Research On Drought Monitoring Model Based On Multi-source Data

Shandong province is a big agricultural province.It is of great significance to the social and economic development of Shandong province to accurately monitor the occurrence of drought,the level of drought and the affected area of drought.Drought occurrence reason is relatively complex,comprehensive consideration of precipitation,vegetation cover,the surface temperature and soil moisture data,at the same time considering the atmosphere,vegetation and soil moisture and energy exchange between the problem,add MOD16A2 evapotranspiration data calculation of Crop Water Stress Index(CWSI),considering each factor to drought between linear and nonlinear relations.Taking the drought monitoring index calculated from multi-source remote sensing data from January to December of 2000 to 2019 in Shandong Province as independent variable,meteorological data calculation of drought monitoring index of target variable,deviation correction based on multivariate linear regression method,the random forest algorithm,support vector regression method and artificial neural network algorithm and so on the many kinds of model building method,At the same time,considering the difference of precipitation and temperature,a comprehensive drought monitoring model was constructed in Shandong province on a monthly basis.In order to verify the influence of the CWSI index on the construction model,four comprehensive drought monitoring models(DCI?LR,DCI?BRF,DCI?SVR and DCI?ANN)were constructed with or without CWSI index in the independent variables.In order to evaluate the drought monitoring ability of the model,the correlation analysis between the model results and the Standardized Precipitation Evapotranspiration Index(SPEI)was conducted.The drought grade was classified with the measured value CI of meteorological station to evaluate the monitoring ability of the model to the drought grade of meteorological station.The model simulated the spatial distribution of drought from April to October in 2002,a typical dry year.The optimal comprehensive drought monitoring model was selected and used to monitor and analyze the spatial distribution of drought in Shandong province from 2010 to2019.The main conclusions are as follows:1.The inclusion of CWSI index improves the fitting accuracy of DCI?LR ?DCI?BRF ? DCI?SVR and DCI?ANN models to varying degrees,among which DCI?BRF model has the highest accuracy.2.The correlation between DCI?BRF and SPEI was the highest,with the correlation coefficient ranging from 0.433 to 0.908,and the correlation of the four models from April to October was higher than that of other months.DCI?BRF model has the highest average agreement rate of drought grade reaching 92.06%.The four comprehensive drought monitoring models can accurately reflect the drought-affected areas in Shandong Province,and can reflect the drought degree and change trend of different regions to a certain extent.Based on the evaluation of the above monitoring capability,the optimal comprehensive drought monitoring model DCI?BRF was selected.3.Based on the cultivated land area in Shandong Province,the disaster area of cultivated land from April to October in the 2019 crop growing season in Shandong Province was analyzed by superposition statistics,among which May and July were the most severely affected.Based on DCI?BRF,the temporal and spatial characteristics of drought in different seasons and in different years in Shandong province were analyzed.The annual drought in Shandong province was mainly light drought,and the drought in spring,autumn and winter was more serious and frequent drought occurred.Meanwhile,M-K trend test was used to analyze the change of drought trend from 2010 to 2019.The results showed that there was no significant change in the drought trend,but the inter-annual drought trend worsened.The drought in spring,summer and winter showed a trend of worsening,and the drought in autumn showed a trend of easing.