摘要
随着气候的异常变化,落叶松早落病呈逐年严重的态势发展。为了有效预防早落病的大面积发生,笔者利用气象资料建立2种模型对落叶松早落病的发生面积进行预测预报。结果表明:落叶松早落病喜低温高湿的气候环境,不耐高温干旱。通过逐步回归和神经网络2种模型预报出落叶松早落病的发生面积的结果显示,逐步回归模型预报的平均误差为18.8%,神经网络模型预报的平均误差为8.6%,再利用2012年和2013年进行检验,逐步回归模型预报的误差分别为7.2%和11.4%,神经网络模型预报的误差分别为5.3%和2.4%,较逐步回归模型的预测结果的误差分别低1.9%和9.0%。因此得出结论即神经网络模型预报效果较好,可以为林业部门提供科学依据。
Mycosphaerella larici-leptolepis lto et al increasingly occurs across global forest in recent years withthe climate change. To effectively prevent large outbreak of the disease, two models were constructed by theauthors to monitor and forecast the occurrence of the disease via meteorology data analysis. The results showedthat Mycosphaerella larici-leptolepis lto et al could easily survive in the climate condition of low temperaturewith high humidity, but was venerable to high temperature with low humidity. Results from the two models ofstepwise regression and neural network showed that the average error of stepwise regression model and neuralnetwork model was estimated to 18.8% and 8.6%, respectively. In addition, average errors generated fromstepwise regression model that tested by meteorology data of 2012 and 2013 were estimated to 7.2% and 11.4%while average errors generated from neural network model were estimated to 5.3% and 2.4%, respectively.Errors obtained from neural network model were 1.9% and 9.0% lowered than that from stepwise model,respectively. The results suggested that the neural network model exhibited better performance in monitoringthe disease compared with the stepwise regression model, and could be employed in scientific forecasting of Mycosphaerella larici-leptolepis lto et al in forest of northeast China.
出处
《中国农学通报》
2015年第25期19-23,共5页
Chinese Agricultural Science Bulletin
基金
内蒙古自治区气象局科技创新基金项目"呼伦贝尔市森林病虫害松鞘蛾和早落病气象预报方法研究"(nmqxkjcx201407)
关键词
气候
落叶松早落病
神经网络模型
逐步回归模型
climate
Mycosphaerella larici-leptolepis lto et al
neural network model
stepwise regression model
