Study On Rainfall Prediction Based On Wavelet Neural Network And Its Application In Agricultural Production

Owning to the special geographical position and climate environment,the frequent droughts and floods in Yueyang City has become a major obstacle refraining the development of local agriculture.Therefore,a deep understanding about the cause and regular pattern of droughts and floods occurrence,and an accurate objective prediction method for droughts and floods are beneficial in establishing the relevant warning system and can lay a foundation for reasonable allocation of water resources.This paper first analyzes the trend of precipitation change and identifies the condition of droughts and waterlogging based on the annual precipitation data and monthly average data in Yueyang City from 1986 to 2017.Then,we use the precipitation data from the station of Yueyang Building District to establish a precipitation prediction model for this district.The specific research contents and results of our study are as follows:(1)Characteristic analysis of precipitation variationAnnual and seasonal precipitation changes in Yueyang City are analyzed by means of sliding average method,cumulative anomaly method,precipitation trend rate,m-k(mann-kendall)significance test and m-k mutation test.From the annual perspective,the precipitation of Yueyang City showed a downward trend,and the distribution of rainfall within the year was uneven.From the seasonal perspective,the precipitation of Yueyang City shows an upward trend in spring and autumn,and a downward trend in summer and winter.The years with abrupt change in spring,summer and autumn are respectively 2004,1997 and 2011,nevertheless there is no abrupt change in winter.(2)Identification and evaluation of droughts and waterloggingThree indicators of drought and flood,namely precipitation anomaly percentage,Z index,and humidity index,were used to identify and evaluate the rainfall of Yueyang City in our research set 32 years.The results show that the frequency of droughts and floods occurred during the 32 years in Yueyang City was higher both in Z index and humidity index,and the frequency of droughts was basically the same as that of floods.Finally,according to the statistics of actual droughts and waterlogging disasters in Yueyang City,we found that Z index is the most suitable evaluation index for drought and waterlogging for Yueyang City.(3)Establishment of precipitation prediction modelBased on the nearest neighborhood sampling regression,wavelet neural network model,we effectively predicate and test the annual and monthly precipitation in the site of Yueyang Building District from 1986 to 2017 and conclude there are two methods are feasible in Yueyang Area rainfall prediction.However,the predict effect via wavelet neural network model is obviously better in general than that of the nearest neighbor model.The prediction results reflect the variation of precipitation in the process more truly.The wavelet neural network model is finally used to predict the annual and monthly precipitation in 2019-2021,and to evaluate the droughts and floods grade.(4)Causes of climate disasters and related anti-disaster policiesAccording to the comprehensive evaluation results of drought and waterlogging in Yueyang City in the past period,the main factors of droughts and waterlogging in this area were analyzed,and the corresponding suggestions are proposed for the coming droughts and waterlogging in 2019-2021.