Study On Spatial-temporal Characteristics And Prediction Model Of Reference Crop Evapotranspiration In Xinjiang

Reference crop evapotranspiration(ET0)is a key factor to determine crop water requirement.As a major agricultural province in China,Xinjiang is an arid and semi-arid region with extreme lack of water resources.The shortage of Crop supply and water resources puts forward higher requirements for accurate prediction of crop water.Accurate prediction of ET0 is the key to optimize the allocation of agricultural water resources,formulate accurate irrigation system and improve the utilization efficiency of water resources.In this paper,the daily meteorological data of 11 stations in Xinjiang from 1966 to 2016 and the solar radiation(Rs)data from 1993 to 2016 were collected.The ET0 calculated by FAO-56 Penman-Monteith(FAO-56 PM)formula was used as the standard value to analyze the temporal and spatial evolution of ET0 and meteorological factors and the meteorological influencing factors of ET0.A time series hybrid prediction model combining decomposition method and back propagation neural network model(BPNN)was proposed to predict ET0 in Xinjiang in short term.Using the hybrid model established above,five meteorological parameters,average temperature(Tmean),sunshine hours(n),solar radiation(Rs),relative humidity(RH)and wind speed(u),were selected to predict daily ET0,and the optimal meteorological parameters were determined to predict ET0 in short-term and evaluate the performance of the model.The main research results were as follows:(1)The variation of Tmean,Rs,n,u and ET0 during the year showed a single peak trend of first increase and then decrease.Except for the peak of u in spring,other meteorological factors and ET0 peak appeared in summer.The change trend of RH during the year was opposite,and the lowest RH was in spring.From 1966 to 2016,Tmean showed an overall upward trend,while n,RH and u showed an overall downward trend.From 1993 to 2016,Rs showed a decreasing trend and ET0 showed a significant upward trend.The spatial distribution characteristics of Tmean and Rs in Xinjiang were southeast high,northwest low,southern Xinjiang was greater than northern Xinjiang,and RH is the opposite.The main distribution characteristics of n and u were east high and west low.ET0 had the strongest periodic change around 6 a.There was a significant positive correlation between Tmean,Rs,u,n and ET0 in Xinjiang,and a significant negative correlation with RH.(2)The results indicated the superiority of the VMD-BPNN hybrid prediction model to EMD-BPNN and EEMD-BPNN in terms of accuracy and stability,with root mean square error(RMSE)=0.405 mm/d,mean absolute error(MAE)=0.268 mm/d,and coefficient of determination(R2)=0.979.When employing the VMD-BPNN model to forecast ET0 for seven days,the RMSE and Nash-Sutcliffe efficiency coefficient(NSE)of the VMD-BPNN model were 0.588 mm/d and 0.952,respectively,and the prediction results indicated high precision and reliability.The prediction accuracy of the VMD-BPNN model was significantly higher than that of single machine learning models,such as BPNN,SVR,and GBRT.The RMSE and MAE values of the VMD-BPNN model were more than 60%smaller than BPNN,SVR,and GBRT models,and R2 and NSE were approximately 18%higher than BPNN,SVR,and GBRT models,respectively.This demonstrates the effectiveness of the VMD method in reducing the nonstationarity of the original daily ET0 data.The BPNN model predicted the decomposed data series,and the prediction accuracy and stability were significantly enhanced.This indicates the high reliability of the VMD-BPNN model and its capability for ET0 prediction in Xinjiang.(3)Using the VMD-BPNN model,the prediction performance of Tmean,Rs,and n was relatively good,followed by relative humidity and wind speed.Tmean had the highest prediction accuracy and stability,with the variation ranges of R2,NSE,RMSE,and MAE were 0.88-0.93,0.86-0.93,0.57-0.89 mm/d,0.40-0.57 mm/d,respectively.The prediction error fluctuation was small,and the prediction performance was good at all stations.The prediction accuracy of meteorological factors in spring and autumn was higher than that in winter and summer.As the foresight period increases,the predictive performance of the model decreased.Comparing the VMD-BPNN model with three single machine learning models,BPNN,SVR,and ELM,the VMD-BPNN model had the best performance.In general,the forecast duration and season will have a certain impact on the accuracy of single meteorological factor forecast ET0.In the absence of meteorological data in Xinjiang,the VMD-BPNN hybrid model can be used to make a shortterm prediction of ET0 in Xinjiang using Tmean.