The spatially-distributed ANN-optimization approach for water-agriculture-ecology nexus management under uncertainties and risks

Abstract

water-agricultural-ecology nexus system has spatial heterogeneous characteristics owing to heterogeneities of soil, crop, and weather types. Such complexes aggravate the difficulties in agricultural water, land, and fertilizer resources optimization. To address the above problems, this paper developed the spatially-distributed ANN-multi-objective multi-preference fuzzy credibility constrained programming (distributed ANN-MOMPFCCP) for optimizing water, land, and fertilizer resources. It incorporates the multi-objective programming (MOP), multi-preferences credibility constrained programming (MPFCCP) with the distributed ANN model, and has obvious advantages in improving the calculation efficiency of the traditional spatially-distributed crop simulation-optimization model. Besides, it can address uncertainties of the crop simulation model by introducing stochastic programming. It builds connections between food yields and agricultural ecological effects by introducing negative and positive ecological benefits in study system. Through proposed model, optimal distributed schemes considering spatial heterogeneous net economic and ecological benefit, and associated integrated risks can be generated. Optimal results show that water shortage risk occupied about 20 % of total risks, economic and ecological loss occupied about 20~20.2 % and 20~21 % of total economic and ecological benefits, respectively. Considering uncertainties of crop simulation model was a high effective way to improve robustness of simulation model. The model could offer insight alternatives for regional managers in measuring distributed agricultural-ecological benefit as well as making tradeoff among economic, ecological benefits and risks. Moreover, this approach can help manage multi-type resources under uncertainties, and provide multiple groups of Pareto solutions for managers, so that managers could select best management practices based on respective risk attitudes, and preferences.