Environmental Decision Support Systems

Fedra, K. (2000)
Environmental Decision Support Systems: A conceptual framework and application examples. Thése prèsentèe á la Facultè des sciences, de l'Universitè de Genéve pour obtenir le grade de Docteur és sciences, mention interdisciplinaire. 368 pp., Imprimerie de l'Universitè de Genéve, 2000.				On-line reports, papers On-line interactive DEMOS ASP/SaaS solution examples

Summary

Environmental information and decision support systems (DSS) have emerged over the last decades as important tools for environmental planning and management.

Environmental problems, from urban and industrial pollution to natural and technological hazards keep growing, driven by local and global population growth and ever growing consumption of energy and materials. However, especially in the industrialized countries, most simple decisions with large pay-offs have already been taken: what remains in most cases is the fine tuning of the relationship between technology, economy, and the environment.

These problems are complex in the physical domain, and usually controversial in the socio-economic domain. Environmental systems are complex, dynamic, spatially distributed, and highly non-linear. Their coupled processes operate on a multitude of interdependent scales in time and space. In addition, many of the governing processes are not directly observable and therefor not easily understood.

On the socio-economic side, all decisions related to environmental planning and management are characterised by multiple and usually conflicting objectives, and multiple criteria; one also faces the problem of uncertainty, data versus perceptions, believes, and fears, hidden agenda and plural rationalities, which are a necessary consequence of the increasingly wide public participation in the decision making processes.

Environmental awareness has grown sharply over the last decades, and environmental legislation is introducing and tightening standards in many fields. Public participation and the right to know are mandated by law in many cases. This makes environmental information an important element of the policy making process in a civic society.

The history of pollution and pollution control dates back to early civilization with documented environmental regulations of Greek and Roman times, and a number of examples from the middle ages and subsequent periods. With the industrialization in the late 18^th and 19^th centuries the effects of pollution were becoming more noticeable, affecting ever larger proportions of the population. However, the biggest push towards a new attitude to the environment started with the environmental movement in the United States in the nineteen sixties, culminating in Rachel Carsons Silent Spring, Earth Day, and the US National Environmental Policy Act of 1971. The Club of Rome, Limits to Growth the foundation of the United Nations Environment Programme (UNEP), the 1972 Stockholm Conference, UNCED's 1992 Rio Conference and its Agenda 21, and a growing list of international accords on the environment illustrate and document this development.

Societies response to the perceived (and some very obvious) environmental problems is the introduction of laws and regulations. Regulatory instruments or control policies take a variety of forms. These include a range of monetary instruments from taxes to subsidies, laws and regulations including planning requirements and process oriented controls such as the requirements of best available technology (BAT), and more recently, mechanisms of self-regulation, the reliance on voluntary compliance and control, basically achieved through social and eventually market mechanisms and pressures rather than any central government administered enforcement policy. Examples from the extensive and continuously developing environmental legislation of the European Union are presented as illustrations of these basic boundary conditions for any environmental decision support system. Another important aspect of environmental legislation, however, are the provisions for free access to environmental information and public participation in decision and policy making processes, as they directly affect the design and implementation of environmental decision support systems.

Having established that there is a regulatory framework, and thus real need for environmental information (in the broadest sense), the main issue is how to obtain - and communicate - policy and decision relevant information in a reliable, timely and cost effective way to the different audiences and participants in the decision making processes, ranging from the technical specialist to the general public.

At the same time, information is becoming a commodity and a service in a explosively developing information society. Computing power abundant and cheap, and access to broad-band communication ubiquitous. These trends define the framework for the potential, and role of environmental information and decision support systems in the future.

Environmental information and decision support systems have to be understood, and developed, in a context that includes environmental problems as the primary focus. Their development must equally consider the nature of planning and decision making processes and thus the users and audiences of such systems. And it can exploit the information technology that provides the tools for their implementation.

The tools that provide the conceptual framework are drawn from the domain of applied systems analysis; models in a general sense are at the core. This leads to problems of uncertainty, and reliability, and thus usefulness, of model based results for decision support. The challenge is to incorporate and exploit information on uncertainty.

Due to the nature of environmental models, i.e. their complexity, nonlinearity, dynamics and spatially distributed nature, many of the classical methods of error analysis are difficult to apply as they require differentiable models. As an alternative, an approach based on Monte Carlo methods, that is computerized trial and error can be used for model identification, parameter estimation, and error propagation in forecasting.

Uncertainty in the context of environmental modeling as the main tool underlying environmental decision support systems is discussed in terms of sources of uncertainty; model identification and calibration; the propagation of uncertainty in forecasting and extrapolation, sensitivity and robustness, followed by a discussion of practical implementation based on Monte-Carlo methods.

The basic principle of Monte Carlo as used and presented here is a simple and straight-forward trial-and-error approach for problem solving. The problem is to identify an acceptable model structure, and to determine and acceptable parameter vector for the model. The conceptual framework used for the practical implementation is based on a simple structure of the modeling process. It involves:

1. A hypothesis or universal statement (the model structure);
2. A set of initial conditions, including:
   • the initial conditions sensu stricto, i.e., the state vector at time t=0;
   • the parameters, i.e., the constants describing the relationships between the elements of the state vector;
   • and in the case of dynamic models, time variant inputs or forcings;
3. A set of singular statements (the model output)

The practical procedure is as follows: For a given model structure, a parameter vector is sampled randomly from an a priori defined parameter space; the model is solved, and the result evaluated according to a set of rules,i.e. an objective function. The approach resembles that of natural selection fueled by (more or less) random mutation. The example presented demonstrates the application of the Monte Carlo approach to the identification of an appropriate model structure, combined with simultaneous parameter identification. Using a 15 year data set from the North Sea, a description of plankton dynamics (in terms of phosphorus and organic carbon) is formulated as a set of constraint conditions. A series of model formulations of increasing complexity, starting with a simple two compartment model, is then tested against this description, using a set of a priori parameter ranges. Model structure and parameter sets estimated by this method can then be used within the generic decision support framework introduced below as a mechanism to, on the one hand, represent uncertainty in a simple and straight forward manner, and on the other to solve some classes of inverse problems where traditional optimization methods may be difficult to use.

Decision support systems are a loosely defined group of tools, and include information systems,tools for scenario analysis, and optimization approaches. Again, as the complexity of environmental systems often leaves simulation as the only feasible approach, solving invers optimization problems analytically for the typical multi-criteria situations found in environmental problems is difficult to impossible, and would require drastic simplifications of problem representation. As an alternative, discrete multi-criteria methods can be used on scenarios generated with simulation models of arbitrary complexity.

Environmental problems require a new approach to decision support for two fundamental reasons:

1. it is impossible to solve the inverse problem (HOW TO) directly due to the complexities of the systems;
2. it is impossible to solve decision problems unequivocally due to the complexities and changing nature of the decision making process itself.

Bringing together sectoral models within an economic framework to find feasible and efficient solutions, and sectoral and global optimization approaches, the latter based on a discrete multi-criteria approach, provides a flexible tool to address dynamic and spatially distributed environmental management problems.

This leads to a multi-tiered, iterative, adaptive and interactive and in part and heuristic approach that combines simulation and optimization together with spatial analysis through geographic information systems (GIS).

The generic approach can be summarised as follows:

1. Obtain a description of the input change, usually but not necessarily a stress such as emission scenarios for different sectors (energy, transportation, industrial activities, development projects) through a range of tools and approaches such as environmental impact analysis or optimization models maintaining economic efficiency and meeting sectoral objectives and constraints. This is based on rational economic behavior of realistic agents at the sectoral level.

   In the case of air quality management, the sectoral optimization approach is based on the use of models for energy, e.g., MARKAL optimizing an energy demand-supply balance for the city including bounds on pollution emissions, and for transportation, e.g., EMME/2 simulating traffic equilibria which again constitute an emission scenario.

2. Get a representation of the resulting environmental systems response, for example, ambient air quality as a function of these emissions for different averaging periods, relating to air quality standards through solving the dispersion equations with spatially distributed air quality models or the equivalent simulation models for other environmental domains.

3. Obtain spatially distributed measures of impacts such as exposure, in general environmental and public health impacts based on land use information and a population distribution, resulting in a spatially distributed measure of vulnerability or damage function (or, more generally, a cost function), for different classes of stress levels or pollutants.

4. Minimize this distributed impact function subject to economic constraints by distributing the maximum acceptable costs. Alternatively, subject to environmental quality constraints, allocate the maximum permissible emission levels to traffic and energy uses while maintaining economic efficiency. In the case of air quality management, this global optimization takes emission scenarios (un-mitigated or sectorally optimized) as its starting point and is based on a source-receptor matrix computed by a long-term air quality model.

5. Use the permissible stress or emission levels obtained as a constraint on the previous sectoral optimization. Feedback loops from the impact analysis and global optimization can help to redefine objectives and constraints of the sectoral models.

6. Repeat the above steps to obtain a number of (sectorally) optimal and thus feasible scenarios. Where feasible, this may be done with a Monte Carlo approach as outlined above.

7. Use a discrete multi-criteria tool to find a preferred compromise solution (trade-off between sectoral aspirations) that satisfies the objectives of all groups of actors.

The generic framework presented must be implemented as an operational set of tools for any practical application. Systems architecture, based on distributed client-server system and an object-oriented design provide the framework, multi-media formats the user interface that can reach the diverse audiences involved in environmental decision making processes.

This architecture implements, and thus reflects the main features of the DSS approach presented above: multi-layered (and distributed), open and flexible (adaptive), and interactive with a multi-media user interface. The implementation must also support the integration of a variety of tools including models, GIS and expert systems, and their shared data bases.

The system architecture proposed for environmental information and decision support systems is presented at two levels:

1. On a conceptual level that describes the logical relationship of the elements (objects) used to represent, and manage, an environmental problem in a decision support system; this is based on an object-oriented design;
2. On a physical implementation level, that provides the actual operational hardware and software environment for running the decision support system; this is based on a client-server architecture that integrates various distributed information sources and supports different types of (display) clients.

The core of the proposed approach are environmental simulation models that translate stress like emissions from techno-economic processes into environmental impacts. Classical modeling approaches are described for atmospheric and aquatic systems, where transport (dispersion) processes represented by partial differential equations are used. Probabilistic extensions are used for risk assessment, and all the primary tools can be linked and aggregated into policy oriented models.

Different types of environmental models are introduced, with a detailed presentation of classical dispersion type models: basic air quality models and water resources models, including water allocation and groundwater flow and transport modeling.

The basic principles behind these models are the conservations laws, resulting in systems of partial differential equation describing the flow of environmental media (air, water), which in turn transport and diffuse pollutants. The control of this basic process transforming emissions into ambient concentrations (immission) is at the core of a large class of environmental problems.

Going beyond the classical numerical environmental models which are primarily process oriented, geographic information systems provide the tools for spatial analysis. They can be linked to dynamic and spatially distributed simulation models to extend their basically static analytical capabilities.

Geographic information systems are tools to capture, manipulate, process and display spatial or geo-referenced data. They contain both geometry data (coordinates and topological information) and attribute data, ie., information describing the properties of geometrical spatial objects such as points, lines, and areas.

In GIS, the basic concept is one of location, of spatial distribution and relationships, the basic elements are spatial objects. In environmental modeling, by contrast, the basic concept is one of state, expressed in terms of numbers, mass, or energy, of interaction and dynamics; the basic elements are ``species'', which may be biological, chemical, and environmental media such as air, water or sediment.

Another class of tools are rule-based expert systems: they use symbolic logic for deductive inference, and can express qualitative concepts and conditional relationships that are difficult to formulate in purely numerical terms.

Expert systems are empirical and heuristic systems, based on a more or less explicit, and usually qualitative, understanding of how things work. A perfect example of an ideal application area is law, or in the context of water resources, water rights and allocation problems. However, it is important to realize that expert systems are certainly no substitute for many time-tested quantitative and physically based methods and models, but should be seen as complementary techniques which can improve many of these models. Obvious applications related to numerical models are in data pre-processing, parameter estimation, the control of the user interface, and the interpretation of results. There are certainly enough arts and crafts components in numerical modeling that open attractive opportunities for AI techniques. Within the multi-tiered approach to decision support presented, expert systems can play an important part for many relationships that are difficult to quantify.

The integration of these different tools, i.e. models, GIS, and expert systems provides the basis for a new generation of powerful information and decision support systems. This is illustrated for a set of application domains.

State-of-the environmental reporting, and its extension to systems for early warning provides a classical example for environmental information systems and a rather passive decision support paradigm through awareness raising.

State-of-the-Environment reports address a broad range of diverse themes or topics. They need to be presented (but also compiled) within a consistent framework that ensures completeness and consistency of the information presented. Different organizational frameworks have been developed, including:

• issues such as environmental problems, e.g., land degradation or various forms of pollution;
• environmental media such as air, water land, fauna and flora;
• resource sectors such as agriculture, forestry, fisheries, mining, and recreation;
• environmental processes, such as stress and response as a consequence of human activities;
• various conceptual or formal models, emphasising interactions and feed backs;
• combinations of more than one approach.

Environmental impact assessment, and its probabilistic extension into risk assessment are typical examples of scenario analysis and basic decision support strategies such as ranking and benchmarking, and compliance monitoring.

There is a well defined regulatory framework for, and an extensive literature on impact assessment, with abroad range of methods to ensure completeness and consistence for the comparative evaluations. However, an environmental impact statement in itself does not yet contain a direct decision support component, since it only analyses, and presents, alternatives. The actual decision about the selection, or rejection, of the alternatives evaluated is part of the regulatory and administrative framework involving public hearings, and thus the political domain.

Finally, the power and flexibility of integrated decision support systems is demonstrated in three case studies. They cover an example of technological risk assessment, where the decision support is explicit and based on a rule-based expert system that provides real-time instructions for emergency management. The example includes a case study in Switzerland for hazardous goods transportation by rail along the North-South Alpine corridors.

In the example, decision support is provided based on a combination of a forward-chaining rule-based expert system and simulation models. This set of tools represents an evolving emergency as a combination of discrete events and continuous simulations. The latter use the Monte Carlo approach introduced above to provide representations of uncertainty. The objective of the system is to aide emergency managers and field personnel by providing both background information on the emerging incident as well as forecast on the likely evolution and the expected impact as well as the effectiveness of specific intervention measures.

An urban air quality management system uses both scenario analysis and simple comparison as well as classical optimization for a case study of emission control and district heating within the regulatory framework of environmental impact assessment in the City of Vienna and in Berlin.

Several layers of decision support within a well defined regulatory framework for urban air quality assessment and management are illustrated here: Analyzing continuous monitoring data provides information on compliance with national and European standards, and thus the basis for any intervention need. Data analysis is augmented by simulation modeling that provides the necessary spatial coverage and resolution for a city-wide assessment. Managing a consistent emission inventory is the basis for a range of regulatory decisions regarding operating permits for industries and commercial entities generating emissions.

For specific projects, like the extension of a major power plant to provide not only electricity, but also heat for district heating, impact assessment is based on the simulation and comparative analysis of a range of design and implementation scenarios. Simply comparing the resulting computed concentration maps, and showing the difference between scenarios is improvement or deterioration, provides a very intuitive basis for policy level decisions.

Finally, models together with the monitoring data are used for short-term forecasting of summer ozone concentrations. Within a well defined regulatory framework of thresholds alarm levels, and associated actions including the use of alternative fuels for industries and traffic restrictions, these tools provide direct input for operational decisions.

The examples also demonstrates the integration of a range of alternative models for different aspects of the overall air quality management problem in a city. The integration of these different models within a common framework of data bases and GIS turn a set of analytical tools into a practical decision support system.

An a regional development planning system finally brings together a range of data bases and analysis tools, simulation and optimization models in a case study of five year planning for Shanxi Province in China.

This study again is based on the use of a diverse set of models and expert systems that share a common data base and user interface. The coordinated (if not optimal) development of a region, and its industrial structure in particular, requires the simultaneous consideration of numerous inter-relationships and impacts, eg., resource requirements, environmental pollution, and socio-economic effects.

Plans and policies for a rational and coordinated development need a large amount of background information from various domains such as economics, industrial and transportation engineering, and environmental sciences. To be useful, this information must be presented in a readily available format, directly usable by the planner and decision maker. However, the vast amount of complex and largely technical information and the confounding multitude of possible consequences and actions taken on the one hand, and the complexity of the available scientific methodology for dealing with these problems on the other hand, pose major obstacles to the effective use of technical information and scientific methodology by decision makers. No single tool can address this range of problems, so that the generic decision support approach presented above fully applies.

The integrated set of software tools developed for this project is designed for a broad group of users with diverse, including non-technical, backgrounds. Its primary purpose is to provide easy access and allow efficient use of methods of analysis and information management which are normally restricted to a small group of technical experts. By combining numerical and symbolic methods, a synthesis of rigorous mathematical treatment and human expertise and judgement is obtained in a new generation of hybrid information and decision support systems.

The decision support system (DSS) is based on information management and model-based decision support. It envisions experts as its users, as well as decision and policy makers, and in fact, the computer is seen as a mediator and translator between expert and decision maker, between science and policy. The computer is thus not only a vehicle for analysis, but even more importantly, a vehicle for communication, learning, and experimentation.

The three basic, interwoven elements of the DSS are:

1. to supply factual information, based on existing data, statistics, and scientific evidence;
2. to assist in designing alternatives and to assess the likely consequences of such new plans or policy options, and
3. to assist in a systematic multi-criteria evaluation and comparison of the alternatives generated and studied.

The examples demonstrate that rather complex environmental decision support systems are feasible. This is due, in part, to the explosive development of information technology. Broadband communication, and the Internet are further changing the role and nature of environmental information.

The discussion analyses the trends that provide the context for environmental decision support systems. Not only do environmental problems continue to pose challenges, they get more difficult as the technological and economic constraints as well as the regulatory framework get tighter.

At the same time, the nature of policy and decision making changes into a more participatory style. Different audiences and the direct public use of scientifically based information in increasingly open decision making processes call for new presentation formats, a new style of argumentation, but also a new role for the scientist.

The main change in attitude and approach is the integration of scientifically based analysis with public policy and decision making processes. This forces scientists to leave the fabled ivory tower, and present their findings to a lay audience. This requires effective communication beyond a peer group, and increasingly this is achieved world-wide over the Internet.

This changes the rules of the traditional academic discourse radically. The detached, objective analysis subject only to peer review and judged on a fine balance of innovation and a solid foundation in tradition as well as elegance, becomes embroiled in daily politics and is judged by public acceptance, usefulness, efficiency and effectiveness.

Concepts of correctness, precision, observability, experimental verification, and least-square correspondence with data, completeness, convergence, formal proof, or optimality are confronted with concepts of political feasibility, cost efficiency, expediency, acceptability (that is the ability to sell it to a majority), or simply good enough.

In the context of multi-criteria decision support, it is easy to demonstrate that given a set of feasible alternatives, the efficient solution depends on the choice of criteria. This choice is ultimately a political one. What a decision support system then contributes is not so much an efficient mechanism to find an optimal solution given any set of (more or less) debatable preferences, but a mechanism to make the entire process more accessible, open and transparent.