Integration Of ACA,RS And GIS For Spatial Optimal Allocation Of Water Resources

Spatial optimal allocation of water resources (SOAWR) plays a very important role in therational and effective use of water resources. It can promote the coordinated development ofpopulation, resources, environment and economy and the shift from engineering water toresources water. Also, the SOAWR can deepen and expand the optimal theory and method withmulti-objective and large-scale systems.The SOAWR is a complex problem with a multi-objective and multi-constraint model. Themodel must be solved and optimized in a large-range and large-scale region. And it needs toprovide a variety of alternative allocation schemes based on large amounts of data. However, howdo the optimal and spatial allocation schemes obtain to produce the maximum benefit? Therefore,the object of the study is to solve the problem of the SOAWR.In this study, an integration method of ant colony algorithm (ACA), remote sensing (RS), andgeographic information system (GIS) was used to resolve the SOAWR problem. Study area wasselected in Zhenping County, Henan Province, China. Water demand was classified, assessed andforecasted. And the model of SOAWR was established and solved based on the integration methodof ACA, RS and GIS. Different schemes of SOAWR were proposed on the micro unit of pixel.Then example simulations were completed and different intelligent optimization algorithms werecompared to verify the feasibility and effectiveness of the integration method of RS, GIS, andACA for the optimal allocation problem of water resources.The main conclusions of the study are as follows:(1) Pareto ant colony algorithm (PACA) was improved according to self-adaption dynamic updates for the local and global pheromones and was designed according to weight low-pass filter.The improved PACA makes the ant move towards the optimal border with stronger pheromones.Pheromones on the pixel where ant forage were adjusted based on multi-objective function value.And multi-objective optimization solutions were determined according to maximum fitness value.Comparisons of PACA with other intelligent optimization algorithms were done andevaluations of different allocation schemes were completed. The optimal performance, timeperformance and robust performance of the improved PACA are0.398,21.6and5.38, respectively.The results showed that the PACA is significantly superior to the evaluated results of the basicACA (2.108,36.8and8.16).Number of solutions, spacing and the maximum obtained from the improved PACA are389,0.68and183.58, respectively, which are superior to those obtained from genetic algorithm (GA)and BP artificial neural network (BPANN). It shows that the PACA could find the optimal orapproximate optimal solutions and cope with the contradiction among the convergence, prematureand stagnation problems. The PACA improved the global search ability and convergence speed.Different level years (2010,2020and2030), different guaranteed rates (50%,75%and95%)and different water sources (surface water, groundwater and water transfer) were comparedbetween the original water and optimized water. It indicates that the optimized results arereasonable and feasible and that the improved PACA has a high optimization performance.(2) Integration of RS, GIS and ACA can accurately determine the types and amount of waterdemand, so as to provide data support and decision making for optimal allocation, optimalscheduling, management and planning of water resources.Firstly, water demand types were classified based on Ant Colony Clustering Algorithm (ACCA) on the remote sensing image. Secondly, they were reclassified more accurately in detailaccording to field surveys and RS technology such as NDVI, NDII, NDBI and MNDWI. And inthe end, the amount of water demand was obtained using GIS technology. Total F-measure valuecalculated from the ACCA is0.918, which is more than those from the maximum likelihoodmethod (0.884) and the minimum distance method (0.851). Bigger F-measure value indicates thatthe classified results from the ACCA are more accurate and its noise immunity is stronger.(3) Integration of the ACA and PP prediction model of water demand can accurately obtainthe optimal parameter combination of the high-dimensional and non-linear PP prediction model.This method improves the fitting precision and prediction accuracy of water demand, whichprovide more accurate reference for the optimal allocation and scheduling of water resources. ThePP model is suitable for the regions where the system mechanism is not clear enough or hydrogeological data are lacked. Ant colony algorithm for continuous domain was designed andimproved to solve the optimal parameter combination of the PP prediction model. The algorithmcan better converge to global optimal solution, and have a strong resistance to noise and avoidpremature convergence.Results of case simulation show that the fitting accuracy of the prediction model obtainedfrom the ACA is very high. The absolute values of the relative error are all less than2%, mostly inless than1%. It is significantly better than those from the artificial immune algorithm (10%) andBP neural network (12%). The method can be extended to the solution of other high-dimensionaland nonlinear problems.(4) The spatial optimal allocation model of water resources is based on the ecologicaleconomics theory and sustainable development theory. The objective function of the model is to obtain maximization of social, economic and environmental benefits. The constraints of the modelinclude the amounts of various water supplies, the amounts of industrial water demands and thestandards of water quality. Therefore, the model integrates water quality and water quantity andcombines the economic, social and ecological benefits on the micro-unit pixels of the rasterimages, which is of multi-objective, multi-constraint, multi-level and multi-user characteristics.(5) Integration of RS, GIS and ACA can solve the complex SOAWR model and achieve avisual expression and validation of the reconstruction schemes and allocation results.Different allocation schemes upon pixels were achieved from solved model. These schemesinclude water demand, available surface water, available ground water, available transfer water, aswell as economic benefit, social benefit, ecological benefit and comprehensive benefit.Water balance analysis in different level years shows that the integration of RS, GIS and thePACA can solve large-scale, multi-objective optimal allocation model of water resources. Themethod makes the running time relatively shorter and makes the allocation results are closer toreality. Thus, these allocation results can provide a reference for the formulation of the watershedwater policy.Integration method of ACA, RS and GIS expands and deepens the optimal allocation theoryand method of water resources and the field of application for RS and GIS. It provides newapproach to the solution of the similar multi-objective and multi-constraint model.