Evolutionary Bayesian Belief Networks for Participatory Water Resources Management under Uncertainty

Introduction

Water resources management are rarely well structured (Simonovic, 1996) as there is no definitive formulation, no true or false solution, and no easy test, screening procedure or single optimal solution for these problems (Chatttopadhyay & Chattopadhyay, 2008; Sivakumar, 2005). Therefore, successful environmental decision making in complex settings will depend on the extent to which stakeholders, processes and tools are integrated. Environmental decision analysis process involves several steps. In the first step the problem and objectives are identified. The boundaries should also be identified, however as decision making process is iterative some boundaries might be modified or new ones introduced as learning takes place over time. This will result in redefinition of the problem. In the second step a qualitative decision tree is developed based on contribution from experts (stakeholders, decision makers or scientists) that can help in identification of alternative options and their uncertainties. This will be followed by assigning probability to those variables in the decision tree whose states are not certain. These assignments are based on experts’ knowledge, existing data, modelling and monitoring. The next step is assignment of value (benefits, costs, risk etc.) to consequence of different alternative options that needs to be optimised. The final step is to find optimum strategy or strategies. Although in simple cases, the main steps of the process will easily be constructed, in complex cases, such as environmental decision analysis, this is not a straight forward task. Traditionally, in order to reduce level of complexity in systems, two very important elements of the systems were overlooked. Firstly, not all the elements influencing the decision making process are considered (over-simplifying the system), especially feedback loops. Secondly, the temporal nature of the problem is usually ignored despite the fact that the consequences of a decision often do not all occur simultaneously. Consistency check can help to develop some confidence in representation of the decision maker’s preferences. In checking for consistency it is important to detect errors in the decision making utility function. This is the most common source of error, meaning that it does not represent the decision maker’s true preferences. Methodologies to check the properties of completeness and redundancy of developed models based on their utility function are not well developed. Traditionally the decision maker will be asked to suggest a number of possible scenarios to check the expected utility of the preferred situation in order to be consistent. Often in practice, the decision maker may seem not to be interested in the entire possible range of the attributes except for a small variable space. For utility functions implying a complex preference structure, there is a greater need and opportunity for meaningful consistency checks. Another shortcoming of the existing techniques is the way circular relationships or feedback loops are handled. Management of real world problems requires approaches that take into account full complexity of the systems to be managed and the need to develop adaptive and integrated approaches. In order to succeed, it is important to take into account a wide range of factors that impact on the problem. It is equally important to identify the best way of linking these factors together and to simulate the interactions between them. Uncertainty is another important problem, which is an inherent feature of complex systems.