Monitoring And Prediction Of Wheat Stripe Rust

Stripe rust caused by Pucciania striiformis f.sp.tritici seriously affects the safety of wheat production and often results in huge yield loss.At present,chemical pesticide spraying is the main means to control wheat stripe rust,but excessive application will lead to the increase of control cost,pesticide residue,pathogen resistance and other problems.Monitoring and early predicting technology is the premise of scientific and reasonable guidance for the green control of stripe rust,which is of great significance for reducing the control cost,reducing pesticide residues,and slowing down the resistance of pathogens.At present,the monitoring and early predicting of wheat stripe rust is mainly used in the oversummer,autumn seedling and early spring disease survey,combined with the weather forecast and experts experience.There are some problems,such as time-consuming and laborious,the accuracy of the prediction results is not high.With the development of science and technology,the Internet of things,network communication,sensor and artificial intelligence technology are gradually applied in the monitoring of crop diseases and insect pests.Intelligent and information-based stripe rust monitoring and early predicting technology has become the future development trend.In this study,spore trap,environmental monitoring,Internet of Things,big data and molecular biology were used to monitor and predict wheat stripe rust.The spore capture technology and multidimensional environmental meteorological monitor were used to automate the collection of urediniospores and field meteorological factors in several detection sites in wheat stripe rust epidemic areas.The prediction model based on the density of urediniospores and key meteorological factors was established,and monitoring and early predicting of wheat stripe rust were realized.The main results of this study are as follows:1.The dynamic changes of urediniospore density in different epidemic areas of wheat stripe rust were clarified.Seven Burkard spore traps were installed in the main epidemic region of wheat stripe rust in China,including Tianshui and Pingliang in Gansu,Mianyang in Sichuan,Xiangyang in Hubei,Xianyang(Wugong,Yongshou,and Changwu country)in Shaanxi.Among them,Tianshui and Pingliang belong to the over-summering area,Mianyang and Xiangyang belong to the winter propagation area,and Central of Shaanxi(Wugong,Yongshou,and Changwu country)belong to the over-wintering area.Through continuous monitoring for many years,the annual dynamic of urediniospores in the main epidemic areas were clarified.Among them,the peak of urediniospore density in Tianshui and Pingliang regions of Gansu generally appeared from May to June,the peak of urediniospore density in Xiangyang,Hubei generally appeared in April-May and October-November,The peak of urediniospore density in Mianyang,Sichuan generally occurs from March to April.The peak of urediniospore density Central of Shaanxi Province generally appears in May-June and November-December.Sporadic Pucciania striiformis f.sp.tritici urediniospore can be captured during ver-summering stage in all regions.2.The main influencing factors of wheat stripe rust epidemic were analyzed.By analyzing the factors affecting the occurrence degree of wheat stripe rust,it was found that the factors affecting the disease index of wheat stripe rust in China were Spo-4(cumulative spore density at flowering stage),SH-17(sunshine hours of autumn seedling stage),R-14(cumulative rainfall in wintering stage)and R-9(cumulative rainfall in jointing stage).The main factors affecting the disease index of wheat stripe rust in Tianshui region of Gansu Province were Spo-11(cumulative urediniospore density in early March),RH-15(relative humidity in late January)and R-13(cumulative rainfall in middle February).The disease index of wheat stripe rust in Pingliang region of Gansu Province was mainly related to Spo-10(cumulative urediniospore density in middle March)and AT-14(air temperature in early February).The disease index of wheat stripe rust in Xiangyang,Hubei was mainly related to Spo-19(cumulative urediniospore density in mid-November of last year)and R-11(cumulative rainfall in early February).The disease index of wheat stripe rust in Mianyang,Sichuan was mainly related to Spo-9(cumulative urediniospore density in middle February)and SH-14(sunshine hours in late December of last year).The disease index of wheat stripe rust in Guanzhong region was mainly related to Spo-14(cumulative urediniospore density in middle January)and R-11(cumulative rainfall in middle February).3.The prediction model of wheat stripe rust was established and evaluated.By screening the key factors affecting the occurrence degree of wheat stripe rust,a prediction model of wheat stripe rust based on key urediniospore density stage and meteorological factors was established.Among them,the prediction accuracy of random forest and decision tree models for wheat stripe rust in Tianshui was 100% and 87.5%,respectively.The prediction accuracy of multiple linear regression model,random forest model and decision tree model for the occurrence degree of wheat stripe rust in Pingliang of Gansu Province was 100%,100%and 87.5%,respectively.The prediction accuracy of random forest model and decision tree model for wheat stripe rust in Xiangyang was 90% and 77.5%,respectively.The prediction accuracy of multiple linear regression model,random forest model and decision tree model for wheat stripe rust in Mianyang,Sichuan was 100%,100% and 87.5%,respectively.The prediction accuracy of each model in Guanzhong of Shaanxi is above 85%.In addition,special models for wheat stripe rust in Tianshui of Gansu,Pingliang of Gansu,Xiangyang of Hubei,Mianyang of Sichuan and Guanzhong of Shaanxi were established.Among them,the prediction accuracy of BP neural network and decision tree model for the occurrence degree of wheat stripe rust in Tianshui was 75 % and 100 % respectively.The prediction accuracy of BP neural network model,random forest model and decision tree model for the occurrence degree of wheat stripe rust in Pingliang of Gansu Province was 75%,75% and 100%,respectively.The prediction accuracy of multiple linear regression model,BP neural network model,random forest model and decision tree model for the occurrence degree of wheat stripe rust in Xiangyang was 80%.The prediction accuracy of the four prediction models for Mianyang in Sichuan was 75%,75%,100% and 100%,respectively.The prediction accuracy of all model in Guanzhong of Shaanxi Province was 100%.The prediction model of wheat stripe rust based on the key period of urediniospore density and meteorological factors established by the above different methods provides scientific guidance and theoretical support for the accurate and effective prevention and control of wheat stripe rust,and lays a foundation for the high and stable yield of wheat industry in China.