Solving large-scale spatial optimization problems in groundwater management

spatial evolutionary algorithms;ecosystem management;vegetation-groundwater interactions;genetic algorithms;spatial optimization;large-scale;pumping management;spatial information

Wang, Jihua

关键词: spatial evolutionary algorithms;    ecosystem management;    vegetation-groundwater interactions;    genetic algorithms;    spatial optimization;    large-scale;    pumping management;    spatial information;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/44422/JIHUA_WANG.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Large-scale groundwater management problems pose great computational challenges for decision making because of the spatial complexity and heterogeneity. The major output of this thesis is a modeling method to solve large-scale groundwater management problems using a newly-developed spatial evolutionary algorithm (SEA). SEA incorporates the spatial information of hydrological conditions with the design of evolutionary algorithm (EA). The algorithm employs a hierarchical tree structure to represent large-scale spatial variables. It is designed to capture spatial characters with reduced data volume by focusing on the important subsets of the entire system. This focusing results in an efficient representation and reduced computing time. Furthermore, special crossover, mutation and selection operators are designed to accommodate hydrological patterns and are in accordance with the tree representation.A hypothetical optimization problem is used to illustrate the encoding of spatial dataset and the detailed procedures of the SEA operators in Chapter 2. This chapter discusses how SEA employs a hierarchical tree structure to represent a spatial dataset in a more efficient way. It illustrates the SEA crossover, mutation and selection operators in details with an example.Chapter 3 shows how this method is applied to searching for the maximum vegetation coverage associated with a distributed groundwater system in an arid region. Vegetation in arid riparian zones heavily depends on groundwater availability, while at the same time the distribution of vegetation impacts groundwater flow. This chapter describes a methodology to characterize these groundwater-vegetation dynamics using the newly developed SEA. This method incorporates spatial patterns of groundwater and vegetation distribution to facilitate the optimal search of vegetation distribution compatible with groundwater depth. The SEA is applied to searching for maximum vegetation coverage associated with a distributed groundwater system in an arid region. Computational experiments demonstrate the effectiveness of SEA for large-scale spatial optimization problems.Chapter 4 discusses how this method is extended to a discrete spatial optimization problem and applied to the operation management of irrigation pumping wells in the Republican River basin, Nebraska. Sustainable management of groundwater resources is of crucial importance to irrigated agriculture in arid regions. This chapter focuses on optimizing the pumping strategy, including the placement and operations of a large number of pumping wells, to alleviate flow depletion and associated ecological damages in streams. The SEA is employed to optimize decisions on operating a large-scale irrigation pumping plan. The case study is based on the Republican River basin (RRB), where excessive irrigation pumping has led to both ecological damages in the streams and legal conflicts over water rights in this basin. More than 10,000 pumping wells are optimized simultaneously. The pumping yield of all the wells can be determined within the modeling framework of SEA. The physical system of coupled groundwater-surface water is simulated using a transient MODFLOW model that contains more than 215,160 grids and 2,903 stream reaches. The groundwater management problem is defined as a single-objective optimization problem to maximize total pumping yield under the regulations of ecological streamflow requirements. The results from the case study basin show that the problem with large-scale groundwater management model can be effectively solved by the SEA.This chapter includes some results with different streamflow requirements.Chapter 5 summaries the major research findings in this thesis. The developed SEA framework is efficient and effective for the spatial optimization of large-scale groundwater management. Two case studies are illustrated in chapters 3 and 4.However, it has some limitations and can be refined and extended by integrating advance spatial regression models and sophisticated management models in order to solve very complex systems. This chapter also discusses the intellectual merits and broad impacts on other large-scale water resources management problems.