摘要
【目的】构建四川省主要稻区二化螟(Chilo suppressalis)灯诱和性诱的年诱虫量预测模型。【方法】基于四川主要稻区2018-2023年灯诱和性诱蛾量,采用Pearson分析二化螟年诱虫量与气象因子的相关性,通过逐步回归和BP神经网络构建相关预测模型。【结果】在灯诱和性诱2种诱捕法下,各稻区二化螟年诱虫量与气象因子(温度、湿度、降雨量和气压)密切相关;在灯诱捕法下,成都平原稻区的8月平均气温与二化螟年诱虫量正相关性显著(R=0.701),且是成都平原稻区逐步回归模型构建因子;在性诱捕法下,川东稻区的6月气压与二化螟年诱虫量负相关性显著(R=−0.840);比较不同诱捕法下各稻区的逐步回归模型和BP神经网络模型的预测值、回归拟合值、平均绝对误差(mean absolute error,MAE)和均方误差(mean square error,MSE)后,灯诱法下四川主要稻区的BP神经网络模型预测精准度较好,对成都平原稻区、川东稻区和川南稻区的回归拟合值分别稳定在78.65%~99.59%、92.38%~99.88%和76.97%~99.96%之间。2023年灯诱虫量用于BP神经网络模型独立检验的结果表明,大部分稻区在该模型下的预测值与实际诱集量基本一致。【结论】灯诱法下的BP神经网络模型比逐步回归有更好的预测和拟合效果。
【Objective】To construct an optimal prediction model under the light and sex induced condi⁃tions of the annual population of stem borer(Chilo suppressalis)in major rice areas of Sichuan.【Method】According to the adult moth numbers under the light-and sex-induced conditions from 2018-2023.Firstly,the correlations between the annual population of C.suppressalis and meteorological factors were analyzed using Pearson correlation analysis.Then,the prediction models were further built through the stepwise regression and back propagation(BP)neural network.【Result】Under the lightand sex-induced conditions,the annual population of C.suppressalis in each rice area was closely corre⁃lated with meteorological factors(i.e.,temperature,humidity,rainfall,and barometric pressure).The average temperature in August in rice area of Chengdu Plain was the most correlated with the annual population of C.suppressalis under the light induced condition(R=0.701).On the contrary,under the sex-induced condition,the barometric pressure in June in the eastern Sichuan rice area showed the larg⁃est negative correlation with the development dynamics(R=−0.840).After comparing the predicted val⁃ues,regression fitted values,mean absolute error(MAE)and mean square error(MSE)of the stepwise regression model and the BP neural network model under different induced methods,it was determined that the BP neural network model under the light-induced condition yielded the most accurate predictions in the primary rice-growing areas of Sichuan,and the regression fit values were stable between 78.65%-99.59%,92.38%-99.88%,and 76.97%-99.96%for Chengdu Plain Rice Area,East Sichuan Rice Area,and South Sichuan Rice Area,respectively.The results of the 2023 light induction for independent testing of the BP neural network model showed that the predicted values were basically the same as the ac⁃tual induction in most of the rice districts.【Conclusion】The BP neural network model under the light in⁃duction has better prediction and fitting ef
作者
徐翔
李祥松
王浩
张林
蒲颇
王学贵
XU Xiang;LI Xiangsong;WANG Hao;ZHANG Lin;PU Po;WANG Xuegui(Sichuan Provincial Plant Protection Station,Sichuan Provincial Department of Agriculture and Rural Affairs,Chengdu 610041,China;State Key Laboratory of Crop Gene Exploration and Utilization in Southwest China,Sichuan Agricultural University,Chengdu 611130,China;College of Agriculture,Sichuan Agricultural University,Chengdu 611130,China)
出处
《四川农业大学学报》
CSCD
北大核心
2024年第5期1112-1122,1151,共12页
Journal of Sichuan Agricultural University
基金
四川省科技厅重点研发项目(2022YFN0044)。
关键词
二化螟
气象因子
灯诱和性诱
逐步回归
BP神经网络
预测模型
Chilo suppressalis
meteorological factors
light and sex induction
stepwise regression
BP neural network
prediction model
