Development of a Decision Support System for Groundwater Pollution Assessment

N. Kukuric

• Publisher: CRC Press
• Year: 2020
• Print ISBN: 9789054104186
• eBook ISBN: 9781000674668
• Topic: Business
• Subtopic: Human Resource Management
• Index: Business

This book discusses the development of the decision support system for groundwater pollution assessment, one of the first integrated information systems in the field of hydrogeology, reflecting the purpose of knowledge encapsulation in the field of groundwater quality management.

1.  INTRODUCTION

One of the first thoughts related to the possible (research on) development of a Decision Support System (DSS) was: ‘it will not be about interfacing’. In other words, the research should not concentrate on connecting software into a DSS, because that is, in principle, a pure engineering task.¹ The research should be more about electronic Decision Support, and less about forming the operational System.

Decision Support Systems are the systems that support decision-making. In practice, the support just to be provided mostly through the electronic storage, processing (modelling) and visualisation of data (numerical information). Although a knowledge base was often mentioned (next to a model and a database) as a basic DSS component, very few DSSs were found that contained any kind of knowledge component. This is a notable result, because both data (quantitative information) and knowledge (qualitative information) are needed to support decision-making. Without a knowledge component DSSs remain integrated (interfaced) software systems, whereas they should be integrated information systems. This conclusion led to clarification of research objectives; the research should be (and has been) dedicated to a DSS knowledge component, or, in somewhat broader terms, to the ‘electronic encapsulation of knowledge’.

Knowledge about what? In a time of ‘interdisciplinary problems’ none of the possible terms seemed to be exact enough. At the beginning of the research, knowledge about groundwater problems in general was targeted, with the intention of narrowing the scope during the research. Knowledge coming from various disciplines (hydrogeology, geochemistry, geophysics, ecology, economy, legislation, etc.) is involved in the management of groundwater problems. Therefore the first phase of the research was conducted under the umbrella ‘electronic encapsulation of knowledge for Groundwater Quality Management’ (GWQM).² It included:

–  a review of current achievements in software integration and knowledge encapsulation;

–  a review of postulates and techniques of Artificial Intelligence (AI) related to knowledge encapsulation and DSS development, and

–  an analysis of Groundwater Quality Management in the light of knowledge encapsulation.

The purpose was to define a possible realm and the best way(s) of applying AI in GWQM. The research phase was completed at the end of 1996 and results were published as an extended article in Water Resources Management (Kukurić and Hall, 1998). The article has been included in the thesis in its original form (Chapter 2), showing that no major conclusions needed to be reconsidered, this in spite of breath-taking developments in the field of AI and Information Technology (IT) in general.

The results obtained in the first research phase allowed the concept ‘DSS - an integrated information system’ to be worked out (Chapter 3). The concept and defined application realm were subsequently used to establish the framework of a DSS for Groundwater Pollution Assessment at the local scale (Chapter 3). The DSS consists of the DSS kernel (that is a fully integrated powerful rational data base and a GIS), an application environment and a knowledge component.³ Since the knowledge component was the prime target of the research, Chapter 3 also contains general considerations on knowledge encapsulation, knowledge processing and the DSS knowledge component.

The second research phase was dedicated to the design and development of the DSS knowledge component. The component consists, at the time of writing, of three task-oriented Knowledge-Based Modules (KBMs), namely:

–  a Site Characterisation Module (SCM);

–  a Vulnerability Assessment Module (VAM), and

–  a Groundwater Modelling Module (GMM).

Characterisation of an investigated site is the first task to be carried out when dealing with pointsource groundwater pollution problems. The main objectives of the characterisation are conceptualisation of the groundwater system and diagnosis of groundwater pollution. General knowledge on site characterisation has been acquired, systematised and encapsulated in the Site Characterisation Module (Chapter 4). The SCM has been encoded in HyperText Markup Language (HTML) that allows encapsulation of large amounts of semi-structured knowledge in an easily adaptable form. Besides knowledge, the SCM contains the links with procedures for data storage, processing and presentation located elsewhere in the DSS. The module was presented at the international conference ‘Hydroinformatics 98’ (Copenhagen, Denmark) and the paper was published in the conference proceedings (Kukurić et al, 1998a).

In general terms, groundwater Vulnerability Assessment (VA) can be described as a procedure for the quick assessment of groundwater pollution potential. It is based on intrinsic aquifer characteristics, though contaminant characteristics and management practice can also be taken into account. In the context of investigations of groundwater pollution potential (at a local scale), VA is seen as a step that follows characterisation of the investigated site. Accordingly, VA uses the results of the site characterisation, yielding a first assessment of the pollution potential. The results of VA provide a basis for further investigations and/or assessment, and a means for comparison of pollution potentials. A new ranking-based VA methodology has been developed and encapsulated in the VAM-Vulnerability Assessment Module (Chapter 5). The VAM has been developed in object-orientated Borland Delphi Developer, a tool based on Object Pascal as a programming language. Integration of the VAM into the DSS involved (inter alia) DSS kernel, the SCM and a Chemical Database (CDB). The module was presented at the international conference ‘Hydrology in a Changing Environment’ (Exeter, United Kingdom) and the paper was published in the conference proceedings (Kukurić et al, 1998b).

The GMM has been developed to support modelling of point-source pollution problems (Chapter 6). It consists of several software applications developed in various programming environments. The core of the GMM is an electronic Modelling Protocol guidance composed of sets of hypertext-based topics. The knowledge on general model complexity is encapsulated in a rulebased Model Complexity Module by using a knowledge base editor. The GMM also contains two additional modules (developed in Delphi) that assist in estimation of the retardation factor and dispersivity. The Modelling Protocol acts as a platform that integrates the software applications in a unique DSS module. In the framework of GMM development, MODFLOW, a modular 3D groundwater flow model, has been fully integrated with the DSS kernel. The GMM was presented at the international conference ‘MODFLOW 98’ (Golden, Colorado) and the paper was published in the conference proceedings (Kukurić et al, 1998c).

Development of above-mentioned KBMs (described in chapters 4, 5 and 6) completed the second phase of the research. Concluding remarks about the conducted research are given in Chapter 7.

Although not being explicitly stated in the title of the thesis, the research described in the following chapters was primarily about knowledge encapsulation. The title reflects the purpose of knowledge encapsulation: to provide ‘decision support’ for ‘groundwater pollution assessment’. Still, the emphasis was on conceptualisation and prototyping of a knowledgecontaining DSS. Decisions are, in principle, based on available information, meaning that an electronic DSS should integrate both quantitative and qualitative information (data and knowledge), in order to provide adequate support to the decision-making process. Therefrom arises the concept of a DSS as an ‘integrated information system’. Knowledge in the DSS is partly about numerical information, for example: which method should be used to define a parameter ‘A’ in the present situation, or which value of the parameter ‘A’ is appropriate for present situation? Additionally, knowledge allows the DSS to act as a process-oriented or a task-oriented system; in other words, the DSS can support decisions on steps that should be taken (e. g.with respect to data acquisition, processing, presentation, etc.) in order to carry out the task (to solve a posed problem).

Knowledge encapsulation in the field of groundwater quality management appears to be a real challenge, as thoroughly discussed in Sections 2.4 and 3.5. Until recently, AI has dealt exclusively with fairly structured knowledge coming from narrow, well-defined domains. Accordingly, AI techniques and procedures for knowledge acquisition and representation are developed to ‘capture’ this kind of knowledge. Numerous interviews have been conducted with the groundwater experts (as a part of this research), in an attempt to locate knowledge domains suitable for electronic encapsulation. Apparently, those domain exist and could be defined (e.g. expertise on validity of equilibrium sorption isotherm or on trade-off between network density and number of particles - see Section 2.4.3). That was, however, not the only conclusion that emerged from the conducted interviews; much more striking was the revelation of ambiguousness and rigidness of a ‘global picture’ in which narrow local domains play a relatively minor role. In other words, global approaches to groundwater pollution problems have not been worked out sufficiently (e.g. for Site Characterisation), or various approaches have been introduced and used (e.g. Vulnerability Assessment) without thorough comparison of their applicability in various situations. The same holds for various methods and tools that, once made available and implemented, are continuously used without reconsideration. Large portions of what is considered as common, general knowledge is not sufficiently structured, widely scattered among various sources of information (books, reports, guidelines, etc), often incomplete and sometimes even inconsistent. Processing knowledge within narrow domains can be successful only if input information coming from the (more) global domain is correct; e.g. what is the sense of adjusting network density and number of particles (in a groundwater model) if the modelling objective or the conceptual model are not set up properly? Moreover, the most important decisions (and the most serious errors) are made at a global level.

Gathering and ordering of general knowledge on groundwater pollution problems (as done in this research) is very labourious task, but is certainly worth doing. Some of the encapsulated knowledge might, in eyes of a field expert, look like no more than a set of ‘standard receipts’. Not everyone is, however, expert in a particular field. Besides, encapsulated knowledge can always serve as a useful reminder. Currently, much knowledge is not used in daily practice simply because it is not available (in consistent, systematic form) at the moment when it is needed. Contemporary IT offers possibilities for systematic storage and transfer of large portions of electronically encapsulated information; once available in a computer (in an appropriate form), it will most certainly be used.

General knowledge is by definition explicit and its acquisition does not ask for either ‘classical’ interviews with experts or other common AI acquisition procedures; the need for an AI acquisition procedure becomes evident when entering deeper domains that contain much more heuristic, tacit knowledge. Only one deeper knowledge domain has been entered during the research: (groundwater) model complexity (Chapter 6). An extensive (electronic) questionnaire was prepared and sent to experts. Response to the questionnaire was very limited, reflecting a lack of motivation for knowledge sharing. Cooperation has also proved to be crucial for systematisation of general knowledge, this time having ‘consensus’ as its most important postulate (see Section 3.5). The general lack of unwillingness of experts to cooperate has triggered thoughts on ethical and philosophical aspects of knowledge sharing. It was very tempting to continue research purely in this direction, but eventually a sense of practicality prevailed.

It was no less tempting to enter deeper into the innovative and sometimes fascinating technological world of Artificial Intelligence, where knowledge-based agents cooperate without hesitation and where DSSs are, at least conceptually, really ‘smart’ systems. Again, this was not such a good idea, knowing how conservative the hydrogeological community is (generally speaking). Eventually, this research was about implementation of AI and IT in hydrogeology into an extent (‘just one step from reality’) that might be accepted (appreciated?) by hydrogeologists. This research was not about generating new hydrogeological knowledge and also not about developing new IT technologies. It was about bringing new technologies (and maybe a new way of thinking) to hydrogeology.

All the steps of the knowledge encapsulation process were carried out in order to create a hydrogeological integrated information system:

–  major tasks related to groundwater pollution assessment were worked out in detail;

–  general knowledge required for accomplishment of the tasks was acquired and systematised;

–  AI knowledge representation forms were applied in formalisation of knowledge and design of knowledge-based modules,

–  the modules were developed and integrated into the DSS using state-of-the-art IT techniques.

The developed (prototype) DSS for groundwater pollution assessment appears to be one of the first attempts in the field of hydrogeology to encapsulate and actively to integrate knowledge in a comprehensive task-oriented DSS. Its development will be justified if it inspires further electronic encapsulation and integration of information on groundwater problems.

At this point, the introduction can be considered completed because it has outlined (mainly by introducing the following chapters and some follow-ups) what this thesis is about (and not about). There are, however, still those ‘tempting’ aspects of knowledge that (with a few exceptions) have not been mentioned explicitly in the thesis, although they were continuously dealt with during the research. Therefore an annex to the thesis is added, containing some thoughts about technological, philosophical and ethical aspects of knowledge.

¹  Already at that time, interfacing seemed to became a ‘common practice’. Four years later, it appears that the epithet ‘fully integrated’ can be assigned to a very few DSSs. This is despite all advances in software and hardware, and an immeasurable amount of work on interfacing done in the meantime. Interfacing is the engineering task, but - apparently - a tough one.

²  The term ‘Quality’ was added to emphasise consideration of both quantitative and qualitative aspect of groundwater problems.

³  The DSS kernel was provided by the REGIS package, a regional geohydrological information system developed by Netherlands Institute of Applied Geoscience TNO, Although some assistance was received from TNO (gratefully acknowledged) with interfacing REGIS and the knowledge components, the programming work associated with the latter was carried out entirely by the author.

2. THE ELECTRONIC ENCAPSULATION OF KNOWLEDGE ...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• Abstract
• Acknowledgements
• Frequently used acronyms
• 1. Introduction
• 2. Encapsulation of knowledge for groundwater quality management
• 3. Development of a Decision Support System (DSS): main considerations and framework
• 4. Knowledge-Based Module for Site Characterisation (SCM)
• 5. Knowledge-Based Module for Vulnerability Assessment (VAM)
• 6. Knowledge-Based Module for Groundwater Pollution Modelling (GMM)
• 7. Concluding remarks
• References
• Annex: Beyond engineering
• Samenvatting
• Curriculum Vitae