HR: 08:45h
AN: IN41B-03    [Abstracts]
TI: Solving Large-scale Spatial Optimization Problems in Water Resources Management through Spatial Evolutionary Algorithms
AU: * Wang, J
EM: jwang41@uiuc.edu
AF: Hydrosystems Laboratory, Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 2522 Hydrosystems Laboratory, 205 N. Mathews Avenue, Urbana, IL 61801, United States
AU: Cai, X
EM: xmcai@uiuc.edu
AF: Hydrosystems Laboratory, Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, 2522 Hydrosystems Laboratory, 205 N. Mathews Avenue, Urbana, IL 61801, United States
AB: A water resources system can be defined as a large-scale spatial system, within which distributed ecological system interacts with the stream network and ground water system. Water resources management, the causative factors and hence the solutions to be developed have a significant spatial dimension. This motivates a modeling analysis of water resources management within a spatial analytical framework, where data is usually geo- referenced and in the form of a map. One of the important functions of Geographic information systems (GIS) is to identify spatial patterns of environmental variables. The role of spatial patterns in water resources management has been well established in the literature particularly regarding how to design better spatial patterns for satisfying the designated objectives of water resources management. Evolutionary algorithms (EA) have been demonstrated to be successful in solving complex optimization models for water resources management due to its flexibility to incorporate complex simulation models in the optimal search procedure. The idea of combining GIS and EA motivates the development and application of spatial evolutionary algorithms (SEA). SEA assimilates spatial information into EA, and even changes the representation and operators of EA. In an EA used for water resources management, the mathematical optimization model should be modified to account the spatial patterns; however, spatial patterns are usually implicit, and it is difficult to impose appropriate patterns to spatial data. Also it is difficult to express complex spatial patterns by explicit constraints included in the EA. The GIS can help identify the spatial linkages and correlations based on the spatial knowledge of the problem. These linkages are incorporated in the fitness function for the preference of the compatible vegetation distribution. Unlike a regular GA for spatial models, the SEA employs a special hierarchical hyper-population and spatial genetic operators to represent spatial variables in a more efficient way. The hyper-population consists of a set of populations, which correspond to the spatial distributions of the individual agents (organisms). Furthermore spatial crossover and mutation operators are designed in accordance with the tree representation and then applied to both organisms and populations. This study applies the SEA to a specific problem of water resources management- maximizing the riparian vegetation coverage in accordance with the distributed groundwater system in an arid region. The vegetation coverage is impacted greatly by the nonlinear feedbacks and interactions between vegetation and groundwater and the spatial variability of groundwater. The SEA is applied to search for an optimal vegetation configuration compatible to the groundwater flow. The results from this example demonstrate the effectiveness of the SEA. Extension of the algorithm for other water resources management problems is discussed.
DE: 0430 Computational methods and data processing
DE: 0520 Data analysis: algorithms and implementation
DE: 1805 Computational hydrology
DE: 1813 Eco-hydrology
DE: 6339 System design
SC: Earth and Space Science Informatics [IN]
MN: 2007 Fall Meeting