基于概率模型的冬小麦白粉病监测研究

Wheat powdery mildew monitoring using probabilistic model

关键词:  小麦  白粉病  多源数据  概率模型  遥感

Key words:  wheat  powdery mildew  multi-source information  probabilistic model  remotesensing.

摘要:

【 目的】利用多源数据对白粉病易发区进行病害发生监测研究能够提供大面积、快 速、客观的病害发生信息，为农业植保部门开展科学防控提供有效的指导。研究以结合遥 感与气象数据监测冬小麦白粉病并获取其精细空间信息为目的。【方法】利用中国高分辨率 对地观测系统高分一号卫星（GF-1/WFV）遥感影像提取了研究区小麦种植区及归一化植被 指数（normalized difference vegetation index，NDVI）、增强型植被指数（enhanced vegetation index，EVI），通过气候灾害组站点红外雨量数据（climate hazards group infraRed precipitation with station data，CHIRPS） 及美国中分辨率成像光谱仪（moderate-resolution imaging spectroradiometer，MODIS）获取了研究区气象相关数据。采用基于概率模型的方法筛选并定 量化表达了特征因子与小麦患病情况间的关系，生成了研究区小麦患病概率分布数据、研究 区小麦白粉病监测结果、监测结果错分概率分布数据。【结果】基于概率模型的白粉病监测 方法总体精度为81.25%，与目前较为流行的分类与回归树（classification and regression tree， CART）和随机森林（random forests，RFs）2 种分类监测方法相比具有较高的监测精度；小 麦患病概率分布数据能够详细地展现研究区小麦患病概率的空间分布及高患病概率麦区向低 患病概率麦区的过渡情况；监测结果的错分概率分布数据与实际错分情况具有较好的一致 性。【结论】基于概率模型的白粉病监测方法能够应用于区域尺度冬小麦白粉病发生发展状 况监测研究。

Abstract:

［Purpose］Wheat powdery mildew is one of the serious crop diseases that affect the food safety in China. Integrating multi-source information（Earth Observation-EO，meteorological，etc. ） to support decision making in the sustainable management of wheat powdery mildew in agriculture is demanded. The purpose of this study is to acquire detailed spatial information of wheat powdery mildew using remote sensing data and meteorological data. ［Method］The GF-1/WFV remote sensing image of March 14th，May 18th and June 17th in 2014 were used to extract two vegetation indices including normalized difference vegetation index （NDVI） and enhanced vegetation index（EVI）. Meteorological data including Climate Hazards Group InfraRed Precipitation with Station data from April 1st to May 11th and level-3 MODIS global Land Surface Temperature and Emissivity 8-day data from April 7th to May 9th were also used to get the parameters delineating the environment condition such as average land surface temperature（LSTper-8-days） and average precipitation（Precipitationper-5-days）. In the field work， 42 field sites were collected in study area and the degree of wheat powdery mildew in these sites was recorded. An independent t-test analysis was used to test the difference between disease and healthy sites based on calibration data. Those vegetation indices and environmental factors showed a statistical significant（p<0.01） and were identified as optimal explanatory variables for developing the powdery mildew monitoring methodology. The powdery mildew monitoring methodology based on probabilistic model was established to monitor powdery mildew occurrence of wheat in Guanzhong Plain，Shaanxi Province. The result obtained from the probabilistic model consists of the distribution of possibility of infection，the monitored map of powdery mildew and the distribution of possibility of misclassification. The distribution of possibility of disease infection could represent the wheat landscapes as continua. The monitored map of powdery mildew was compared with results of monitoring model developed using classification and regression tree （CART） and random forests（RFs）. The accuracies of models respectively based on validation samples were obtained to evaluate the difference of performance of the models.［Result］The results showed that the overall accuracy of the probabilistic model was 81. 25%，which is higher than RFs model（77. 08%）. Although the present accuracy of probabilistic model was relatively lower than the CART model’s，it was more stable when using different training points set，which indicates probabilistic model has a better performance in recognizing powdery mildew. Besides， the distribution of possibility of misclassification was consistent with actual classification of validation points，which means it could provide references for evaluating the monitored map. ［Conclusion］All these results reveal that probabilistic model could be used to monitor the occurrence of wheat powdery mildew.