)

the normalized road salt loadings were to have decreased by

and the period after implementation of the Road Salt Code

10% post implementation of best practices, it was assumed

of Practice (2004–2007). The percent change in normalized

that chloride levels in surface waters in Toronto area

road salt application rate was computed from before to after

streams would be roughly 10% lower than if the Code of

implementation of the Code.

Practice had not been implemented. It is recognized that there are various lags when chlorides transport to streams,

Quantifying ecological benefit

and that some of the lags are considerable (e.g., decades in some cases). For the purpose of this paper, it was assumed

CCME () reviewed the existing chloride toxicity litera-

that chloride loadings to watercourses responded immedi-

ture and developed a species-sensitivity distribution (SSD),

ately to reductions in application of road salt to roadways.

which describes the expected distribution of species toler-

The ecological implication of that reduction was estimated

ances when exposed to chloride. CCME developed two

considering the magnitude and form of the short-term and

curves: (1) the first SSDa was for acute, or short-term (24–

long-term SSDs. We computed the percentage of species

48 h) exposure to chloride; and (2) the second SSDc was

that would be anticipated to benefit from reductions in

for chronic, or long-term (96 h or longer) exposure to chlor-

chlorides (as per the estimated reduction in normalized

ide. Both SSDs were based on the set of taxa for which

road salt application rates), using both the short-term and

reliable toxicity data are available; data are not available

long-term SSDs.

46

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

RESULTS

Water Quality Research Journal of Canada

|

49.1

|

2014

was related to cumulative snowfall (Figure 1). A little less than 5% of the variation in salt application rate was related

Reduction in normalized road salt application rates

to road density (Figure 2). Salt application rates normalized for both road density and snowfall are illustrated in Figure 3.

Road salt application in the City of Toronto varied by an

There was a subtle but distinctive reduction in road salt

order of 3× from a low of ∼57,000 tonnes in the winter of

application rates in the period defined as being after the

2001/2002 to ∼200,000 tonnes in the winter of 2007/2008

Road Salt Code of Practice was implemented. The difference

(Table 1). Over 50% of the variation in road salt application

(26% reduction in mean normalized salt application rate)

Table 1

|

Normalized road salt application rates

Total amount of road

Cumulative

Length of roadway

Normalized road salt application

Winter period

salt applied (tonnes)

snowfall (cm)

lanes (km)

(tonnes/cm/km)

1995/06

128,000

150

12,343

0.069

1996/97

157,600

167

12,415

0.076

1997/98

101,900

112

12,493

0.073

1998/99

140,400

124

12,493

0.091

1999/00

142,900

88

13,846

0.117

2000/01

176,600

181

13,800

0.071

2001/02

56,900

77

13,800

0.054

2002/03

208,200

145

13,800

0.104

2003/04

108,200

108

15,052

0.067

2004/05

147,400

166

15,052

0.059

2005/06

94,700

108

15,052

0.058

2006/07

89,100

83

15,052

0.071

2007/08

195,600

233

15,052

0.056

2008/09

147,100

195

15,052

0.050

Figure 1

|

Relationship between cumulative snowfall and tonnes of road salt applied in the Toronto area between the winters of 1995/1996 and 2008/2009.

47

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

Water Quality Research Journal of Canada

Figure 2

|

Relationship between length of roads (km) and tonnes of road salt applied in the Toronto area between the winters of 1995/1996 and 2008/2009.

Figure 3

|

Variations in normalized road salt applications rates, City of Toronto.

|

49.1

|

2014

was statistically significant at a probability level of 0.03%

benefit was greatest within the zone of the SSD in which

(for a one-sided t-test).

the slope of the relationship between species affected and concentration was the most extreme, for both the short-

Ecological benefit

term and long-term relationships.

The relationship between percent of taxa affected and chloride concentration was sigmoid in shape (Figure 4). The

DISCUSSION

percentage of species benefiting from a 26% reduction in chloride concentrations, therefore, varies from a negligible

This analysis demonstrated a potential 26% reduction in

fraction when the chloride concentrations are already low

chloride loads to watercourses in the Toronto area after

(say <200 mg/L), to ∼14% when chloride concentrations

implementation of the Code of Practice. That loading

decrease from ∼6,000 to 4,400 mg/L in a short-term acute

should, in the longer term, result in chloride concentrations

exposure (Table 2; Figure 4). The percentage of species ben-

being ∼26% less than what they otherwise would have been,

efiting also depends on whether the exposure is short- or

had the Code of Practice not been implemented. Those

long-term, with a greater benefit occurring under short-

reductions in chloride concentrations, further, have the

term acute exposure (Figure 4; Table 2). The ecological

potential to benefit as much as 14% of potential freshwater

48

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

Water Quality Research Journal of Canada

|

49.1

|

2014

and that aquatic species will more likely continue to be at an increasing risk; and (4) recognizes that other pollutants and stressors may override the influences of chlorides and further limit the distributions of aquatic organisms in heavily urbanized centers. Each of these points is discussed in greater detail below.

The estimated reduction in chloride concentrations in Toronto-area streams seems to be a reasonable observation based on other published works. In a review of the anticipated benefits of best practices for road salt applications, Gartner Lee Limited (GLL ) predicted about a 20% Figure 4

|

Models of species sensitivity distributions for short-term and long-term exposures to chloride in surface waters. Models are from CCME (2011). Vertical lines indicate 2,000 and 1,400 mg chloride/L; horizontal lines indicate fraction of species affected by 2,000 and 1,400 mg/L in short-term and long-

reduction in chloride loads and chloride concentrations.

The City of Waterloo was further able to reduce total road salt application by 10% in the broader urban road network, and by 25% in the vicinity of well fields known to be

term exposures.

susceptible to road salts (Stone et al. ). Municipal agencies, then, appear to be targeting a general reduction Table 2

|

Percent of taxa benefiting from a 20% reduction in chloride concentrations in surface water

in the use of road salt by some 20–25%, and that level of reduction from historical loadings appears to be a reason-

Initial

Final chloride

% benefiting under

% benefiting under

chloride

(mg/L) with 26%

short-term

long-term

(mg/L)

reduction

exposures

exposures

10,000

7,400

8

2

(), road salt loadings are in part weather dependent,

9,000

6,660

9

2

while changes in climate may result in a requirement to

8,000

5,920

11

3

increase salt use per precipitation event, depending on air

7,000

5,180

12

3

temperatures.

6,000

4,440

14

4

Despite the anticipated loading reductions, it is

5,000

3,700

14

6

considered unlikely in the near term that chloride concen-

4,000

2,960

14

7

trations will significantly decrease in surface waters.

3,000

2,220

13

9

Analysis of Ontario’s long-term data set shows an increas-

2,000

1,480

10

10

ing trend in chloride concentration in river waters in

1,000

740

4

10

almost every region (Todd et al. ). Time trends in

800

592

3

9

four watersheds, two rural and two urbanized, were exam-

600

444

2

8

ined, in particular by Todd et al. (). Chloride

400

296

1

6

concentrations in the Skootamatta River near the village

200

148

<1

3

of Actinolite (surrounded by natural and agricultural

100

74

<1

1

able expectation assuming that weather patterns also remain consistent into the future. As per Perera et al.

lands) have increased from a 1980s baseline of ∼2 mg/L to a present-day value of ∼3 mg/L. Concentrations in the Sydenham River near Owen Sound (population ∼15,000)

taxa. The estimated benefit: (1) assumes that the anticipated

have increased from a 1970s baseline of ∼8–9 mg/L to a

reduction in chloride loads is accurate; (2) recognizes that

present-day value of ∼12 mg/L. Chloride levels in Fletchers

there are time lags between changes in application rates

Creek in Brampton (outside Toronto) have increased from

and concentrations in surface waters; (3) recognizes that

an average of ∼100 mg/L in the 1970s to an average of

chloride loads can be expected to generally increase in the

almost 500 mg/L in 2008. Chloride concentrations in Sher-

future, regardless of implementation of the Code of Practice,

idan Creek in Mississauga have increased from a 1970s

49

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

Water Quality Research Journal of Canada

|

49.1

|

2014

baseline of almost 300 mg/L to a present day average of

urban areas where high chloride levels presently occur

about 800 mg/L. Todd et al. () further tested for stat-

(Mayer et al. ; Morin & Perchanok ). The ecologi-

istical significance in trends over time across the

cal benefits of the Code of Practice are, thus, most likely

province. They compared chloride concentrations in the

to occur in a relatively small fraction of the total land area

period 1980–1985, to those in the period 2000–2004. For

in Canada.

those stations for which there were adequate data to

The magnitude of the ecological benefit estimated here

make the comparison, over 90% demonstrated a statisti-

(14% of taxa) can be re-expressed in real numbers if we con-

cally significant increase in chloride concentration. The

sider the number of species that might naturally occur in a

data, thus, overwhelmingly indicate that chloride concen-

watercourse. Of the some 10,000 aquatic species that

trations in ground and surface waters are increasing, and

occur in watercourses in North American (Morton & Gale

those increases appear to be related to increasing densifica-

; Pennack ), inventories in Canadian waters typi-

tion of road networks.

cally produce about 100–200 individual taxa in a ‘healthy’

The lack of reduction in chloride concentrations in sur-

system if we consider fish, benthic invertebrates and macro-

face waters is in part related to historical loadings that are

phytes. If the calculations are correct, and implementation

now resident in groundwaters which provide a base flow

of the Code of Practice resulted in a reduction in chloride

to surface waters. Groundwater is a storage compartment

load of some 26%, and there was a benefit to some 14% of

for road salts (Kincaid & Findlay ; Mullaney et al.

possible taxa, then the number of taxa in the watercourse

), and has been identified as a particular challenge to

may increase by as many as 14–28 taxa. Such an increase

short-term recovery of surface water concentrations by

in diversity is clearly measurable assuming a statistically

both Canadian and US researchers (Ramakrishna & Virara-

robust study design (EC ).

ghavan ; Wenck Associates Inc. ; Howard & Maier

The ecological benefits of the Code of Practice may how-

; Cooper et al. ; Kincaid & Findlay ; Rubin

ever be masked by other stressors. The ecological benefit

et al. ). The University of Guelph, in association with

from

the City of Toronto, has monitored chloride concentrations

‘urbanization-related’ stressors are not also limiting the eco-

in rivers (Highland Creek, Rouge River, Don River, Humber

logical diversity of a surface-water feature. Many urban

River) in Toronto. Despite an estimated 26% reduction in

centers, however, do release contaminants other than road

normalized road salt application loadings (as estimated in

salt into aquatic receiving environments. Metals and hydro-

this paper), concentrations of chlorides in the Toronto-

carbons, in addition to chlorides, contaminate runoff from

area rivers has not measurably declined (Perera et al.

roadways, at concentrations that are toxic to aquatic ecologi-

reducing

chloride

levels

assumes

that

other

). Road densities increased over that period, while

cal receptors (Murakami et al. ). Nutrients, suspended

weather patterns varied, and total loading of chloride to

particulate material, and pesticides from urban areas also

streams has continued to increase. The long-term ground-

enter surface waters (via stormwater runoff and treated

water concentration of chloride was, further, estimated to

sewage) at concentrations that pose additional risks to

be approximately 275 mg/L, or high enough to pose risks

aquatic organisms (EC ). Storm flows from urban

to aquatic organisms under long-term exposure scenarios.

areas can also be erosive, leading to changes to physical

There has been no calculation of the fraction of

stream habitats, and associated losses of the critical habitats

Canadian surface waters that will benefit from the

of some species. The loss of riparian cover within water-

implementation of the Code of Practice. Impacts are, cur-

sheds leads to alterations in the hydrological cycle (greater

rently, anticipated in heavily urbanized areas with high

storm flow volumes, increase in frequency of stream flashi-

road densities, and in particular in surface waters that

ness), which can lead to an increase in erosivity within

have low dilution (i.e., small watersheds or catchments)

watercourses (Booth & Jackson ). The lack of riparian

draining major roadways. The ecological benefits of

zones further leads to increasing solar inputs to streams,

implementation of the Code of Practice, therefore, are

resulting in warmer summer temperatures (Barton et al.

anticipated to primarily occur in the most densely populated

). There is thus the real potential that reductions in

50

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

Water Quality Research Journal of Canada

|

49.1

|

2014

chloride levels will not result in real, measurable ecological

generally <100 mg/L when percent imperviousness was

benefit because of masking by other stressors. The masking

<10–15%. That relationship could be used as one of several

effect is, however, hard to predict or quantify because we

potential rules of thumb to identify areas where road salts

have a generally limited specific understanding of the toler-

are likely to pose a negligible risk to aquatic receptors.

ances of individual aquatic species to the numerous and various stressors that are present in urban system (see Stanfield & Kilgour (), for example).

We predict, herein, in the short term it is unlikely that we will observe any measurable and apparent ecological benefit of the Road Salt Code of Practice. Many watercourses, in the most salt-impacted regions (i.e., Torontoarea watershed, Mayer et al. ()), are already additionally impacted by other various urbanization related stressors.

Increasing urban expansion and road densification is, further, leading to overall increases in chloride concentrations in surface waters. Further, although the Code of Practice is being implemented by many municipalities in Canada, it does not apply to commercial snow-plowing operations. Environment Canada considers it likely that commercial operators use larger amounts of salt to clear snow and ice from parking lots than typical municipal agencies would or do, in part because the commercial operators are compensated on the basis of use (Stone & Marsalek ). The lack of immediate ecological benefit should, however, not be used as an excuse to not implement the Code. The analyses here demonstrate that implementation of the Code of Practice will lessen the effects from what they might otherwise become as urban areas expand and road densities increase. Second, arguing that we should not address known risks associated with one substance because there are other known risks associated

CONCLUSIONS The Code of Practice appears to have contributed to a reduction in the ‘normalized’ road salt application by about 26%. Despite increasing urbanization and densification of road networks, a 26% reduction in normalized application means in the long term that chloride levels will at least not be increasing by that amount. SSD in contrast predicted between 1 and 14% of taxa would benefit from a 26% reduction in chloride concentrations in surface waters. Thus, the Road Salt Code of Practice can be expected to benefit up to 14% of potential freshwater taxa over the long term. We predict that we will not in the near term observe ecological benefits of having implemented the Code of Practice because: (1) road salts pose ecological risks in limited areas (i.e., densely populated urban areas in southern parts of the country); (2) other urban stressors are expected to mask small ecological benefits associated with small reductions in chloride loads; (3) chloride loadings and concentrations are generally increasing in association with increasing urbanization. The lack of observed ecological benefit should not be used as an argument to not implement the Code of Practice in areas where road salts clearly pose risk to aquatic organisms.

with another substance sets a dangerous precedence that could perpetuate risk legacies. The areas within which road salts pose significant ecological risks in Canada are

ACKNOWLEDGEMENTS

relatively small and localized (i.e., to those areas that are heavily urbanized like the City of Toronto; Mayer et al.

The concept for this manuscript was developed while the

). There are large areas in Canada where the risks

authors were contracted by Environment Canada (Lise

associated with road salts can be considered negligible to

Trudel) to explore the notion of ecological benefit related

the point that implementation of the Code of Practice

to the Code of Practice. Other contributors, either in-kind

would have a negligible benefit to ecological receptors.

or intellectually included Scott Jarvie (Toronto and Region

Kaushal et al. () recently examined the association

Conservation Authority), Peter Noehammer (City of

between chloride concentrations and percent impervious-

Toronto), Tim Fletcher and Monika Nowierski (Ontario

ness. They demonstrated, for watercourses in Baltimore

Ministry of Environment), and Steve Struger (Ontario

that chloride concentrations in surface waters were

Ministry of Transportation).

51

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

REFERENCES Barton, D. R., Taylor, W. D. & Biette, R. M.  Dimensions of riparian buffer strips required to maintain trout habitat in southern Ontario streams. N. Am. J. Fish. Manage. 5, 364–378.

Benbow, M. E. & Merritt, R. W.  Road salt toxicity of select Michigan wetland macroinvertebrates under different testing conditions. Wetlands 24, 68–76.

Bester, M. L., Frin, E. O., Molson, J. W. & Rudolph, D. L.  Numerical investigation of road salt impact on an urban wellfield. Ground Water 44, 165–175.

Booth, D. B. & Jackson, C. R.  Urbanization of aquatic systems: degradation thresholds, stormwater detection, and the limits of mitigation. J. Am. Water Resour. Assoc.

33, 1077–1090.

Bringolf, R. B., Cope, W. G., Eads, C. D., Lazaro, P. R., Barnhart, M. C. & Shea, D.  Acute and chronic toxicity of technical-grade pesticides to glochidia and juveniles of freshwater mussels (unionidae). Environ. Toxicol. Chem.

26 (10), 2086–2093.

CCME (Canadian Council of Ministers of the Environment)  Scientific Criteria Document for the Development of the Canadian Water Quality Guidelines for the Protection of Aquatic Life, Chloride Ion, PN 1460. Canadian Council of Ministers of the Environment, Winnipeg, MB. City of Toronto  Salt Management Plan. Prepared by Transportation Services Division, Toronto, ON. Cooper, C. A., Mayer, P. M. & Faulkner, B. R.  The influence of road salts on water quality in a restored urban stream.

16th National Nonpoint Source Monitoring Workshop, September 14–18, 2008, Columbus, OH. COSEWIC (Committee on the Status of Endangered Wildlife in Canada) a COSEWIC assessment and status report on the Wavy-rayed Lamp mussel Lampsilis fasciola in Canada.

COSEWIC. COSEWIC (Committee on the Status of Endangered Wildlife in Canada) b COSEWIC assessment and status report on the Northern Riffle shell Epioblasma torulosa rangiana in Canada. COSEWIC. Crowther, R. A. & Hynes, H. B. N.  The effect of road deicing salt on the drift of stream benthos. Environ. Pollut. 14, 113–126.

Dowden, B. F. & Bennet, H. J.  Toxicity of selected chemicals to certain animals. Water Pollut. Control Fed. 30 (9), 1308–1316.

EC (Environment Canada)  Threats to Sources of Drinking Water and Aquatic Ecosystem Health in Canada. National Water Research Institute, Burlington, Ontario. NWRI Scientific Assessment Report Series No. 1, 72.

EC (Environment Canada)  Metal Mining Guidance Documents for Aquatic Environmental Effects Monitoring.

National EEM Office, Environment Canada, Ottawa, Ontario.

EC (Environment Canada)  Code of Practice for the Environmental Management of Road Salts. Environmental Protection Services, Ottawa, ON.

Water Quality Research Journal of Canada

|

49.1

|

2014

EC & HC (Environment Canada & Health Canada)  Priority Substances List Assessment Report – Road Salts. Minister of Public Works and Government Services, Ottawa, ON. Elphick, J. R. F., Bergh, K. D. & Bailey, H. C.  Chronic toxicity of chloride to freshwater species: effects of hardness and implications for water quality guidelines. Environ. Toxicol.

Chem. 30, 239–246.

Evans, M. & Frick, C.  The effects of road salts on aquatic ecosystems. NWRI Contribution Series No. 01-000, August 2001.

Gillis, P. L.  Assessing the toxicity of sodium chloride to glochidia of freshwater mussels: implications for salinization of surface waters. Environ. Pollut. 159, 1702–1708.

Gillis, P. L., Mitchell, R. J., Schwalb, A. S., McNichols, K. A., Mackie, G. L., Wood, C. M. & Ackerman, J. D.  Sensitivity of the glochidia (larvae) of freshwater mussels to copper: assessing the effect of water hardness and dissolved organic carbon on the sensitivity of endangered species.

Aquat. Toxicol. 88, 137–145.

GLL (Gartner Lee Limited)  Review of ecological/ environmental monitoring opportunities for assessing MTO’s salt management initiatives. Report prepared by Gartner Lee Limited, Guelph, Ontario, for the Ontario Ministry of Transportation, St. Catharines, Ontario.

Howard, K. W. F. & Maier, H.  Road de-icing salt as a potential constraint on urban growth in the Greater Toronto Area, Canada. J. Contam. Hydrol. 91, 146–170.

Kaushal, S. S., Groffman, P. M., Likens, G. E., Belt, K. T., Stack, W. P., Kelly, V. R., Band, L. E. & Fisher, G. T.  Increased salinization of fresh water in the northeastern United States.

Proc. Natl. Acad. Sci. USA 102, 12517–13520.

Keummel, D. A.  The public’s right to wintertime traffic safety.

In: Transportation Research Board 3rd annual International Symposium on Snow Removal and Ice Control Technology, Minneapolis, MN. Kincaid, D. W. & Findlay, S. E. G.  Sources of elevated chloride in local streams: Groundwater and soils as potential reservoirs. Water Air Soil Pollut. 203, 335–342.

Marsalek, J.  Road salts in urban storm water: an emerging issue in storm water management in cold climates. Water Sci. Technol. 48 (9), 61–70.

Mayer, T., Snodgrass, W. J. & Morin, D.  Spatial characterization of the occurrence of road salts and their environmental concentrations as chlorides in Canadian surface waters and benthic sediments. Water Qual.

Res. J. Can. 34, 545–574.

Morin, D. & Perchanok, M.  Road Salt Loadings in Canada, Supporting Document for Road Salts, Submitted to the Environmental Resource Group for Road Salts, Commercial Chemicals Evaluation Branch, Environment Canada.

Morton, W. B. & Gale, G. E.  A Guide to Taxonomic Standards for Identification of Ontario Freshwater Invertebrates. Ontario Ministry of Natural Resources, Fisheries Branch, Ontario, Canada, Version 85.

52

B. W. Kilgour et al.

|

Ecological benefit of road salt code of practice

Mullaney, J. R., Lorenz, D. L. & Arntson, A. D.  Chloride in Groundwater and Surface Water in Areas Underlain by the Glacial Aquifer System, Northern United States, National Water-Quality Assessment Program, Scientific Investigations Report 2009–5086. US Department of the Interior, US Geological Survey.

Murakami, M., Sato, N., Anegawa, A., Nakada, N., Haraada, A., Komatsu, T., Takada, H., Tanaka, H., Ono, Y. & Furumai, H.  Multiple evaluations of the removal of pollutants in road runoff by soil infiltration. Water Res. 42, 2745–2755.

Novotny, V., Smith, D. W., Keummel, D. A., Mastriano, J. & Bartosova, A.  Urban and Highway Snowmelt: Minimizing the Impact on Receiving Water. Water Environment Research Foundation, Alexandria, VA. Pennack, R.  Fresh-water Invertebrates of the United States: Protozoa to Mollusca, 3rd edn. Wiley Interscience, New York.

Perera, N., Gharabaghi, B., Noehammer, P. & Kilgour, B.  Road salt application in Highland Creek watershed, Toronto, Ontario – chloride mass balance. Water Qual. Res. J. Can. 45, 451–461.

Posthuma, L., Suter, G. W. & Traas, T. P.  Species Sensitivity Distributions in Ecotoxicology. Lewis Publishers, Boca Raton.

Ramakrishna, D. M. & Viraraghavan, T.  Environmental impact of chemical deicers – a review. Water Air Soil Pollut.

166, 49–63.

Rubin, J., Garder, P. E., Morris, C. E., Nichols, K. L., Peckenham, J. L., McKee, P., Stern, A. & Johnson, T. O.  Maine Winter Roads: Salt, safety, environment and cost, a report by the Margaret Chase Smith Policy Center. The University of Maine.

Water Quality Research Journal of Canada

|

49.1

|

2014

Rutherford, J. C. & Kefford, B. J.  Effects of salinity on stream ecosystems: Improving models for macroinvertebrates.

CSIRO Land and Water Technical Report 22/05.

Stanfield, L. W. & Kilgour, B. W.  Effects of percent impervious cover on fish and benthos assemblages and instream habitats in Lake Ontario. In: Influences of Landscape on Stream Habitats and Biological Assemblages, R. M. Hughes, L. Wang & P. W. Seelbach (eds). American Fisheries Society Symposium, Bethesda, MD, 48, 577–599.

Stone, M., Emelko, M. B., Masalek, J., Price, J. S., Rudolph, D. L., Saini, H. & Tighe, S. L.  Assessing the efficacy of current road salt management programs. Report by the University of Waterloo and the National Water Research Institute to the Ontario Ministry of the Environment and the Salt Institute.

Stone, M. & Marsalek, J.  Adoption of best practices for the environmental management of road salt in Ontario. Water Qual. Res. J. Can. 46, 174–182.

Todd, A., Kaltenecker, G. & Sunderani, S.  Chloride concentrations in Ontario’s streams. 1st International Conference on Urban Design and Road Salt Management in Cold Climates, University of Waterloo, May 26, 2009.

USEPA (United States Environmental Protection Agency)  Ambient Water Quality Criteria for Chlorides. Prepared by the Office of Water, Regulations and Standards Criteria and Standards Division, Washington, DC. Wenck Associates Inc.  Shingle Creek Chloride TMDL Report. Prepared for Shingle Creek Water Management Commission and the Minnesota Pollution Control Agency, Wenck Associates Inc., Maple Plain, MN.

First received 15 April 2012; accepted in revised form 2 December 2012. Available online 27 August 2013

53

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Comparison of CANWET and HSPF for water budget and water quality modeling in rural Ontario Syed I. Ahmed, Amanjot Singh, Ramesh Rudra and Bahram Gharabaghi

ABSTRACT This study comparatively evaluates the Hydrological Simulation Program-FORTRAN (HSPF) model and the Canadian ArcView Nutrient and Water Evaluation Tool (CANWET) for non-point source pollution (NPS) management in rural Ontario watersheds. Both models were calibrated, validated, and applied to a 52 km2 headwater rural watershed known as the Canagagigue Creek near Elmira in the Grand River basin, Ontario, Canada. A comparison of the simulated and observed values for stream flow, surface runoff, subsurface runoff, evapotranspiration, and sediment yield showed that (Better Assessment Science Integrating Point and Nonpoint Sources) BASINS/HSPF and CANWET models have similar capabilities to simulate various hydrological processes at the watershed scale. The seasonal stream flow comparison between observed and simulated values from HSPF and CANWET showed Nash-Sutcliffe efficiency (Nash-E) coefficients of 0.80 and 0.72, respectively. The monthly

Syed I. Ahmed (corresponding author) Ramesh Rudra Bahram Gharabaghi School of Engineering, University of Guelph, Thornborough Building, 50 Stone Road, Guelph, Ontario, N1G 2W1, Canada E-mail: sahmed@uoguelph.ca Amanjot Singh Credit Valley Conservation Authority, 1255 Old Derry Road, Mississauga, Ontario, L5N 6R4, Canada

comparison between the simulated and observed stream flow yielded Nash-E coefficients of 0.88 and 0.94 for HSPF and CANWET, respectively. Overall, both models predicted the components of the annual, seasonal, and monthly water budget accurately. There was a considerable difference in the monthly simulated sediment yield by both models. This difference is consistent with the surface runoff variation predicted by both models. Both models predicted sediment yield with early winter and spring storms which is typical for southern Ontario.

Key words

| modeling, sediment, water budget, water quantity

INTRODUCTION In the last few years, researchers have attempted to address

In the last few decades, the modeling approach has

the problem of non-point source (NPS) pollution by combin-

become more common to address the issue of water

ing the relationship between land management practices

resources pollution from NPSs. Field monitoring studies

and water quality degradation. Under Tier 1 of the Ontario

have been limited due to their requirements of high

Source Water Protection Act by Ministry of Environment

financial and time investment. Hydrologic modeling

(MOE), Ontario, conservation authorities, and other govern-

approaches have been proven to be more versatile, as

ment agencies are currently involved in assessment of water

they have been effectively used to simulate a variety of

budget at watershed scale to quantify drinking water sources

environmental conditions and soil-water management

(MOE ). Two systems, the surface water system and

practices needed to prevent water quality degradation

groundwater system, are being analyzed and different tools

(Saleh & Du ; Wang & Linker ; Walton &

have been researched for quantifying elements of both sys-

Hunter ). Some of the widely used watershed scale

tems individually and in integration. Understanding the

models are: AnnAGNPS (Bingner & Theurer ), Soil

environmental conditions and watershed hydrological

and Water Assessment Tool (SWAT) (Arnold et al. ),

characteristics is the first step to quantify and protect the

Hydrological

water resources of the area.

(Bicknell et al. ), and Canadian ArcView Nutrient

doi: 10.2166/wqrjc.2013.044

Simulation

Program-FORTRAN

(HSPF)

54

S. I. Ahmed et al.

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

and Water Evaluation Tool (CANWET) (Anonymous

interactions (Fontaine & Jacomino ; Whittemore &

).

Beebe ). HSPF is also applicable to many watersheds

The two models, HSPF and CANWET, were selected for

and the adjustment of various parameters is possible for

comparison as well as application of these models for cli-

various climatic and site-specific conditions (Carrubba

matic conditions in Ontario. HSPF is a complex model for

). A study by Chung et al. () was conducted to

hydrologic and water quality simulations; whereas the

evaluate the effects of climate change and urbanization

CANWET model is considered as a relatively simple

on hydrologic behavior and water quality using the HSPF

model for hydrologic processes and water pollutants. For

model in Anyangcheon watershed in Korea. The model

example, HSPF calculates the surface runoff using hourly

was able to simulate the significant effect of urbanization

time step as a function of infiltration computed using Phi-

on water quality as compared to low flow conditions;

lip’s equation (Philip ), and uses a storage routing

whereas climate change has a major effect on stream

technique to transport water from one reach to the next

flow rate.

during stream processes. The CANWET model provides a

BASINS (Better Assessment Science Integrating Point

continuous stream flow simulation using daily time steps

and Nonpoint Sources) (US Environmental Protection

using climatic data for water budget. It calculates the surface

Agency (USEPA) ) integrates GIS (Geographic Infor-

runoff by the Soil Conservation Services (SCS) Curve

mation System), data analysis, and a modeling system to

Number (CN) method (United States Department of Agri-

support watershed based analysis and TMDL development

culture (USDA)-SCS ).

(Lahlou et al. ). HSPF/BASINS have also been used

The surface water hydrology and sediment simulations

as prediction tools for in-stream bacterial concentration in

of both the models are distributed based on multiple land

the watershed by Paul et al. (). A sensitivity analysis

use/cover scenarios; however, each land use area is con-

to determine the parameters that most influence the

sidered as homogenous in regard to various attributes

coliform predictions showed that maximum storage of coli-

considered by the model. These two models do not spatially

form bacteria over the pervious land segment and amount

distribute the source areas, but simply aggregate the loads

of surface runoff affected the bacterial concentration. This

from each area into a watershed total at the specified outlet.

study emphasized the importance of parameterization of

Both models operate under different complexities and

sensitive parameters to obtain better results (Geurink &

are currently being used by various agencies and consultants

Ross ). Also, the method used for the calibration of

in Ontario to address water budget issues for source water

HSPF is important as reported by Gutierrez-Magness &

protection.

McCuen ().

HSPF is a comprehensive, conceptual, and continuous

The CANWET model is the Canadian version of the

watershed model designed for simulation of watershed

Generalized Watershed Loading Function (GWLF) model

hydrology and water quality. It is an analytical tool with

developed by Haith et al. (). The model uses the concept

application in planning, designing, and operating of water

of HSPF in a simplified form (Haith ) as well as a land

resources systems (Bicknell et al. ). HSPF has also

use based approach for hydrologic simulations with limited

been extensively used for hydrological and total maximum

parameters, making the application of the model simple.

daily load (TMDL) modeling (Al-Abed & Whiteley ;

The CANWET/GWLF models have been used for hydrol-

Singh et al. ). HSPF has generally been used to

ogy and NPS pollution simulations from watersheds

assess the effects of land-use change, reservoir operations,

(Benham et al. ; ShuKuang et al. ). A theoretical

point or NPS treatment alternatives, flow diversions, etc.

comparison of a number of watershed models including

(Van Liew et al. ; Saleh & Du ; Mishra

HSPF and GWLF was done by Borah & Bera (). They

et al. ). It is the only comprehensive model of

concluded that the HSPF model is more comprehensive

watershed hydrology and water quality that allows the

compared to CANWET, involves a large number of

integrated simulation of land and soil contaminant runoff

variables, and uses complex approaches in simulating

processes with in-stream hydraulic and sediment–chemical

watershed hydrology and pollutants.

55

S. I. Ahmed et al.

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Saleh & Du () evaluated and compared HSPF and

A recent study by Singh et al. () using the CANWET

SWAT models within BASIN 4.0 using daily and monthly

model on the Canagagigue Creek watershed stated that the

measured flow, sediment, and nutrient loading for a water-

hydrologic processes and water budget components need

shed in Texas. The results showed that SWAT simulated

more detailed and accurate calculation. However, the

average daily flow, sediment, and nutrient loading better

CANWET model was found to be a useful tool to simulate

than HSPF when compared with the measured values for

hydrologic processes at watershed scale on an annual

the calibration and verification periods. However, HSPF

(R 2 ¼ 0.89), seasonal (R 2 ¼ 0.68), and monthly (R 2 ¼ 0.68)

was found to be a better predictor of temporal variations

basis.

of daily flow and sediment. HSPF under-predicted nutrient

Now with the availability of models with varied com-

loads due to its limited ability to incorporate detailed infor-

plexities in use and approach, the question of how much

mation regarding farm management.

complexity do we need for water budgeting and NPS

The performances of HSPF and SWAT were also com-

pollution assessment arises? The present study was

pared by Van Liew et al. () for eight nested

conducted to compare the hydrology and sediment

watersheds in southern Oklahoma, to assess simulated

components of a complex model (HSPF) and a relatively

stream flow under various climatic conditions. This study

simple model (CANWET) to describe water budget and

showed that HSPF performed better on the watersheds

sediment yield, and to identify the strengths and weak-

used for calibration and SWAT produced better results on

nesses of each model.

the validation watersheds when compared with the

In this study, the integrated watershed models BASINS

observed flow data. Overall, SWAT simulated better results

4.0/HSPF and CANWET were applied to simulate the flow

( 1.3%) for stream flow volumes than HSPF ( 8.2%). The

and sediment transport for the Canagagigue Creek water-

difference in the simulated results by SWAT and HSPF

shed in southern Ontario. Basic hydrologic units in the

was attributed to approaches used for the simulation of

models were constructed from the combination of land

the runoff process.

use groups, soil type, and hydrologic characteristics. The

HSPF presents volume-dependent discharge from a

main objective of this study was to provide a watershed-

stream based on stage, surface area, volume, and discharge

modeling framework for estimating hydrological processes

entries in a function table (FTABLE). A study was con-

and sediment loads from watershed, and to compare results

ducted by Staley et al. () to evaluate the flow

from both models to provide the reductions needed to meet

predictions of HSPF in FTABLEs using computer-based

the NPS pollution standard in Ontario climatic conditions.

data and field measurements for Pigg River watershed in southern Virginia. The simulated FTABLEs showed similarities in stage-discharge curves for the bankfull range and

MATERIALS AND METHODS

differences at floodplain depths. In addition, use of field data did not show any improvement in the simulated aver-

The watershed for this study is located in the Grand River

age daily outflows as compared to use of the computer-

basin in Ontario, Canada. The Grand River starts from

based data. This study suggests that computer-based data

south of Georgian Bay and ends in Lake Erie, winding

can be used for similar conditions without loss of accuracy;

300 km through heartland of southern Ontario between

however, it was also emphasized that it should be deter-

longitude 79 300 W, and 80 570 W and latitude 42 510 and

mined if use of field data could improve the accuracy for

44 130 N. The draining area of Grand River and its tribu-

various conditions and watershed sizes. Shenk et al. ()

taries

also found that application of the HSPF model to simulate

watershed selected for this study is a subwatershed of

W

W

W

W

is

about

6,965 km2.

The

Canagagigue

Creek

nutrient and sediment load poses challenges for efficient

Grand River. The upper Canagagigue Creek watershed,

simulation of Best Management Practices (BMPs) in large-

upstream of the Floradale reservoir, is located between

scale watersheds, and can be improved by combined upgrad-

43 360 N–43 420 N latitude and 80 330 W–80 380 longitude,

ing of segmentation, input data, and model calibration.

and drains approximately 52 km2 (Figure 1). About 79.4%

W

W

W

W

56

S. I. Ahmed et al.

Figure 1

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Map of study area at the Canagagigue Creek watershed, Ontario.

of the watershed area is under agriculture land use, 11.3%

point in a watershed. HSPF simulates three types of sedi-

forest and wetlands, 9% pasture, 0.2% built up, and 0.1%

ment classes (sand, silt, and clay), water quality processes

is open water. About 95% of the area has slope less than

on pervious and impervious land surfaces, in the streams,

5%. The dominant soil in the watershed is categorized as

and well-mixed impoundments (Donigian et al. ).

silt and silt loam.

HSPF also simulates surface runoff, interflow, base flow,

The two models, HSPF and CANWET, selected for com-

snowpack

depth,

snowmelt,

evapotranspiration

(ET),

parison of water budget analysis and sediment simulations

groundwater recharge,

were evaluated for the Canagagigue Creek watershed. The

oxygen demand (BOD), pesticides, pH, ammonia, nitrate-

dissolved oxygen, biochemical

surface water loadings for both models are distributed in

nitrite, organic nitrogen, phosphorus, and sediment detach-

the sense that they allow multiple land use/cover scenarios,

ment and transport.

but each area is assumed to be homogenous in regard to var-

The HSPF has three main modules, PERLAND, IMP-

ious attributes considered by the model. Additionally, the

LAND, and RCHRES. PERLAND represents permeable

models do not spatially distribute the source areas, but

(non-urban agricultural and forest) lands, IMPLAND

simply aggregate the loads from each area into a watershed

impermeable

total. The brief description of models on hydrologic and

streams/rivers in the watershed. In the HSPF, the surface

sediment simulation approaches is presented in the follow-

runoff is estimated using an hourly time step and the Philip’s

ing sections.

infiltration equation. The model uses a storage routing tech-

(urban/developed)

lands,

and

RCHRES

nique to route water from one reach to the next during Hydrological Simulation Program-FORTRAN (HSPF)

stream processes. For the sediment simulations, it uses the model developed by Negev ().

HSPF uses information such as the time history of rainfall,

The BASINS has made HSPF input sequences easier

temperature, evaporation, and parameters related to land

and provided advanced interaction with input sequence

use patterns, soil characteristics, and agricultural practices

and graphical representation of the output. WinHSPF assists

to simulate watershed processes. The result of the simu-

the user in the building of a user control input (UCI) file

lation is a time history of runoff rate, sediment load, and

from GIS data, especially from the BASINS system. Also,

nutrient and organic chemical concentrations at any

HSPF may be run from within WinHSPF and input

57

S. I. Ahmed et al.

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

sequences may be modified, thus creating simulation scen-

Canadian ArcView Nutrient and Water Evaluation Tool

arios. WinHSPF also assists the user in building the

(CANWET)

necessary dataset for hydrologic calibration using the United States Geological Survey (USGS) Expert System

The CANWET (Anonymous ) Version 3.0 is a GIS-

(HSPEXP). WDMUtil, a tool built inside BASINS, is used

based model and has two main modules, Rural and Urban.

to prepare input data for WinHSPF and manages WDM

The Rural module is similar to the PERLAND module of

files which contain input and output time-series data

HSPF and the Urban is similar to the IMPLAND. The

needed for HSPF. For the multiple runs, HSPF requires

CANWET model provides a continuous stream flow simu-

manual tracking of time-series datasets from all scenarios

lation using daily time steps for weather data and water

at multiple locations for several constituents. GenScn (Gen-

balance calculations. It uses SCS CN approach for surface

eration and analysis of model simulation Scenarios),

runoff generation and the universal soil loss equation

developed by the USGS, helps to facilitate the display and

(USLE) (Wischmeier & Smith ) for soil erosion calcu-

interpretation of output data derived from model appli-

lations. Monthly calculations are made for sediment and

cations (Kittle et al. ). GenScn also provides advanced

nutrient loads based on the sum of daily water balance

interaction with the HSPF input sequence and integrated

values for a given month. The sediment yield is computed

analysis capabilities.

by using an area based delivery ratio approach. Spatial rout-

HSPF requires six meteorological time-series datasets to

ing is not available in Version 3.0 of CANWET. For

simulate stream flow. These include precipitation, minimum

subsurface loading, the model acts as a lumped parameter

and maximum air temperature, dew point temperature,

model using a daily water balance methodology. Daily

evaporation, wind speed, and solar radiation. In this study,

water budgets are computed for the unsaturated zone, as

the input time series were developed for 9 years (1991–

well as the saturated subsurface zone, where infiltration is

1999). Recorded daily precipitation, air temperature, and

simply computed as the difference between precipitation

wind speed data were obtained from the Grand River Con-

and snowmelt minus surface runoff and ET. ET is deter-

servation Authority (GRCA), Ontario. The potential ET

mined by the Hamon method (), using daily weather

time series provides a maximum limit for the ET demand.

data and a cover factor dependent upon land use.

The ability of the model to simulate stream flow is limited

Version 3.0 of the CANWET model allows monthly and

by the accuracy of both precipitation and ET time-series

seasonal variation in CN, ET coefficient, recession coeffi-

inputs.

cient, and seepage coefficient. The seepage coefficient

For the application of CANWET and HSPF the Canaga-

provides the possibilities of discharge and recharge from

gigue Creek watershed, upstream of the Floradale reservoir

neighboring aquifers. This version can also be run for a

and downstream of the confluence of the two tributaries

single basin or aggregated basin simulation; however, spatial

(draining eastern and western parts of the watershed) was

routing is not yet available.

delineated and discretized into sub-basins using the automatic delineation tool available in the EPA-BASINS

Model calibration

(United States Environmental Protection Agency (USEPA) ). The three GIS layers, 10 m resolution digital elevation

Prior to the application of a hydrological simulation model

data, land use grid layer, and soils grid layer were obtained

to a specific area, calibration of the model is very important

from GRCA. A 100 ha threshold area was selected for

because most of the hydrological models are developed

stream definition. The land uses were classified into agricul-

mainly from statistical analysis of hydrologic/pollutants

ture, hay/pasture, forest, and urban. The calibration

data collected from a specific location and they may not

parameters for both the models and the possible range

effectively perform for other regions. Existing land use, soil

were adopted from the previous studies completed on

conditions, and management practices significantly affect

these models (Haith et al. ; USEPA ; Anonymous

the hydrology and soil erosion rates of an area (Drohan

).

et al. ; Kaleita et al. ). Also, long-term simulations

58

S. I. Ahmed et al.

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

of hydrologic models have been helpful in evaluation of

controls the availability of sediment on the land surface.

management practices as they capture the temporal variabil-

KRER is related to the erodibility of soil type, soil surface

ity of NPS loads by incorporating a wide range of climate

conditions, and to the K factor in the USLE.

and land conditions (Borah & Bera ).

Calibration is an iterative procedure of parameter evalu-

CANWET

ation and refinement. Good calibration should result in parameter values that are within an acceptable range for

The model was set up for the upper Canagagigue Creek

the watershed and climatic conditions and produce the

watershed in such a way that the basic parameters (e.g.,

best overall agreement between simulated and observed

CN) were not supplied directly and the values extracted by

output response. HSPF and CANWET calibrations were

the model’s ArcView interface were used. However, the

accomplished by adjusting sensitive input parameters to

adjustment factors associated with these parameters were

reproduce realistic watershed behavior well enough to

then fine-tuned to mimic temporal variability in the par-

meet the modeling objectives.

ameters for the study watershed. The key parameter associated with surface runoff estimation and used by the

HSPF

model for the considered land uses was CN. The model extracted CN values from the three GIS layers using the Arc-

The main dominant parameters governing the water balance

View interface provided with the model, and the values were

are LZSN (lower zone soil moisture storage), INFILT (index

75, 63, 34, and 80 for crop land, hay/pasture, forest, and

to infiltration capacity), and LZETP (lower zone ET). If the

urban land, respectively.

amount of percolation to groundwater is small, actual ET is

CANWET also has an option that allows the user to

adjusted to change the runoff component of the water bal-

adjust CN values monthly to accommodate seasonal effects.

ance. LZSN represents the primary soil moisture storage

Such adjustments are essential for proper simulation of

and root zone in the soil profile, and the major portion of

watershed hydrologic behavior in Ontario. For example,

actual ET occurs from the LZSN. Increasing LZSN resulted

frozen soil conditions, prevalent during winter seasons,

in increased actual ET and decreased surface runoff. There-

and freeze-thaw cycles, common during spring seasons,

fore, LZSN is a sensitive parameter for estimation of annual

tend to reflect very poorly drained conditions and can gener-

water balance. INFILT governs the division of precipitation

ate significant runoff. On the other hand, the soils drain and

into upper zone soil moisture storage (UZSN), LZSN, and

dry out significantly during summer and early fall periods

to the groundwater. An increase in INFILT resulted in an

and produce little or no surface runoff. In various modeling

increase in water infiltrating into the lower zone and

approaches, using an infiltration approach to estimate sur-

groundwater zone resulting in a decrease in surface runoff,

face runoff, the seasonal conditions have been represented

and an increase in actual LZSN storage and ET. Therefore,

by temporally varying saturated hydraulic conductivity

INFILT is one of the most sensitive parameters for cali-

(Schroeter ; Al-Abed & Whiteley ). A similar

bration of HSPF for water balance and shape of hydrograph.

approach has been adopted in CANWET for the monthly

LZETP was another important parameter evaluated

and seasonal adjustment of CN values. The monthly CN

carefully for calibration. An increase in LZETP resulted in

adjustment factors used in this study to represent seasonal

an increase in actual ET from the lower zone; therefore,

changes in CN are presented in Table 1.

any change of LZETP will result in a corresponding

A groundwater recession coefficient is used in CANWET

change in runoff amount. INTFW, the Interflow parameter,

to estimate the subsurface (groundwater) contribution to

has minimum effect on runoff amount but it has significant

stream flow. This coefficient allows the removal of a fraction

impact on the shape of the hydrograph. An increase in the

of water available in saturated storage. The recession

INTFW value resulted in a decrease in peak flow and pro-

coefficient was also varied seasonally, as shown in Table 1.

longed recession of the hydrograph. KRER (coefficient in

The model also provides an option to vary groundwater

the soil detachment equation) is the major parameter that

seepage coefficient monthly for simulating the possible

59

S. I. Ahmed et al.

Table 1

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

Temporal variation of adjustment factors for the CANWET model

|

49.1

|

2014

Graphical and statistical procedures applied in this calibration process include the following:

Adjustment factors

1. Time-series plot of observed and simulated values for

Month

ET

CN

GW recession

April

1.0

1.09

0.04

May

1.0

0.9

0.01

2. Observed vs. simulated scatter plots, with a 45 linear

June

1.4

0.8

0.01

July

1.4

0.8

0.01

regression line displayed, for values of stream flow. Scat-

August

1.4

0.8

0.01

along with the slope and intercept of the linear regression

September

1.4

0.8

0.02

line; thus the graphical and statistical assessments are

October

1.1

0.85

0.03

combined.

November

1.0

1.05

0.04

December

1.0

1.08

0.04

January

1.0

1.08

0.04

February

1.0

1.08

0.04

March

1.0

1.08

0.04

flow. The plot visually evaluates the agreement between the simulated and observed values.

W

ter plots include calculation of a correlation coefficient,

3. Percent error difference (monthly and annual) for the observed and simulated stream flow.

The outputs of the calibrated models were compared for the water budget components, stream flow, and total suspended sediment load (TSS) on annual, seasonal, monthly, and daily time frames. The annual analysis was based on

movement to or from a deep aquifer system. In this study,

the year starting from December of the previous year to

monthly values of groundwater seepage coefficient were set

November of the current year. For the seasonal analysis,

to zero for all the months other than November and December

the year was divided into four seasons: Spring from March

( 0.3) to account for the threshold base flow in the stream.

1st to May 31st, Summer from June 1st to September 30th,

The ET was also adjusted temporally to represent realis-

Fall from October 1st to November 30th, and Winter from

tic ET conditions for the study area as shown in Table 1.

December 1st to February 28th or 29th. The comparative

Based on the soil texture and depth of soil profile, the

evaluation used both statistical and graphical approaches.

value of unsaturated available water storage was assumed

Statistical tests include the Nash-Sutcliffe efficiency coeffi-

to be 10 cm. The initial saturation on April 1 (the starting

cient (Nash-E) (Nash & Sutcliffe ) and correlation

date for the simulation) was estimated to be 3 cm. The begin-

coefficient (R 2), and the graphical approach used a scatter

ning of April is generally snow melting time and the soil

diagram. The annual results were also compared using a per-

conditions are close to field capacity.

cent error method.

Analysis

RESULTS AND DISCUSSION Calibration of flow and sediment components was undertaken by adjusting certain model parameters to obtain an

Comparison for annual water budget and stream flow

agreement of within ±10%, as suggested in the BASINS (USEPA ), between observed and simulated output

The comparison of averaged annual components of water

responses. In this study, the calibration focused on the com-

budget (surface runoff, subsurface runoff, and ET) simulated

parison between observed and simulated daily, monthly, and

by HSPF and CANWET are shown in Figure 2. The average

annual stream flow. All of these comparisons were per-

annual rainfall for the 9-year study period (1991–1999) was

formed for a proper calibration of hydrology and sediment

820 mm. The wettest year was 1992 (1,036 mm), and the

parameters. The available observed flow data include con-

driest year was 1998 (549 mm). The HSPF and CANWET

tinuous flow records at the gauge sites for the entire time

simulated water budget was very similar to the observed

period.

data. The models’ predicted annual ET values were also

60

S. I. Ahmed et al.

Figure 2

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Average annual water budget simulated by HSPF and CANWET (1991–1999) for Canagagigue watershed, Ontario.

very similar to the observed ET values. However, HSPF

over predicted annual stream flow (1992) or under predicted

simulated surface runoff was 21% of the precipitation as

(1998 and 1999) during wet years. The models slightly over-

compared to 12% by CANWET. Also, CANWET simulated

predicted (CANWET 10% and HSPF 19%) the stream flow

28% subsurface flow (interflow þ groundwater flow) as com-

in the wettest year (1992). However, these models did well

pared to 20% by HSPF. Overall, the outputs of both models

in predicting the stream flow for the driest year (1998) of

revealed that the water budget components were compar-

the study period. In addition, for the driest year CANWET

able with the observed data. The results reported by

simulated (230 mm) annual stream flow was very close to

Dickinson & Rudra () also found similar division of

the observed annual stream flow (228 mm); and HSPF simu-

rainfall into ET, surface runoff, and subsurface from the

lated 219 mm of annual stream flow.

watersheds with medium textured soils in Ontario.

The comparison between the annual observed and simu-

Figure 3 shows the comparison of the annual stream

lated stream flow yielded a Nash-E of 0.77 for HSPF and

flow simulated by HSPF and CANWET with the observed

0.82 for CANWET. Also, the determination coefficient

stream flow. These data show that the stream flow simulated

between the annual simulated and observed daily stream

by both the models compares well with the observed stream

flow was 0.77 for HSPF and 0.83 for CANWET. These

flow over the 9-year period (1991–1999). When averaged

data indicate that models’ results are similar for the simu-

over the 9-year period, HSPF overestimated the annual

lation of water budget and stream flow; however, the

stream flow by 2.5% and CANWET underestimated by

simulation results on an annual basis were slightly better

1.3%. This indicates that the models are similar in simulat-

for the CANWET model than for the HSPF model.

ing average annual stream flow. Both the models either Comparison for seasonal water budget and stream flow Figure 4 presents the comparison of average seasonal water budget simulated by both models. The data given in Figures 4(a) and 4(b) clearly show that during the winter period HSPF predicted higher surface runoff (8.1 cm) and ET (2.2 cm) than that predicted by CANWET, surface runoff (4.2 cm) and ET (0.5 cm). Overall, CANWET predicted higher groundwater recharge for all the seasons than simulated by HSPF except for summer. Also, CANWET predicted a higher amount of ET (35.5 cm) than Figure 3

|

Comparison of annual and averaged annual observed stream flow with stream flow simulated by HSPF and CANWET.

HSPF (27.1 cm) during summer. For the fall season, both models have similar predictions. Figures 4(c) and 4(d)

61

Figure 4

S. I. Ahmed et al.

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Comparison of seasonal water budget simulated by HSPF and CANWET (1991–1999).

describe the percentage contribution of various components

flow and surface runoff predicted by these models were typi-

of the water budget simulated by HSPF and CANWET

cal for southern Ontario.

during various seasons. The amount of averaged annual

To further evaluate the performance of the HSPF and

ET simulated by HSPF (467 mm) and CANWET (499 mm)

CANWET models, simulated seasonal stream flows were

compared well with the ET value (500–567 mm) available

compared with the observed seasonal stream flow for the

in literature for southern Ontario conditions (Dickinson &

entire simulation period (Figure 5). This comparison gener-

Diiwu ; Rudra et al. ; McLaughlin ). These

ated Nash-E of 0.80 for HSPF and 0.72 for CANWET.

comparisons suggest that both models slightly underesti-

This indicated that both models showed similar perform-

mated actual ET. The overall analysis of the seasonal

ance for simulation of seasonal hydrology in Ontario

water budget shows good agreement among ET, surface

conditions. These data also show that both models did not

runoff, and subsurface runoff simulated by both models.

exhibit a consistent pattern for the winter season. For most

Comparison of groundwater flow simulated by these

of the years, HSPF under-predicted the stream flow.

models for various seasons shows that groundwater flow

CANWET performed better than HSPF for the prediction

was a major part of the water budget in spring (32.7%)

of stream flow values for spring and summer seasons.

and summer (35.9%) for HSPF (Figures 4(c) and 4(d)).

During wet summer seasons (1992 and 1993), HSPF over-

CANWET-simulated groundwater flow showed significant

predicted and CANWET slightly under-predicted the

contribution during winter (28.1%) and spring (43.2%).

stream flow which affected the overall performance of the

The surface runoff values of the HSPF and CANWET

models for the summer season. For the fall season, the pre-

models show that the summer season was predicted better

dicted stream flow by HSPF matched better with the

by HSPF (8.9%) as compared to the CANWET simulated

observed stream flow compared to the stream flow predicted

value (0.4%). Also, HSPF showed a consistent trend in pre-

by CANWET.

dicting the surface runoff for the study period (Figure 4(c)).

To evaluate the overall seasonal performance of these

Overall, spring and winter contributions of groundwater

models, averaged over the entire study period, predicted

62

S. I. Ahmed et al.

Figure 5

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Comparison of seasonal observed and predicted stream flows simulated by HSPF and CANWET (1991–1999); W ¼ December, January, and February; Sp ¼ March–May; Su ¼ June–September; F ¼ October–November.

and observed stream flows are shown in Table 2. These data

coefficient (R 2) of 0.99 and 0.94 between the simulated

show that the percent difference between averaged observed

and observed CANWET and HSPF simulated flows, respect-

and predicted stream flow were 2.2 and 3.9 for CANWET

ively, support the better performance of CANWET. The

and HSPF, respectively, indicating that both models have

observations in southern Ontario reveal that most of the

similar seasonal stream flow prediction capabilities. These

stream flow during summer results from the subsurface

values also show that CANWET did better on a seasonal

(groundwater) contribution. Therefore, the observations

basis than HSPF. In addition, HSPF could not adequately

imply that CANWET performed more realistically in simu-

simulate stream flow during the spring ( 29.3%) and

lating the subsurface flow in spring and summer than HSPF.

summer (47.4%) seasons. This could be due to the higher ET in spring and lower ET in summer simulated by HSPF

Comparison of monthly water budget and stream flow

as shown in Figure 4(c). However, HSPF performed better than CANWET for simulation of stream flow for fall

The performance of HSPF and CANWET was also evalu-

(0.5%) and winter ( 2.9%) periods. Overall, CANWET per-

ated for water budget components on a monthly basis

formed better than the HSPF for the simulation of stream

from 1991 to 1999. However, the data shown in Figure 6(a)

flow. The Nash-E between the average observed and simu-

and 6(b) are for only 2 years (1991–1992). The analyses of

lated stream flow by the CANWET and HSPF models was

monthly results indicate that ET is at minimum during

0.99 and 0.92, respectively. Also, the determination

winter and fall months; and monthly simulated ET varied

Table 2

|

Seasonal analysis of HSPF and CANWET models for the study period (1991–1999) Stream flow (cm)

% Difference

Seasona

Precipitation (cm)

Observed

CANWET

HSPF

Winter

18.65

11.41

10.76

11.08

6.03

2.96

Spring

18.57

15.14

15.55

11.70

2.64

29.35

CANWET

HSPF

Summer

32.38

4.16

4.11

7.92

1.27

47.46

Fall

12.38

3.23

3.10

3.25

4.17

0.55

2.21

3.92

Overall a

Winter ¼ December, January, and February; Spring ¼ March–May; Summer ¼ June–September; Fall ¼ October–November.

63

S. I. Ahmed et al.

Figure 6

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Comparison of water budget on a monthly basis simulated by HSPF (a) and CANWET (b) (1991–1999).

from 0 to 3.9 cm during these months. However, ET started

For the fall months (October–November) water budget data

accumulating more during spring and reached up to 6.4 cm

show that CANWET predicted consistently higher ground-

in May 1996 by HSPF. CANWET also showed up to 9.5 cm

water flow than the HSPF simulated groundwater flow.

ET in May 1991 (a wet month, above average rainfall). As

However, HSPF predicted some decent contributions in

anticipated, summer months were found to be the highest

these months for the above average rainfall years (1991,

ET producing months for both models. HSPF simulated

1992, and 1996) (Figures 3, 6(a) and 6(b)).

peak ET (9.8 cm) in August 1991 and CANWET gave the

Analysis of HSPF simulated surface runoff showed a

peak ET values of 6.7–17.1 cm in July 1993. The trends for

general trend of higher flows during January (1.5–5.4 cm)

ET were found to be similar for both models during wet

and March–April (0.1–10.1 cm) (Figures 6(a) and 6(b)).

and dry months; however, CANWET predicted higher ET

The CANWET model also produced a similar pattern of sur-

during summer months when compared with the predicted

face runoff in these months; however, CANWET simulated

values from HSPF for these months.

surface runoff values were comparatively lower than the

The trends of simulated groundwater contribution to stream flow by CANWET and HSPF were similar and realis-

HSPF simulated values.

An analysis of the wettest and driest years in the study

groundwater

period indicated that these models successfully simulated

contribution in November, December, March, and April,

the water budget components (Figure 6). For the year

and minimum contribution during summer months. The

1992 (wet year), HSPF predicted a typical partition of

range of groundwater flow for summer months was 0.1–

precipitation into ET, groundwater, and surface runoff contri-

3.8 cm for HSPF and 0.2–3.5 cm for CANWET, respectively.

bution of southern Ontario. In the case of CANWET, water

tic.

Both

models

produced

maximum

64

S. I. Ahmed et al.

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

budget was more dependent on ET and groundwater contri-

and February). CANWET over-predicted ET by approxi-

bution. For the dry year of 1998, both models showed

mately 12% more than the ET simulated by HSPF during

similar trends as ET was the significant part of the water

summer months (June–September). Nevertheless as dis-

budget from April to September. However, there was also

cussed earlier, the annual ET simulations of both models

some contribution of groundwater flow in March and April.

are within the permissible range.

In addition, there was minimum contribution of surface runoff in all the months of this dry year other than April.

The results shown in Figure 7 also indicate that there are variations in simulated monthly subsurface flow within the

Examination of the components of the monthly water

seasons. CANWET simulated a higher amount of subsurface

budget also showed that simulated ET contributed more

flow during late winter and early spring months (March and

than 80% of the water budget during months from July

April) and less subsurface flow during summer months. This

and September; surface runoff ranging from 30 to 46% in

could be due to the adoption of the single-tank approach in

February, November, and March; groundwater flow contri-

the CANWET model. For subsurface flow computation, ET

buting from 70 to 80% in January, April, and December by

uses the moisture that is available in the tank which results

both models.

in limited moisture remaining to contribute to subsurface

The monthly comparison of surface runoff simulated by

flow, whereas HSPF uses a two-tank approach and the

HSPF and CANWET, when averaged over the years,

amount of groundwater being used for ET can be calibrated.

showed that proportionate monthly surface runoff was pro-

The comparison of the monthly stream flows simulated

duced by both models (Figures 7(a) and 7(b)); however,

by HSPF and CANWET were compared with the observed

there was discrepancy in the amount produced. These differ-

stream flow for the study period (1991–1999) (Figure 8).

ences could be due to the use of one CN in the CANWET

These results showed a good agreement between observed

model for one type of land use for the entire simulation

and simulated stream flow, and generated a Nash-E of

period. However, it is possible to change the parameters

0.70 and 0.67 for HSPF and CANWET, respectively. How-

related to the infiltration rate temporally during the simu-

ever, HSPF performed better than the CANWET because

lation period in the HSPF model. In this study, the

of more temporal control over the variables that mimic

infiltration parameters in the HSPF model were adjusted

monthly variations in the processes. Also, the correlation

to reduce the infiltration rate during winter and early

coefficients of the simulated and observed stream flows

spring periods (to accommodate frozen conditions) resulting

were found to be 0.80 and 0.70 for HSPF and CANWET,

in a higher amount of runoff during these months as shown

respectively.

in Figure 7(a).

Figure 9 shows the comparison of simulated monthly

The monthly simulated water budget averaged over the

stream flow, averaged over the 9-year study period, with

9-year study period showed similar results for both models

the observed stream flow. These results indicate that

(Figure 7) with peak ET demands during June or July and

both models did a good job in simulating stream flow.

minimal ones during winter months (December, January,

Specifically, CANWET results showed a consistency

Figure 7

|

Comparison of average monthly water budget simulated by HSPF (a) and CANWET (b) (1991–1999).

65

S. I. Ahmed et al.

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

Figure 8

|

Comparison of monthly observed and simulated stream flow by HSPF and CANWET (1991–1999).

Figure 9

|

Comparison of averaged monthly stream flow simulated by HSPF and CANWET with observed flow (1991–1999).

|

49.1

|

2014

throughout the year with the observed stream flow data.

and CANWET, respectively. The correlation coefficients

HSPF underestimated stream flow for the months of

between the simulated and observed daily stream flow

March and April, and overestimated for the summer

were also low, 0.35 for HSPF and 0.28 for CANWET. The

months. The comparison also yielded a Nash-E of 0.88

scatter diagram between daily observed stream flow and

and 0.94 when simulated monthly stream flow by HSPF

simulated daily HSPF and CANWET stream flows

and CANWET, respectively, was compared with the

(Figure 11) did not show strong relation between daily simu-

observed monthly stream flow.

lated and observed data; however, both models produced similar trends. HSPF performed reasonably due to better

Daily stream flow analysis

control of simulation of temporal variability of subsurface

An analysis was conducted for the evaluation of the models

peaks of daily stream flows were satisfactorily captured by

for the simulation of daily stream flow. The comparison

both models.

flow contribution to the stream flow. In addition, the

between daily observed and simulated stream flow by HSPF and CANWET were conducted for the 9-year simu-

Sediment simulation

lation period (1991–1999), and were also statistically analyzed. Due to the extensive amount of daily data for

The annual soil erosion by HSPF was compared with the

this analysis, only 3-years (1991–1993) of the study are

annual soil erosion simulated by USLE and multiplied by

shown in Figure 10. The Nash-E for the daily observed

the sediment delivery ratio (SDR) of 0.36 based upon the

and simulated stream flow was 0.33 and 0.17 for HSPF

model given by Renfro (1975). HSPF simulates upland

66

S. I. Ahmed et al.

Figure 10

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Comparison of daily stream flow simulated by HSPF and CANWET with observed flow (1991–1999).

close agreement with HSPF producing slightly higher sediment than CANWET. Furthermore, the monthly sediment yields simulated at the outlet by HSPF and CANWET were compared, and considerable differences were found in the monthly simulated sediment yield by both models. The difference is consistent with the surface runoff variation predicted by both models (Figure 12). In addition, HSPF has a module for in-stream sediment deposition and scouring which makes the simulation dynamic for in-stream sediment routing. However, in the absence of continuous observed data, it is not possible to predict which model is simulating better monthly sediment yield. Both models predicted sediment with early winter storms (January) and spring storms (March–May), which is typical behavior of watersheds in southern Ontario. The sediment yield predicted by both models was due to the wet conditions in the watershed which increased the contributing area during these periods. The simulated sediment yield is Figure 11

|

Scatter plot showing comparison of daily observed and simulated stream flows by HSPF and CANWET (1991–1999).

minimal during the rest of the year because of lower flows in these periods. Overall, this modeling study for sediment shows that the hydrologic behavior of the

erosion as delivered to the stream. Since CANWET uses

watershed controls the sediment transport and sediment

USLE for simulation of upland soil erosion, the erosion

yield.

output produced by CANWET was multiplied by the SDR

The HSPF model also simulates TSS on a daily basis

of the watershed (0.36) to achieve upland sediment deliv-

which was compared with the available 18 observations

ered to the stream as suggested by USEPA (USEPA ).

for the study period. Figure 13 reveals that observed data

Figure 12 shows the comparison of annual upland erosion

points fall on the non-peak flow days and therefore, give

simulated by HSPF and CANWET from 1993 to 1995. The

the base suspended sediment values in the stream. The simu-

outputs for upland sediment delivered to the stream are in

lated TSS also showed base values during the times when

67

S. I. Ahmed et al.

Figure 12

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

Comparison of observed and HSPF simulated TSS concentration (a) and load (b) for the period 1994–1995.

observed TSS data were available. Figure 13 also shows that

used in this study, it may be the reason that CANWET pre-

the peaks in TSS were simulated accordingly when there

dicted results are underestimated. Therefore, CANWET

were peaks in stream flow. Therefore, the trend for TSS is

cannot be used for TMDL evaluation whereas HSPF has

justified, whereas the quantitative verification needs further

the capability of evaluating TMDL at the watershed scale.

investigation with the observed data. Figure 14 gives the

The study has shown the ability of these models to pre-

comparison of annual erosion and sediment yield simulated

dict the hydrologic behavior and sediment yield for climatic

by HSPF and CANWET for the period 1993–1995. It is

conditions in Ontario with limited observed data. It is antici-

obvious from Figure 14(a) and (b) that HSPF simulated ero-

pated that the availability of continuous observed sediment

sion and sediment yield was higher than the erosion and

data and comparison with simulated sediment results by

sediment yield simulated by CANWET. Since CANWET

these models would produce better guidelines for manage-

does not produce daily sediment loads in the version we

ment practices in the area.

68

S. I. Ahmed et al.

Figure 13

|

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

49.1

|

2014

Comparison of monthly erosion (a) and sediment yield (b) simulated by HSPF and CANWET for the period 1993–1995.

CONCLUSIONS The study concludes the following:

•

|

of the models may be used for analysis of annual water

•

budget.

Both models simulated components of seasonal water budget compatible with observed components (Nash-E

Both HSPF and CANWET effectively partitioned annual

0.80 and 0.72 for HSPF vs. observed stream flow and

precipitation into ET, surface runoff, and subsurface flow

CANWET vs. observed stream flow, respectively). The

which is representative of the watersheds with medium

simulated groundwater flow and surface runoff by these

soils in southern Ontario conditions. Therefore, either

models also show a typical pattern for southern Ontario.

69

S. I. Ahmed et al.

|

Modeling water budget and water quality

Water Quality Research Journal of Canada

|

49.1

|

2014

be preferred over CANWET for daily simulations depend-

•

ing upon the accuracy level required.

The upland erosion simulation is comparable for both models with HSPF relying on USLE for calibration.

HSPF has a strong in-stream component for sediment routing and therefore, sediment yield prediction may be more reliable. The conclusion could not be drawn between the two studied models for the prediction of sediment yield because few observed data were available

•

to verify the models’ outputs.

HSPF needs a higher level of expertise for its application compared to that required for the application of CANWET. The number of variables controlling hydrology and sediment in HSPF outnumber the variables needed for CANWET.

ACKNOWLEDGEMENTS The study was sponsored by the Ministry of Agriculture and Food, Ontario, and Greenland International Consulting, Collingwood, Ontario, Canada, through a grant provided Figure 14

|

Comparison of annual erosion (a) and sediment yield (b) simulated by HSPF and CANWET for the period 1993–1995.

to the University of Guelph, Guelph, Ontario. The study was

Also, averaged seasonal simulated stream flow over the

accomplished

by

joint

venture

of

Greenland

International and the University of Guelph.

9-year period showed high Nash-E of 0.91 and 0.99 for HSPF and CANWET, respectively. Therefore, seasonal comparison also supports that either of the two models

•

can be used for seasonal water budgeting.

These

models

components

predicted

realistically

monthly

water

representing

budget

hydrology

of

watersheds in southern Ontario. The monthly simulated stream flow comparison with observed stream flow rendered good correlation (Nash-E ¼ 0.70 and 0.67 for HSPF

vs.

observed

stream

flow

and

CANWET

vs. observed stream flow, respectively). Apparently HSPF better predicts monthly water budgeting than CANWET as it has more temporal control of hydrologic

•

parameters.

The daily comparison of HSPF and CANWET simulated stream flow with observed stream flow showed a Nash-E of 0.33 and 0.17, respectively. Since correlation of HSPF with observed stream flow is more promising, HSPF may

REFERENCES Al-Abed, N. A. & Whiteley, H. R.  Calibration of the Hydrological Simulation Program FORTRAN (HSPF) model using automatic calibration and geographical information systems. Hydrol. Process. 16 (16), 3169–3188.

Anonymous  CANWET User’s Guide. Greenland International Consulting, Collinwood, Ontario, Canada.

Arnold, J. G., Srinivasan, R., Muttiah, R. S. & Williams, J. R.  Large area hydrologic modeling and assessment part I: model development. J. Am. Water Res. Assoc. 34 (1), 73–89.

Benham, B. L., Brannan, K. M., Yogow, G., Zeckoski, R. W., Dillaha, T. A., Mostaghimi, S. & Wynn, J. W.  Development of bacteria and benthic total maximum daily loads: a case study, Linville Creek, Virginia. J. Environ. Qual.

34 (5), 1860–1872.

Bicknell, B. R., Imhoff, J. C., Kittle Jr, J. L., Jobes, T. H. & Donigian Jr, A. S.  Hydrological Simulation Program – FORTRAN (HSPF), User’s Manual for Version 12.0. USEPA, Athens, Georgia.

70

S. I. Ahmed et al.

|

Modeling water budget and water quality

Bingner, R. L. & Theurer, F. D.  AnnAGNPS Technical Processes Documentation: Version 3.2. USDA-ARS National Sedimentation Laboratory, Oxford, Mississippi.

Borah, D. K. & Bera, M.  Watershed-scale hydrologic and non-point-source pollution models: review of mathematical bases. Trans. ASABE 46 (6), 1553–1566.

Borah, D. K. & Bera, M.  Watershed-scale hydrologic and non-point source pollution models: review of applications.

Trans. ASABE 47 (3), 789–803.

Carrubba, L.  Hydrologic modeling at the watershed scale using NPSM. J. Am. Water Res. Assoc. 36 (6), 1237–1246.

Chung, E. S., Park, K. & Lee, K. S.  The relative impacts of climate change and urbanization on the hydrological response of a Korean urban watershed. Hydrol. Process. 25, 544–560.

Dickinson, W. T. & Diiwu, J.  Water Balance Calculations in Ontario (unpublished). School of Eng., University of Guelph, Guelph, ON. Dickinson, W. T. & Rudra, R. P.  Key components of Ontario hydrography. Presentation at A.D. Latornell Conservation Symposium, Alliston, Ontario, Canada, November 15–17.

Donigian Jr, A. S., Imhoff, J. C., Bicknell, B. R. & Kittle Jr, J. L.  Application guide for Hydrological Simulation Program–FORTRAN (HSPF): EPA-600/3-84-065. ERL, Athens, GA. Drohan, P. J., Ciolkosz, E. J. & Petersen, G. W.  Soil survey mapping unit accuracy in forested field plots in Northern Pennsylvania. Soil Sci. Soc. Am. J. 67, 208–214.

Fontaine, T. A. & Jacomino, V. M. F.  Sensitivity analysis of simulated contaminated sediment transport. J. Am. Water Res. Assoc. 33 (2), 313–326.

Geurink, J. S. & Ross, M. A.  Time-step dependency of infiltration errors in the HSPF model. J. Hydrol. Eng. 11 (4), 296–305.

Gutierrez-Magness, L. & McCuen, R. H.  Effect of flow proportions on HSPF model calibration accuracy. J. Hydrol.

Eng. 10 (5), 343–352.

Haith, D. A.  Personal Communication. Cornell University, Ithaca, NY, USA. Haith, D. A., Mandel, R. & Wu, R. S.  GWLF. Generalized Watershed Loading Functions, Version 2.0. User’s Manual.

Dept. of Agric. and Bio. Eng., Cornell Univ., Ithaca, NY. Hamon, W. R.  Estimating potential evapotranspiration.

J. Hydraul. Div. ASCE 87 (HY3), 107–120.

Kaleita, A. L., Hirschi, M. C. & Tian, L. F.  Field-scale surface soil moisture patterns and their relationship to topographic indices. Trans. ASABE 50 (2), 557–564.

Kittle Jr, J. L. , Lumb, A. M., Hummel, P. R., Duda, P. B. & Gray, M. H.  A tool for the generation and analysis of model simulation scenarios for watersheds (GenScn). US Geological Survey Water Resources Investigations, Report 98–4134, 152 pp.

Lahlou, M., Shoemaker, L., Choudhury, S., Elmer, R., Hu, A., Manguerra, H. & Parker, A.  Better Assessment Science Integrating Point and Non-point Sources. BASINS, Version 2.0 User’s Manual. EPA-823-B-98-006. USEPA, Washington, DC.

Water Quality Research Journal of Canada

|

49.1

|

2014

McLaughlin, N.  Improving Nitrogen Leaching predictions within the MCLONE4 Program. MSc Thesis, University of Guelph, Guelph, Ontario.

Ministry of the Environment  Assessment Report: Draft Guidance, Module 7 Water Budget and Water Quantity Risk Assessment. Ontario Ministry of the Environment, Toronto, Ontario.

Mishra, A., Kar, S. & Singh, V. P.  Determination of runoff and sediment yield from a small watershed in subhumid subtropics using the HSPF model. Hydrol. Process. 21 (22), 3035–3045.

Nash, J. E. & Sutcliffe, J. V.  River flow forecasting through conceptual models part I – a discussion of principles.

J. Hydrol. 10 (3), 282–290.

Negev, M.  A sediment model on a digital computer.

Technical Report No. 76, Department of Civil Eng., Stanford University, Stanford, CA, p. 109.

Paul, S., Haan, P. K., Matlock, M. D., Mukhtar, S. & Pillai, S. D.  Analysis of the HSPF water quality parameter uncertainty in predicting peak in-stream fecal coliform concentrations. Trans. ASABE 47 (1), 69–78.

Philip, J. R.  The theory of infiltration: 1. The infiltration equation and its solution. Soil Sci. 83, 345–357.

Rudra, R. P., Dickinson, W. T., Whiteley, H. R., Gayner, J. D. & Wall, G. J.  Selection of an appropriate evaporation estimation technique for continuous modeling. J. Am. Water Res. Assoc. 36 (3), 585–594.

Saleh, A. & Du, B.  Evaluation of SWAT and HSPF within BASINS program for the Upper North Bosque river watershed in central Texas. Trans. ASABE 47 (4), 1039–1049.

Schroeter, H. O.  GAWSER: Guelph All-Weather SequentialEvents Runoff Model, Version 6.5, Training Guide and Reference Manual. Submitted to the Ministry of Natural Resources and the Grand River Conservation Authority, Schroeter and Associates.

Shenk, G. W., Wu, J. & Linker, L. C.  An enhanced HSPF model structure for Chesapeake Bay watershed simulation.

J. Environ. Eng. 138, 949–957.

ShuKuang, N., Chang, N. B., KaiYu, J. & YiHsing, T.  Soil erosion and non-point source pollution impacts assessment with the aid of multi-temporal remote sensing images. J. Environ. Manage. 79 (1), 88–101.

Singh, J., Knapp, H. V., Arnold, J. G. & Demissie, M.  Hydrological modeling of the Iqoquois river watershed using HSPF and SWAT. J. Am. Water Res. Assoc. 41 (2), 343–360.

Singh, A., Rudra, R. P. & Gharabaghi, B.  Evaluation of CANWET Model for Hydrologic Simulations for Upper Canagagigue Creek Watershed in Southern Ontario.

Canadian Society of Biosystems Engineering, Ontario, vol. 54.

Staley, N., Bright, T., Zeckoski, R. W., Benham, B. L. & Brannan, K. M.  A comparison of HSPF simulations using FTABLEs generated by field and computer-based data. Paper 052158, ASABE Annual International Meeting, July 17–20, Tampa, Florida.

71

S. I. Ahmed et al.

|

Modeling water budget and water quality

USDA Soil Conservation Service  National Engineering Handbook, Section 4: Hydrology, Chapters 4–10, Washington, DC, 15-7–15–11.

USEPA (US Environmental Protection Agency)  BASINS Technical Note 6: Estimating Hydrology and Hydraulic Parameters for HSPF. USEPA, Office of Water, EPA-823-R99-013, Washington, DC. USEPA  Better Assessment Science Integrating Point and Nonpoint Sources – BASINS Version 3.0, User’s Manual. USEPA, Office of Water, EPA-823-B-01-001, Washington, DC. USEPA  EPA BASINS Technical Note 8. Sediment Parameter and Calibration Guidance for HSPF. USEPA, Office of Water, 4305, Washington, DC. Van Liew, M. W., Arnold, J. G. & Garbrecht, J. D.  Hydrologic simulation on agricultural watersheds: choosing between two models. Trans. ASABE 46 (6), 1539–1551.

Water Quality Research Journal of Canada

|

49.1

|

2014

Walton, R. S. & Hunter, H. M.  Isolating the water quality responses of multiple land uses from stream monitoring data through model calibration. J. Hydrol. 378, 29–45.

Wang, P. & Linker, L. C.  A correction of DIN uptake simulation by Michaelis-Menton saturation kinetics in HSPF watershed model to improve DIN export simulation.

Environ. Model Softw. 21 (1), 45–60.

Whittemore, R. C. & Beebe, J.  EPA’s BASINS Model: good science or serendipitous modeling. J. Am. Water Res. Assoc.

36 (3), 493–499.

Wischmeier, W. H. & Smith, D. D.  Predicting Rainfall Erosion Losses from Cropland East of the Rocky Mountains: Guide for Selection of Practices for Soil and Water Conservation Planning. USDA Agriculture Handbook. US Government Printing Office, Washington, DC, p. 282.

First received 23 August 2012; accepted in revised form 12 December 2012. Available online 27 August 2013

72

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Performance of reverse osmosis and manganese greensand plants in removing naturally occurring substances in drinking water O. S. Thirunavukkarasu, T. Phommavong, Y. C. Jin and S. A. Ferris

ABSTRACT The Water Security Agency has a legislative authority to regulate water treatment systems and enforce standards with respect to drinking water quality in the Province of Saskatchewan. A number of communities in Saskatchewan which depend on groundwater as a source for drinking water have reported high levels of naturally occurring substances, such as arsenic, uranium and selenium, in their raw water. These communities continue to upgrade their systems by installing new or retrofitting with treatment units, such as reverse osmosis (RO) and manganese greensand (MGS) filters to reduce the levels of naturally occurring substances in finished water. In order to assess the treatment performance of these systems, a study was initiated to collect samples from 20 communities across Saskatchewan and analyse naturally occurring substances in raw and finished water. The study focused on the removal efficiency and the effect of parameters such as sulfate, total dissolved solids, and hardness on the removal efficiency. The paper includes discussion on the results and analysis of sampling/research studies conducted to assess the performance of treatment systems. Results showed that RO plants are effective in removing uranium and MGS are effective in

O. S. Thirunavukkarasu (corresponding author) S. A. Ferris Drinking Water & Wastewater Management Division, Water Security Agency, 420-2365 Albert Street, Regina, Saskatchewan, Canada S4P 4K1 E-mail: o.tarasu@wsask.ca T. Phommavong Municipal Branch, Saskatchewan Ministry of Environment, 3211 Albert Street, Regina, Saskatchewan, Canada S4S 5W6 Y. C. Jin Faculty of Engineering and Applied Science, University of Regina, Regina, Saskatchewan, Canada

removing arsenic from drinking water.

Key words

| arsenic, drinking water, manganese greensand, reverse osmosis, treatment, uranium

INTRODUCTION In the Province of Saskatchewan, Canada, the quality of

Saskatchewan has adopted these guidelines as legally

drinking water, the conditions of systems that produce it

enforceable standards.

and the protection of source water is one of the top priori-

Drinking water health and toxicity parameters include

ties and continues to be an important public health and

a range of naturally occurring substances (arsenic,

environmental goal. As such, ensuring safe drinking

barium, boron, lead, nitrate, selenium, uranium, etc.),

water is a shared responsibility among a number of pro-

and other substances such as trihalomethanes, which

vincial agencies and the Water Security Agency (WSA)

may be produced during chlorine-based disinfection pro-

is a lead agency in implementing the Safe Drinking

cesses. These substances represent a small potential for

Water Strategy in the province. Nearly 50 per cent of

adverse health effects over longer time periods. While

the people who live in Saskatchewan depend on ground-

the safety gains associated with eliminating microbial

water as a source for drinking water and the remaining

threats far outweighs any possible adverse health risks

population use surface water as a source. The Guidelines

associated with disinfection by-products, it is important

for Canadian Drinking Water Quality (Health Canada

to monitor and to ensure they remain within safe levels.

) are used in Canada as the definitive measure

Nearly 10 to 15 per cent of communities in Saskatchewan

of science-based safety criteria for drinking water.

who depend on groundwater as a source of drinking

doi: 10.2166/wqrjc.2013.001

73

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

Water Quality Research Journal of Canada

|

49.1

|

2014

water have reported increased levels of naturally occur-

groundwater containing arsenite (As III) is oxidized to

ring substances, such as arsenic and uranium, in their

arsenate (As V) by the addition of an oxidant such as

raw water.

KMnO4 solution and subsequently arsenate is removed

Arsenic, a potential carcinogenic element is present in

in MGS filtration systems. The MGS filtration system is

natural water systems as a result of both natural and

initially used to remove iron and manganese from

anthropogenic activities. The natural weathering processes

the water, but the iron that has been adsorbed onto the

contribute approximately 40,000 tons of arsenic to the

greensand filter is capable of removing arsenic from

global environment annually, while twice this amount is

the water. Arsenic removal efficiency is based on

being released by human activities (Paige et al. ).

the amount of iron present in the source water and

Arsenic concentrations are generally higher in ground-

studies showed that iron to arsenic ratio played an

water due to increased contact levels with these arsenic-

important role in removing arsenic in MGS filtration

containing deposits. Arsenic can, however, find its way

systems (Subramanian et al. ; Viraraghavan et al.

into water sources through industrial and agricultural pro-

).

cesses. There are both organic and inorganic forms of

Iron oxides, oxyhydroxides and hydroxides (all are

arsenic that exist in water sources, but inorganic arsenic

called ‘iron oxides’) play an important role in a variety of

is the most likely to exist in concentrations high enough

industrial applications, including pigments for the paint

to cause concern for drinking water quality (Thirunavuk-

industry, catalysts for industrial synthesis and raw materials

karasu et al. ). The health effects of arsenic have

for the iron and steel industry (Cornell & Schwertmann

been widely studied in humans, most notably in Taiwan.

). Studies showed that adsorption and filtration treat-

The health effects of arsenic in humans vary depending

ment systems using iron oxide-coated media are effective

on the compound and form. The maximum acceptable

in removing arsenic to a level below the arsenic guideline

concentration (MAC) for arsenic in drinking water is

(Joshi & Chaudhuri ; Driehaus et al. ; Thirunavuk-

10 μg/L in Canada and it was established based on the

karasu et al. ), and are suitable technologies for small-

incidence of internal (lung, bladder, and liver) cancers in

scale water treatment utilities.

humans, through the calculation of a lifetime unit risk (Health Canada ).

Uranium is a naturally occurring element that can be present in water supplies. In Saskatchewan, uranium in

Arsenic can be effectively treated in municipal-scale

groundwater typically occurs as a result of leaching of the

treatment facilities through a number of well-documented

element from soils and rocks. The interim MAC for uranium

methods, which typically include both a pretreatment step

in drinking water in Canada is 20 μg/L (Health Canada

and a final polishing step (Health Canada ). Several

). Regarding treatment, laboratory studies and pilot

studies have demonstrated that arsenic removal can be

plant tests have shown that conventional anion exchange

achieved by various technologies, such as coagulation/fil-

resins are capable of removing uranium from drinking

tration, lime softening, activated alumina, ion exchange,

water supplies to concentrations as low as 1 μg/L (Clifford

reverse osmosis (RO), and manganese greensand filtration

& Zhang ). Favre-Re’guillon et al. () demonstrated

(Viraraghavan et al. ; US EPA ); in particular,

that nanofiltration membranes are capable of removing

coagulation with ferric salts was found to be the most effec-

uranium to a level lower than the World Health Organiz-

tive method in the case of large-scale water utilities (Cheng

ation guideline for uranium. The purpose of this study is to

et al. ; Scott et al. ). Fixed bed and filtration treat-

evaluate the performance of MGS and RO treatment

ment systems are becoming increasingly popular for

plants in removing naturally occurring substances from

arsenic removal in small-scale treatment systems because

raw water of the communities located in Saskatchewan,

of their simplicity, ease of operation and handling, regener-

Canada. This paper also outlines some of the water manage-

ation capacity and sludge-free operation.

ment activities undertaken by the WSA to implement the

In manganese greensand (MGS) filtration treatment systems that are suitable for small-scale communities,

drinking water standards and manage drinking water in the province.

74

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

SAMPLING, RESULTS AND DISCUSSION

Water Quality Research Journal of Canada

|

49.1

|

2014

eight human consumptive systems. Table 2 provides a list of the parameters and number of excursions at all WSA

The WSA has water quality standard and monitoring guide-

regulated waterworks.

lines for the province’s drinking water quality and all 673

A study was designed to evaluate the performance of

communities in Saskatchewan are required to monitor

MGS and RO treatment plants, and in this study raw and

and achieve the physical, chemical, health and toxicity,

treated water samples from 20 groundwater communities

and biological standards as specified in the operating per-

in Saskatchewan were collected and analysed for naturally

mits issued to the communities. Table 1 shows the details

occurring substances. During the summer of 2012, samples

of compliance with sample submission requirements and

were collected from 10 MGS and RO plants, respectively,

testing compliance for health and toxicity parameters

and analysed at the Saskatchewan Centre for Disease Con-

during the 2011–12, 2010–11, and 2009–10 fiscal years

trol (Provincial) Laboratory. Figures 1 and 2 show the

based on routine samples submitted by WSA regulated

results of samples collected from MGS filtration plants of

communities in Saskatchewan. The decrease in sample

10 communities; the arsenic concentration in the raw

submissions in 2011–12 is the result of decreased monitor-

water of these communities varies from 4 to 38 μg/L.

ing by some smaller existing waterworks to determine

Arsenic removal efficiency of these plants ranged between

compliance with the health and toxicity standards that

70 and 94 per cent and the highest removal efficiency was

took effect in December 2010. WSA has and will continue

achieved in the MGS plant of the community Kelliher.

to follow up on a quarterly basis with waterworks owners

The results showed that all the plants removed arsenic to

who have not submitted the required samples as a means

a level well below the drinking water guideline of 10 μg/L.

to help ensure compliance with monitoring and drinking

The raw water uranium levels in these communities ranged between 1 and 14.3 μg/L and the results showed

water quality standards.

In 2011–12, there were 100 of 673 human consumptive

that the performance of MGS plants was poor in removing

waterworks that exceeded at least one health and toxicity

uranium from raw water. Selenium was not detected in

related chemical standard, resulting in a total of 128 excee-

the raw water of all these plants. High levels of iron and

dences.

toxicity

manganese were detected in most of the raw water of

parameters, such as arsenic or uranium, were encountered

these communities and all the MGS plants removed iron

and would represent a short-term health risk, waterworks

and manganese well below the guideline.

When

exceedences

for

health

and

owners were advised of the results and Precautionary Drink-

Figures 3 and 4 show the results of samples collected

ing Water Advisories were issued for the affected water

from RO plants of 10 communities; raw water uranium

supplies. Forty-six arsenic exceedences occurred in 23

levels of these communities ranged between 1 and 43 μg/L

human consumptive systems. Additional arsenic testing

and from the treated water results it was observed that

was conducted by 10 human consumptive systems. Sixty uranium exceedences occurred in 26 human consumptive systems. Additional uranium testing was conducted by Table 1

|

Health and toxicity sample submission and parameter result compliance 2011–12, 2010–11 and 2009–10a

Fiscal year

2011–12

Health and toxicity sample submission compliance rate (%)

75

Parameter standards compliance rate (%)

80

2010–11

89

84

2009–10

86

88

a

Health and toxicity parameters include: aluminum, arsenic, barium, boron, cadmium,

chromium, copper, iron, lead, manganese, selenium, uranium and zinc.

Table 2

|

Health and toxicity parameter specific excursion totals for WSA regulated waterworks during 2011–12 and 2010–11

Number of excursions

Number of excursions

Parameter

in 2011–12

in 2010–11

Arsenic

46

55

Barium

0

1

Copper

2

0

Nitrate

0

0

Lead

3

2

Selenium

3

8

Uranium

60

62

75

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

Figure 1

|

MGS plants: levels of naturally occuring substances in raw water.

Figure 2

|

MGS plants: levels of naturally occuring substances in treated water.

more than 99 per cent uranium was removed by these RO

Water Quality Research Journal of Canada

|

49.1

|

2014

that the presence of sulfate, total dissolved solids (TDS)

plants. However, the results showed that RO plants are not

and hardness in water (Figures 5 and 6) affects or inhibits

effective in removing arsenic from the raw water of some

arsenic removal. In the case of communities such as

communities. There may be many reasons why RO plants

Arlington Beach, Balcarres, and Foam Lake, the TDS,

could not remove arsenic, perhaps because of the molecu-

hardness and sulfate levels in raw water are high and

lar weight cut off (MWCO) of the membrane (used to

that may be the reason for the poor performance of RO

describe the pore size: the smaller the MWCO the tighter

plants in these communities in removing arsenic. Raw

the membrane size), but in this study it was observed

water TDS, hardness and sulfate levels in communities of

76

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

Figure 3

|

RO plants: levels of naturally occuring substances in raw water.

Figure 4

|

RO plants: levels of naturally occuring substances in treated water.

Water Quality Research Journal of Canada

|

49.1

|

2014

White Fox and Kamsack are low and the results showed

RO and MGS plants is blended (uranium level less than

that RO plants of these communities removed arsenic to

the guideline in blended water) and distributed to the com-

well below the drinking water guideline. The community

munity. Selenium was also detected (5.3 Îźg/L) in raw water

of Kenaston has two wells, well 1 has high uranium

from well 1 and was removed by the RO plant. The results

(30 Îźg/L) and the other has none, water from well 1 goes

also showed that both RO and MGS plants removed iron

to the RO plant and the water from well 2 is treated in

and manganese (Figure 7) to levels below the drinking

the MGS plant and ďŹ nally the treated water from both

water guideline.

77

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

Figure 5

|

RO plants: levels of sulfate, hardness and TDS in raw water.

Figure 6

|

RO plants: levels of sulfate, hardness and TDS in treated water.

TREATMENT STRATEGIES AND COST ECONOMICS

Water Quality Research Journal of Canada

|

49.1

|

2014

recently (2012) the community has upgraded the existing system by adding an ion-exchange (IE) resin treatment

Communities in Saskatchewan also adopt different treat-

system. A portion of treated water from the MGS filtration

ment strategies and/or multiple treatment systems to meet

system is further treated in the IE system, and the final

the drinking water guideline and reduce the cost of treat-

blended water from both MGS and IE plants has a uranium

ment. The raw water from the well of community Wapella

level lower than the guideline of 20 μg/L. The total upgrade

(population close to 500) has a uranium level higher than

cost of the system including addition of some distribution

the MAC, the community has a MGS filtration system and

system infrastructure is close to Can$1 million. The source

78

O. S. Thirunavukkarasu et al.

Figure 7

|

|

Groundwater treatment technologies: assessment

Water Quality Research Journal of Canada

|

49.1

|

2014

Iron and manganese levels in water.

water for Grenfell was from both surface and groundwater

arsenic if it coexists with high levels of iron in water. The

and the existing system (MGS filtration plant) could not

removal efficiency of the system as claimed by the manufac-

meet the treated water turbidity guideline; the town came

turer is greater than 90 per cent; however, it should be noted

up with an alternative source, i.e. two new groundwater

that this removal efficiency relies on an iron to arsenic ratio

wells; however, sampling results showed that the uranium

of at least 30:1. Additionally, the operating range for pH is

levels in these wells are high and the community built a

6.5 to 9.0 and water outside this range may require

RO plant in 2012 to remove uranium from raw groundwater.

additional pre-treatment to ensure the effectiveness of

The total upgrade cost including wells, distribution system

AD26 media.

and pumps is $1.8 million.

The AD26 treatment process often includes using chlor-

A group of University of Regina engineering students

ine as an oxidizer. Chlorine is injected into the water as a

developed cost equations that can be used by owners of Sas-

pre-treatment and oxidizes As(III) to As(V) and Feþ2 to

katchewan water supplies with high naturally occurring

Feþ3. The water is then filtered through the media where

arsenic levels. Ten arsenic-affected Saskatchewan water

ferric arsenate is able to form on the surface of the catalyti-

supplies were considered in this study. Three different

cally active media. Backwashing, a process which effectively

filter media (AdEdge, MGS and Media G2) that are capable

removes oxidized precipitated iron, manganese and arsenic

of removing arsenic levels to below that of the arsenic stan-

from the media bed, is necessary to maintain this system’s

dard were used in this study, and based on available raw

efficiency. This process typically required backwashing one

water quality data, cost analysis was conducted and cost

to three times per week, and a percentage of the backwash

equations developed.

water can be re-filtered through the system. In Kannata

AdEdge AD26 Series Systems are stand-alone systems

Valley,

a

Saskatchewan

community

that

recently

designed specifically for well head use. The system utilizes

implemented an AdEdge AD26 system, backwash water is

a dry granular form of manganese dioxide, which is an

collected, stored and left to settle in an underground storage

NSF 61 certified solid phase oxidation mineral media.

tank. After settling, up to 90 per cent of the supernatant from

Through co-precipitation and filtration, the system effec-

this backwash water has been sent back through the system

tively removes iron, manganese and sulfide, as well as

for treatment. Air wash is also typically implemented in the

79

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

Water Quality Research Journal of Canada

|

49.1

|

2014

systems and occurs before the backwash cycle is initiated.

oxidant should be fed at least 10–20 seconds upstream of

This process allows captured precipitated material to be dis-

the filter. This will oxidize the iron in order to convert

lodged from the filter media and allows for a shorter overall

arsenite to arsenate and is removed in the filter. Greensand-

backwash cycle.

Plus has an approximate life expectancy of 10–15 years. This

GreendsandPlus by Inversand Company is an advanced

ensures that the annualized costs of the media and system

filter media system used to remove soluble arsenic, manga-

are relatively low compared with other treatment systems

nese, iron, radium and hydrogen sulfide that are found in

(Inversand ).

municipal or industrial groundwater. GreensandPlus is a

Media G2, a filter media developed by ADI Inter-

purple charcoal filter media, which is similar to the original

national Inc., is made up of a material called diatomite.

MGS media. Both media have the same effective size,

This material is effectively the skeleton of diatoms, and is

uniformity coefficient, density, weight, capacity, and back-

used often in filtration for treatment of groundwater. It

wash and pressure drop curve. They also use manganese

resembles sand, and the basic particles are obtained from

dioxide as their substrate media. The difference is found in

ancient dried sea beds. This media works in a similar

the core of the media; the GreensandPlus core is made of

way to granular ferric hydroxide (GFH). Media G2 is

silica sand, which allows the manganese dioxide to coat

capable of removing both arsenate and arsenite between

the surface and act as a catalyst in the oxidation-reduction

a pH of 5.5 and 7.5, but is not affected by high levels of

reaction of iron and manganese.

iron. Regeneration of Media G2 is possible up to four or

The ideal parameters for the system include a pH range

five times before replacement, and regeneration is depen-

of 6.8–7.2, iron concentration of 3 mg/L and a manganese

dent on specific pH levels and arsenic levels in raw

concentration of 0.3 mg/L. GreensandPlus can withstand

water. Media G2 can be placed in any existing pressure fil-

water with low silica, total dissolved solids and total hard-

ters, which can simplify the retrofitting process for many

ness without any degradation and is effective at high

communities, as well as potentially decreasing retrofitting

temperatures and higher differential pressures allowing for

costs (ADI ).

longer run times. The GreensandPlus media system is oper-

The design flow that was used to develop the cost

ated using a catalytic oxidation process, which involves pre-

equations was based on the maximum daily flow and fire

injecting an oxidant, such as chlorine or potassium per-

flow demand. The filter or vessel dimensions and media

manganate, directly into the raw water source. The

volume are based on the design flow, empty bed contact

Figure 8

|

Cost equations for three media for different flow requirements.

80

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

Water Quality Research Journal of Canada

|

49.1

|

2014

time (EBCT), surface loading rate (SLR), media height

also developed based on the cost equations, which was

(Lcritical), and an expansion coefficient value (E). The

useful for the communities to determine approximate treat-

values for EBCT, SLR and Lcritical provided by respective

ment cost and select the appropriate filter media to

media companies were used in developing the cost equation.

remove arsenic from drinking water. The cost analysis

The cost analysis conducted for each media and system con-

showed that in communities with a design flow below

sidered both capital and operating costs. The capital cost

500 m3/day, the cost of the GreensandPlus system is lower

was determined based on media, volume, type of vessel

compared with other systems; however there may be other

and quantity, construction and auxiliary cost. The oper-

factors that may influence selection more than cost. For

ational cost requirements are based on media replacement,

example, if pre-treatment is necessary, the cost and

chemical costs, etc. The cost of chemicals includes the

additional treatment system may make another treatment

cost of KMnO4 and chlorine; the demand of KMnO4

choice more desirable. In communities with design flow

includes the concentrations of iron and manganese in the

more than 500 m3/day, the cost of Media G2 system is

raw groundwater and chlorine demand is based on residual

lower; however, the costs of GreensandPlus and AD26

chlorine levels in the distribution system. The auxiliary cost

system are still comparable and once again other factors

includes cost for engineering and management, additional

including site-specific conditions and system footprint may

pumps, pipes and valves, and monitoring equipment.

influence the selection of an appropriate treatment system.

The cost equation (Figure 8) for each media for different flow requirements was developed and adjusted with inflation as per Kawamura & McGivney (). A Microsoft

CONCLUSION

Excel based user interface with a cost template (Table 3) was Study results showed that MGS filtration plants are capable Table 3

|

of reducing arsenic levels in finished water to a level below

User interface cost template

the arsenic drinking water guideline and RO plants are effec-

Community info

AdEdge or Greensand or ADI

Name

EBCT (min)

××

Source water

SLR (m/min)

××

Treatment goals

Lcritical (m)

××

System parameters

Media volume (m3)

##

Population served

Area required (m2)

##

Average daily flow (m3/day)

No. of required vessels

Fire flow (m3/day)

16"

##

Design flow (m3/day)

21"

##

Existing pre-treatment

26"

##

Existing disinfection

System costs

Existing facility sq ft

Media

Water analysis

Pressure tank

pH

16"

##

Total As (ug/L)

21"

##

Treated As (ug/L)

26"

##

Iron (mg/L)

Auxiliary equipment

##

Manganese (mg/L)

New construction

##

Media lifetime cost

##

Annualized replacement

##

Total estimated cost

##

##

tive in removing uranium to less than 1 μg/L in finished water. Results also showed that MGS plants are not effective in removing uranium from drinking water. RO plants are not effective in removing arsenic from raw water of some communities and this may be due to the presence of high levels of sulfate, TDS and hardness in raw water that affects or inhibits arsenic removal. Communities in Saskatchewan, Canada adopt different treatment strategies and/or multiple treatment systems to reduce arsenic and/or uranium in treated water. The cost equations and user friendly interface model developed in this study are useful to arsenic-affected water supplies in Saskatchewan as a ‘decision making tool’ and help the communities in selecting appropriate filter media to remove arsenic from drinking water.

ACKNOWLEDGEMENTS The authors acknowledge the Government of Saskatchewan for funding and support to this initiative and research study.

They would also like to acknowledge the municipal

81

O. S. Thirunavukkarasu et al.

|

Groundwater treatment technologies: assessment

authorities, students and operators of the treatment plants who assisted in sample collection.

REFERENCES ADI  MEDIA G2 (http://www.indachem.com/ Manufacturers_ADI_International_MediaG2.htm).

Cheng, R. C., Liang, S., Wang, H. C. & Beuhler, M. D.  Enhanced coagulation for arsenic removal. J. Am. Water Works Assoc. 86, 79–90.

Clifford, D. A. & Zhang, Z.  Combined uranium and radium removal using ion-exchange. J. Am. Water Works Assoc. 86 (4), 214–227.

Cornell, R. M. & Schwertmann, U.  The Iron Oxides: Structure, Properties, Reactions, Occurrence and Uses. VCH Publishers, New York, p. 573.

Driehaus, W., Jekel, M. & Hildebrandt, U.  Granular ferric hydroxide – A new adsorbent for the removal of arsenic from natural water. J. Water Suppl. Res. Technol. – AQUA 47, 30–35.

Favre-Re’guillon, A., Lebuzit, G., Murat, D., Foos, J., Mansour, C. & Draye, M.  Selective removal of dissolved uranium in drinking water by nanofiltration. Water Res. 42, 1160–66.

Health Canada  Guidelines for Canadian Drinking Water Quality: Uranium (http://www.hc-sc.gc.ca/ewh-semt/pubs/ water-eau/uranium/index-eng.php).

Health Canada  Guidelines for Canadian Drinking Water Quality: Arsenic (http://www.hc-sc.gc.ca/ewh-semt/pubs/ water-eau/arsenic/index-eng.php).

Health Canada  Guidelines for Canadian Drinking Water Quality (http://www.hc-sc.gc.ca/ewh-semt/water-eau/drinkpotab/guide/index-eng.php).

Water Quality Research Journal of Canada

|

49.1

|

2014

Inversand  Manganese greensand (http://www.inversand. com/maggreen.htm).

Joshi, A. & Chaudhuri, M.  Removal of arsenic from ground water by iron oxide coated sand. J. Environ. Eng. 122 (8), 769–771.

Kawamura, S. & McGivney, W. T.  Cost Estimating Manual for Water Treatment Facilities. Wiley InterScience, Hoboken, NJ, USA. Paige, C. R., Snodgrass, W. J., Nicholson, R. V. & Scharer, J. M.  The crystallization of arsenate contaminated iron hydroxide solids at high pH. Water Environ. Res. 68, 981–987.

Scott, N. K., Green, F. J., Do, D. H. & McLean, J. S.  Arsenic removal by coagulation. J. Am. Water Works Assoc.

87, 114–126.

Subramanian, K. S., Viraraghavan, T., Phommavong, T. & Tanjore, S.  Manganese greensand for removal of arsenic in drinking water. Water Qual. Res. J. Canada 32, 551–561.

Thirunavukkarasu, O. S., Viraraghavan, T. & Subramanian, K. S.  Removal of arsenic in drinking water by iron-oxide coated sand and ferrihydrite – Batch studies. Water Qual.

Res. J. Canada 36 (1), 55–70.

Thirunavukkarasu, O. S., Viraraghavan, T. & Subramanian, K. S.  Arsenic removal from drinking water using iron-oxide coated sand. Water Air Soil Pollut. 142, 95–111.

US EPA  Technologies and Costs for Removal of Arsenic from Drinking Water. EPA-15-R-00-028, Office of Water, US Environmental Protection Agency, Washington, DC. Viraraghavan, T., Subramanian, K. S. & Swaminathan, T. V.  Drinking water without arsenic: a review of treatment technologies. Environ. Syst. Rev. 37, 1–35.

Viraraghavan, T., Subramanian, K. S. & Aruldoss, J. A.  Arsenic in drinking water problems and solutions. Water Sci. Technol. 40 (2), 69–76.

First received 7 January 2013; accepted in revised form 14 June 2013. Available online 27 August 2013

82

© IWA Publishing 2014 Water Quality Research Journal of Canada

|

49.1

|

2014

Computational modelling techniques in the optimization of corrosion control for reducing lead in Canadian drinking water C. R. Hayes, T. N. Croft, A. Campbell, I. P. Douglas, P. Gadoury and M. R. Schock

ABSTRACT Compliance modelling has been used to good effect in the optimization of plumbosolvency control in the UK and was evaluated in the Canadian and US contexts via three case studies. In relation to regulatory compliance, supplementary orthophosphate dosing could be justified in one water supply system but not in one other. Compliance modelling indicated that Health Canada’s Tier 1 protocol is much less stringent than its Tier 2 protocol and that optimization based on 6þ hour stagnation samples vs 15 μg/l is likely to be more stringent than that based on 30 min stagnation samples vs 10 μg/l. The modelling of sequential sampling for an individual home indicated that sample results could be markedly affected by the length of the lead service line, by the length of the copper premise pipe and by pipe diameters. The results for sequential sampling were also dependent on flow characteristics (plug vs laminar). For either regulatory compliance assessment or for the optimization of plumbosolvency control measures, routine sequential sampling from the same houses at a normalized flow will minimize these variable effects.

Key words

| corrosion control, lead in drinking water, modelling, optimization, sampling

C. R. Hayes (corresponding author) T. N. Croft Civil & Computational Research Centre, College of Engineering, Swansea University, Singleton Park, Swansea, SA2 8PP, UK E-mail: c.r.hayes@swansea.ac.uk A. Campbell I. P. Douglas City of Ottawa Water, 1 River Street, Ottawa, Ontario, Canada P. Gadoury Providence Water, Engineering, 430 Scituate Avenue, Cranston, Rhode Island 02920, USA M. R. Schock Environmental Protection Agency, 26 W. Martin Luther King Drive, Cincinnati, Ohio 45220-2242, USA

INTRODUCTION The publication of the World Health Organization’s Book-

However, there are a range of potential problems to be

let on Childhood Lead Poisoning (WHO ) has

overcome in optimizing corrosion control for reducing

heightened concerns about lead ingestion and its potential

lead in drinking water (IWA , ):

impact on health, particularly reductions in IQ. It followed the withdrawal, earlier in 2010, of the guideline value for a

(i) there is no simple, rapid control loop for linking cor-

tolerable lead intake by the Joint FAO/WHO Committee

rosion

on Food Additives (WHO ) and their conclusion that

concentrations of lead in drinking water at consumers’

there is no safe level for lead ingestion by children. More

control

treatment

to

reductions

in

the

faucets;

recently, the US CDC has revised its action level for lead

(ii) the definition of the term ‘optimization’ is vague and tends

in blood from 10 μg/dl down to 5 μg/dl (CDC ). These

to rely too heavily on regulatory compliance (which may

increased health concerns put renewed pressure on regula-

not be the same as treatment process control);

tors to minimize lead in drinking water within their areas

(iii) regulatory compliance systems are prone to distortion

of jurisdiction and on water utilities to optimize their cor-

as a consequence of the sampling protocols used and

rosion control systems for lead control.

variability due to a range of influencing factors.

doi: 10.2166/wqrjc.2013.009

83

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

These problems in optimizing corrosion control for

The option of sampling after 6þ hours’ stagnation

reducing lead in drinking water have largely been over-

involves Tier 1 and Tier 2 sampling. Tier 1 surveys relate

come in the UK over the past 10 years and the

to the first litre drawn from the faucet after the stagnation

experience gained should be a useful reference point for

period; if the Action Level of 15 μg/l is exceeded at more

Canadian utilities and regulators, including the use of com-

than 10% of the homes sampled then corrective action is

putational modelling techniques that were used to good

required as well as supplementary Tier 2 sequential

effect by many UK water companies. This paper outlines

sampling of the following second, third and fourth litres.

what was achieved in the UK and summarizes the results

The prompting of corrective action on the basis of Tier 1

of a recent collaborative research project that aimed to

sampling is flawed because most first litre samples will con-

demonstrate the potential uses of computational modelling

tain water that has stood in non-lead pipework adjacent to

in the Canadian and US contexts; it also summarizes the

the faucet, not the lead service line, and in consequence pro-

results of an investigation into the behavioural character-

blems with lead in drinking water will be greatly under-

istics of sequential sampling.

estimated (IWA ). While Tier 2 samples will more likely contain water that has stood in a lead service line, this is not certain and will vary depending on the volume

CANADIAN GUIDELINES FOR LEAD IN DRINKING WATER The Canadian guideline (Health Canada , ) for lead in drinking water, as a maximum acceptable concentration (MAC), has been 10 μg/l since 1992, based on flushed samples, until recently. As a consequence of flushing prior to sampling, the samples could neither determine the

of non-lead pipework. Distortion of the results from Tier 1 and Tier 2 sampling will also vary due to changes in the sampling pool, due to the proportion of homes sampled that have a lead service line, if Tier 1 and Tier 2 sampling is done at different times, and as a consequence of any deviations from the sampling protocol by home-owners (IWA ). The use of these sampling and assessment protocols in the optimization of corrosion control must be questioned.

extent of lead in drinking water problems within a water

The second option involves taking four sequential

supply system nor provide any basis for optimizing cor-

samples (each of 1 l volume) after 5 min flushing and

rosion control. The guideline for lead was, in essence,

30 min stagnation, from properties with lead service lines.

revised in August 2012 (Health Canada ); while the MAC of 10 μg/l remains the same, reference is no longer

Corrective action is required if the average result from the four samples exceeds 10 μg/l at more than 10% of the

made to flushing prior to sampling; instead, reference is

homes surveyed. This option is also prone to variable distor-

made to applying the MAC as an average concentration

tion from water stood in non-lead pipework as well as

over extended periods. Recognizing the practical limitations

variation from changes in the sampling pool; further, the

of the earlier guideline, Health Canada issued ‘Guidance on Controlling Corrosion in Drinking Water Distribution Systems’ (Health Canada ), which recommended two options for assessing lead in domestic drinking water:

averaging of results is questionable. Health Canada () does not recommend this option for optimizing corrosion control due to uncertainties about sensitivity and behavioural characteristics.

(i) first draw samples after 6þ hours’ stagnation (at least 50%

To complicate matters, some provinces have their own

of the homes sampled must have a lead service line),

standards for lead in drinking water. For example, Ontario

based on the US Lead Copper Rule (US EPA ), with

has a standard of 10 μg/l based on the highest lead concen-

further sequential samples in some circumstances; or

tration from two sequential samples (each of 1 l volume)

(ii) if sampling after a 6 þ hour stagnation time is not

taken after flushing then 30 to 35 min stagnation. Corrective

practical or is restricted by regulatory obligations,

action is required if the standard is exceeded at more than

sequential samples after 30 min stagnation from proper-

10% of the homes surveyed. The Ontario standard is no

ties with lead service lines.

less susceptible to variable distortion effects. Overall, it

84

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

can be concluded that there is considerable scope to develop

solubility or computational models, if an increase in orthopho-

the guidelines further, with the aim of providing a stronger

sphate dose produced no further worthwhile improvement or

basis for quantifying lead in drinking water problems and

if a sufficient number of random daytime (RDT) samples had

for optimizing corrosion control.

been taken and less than 2% of samples exceeded 10 μg/l.

This numeric criterion was adopted by most water companies as their target for optimization.

The DWI followed up the progress being made by water

OPTIMIZING CORROSION CONTROL FOR REDUCING LEAD IN DRINKING WATER

companies with technical audits. Optimization schemes were

A generic definition of the term ‘optimization’ as it relates to

were reported formally. Arrangements in Northern Ireland

subject to legally binding agreements and once concluded

the control of lead in drinking water has been proposed

and Scotland were broadly similar. Across the UK, 95% of

in the International Water Association’s Code of Practice

public water supply systems are now dosed with orthopho-

for the Internal Corrosion Control of Water Supply Systems

sphate. The concentration of orthophosphate that is dosed

(IWA ):

varies from 0.5 to 2.0 mg/l (as P), most typically between 1.0 and 1.5 mg/l (as P) and is water supply system specific,

‘The application of best available techniques, not entailing

as determined by both water quality and the extent of occur-

excessive cost, to reduce lead concentrations in drinking

rence of houses with lead pipes. For England and Wales in

water to the minimum that is practical to achieve.’

2009/10, following the optimization of plumbosolvency control treatment, 99.0% of RDT samples complied with the new

This definition implies a holistic approach that is robust

standard of 10 μg/l for lead in drinking water (that applies

scientifically, that incorporates risk assessment and that

from December 2013), compared to 80.4% before orthopho-

matches mitigation measures specifically to the circum-

sphate dosing became widespread. In some regions, 99.5%

stances of the water supply system. It is consistent with

compliance has already been achieved intermittently and

Health Canada’s guidance () that delivered water

could be considered as a national target. If optimized plum-

should ‘not be aggressive’ to distribution systems, including

bosolvency control was reinforced by selective lead pipe

domestic pipework. The definition in the Code of Practice

replacement (DWI ), it might even be possible to achieve

is also consistent with the approach taken in the UK.

99.8% compliance without the widespread replacement of

In England and Wales, optimization was defined as the ‘best practical reduction in lead concentrations’ (DWI , ) and was taken to mean ‘maintaining an optimum ortho-

lead pipes (Hayes & Hydes ).

The optimization of plumbosolvency control in the UK was mostly achieved in one of two ways:

phosphate dose throughout a water supply system, within an optimum pH range’. If pH and alkalinity control were pro-

(i) step changes in orthophosphate dose until optimum

posed as the only treatment measures, water companies had

reductions in lead had been demonstrated by in situ

to demonstrate that an optimum dose of orthophosphate

lead pipes at consumers’ houses; however, in situ lead

could not achieve a further significant reduction in lead con-

pipes can take up to 2 or 3 years to equilibrate with a

centrations (in practice, none did so). Water companies were

new orthophosphate dose (IWA ) and dose

also expected to take into account any organic or iron discolor-

responses may not have been fully established within

ation interference effects. In essence, the Drinking Water

the timescales that were available (pipe rigs using old

Inspectorate (DWI) implemented an optimization framework,

exhumed lead pipes will be similarly affected); or

recognizing that the precise definition of an optimum ortho-

(ii) laboratory plumbosolvency testing coupled with compli-

phosphate dose was necessarily vague, at least initially. It

ance modelling to quickly determine the likely optimum

was stated (DWI , ) that the optimum dose of ortho-

dose, which was then confirmed (and adjusted if necess-

phosphate could be determined from laboratory tests, from

ary) by routine monitoring of in situ lead pipes at

full scale or pilot scale trials, by practical experience, from

consumers’ houses; this approach minimized the

85

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

number of iterative changes to water treatment con-

described by curves of the concentration of lead that

ditions and saved both time and money.

increases over time when the water stands (stagnates)

In most water companies, RDT sampling was the principal means used for demonstrating the success of the plumbosolvency control measures, supplemented variously by fixed point monitoring (typically, 30 min stagnation sampling) at selected houses and the use of lead pipe test rigs.

The success of the second approach (Hayes et al. , ) was the motivation for the ‘proof-of-concept’ project to demonstrate the feasibility of an amended modelling system being used in the optimization of plumbosolvency control in Canadian and US drinking water supply systems.

within a lead pipe. Curves A1 and A2 describe waters with a higher plumbosolvency than curves B and C and the shape of the curves can vary (A1, A2). These simple exponential curves are defined by the initial mass transfer rate of lead M (μg/m2/s) from the internal lead pipe surface (which determines the initial slope) and the equilibrium concentration of lead E (μg/l). Such curves are an adequate representation of the lead dissolution curves generated by Kuch and Wagner’s diffusion model () and offer computational advantage (much quicker computer run-times). The values of M and E can be determined experimentally by laboratory plumbosolvency testing using the method developed

by Colling et al. (). This involves pumping test water continually through short sections of new lead pipe at a con-

COMPLIANCE MODELLING TECHNIQUES

W

stant 25 C and a constant contact time of 30 min for a period of around one month; at the end of this period the

Describing plumbosolvency

test water is allowed to stand for 16 hours before sampling.

The concentration of lead after 30 min contact enables the

The schematic diagram given in Figure 1 illustrates (top

value of M to be determined from the numerical relationship

right) how the plumbosolvency of drinking water can be

determined by Hayes ().

Figure 1

|

Compliance modelling schematic.

86

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

In Figure 1 (top left) are the actual results, as median

left). These distributions can be based on generally applied

concentrations, from plumbosolvency testing for two test

assumptions or on survey data provided by the water utility.

waters at 30 min contact. The test series in these cases

Normally, the plumbosolvency characteristics are applied as a

both show the effect of increasing orthophosphate over a

constant, but it is possible to apply a range if water quality (e.

zero to 1.5 (mg/l P) range. Alternatively, a sequence of pH

g., pH) was known to vary significantly. The single pipe model

values could be investigated by the test procedure. Such

then generates the lead emission profile at each simulated

data generate curves of different magnitude (top right)

home for a 24-hour period. The daily average lead concentration

depending on the treatment condition and its associated

at each simulated home can be readily calculated.

extent of lead reduction. M and E can also be estimated from sequential sampling after 30 min and 6þ hour stagna-

Sampling models

tion (respectively) at homes with a lead service line.

A sampling model can then be used to investigate lead emisSingle pipe model

sions across the water supply system, in terms of probability, based on a range of sampling protocols. These include: (a)

The single pipe model consists of a lead pipe, non-lead pipe

random daytime samples, (b) 30 min stagnation samples

and an imaginary faucet that is represented by a 1 l sample

(first litre drawn and sequential), and (c) 6 hour stagnation

volume. The pipes are broken down into elements (as illus-

samples (first litre drawn and sequential). In the case of stag-

trated in Figure 1 – middle right) and when assuming plug

nation sampling methods, the lead concentration in the

flow each element is treated as a stirred tank. During periods

simulated pipes is assumed to be zero immediately prior to

of flow, the contents of one stirred tank are passed to the

the stagnation period.

next, and so on, at a time increment of 1 s. The concen-

It is therefore possible to investigate the link between

trations of lead that are computed at the imaginary faucet

the plumbosolvency of the water in the system and probable

for every second of flow are determined by: (a) the length

compliance with regulatory sampling protocols. Simulated

and diameter of the lead and non-lead pipes, (b) the contact

samples are taken randomly from the probabilistic frame-

time of the water, as determined by the assumed pattern of

work, typically 100 samples in a simulated survey,

water use, and (c) the plumbosolvency of the water. The

repeated 1,000 times so as to gain an appreciation of var-

assumption of plug flow is another simplification and

iance. Mapping on to regulatory sampling protocols also

describes adequately the turbulent flow that is most likely

involves setting the percentage of the homes that must

in the small-bore pipes in premise plumbing (commonly ½

have a lead pipe in the sampling simulations (e.g., 50% for

and ¾ inch) at the flow rates expected in normal domestic

the US Lead Copper Rule (LCR), 100% for the Canadian

use (around 0.1 l per second, or 1.6 US gallons per minute).

30 min stagnation (30MS) sampling protocol).

Laminar flow can also be investigated but is computationally much slower. The mathematical equations used for both flow

Limitations, simplifications and validation

types are given in Van der Leer et al. ().

The deterministic models described above relate only to the Using a probabilistic framework for zonal modelling

dissolution of lead from lead pipes and no allowance is made for lead leaching from brass or for galvanic corrosion

To investigate lead emissions across an entire water supply

or particulate lead effects, although it is possible to include a

system, a probabilistic framework is established by creating a

correction factor for lead from premise plumbing if data are

large number (normally 10,000) of simulated homes that

available to characterize this. Simplifications are made in

make up the simulated system, as defined by lead and non-

the way that plumbosolvency and flow are defined and the

lead pipe and water use charcteristics. The variables that

calibration matrix can be limited to the application of gener-

define each simulated home derive from a series of statistical dis-

alized assumptions (such as non-lead pipe lengths) when

tributions, applied randomly, as illustrated in Figure 1 (bottom

detailed survey data are not available.

87

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

Despite these limitations and simplifications, excellent

basis. The results of this optimized orthophosphate

validation of predicted RDT sampling results was obtained

dosing were confirmed over a 2 year period by a combi-

in numerous case studies in the UK (Hayes et al. ,

nation of RDT sampling and the 30 min stagnation

) by comparison with the results of actual RDT

sampling of lead pipe test rigs. This approach successfully

sampling by the water companies. As a case example, data

delivered 99% compliance with the future UK lead stan-

are shown in Figure 2 for pre- and post-orthophosphate

dard of 10 μg/l.

dosing conditions in Cambridge (UK); systems before ortho-

It can be concluded from the successful validation

phosphate dosing commenced are numbered 1 to 6; the

achieved in these two major case studies that the simplifica-

system numbered 7 was after orthophosphate dosing in the

tions inherent in the modelling procedure do not hinder its

aggregate of the city’s four major systems. In this case

use in operational terms. The main difference between the

example, comprehensive pipe surveys, knowledge of water

UK and US/CA case studies are the sampling protocols

use and plumbosolvency testing enabled the compliance

used for assessing regulatory compliance.

model

to

be

calibrated

without

using

generalized

assumptions.

In Wales (Hayes et al. ), orthophosphate dosing was

CANADIAN AND US CASE STUDIES

optimized by a combination of laboratory plumbosolvency testing and compliance modelling for 29 systems that were

Three case studies (two US, one Canada) involved the use of

subject to a Regulatory Programme of Work (as agreed

modelling techniques in various ways and the results are

with the Drinking Water Inspectorate). Of these 29, plumbo-

presented below in a Canadian context. Further details of

solvency testing data were not available for six schemes and

this modelling are available from the project report (Hayes

the compliance model had to be fitted to the observed data

& Croft ).

that were available, although the benefit of orthophosphate dosing could still be predicted. Generalized assumptions

Case study 1 (US)

had to made for pipe lengths and water use for all systems.

Despite these limitations, the predicted RDT sampling

The water supply system serves a population of around

results, for the pre-orthophosphate condition, were closely

300,000 people and 35% of the customer connections are

matched by the water company’s sampling for the 23 sys-

lead service lines, including the privately owned side. In

tems in which primary validation was possible.

order to control plumbosolvency, it has been long-standing

The optimum orthophosphate doses that were deter-

practice to elevate pH in order to suppress lead solubility.

mined were subsequently applied on a system-specific

The pH within the system at the present time is typically between 9.5 and 10.0. Corrosion inhibitors such as orthophosphate are not used. The system was marginally noncompliant with the criteria for lead set by the US LCR in seven out of nine surveys over the period 2007 to 2011.

In calibrating the zonal compliance model, very comprehensive data were available on the length and diameter of service lines but very little on premise plumbing. The plumbosolvency factors M (0.047) and E (70) were estimated by inspection of the results from LCR surveys and from nine sequential sampling exercises, as data from laboratory plumbosolvency testing were not available. The assumptions about non-lead premise plumbing were initially based on those used in UK case studies, with subsequent

Figure 2

|

Validation data from Cambridge Water (UK).

amendment. Additionally, a correction factor was used for

88

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

lead emissions from non-lead premise plumbing; the need

had sampled beyond the first litre, it is possible that the cali-

for this was identified from the detailed inspection of LCR

bration of the compliance model may have been adjusted.

survey results from premises with copper service lines. The

The orthophosphate dosing scenario (A) defined by

simulated and observed LCR survey results are shown in

M ¼ 0.02 and E ¼ 30 is typical of many optimized systems

Table 1 and indicate a good match based on the first litre

in the UK and should be achievable with a relatively low

sampled after stagnation, equivalent to Tier 1 sampling in

orthophosphate dose. It was predicted that LCR compliance

Canada.

would be achieved, based on the first litre drawn after stag-

The potential benefits of orthophosphate dosing were

nation (Tier 1), but not if compliance was based on further

then investigated, using the compliance model, by reducing

sequential samples (Tier 2). This is also the case when

the values of M and E that define plumbosolvency. As these

M ¼ 0.015 and E ¼ 22.5. To achieve Tier 2 compliance

values were reduced, an equivalent reduction factor was

would require M ¼ 0.01 and E ¼ 15, likely to be at the

applied to the premise plumbing correction. The results for

extreme range of orthophosphate’s ability to suppress lead

three orthophosphate dosing scenarios are shown in

solvency. These predicted results indicate that optimization

Table 2 for four sequential litre samples and are compared

based on Tier 1 sampling would be different to that based on

to the present non-orthophosphate dosed condition. Only

Tier 2 sampling, if judged by the same action level of 15 μg/l.

data for the first litre were available from the LCR compli-

Laboratory plumbosolvency testing would be necessary to

ance monitoring, in order to calibrate the model (Table 1);

confirm the orthophosphate dosing response of the water

the predictions for the second, third and fourth litres are

and to determine the orthophosphate doses associated

more speculative, but do show a significant difference

with each scenario.

between the first litre and the litres that follow. If the utility Case study 2 (CA) Table 1

|

Matching the model to observed LCR results

a

The water supply system serves a population of around Average 90th percentile

Average percentage of samples

concentration and range (μg/l)

exceeding 15 μg/l and range

Predicted: 20.5 (6.3–37.8)

Predicted: 14.6 (4.0–27.0)

Observed: 20.1 (13.6–30.0)

Observed: 16.2 (8.0–29.4)

a

800,000 and 14% of customer connections are believed to have lead service lines. In order to control the plumbosolvency of the water supplies to consumers, it has been the utility’s long-standing practice to elevate pH in order to sup-

Predicted results based on simulated first litre samples after 6 hours’ stagnation, equiv-

alent to Tier 1 sampling in Canada; 56.3% of the simulated houses had a lead service line,

press lead solubility. The treated water has a pH of 9.2 to 9.4 and this has been successful in meeting regulatory compli-

consistent with the utility’s LCR surveys.

ance with Provincial standards, which are based on sequential Table 2

|

Predicted LCR compliance for orthophosphate dosing scenarios: average 90th

sampling after 30MS. Corrosion inhibitors, such as orthopho-

percentile concentrations from Tier 1 and Tier 2 samplinga

sphate, have not been used.

Average 90th percentile concentrations Modelling scenario

M

E

Detailed information on the lengths of lead service lines was not available. Instead, reference was made to infor-

(μg/l) 1st

2nd

3rd

4th

Litre

Litre

Litre

Litre

Tier 1

Tier 2

Tier 2

Tier 2

mation from another Canadian city (Cartier et al. ) with minor adjustments to help fit predicted 30MS survey results to those observed. For lead service line diameters, a

Without o-PO4

0.047

70.0

20.5

63.8

67.0

59.1

consensus view from supply network engineers was used.

o-PO4 – (A)

0.020

30.0

10.1

27.8

28.8

25.9

No data were available for premise plumbing other than esti-

o-PO4 – (B)

0.015

22.5

9.6

20.8

21.6

18.5

mates of the pipe materials involved: >90% copper, 2%

o-PO4 – (C)

0.010

15.0

8.7

14.2

14.5

13.1

galvanized iron, 8% plastic. Again, reference was made to

a

Predicted average 90th percentiles from 1,000 simulated surveys, each of 100 simulated samples after 6 hours’ stagnation, using the plug flow model; M ¼ initial mass transfer rate (μg/m2/s); E ¼ equilibrium concentration (μg/l); 56.3% of the simulated houses had a lead service line.

information from the other Canadian city with minor adjustments. Estimates of E (31) and M (0.026) were made in a similar manner to the first case study and fine-tuned by

89

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

calibration trials. A correction for lead release from premise

the utility compared the lead concentrations in sequential

plumbing was not considered necessary as the utility had

samples after stagnation for 30 min and 6 hours at 22 homes.

indicated that in their experience, lead was very rarely

On average, for each litre sampled sequentially, the lead concen-

detected at homes that did not have a lead service line.

trations after 6 hours’ stagnation were 1.1 to 1.4 times higher

The match between simulated and observed 30MS survey

than the lead concentrations after 30 min stagnation, much

results is shown in Table 3, based on the first four litres

closer than expected. The ratio between lead concentrations

sampled after flushing and stagnation.

after 30 min and 16 hour stagnation is normally in the range

The slight reductions in M to 0.020 and E to 30, the con-

3 to 5 but can be as low as 1.8 and as high as 15.2 for waters with-

ditions typical for optimized orthophosphate dosing in the

out orthophosphate (Hayes ); on the basis of the lead

UK, resulted in a very substantial change in compliance

dissolution curves generated by the model, lead concentrations

(Table 3). This is because 30MS values of 10 μg/l cannot

after 6 and 16 hours’ stagnation are expected to be fairly similar.

be exceeded by the lead dissolution curve generated by the

In consequence of the wide range in this ratio, the shape of the

model using these values for M and E. It can be concluded

lead dissolution curve can differ as illustrated in Figure 3 and is

that the case for supplementary orthophosphate dosing is

specific to individual waters and scale surface chemistries. The

weak, assuming that the estimates of M (0.026) and E (31)

predicted data in Tables 4 and 5 are based on lead dissolution

are correct (or close) for this water supply system.

curves that had 6 hour to 30 min stagnation ratios of 2.4 to 3

Simulated results for sequential sampling after 6 hours’

and are therefore illustrative of the more average condition.

stagnation are shown in Table 4 (US LCR basis) and

Further work will be required to reconcile the model with the

Table 5 (Canadian Tier 1 and 2 basis) for surveys in which

utility’s limited survey results.

50% of simulated houses had a lead service line, in order to investigate the protocol specified by Health Canada

Case study 3 (US) and the modelling of sequential

() guidelines. With M ¼ 0.026 and E ¼ 31, compliance

sampling

was predicted to be achieved with Health Canada’s guidelines for Tier 1 sampling but not for Tier 2 sampling, the

The emphasis of this case study was the sequential sampling

difference being the effect of water stood in non-lead pipe-

of homes after 6þ hours’ water stagnation to investigate the

work. Similar results were obtained for M ¼ 0.020 and

behavioural characteristics of the sampling protocol used

E ¼ 30. A reduction to M ¼ 0.010 and E of 15 would be

by the US LCR. It is also relevant to Tier 1 and 2 assessments

required for Tier 2 compliance, likely to be at the limit of

in Health Canada’s guidelines. Sequential sampling survey

what plumbosolvency control treatment can achieve.

data were provided from two surveys by the US Environ-

It appears that compliance with Health Canada’s guidelines

mental Protection Agency. After flushing for 5 min and at

() is less stringent if based on 30 min stagnation samples if

least 6 hours’ stagnation, 12 or more 1 l samples were taken

the results in Table 3 are compared to those in Tables 4 and 5

in sequence from each site. The lead concentrations that

for M ¼ 0.02 and E ¼ 30. However, in an earlier investigation,

were observed varied from 2 to 37 μg/l and all first litre

Table 3

|

Simulated and observed 30MS survey resultsa Average % 30MS samples > 10 μg/l: 1st Litre

2nd Litre

3rd Litre

4th Litre

Predicted with M ¼ 0.026 and E ¼ 31 (range in brackets)

0.21 (0.00–3.00)

1.92 (0.00–7.00)

7.79 (1.00–18.00)

15.23 (4.00–25.00)

Observed over 8 surveys from 2008 to 2011 (range in brackets)

1.33 (0.00–2.56)

2.36 (0.00–5.98)

8.55 (2.42–16.24)

11.95 (1.92–26.50)

Predicted with M ¼ 0.020 and E ¼ 30 (range in brackets)

0.00 (0.00–0.00)

0.00 (0.00–0.00)

0.00 (0.00–0.00)

0.00 (0.00–0.00)

a Predicted average percentages from 1,000 simulated surveys, each of 100 simulated samples; in the modelling, 95% of simulated houses were assumed to have a lead service line, reflecting the utility’s 100% objective and their indication that a few of the earlier survey samples were taken from homes without a lead service line.

90

Table 4

C. R. Hayes et al.

|

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

Average 90th percentile concentrations (μg/l) after 6 hours’ stagnation

M

E

1st Litre

2nd Litre

3rd Litre

4th Litre

0.026

31

0.09

20.20

30.54

30.62

0.020

30

0.12

19.54

29.09

29.08

0.010

15

0.01

9.22

14.54

14.57

a

49.1

|

2014

June 2011 survey, 13 out of the 28 sites exceeded 15 μg/l in

Predicted LCR compliancea

Plumbosolvency

|

In the USA, corrective action is required if the 90th percentile concentration exceeds

15 μg/l based on the first litre; the modelling assumed that 50% of simulated houses had a lead service line.

at least one sequential sample and in the September to October 2011 survey, 16 out of the 29 sites exceeded 15 μg/l in at least one sequential sample. At some sites, elevated lead concentrations were observed for greater parts of the 12-sample sequence (up to twice) than could be explained by the volume of the lead service line, implying laminar flow influences during sampling. Similar observations have been made in Rhode Island (RIH ).

The single pipe model was used (after validation) to investigate the lead emission characteristics of different pipework

Table 5

|

circumstances at a single home, comprising a simulated lead

Predicted Tier 1 and 2 compliancea

Plumbosolvency

pipe (service line) and a simulated copper pipe (premise

Average percentage samples >15 μg/l after 6 hours’

plumbing) of various lengths and diameters, for both plug

stagnation

and laminar flow conditions. The detailed results are reported

M

E

1st Litre

2nd Litre

3rd Litre

4th Litre

0.026

31

1.41

13.62

34.90

31.62

0.020

30

1.45

14.01

35.11

30.87

0.010

15

0.00

0.00

0.00

0.00

a

in Hayes & Croft () and are summarized below. (i) Length of lead pipe:

•

In Canada, corrective action is required if more than 10% of sites exceed 15 μg/l; the

modelling assumed that 50% of simulated houses had a lead service line.

•

for plug flow: the longer the lead pipe, the greater were the number of sequential samples with an elevated lead concentration; for laminar flow: lead concentrations were lower generally than for plug flow, the elevation of lead was spread over a greater number of samples, and the longer the lead pipe, the higher was the peak lead concentration in the sample sequence.

(ii) Length of copper pipe:

• •

for plug flow: the longer the copper pipe, the later was the elevated lead in the sample sequence; for laminar flow: lead concentrations were lower generally than for plug flow, the elevation of lead was spread over a greater number of samples, and the longer the copper pipe, the lower was the peak lead concentration in the sample sequence.

(iii) Pipe diameter:

• Figure 3

|

Variation in lead dissolution characteristics. The three illustrative curves relate to different ratios of lead concentration after 30 min stagnation (30MC) and after 8 hours when equilibrium (E) has been assumed to be reached.

samples had a lead concentration below the Action Level of

for plug flow: the larger the diameters, the more extensive was the sequence of lead concentrations associated with the lead pipe and the lower was the

•

concentration of peak lead in the sample sequence; for laminar flow: lead concentrations were lower generally than for plug flow, the elevation of lead was

15 μg/l. Significantly, over 90% of houses had a premise

spread over a greater number of samples, and the

plumbing volume greater than 1 l, helping to explain the

larger the diameter of the lead pipe, the lower was

observed compliance with LCR criteria. However, in the

the peak lead concentration in the sample sequence.

91

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

This modelling suggests that sequential sampling results

Health ; Hayes & Croft ). These results prompted a

can be markedly affected by the length of lead service pipe,

modelling investigation using the single pipe model and both

by the length of copper premise pipe and by pipe diameters.

plug flow and laminar flow for a range of pipework character-

In consequence, it appears that sequential sampling may be

istics. The results have been summarized in case study 3

too subject to variable influences to be used in definitive

(above) and reported in detail by Hayes & Croft () and

terms, for either regulatory compliance assessment or for

show significant differences between the two flow regimes. Pre-

the optimization of plumbosolvency control measures, if

sently, it is difficult to know to what extent laminar flow occurs

sampling is undertaken from different sets of houses from

under normal home use and under what flow and pipework

within a changing sampling pool.

conditions; further research is planned that will endeavour to

Sequential sampling can be used for diagnostic purposes, but differentiating specific sources of lead within

describe the behaviour of transitional flow regimes in the context of sequential sampling for lead in drinking water.

the premise plumbing may be affected by skewing and low-

The new interpretation of the Canadian guideline for

ering effects. While sequential sampling appears to have

lead in drinking water, that the MAC of 10 μg/l should be

limitations, it will provide a much better insight into lead

applied as an average over extended periods (Health

emission characteristics than the sampling protocols that

Canada ) creates new and interesting challenges,

only utilize the first litre drawn after the stagnation period.

namely how to determine the average concentration of lead at a home. The most accurate sampling method will be split-flow composite sampling (Van den Hoven et al.

DISCUSSION

) but this has logistic limitations. Although outside the scope of the case studies reported here, the single pipe

The extent of calibration data in these Canadian and US

model can readily predict daily average lead concentrations

case studies was limited. The plumbosolvency of the water

and the compliance model can predict RDT sample results

could only be estimated as test data were not available and

in support of zonal risk assessment.

data on premise plumbing were mostly absent. It would, however, be relatively easy to strengthen the calibration of the compliance model by laboratory plumbosolvency testing

CONCLUSIONS

(quick and affordable) and by fuller pipework inspections when homes are sampled for compliance assessment pur-

1. Compliance modelling has been demonstrated to have a

poses. For the modelling approach to gain broader

potential role in the optimization of plumbosolvency

acceptance, it would be beneficial to conduct case studies

control, in the Canadian and US contexts, and to pro-

in cities which are already using orthophosphate or are con-

vide

templating its use. That said, there is no reason why the

a

deeper

appreciation

of

the

behavioural

characteristics of related sampling protocols.

modelling approach cannot be applied to systems that con-

2. A rapid, low cost approach to optimization could com-

trol internal corrosion through pH elevation alone or to

prise laboratory plumbosolvency testing, the gathering

systems that use other corrosion inhibitors.

of basic information about the water supply system and

The compliance modelling was undertaken assuming plug

compliance modelling, prior to confirmatory monitoring.

flow, similar to earlier UK studies, and is considered reasonable

Such an approach has already been used widely in the

on the basis that most premise plumbing is ½ or ¾ inch diam-

UK to good effect and can be accommodated for use in

eter and the flow rates are likely be around 0.1 l per second

Canada and the USA by an amended compliance

(1.6 US gallons per minute) in home use. However, the results

model that is able to simulate the relevant sampling

of actual sequential sampling in the city relating to case study 3

protocols.

(and elsewhere) have often been found to be skewed, whereby

3. In this context, compliance modelling can help to quan-

the lead concentrations in these sequential samples can only be

tify what is meant by ‘best practical reductions in lead in

explained by laminar flow effects (Rhode Island Department of

drinking water’ for a water supply system, by exploring

92

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Water Quality Research Journal of Canada

|

49.1

|

2014

more deeply the relationship between corrosion control

responsibility for the interpretation and use of the material

treatment and regulatory compliance.

lies with the reader. In no event shall the publisher or

4. The case studies indicated that Health Canada’s Tier 1

authors be liable for damages arising from its use. The

protocol is much less stringent than its Tier 2 protocol

views expressed by the authors do not necessarily represent

(if benchmarked against the same action level) and

the decisions or the stated policies of any organization

that optimization based on 6þ hour stagnation samples

referred to in this publication.

vs 15 μg/l is likely to be more stringent than that based on 30 min stagnation samples vs 10 μg/l.

5. In relation to regulatory compliance, the case studies indicated that supplementary orthophosphate dosing might be

REFERENCES

justified in one water supply system but not in one other.

6. Compliance modelling can assist the evaluation of sampling and assessment protocols in the further development of regulatory criteria, by examining far more scenarios than it is feasible to do experimentally.

7. The modelling of sequential sampling for an individual home indicated that sample results could be markedly affected by the length of the lead service line, by the length of the copper premise pipe and by pipe diameters. The results for sequential sampling were also dependent on flow characteristics (plug vs laminar).

8. In consequence, it appears that sequential sampling may be too subject to variable influences to be used in definitive terms, for either regulatory compliance assessment or for the optimization of plumbosolvency control measures, if sampling is undertaken from different sets of houses from within a changing sampling pool.

9. While sequential sampling appears to have limitations, it will provide a much better insight into lead emission characteristics than the sampling protocols that only utilize the first litre drawn after the stagnation period.

10. Sequential sampling under a normalized flow condition could be used in the routine benchmarking of lead concentrations at selected properties, as part of an optimization procedure for plumbosolvency control measures.

DISCLAIMER All reasonable precautions have been taken by the authors to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either express or implied. The

Cartier, C., Laroche, L., Deshommes, E., Nour, S., Richard, G., Edwards, M. & Prevost, M.  Investigating dissolved lead at the tap using various sampling protocols. J. Am. Water Works Ass. 103 (3), 55–67.

CDC  What Do Parents Need to Know to Protect their Children. Centers for Disease Control and Prevention, Atlanta, GA. Available at: www.cdc.gov/nceh/lead/ ACCLPP/blood_lead_levels.htm.

Colling, J. H., Whincup, P. A. E. & Hayes, C. R.  The measurement of plumbosolvency propensity to guide the control of lead in tapwaters. J. Inst. Water Environ. Manage.

1 (3), 263–269.

Drinking Water Inspectorate  Determination of requirements to meet new lead standards. Information Letter 12/2000.

Available at: http://dwi.defra,gov.uk.

Drinking Water Inspectorate  Further guidance on requirements to meet new lead standards. Information Letter 3/2001. Available at: http://dwi.defra,gov.uk.

Drinking Water Inspectorate  Guidance document. Guidance on the implementation of the Water Supply (Water Quality) Regulations 2000 (as amended) in England. September 2010.

Available at: http://dwi.defra,gov.uk.

Hayes, C. R.  Computational modelling of lead in drinking water. PhD Thesis, University of Wales.

Hayes, C. R.  Optimisation tools for achieving the lead standard of 10 μg/l in drinking water. Proceedings of an International Conference on Metals and Related Substances in Drinking Water, October 2007, Antalya, Turkey. COST Action 637.

Hayes, C. R. & Croft, T. N.  Collaborative Project to Determine Proof of Concept. Optimisation of Corrosion Control for Lead in Drinking Water using Computational Modelling Techniques. International Water Association, Specialist Group on Metals and Related Substances in Drinking Water, Research Report Series, IWA Publishing, London.

Hayes, C. R. & Hydes, O. D.  UK experience in the monitoring and control of lead in drinking water. J. Water Health 10 (3), 337–348.

Hayes, C. R., Bates, A. J., Jones, L., Cuthill, A. D., Van der Leer, D. & Weatherill, N. P.  Optimisation of plumbosolvency control using a computational model. Water Environ. J. 20, 256–264.

93

C. R. Hayes et al.

|

Optimization of plumbosolvency control

Hayes, C. R., Incledion, S. & Balch, M.  Experience in Wales (UK) of the optimisation of ortho-phosphate dosing for controlling lead in drinking water. J. Water Health 6 (2), 177–185.

Health Canada  Guidelines for Canadian Drinking Water Quality: Guideline Technical Document – Lead. Water Quality and Health Bureau, Healthy Environments and Consumer Safety Branch, Health Canada, Ottawa, Ontario.

Health Canada  Guidance on controlling corrosion in drinking water distribution systems.

Health Canada  Guidelines for Canadian Drinking Water Quality. Federal-Provincial-Territorial Committee on Drinking Water of the Federal-Provincial-Territorial Committee on Health and the Environment, Ottawa, December 2010.

Health Canada  Guidelines for Canadian Drinking Water Quality, Summary Table. Federal-Provincial-Territorial Committee on Drinking Water of the Federal-ProvincialTerritorial Committee on Health and the Environment, Ottawa, August 2012.

International Water Association  Best Practice Guide on the Control of Lead in Drinking Water. IWA Publishing, London.

International Water Association  Code of Practice on the Internal Corrosion Control of Water Supply Systems.

Water Quality Research Journal of Canada

|

49.1

|

2014

Specialist Group on Metals and Related Substances in Drinking Water. IWA Publishing, London.

Kuch, A. & Wagner, I.  A mass transfer model to describe lead concentrations in drinking water. Water Res. 17 (10), 1303–1307.

Rhode Island Department of Health  Effect of Partial Lead Service Line Replacement on Total Lead at the Tap. Office of Drinking Water Quality, Providence (RI), April 2011.

US EPA  Drinking water regulations; maximum contaminant level goals and national primary drinking water regulations for lead and copper, final rule. Federal Register, 40CFR parts 141 and 142. 56 (110), 26505.

Van den Hoven, T. J. L., Buijs, P. L., Jackson, P. J., Gardner, M., Leroy, P., Baron, J., Boireau, A., Cordonnier, J., Wagner, I., do Mone, H. M., Benoliel, M. J., Papadopoulos, I. & Quevauviller, P.  Developing a new Protocol for the Monitoring of Lead in Drinking Water. European Commission, Brussels, BCR Information, Chemical Analysis, EUR 19087.

Van der Leer, D., Weatherill, N. P., Sharp, R. J. & Hayes, C. R.  Modelling the diffusion of lead into drinking water.

Appl. Math. Modell. 26 (6), 681–699.

World Health Organization  Booklet on Childhood Lead Poisoning. WHO, Geneva, Switzerland.

World Health Organization  Guidelines for Drinking Water Quality, 4th edn. WHO, Geneva, Switzerland.

First received 11 February 2013; accepted in revised form 18 September 2013. Available online 15 October 2013