Study On Evolution And Prediction Of Groundwater Depth In Huaibei Plain Of Anhui Province

As an important grain producing area in Anhui Province,Huaibei Plain had promoted the sustainable development of agriculture in our province and played a major role in the protection of water resources in the region.In order to avoid groundwater in the Huaibei Plain,there were a series of problems such as uneven distribution,serious mining,and formation of groundwater dropping funnel.The Kriging interpolation method was used to generate the average annual buried depth distribution map of shallow groundwater in the Huaibei Plain and the spatial distribution map of groundwater depth in six periods,which were 1978-1982,1983-1987,1988-1992,1993-1997,1998-2002 and 2003-2007.The SPA-GA-BPNN model predicted the short-term groundwater depth in Mengcheng County and Suzhou City.It was compared with five models,including full-variable Linear Regression,full-variable BPNN,full-variable GA-BPNN model,and SPA-BPNN and SPA-LR.It could be seen from the average buried depth distribution map for many years.The groundwater level in the Huaibei Plain was gradually decreasing from the southeast to the northwest.The groundwater depth was the shallowest in the southern part and the deepest in the north.The groundwater in the Huaibei Plain was extremely rich from 1978 to 1982,and the most dry from 1983 to 1987.Since 1988,the groundwater in the Huaibei Plain had entered a period of continuous recovery.The hydraulic gradient had gradually decreased from northwest to southeast combined with the groundwater depth of the six periods.The intergenerational dynamic changed of groundwater depth were affected by precipitation,and they were more affected by human activities from 6 key observation wells in Mengcheng County and Suzhou City.The SPA-GA-BPNN model predicted the result that were evaluated using MPAE,MSE and NSE.The results showed that the prediction accuracy of SPA-GA-BPNN model reached a high standard.The values of MPAE,MSE and NSE in the prediction stage of Suzhou were 0.077,0.103 and 0.848 respectively.The values of MPAE,MSE and NSE in the forecasting stage of Mengcheng County were 0.088,0.068 and 0.906 respectively.The model had good generalization ability,high prediction accuracy and good stability.In the SPA model screening independent variables,the SPA model screening independent variables was an important aspect to improve the prediction performance,which layed a foundation for improving the prediction ability and prediction accuracy of the model.Continuous optimization of the model could greatly improve the stability and accuracy of prediction through the comparison of SPA-LR,SPA-BPNN and SPA-GA-BPNN model.The overall stability and precision of the model applied in Mengcheng County with inter-generational variation of groundwater depth being small was better compared by Mengcheng County and Suzhou City.The SPA-GA-BPNN model showed high accuracy and stability in groundwater depth prediction,and provided an effective method and reference for groundwater depth prediction.