全球变化知识资源中心

Due to the widespread uncertainties in agricultural water resources systems and climate change projections, the traditional optimization methods for agricultural water management may have difficulties in generating rational and effective optimal decisions. In order to get optimal future agricultural water allocation schemes for arid areas with consideration of climate change conditions, the model framework established in this paper integrates a statistical downscaling model, back propagation neural networks, and an evapotranspiration model (the Hargreaves model) with inexact irrigation water allocation optimization model under future climate change scenarios. The model framework, which integrates simulation models and optimization models, considers the interactions and uncertainties of parameters, thereby reflecting the realities more accurately. It is applied to the Yingke Irrigation Area in the midstream area of the Heihe River Basin in Zhangye city, Gansu Province, northwest China. Then, water allocation schemes in planning year (2047) under multiple future Representative Concentration Pathways (RCP) scenarios and the status quo (2016) are compared, in order to evaluate the practicability of generated water allocation schemes. The results show that the water shortages of economic crops are improved compared with the status quo under all RCP scenarios while those of the grain crops present opposite results. Meanwhile, the economic benefits decrease from the status quo to planning year under all future scenarios. This phenomenon is directly related to the amount of irrigation water allocation and is indirectly related to the changes of meteorological conditions. The model framework can reveal the regular pattern of hydro-meteorological elements with the impact of climate change. Meanwhile, it can generate irrigation water allocation schemes under various RCPs scenarios which could provide valuable decision support for water resources managers.