Modeling watershed management with an ecological objective - a multi-agent system based approach

Hydroecology;Multi-Agent System Modeling;Water Resources Management;Decentralized Optimization;Human Impact

Yang, Yi-Chen

关键词: Hydroecology;    Multi-Agent System Modeling;    Water Resources Management;    Decentralized Optimization;    Human Impact;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/16119/1_YANG_YI-CHEN.pdf?sequence=2&isAllowed=y

美国|英语

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

【 摘 要 】

Watershed management is a practice that administers water and land resources within a watershed context for different users under spatial heterogeneity and temporal variability.Meeting water demands for both human and ecosystem and maintaining the harmonic balance between those two is one of the ultimate challenges for watershed managers. This study quantitatively defines both human and natural water users as agents in a watershed and then uses a multi-agent system (MAS) modeling framework to formulate the water allocation issue among the agents.A decentralized optimization algorithm is developed to evaluate the complexity of the watershed level outcomes resulting from different agents’ decisions.The construction of ecology-hydrology quantitative connection is the very first step to incorporate the ecological concern into the watershed planning.A data-mining approach: genetic programming (GP) is applied in this study to addressing this task. Using the Indicators of Hydrologic Alteration (IHAs) to represent the environmental flow condition, genetic programming generated a quantitative relationship (equation) between natural flow regime metrics (hydrology) and fish community indices (ecology).After a robustness test shows that each metrics in the equation had a consistent relationship with the ecological index, the GP generated equation is proofed as a direct linkage between hydrology and ecology and can reflect the ecosystem status under given flow situation. Meanwhile, the verification test shows a consistent outcome between GP, a traditional statistic method: principal component analysis (PCA) and an ecological approach: autecology matrix (AM).This common outcome provides confidence in using existing and new approaches and observational data to build solid hydroecological relationships. This method is applied to a real world case study regarding using daily reservoir operation for ecosystem restoration purpose.A multi-objectives optimization model is constructed for a case study reservoir in Illinois, USA to evaluate the tradeoff between human economic purpose and ecological conservation purpose.In general, adding the ecological objective into case study reservoir operation will not jeopardize the original economic objective, which is the major concern of current reservoir operation. Meanwhile, the result can also improve the downstream fish habitats by providing a flow regime that close to the natural condition.Due to the spatial heterogeneity involved in the management issue, dividing the watershed into interconnected subsystems (agents) should be a more realistic approach for analyzing water resources management problems.Agent can be flexibly defined in any appropriate format to characterize the heterogeneous water usage.Agents in the watershed are different types of water users that include offstream human land and water users and instream water “users,” such as riparian aquatic ecosystems.Under the MAS framework, both individual agents’ objectives and the interconnection among these objectives would be addressed using mathematical format.A penalty-based decentralized optimization algorithm is used to solve the formulated mathematical problem.By introducing the local interest factor (βi), this algorithm allows agents to search for the solution in the infeasible solution space when system-wide constraints are violated.The use of βi values enables the analysis of a water resource management problem to include the impact of the various agent behaviors that βi values reflect. Aggressive agents with large βi values affect other agents, particularly the reactive ecosystem agents, and even the entire system.In the real world case study of the Yellow River Basin, China, the concept of βi has been translated into local water price (pi).The local water price, served as an institutional arrangement, is used to guide the convergence of the decentralized management.The results of different scenarios analysis shows that equilibrium local water prices can be achieved under physical constraints in the watershed that will result in lower water consumptions and higher economic benefits compare to the baseline scenario.The subsidies from the government can not only guarantee the environmental flow requirement, but also increase the system GDP at the same time.Data limitation is always a major concern for watershed management modeling analysis.The construction of quantitative hydroecological relationship using genetic programming requires long-term fish species data.The utility function for the human water agent also needs actual water consumption and economic benefit data.The modeling results can be further improved if more data become available.Incorporating the concept of uncertainty into the MAS framework is the reasonable next step of this study.Agents' decisions will become dynamic and the results should be more close to the real world situation.