Study On Drought Vulnerability Of Summer Soybean In Huaibei Plain Based On Drought Test And BP Neural Network

Droughts have been one of the major natural disasters facing China and the world since ancient times.With the global climate change and intensification of human activities,the damage caused by drought has also increased.There are many factors influencing the drought,and the vulnerability of the drought itself is the key to the risk.Based on this,the thesis combines the basic theory of drought risk assessment with practical application,establishes the experimental index,constructs the BP neural network model,combines the qualitative chart analysis with the quantitative calculation,and discusses the drought vulnerability research method based on BP neural network.In the study area,the Huaibei Plain was selected.The research subjects selected summer soybean in the Huaibei Plain,established a drought vulnerability research method based on BP neural network,and conducted applied research.The following conclusions were obtained:(1)According to the drought-testing data of the summer soybean pots in the Fuxin Maqiao Agricultural Comprehensive Testing Station of the Huaihe River Water Conservancy Commission,Anhui Province,the experimental indicators were calculated and calculated,and the relationship between the indicators and the yield was established.Diagrams and tables for correlation analysis.The results showed that the correlation between drought duration,dry matter quality,drought intensity index and yield was good in each growth period.Among them,the relationship between drought intensity and yield is good,indicating that the drought intensity has a great influence on the yield of soybeans affected by water stress.(2)A model was established using the method of BP neural network to measure the yield of summer soybean in different growth stages under different drought conditions.Through the comparison between different models,the impact of each index on the yield at different stages of production was analyzed,and the impact factors were established.The results showed that the correlation between indicators and yield was good.The impact factors of yield at each growth stage were drought duration,drought intensity and dry matter quality.The difference was that the impact factors of the first three growth periods included the leaf area index,and the last one.The impact factors of the growth period include 1000-grain weight and the actual rate.The correlation between indicators and yield is good and the fitting error and verification error are small.(3)The fitted value(predicted value)of the yield was obtained by the BP neural network model,and the reduced yield of light drought and heavy drought was established together with the measured value.The results showed that the relative error between the predicted value and the measured value of the reduction rate of light drought and severe drought was not significant under the same treatment conditions in different growth stages or under the same treatment conditions at different growth stages.The forecast of reduced drought yields is more accurate.(4)From the perspective of the mechanisms of disaster science and natural disaster risk formation,the impact of drought duration and drought intensity on the yield reduction(loss rate)of summer soybean was studied to study vulnerability,make vulnerability curves and conduct correlation analysis.At the same time,the effects of drought duration and drought intensity on nutrient substances were studied,and correlation diagrams were made on the graphs to examine vulnerability from another perspective.The results showed that the correlations of drought intensity,drought duration,dry matter quality,and yield reduction were better in each growth period,that is,the vulnerability under water stress was significant.The relationship between drought intensity and yield reduction in the four growth periods is very good,indicating that the drought intensity of soybean vulnerability to water stress is very important.(5)A model was established using the method of BP neural network to measure the vulnerability and nutrients under different drought intensity and drought duration conditions.Different models were compared to investigate vulnerability,and the effects of drought intensity and drought duration on nutrients were also studied.Finally,it is concluded that the vulnerability is relatively strong and the impact on production reduction is greater.At the same time,it is also found that the relationship between vulnerability and nutrients in the whole growth period is good and the error is also small.It also shows that the vulnerability is significant under water stress.