BP网络模型和LOGIT模型在森林害虫测报上的应用初报——以安徽省潜山县马尾松毛虫为例

以安徽省潜山县马尾松毛虫测报为例，在潜山县全县设置102个诱捕点，诱集越冬代成虫，并同时记录环境因子数据，例如树高、树龄、坡向、植被等。以这些环境因子为测报因子，以性外激素诱集到的雄蛾数代表田间实际发生情况，构建了森林重要害虫-马尾松毛虫的分类测报模型-BP人工神经网络和概率测报模型。BP网络模型的拟合率为100%，预留样检验报准率为100%。LOGIT模型的一致对率为74%。两类模型的拟合结果和预测结果基本吻合。BP网络的测报能力优于LOGIT模型。

The application of BP model and logit model to prediction of forest insect pests

The prediction of masson pine moth,Dendrolimus punctatus,in Qianshan County of Anhui Province was studied and 102 trapping locations was chosen to place traps-baited with sex pheromone throughout the county.The number of adult males trapped of the overwintering generation was counted and recorded.Some environmental factors such as the height of the pine tree,slope direction,plant type were recorded at the same time.The independent variables consist of these environmental factors and the number of adult males trapped stands for the occurrence of D.punctatus in the field.Two models are developed for the classification and the probability prediction of forest pests using BP artificial neural network and LOGIT model with factors related to forest pests.The results of prediction to masson pine moth show that the qualified rates of fitting and prediction using BP model acquire 100%.The rate of concordant in LOGIT model is 74%.The prediction result of BP model agrees with that of LOGIT model.BP model is better than LOGIT model in prediction.