groundwater vulnerability assessment using process based...

GROUNDWATER VULNERABILITY

ASSESSMENT USING PROCESS-BASED

MODELS

Riitta Lindström

April 2005

TRITA-LWR PhD Thesis 1022 ISSN 1650-8602 ISRN KTH/LWR/PHD 1022-SE ISBN 91-7178-084-X

Riitta Lindström TRITA LWR PhD Thesis 1022

ii

Groundwater vulnerability assessment using process-based models

ACKNOWLEDGEMENTS

I wish to thank all those who have helped me in different ways with this work throughout the years. I am especially grateful to my supervisors Professor Roger Thunvik and Professor Per-Erik Jansson at the Department of Land and Water Resources Engineering. Thanks are also due to my former supervisor during the first phase of the project, Professor Emeritus Gert Knutsson, for his support and encouragement. A special thanks also to the late Bengt Espeby for all his help with computer techniques and for being such a good friend. During the first phase of my work, the technical section of Haninge municipality was most gen-erous with practical support. VBB Viak was also helpful, especially Håkan Djurberg who supplied me with data and imparted some of his experience. Over several years, parallel to my studies at KTH, I have had the opportunity to work with dif-ferent research issues at the Swedish Aggregates Producers Association (SBMI). The work at SBMI has widened my views and perspectives, and has helped me to improve my understanding of what groundwater vulnerability may be in practice. It has also afforded me the opportunity of meeting new colleagues and friends, whom I highly appreciate. I have also greatly enjoyed the cheerful and warm atmosphere among the staff at the Department of Land and Water Resources Engineering. It has provided me with comfort, support, and many inspiring discussions during lunches and coffee breaks. I have also received invaluable help in the final stages. In particular I would like to thank Lena Maxe and Per-Olof Johansson for comments on the manuscript, my fellow researcher Annika Lundmark for fruitful co-operation and for reviewing one of my papers, Estifanos Haile for helping me with the manuscript format for Paper II, Duncan Mc Connachie for improving my English, and my daughter Ilma who has taken care of her little sister more than usual and who also helped me to check my thesis references. Last, but of course most importantly, I am most grateful to my children for their affection, for being around and providing such a warm atmosphere at home – and Per-Olof, my beloved hus-band, for being with me and loving me and giving me support in many ways. Heartfelt thanks also go to my mother for spending time with the children and taking care of the household whenever required. Financial support for the research has been provided by the Geological Survey of Sweden (SGU), the National Road Administration, Haninge municipality, VBB Viak, and the Royal Institute of Technology (KTH), which is gratefully acknowledged. I am also most grateful to SGU for gener-ously allowing me to use and study the data in the municipal parameter database. Stockholm in April 2005 Riitta Lindström

iii iii


iv


TABLE OF CONTENT

Acknowledgements........................................................................................................................................... iii Table of Content .................................................................................................................................................v List of papers.....................................................................................................................................................vii Abstract ............................................................................................................................................................... 1 Introduction ........................................................................................................................................................ 1 Research issues ...................................................................................................................................................3 Objectives and Delimitations .............................................................................................................................4 Scientific background .........................................................................................................................................4

General concepts of vulnerability assessment...................................................................................................... 4 The purpose of vulnerability assessments............................................................................................................ 6 Approaches to vulnerability assessment .............................................................................................................. 6

Index and overlay methods........................................................................................................................................................ 6 Process-based simulation models ................................................................................................................................................ 7 Statistical methods.................................................................................................................................................................... 7

Vulnerability assessment using process-based simulation models ........................................................................ 8 Models for the unsaturated zone................................................................................................................................................ 8 Models for the saturated zone including coupled models ............................................................................................................ 10

GIS in vulnerability assessments ....................................................................................................................... 11 Decision support systems and expert systems ................................................................................................... 11 Influence of uncertainty in vulnerability assessments......................................................................................... 12

Methods............................................................................................................................................................. 13 A modelling system for groundwater vulnerability assessment .......................................................................... 14 Generic data for modelling ............................................................................................................................... 16 Modelling study for diffuse contamination........................................................................................................ 17 Modelling study for liquid spills ........................................................................................................................ 18

Results ............................................................................................................................................................... 19 Testing the modelling system for the unsaturated and saturated zone................................................................ 19 Hydrogeological database for modelling............................................................................................................ 19

The HPAR database ............................................................................................................................................................ 20 Hydrogeological environments for modelling the soil water system................................................................... 20 Influence of different soil and vegetation types on mean turnover times in the unsaturated zone ...................... 21 The role of soil hydraulic properties for infiltration of liquid spills .................................................................... 22

Discussion and Conclusions............................................................................................................................. 24 Recommendations for future research ............................................................................................................. 27 References ......................................................................................................................................................... 29 Other references................................................................................................................................................ 36

v v


vi


LIST OF PAPERS

This thesis summarises and concludes four research papers presenting possible methods to per-form and evaluate vulnerability assessment with the help of processed-based simulation model-ling. A state-of-the-art review, covering approaches to the assessment of groundwater vulnerabil-ity, formed the platform from which the research was developed (Lindström and Scharp, 1995). Since this report is ten years old, a short synopsis of the recent research in the field of vulnerabil-ity assessment is given in the summary of the thesis. References to the papers within the body of the text are referred to using roman numerals.

I. Lindström, R. 2005. A system for modelling groundwater contamination in water supply areas - chloride contamination from road de-icing as an example (accepted for publication in Nordic Hydrology)

II. Lindström, R. 2005. A hydrogeological database for flow and transport modelling with em-phasis on Swedish conditions (submitted to Water Resources Management April 2005)

III. Lundmark, A., Lindström, R. Jansson, P.-E. 2004. Effects of soil, vegetation and ground-water level on hydrogeological turnover times (submitted to Journal of Hydrology Decem-ber 2004)

IV. Lindström, R and Jansson, P.- E. 2004. Prediction of high rate infiltration in typical Swed-ish soil profiles (submitted to Journal of Hydrology April 2005)

vii vii


viii


ABSTRACT

The focus of this thesis is on groundwater vulnerability assessment by process-based simulation models and data acquisition for these assessments. A modelling system for intrinsic groundwater vulnerability assessment in water supply areas was developed, consisting of flow- and transport models for the unsaturated zone and the groundwater zone, coupled to a geographical informa-tion system. The system was applied to a water supply area located close to a major road south of Stockholm. Chloride was used as an indicator in determining the vulnerability for groundwater contamination from the road. The approach was useful to illustrate the dynamic change of chlo-ride concentrations both during the stage of continuous application and after the applications was terminated. A structure and content of a database for flow and transport modelling, based on hydrogeological environments, was outlined. An existing hydrogeological parameter database, HPAR at the Geological Survey of Sweden (SGU), was examined as a potential source of data for the new database. Values for some important parameters needed for groundwater modelling, such as hydraulic conductivity and effective porosity, were lacking in the three municipal HPAR databases that were studied. It was suggested that these data should be added, together with information on the hydrogeological environments, for all geographical positions of interest. Without such minimum information, the efficient use of modelling tools could not be expected. Typical profiles of three common Swedish hydrogeological environments (sand deposits, glacial till and clay covered areas) were used to represent generic input data to model simulations in the unsaturated zone so that the importance of soil, vegetation type and groundwater levels on turn-over times of conservative contaminants transported by natural recharge could be examined. The same profiles were used to predict the penetration depth of accidental liquid spills that occur at the land surface level. In the case of contaminant transport by natural recharge, water storage in the soil profile and vegetation type played an important role for turnover times. For liquid spills, the hydraulic con-ductivity was found to be of major importance, while the water retention properties were of less importance. Modelling, together with available data sources, were successfully used to demon-strate the vulnerability of different environmental conditions.

Keywords: Database; Groundwater vulnerability; Hydrogeological environment; Process-based models; Penetration depth; Turnover time

INTRODUCTION

Groundwater is a vital natural resource for the economic and secure provision of drink-ing water, and plays a fundamental role in human well-being. However, the pressure on groundwater, in terms of both quantity and quality, has increased to an extent whereby not only drinking water sources but also sensitive ecosystems are threatened by con-tamination through human exploitation. Increasing demand for water, a growing use of pesticides and fertilisers, atmospheric deposition and numerous point sources of contamination constitute a threat to the qual-ity of groundwater. The present groundwater

quantity and quality in Europe is summarised in a report by European Environment Agency (EEA) (1999), where it is concluded that significant problems with pesticides and nitrates in groundwater occur in several European countries. Furthermore, acidifica-tion commonly occurs in northern European countries, resulting in low pH values in groundwater. By international standards, the status of Sweden’s groundwater is good, both in terms of quantity and quality. This may be one reason why, up to now, groundwater has been a secondary consideration in the context of land use planning and environmental protection in Sweden (SEOC, 2004).

1 1


Despite groundwater’s important function to society, these resources have generally not been provided with adequate protection (Foster et al., 2002). This is mainly due to the geological formation's natural ability to at-tenuate many water contaminants and hence the soil cover has long been considered as potentially effective for the protection of underlying groundwater. The groundwater system responds slowly to contamination events and the travel times to reach the groundwater zone are often long. Cleaning and restoring contaminated groundwater is often technically problematic and costly, and finding alternative sources for water supply is not always possible (Lind-ström and Scharp, 1995). Therefore, the most effective and realistic solution is to prevent the contamination of groundwater through adequate land-use planning and groundwater management. All groundwater resources are not valued equally though, and the willing-ness to protect a groundwater resource de-pends on several issues, such as the actual use of groundwater, the potential for future water supply, alternative for water supply, and the importance of the groundwater resource for sensitive ecosystems and landscapes (Scharp et al., 1997). High dependence on groundwa-ter for water supply usually implies a strong willingness to afford the resource protection; the willingness being even higher if alterna-tive sources are not available. However, it is often necessary to define groundwater pro-tection strategies that accept trade-offs be-tween competing interests. The U.S. National Research Council, NRC (1997) provides a framework for calculating the economic value of groundwater and evaluating trade-offs between competing uses of it. Valuation of groundwater resources is critical in deter-mining an efficient outcome in many groundwater development, protection, and remediation projects, programs or policy decisions. In Sweden, the value of water has been discussed, e.g., by Sandström (1998) and Johansson et al.(2002). In recent years, groundwater vulnerability assessments have been conducted in many countries as an essential part of comprehen-sive protection strategies. Vulnerability as-

sessment provides a basis for initiating pro-tective measures for important groundwater resources and will normally be the first step in a groundwater pollution hazard assessment and quality protection, when the interest is at the municipal or provincial scale (Foster et al., 2002). However, the concept is also valid for vulnerability assessment at a more local level, within individual groundwater supply catch-ment areas. The basis for vulnerability as-sessments is the heterogeneity of geological formations, implying that aquifers are not equally sensitive to disturbances by human activities or changes in the environment. Examples of planning situations where a vulnerability assessment can be used are: general land use planning, land use restriction within well-head protection zones, localisa-tion and operation of activities endangering the groundwater quality, and design and management of groundwater monitoring networks The European Water Framework Directive (WFD) (2000/60/EC) aims to establish a framework for community action in manag-ing the water environment and sets out a common approach to protection by defining environmental objectives for all ground and surface waters within the European Union. The WFD demands sustainable water use based on the long-term protection of water resources. The term “vulnerability” is only used in relation to coastal aquatic ecosystems. However, the idea of groundwater vulnerabil-ity assessment is indirectly included in the directive. The WFD requires an initial charac-terisation of all groundwater bodies to assess their uses and the degree to which they are at risk. As part of this initial characterisation, the monitoring, evaluation and presentation of the quantitative and qualitative status of ground and surface water are mandatory. In several European countries, vulnerability studies are an integral element of the work being undertaken to implement the WFD (for application examples see Witkowsky (ed.), 2004). In April 1999, the Swedish Parliament adopted fifteen national environmental qual-ity objectives for the development of an ecologically sustainable society. The Geologi-

2


cal Survey of Sweden (SGU) is responsible for the environmental objective Good-Quality Groundwater. The Good-Quality Groundwater objective requires that groundwater must provide a safe and sustainable supply of drinking water and contribute to viable habi-tats for flora and fauna in lakes and water-courses. Adopted approaches for vulnerability as-sessment range from empirical classifications of key properties to detailed process-based simulation models. The purpose of the as-sessment governs the level of complexity required. In the present work, a process-based model approach was selected for the vulnerability assessment studies. The main concern in the research work was the vulner-ability assessment of important groundwater resources in Sweden used for drinking water production. The great importance of these resources implies that it is usually desirable to develop quantitative assessments, e.g., in respect to transport times and contamination concentrations. Due to the large data needs for process-based models, data acquisition as input to these models is crucial.

RESEARCH ISSUES

The usefulness of qualitative results from stratigraphic zoning and from index and overlay methods have been questioned. On the other hand, quantitative assessments, expressed in terms that are relevant and understandable for the user, have been re-quested (Maxe and Johansson, 1998). Proc-ess-based quantitative vulnerability assess-ments are needed for more accurate assessments. This is particularly important for safeguarding of the present and future municipal drinking water supplies. Thus, the main focus of the present research work was on the development of a vulner-ability assessment, using process-based mod-elling, for the important groundwater re-sources in Sweden. The review of current literature revealed that basic questions that need to be answered in quantitative vulnerability assessment include: - How deep/far does the contaminant reach in the subsurface;

- How long does it take until the contamina-tion reaches the target; - What will the concentration level at the target be; and - For how long will the contamination last? As the intrinsic vulnerability of groundwater to contamination is, by definition, independ-ent of both the contaminant nature and the contamination scenario, the basic concepts to be considered in order to answer the ques-tions above include: - Advective transport time from the origin to the target; - Physical attenuation, e.g. by dispersion, dilution and dual porosity effects; and - Relative concentration of contaminants, which can reach the target. An important issue in process-based model-ling is the acquisition of relevant input data. Ideally, all data used for contaminant trans-port modelling should be site-specific. How-ever, this is often unrealistic due to the costs involved in the data collection process. One possible means of coping with this lack of data is to instead use generic data from simi-lar hydrogeological environments, stored in existing databases. These data could be ana-lysed and presented as probability distribu-tions. Establishment of a database based on hydrogeological environments would pre-sumably offer an easily available source of hydrogeological information for modellers. Modelling of contaminant transport through the unsaturated zone is a critical part of any vulnerability assessment. The behaviour of the soil system is very dynamic and depends on a complex interaction between soil, plant and atmosphere. These properties vary natu-rally between different hydrogeological envi-ronments. The hydrogeological environment with, e.g., geology, position in landscape, drainage conditions, groundwater table and vegetation, play an important role in the contaminant transport process. In most cases there is a strong link between these factors, and a better understanding of the relation-ships between them could serve as useful knowledge when available data are insuffi-cient.

3


In light of the above, investigating the fol-lowing questions have been the main focus of research in the present work: • How can subsurface flow and transport

models be used in vulnerability assess-ment? (Paper I);

• How can hydrogeological environments be used by process-based models in vul-nerability assessment? (Paper II);

• How can data be derived for vulnerability studies with process-based models when site-specific input data are scarce or miss-ing? (Paper II);

• How do the physical and hydraulic prop-erties of soil, vegetation and the presence of groundwater effect the turnover times in the unsaturated zone when considering natural recharge? (Paper III); and

• Which soil hydraulic properties are most important for the contamination spread-ing pattern with dealing with large spills? (Paper IV)

OBJECTIVES AND

DELIMITATIONS

The overall objective was to develop meth-ods for the quantitative assessment of groundwater vulnerability using process-based models, with specific reference to Swedish conditions. The specific objectives were to: • Develop and demonstrate the use of a

process-based modelling system in groundwater vulnerability assessment (Paper I);

• Outline a framework for a hydrogeologi-cal database for groundwater modelling, based on hydrogeological environments and an existing Swedish database, by identifying existing data that is suitable for vulnerability studies using process-based modelling (Paper II); and

• Evaluate the impact of soil hydraulic and physical properties on turnover times re-garding contaminant transport with natu-ral recharge (Paper III) and on penetra-tion depths of contaminants due to liquid spills (Paper IV).

The research focused on intrinsic vulnerabil-ity assessment, implying that the geological, hydrological and hydrogeological characteris-tics of an area are taken into account, but the nature of the contaminants is not included. Thus, the specific interaction of different contaminants with topsoil, subsoil and the unsaturated soil has not been considered. Furthermore, only vulnerability assessment using process-based simulation models were applied.

SCIENTIFIC BACKGROUND

In this chapter, the evolution of the aquifer vulnerability concept is discussed. Since the focus of the present research has been on process-based simulation models as a tool for groundwater vulnerability assessment, the literature review emphasises studies in which such tools have been applied.

General concepts of vulnerability assess-ment

In hydrogeology the term vulnerability was first used in the late 1960s by the French hydrogeologist J. Margat, and since then it has been used more widely in the 1980s (Haertle, 1983; Aller et al., 1987; Foster and Hirata, 1988). Currently, the term is com-monly used all over the world. A common definition of groundwater vulnerability has not been agreed upon and various definitions of vulnerability have been proposed. Most of them are quite similar. The report COST 65 (1995) from European Commission presents an overview of the various definitions. One often-used definition is from (NRC, 1993): “Groundwater vulnerability is the tendency of or likelihood for, contaminants to reach a specific posi-tion in the groundwater system after introduction at some location above the uppermost aquifer”. The basic premise underlying the concept of aquifer contamination vulnerability is the variation of groundwater recharge mecha-nisms and the natural attenuation capacity of soil and subsoil profiles. Thus, instead of applying universal controls over potentially contaminating land uses and effluent dis-charges, it is more cost effective to vary the type and level of control according to this attenuation capacity (Foster et al., 2002).

4


Groundwater vulnerability, as defined in most studies, is not a characteristic that can be directly measured in the field (Gogu and Dassargues, 2000). It is an idea based on the fundamental concept “that some land areas are more vulnerable to groundwater con-tamination than others” (Vrba and Zaporo-zec (eds.), 1994), according to the attenuation capacity. Therefore, it is often defined as an intrinsic property of the groundwater system that takes into account the inherent hydro-geological characteristics of an area. As in-trinsic vulnerability does not refer to a spe-cific contaminant, only the properties that are relevant for all types of contaminants are considered. The main attributes used for the intrinsic vulnerability assessment are groundwater recharge, soil properties and characteristics of unsaturated and saturated zones (Foster, 1987 and 2002; Gogu and Dassargues, 2000). Daly et al. (2002) and Zwahlen (ed.) (2003) presented three basic aspects, proposed by the European working group in the field of scientific and technical research, COST Action 620, that should be considered in order to quantify the intrinsic vulnerability: the advective transport time, the relative quantity of contaminants which can reach the target, and the physical attenua-tion (dispersion, dilution, dual porosity ef-fects, etc.). Some vulnerability concepts also include fluxes, concentrations and characteristics of contaminants, and are referred to as specific groundwater vulnerability (Lindström and Scharp, 1995; Daly et al., 2002; Zwahlen (ed.), 2003). Furthermore, the vulnerability of an area can either be determined solely as an effect of the vertical transport of contami-nants in the unsaturated zone, or may also include the horizontal transport within the saturated zone. In the latter case, vulnerability is related to water quality at a specific loca-tion in the groundwater zone, such as a well (Johansson et al., 1993, Zwahlen (ed.) 2003). In COST Action 620 (Zwahlen (ed.), 2003), the concept of vulnerability is proposed to be based on an origin-pathway-target model for environmental management. For resource protection, the groundwater table in the aquifer is regarded as the target, and the

pathway consists of the most vertical passage possible from the ground surface to the groundwater table. For source protection, the water in the well or spring is the target, while the pathway additionally includes the most horizontal flow route in the aquifer. Recently, new terminology in connection to vulnerability assessments has started becom-ing formalised. A groundwater resource pollution hazard relates to the probability that groundwater in the aquifer will become con-taminated with concentrations above the corresponding guideline value for drinking water quality (Foster et al., 2002). Hazards are the activities and developments, which pose a threat to groundwater (Daly et al., 2002). The hazard is normally assumed to originate at the ground surface (potential release of a contaminant). The risk of contamination of groundwater also includes the potential con-sequences of a contamination event (Zwahlen (ed.), 2003). SRA (1998) and Back and Rosén (2001) described the groundwater risk assessment as a way of considering the probability and economical consequences of contamination, the intrinsic vulnerability being a part of the assessment. Fedra (1998) presented an overview of integrated risk assessment and management using dynamic models and Geographical Information Sys-tem (GIS). Wladis et al. (1999) applied a monetary risk-based decision analyses (on a national scale) to compare alternative actions designed to protect groundwater from fur-ther degradation. Comprehensive reviews of the vulnerability assessment methods are presented in reports by the NRC (1993), U.S. EPA (1993), Vrba and Zaporozec (1994) and Lindström & Scharp (1995). More recent summaries of existing vulnerability mapping methods can be found in Gogu and Dassargues (2000) and Zwahlen (ed.) (2003). A comparison of groundwater vulnerability assessment tech-niques was performed by Gogu et al. (2003). A large number of different applications using vulnerability assessment and mapping are presented in Witkowsky (ed.) (2004). Burkart et al. (1999) present examples of the application of index and overlay, process-based modelling and statistical methods for

5


groundwater vulnerability assessments for a variety of data from Midwest U.S.

The purpose of vulnerability assessments

Vulnerability assessment is a general planning and decision-making tool. The objective of vulnerability assessment is to direct regula-tory, monitoring, educational and policy development efforts to those areas where they are most needed for the protection of groundwater quality. Often the purpose of groundwater vulnerability assessment is to differentiate between areas that need protec-tion from potential contaminating activities, and areas where such activities would consti-tute a minor threat to the groundwater. Vul-nerability assessments can be included within the traditional efforts for groundwater pro-tection. Hence, they are meant to be included within a protection strategy and not consti-tute a single tool (Lindström and Scharp, 1995, Foster et al., 2002). Combined with information on the pollution hazard from different sources, areas where more in-depth studies will be carried out and plans for pre-ventive actions to be taken can be pointed out. Many groundwater professionals have argued about the usefulness of vulnerability assess-ments. It has been claimed that the hydro-geological conditions are too complex to be encapsulated by any vulnerability tool. It has also been questioned if it is possible to pre-sent a single, integrated vulnerability index or if it is necessary to work with specific vulner-abilities for individual contaminants. Scien-tifically, it is more consistent to evaluate vulnerability to contamination by each con-taminant or group of contaminants. How-ever, the implication would be an atlas of maps for any given area, which would be difficult to use in most applications (Foster et al., 2002). Moreover, there will normally not be adequate data and/or sufficient human resources to achieve this ideal. The NRC (1993) outlined three “laws” of groundwater vulnerability that should be spelled out explicitly with every vulnerability assessment:

• All groundwater is to some degree vul-nerable;

• Uncertainty is inherent in all vulnerability assessments; and

• In the more complex systems of vulner-ability assessment, there is a risk that the obvious may be obscured and the subtle may become indistinguishable.

The latter point refers to the danger, espe-cially when using complex vulnerability as-sessment tools, that in light of the final vul-nerability ranking one may lose sight of the data used for the analysis and of the assump-tions underlying vulnerability assessment schemes. However, in spite of these reservations, vul-nerability assessments are often recom-mended as an initial step in groundwater protection (e.g. by NRC, 1993; U.S. EPA, 1993; Lindström and Scharp, 1995; Vrba and Zaporozec, 1995; and Zwahlen (ed.), 2003).

Approaches to vulnerability assessment

There is no universal methodology for groundwater vulnerability assessment, al-though a number of different approaches exist. The methods are usually grouped into three major categories: (1) index and overlay methods, (2) methods employing process-based simulation models and (3) statistical models. The same categories will be used here. Each category has advantages and limi-tations, and none are considered to be most appropriate for all situations. They serve different purposes and have different aims.

Index and overlay methods Index and overlay methods are based on the assumption that a few major parameters largely control groundwater vulnerability, and that these parameters are known and can be evaluated. These methods are, in general, based on limited basic data, used in regional studies, and usually cover extensive areas (Lindström and Scharp, 1995). The ground-water vulnerability evaluated is qualitative and relative. Scoring, integrating or classifying to produce an index, rank or class of vulnerabil-ity, interprets the information. The simplest overlay systems identify areas where parame-

6


ters indicating high vulnerability coincide, e.g. shallow groundwater and sandy soil. More sophisticated systems assign numerical scores based on several parameters. The most commonly used of these methods, DRASTIC (Aller et al., 1987), uses a scoring system based on seven hydrogeological char-acteristics of a region. Several other overlay and index systems for groundwater vulner-ability exist. Examples of these methods are GOD (Foster, 1987), SINTACS (Civita, 1994) and EPIK (Doerfliger et al. 1999). Typically, such systems include variables related to groundwater recharge rate, depth to the groundwater table, and soil and aquifer properties. In general, index and overlay methods rely on simple mathematical repre-sentations of expert opinion and not on process representation. The advantage of these methods is that they provide relatively simple algorithms or decision trees to inte-grate a large amount of spatial information into maps of vulnerability classes or indexes. Examples of vulnerability studies by index and overlay methods, conducted in Europe and elsewhere, are presented in Witkowski (ed.) (2004). The methods are particularly suitable for use with GIS. However, the lack of a physically-based, precise definition also has some drawbacks. The results tend to be subjective. If various methods are tested in one area, the resulting maps are often differ-ent and sometimes contradictory (Ferreira and Oliveira, 2003; Gogu et al., 2003).

Process-based simulation models Process-based simulation models are used for examining vulnerability from a quantitative point of view and for establishing clearly identified reference criteria for quantification, comparison and validation purposes. Proc-ess-based models use current scientific un-derstanding to incorporate the most impor-tant and relevant processes, using the governing equations for water flow and sol-ute transport. The focus is on computing travel times or concentrations of a contami-nant in the unsaturated and groundwater zones. Most modelling efforts are aimed at predicting the consequences of a proposed action and, thus, can be used for making land use planning decisions. Models can also be

used in an interpretative sense to gain insight into the controlling parameters in a vulner-ability assessment. Additionally, models can be used to study processes in generic hydro-geological settings. For example, generic models can be used to analyse flow in hypo-thetical hydrogeological systems and they may be useful in helping to define regulatory guidelines for a specific region (Anderson and Woessner, 1992). They are also useful for different predictions involving contamination hazards at specific sites. However, their need for extensive data input and the expertise required to implement them may, in some cases, limit their use over large areas (e.g. Lindström and Scharp, 1995 and Refsgaard et al., 1999).

Statistical methods Statistical methods are the least common category of vulnerability assessment methods found in the literature. Although statistical studies are used as tests for other methods and geostatistical methods, such as kriging, are frequently used to describe the distribu-tion of water quality parameters, very few vulnerability assessment methods are directly based on statistical methods. Statistical meth-ods are used to quantify the vulnerability of groundwater contamination by determining the statistical dependence or relationship between observed contamination, observed environmental conditions that may or may not characterise vulnerability (e.g. unsaturated zone properties or recharge) and observed land uses that are potential sources of con-tamination (e.g. fertiliser application and septic tank occurrence). Once a model of this dependence or the relationship has been developed with statistical analysis, the prob-ability of contamination can be evaluated. Knowledge of significant environmental conditions is required for the area in ques-tion. In statistical methods the vulnerability is expressed as contamination probability. The higher the contamination probability, the higher the vulnerability. Furthermore, the statistical significance of the results can be explicitly calculated. This provides a measure of uncertainty in the model. The disadvan-tage is that statistical methods are difficult to develop and once established, can only be

7


applied to regions that have similar environ-mental conditions to the region for which the statistical model was developed. Teso et al. (1996) used a logistic regression model and GIS to predict groundwater vulnerability to pesticides. A statistical method, CALVUL (Troiano et al., 1999), is used to determine the groundwater vulnerability due to pesticide leaching in California. In Texas, Evans and Maidment (1995) developed a similar statisti-cal analysis using nitrate in groundwater as the basis for delineating vulnerable ground-water areas.

Vulnerability assessment using process-based simulation models A merit of process-based simulation models in vulnerability assessment, compared to index and overlay methods, is that the results are quantitative in terms of, e.g., travel times, leachate concentrations and critical loads. Process-based models are efficient tools for the analysis of groundwater problems and groundwater resources management. Models help groundwater resources managers to develop a shared conceptual understanding of complex natural systems, allow testing of management scenarios, predict outcomes of high risk and high cost environmental ma-nipulations and set priorities (Caminiti, 2004). A large spectrum of models exist today for predicting flow and contaminant transport in the subsurface. Addiscott and Wagenet (1985) made a distinction between mechanis-tic and functional models (also called box-models). A mechanistic model uses current scientific understanding to incorporate the most fundamental (i.e. mathematical-physical) descriptions of the important and relevant processes using the governing equa-tions for water flow and contaminant trans-port. A functional model uses simplified, less mechanistic treatments of some or all the relevant processes. By using a simplified, less mechanistic approach, the number of input parameters for the model may be reduced. In some cases, a clear distinction between mechanistic models and functional models may be difficult to make. Thus, in practice, the distinction is made between more or less mechanistic models.

Limiting factors for the effective use of simu-lation models for groundwater resources management purposes, e.g. vulnerability studies, and decision-making include the difficulty to manage large volumes of soil and groundwater data and to account for the spatial heterogeneities that pervade in natural environmental systems. The dependence on the existence of abundant input data implies that when data is scarce, the obtained results will suffer from substantial uncertainty. The simple functional models are attractive be-cause they require relatively less data, while the predictive capability of these models is questionable due to the semi-empirical nature of the process description. On the contrary, the key problem in using the more complex models operationally lies in their generally large data requirements (Refsgaard et al., 1999). Hence, applying transport models for management purposes implies a trade off between the wish to describe the involved processes as well as possible, and the need to design a not too complicated management tool. Consequently, important concerns in the use of models are the choice of modelling approach and the choice of the degree of complexity. The purpose of the study should determine what type of model to develop or to choose. The customary division of the subsurface into two zones separated by the groundwater table (in the case of unconfined aquifers) is artificial because there is a continuous ex-change of water between the unsaturated and saturated zones. However, in most situations it is common to model the two zones sepa-rately. Unsaturated flow and transport mod-els are more difficult to apply than models for the saturated zone, owing not only to the complicated plant-soil-water interactions and the complexity of unsaturated flow, but also to the numerical problems that usually are encountered in solving the highly non-linear governing equations for unsaturated flow (Andersson and Woessner, 1992).

Models for the unsaturated zone A number of models of varying complexity describing the contaminant transport in the unsaturated zone have been developed since the 1970s and applied in numerous pollution

8


studies. Models for the unsaturated zone are often one-dimensional, which means that they only consider vertical flow. Many are based on semi-empirical process descriptions of the lumped functional type while others use more physical descriptions based on Richards’s equation of flow through porous media and the advection-dispersion equation for movement of inorganic pollutants. For vulnerability assessments, these models are sometimes applied to evaluate the travel time through and attenuation capacity within the unsaturated zone above the ground water table. Subregions or areas with similar un-saturated zone properties are selected and grouped. One simulation is performed for each group or section and the results are mapped in various ways, e.g. a map of the travel time to the water table, a map of the percent removal of a contaminant within the unsaturated zone, etc. Alternatively, the dif-ferent travel times can be transformed to various vulnerability classes. The flow and transport models can be used both in regional and site-specific studies. However, because of the difficulties in find-ing relevant data for large areas, they have mostly been applied at the point, field and catchment scale, even though examples of both the functional and mechanistic models being applied at the regional scale exist as well. Point source and non-point source (NPS) pollutants differ in the scale of the areal ex-tent of their source. Point source pollutants are those isolated to a single point location, such as a hazardous liquid spill. In contrast, NPS pollutants are spread across broad areas. Obviously, land use decisions related to NPS groundwater contamination are very different from those for point sources. Large attention in vulnerability assessments by process-based models has been given to the assessment of NPS groundwater con-tamination from past, present and future use of agrochemicals. These chemicals have the potential to pose problems that have signifi-cantly greater economic effects than those that have long been recognised from point

sources (Loague et al., 1996), as large areas can be negatively affected. The simplified functional models include NPS models, such as the pesticide leaching models CMLS (Nofzinger and Hornsby, 1986), GLEAMS (Leonard et al., 1986; Knisel et al., 1989), CREAMS (Knisel, 1980; Knisel and Williams, 1995), PELMO (Klein, 1995) and PRZM3 (Carsel et al., 1998). Groundwa-ter contamination from nitrate leaching has been studied in Holland using a simplified regional nitrate-leaching model RENLEM2 (De Vries et al., 1987; Kragt and Hack-ten Broeke, 1991). In Germany, regional scale modelling of vulnerability to nitrate contami-nation was conducted by Wendland et al. (1998) and Gömann et al. (2005), who inte-grated an agricultural economic model with two hydrological models. Another model developed at the regional scale is the phos-phate transport model, REPTRAM (Schou-mans et al., 1989), that predicts the impact of application of animal manure on phosphate transport in the unsaturated zone. TETrans (Corwin et al., 1999) is a one-dimensional functional model for solute transport in the unsaturated zone, developed for prediction of salt loading to groundwater from agricultural land below the root zone. VULK (Jeannin et al. 2001) is an analytical model developed under the scope of COST Action 620 as a tool for intrinsic vulnerability assessment. The acronym VULK stands for vulnerability and karst. The conceptual framework underlying VULK comprises a simple method for transfer time mapping, both for resource and source vulnerability. Mechanistic models used for vulnerability assessment are most commonly one-dimensional NPS pollution models, such as LEACHM (Wagenet and Hudson, 1987), MACRO (Jarvis, 1994), WAVE (Vanclooster et al., 1994, 1995), UNSATCHM (Simunek et al., 1996), HYDRUS (Simunek et al., 1998), UNSAT-H (Fayer, 2000), Daisy (Abrahansen and Hansen 2000), PEARL (Tiktak et al., 2000) and the CoupModel (Jansson and Moon, 2001). For the prediction of ground-water acidification, an integrated dynamic model was developed by combining a modi-fication of hydrological model PULSE with

9


two hydrochemical models, SAFE and PROFILE (Knutsson et al., 1995). The mod-els were used for predicting the response in the groundwater chemistry to different future acid deposits scenarios at the regional scale. Maxe (1999) examined three different meth-ods to assess the groundwater vulnerability to acidification and found that due to the grid scale, a mechanistic model used in regional vulnerability assessment failed to assess the least-sensitive areas, mostly because of the inability to adequately use the information and expertise available. The spatially distrib-uted leaching model for plant protection products, EuroPEARL (Tiktak et al., 2004), was tested in agricultural areas at the scale of the European Union (EU) and combined the PEARL model and European soil and cli-mate databases. Seventy five % of the total agricultural area in the EU was parameter-ised. However, Austria, Finland and Sweden could not be included in the simulations, because there was insufficient soil profile information for these countries (Tiktak et al., 2004). Models that account for preferential flow paths, such as roots and wormholes, cracks, joints and solution channels, are not com-monly found in vulnerability assessments. Yet, preferential flow can be a dominant transport phenomenon under certain circum-stances. A model of water movement and solute transport in macro porous soils, MACRO, was presented by Jarvis (1994), and Jarvis and Larsson, (1998). Beverly et al. (1999) developed a simplified unsaturated model, SMILE (simple macro pore infiltra-tion, leakage and evaporation), for providing recharge estimates to groundwater models, while a physically based approach for macro pore flow at the catchment scale was devel-oped by Christiansen et al. (2004), using a coupling of the three-dimensional MIKE-SHE (Danish Hydrological Institute, 2000 a, b) and one-dimensional Daisy models.

Models for the saturated zone including cou-pled models Flow and transport are better understood in the saturated zone than in the unsaturated zone. A number of generalised computer codes, which can be used for the simulation

of most aquifer systems, are available. For vulnerability assessments, the groundwater models are used for source protection studies assessing the travel times from the ground-water table to the well or spring and con-taminant concentration in the well or spring. Groundwater models are usually two or three-dimensional. Two examples of two-dimensional computer software, that are widely used and well documented, are the MOC model (Konikow and Bredehoef, 1978, Konikow et al., 1994) and SUTRA (Voss, 1984). The most used computer programme for groundwater modelling in three dimen-sions is MODFLOW (McDonald and Har-baugh, 1988). FEFLOW (Diersch, 1998) is an example of a very advanced model, includ-ing both unsaturated and saturated flow and transport. The three-dimensional MIKE-SHE model is an example of an integrated surface and groundwater model. In the model, all major hydrological processes occurring in the land phase of the hydrological cycle can be simu-lated. One-dimensional flow and transport in rivers and in the unsaturated zone is in-cluded. A variety of applications for different groundwater flow and transport problems using assorted computer codes are found. In some cases, computer code for the unsatu-rated zone and saturated zone are coupled, most often using a one-dimensional model for the flow and transport in the unsaturated zone and two or three-dimensional models for the saturated flow and transport. In the coupled applications, the two codes are not integrated but sequentially executed. This implies that there is no feedback from the groundwater zone to the unsaturated zone (e.g. Refsgaard et al., 1999 and Lindström, 2005). Granlund and Nystén (1998) modelled chlo-ride transport in typical Finnish aquifers vulnerable to chloride contamination, using the two-dimensional MOC model. Lindström (2005) used the one-dimensional, unsaturated MACRO model and the two-dimensional MOC model to assess the groundwater vul-nerability due to salt from road de-icing in a water supply area of Sweden.

10


Eliasson (2001) used the FEFLOW model to predict the impact of de-icing salt and acci-dental spills from a large road to an impor-tant aquifer for water supply. The model results were used in the context of a Multiple Criteria Decision Aid System. Refsgaard et al. (1999), in a study concerning groundwater contamination from nitrate leaching at a large scale, used the MIKE SHE/Daisy coupled modelling system and Christansen et al. (2004), used the same pack-age for modelling macropore flow and trans-port processes at the catchment scale. In Sweden, MIKE SHE has been used, e.g. by Gustafsson et al. (1997) and Kylefors et al. (1999), for operational water management. Demetriou and Punthakey (1999) evaluated different options for sustainable groundwater management using MIKE SHE modelling package. Recent applications of MODFLOW in vul-nerability assessments include Gumbricht and Thunvik (1996) Beverly et al. (1999), Lasserre et al., (1999), Gogu et al. (2001), and Thunqvist (2003).

GIS in vulnerability assessments For all three categories of vulnerability as-sessment methods, i.e. index and overlay, process-based simulation models and statisti-cal methods, GIS technology may provide greatly increased efficiency in data handling, analytical capability and display flexibility. Examples of areas where GIS has been used in groundwater vulnerability assessment are: (i) integration of various data layers that are involved in the vulnerability assessment, (ii) support of analysis and modelling of spatial and physical relationships of critical envi-ronmental variables and parameters and (iii) display of results in the form of maps. Ex-amples of the application of index and over-lay methods, process-based simulation mod-els and statistical methods for groundwater vulnerability assessment to a variety of data from the Midwest U.S. are presented in Burkart et al. (1999). The spatial nature of most of the vulnerabil-ity assessments make the evaluation well suited to the integration provided by GIS. However, all data used in the production of a

vulnerability map also imposes limitations. For example, the interpolation of a small number of data spread across a relatively large area can have a significant impact on the accuracy of the vulnerability map (e.g. permeability and layer thickness). The linkage of GIS and models is becoming more and more attractive due to the increas-ing capability of GIS to store, retrieve, ana-lyse and present geographically referenced spatial data. When coupled to a user-friendly graphical interface, these tools become inter-esting to users such as personnel at local, regional and national planning authorities and private consultants, and can be very useful for decision support in planning situations. In many cases, where process-based models have been used to determine groundwater vulnerability, a GIS has also been used to organise the input data and to determine homogeneous biophysical units (e.g. Sokol et al., 1993; El Kadi et al., 1994; Morse et al., 1994; Zhang et al,. 1996, Loague and Corwin, 1998; Corwin et al., 1999; Lasserre et al., 1999; Lord and Anthony, 2000; Tiktak et al., 2002; Wendland et al., 1998, 2002; and He, 2003. Corwin et al, (1997) presented a review con-cerning the modelling of non-point source pollutants in the unsaturated zone with GIS. In Sweden, simulation models together with GIS in a management context have been used, e.g., by Gumbricht and Thunvik (1996), Eliasson (2001), Sivertun and Prange (2003), Thunqvist (2003) and Lindström (2005). An example of a recent Swedish application of an expert system using key parameters and GIS in risk mapping of groundwater salinisa-tion is presented by Lindberg et al. (1996), and Gontier and Olofsson (2005).

Decision support systems and expert systems

The basic idea of an expert system (ES) is to use applied artificial intelligence technology to transfer expertise from an expert to a computer. The software users can then ob-tain advice and explanations from the system. While expert systems are usually applied to narrow and specific problems, decision sup-port systems (DSS) are more like a collection of tools and data that are used to solve prob-

11


lems that are often very broad. The integra-tion of data, models and a sophisticated user interface is a key characteristic of a DSS (Newell et al., 1990). The user can experiment with the tools and attempt to find a solution. Essentially the decision process is controlled by the user (Newell et al., 1990). In a groundwater management context, DSS seem to be more suitable to use than ES due to the large scope of the problem with various types of modelling studies, different scales of prob-lems and several types of data involved in using a model. Several DSS within the groundwater man-agement field have been developed during the past 15 years. Newell et al. (1990) de-scribed a Graphical Decision Support System for Groundwater Contaminant Modelling (OASIS). Three technologies were applied to build OASIS: graphical interfaces, hypertext and object oriented programming. Fedra (1993, 1998, and 2002) presented ap-plication examples of modelling practices, which use decision support systems and discusses the benefits of interactive decision support software for water resources in the context of risk management. An example of a DSS that draws together models, spread-sheet, expert systems, neural network, GIS and databases in a programmable system is RAISON (Lam and Swayne, 1993, Booty et al., 2001). The U.S. EPA’s Pesticide Assess-ment Tool for Rating Investigations of Transport (PATRIOT) is a DSS for provid-ing rapid analysis of groundwater vulnerabil-ity to pesticides on a local, regional or state level. MACRO-DB (Jarvis et al., 1997) de-scribes a decision support tool for assessing pesticide fate and mobility in soils. MACRO-DB consists of soil, pesticide, climate and crop databases linked to parameter estima-tion routines and a simulation model, MACRO (Jarvis, 1994). An often-cited source on groundwater deci-sion support is a set of papers by Freeze and co-workers (Freeze et al., 1990, Massmann et al., 1991, Sperling et al., 1992, Freeze et al., 1992), where a decision model is presented together with example situations where it has been applied. The system helps to systemati-cally evaluate risk, uncertainty and costs for

various management alternatives. The total costs are calculated via an objective function to find the cheapest alternative.

Influence of uncertainty in vulnerability assessments Since a model is only an approximation of reality, and also because the inputs to the model are rarely, if ever, exactly known, the output of the model is also likely to deviate from reality. Hence, uncertainty assessments are necessary and should include uncertainty on model structure, parameter values, etc. (Refsgaard and Henriksen, 2003). It is impor-tant to know how large the uncertainties in the model outputs are, particularly when the model is used for predictive purposes (Heu-velink, 1998, Beven 2002). Uncertainty analyses can help to identify which attributes require more accurate meas-urements in order to reduce overall uncer-tainty, identify attributes for which less pre-cise information is required and thereby reduce data collection efforts, and determine whether a simpler approach would be suffi-cient or if a more sophisticated approach is required for better reliability (Heuvelink, 1989). Examples of uncertainties inherent in all approaches to groundwater vulnerability studies are presented in, e.g., NRC (1993), Loague and Corwin (1998), Dubus et al. (2003), and Tiktak et al. (2004). These are model related errors, which include uncer-tainty resulting from inadequate or incom-plete representation of the system processes, and data related errors, which include uncer-tainty resulting from errors in inputs and parameters. Input error can arise from meas-urement, position and/or synchronisation errors. Parameter error has two possible origins. For models requiring calibration, parameter errors are usually the result of model parameters that are highly interde-pendent and non-unique. For models with physically based parameters, parameter error results from an inability to represent areal distributions on the basis of a limited number of point measurements (Loaque and Corwin, 1998). In process-based modelling, the model upscaling may also be a source of errors

12


(Heuvelink, 1998). Many of the processes that are studied, and for which models are built, take place at various temporal and spatial scales (Refsgaard et al., 1999). Differ-ent scales for modelling, adopted from

Refsgaard et al. (1999), include point (local) scale, field scale (50-200 m), grid scale (1-5 km), catchment scale (10-50 km) and larger scales, such as a regional or continental scale. Models for vulnerability assessments are used at different scales and scale-specific errors may occur because different processes are important at different scales. The input data density is reduced at large scales and the support of the inputs and outputs of a model changes with a change in scale, and this af-fects the relationships between them (Heu-velink, 2003). The use of GIS is also associated with differ-ent kinds of errors. Sources of errors result-ing from the inappropriate use of GIS are presented by Burrough (1986). In recent years, a great deal of attention has been paid to the propagation of errors in process-based modelling and several tech-niques are available to carry out uncertainty analyses (e.g. Janssen et al., 1994 and Wingle et al., 1996). One can, e.g., use classical sensi-tivity analysis to measure the effect that changing one factor has on another or, first-order uncertainty analysis (FOUA) based on Taylor series and statistical sampling methods that utilise a range of likely values for input parameters to assess the probable range of output parameters (e.g. Monte Carlo simula-tion). Furthermore, one can use stochastic modelling approaches that directly incorpo-rate the parameter or process uncertainties into the model itself and provide direct un-certainty estimates of model outputs (Heu-velink, 1998). Loague and Corwin (1996) provide examples of sensitivity analysis, first-order uncertainty analysis and Monte Carlo simulation. Bayesian methods may be used when uncertainties in input parameters can be specified by either judgment, or estimated from existing data bases from which input parameter values have been determined (Freeze et al., 1990, Massmann et al., 1991, Sperling et al., 1992, Freeze et al., 1992, NRC, 1993, Rosén and Gustafson, 1996).

In vulnerability assessments, the most fre-quently used methods for evaluating uncer-tainty are the FOUA-technique and stochas-tic modelling. Uncertainty analysis using different methods have been performed by, e.g., Loaque and Corwin (1998) Tiktak et al. (1999), Crosetto et al. (2000), Bärlund and Tattari, 2001, Mailhot and Villeneuve (2003), Borman and Diekkruger (2004), and Baben-dreier and Catleton (2005). Dubus et al. (2003), reviewed sources of uncertainty in pesticide fate modelling. The review suggests that the probabilistic ap-proaches, which are typically being used to account for uncertainty in pesticide fate modelling, such as Monte Carlo modelling, ignore a number of key sources of uncer-tainty. The result of doing so is likely to have a significant effect on the prediction of envi-ronmental concentrations for pesticides (e.g. model error or modeller subjectivity). Progress in the domain of uncertainty analy-sis is crucial if robust estimates of uncertainty in model predictions are to be obtained (Du-bus et al., 2003). Uncertainty will limit the use of a model for making regulatory decisions unless the uncertainty is included into the decision making process (e.g. Jury et al., 1987; Greig-Smith, 1992; and Costanza et al., 1992). Disclosure of the sources and magnitude of the uncertainty is needed for the decision maker to understand with what confidence the decisions are being taken.

METHODS

The research work has a general and applied character, and strives to shed light on the potential of process-based models for groundwater vulnerability assessment. A literature review of current approaches to groundwater vulnerability assessment was the starting point of the research (Lindström and Scharp, 1995). The introductory part of the thesis summarises the concept of vulnerabil-ity assessment of groundwater, providing an updated literature review of process-based models used in groundwater vulnerability assessment. Paper I describes a study in which a model-ling system for assessing groundwater vulner-

13


ability was developed and tested in a munici-pal water supply area. The system included two different models for the unsaturated zone and the groundwater zone. GIS was used as a pre-and post-processor for the groundwater part of the system and a graphi-cal user interface was developed to make the system user-friendly. Paper II outlines the structure and content of a new hydrogeological database for flow and transport modelling based on hydrogeologi-cal environments. An existing hydrogeologi-cal parameter database is explored as a poten-tial source of data and a user interface for the parameter database is presented. In Paper III, typical soil profiles for three different soils and vegetation properties for four vegetation types are described, thereby supplying generic input data to model simula-tions in the unsaturated zone. The profiles were defined for three typical hydrogeologi-cal environments in Sweden, including the groundwater table and vegetation. Moreover, the paper discusses the impact of the soil hydraulic properties, vegetation properties and different groundwater levels on the turn-over times of water conservative contami-nants in the unsaturated zone. Paper IV demonstrates the use of the typical soil profiles characterised in Paper III when predicting the penetration depth of liquid spills that occur at the land surface. The relative importance of soil hydraulic proper-ties for contamination spreading regarding large spills is examined. Moreover, estimates of penetration depths for comparatively large volumes of liquid in three typical soil profiles (sand, glacial till and clay) are presented. Thus, the thesis had two main themes: (1) vulnerability assessment using process-based models, and (2) data acquisition for these assessments.

A modelling system for groundwater vulnerability assessment

The potential contamination is assumed to originate at the land-surface. For the purely intrinsic vulnerability assessment, a conserva-tive contaminant is taken as a reference. The main target of the vulnerability assessment is

the groundwater table (resource protection). This enables a consistent investigation of the total recharge area. For the protection of a particular well or spring (source protection), the lateral movement in the saturated zone to the source is also considered. A general, conceptual modelling of the transport of contaminants in the subsurface is described in Figure 1, which involves: (1) Computation of a water budget at the land surface based on input/output fluxes in order to assess actual infiltration; (2) Use of the infiltration (and associated mass of contaminant) as an upper boundary condition for modelling one dimensional vertical transport of a dissolved contaminant through different layers of the unsaturated zone; and (3) Use of the computed fluxes and concen-trations at the base of the unsaturated zone as input to a two or three-dimensional groundwater flow and transport model of the aquifer. An outline for a groundwater vulnerability assessment centered on process-based simu-lation models, including a GIS and a hydro-geological database, is illustrated in Figure 2. In order to make the use of the included components easier, they may be supported by a graphical user interface. The primary components and features of a system for assessing groundwater vulnerabil-ity in water supply areas was developed and tested (Paper I). The system can be used, e.g., for obtaining insight into the potential groundwater contamination from roads, such as de-icing salt. The system consists of the flow and transport models, MACRO (Jarvis, 1994, Jarvis and Larsson, 1998) for the unsaturated zone, and MOC (Konikow and Bredehoeft, 1978; Konikow et al., 1994) for the groundwater zone, coupled to the GIS, IDRISI (Eastman, 1992, 2001). The general simulation procedure is illustrated in Figure 3. MACRO is a one-dimensional model includ-ing a number of options allowing the user flexibility in matching the simulation to the particular conditions that are of interest. The most important options relate to pre-

14


1D unsatmodel

2D or 3D saturatedmodel

RESOURCE

SOURCEOrigin of potential contaminationland-surface

1st targetgroundwater surface

2nd targetspring, well

2nd pathway

1st pathway

I

PE&T

R

f,

Figure 1. Illustration of a groundwater vulnerability assessment, based on an origin-pathway-target conceptual model, showing the main conceptual processes affecting transport of dis-solved contaminant from the soil surface to the spring or pumping well: (modified a ter Cost action 620 2003).

cipitation input (hourly or daily rainfall data), the type of solute, the type of vegetation and the soil type. MACRO includes an option for preferential flow. MOC is two-dimensional and used for steady-state or transient simula-tions. The model incorporates a first-order, irreversible rate-reaction, linear and non-linear equilibrium-controlled sorption, and equilibrium-controlled ion exchange. Owing to the one-dimensional nature of the unsatu-rated zone model, multiple realisations are needed to account for the heterogeneity in an area. The study area has to be divided into homogeneous sub-areas with respect to the land use and the soil type. The resulting time series of fluxes and concentrations from the unsaturated zone are used as input to the groundwater zone model. The GIS contains all the spatially distributed data. It is mainly used to prepare and to transform, e.g., the model input data into a form suitable for groundwater modelling. In

addition, the GIS is used to present and to further analyse the model results. The data prepared by the GIS and the data from the unsaturated zone model (daily fluxes and concentrations), are linked to the groundwa-ter model to simulate flow and transport in the groundwater zone. The results can be viewed using the interface.

Process-basedmodels

Site-specific and Genericdatabases

GIS

User interface

Figure 2. Outline of a computerised groundwater vulnerability assessment system.

15


BOUNDARY

RECHARGE

TRANSMISSIVITY

SAT. ZONE THI

INITIAL HEADS

INITIAL CONC

MACRO

MOC

PE

FLUXES

GROUNDWATER HEADS

PARAMETER DATA

CONCENTRATIONSTRAVEL TIMES

GIS,IDRISI

MAPS

MAPSSOLUTE CONC

dy.

Figure 3. Graphical illustration of themodelling system used in the stu

For the simulations, the MOC-model was modified to facilitate the data exchange be-tween MOC and IDRISI. Furthermore, a particle tracking routine, allowing both for-ward and backward determination of particle path lines, was developed. The routine en-ables the delineation of capture zones for water supply wells and presentation of the advective transport from a contamination source. In order to facilitate the exchange of data between the two models (MACRO and MOC), the MOC source code was altered. The system was applied to a water supply area located close to a major road south of Stockholm (Paper I). The Jordbro study area (Figure 4), which consists of gravel and sand, is highly permeable. The total thickness of the esker deposits that underlie the entire study area varies between 10-30 m across most of the area. The unsaturated thickness varies between 2-10 m. Nynäsvägen road, which transverses the study area over a dis-tance of 1.5 km along the esker formation, constitutes a potential risk for contamination of the groundwater resource as a whole and

also the drinking water source (i.e. the water supply well that is located a few hundred meters away from the road). Chloride was used as an indicator in determining the risk for groundwater contamination from the road. The future chloride concentration in the aquifer was predicted and the effects of different pumping rates on the chloride con-centrations in the water supply well were tested.

Generic data for modelling

The structure and content of a generic data-base for flow and transport modelling based on hydrogeological environments was out-lined (Paper II). Typical data needed for process-based modelling were presented, including different key hydrogeological pa-rameters in different hydrogeological envi-ronments and data relating to contaminant transport. An existing hydrogeological digital database, the Hydrogeological Parameter Database (HPAR), which is located at the Swedish Geological Survey, was examined as a poten-

16


Stockholm

Haninge,

pilot area

Haningestudy area

Well

Road

Lake Lillsjön

N

Meters500

ResidentialIndustrial

Sweden

StreamNytorpsbäcken

Figure 4. The Jordbro study area.

tial source of parameter data for the outlined database. A graphical user interface for the HPAR was developed. The interface can be used for viewing the parameter data, per-forming statistical analyses of the parameter data and, finally, storage of the data in a new hydrogeological database for flow and trans-port modelling. The interface was designed using Visual Basic. SGU’s digital databases from Heby, Västerås and Hallstahammar municipalities were used to test the new interface. The concept of hydrogeological environ-ments was presented as a potential primary basis for obtaining generic data for ground-water modelling. A hydrogeological environ-ment is defined as an area delineated by using its characteristic hydrogeological conditions. It is therefore a composite description of all the major geological and hydrological factors, which affect and control groundwater move-ment into, through and out of an area. In the context of this work, hydrogeological environments are defined as groups, includ-ing hydrogeological settings or subunits (Fig-ure 5). Hence, a hydrogeological environment can consist of several different settings with local differences in stratigraphy.

Modelling study for diffuse contamina-tion

The influence of different soil hydraulic and vegetation properties on the mean turnover times for conservative contaminants trans-ported by natural recharge in the unsaturated zone, was evaluated for different groundwa-ter levels (Paper III). Instances where groundwater levels were 0.5 and 2.5 m below the soil surface were compared. The com-parison also included unsaturated conditions, i.e. the groundwater table below the model-ling region (4 m). The contaminants moving with water were considered to be transported by advection only, i.e. dispersion and diffu-sion were not accounted for. When saturated conditions existed in the soil profile, dis-charge of groundwater was calculated. The water was discharged at the depth of the groundwater level. Three different soils (sand, glacial till and clay) were combined with four vegetation types (grass, pine, spruce and arable plants) to represent typical Swedish hydrogeological environments. Each hydrogeological envi-ronment comprised a description of a repre-sentative soil profile, vegetation properties and depth to the groundwater. The descrip-tion of a soil profile included the most prob-

17


A) Delta

B) Subaquatic esker

C) Supraaquatic esker

D) Fluvial deposits

Sand

ClayGlaciofluvialmaterialTillBedrock

Glaciofluvialdeposits

i r ,Figure 5. Glaciofluvial deposits w th subunits (afte SEPA, 1999 and SRA 1998).

able layering of the soil profile and the hy-draulic properties within each layer. Carrying out a statistical analysis of soil hydraulic properties, using a Swedish soil database, allowed for the most probable layering and properties within the layers to be determined. The soil database, included in the Coup-Model, contains data from approximately 2200 soil layers, representing 260 soil profiles from different parts of Sweden. For the purpose of this study, a subset of 171 soil profiles, representing sand, glacial till and clay, were selected for the statistical analysis. The numerical calculations were performed using a physically based, one-dimensional coupled heat and mass transfer model for the plant-atmosphere system, namely the Coup-Model (Jansson and Moon, 2001). The basic structure of the model consists of a depth profile of the soil. The soil profile is divided into layers with certain thicknesses and soil properties (in the present case a 4 m deep soil profile, divided into 12 model layers, was used). Two coupled differential equations are employed to describe water and heat flow in the vertical soil profile. Sub models account-ing for interception, snow, surface water, evaporation and transpiration define the upper boundary conditions for the model.

Different options exist for the lower bound-ary, depending on whether saturated or un-saturated conditions are assumed to exist. When groundwater is present in the soil profile, water discharge from different parts of the soil profile can be calculated. An extensive description of the model can be found in Jansson and Karlberg (2004). Im-portant parts of the model that have rele-vance for this work are described in Paper III.

Modelling study for liquid spills

The role of soil hydraulic properties for con-servative contaminant transport at high infil-tration rates (liquid spills) was examined using the dynamic CoupModel (Paper IV) together with parameter values defined for typical Swedish soil profiles (Paper III) as input data. It was observed how typical soil properties, including their variability due to the hydraulic conductivity, influence the prediction of the penetration depth of liquid spills occurring at the land surface. Wet (-0.60 m initial pressure head) and dry (- 150 m initial pressure head) initial conditions in the soil profiles were considered in connection to the spill. A salt tracer was used in the simula-tions in order to evaluate how deep down the

18


i ii

Figure 6 a (left). Chloride concentration in the aquifer after applying de- c ng salt over a 10 years period. 6 b (right) Chloride concentrations when de-ic ng salt has been applied over a 10 year period and then stopped for a period of 10 years.

spill penetrated from the ground surface. The variation of the simulated penetration depths was studied using standard deviations for K-values for the different soil profiles. Due to the high infiltration rates that occur in con-nection to large spills, effects of vegetation and evaporation were assumed to be of mi-nor importance and therefore neglected The spreading area of the spill at the land surface was calculated by assuming that the area affected by the spill can be defined as the ratio between the total volume of the spill liquid on the ground surface and the total volume infiltrated.

RESULTS

Testing the modelling system for the unsaturated and saturated zone The test of the developed modelling system in the Jordbro area (Paper I) showed that the chloride concentration in the studied aquifer will increase substantially due to road de-icing and that it may take decades to lower the chloride concentration to the original back-ground values after halting the use of de-icing salt (Figure 6 a, b). The technical drinking water limit for chloride of 100 mg/L was exceeded at a distance of 100 m from the road; however, the water supply well was not affected. In the vicinity of the road, chloride concentrations were even higher. Ten years application of de-icing salt, at a rate of 10 kg/m/yr, followed by an equivalent period

with no salt application, was studied. An initial chloride concentration of 25 mg/L was assumed and the abstraction rate from the water supply well was set to 7.5 L/s. The water supply well is located approximately 600 m from the road and was affected by the de-icing salt after approximately 8 years. In the second stage, where no salt was applied, a maximum chloride concentration of about 60 mg/L was found in the water supply well nearly 5 years after halting the application of de-icing salt.

Hydrogeological database for modelling

Figure 7 outlines the basic elements of the database. The key part of the database con-sists of the probability distributions of sig-nificant hydrogeological parameter data de-fined for hydrogeological environments in Sweden. Hydrogeological environments described by the Swedish Environmental Protection Agency (SEPA 1999) are sug-gested to be used. Key parameter data on the properties of contaminants, as well as pedo-transfer functions to derive parameter esti-mates from more easily available data, e.g. for K-values, were proposed to be included in the database. This means that the database is a combination of data descriptions from both measurements and data that are part of mod-els to convert from one property to another property of the environment. Information from other data sources, e.g. on climate data, soil, vegetation and contaminant properties,

19


Figure 7. Schematic structure of the outlined database.

should be linked to the hydrogeological pa-rameter database in order to make it opera-tional and easy to use without too much data processing. In order to view and analyse the data in the database, a graphical user interface was designed. Nowadays, these user inter-faces are standard features and are often requirements for databases.

The HPAR database The study of the HPAR database (Figure 8) revealed that the structure is feasible for extraction of data to the outlined database for groundwater modelling. Any SQL-type lan-guage should be useful to handle a number of the tables. However, values for some impor-tant parameters needed for groundwater modelling, e.g. hydraulic conductivity and effective porosity, were not found in the three municipal HPAR databases that were studied. The HPAR database does make provision though for entering such data. It was therefore suggested that these data should be extracted from other sources, such as from SGU and elsewhere, and included in the HPAR database. In order to use the concept of hydrogeological environments as a basis for the retrieval of generic data for groundwater modeling, it was proposed that the hydrogeological environments should be identified and registered in the HPAR data-base for all data points.

Hydrogeological environments for modelling the soil water system

The soil hydraulic properties, vegetation properties and groundwater levels of three typical hydrogeological environments were identified and described (Paper III). Each hydrogeological environment in this study comprised a description of a representative soil profile, vegetation properties and typical depths to the groundwater. The description of the soil profile included the most probable layering and the hydraulic properties within each layer. The results from the statistical analysis of the soil hydraulic properties for the three soil types (sand, glacial till, clay) are presented in Table 1. Clay and glacial till soils in the database generally displayed a more pronounced layering with depth than sand profiles, which were more or less homogene-ous throughout the soil profile. The vertical heterogeneity of the soil hydraulic properties was most pronounced for the glacial till soil. The porosity (θs) and the hydraulic conductiv-ity (Ksat) decreased significantly with increas-ing depth, and the variation in the water retention curve within each type of soil was identified (Figure 9) and estimated according to the corresponding variation in the water content at saturation and residual levels.

20


iFigure 8. The HPAR database structure and some of the available data The boxes depict tables and their attribute fields. ID fields are used to l nk the tables.

Influence of different soil and vegetation types on mean turnover times in the un-saturated zone

The influence of different soil hydraulic properties and vegetation properties on the mean turnover times of water in the unsatu-rated zone was evaluated for different groundwater conditions (Paper III). An in-crease in the mean turnover times was shown when moving from coarser to finer textured soils and from arable land to a more water-demanding spruce forest (Figure 10). This pattern was most pronounced for unsatu-rated conditions and for the deeper (2.5 m) groundwater level, where the sensitivity to different soils was twice as large as the sensi-tivity to different vegetation types. Higher soil water storage resulted in longer turnover times. In the case of the different vegetation

types, the water storage was of minor impor-tance. However, forest vegetation generally has larger evapotranspiration values than grass and arable land vegetation. This leads to a smaller amount of water discharge from the soil profile, which in turn leads to longer turnover times. To characterise a hydrogeological environ-ment, a specific soil type, vegetation type and groundwater table were combined (in the case of glacial till soil, both shallow and deep groundwater levels were considered for the same hydrogeological environment). Sand deposits were made up of a combination of sand soil, pine vegetation and a deep groundwater table (2.5 m). The resulting mean turnover time for this hydrogeological environment was 230 days. A reasonable approximation of this mean turnover time could be obtained for deeper groundwater

21


Sand Glacial till Clay Depth (m) 0-0.1 0.1-0.5 >0.5 0-0.1 0.1-0.5 >0.5 0-0.2 0.2-0.3 0.3-0.8 >0.8 Brooks and Corey Pore size distribution, λ 0.32 0.42 0.56 0.26 0.22 0.16 0.06 0.045 0.035 0.03 Air entry, ψa (cm) 13.5 12.5 10.5 1.8 8.0 10.0 5.0 5.3 2.5 5.2 Saturation, θs (vol%) 48 43 40 76 50 33 50 47 49 48 Residual water, θr (vol%) 2.0 4.8 5.2 7.0 0.5 0.3 0.5 0.5 0.5 0.5 Wilting point (vol%) 2.5 2.5 2.5 4 4 4 16.8 16.8 16.8 16.8 Hydraulic conductivity, Ksat (mm/day) 2113 2018 1528 1291 737 34 704 411 996 304

Standard deviation Ln Ksat (mm/day) 1.64 1.62 2.09 2.52 2.20 1.79 3.03 2.90 3.32 3.31

i tt

Table 1. Results from the statistical analysis of the soil hydraulic properties for three dif-ferent soil types. Coeffic ents for the Brooks and Correy reten ion function, fitted to the arithmetic mean values of the water content at different pressure heads, and the geome -ric mean for the hydraulic conductivity are presented.

levels by scaling the turnover time for the unsaturated soil to the desired depth (e.g. a groundwater table at 10 m would have a mean turnover time of 920 days). The glacial till environment comprises a combination of glacial till soil and spruce vegetation. Both shallow (0.5 m) and deeper (2.5 m) ground-water tables were assumed to occur in glacial till, depending on the position in the land-scape. The mean turnover time for the deep groundwater table was simulated to be 430 days. For the shallow groundwater table, the mean turnover time decreased to 200 days. Clay covered areas are described as a combi-nation of clay soils, arable land vegetation and a shallow groundwater table (0.5 m). The turnover time in this environment was ap-proximately 150 days.

The role of soil hydraulic properties for infiltration of liquid spills

The relative importance of soil hydraulic properties for contamination spreading from large spills was examined, and estimates of penetration depths for liquid spills were presented (Paper IV). As expected, large variations in the simulated mean infiltrated volumes, spreading areas and penetration depths were obtained between the different soils (Figure 11). The range of variations was also substantial within a soil type as a result of the variation in hydraulic conductivity. In addition, the initial conditions were of some importance, but this was much less pro-nounced when compared to the variation due to soil type. In general, initial conditions were

0123456

0 10 20 30 40 50 600123456

0 10 20 30 40 50 600123456

0 10 20 30 40 50 60

Pres

sure

head

,pF

log(

cm w

ater

)

Water content (vol %) Water content (vol %) Water content (vol %)

Figure 9. The water reten ion curves (WRC) for different soils: till (left), sand (middle) and clay(right), at a depth of 0.5 m. The curves within the same graph represent the mean WRC and theWRCs representing one standard devia ion of the mean, calculated according to Paper III.

t

t

22


0200400600800

10001200

0200400600800

10001200

050

100150200250300350

Unsaturated 2.5 m

Deep grw2.5 m

Shallow grw 0.5 m

Drainage condition

050

100150200250300350

Unsaturated 2.5 m

Deep grw2.5 m

Shallow grw 0.5 m

Drainage condition

0.00.20.40.60.81.01.21.4

Sand Till Clay

0.00.20.40.60.81.01.21.4

Arable Grass Pine Spruce

Turn

over

tim

e (y

ears

)W

ater

sto

rage

(mm

)W

ater

dis

char

ge

(mm

/yea

r)

Turn

over

tim

e (y

ears

)W

ater

sto

rage

(mm

)W

ater

dis

char

ge

(mm

/yea

r)

t lFigure 10. The mean turnover time, water storage and mean water discharge for the different groundwater conditions with respect to the various soil ypes ( eft) and the different vegeta-tion types (right).

demonstrated to be important for soils with a low hydraulic conductivity, i.e. glacial till and clay. The variation of penetration depths between different soil types followed the expected pattern; the coarser the soil, the deeper the maximum penetration depth. An exception was clay soil with a high hydraulic conductiv-ity, which had a maximum penetration depth of approximately 5.7 m. for dry initial condi-tions. It was also observed that, although the sand and the clay soil with high hydraulic conductivities had approximately the same infiltration volume, the penetration depth in the clay was approximately 50 % that of the penetration depth in the sand soil. The dy-namic behaviour (Figure 12) showed that the greater penetration depth for the sand was obtained continuously during a 20 day-

period. Initially, infiltration depths were within the same range after the first day. Within the sand and glacial till profiles, the variation patterns were similar to the penetra-tion depth increasing in the same way with increasing hydraulic conductivity. The differ-ence in penetration depth was most pro-nounced within the clay profile and the varia-tion pattern differed from the other two soils. The clay soil, with low K-value (- 1 std), showed significant variation between wet and dry initial soil water conditions. Impacts from macro pore flow paths were shown for the clay soil when high K-values (+1 std) were used, resulting in penetration depths of up to -5.7 m. When mean and high K-values were applied, however, there was no variation between wet and dry conditions.

23


t

t

i

Figure 11. Accumulated infiltra-tion in the three soil profiles for different ini ial conditions and hydraulic conductivities, due to the hydraulic load of 5000 mm and the dep h of the infiltrated waterfront after a 20 day simula-tion per od when the infiltration has stopped and field capacity is reached.

DISCUSSION AND CONCLUSIONS

There is an obvious need and interest for vulnerability assessments as a tool in groundwater resources management. Decid-ing which method to use will depend primar-ily on the objectives of the vulnerability as-sessment. The use of process-based models for vulnerability assessments has increased during the past ten years, due to the desire to generate quantitative assessments and be-cause of the progress made with model de-velopment and database technology. The scales in which the process-based models are used have become larger, now ranging from assessments at the local (point) scale to larger

scales, and even up to nation (and continent) wide assessments. Consequently, the use of GIS to facilitate process-based modelling has also increased. GIS is now widely used to create digital geographic databases, to ma-nipulate and prepare data as input parameters for modelling and to display results. The coupling of the process-based models and GIS offers possibilities to develop new pro-cedures for the quantification of physically significant parameters using hydrogeological databases. An important concern in vulner-ability assessments, using process-based modelling, is in the area of data availability, scale and uncertainty. A great challenge exists in vulnerability assessments to correctly tackle these issues and clearly express the uncertainties in the results.

-8

-7

-6

-5

-4

-3

-2

-1

01 3 5 7 9 11 13 15 17 19 21

Day number

Pene

trat

ion

dept

h (m

)

Glacial till Sand Clay

i

Figure 12. The mean simulated depths of the contamination front (m) for the different soil types at dry initial soil condit ons.

The process-based modelling approach has a good potential to serve as a tool for ground-water vulnerability assessments. The merit with the approach is that the results pro-duced are quantitative, e.g. in terms of con-centrations and/or travel times. The model-ling system, as well as the outline of a new hydrogeological database for groundwater modelling presented in the current work, is intended to be a practical modelling aid in vulnerability assessments and can be used, e.g., in predictive modelling to simulate con-tamination scenarios and changes of groundwater quality due to protective or remedial measures.

24


Results from the Jordbro study area demon-strate the potential applicability of the pre-sented system as a tool for the evaluation of the impact of road de-icing on groundwater quality, and for assessing changes in ground-water quality due to different remedial meas-ures, such as reducing salt application. Be-cause of the conservative nature of chloride, it is a good tracer and hence it is able to show potential contamination risk in the subsur-face.As showed in other studies, e.g. Maxe and Johansson (1998) and Eliasson (2001), accidental spills on roads may also pose a severe risk for highly permeable deposits. The penetration depths for liquid spills in sand soils, as simulated in Paper IV, indicated that in the Jordbro case, where the ground-water level was located at a depth of 5 m, a conservative contaminant would reach the groundwater level within 24 hours. The modelling system is expected to be a useful tool for evaluating different groundwa-ter protection strategies. However, the mod-elling approach is data demanding. When input data are scarce, the results obtained will suffer from substantial uncertainty. The uncertainties are related to, for instance, the estimation of the recharge rates, the thickness of the unsaturated zone, and the hydraulic conductivities of the unsaturated and satu-rated zones if the number of measurement points are scarce. Using GIS to deal with the point values and thereby expressing the spa-tial representation of the modelling, is a pow-erful method. However, procedures for carrying out uncertainty assessment should be included and also expressed by the GIS methods. Moreover, thorough sensitivity analysis of modelling results is strongly rec-ommended before using the results for plan-ning and management purposes. In a specific modelling situation, it is important to be aware of the models limitations and their consequences for the results. The value of modelling applications ultimately depends on the people who apply the models, their knowledge of the site hydrogeology and their understanding of the models capabilities. An important aspect for the successful use of any process-based model in vulnerability assessment is the acquisition of relevant data

for the assessments. For this purpose, the structure and content of a generic database was outlined. The data in the database may primarily be used in preliminary simulations, and more detailed studies should be supple-mented with site-specific data. The use of hydrogeological environments is expected to provide a useful framework for the outlined database. Within planning and prevention scenario studies, hydrogeological environ-ments will contribute with hydrogeological data gathered from similar environments. These data may then be used for simulating potential risks and benefits, in the absence of measured soil hydraulic characteristics or parameter data for groundwater modelling. Since the parameters of the hydrogeological environments are used, e.g., to evaluate the turnover times in the unsaturated and groundwater zone, the results are associated with uncertainties. Sensitivity analyses should be carried out on parameters taken from the database to determine the significance of uncertainty on the model results. When probability distributions are used for the input parameters, the resulting distributions in model results may serve as a measure of the uncertainty. When hydrogeological envi-ronments are used in the context of vulner-ability assessments, it is important to identify and map all hydrogeological environments within an area because the contaminants in the groundwater zone may pass through several hydrogeological environments. Hence, the hydrogeological environments are not to be used as isolated units, but should be combined (Stejmar Eklund, 2002). To be able to use the concept of hydro-geological environments as a basis for the retrieval of generic data for process-based modelling, it is proposed that all of the hy-drogeological environments are identified and well documented. The structure of the HPAR database offers a sound basis for groundwater modelling data. However, in the HPAR databases of the municipalities studied actual data on some important parameters, e.g. hydraulic conductivity, are missing, al-though space has been allocated in the data-base to register such data. These data should be extracted from other sources at SGU and

25


elsewhere, and included it in the HPAR data-base. Data quality within the HPAR database is relatively good, and it contains information related to the sample ID, the sampling data, the sampling method, geographic position, type of represented entity, some technical characteristics connected to the represented entity, etc. However, some information rele-vant for vulnerability assessments, such as land use and information on potential con-taminations sources in the vicinity of the wells, should be incorporated. In addition, typical soil profiles defined for three com-mon Swedish hydrogeological environments, and statistically analysed and explored by the one-dimensional dynamic model, may be included in the outlined database. The dynamic modelling approach for the unsaturated zone was primarily used for comparative and illustrative purposes, explor-ing processes and parameters that are impor-tant for vulnerability assessments by process-based models. Furthermore, the modelling approach enabled a calculation of turnover times of conservative contaminants trans-ported by natural recharge in the saturated zone. In addition, the model simulations clearly demonstrated the impact of the soil properties, vegetation types and groundwater levels on the transport of a conservative tracer in the soil and the response to envi-ronmental changes. The simulations showed that the mean turn-over times were most dependent on the variation in the water retention properties (variation in hydraulic conductivity was not taken into account). Differences in the total water storage in the soil profile explains the variations in turnover times between soil types. In fine-grained soils this leads to longer turnover times, compared with coarse-grained soils. The glacial till soils become more compacted with increasing soil depth, resulting in a decrease in porosity and hydraulic conductivity, which efficiently prevents the water flowing to deeper layers. Thus, water discharge was more superficial for the glacial till than for the clay and sand soils. An increase in turnover time from arable land, to grass, to pine and finally to more

water-demanding spruce forest, was expected and also shown in the simulations. Forest vegetation generally has larger evapotranspi-ration than grass and arable vegetation, lead-ing to a smaller amount of water discharge from the soil profile, implying longer turn-over times. Groundwater levels generally had a major impact on turnover times. Notably, the re-sults from the simulations with a shallow groundwater level differed from the rest in that the turnover times were not sensitive to the variation of water retention properties. Furthermore, the sensitivity to vegetation types was low for the simulations with a shallow groundwater level. On the other hand, the influence on turnover times was most pronounced when the groundwater level was shallow. Glacial till and sand soils with shallow groundwater levels stored more water in the upper 50 cm of the soil profile compared to the deeper groundwater levels. However, this was not observed for the clay soils. Hence, the capillary transport of water from the groundwater was more pronounced in the coarser soils than in the clay soil. The study showed that it is important to account for the variation in soil properties and groundwater levels within an area if further detailed studies are to be made. Another key consideration in vulnerability assessments is the expected penetration depth of the contaminant in the subsurface in case of hydraulic surcharge. This is especially important when considering liquid spills. It is important to know what possibilities for rapid remediation measures exist after, for example, an accidental spill on a road. As expected, for liquid spills, the hydraulic con-ductivity was found to be of major impor-tance, while the water retention properties were of less importance. The initial conditions in the soil were demon-strated to be vital for the infiltrated volume of liquid spills for soils with low hydraulic conductivity. Except for the clay soil with low hydraulic conductivity, the initial water content did not have any pronounced effect on the penetration depth of the liquid spill. A one-domain approach ignoring preferential flow was used in the simulations. This may

26


question the validity of the present investiga-tion, if only the penetration depth is of inter-est. Nonetheless, for contaminants where the focus is on total amounts, the relative differ-ences obtained between the soils in the study is considered reliable. A one-dimensional model based on Richards’ equation may be unnecessarily complex for simple investiga-tions of penetration depths by infiltration, and often more simple models are used for the prediction of the advective transport in unsaturated soils. On the other hand, the dynamic approach was important to account for the vertical heterogeneity and the redis-tribution phenomena after infiltration. Penetration depths of accidental spills, simu-lated by a dynamic model in the unsaturated zone, may be used in conjunction with typical hydrogeological environments by examining characteristic groundwater levels for different environments. The groundwater level in sand deposits is normally between 2-10 m, and in some eskers more than 20 m. In glacial till, the groundwater level usually varies between 1-3 m depending on the position in the land-scape, whereas in clay the groundwater level is usually shallow (0.5-2 m). The penetration depth of a liquid, when related to the typical groundwater levels of the different soils, provides important information about the possibility for the contaminant to reach the groundwater zone rapidly, and thus give an indication of the vulnerability of the ground-water and of the possibility for successful remedial actions to be taken. The hydraulic properties of the soil profiles and the vegetation properties identified for typical hydrogeological environments in Sweden could prove to be a useful way of obtaining input data for model simulations when detailed measurements are not avail-able. Statistical information of water reten-tion properties and hydraulic conductivities provided a basis for estimating the turnover times and penetration depths, giving an indi-cation of the uncertainty in the predictions. In conclusion, the approach of using process-based modelling for vulnerability assessment is based on relatively loosely coupled models, generic databases and a GIS. It is believed that this will provide an appropriate, flexible

and time efficient means to express vulner-ability of groundwater in quantitative terms. However, more work needs to be carried out on the single components of the approach, as well as methods for integrating them. Fur-thermore, structured routines for uncertainty analysis are crucial for the credibility of the results.

RECOMMENDATIONS FOR

FUTURE RESEARCH

From the examples of the vulnerability as-sessments presented in this thesis, it is possi-ble to point out several specific areas in which future research is desirable. By ad-dressing these areas, better assessments of groundwater vulnerability using process-based models will be achieved. Further development of the Swedish soil database included in the CoupModel will need to be undertaken. In particular, soil profiles with information on several types of glaciofluvial deposits with coarse materials, and deep till profiles are needed. In addition, the user friendliness and accessibility of the database needs to be improved. The outlined hydrogeological database for groundwater modelling also needs to be further developed. Data must be gathered and included from different sources in this regard. Test areas should be chosen where the hydrogeological information may be transformed to hydrogeological environ-ments, with the inclusion of the most impor-tant parameters for groundwater modelling. In addition, the influence of using parameter data in hydrogeological environments on the vulnerability assessments should be studied further, e.g. by performing sensitivity analy-ses. Comparative studies of simulations with process-based models using parameter data from hydrogeological environments and simulations with site-specific data should be performed. Numerous aspects that affect the vulnerability assessments in conjunction with the use of the hydrogeological environments concept need to be studied further. These aspects include land use, vegetation, climate, integrated systems, data access, etc.

27


Methods that permit preferential flow in groundwater vulnerability assessments to be determined should be investigated and the explicit incorporation of effects in the groundwater vulnerability assessments should be considered. In general, one of the future challenges is the establishment of a conceptual and opera-tional basis for an integrated vulnerability assessment concept. This could take the form of, for example, a computerised decision support tool that would be capable of merg-ing the data from spatial and generic data-bases, and combining different vulnerability assessment methods. This might, in principal, be achieved by using the process-based mod-els, GIS packages and interfaces, but the parameter data needs to be collected and

included in databases. Furthermore, spatial variability will be a major problem when applying point models to large areas for vul-nerability assessment. Tools need to be cre-ated for merging data obtained at different spatial and temporal scales into a common scale for vulnerability assessment. Advances in new techniques might result in the user developing a strong belief in the ability of the assessments, using these tech-niques to describe the behaviour of the com-plex subsoil system. However, humans can never be substituted, even if we create the perfect tools. What is needed is the develop-ment of an understanding of the interactions between the different roles of planners, poli-ticians, experts, etc. in groundwater manage-ment.

28


REFERENCES

Abrahansen, P. and Hansen, S., 2000: Daisy: An open soil-crop-atmosphere model. Environment Modelling and Software, 15:313-330.

Aller, L., Bennet, T., Lehr, J.H. and Petty, R.J., 1987: DRASTIC: a standardised system for evaluating groundwater pollution potential using hydrogeological settings. US-Environmental Protection Agency.

Anderson, M.P. and Woesner W.W., 1992: Applied Groundwater Modeling - Simulation of Flow and Advective Transport. Academic Press, Inc. San Diego, California 92101, 381 p.

Babendreier, J.E. And Castleton, K.J., 2005: Investigating uncertainty and sensitivity in integrated, multi-media environmental models: tools for FRAMES-3MRA. Environmental Modelling & Softwa-re, 20(8):1043-1055.

Back, P.E. and Rosén, B., 2001: Riskanalys av områden där järnvägstrafik berör vattentäkter och andra vattenresurser – metodutveckling. Varia 513. Swedish Geotechnical Institute, Linköping. (in Swedish)

Beven, K., 2002: Towards a coherent philosophy for modelling the environment. Proc Roy Soc Lond A, 458(2026) 24:65-84.

Beverly, C.R., Nathan, R.J., Malafant, K.W.J. and Fordham, D.P., 1999: Development of a simplified unsaturated module for providing recharge estimates to saturated groundwater models. Hydrological Processes, 13:653-675.

Bodvarsson, G., Finsterle, S., Liu, H.H., Oldenburg, C.M., Pruess, K., Sonnenthal, E. and Wu, Y.S., 2000: Flow and transport modelling of the vadose zone. In: Looney, B.B., Falta, R.W. (eds), Vadose Zone Science and Technology Solutions. Vol. II. Battelle Press, Columbus, USA, 591-784.

Booty, W.G., Lam, D.C.L., Wong, I.W.S. and Siconolfi, P., 2001: Design and implementation of an environmental decision support system. Environmental Modelling & Software, 16:435-458.

Borman, H. and Diekkruger B., 2004: A conceptual, regional hydrological model for Benin (West Africa): validation, uncertainty assessment and assessment of applicability for environmental change analyses. Physics and Chemistry of the Earth, 29(11-12):759-768.

Burkart, M.R., Koplin D.W. and James, D.E., 1999: Assessing groundwater vulnerability to agrichemical contamination in the Midwest. Water Science and Technology, 39(3):103-112.

Burrough, P.A., 1986: Principles of Geographical Information Systems for Land Resources Assessment. Clar-enden Press, Oxford.

Bärlund, I. and Tattati, S., 2001: Ranking of parameters on the basis of their contribution to model uncer-tainty. Ecological Modelling, 142:11-23.

Caminiti, J.E., 2004: Cathment modelling – a resource manager’s perspective. Environmental Modelling & Software, 19:991-997.

Carsel, R.F., Imhoff, J.C., Hummel, P.R., Cheplick, J.M. and Donigian Jr., A.S., 1998: A Model for Predicting Pesticide and Nitrogen Fate in the Crop Root and Unsaturated Soil Zones: Users Manual for Release 3.0.

Christiansen, J.S., Thorsen, M., Clausen, T., Hansen, S. and Refsgaard, J.C., 2004: Modelling of macropore flow and transport processes at catchment scale. Journal of Hydrology, 299:136-158.

Civitia, M., 1994: La carte della vulnerbilità deli aquiferi all'inquinamento: teoria & pratica. Pitagora Editrice, Bologna, Italy, (in Italian), 325 p.

29


Costanza, R., Funtowicz, S.O. and Ravetz, J.R., 1992: Assessing and communicating data quality in policy relevant research. Environ Manage, 16:21–131.

Corwin, D. L., 1996: GIS applications of deterministic solute transport models for regional-scale assessment of non-point source pollutants in the vadose zone. In: Corwin, D.L. and Logue, K. (eds.), Applica-tions of GIS to the modelling of non-point source of pollutants in the vadose zone. Soil Sci Soc Am, 48:69-100.

Corwin, D.L. and Wagenet, R.J., 1996: Applications of GIS to the modeling of nonpoint source pollutants in the vadose zone: A conference overview. Journal of Environmental Quality, 25(3):403-411.

Corwin, D.L., Vaughan, P.J. and Loague, K., 1997: Monitoring nonpoint source pollutants in the vadose zone with GIS. Environmental Science & Technology, 31(8):2157-2175.

Corwin, D.L., Carrillo, M.L.K., Vaughan, P.J., Rhoades, J.D. and Cone D.G., 1999: Evaluation of a GIS-linked model of salt loading to groundwater. Journal of Environmental Quality, 28 (2): 471-480.

COST 65., 1995: Hydrogeological aspects of groundwater protection in karstic areas, Final report (COST action 65). European Comisson, Directorat-General XII Science, Research and Development, Report EUR 16547 EN, 446 p

Crosetto, M., Tarantola, S. and Saltelli, A., 2000: Sensitivity and uncertainty analysis in spatial modelling based on GIS. Agriculture, Ecosystems and Environment 81: 71-79.

Daly, D., Dassargues, A., Drew, D., Dunne, S., Goldscheider, N., Neales, S., Popescu, C.H. and Zwahlen, F., 2002: Main concepts of the “Eurpean Approach“ for (karst) groundwater vulnerability assessment and mapping. Hydrogeological Journal, 10, 2: 340–345.

Demetriou, C. and Punthakey, J.F., 1999: Evaluating sustainable groundwater management options using the MIKE SHE integrated hydrogeological modelling package. Environmental Model-ling & Software 14: 129-140.

Danish Hydrological Institute, DHI, 2000a: MIKE SHE Water Movement User Manual. DHI Water and Environment, Hörsholm, Denmark.

Danish Hydrological Institute,DHI, 2000b: MIKE SHE Water Quality User Manual. DHI Water and Environment, Hörsholm, Denmark.

Dierrsch, H.J.G., 1998: FEFLOW users manual. Institute for water resources planning and system research Ltd. Waltersdorften Str 105, D-12526 Berlin.

Doerfliger, N. and Zwahlen, F., 1997: EPIK: a new method for outlining of protection areas in karstic environment. In: Günay G., Johnson A.L. (eds), International symposium and field semi-nar on “karst waters and environmental impacts”. Antalya, Turkey. Balkema, Rotterdam. pp 117-123.

Dubus, I.G., Brown, C.D. and Beulke, S., 2003: Sources of uncertainty in pesticide fate modelling. The Science of the Total Environment 317: 53-72.

Eastman, R.J., 1992: IDRISI Version 4.0. User's Guide and Technical Reference. Clark University, Graduate School of Geography, Worcester, Massachusetts, 01610, USA.

Eastman, R. J., 2001: Idrisi 3.2, version 2. Guide to GIS and Image Processing. Clark Labs, Clark Uni-versity, USA, 171 p.

Eliasson, Å., 2001: Groundwater Impact Assessment and Protection - Predictive Simulations for Decision Aid. Licentiate thesis TRITA-AMI LIC 2066. Division of Land and Water Re-sources, De-partment of Civil and Environmental Engineering, Royal Institute of Tech-nology, Stockholm.

El-Kadi, A.I., Oloufa, A.A., Eltahan, A.A. and Malik, H.U., 1994: Use of Geographic Information System in Site-Specific Groundwater Modelling. In: Ground Water July- August, Vol 32, no 4: 617-625.

30


European Water Directive, 2000: 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.

Evans, T. A. and Maidment, D. R., 1995: A Spatial and Statistical Assessment of Vulnerability of Texas Groundwater to Nitrate Contamination. Technical report CRWR 260. Center of Research in Water Resources. The University of Texas at Austin, Austin TX, 257 p.

Fedra, K., Weigkricht, E. and Winkelbauer, L., 1993: Decision Support and Information Systems for Regional Development Planning. IIASA RR-93-13, Laxenburg, Austria.

Fedra, K., 1998: Integrated risk assessment and management: overview and state of the art. Journal of Haz-ardouz Materials 61: 5-22.

Fedra, K. and Winkelbauer, L., 2002: A hybrid expert system, GIS, and stimulation modelling for environ-mental and technological risk management. Computer-aided Civil and Infrastructure Engineer-ing 17 (2):131-146.

Foster, S.S.D., 1987: Fundamental concepts in aquifer vulnerability, pollution risk and protection strategy. In: W. van Duijvenbooden, H.G. van Waegeningh(eds.),Vulnerability of soils and ground-water to pollution, , Proceedings and information, TNO Committee on Hydrological Research, The Hague, No. 38.

Foster, S.S.D. and Hirata, R., 1988: Groundwater pollution risk assessment - a methodology using available data. Pan American center for sanitary engineering and environmental sciences (CEPIS), Lima, Peru.

Foster S., Hirata R., Gomes D., D’Elia M. and Paris M., 2002: Groundwater Quality Protection a guide for water utilities, municipal authorities, and environment agencies. The World Bank Washington, D.C. 116 p.

Freeze, A., Mason, J., Smith, L., Sperling, T. and James, B., 1990: Hydrogeological Decision Analysis: 1. A Framework. In: Ground Water, September-October 1990, Vol. 28, No. 5: 738-766.

Freeze, R.A., James, B., Massmann, J., Sperling, T. and Smith, L., 1992: Hydrogeological Decision Analysis: 4 The Concept of Data Worth and Its Use in the Development of Site Investigation Strate-gies, 30 (4): 574-588.

Gogu, R. and Dassargues, A., 2000: Current trends and future challenges in groundwater vulnerability as-sessment using overlay and index methods. Environmental Geology 39: 549-559.

Gogu, R.C., Carabin, G., Hallet, V., Peters, V. and Dassargues, A., 2001: GIS-based hydrogeological databases and groundwater modelling. Hydrogeology Journal, 9: 555-569.

Gogu, R.C., Hallet, V. and Dassargues, A., 2003: Comparison of aquifer vulnerability assessment tech-niques. Application to the Neblon river basin (Belgium). Environmental Geology 44 (8): 881-892.

Gontier, M. & Olofsson, B., 2005: Areell sårbarhetsbedömning för grundvattenpåverkan av vägförorening. TRITA-LWR Report 3011. Department of Land and Water Resources Engineering, Royal Institute of Technology, Stockholm. (in Swedish)

Granlund, K. and Nystén, T., 1998: The effects of road salt on the Miekkamäki aquifer in central Finland – simulation by a two-dimensional groundwater model. In: Nystén, T. and Suokko, T. (eds), De-icing and dustbinding – risk to aquifers. Proceedings of an International sym-posium, Helsinki, Finland. Nordic Hydrological Programme. NHP report No. 43: 193-200.

Gumbricht, T. and Thunvik, R., 1996: 3D Hydrological modelling with an expert GIS interface. NHP-report No. 40: 176-185. Paper presented at Nordic Hydrological Conference, Akureyri, Iceland

Gustafsson, A.M., Gustafsson, L.G., Winberg, S. and Refsgaard, A., 1997: Safeguarding the Kristian Plain groundwater resource by using MIKE SHE model. In: Refsgaard and Karalis (eds) Opera-tional Water Management. Balkema, Rotterdam.

31


Gömann, H., Kreins, P., Kunkel, R. and Wendland, F., 2005: Model based impact analysis of policy options aiming at reducing diffuse pollution by agriculture – a case study for the river Ems and sub-catchment of the Rhine. Environmental Modelling & Software 20: 261-271.

Haertle, A., 1983: Method of working and employment of EDP during the preparation of groundwater vulner-ability maps. International Association of Hydrological Sciences 142: 1073-1085.

He, C., 2003: Integration of geographic information systems and simulationmodel for watershed management. Environmental Modelling & Software 18: 809-813.

Heuvelink, G.B.M., Burrough, P.A. and Stein, A .,1989: Propagation of errors in spatial modelling with GIS. International Journal of GIS 3: 303–322.

Helton, J.C., Johnson, J.D. and Oberkampf, W.L., 2004: A exploration of alternative approaches to the representation of uncertainty in model predictions. Reliability Engineering & System Safety 85 (1-3): 39-71.

Heuvelink, G.B.M., 1996: Identification of field attribute error under different models of spatial variation. International Journal of GIS 10: 921–935.

Heuvelink, G.B.M., 1998: Uncertainty analysis in environmental modeling under a change of spatial scale. Nutrient Cycling in Agroecosystems 50: 255-264.

Heuvelink, G.B.M., 2003: Spatial data quality. International Journal of Geographical Information Science 17 (8), 816-U2.

Hoosbeek, M.R. and Bryant, R.B., 1992: Towards the quantitative modeling of pedogenesis - a review. Geoderma 55: 183–210.

Jansen, M.J.W., 1998: Prediction error through modelling concepts and uncertainty from basic data. Nutrient Cycling in Agroecosystems 50 (1-3): 247-253.

Janssen, P.H.M., Heuberger, P.S.C. and Klepper, O., 1994: UNCSAM: a tool for automating sensitiv-ity and uncertainty analysis. Environmental Software 9: 1–11.

Jansson, P.E. and Karlberg, L. 2004: CoupModel-Coupled Heat and Mass Transfer Model for Soil-Plant-Atmosphere System. Div. of Land and Water Resources Engineering, Royal Institute of Technology, Stockholm, TRITA-LWR report 3087, 427 p.

Jansson, P.E. and Moon, D.S., 2001: A coupled model of water, heat and mass transfer using object orienta-tion to improve flexibility and functionality. Environmental Modelling & Software 16: 37-46.

Jarvis, N., 1994: The MACRO Model (Version 3.1). Technical Description and Sample Simulations. Swedish University of Agricultural Science. Department of Soil Science. Reports and Dissertations 19, Uppsala.

Jarvis, N. J., Hollis, J. M., Nicholls, P. H., Mayr, T. and Evans, S. P., 1997: MACRO-DB: a decision-support toll for assessing pesticide fate and mobility in soils. Environmental Modelling & Soft-ware, Vol. 12, Nos 2-3: 251-265.

Jarvis, N.J. and Larsson, M.H., 1998: The MARCO model (version 4.1), Technical description, http://130.238.110.143/bgf/Marcohtm/marco.htm

Jeannin, P.Y., Cornaton, F., Zwahlen, F. and Perrochet, P., 2001: VULK: a tool for intrinsic vulner-ability assessment and validation. 7th Conference on Limestone Hydrology and Fissured Media, Besançon 20–22 Sep. 2001, 13: 185-190.

Johansson, P. O., Knutsson, G., Lindström, R., Maxe, L. and Scharp, C., 1993: Groundwater vulner-ability assessment - a short state of the art and an outline for research activities and development of methodologies. Dept. of Land and Water Resources, Royal Institute of Technology, Stock-holm.

Johanssson, P-O., Djurberg, H., Gunnemyr, L., Söderqvist, T and Collentine,D., 2002: Värdering av grundvattenresurser. Metoder ochtillvägagångssätt. Naturvårdsverket, Rapport 5142, 50 p. (in Swedish)

32


Jury, W.A. and Gruber, J.A., 1989: Stochastic analysis of the influence of soil and climatic vari-ability on the estimate of pesticide groundwater pollution potential. Water Resources Re-search 25: 2465 –2474.

Klein, M., 1995: PELMO: Pesticide Leaching Model, version 2.01. Fraunhofer Institut fur Umweltchemie und Onkotoxikologie, 103 p.

Knisel, W.G. (Ed.), 1980: CREAMS: a fiel-scale model for chemicals, runoff, and erosion from agricultural managements systems. US Department of Agriculture, Science, and Education Administra-tion. Conservation Research Report no. 26, 643 p.

Knisel, W. G., Leonard, R.A. and David, F.M., 1989: GLEAMS User manual. Tifton, Georgia. Southeast Watershed Research Laboratory.

Knisel, W.G. and Williams, J.R., 1995: Hydrology component of CREAMS and GLEAMS models. In: Singh, V.P. (Ed.). Computer Models of Watershed Hydrology, Water Resources Publi-cation, pp 1069-1114.

Konikow, L.F. and Bredehoef, J.D., 1978: Computer model of two-dimensional solute transport and disper-sion in groundwater. Techniques of Water Resources Investigation Book 7, Chap. C2, USGS.

Konikow, L. F., Granato, G.E. and Hornberger, G. Z., 1994: User’s guide to revised method of charac-teristics solute transport model (MOC-version 3.1). U.S. Geological Survey. Water Resources Investigations Report 94-4155, 63 p.

Kragt, J.F. and Hack-ten Broeke, M.J.D., 1991: Brief description of RENLEM. In: CEC, 1991. Ni-trate in Soils, Soil and Groundwater Research Report II, EUR 13501 EN, pp 253-255.

Kylefors, L., Gumbricht, T. and Gustafsson, L.G., 1999: Application of hydrological models in water management. Va-Forsk rapport 199-13. Svensk Byggtjänst, Stockholm, Sweden (in Swe-dish).

Lam, D.C.L. and Swayne, D.A., 1993: An expert system approach of integrating hydrological database, models and GIS: application of the RAISON System. In: Application of Geographic Informa-tion Systems in Hydrology and Water Resources Management (eds. Kovar and Nachtnebel). IAHS Publication No. 211: 23-33.

Larsson, M.H. and Jarvis N.J., 1999: A dual-porosity model to quantify macropore flow effects on nitrate leaching. Journal of Environmental Quality 28 (4): 1298-1307.

Lasserre, F., Razack, M. and Banton, O., 1999: A GIS-linked model for the assessment of nitrate contami-nation in groundwater. Journal of Hydrology 224: 81-90.

Leonard, R.A., Knisel, W.G. and Still, D.A., 1986: GLEAMS: groundwater loading effects of agricultural management system. In: Trans. ASAE. Vol 30, No. 5.

Lindberg, J., Olofsson, B. and Gumbricht, T., 1996: Risk mapping of groundwater salinization using Geographical Information Systems. 14th Salt Water Intrusion Meeting (SWIM-96). Swedish Geological Survey (SGU) Reports and notes 87: 188-197.

Lindström, R., and Scharp C., 1995: Approaches to groundwater vulnerability assessments: A state of the art report. Div. of Land and Water Resources, Royal Institute of Technology, Stockholm, 77 p.

Lindström, R. 2005: A system for modelling groundwater contamination in water supply areas - chloride con-tamination from road de-icing as an example. (accepted for publication in Nordic Hydrology).

Loague, K., Bernknopf, R.L., Green, R.E. and Giambelluca, T.W., 1996: Uncertainty of groundwater vulnerability assessments for agricultural regions in Hawaii: Review. Journal of Environmental Quality 25 (3): 475-490.

Loague, K. and Corwin, D.L., 1996: Uncertainty in regional-scale assessments of non-point source pollutants. In: Corwin, D. L. and Loague, K. (eds), Applications of GIS to the Modeling of Non-

33


Point Source Pollutants in the Vadose Zone, Special Publication 48. Soil Science Society of America, Madison: 131-152.

Loague, K. and Corwin, D.L., 1998: Regional-scale assessment of non-point source groundwater contamina-tion. Hydrological Processes 12: 957-965.

Lobo Ferreira, J.P. and Oliveria M.M., 2003: On the Experience of Groundwater Vulnerability Assess-ment in Portugal. Aquifer Vulnerability and Risk, International Workshop.

Lord, E.I. and Anthony, S.G., 2000: MAGPIE: A modeling framework for evaluating nitrate losses at national and catchment scales. Soil Use and Management 16:. 167-174 Suppl.1

Mailhot, A., Villeneuve, J-P., 2003: Mean-value second-order uncertainty analysis method: application to water quality modelling. Advances in Water Resources 26: 491-499.

Massmann, J., Freeze, R.A., Smith, L., Sperling, T. and James, B., 1991: Hydrogeological Decision Analysis: 2. Applications to Ground-Water Contamination. Ground Water 29 (4): 536-548.

Maxe, L. and Johansson, P-O., 1998: Assessing groundwater vulnerability using travel time and specific surface area as indicators. Hydrogeology Journal 6: 441-449.

McDonald, M.G. and Harbaugh, A.W., 1988: A Modular three-dimensional finite differnce ground-water flow model. Techniques of Water-Resources Investigations of the U.S. Geological Survey. Book 6. Chapter A1. Open-File Report 83-875.

Morse, G., Eatherall, A. and Jenkins, A., 1994: Managing agricultural pollution using linked geographical information system and non-point pollution model. In: J. IWEM, Vol 8 June: 277-286.

National Research Council (NRC), 1993: Groundwater Vulnerability Assessment. Predicting Relative Contamination Potential Under Conditions of Uncertainty. National Academy Press.

National Research Council (NRC), 1997: Valuing Ground Water, Economic concepts and approaches. National Academy Press, Washington.

Newell, C. and Haasbeek, J.F., 1990; OASIS: A Graphical Decision Support System for Ground Water Contaminant Modeling. In: Ground Water, March-April 1990. Vol 28, No 2: 224-234.

Refsgaard, J.C., Thorsen, M., Jensen, J.B., Kleeschulte, S. and Hansen, S., 1999: Large scale model-ling of groundwater contamination from nitrate leaching. Journal of Hydrology 221: 117-140.

Refsgaard, J.C. and Henriksen H.J., 2004: Modelling guidelines – terminology and guiding principles. Ad-vances in Water Resources 27: 71-82.

Rosen, L. and Gustafson G., 1996: A Bayesian Markov geostatistical model for estimation of hydrogeological properties. Ground Water 34 (5): 865-875.

Sandström M., 1998: Vad är vattnet värt – värderingsmodell för grundvattentillgångar. Naturvårdverket Rapport 4876. 47 p. (in Swedish)

SEPA, 2000: Environmental quality criteria – Groundwater. Swedish Environmental Protection Agency Report 5051, 140 p.

Scharp, C., Alveteg, T., Johansson, P-O. and Caldera, M.I., 1997: Assigning a groundwater protection value: methodology development. Proceedings of International Association of Hydrogeologists Congress: Groundwater in the urban environment, Problems, Processes and Manage-ment. Nottingham: 659-664.

Schoumans, O.F., Breeuwsma, A. and Thunnissen, H.A.M.., 1989: Use of remote sensing and geo-graphic information systems in soil and groundwater vulnerability assessment. In: Proceedings of the 9th Earsel Symposium, 27 June- 1 July 1989, Espoo, Finland, pp 54-60

Simunek, J., Huang K., and van Genuchten, M. Th., 1998: The HYDRUS code for simulating the one-dimensional movement of water, heat, and multiple solutes in variably-saturated media. Version 6.0, Research Report No. 144, U.S. Salinity Laboratory, USDA, ARS, Riverside, California, 164 p.

34


Sivertun, Å. and Prange, L., 2003: Non-point source critical area analysisin the Gisselö watershed using GIS. Environmental Modelling & Software 18: 887-898.

Sokol, G., Leibungut, Ch., Schulz, K.P., and Weinzierl, W., 1993: Mapping procedures for assessing groundwater vulnerability to nitrate and pesticides. In: HydroGIS 93:Application of geographic information systems in hydrology and water resources, IAHS Publ. no 211: 631-639.

Sperling, T., Freeze, R.A., Massmann, J., Smith, L. and James, B., 1992: Hydrogeological Decision Analysis: 3. Application to Design of a Ground-Water Control System at an Open Pit Mine, 30 (3): 376-388.

SRA, Swedish Road Administration, Vägverket, 1998: Förorening av vattentäkt vid vägtrafikolycka - Hantering av risker med petroleumutsläpp. Vägverket Publ 1998, 64 p. (in Swedish).

Stewart, I.T. and Loague, K., 2004: Assessing ground water vulnerability with the type transfer function model in the San Joaquin Valley, California. Journal of Environmental Quality 33 (4): 1487-1498.

Sun, M. and Zheng, C.M., 1999: Long-term groundwater management by MODFLOW based dynamic optimisation tool. Journal of the American Water Resources 35 (1): 99-111.

Teso, R.R., Poe, M.P., Younglove, T. and McCool, P.M., 1996: Use of logistic regression and GIS modeling to predict groundwater vulnerability to pesticides. Journal of Environmental Quality 25 (3): 425-432.

Thunqvist, E.-L. J., 2003: Estimating chloride concentration in surface water and groundwater due to de-icing salt application. PhD Thesis TRITA-LWR-PHD 1006, Department of Land and Water Resources Engineering, Royal Institute of Technology, Stockholm.

Tiktak, A., Leijnse, A. and Vissenberg, H., 1999: Uncertainty in a regional-scale assessment of cadmium accumulation in the Netherlands. Journal of Environmental Quality 28 (2): 461-470.

Tiktak, A., van den Berg, F., Boesten, J.J.T.I., Leistra, M., van der Linden, A.M.A., van Kraal-ingen, D., 2000: Pesticide Emission Assessment at Regional and Local Scales: User Manual of FOCUS Pearl version 1.1.1. RIVM Report 711401008, Alterra Report 28, RIVM, Biltho-ven, 142 p.

Tiktak, A., de Nie, D., van der Linden, T. and Kruijne, R., 2002: Modelling the leaching and drainage of pesticides in the Netherlands: the GeoPEARL model. Agronomie 22 (4): 373-387

Tiktak, A., de Nie, DS., Piñeros Garcet, JD., Jones, A. and Vanclooster, M., 2004: Assesment of the pesticide leaching risk at the Pan-European level. The EuroPEARL approach. Journal of Hydrol-ogy 289: 222-238.

Troiano, J., Spurlock, F. and Marade, J. 1999: Update of the Californian Vulnerability Soil Analysis for Movement of Pesticides to Ground Water, 1999. Department of Pesticide Regulation, Sacra-mento. CA 95814-3510, Document EH 00-05.

U.S.EPA Environmental Protection Agency, 1993: A review of methods for assessing aquifer sensitivity and groundwater vulnerability to pesticide contamination. Washington, 147 p.

Vanclooster, M., Viaene, P. and Christians, K., 1994: WAVE-a mathematical model for stimulating agrochemicals in the soil and vadose environment. Reference and user’s manual (release 2.0). Institute for Land and Water Management, Katholieke Universiteit Leuven, Belgium.

Voss, C.I., 1984: SUTRA - Saturated Unsaturated Transport - A Finite- Element Simulation Model for Saturated - Unsaturated, Fluid-Density - Dependent Groundwater Flow with Energy Transport or Chemical Reactive Single - Species Solute Transport. WRI 84-4369. U.S: Geological Survey, Reston, Virginia. 409 p.

Vrba, J. and Zaporozec, A. (eds.), 1994: Guidebook on Mapping GroundwaterVulnerability. IAH Inter-national Contributions to Hydrogeology, Vol. 16.Heise Verlag, Hannover. 131 p.

35


Wagenet, R.J. and Hutson, J.L., 1987: LEACHM: A Finite-Difference Model for Simulating Water, Salt and Pesticide Movement in the Plant Root Zone, Continuum 2 Ithaca. New York State Resources Institute, Cornell University.

Wladis, D., Rosen, L. and Kros, H., 1999: Risk-based decision analysis of atmospheric emission alternatives to reduce ground water degradation on the European scale. Ground Water 37 (6): 818-826.

Wendland, F., Bach, M. and Kunkel, R., 1998: The influence of nitrate reduction strategies on the temporal development of the nitrate pollution of soil and groundwater throughout Germany – a regionally differen-tiated case study. Nutrient Cycling in Agroecosystems 50 (1-3): 167-179.

Wendland, F., Kunkel, R., Grimwall, A.., Kronvang, B. and Muller-Wohlfeil, DI., 2002: The SOIL-N/WEKU model system – a GIS-supported tool fort he assessment and managment of diffuse nitrogen leaching at the scale of river basins. Water Science and Technology 45 (9): 285-292.

Witkowski (ed.), 2004: Groundwater vulnerability assessment and mapping, Abstracts. International con-ference, University of Silesia, Poland. 158 p.

Zwahlen, F. (ed.), 2003: COST Action 620. Vulnerability and risk mapping for the protection of carbonate (Karst) Aquifers. Final Report. 297 p.

OTHER REFERENCES

European Environmental Agency, 1999: Groundwater quality and quantity in Europe. Copenhagen, Denmark. www.eea.eu.int, 123 p.

Swedish Environmental Objectives Council (SEOC), 2004: Swedish’s environmental objectives – are we getting there. www.naturvardsverket.se/bokhandeln/pdf/620-1238-X.pdf. 101 p.

U.S. EPA, 1994: PATRIOT: Pesticide Assessment Tool for Rating Investigations of Transport. http://www.epa.gov/ceampubl/gwater/patriot/index.htm.

Wingle W.L., Poeter E.P. and McKenna S.A., 1996: UNCERT: a geostatistical uncertainty analysis package applied to groundwater and contaminant transport modeling. http://www.mines.edu/fs home/wwingle/uncert/.

36

http://www.eea.eu.int/

http://www.naturvardsverket.se/bokhandeln/pdf/620-1238-X.pdf

http://www.epa.gov/ceampubl/gwater/patriot/index.htm

http://www.mines.edu/fs home/wwingle/uncert/

http://www.mines.edu/fs home/wwingle/uncert/

groundwater vulnerability assessment using process based...

Documents