Simulation And Analysis Of Agricultural Drought Disasters In China Based On Artificial Neural Network

Drought not only affects the balance of the natural ecological environment,but also seriously impacts China's agricultural production,which is one of the key factors restricting China's agricultural development.Traditional drought monitoring methods mainly focus on monitoring single drought response factors such as soil moisture or vegetation status.Comprehensive multi-factor drought monitoring research is still very limited,and it has regional limitations and cannot accurately reflect the regional drought dynamics of large-scale spatial continuity.Artificial neural network is a machine learning method,which has the advantages of fast learning process,fast calculation speed,good stability,and high prediction accuracy.Therefore,it is of practical significance to study the large-scale regional drought by using the advantages of neural networks.This study used annual evapotranspiration data from 1950-2015,monthly temperature and precipitation data from 1949-2015,and agricultural drought disaster data sets,using climate factors as independent variables,and agricultural drought loss rates as target variable to construct an Artificial Neural Network(ANN)model of agricultural drought monitoring based on climate factors.This study systematically analyzed the change trend of agricultural drought disaster loss rate in China in the past 60 years and its regional differences in south and north China,and discussed the dependence of agricultural drought loss rate on climate factors such as temperature and precipitation in order to understand the impact of climate factors on agricultural drought.The research results will be beneficial to achieve a wide range of drought monitoring,and provide a scientific basis for government and relevant departments to carry out drought prevention and disaster reduction.The main conclusions of this study are as follows:1.This study is based on the BP artificial neural network algorithm,using the monthly high temperature,low temperature,monthly heavy precipitation,and light precipitation as the input of the model during the crop growth period from January to October,and the agricultural drought loss rate as the model output.To construct temperature,rainfall,and agricultural drought monitoring models for 25 provinces(autonomous regions)of China,the coefficients of determination of the simulated and observed values on the test set are between 0.60?0.91.At the same time,using the heavy precipitation,the area percentage of light precipitation and the annual evapotranspiration from January to October as model inputs,a neural network model of agricultural drought monitoring for precipitation evapotranspiration was established.The coefficient of determination is between 0.62 and 0.96.Overall,the model has outstanding performance,so the model can provide a new method for regional agricultural drought monitoring.2.The spatial and temporal differences of agricultural drought in China are obvious,and the drought distribution has certain temporal and spatial heterogeneity.On the time scale(1949?2015),seasonality is obvious,there are many summer droughts,and the inter-decadal differences are significant.In space,the overall characteristics of the east over the west and the north over the south were presented.Agricultural drought in the northern region was more sensitive to climate change than in the south.Agricultural drought in the northern region can occur with less changes in precipitation conditions.3.The effects of climatic factors on agricultural drought were inconsistent.Due to the different dependence degree of crops on climatic factors in each stage and the uneven seasonal distribution characteristics of climatic factors,the agricultural drought loss rate was mainly affected by the changes of climatic factors such as precipitation and temperature in the critical period of crop growth(April?August),while the changes of climatic factors in other periods have no obvious impact on the agricultural drought loss rate.At the same time,the impact of temperature and precipitation on agricultural drought is not consistent.Nationwide,the difference in average temperature between drought years and non-drought years is 1.1%,and the difference in rainfall is 9.3%.It can be seen that the impact of precipitation on agricultural drought is more significant.4.We provided cumulative precipitation thresholds for agricultural drought in each province.The threshold can provide some decision support for related departments to monitor agricultural drought.