Continuing lines : application of heuristic optimisation techniques for spatial environmental problems with multiple objectives

Management of hydrological and other natural resources is becoming increasingly complex because of their increasing scarcity, the increasing number of actors and objectives involved, and because of the increasing rate of change of technological, environmental and economic conditions. The risks of choosing suboptimal solutions have become more pronounced because of this increasing complexity. The hypothesis of this thesis is that application of heuristic optimisation techniques to complex spatial environmental problems with multiple objectives can improve the identification of (near) Pareto-efficient solutions and thus contribute to more effective decision making. The study consisted of four case studies, of which three comprised the application of genetic algorithms. The fourth case consisted of a sequential game experiment. It was investigated whether genetic algorithms can be applied successfully to a suite of complex optimisation problems in the environmental and hydrologic fields. The pro’s and con’s of the use of these techniques and the conditions for effective application were studied, particularly focussed on handling multiple objectives and validation of results. In case study 1 the calibration of a groundwater model was approached as a multiple objective optimisation problem. Regional drinking water production was formulated as a multiple objective optimisation problem in case study 2 and case study 3 consisted of various optimisation problems concerning allocation of agriculture and nature landuse types. In the fourth study, the optimal strategy for the prioritisation of groundwater quality prediction studies was searched with a set of game experiments that enabled simulation and evaluation of various strategies. The results of the case studies confirm the hypothesis of this dissertation that the application of heuristic techniques to complex optimisation problems in spatial planning and resource management enables better, more efficient decision making. The genetic algorithms that were built specifically for the case studies provided a powerful, stable and flexible optimisation technique. Pareto-optimality and uniqueness of solutions proved to be effective, unbiased fitness criteria for identifying trade-off curves. In these three case studies a certain degree of tuning of the genetic algorithm was necessary for the more complex versions of the problems. Consequently, it is considered essential to validate the results of heuristic optimisation techniques. Although complete validation is principally not possible, several ways of circumstantial validation could be achieved. Three different methods of circumstantial validation were applied: 1. Formulation of ‘dummy’ problems that are similar to the real optimisation problem, but constructed in such a way that one or more properties of solutions along the Pareto front of the ‘dummy’ optimisation problem are known; the results of the optimisation of the real problem can thus be partially validated. 2. Analytical inspection of particularly the extreme ends of Pareto fronts. 3. Application of the genetic algorithm to similar, but simplified problems that allow application of other techniques, such as linear or nonlinear programming, while maintaining a similar degree of difficulty of the modified problem from the viewpoint of optimisation by a heuristic technique