文献类型：期刊文献

中文题名：马尾松毛虫发生相关气象因子筛选及预测

英文题名：Selection and Prediction of Meteorological Factors Correlated with Dendrolimus punctatus Outbreak

作者：费海泽[1] 王鸿斌[1] 孔祥波[1] 张真[1] 张苏芳[1] 宋雄刚[1]

第一作者：费海泽

机构：[1]国家林业局森林保护学重点实验室(中国林业科学研究院森林生态环境与保护研究所)

年份：2014

卷号：42

期号：1

起止页码：136-140

中文期刊名：东北林业大学学报

外文期刊名：Journal of Northeast Forestry University

收录：CSTPCD;;北大核心:【北大核心2011】；CSCD:【CSCD2013_2014】；

基金：国家科技支撑计划(2012BAC19B02;2013BAC09B02)

语种：中文

中文关键词：马尾松毛虫;气候因子;BP神经网络;预测预报

外文关键词：Dendrolimus punctatus;Meteorological factors;BP ANNs;Prediction

分类号：S763.42

摘要：选取71个与广西马尾松毛虫的发生期相关的气象因子,结合广西23个站点10年间的发生情况,用双重逐步回归法进行筛选,确定了不同发生等级最相关的气象因子。利用筛选出的相关气象因子建立人工神经网络,对马尾松毛虫的发生情况进行拟合及验证。BP人工神经网络对预留发生点的预测结果平均准确率达80%。同时,根据马尾松毛虫在广西和福建发生期的不同,调整得出其在福建发生的相关气象因子,利用训练好的网络进行预测,结果显示仅中度发生的预测准确率达68%,其他发生水平预测结果不理想。
From the life history records of Dendrolimus punctatus in Guangxi Province, 71 related meteorological factors with dif-ferent levels were selected to analyze the correlations with D.punctatus population .Dual selective stepwise regression anal-ysis were used to find out the most related meteorological factors contributing to ten-year records of D.punctatus in 23 test-ed points in Guangxi on three damage levels.The selected environmental variables based on three different damage levels were used to implement the artificial neural networks （ANNs） models.Back-propagation （BP） algorithm was applied to predict the reserved points with the average accuracy rate of 80%.But the accuracy were only 68%with the medium level prediction in Fujian Province while adjusted the meteorological factors by regional life cycle differences, and the other lev-els could not show the consistent results.