欢迎访问《麦类作物学报》编辑部网站

作者	单位
郁凌华，邢程，荀静，缪新伟，王军，曹文昕，岳伟	（1.安徽省滁州市气象局,安徽滁州 239000； 2.安徽省滁州市农业农村技术推广中心,安徽滁州 239000； 3.安徽省滁州市南谯区农业农村技术推广中心,安徽滁州 239000； 4.安徽省农业科学院作物研究所，安徽合肥 230031 5.安徽省农业气象中心,安徽合肥 230031）

摘要点击次数: 156

全文下载次数: 132

中文摘要:

为探寻小麦赤霉病病穗率预测方法，基于滁州市2005-2020年小麦赤霉病病穗率资料和对应气象资料，运用相关性及灰色关联分析法(GRA)确定小麦赤霉病主要气象影响因子并作为支持向量回归(SVR)模型的输入向量，再利用粒子群算法(PSO)优化SVR模型的惩罚因子C和核函数参数g，建立基于粒子群算法优化的小麦赤霉病预测支持向量回归模型。同时针对本地不同小麦品种，构建PSO-SVR-SOUTH和PSO-SVR-NORTH的PSO-SVR分模型，应用3种模型对滁州地区小麦赤霉病病穗率进行预测。结果表明，拔节期至灌浆期是影响滁州小麦赤霉病的重要时段，各生育时期内降水量、雨日数、湿度、日照等气象因子与赤霉病有高关联；PSO-SVR赤霉病病穗率预测模型的起报时间越接近灌浆期，其预测精度越高，测试样本的预测值与实测值相关系数最高达0.68，均方根误差最小为9.55%；按照不同小麦品种构建的PSO-SVR-SOUTH和PSO-SVR-NORTH模型的预测效果要优于原PSO-SVR模型，其中最迟起报时间的PSO-SVR-SOUTH和PSO-SVR-NORTH模型的平均绝对误差分别较原PSO-SVR模型减少了63.7%和20.8%，均方根误差RMSE较原有模型分别降低了61.6%和40.6%，相关系数分别提高了38.2%和29.4%，拟合优度R²则分别提高了1.4倍和1.1倍。该模型业务服务效果较好，可用于本地小麦赤霉病预测。

英文摘要:

In order to explore the prediction method of diseased spike rate of wheat scab,the main meteorological factors were selected as the input feature vectors of the support vector regression(SVR) model by correlation and grey relation analysis(GRA) using the data of diseased spike rate of wheat scab and corresponding meteorological data in Chuzhou from 2005 to 2020. The particle swarm optimization (PSO) algorithm was used to optimize the penalty factor C and the kernel function parameter g of SVR. The prediction model of wheat scab was established based on PSO-SVR.At the same time, PSO-SVR sub-models such as PSO-SVR-SOUTH and PSO-SVR-NORTH were constructed according to different local wheat varieties. Three models were applied to predict the diseased spike rate of wheat scab in Chuzhou. The results showed that: jointing to grouting period was the important period affecting wheat scab in Chuzhou.The meteorological factors, such as precipitation, rainy days, humidity, sunshine etc. in each development period were highly associated with wheat scab.The prediction model of diseased spike rate of wheat scab based on PSO-SVR initialized at different time was constructed. The higher accuracy of PSO-SVR model appeared when the initial time of PSO-SVR was close to grain-filling stage.The maximum correlation coefficient between the predicted value and the actual value of the test sample is 0.68, and the minimum root mean square error is 9.55%. The prediction effect of PSO-SVR-SOUTH and PSO-SVR-NORTH models according to different wheat varieties were better than that of the original PSO-SVR model.Compared with the original model, the mean absolute error of PSO-SVR-SOUTH and PSO-SVR-NORTH models with the latest initial time were reduced by 63.7% and 20.8%, respectively;the root mean square error were reduced by 61.6% and 40.6%; the correlation coefficients were increased by 38.2% and 29.4%, and the goodness of fit R²were increased by 1.4 times and 1.1 times.The model is good for operational service, which can be used to predict the occurrence trend of wheat scab in Chuzhou.

查看全文查看/发表评论下载PDF阅读器

关闭