基于气象条件的我国北方马铃薯晚疫病发生预报模型及其验证

The prediction model and its verification of potato late blight in northern China based on meteorological conditions

病害发生与否是一个二分类问题,故本文采用分类模型,将近年北方马铃薯种植区的晚疫病发病数据与气象数据进行结合,探索北方区域马铃薯晚疫病发病模型,以预测马铃薯晚疫病发病时间来指导区域马铃薯晚疫病防治。模型中考虑到气象站到种植区有一定距离,故根据距离将北方马铃薯晚疫病病害数据分为3个样本集。同时结合马铃薯晚疫病发病机理,采用以日为时长单位的气象数据,利用ROC曲线,确定适合马铃薯晚疫病发病的参数阈值与体系,并进行了历史验证。验证结果表明,该模型在北方种植区预报马铃薯晚疫病的整体正确率达到78%,灵敏度为79%,特异性为78%。

The occurrence or non-occurrence of diseases is a binary problem. Combined the data of late blight of potato in northern China and meteorological data,the classification model was used to explore the pathogenesis model of potato late blight. Three sample sets were created according to the distance between the weather station and the planting region. Considering the pathogenesis mechanism of late blight of potato,the meteorological data（ a day length） and receiver operating characteristic curve（ ROC curve） were used in the classification model to select the appropriate parameter threshold and the framework for late blight of potato. Then,the classification model was verified by historical data. The results demonstrated that the overall accuracy of the forecasting model was 78%,the sensitivity was 79%,and the specificity was 78% in the northern planting region.