Using Bayesian Network For The Prediction Of Winter Wheat Powdery Mildew In Hebei Province

Powdery mildew is one of the main diseases and insect pests of wheat in Hebei province.In recent years,with the change of meteorological conditions,the degree of damage of Wheat Powdery Mildew in Hebei has been increasing,which seriously affects the yield and quality of wheat in Hebei province.Therefore,the establishment of effective monitoring and forecasting mechanism is of great significance to the formulation and scientific implementation of prevention and control plans.Traditional prediction methods often neglect the correlation between the pathogenic factors of powdery mildew and wheat,and are not flexible enough to analyze outbreaks of diseases.Therefore,this paper uses Bayesian network to construct the prediction model of wheat powdery mildew,which aims to improve the real-time performance of wheat powdery mildew prediction and enhance the strain ability of the outbreak of powdery mildew.Based on the 1990~2012 year atmospheric circulation index,the North Pacific sea surface temperature index,the real time meteorological factors,the occurrence area and the occurrence degree of Wheat Powdery Mildew in Hebei Province,the key factors were selected by the correlation analysis and path analysis to construct the relationship network of the key climatic and meteorological factors affecting the occurrence of wheat powdery mildew.A Bayesian network for the prediction of Wheat Powdery Mildew in Hebei is established.The occurrence degree of Wheat Powdery Mildew in 5 sites in Hebei province and 2013-2016 years was forecasters.The main conclusions are as follows:(1)The real-time meteorological factors that have significant impact on wheat powdery mildew are screened by correlation analysis,including the humidity in mid March,precipitation in late March,humidity in early April,humidity in mid April and precipitation in late May in May.The correlation coefficients between above meteorological factors and the disaster area of powdery mildew are 0.531**,0.553**,0.600**,0.499* and 0.485* respectively.The key climate factors are selected by path analysis,in which the North Pacific sea surface temperature index includes T272-7,T61-12,T102-3,T102-4,T43-7,T300-7,T296-12,T298-7,and the atmospheric circulation index consists of H56-10,H2-7,H35-7,H17-6,H17-6,etc.(2)The network relationship between key factors and wheat powdery mildew is established by means of path analysis,and the network structure is introduced into Bias network,and the whole factor prediction model,early spring preforecast model and pre winter forecast model of wheat powdery mildew prediction in Hebei province are established by learning the parameters of the wheat powdery mildew in Hebei province.In the process of establishing the climate prediction model of powdery mildew,the climate weather factors are stratified according to the time sequence,the first layer is the key climate factor,the second layer is the key meteorological factor,which connects the target nodes.In order to reduce the number of the parent nodes,improve the prediction efficiency and set up the "precipitation".And the "humidity" of the two virtual nodes;the third level is the target node "disaster grade".(3)Based on the whole factor forecast model of wheat powdery mildew weather forecast in Hebei Province,the grade of Wheat Powdery Mildew in Hebei province was checked and predicted,and the correct rate was 100%.Only in 1994 and 2003,the probability of occurrence of disease prediction grade 2 and 3 is very small.The probability of occurrence of disease in other years is greater than 99%.In the forecast test of early spring forecast model of wheat powdery mildew weather forecast in Hebei Province,the forecast grade of 19 years in 27 years is completely correct and the result of 4 year forecast is basically correct.Only 1 years' forecast grade is not consistent with the actual grade,the other year forecast grade is "uncertain",the correct rate is 86.2%,and the error rate is 3.7%.In the forecast test of winter forecast model of Wheat Powdery Mildew in Hebei Province,the forecast results of 15 years are completely correct,1 year forecast results are basically correct and the other forecast grades are "uncertain",and the correct rate is 59.3%.(4)Based on the whole factor forecast model of wheat powdery mildew weather forecast in Hebei Province,the grade of Wheat Powdery Mildew in 5 counties of Dingzhou,Ningjin,Feixiang,Zhengding and Fucheng was forecasted,and the results of each year were compared with the fluctuating yield.The year of injury is not necessarily a high yield year,but the overall trend is negative correlation.