Prediction Of Agricultural Drought In Jiangxi Province Based On NAR Neural Network

China is a big agricultural country.Since ancient times,agricultural disasters caused by drought have always been the main disasters in the development of agricultural production in China.In order to predict the trend of drought,effectively formulate drought resistance and disaster reduction countermeasures,effectively reduce agricultural losses,agricultural drought prediction research has practical guiding significance.Jiangxi province is a big agricultural province in China,but it is often damaged by drought,mainly agricultural drought.In this paper,the drought law of Jiangxi province is analyzed,and the agricultural drought prediction is made.Jiujiang,Yichun,Nanchang,Ganzhou and Nancheng are selected as the main research areas.The results are as follows:(1)Firstly,the trend and mutation tests of precipitation and temperature in each study area were carried out by using the method of unary linear regression,M-K test and sliding T test.The results show that the precipitation in each study area is sufficient,and the precipitation in northern Jiangxi,western Jiangxi and eastern Jiangxi has a tendency of increasing,but it is not significant.The precipitation in southern Jiangxi showed an insignificant decreasing trend.Only the precipitation in the middle of Jiangxi province increased significantly.The abrupt changes of precipitation in northern Jiangxi and southern Jiangxi were 1968 and 2003.The temperature in each study area increased significantly and increased year by year as a whole.Only northern Jiangxi and central Jiangxi had abrupt changes in temperature in1996 and 2003.(2)SPI and SPEI drought index were used to analyze the drought law of each research area in Jiangxi Province.The results showed that drought frequency was high in each study area from 1960 to 2018,and drought intensity was mainly light and moderate,while severe and special drought occurred less.The drought is more serious in northern Jiangxi,central Jiangxi and southern Jiangxi.In terms of seasonal scale,summer and autumn are more prone to drought than spring and winter.In general,the results calculated by SPEI index are more consistent with the actual situation of Jiangxi Province.This indicates that SPEI index has a strong ability to identify agricultural drought in Jiangxi Province,and is sensitive to agricultural drought,and can make different judgments on drought degree at different time scales.(3)By using NAR artificial neural network model to train and predict sp I-1and SPEI-1 indexes in each study area,the predicted values of SPI-1 and SPEI-1 were compared with the actual values of SPI-1 and SPEI-1,and the overall prediction results of the model were satisfactory.However,once the prediction period is too long,its prediction ability will decline.In a short forecast period,the prediction accuracy of SPI-1 and SPEI-1 index is higher,among which the prediction effect of THE model on SPEI-1 index is better,and the accuracy is up to 81.5% in the forecast period of108 months.The research results can be used for reference to predict the short-term drought trend in Jiangxi Province and help the development of agricultural production.