Study On Dynamic Prediction Model Of Rice Sheath Blight Driven By Meteorological Factors

Rice sheath blight is a common and frequently-occurring disease in rice in China,and it accounts for the first position in all diseases of rice regardless of frequency of occurrence or area of disease.Accurate and effective forecasting of rice sheath blight can guide the rational application of pesticides,and can effectively reduce the amount of pesticides and improve grain yield and quality while ensuring control effect.At present,the prediction of rice sheath blight is mostly based on statistical analysis.It can only give static prediction results in years,and it is difficult to guide the control of rice sheath blight at a fine level.Therefore,the dynamic prediction model of rice growth period is established.It is of great significance to reasonably guide the prediction of sheath blight.The causes of rice sheath blight include soil bacterial sources,meteorological factors,plant nutrient status,management measures,etc.This study based on meteorological factors to establish a rice sheath blight prediction model.Based on the logistic growth curve and the classic disease epidemiology SEIR model,the modeling method predicts the incidence and disease rate of rice sheath blight,respectively,and realizes the multi-temporal dynamic prediction of rice sheath blight driven by meteorological factors.The grid point meteorological data is used as the input of the model to obtain the surface prediction result of the sheath blight,thus realizing the space-time dynamic warning of the sheath blight during the growth period of rice.This research mainly includes the following aspects:(1)Data acquisition and preprocessing.The study area is a rice growing area in five provinces of Anhui,Jiangsu,Zhejiang,Hunan and Hubei.From the National Meteorological Administration China Meteorological Data Network,the surface daily meteorological data including temperature,humidity and other meteorological stations in the study area from 2010 to 2016 are downloaded,as well as the daily surface temperature data of the study area in 2010-2016.From the plant protection station,the rice sheath blight plant protection data of the study area was recorded in 5 days,including information on rice type,growth period,disease grade,and disease plant rate.Focusing on the plant protection station,the plant protection-meteorological data matching based on the inverse distance weighting obtained the modeling data set.(2)Establish a Logistic-RICEBLA prediction model to predict the incidence of late rice sheath blight in Hunan Province.Based on Logistic growth curve,Logistic fitting was applied to the incidence of rice sheath blight.Based on the influence of meteorological factors on the pathogenesis of sheath blight,the influence module of temperature and humidity was designed and incorporated into the Logistic fitting model to construct the Logistic-RICEBLA predictive model.To achieve dynamic prediction of multi-year and multi-temporal phase of rice sheath blight,the model predicts fault-tolerant accuracy of up to 88%,and the prediction results are in good agreement with the actual.(3)Establish the SEIR-RICEBLA prediction model to predict the rate of late rice and medium rice sheath blight in the study area.Based on the classic disease epidemiology model SEIR,combined with the factors and characteristics of the sheath disease including meteorological factors,growth period,and bacterial source information,the relevant modules are constructed and integrated into the structure of the model to construct the SEIR-RICEBLA prediction model.The temporal dynamic prediction of late rice sheath blight,the prediction accuracy of the model R2 is 0.61,and the RMSE is 8.12.The model is expanded in multiple provinces.The corresponding model accuracy R2 is 0.65 and the RMSE is 11.57.The model is expanded by rice type.The corresponding model accuracy R2 is 0.46 and RMSE is 14.23.The prediction effect of the model is ideal.Finally,based on Arc GIS software,the planar grid data and the SEIR-RICEBLA model proposed in this study were combined to obtain the warning information of rice sheath blight in space-time dynamics.