摘要：
Over the past 30 years,climate change has caused obvious impacts on the catastrophic immigration of the brown plant hopper (BPH),Nilaparvata lugens (St(a)l).To further understand the impact of abnormal climate change on BPH immigration,this study collected data on the BPH using light traps at 35 plant protection stations in China from 1980 to 2016,and collected reanalyzed meteorological data from the National Center of Environmental Prediction (NCEP) from 1979 to 2016 to identify correlations between the occurrence grades of BPH and the meteorological factors affecting them;the key predicting factors were screened.The support vector machine (SVM) model,back propagation (BP) neural network,and regression analysis were used to establish medium long-term prediction models of the annual occurrence grades of BPH at the representative stations in south China,and their advantages and disadvantages were compared.The results were as follows:(1) Most of the abnormal climate occurrence areas in the Indochina Peninsula were distributed in the north.The occurrence frequency of abnormal climate in the north was higher than the frequency of the south,with the characteristics of the frequencies progressively descending from north to south in an annular pattern.(2) If the ground temperature of the Indochina Peninsula was higher than the average temperature and the relative humidity was greater than the average relative humidity during the 37 years,it brought about partially heavy or heavy occurrences of BPH immigration in south China.However,if the ground temperature of the Indochina Peninsula was lower than the average temperature and the relative humidity was less than the average relative humidity during the 37 years,it brought about partially light or light occurrences of BPH immigration in south China.(3) By comparing the correct rates of back substitution and prediction accuracy of all three models,we found that all three models had certain capabilities of predicting the occurrence grades of BPH in south China.The predicting capability of the SVM model was the best,the BP neural network was the second best,and the multiple linear regression model was the worst,indicating that the SVM model was more suitable for predicting the occurrence of BPH in rice production.%近30多年来,气候变化对中国褐飞虱的灾变性迁入带来了明显的影响,为了进一步了解虫源地的异常气候变化对我国褐飞虱迁入量的影响,利用1980-2016年中国各植保站提供的褐飞虱虫情资料及同期美国国家环境预测中心(NCEP)提供的全球气象再分析资料,分析了中国褐飞虱境外主要虫源地中南半岛前期异常气候条件对中国南方稻区褐飞虱发生程度的影响,并对褐飞虱发生等级与影响其迁飞的气象因子进行了相关性分析,筛选出关键预报因子,分别应用支持向量机(SVM)、BP神经网络和多元回归分析3种方法对代表站点褐飞虱年发生等级进行了预测,并比较了3种预测模型的优劣.结果表明:(1)中南半岛气候异常区主要分布在北部,异常气候的发生次数在中南半岛呈现出北高南低的特征,并从北向南呈环状递减.(2)中南半岛前期温度偏高(暖冬、暖春)、相对湿度偏大(湿冬、湿春),易引起褐飞虱在中国南方稻区的偏重以上发生;若中南半岛前期气候偏冷(冷冬、冷春)、偏干(干冬、干春),则常导致褐飞虱在中国南方稻区的偏轻以下发生.(3)通过比较3种模型的历史回代率和预测准确率,发现3种模型对褐飞虱的发生程度均有一定的预测能力,其中SVM模型的预测效果最好,BP神经网络次之,多元线性回归模型最差,表明SVM更加适用于生产实际中的褐飞虱发生程度预测.
