Genetic Algorithm based multi-criteria decision support for the rehabilitation of the Gemenc floodplain in Hungary

Nature rehabilitation in general is a multi-objective problem. Besides ecological objectives, the interest of different human activities should also be taken into consideration. It is also important to make nature rehabilitation financially feasible, that is to minimize the costs and possibly to generate some income. Because of the inevitable conflicts between these objectives, the Decision Making Group (DMG) is faced with the difficult task of locating a compromise solution. This study deals with a specific nature rehabilitation problem: the rehabilitation of the standing water system of the Danube riparian Gemenc floodplain. The main objective is to restore habitats for characteristic wet flora and fauna as the area has undergone serious ecological deterioration, due to changes in the hydrological conditions. Rehabilitation is constrained by forestry, which require sustaining alluvial forests. On the basis of these general objectives several evaluation criteria were identified such as: habitat sizes of targeted flora and fauna, percentage of forest cover, but also investment and operation costs. The key to rehabilitation is improving hydrological conditions, since these are the most important abiotic factors for floodplain ecosystems. Modifying the hydrological regime is proposed to be implemented by appropriately transforming the water conveyance infrastructure. This infrastructure consists floodplain depressions (cells) connected to each other and to the river by means of small, usually ill-developed channels (so called ‘foks’). Such cell-system is filled and drained gradually, from cell to cell, according to the regime of the river. The location, threshold levels and geometry of the channels as well as the boundary elevations of the cells determine the water regime. Accordingly, hydrological conditions of the system can be improved by means of appropriate combination of measures like dredging the fok-channels, opening new ones, installing weirs/sluices, erecting summer dikes on cell boundaries etc.. Based on such measures, a decision space can be defined where each point stands for a particular cell-system alternative. Thus, the goal is to locate a compromise solution in the decision space according to the identified criteria. Searching for such solution is aided by a multi-criteria Decision Support System (DSS). The applied DSS is based on the ESEMOPS method. In ESEMOPS each criterion function is associated with one or two aspiration levels expressing the preference structure of the DMG. Based on these levels, certain number of surrogate objective functions are formulated. By optimising these functions, the DMG gains information concerning the set of feasible solutions, and learns about the acceptable interrelationships (trade-offs) between the criteria. This enables the DMG to adjust its preference structure and repeat the optimisations. The iteration stops when the DMG accepts a certain solution as best compromise solution for the problem. The optimisation problems are solved with the help of Genetic Algorithm (GA). The major advantage of GA comes from its ability to rely exclusively on information contained in the input (population of alternative solutions), and output (surrogate objective values) of criteria evaluations throughout the search. Auxiliary information, such as derivatives, is not needed. Evaluation of a particular solution within the DSS is carried out by an embedded system of hydrodynamic-ecological models.