Data-driven Model Of The Water Use And Water Uniform-scarcity Risk Analysis In Shiyang River Basin

The Shiyang river basin is located in the inland arid region of Gansu province ofnorthwest China. The basin is in a typical continental temperate climate zone with low annualprecipitation, high evaporation and strong solar radiation. The limitation of water resources isone of the major challenges to the communities in the arid and semiarid regions of northwestChina. Understandably, an increasing demand for human production and life has always beenconcerning water resources managers. A better understanding of the influencing factors onregional water use is believed to be one of the key steps to address the dilemma of socialeconomical development and the limited water resources. Facing the water scarcity situationand the changing of the social-economic background, the history data information has beenchosen for this study. The study involves several steps, which are (1) to analyze the change ofregional water use by calculating the information entropy;(2) to select the importantenvironment variables using linear correlation coefficient and mutual information methods;(3)to build the model with different selection variables (input) to obtain the best modelsimulating the change of regional water use (municipal, industrial and agricultural water use);(4) with the four design scenarios, Monte-Carlo (MC) simulation were used for the futureprediction;(5) the water uniform-scarcity index was defined for the regional risk of water use,combine the history samples, the system of the risk assessment was build. Which the results:(1) The information entropy of water use represents the change of water use proportionand can be used for the regional comparison; the mutual information score is shown to bevery helpful in building MLRs for modeling water use with very limited data available.(2) Compared to linear and non-linear two models, Ensemble ANN model had lowererror and little range changes of the prediction value, which means Ensemble ANN model hadadvantages in the simulation of net irrigation requirement in this study. To the comparisonabout the upstream, midstream and downstream in Shiyang riverbasin: Minqin had the lowesterror and Liangzhou have the highest error; in contrast, Liangzhou had lowest ranges valuechanges, thus net irrigation requirement can be predicated better using data-based model insuch area, where had more human intervention and larger agriculture area. (3) The Monte Carlo simulation results from several scenarios, the risk of forecast thanthe status quo in water uniform-scarcity risk has reduced. The most significant reduced during2020-2030; the water uniform-scarcity risk varying degrees of recovery during2030-2050.Comparison of the variations in the environmental scenario design, under policy adjustmentreduced the water uniform-scarcity risk significantly; the water uniform-scarcity risk alsoreduced under climate change.It can be seen from the analysis above that the information entropy of water use can beused as a convenient measure to characterize the diversity of water use among differentsectors. Based on this, its evolution characteristics and the identity of the driving force can bebetter expressed. The concept of water uniform-scarcity risk combine the water use uniformand the water scarcity; under the complex situation from the development of regional social,economic, policy and climate change, the scenario analysis, neural network ensemble models,and MC simulation technology are intergraded to achieve the data-driven model of the wateruse and water uniform-scarcity risk analysis in Shiyang River Basin.