水稻纹枯病病情指数预测模型研究

Study on disease index prediction model for rice sheath blight

水稻纹枯病是水稻最重要的真菌性病害之一,准确预测该病的发生对纹枯病的防控具有重要的实践意义。本研究以衡阳地区1995—2013年纹帖病病情指数为基础,基于地统计学分析纹枯病时效性特性,提取时效自变量,并整合气象因子构建支持向量回归非线性预测模型。然后基于非线性多轮末尾汰选,从23个初始自变量中(气温、降水、日照等18个气象因子以及5个时效因子)获得了13个纹枯病诱发因子,并以此构建高精度非线性稻纹枯病发生预报模型。对1997—2013年早、晚稻纹枯病病情指数实施预报,其均方误差MSE为5.97,R^2达到0.831 8。预测结果表明,该模型能准确预报纹枯病的发生程度,可为及时制定纹枯病防治策略提供可靠的指导。

Rice sheath blight was one of the most important fungal diseases of rice and trend prediction of this disease had important practical significance to its control.In this paper,a high precision nonlinear prediction model of rice sheath blight was proposed according to dataset of Hengyang in 1995-2013.Firstly,the dataset was analyzed by geo-statistics and variables of order for time series were extracted.Secondly,nonlinear prediction model of support vector regression was constructed based onorder of time series and meteorological factors.Lastly,13 variables out of 23（18 meteorological factors include the temperature,precipitation,and sunshine and 5 order of time series） were selected by worst descriptor elimination multi-round method.Rice sheath blight disease indexes for 1997-2013 were predicted with new prediction model,andmean squared error and R^2 were S.97 and 0.831 8,respectively.The results showed that the model can accurately forecast occurrence of sheath blight and can provide guidance for prevention strategy of sheath blight.