山东省冬小麦产量动态集成预报方法

在新型统计检验聚类分析 (CAST) 方法对山东省冬小麦种植区进行合理分区的基础上，利用基于作物产量历史丰歉气象影响指数、关键气象因子影响指数、气候适宜度指数、WOFOST (world food study) 作物生长模型分别建立各区域冬小麦产量动态预报方法，利用这4种方法分别对2004—2011年山东省冬小麦产量进行动态预报，在分析历史预报结果平均准确率的基础上，剔除预报准确率低于90.0%的预报方法，确定每种方法的权重系数，采用加权方法建立山东省冬小麦产量动态集成预报方法。结果表明:4种单一产量预报方法在各区域各时段的预报准确率很不稳定，波动范围较大。而集成预报方法对山东省各区域冬小麦产量动态预报准确率相对于4种单一预报方法均有所提高，预报准确率普遍在95.0%以上，且其预报结果稳定性较好，变化比较平稳, 集成预报方法更适合在业务上应用。

Integrated Technology of Yield Dynamic Prediction of Winter Wheat in Shandong Province

Using winter wheat yield and growth data of 17 prefecture-level city, daily meteorological data from 1980 to 2011, and daily 20 cm depth soil moisture data of 14 representative meteorological stations from 1992 to 2011, methods for dynamic prediction of winter wheat yield are established in 4 regions of Shandong Province, considering historical meteorological influence index for bumper or poor harvest of crop yield, key meteorological factors influence index, the climatic suitability influence index and the WOFOST crop growth model, respectively. A newly developed statistical method, cluster analysis of statistical test (CAST), which divides planting areas of winter wheat in Shandong Province into four regions. These four methods are used to predict yield of winter wheat in regions of Shandong Province from 2004-2011. An integrated prediction method is established in which the weight coefficients of each method is determined based on the prediction accuracy, and the prediction method with accuracy lower than 90.0% in each period is removed.The comparison result shows the prediction accuracy in each region and period of four single yield prediction method is very unstable and has a large fluctuation range. Forecast results of the historical meteorological influence index for bumper or poor harvest of crop yield are relatively good in region of C1 and C3. The accuracy of key meteorological factor influence index in region C1 and C2 is relatively consistent, while not quite stable in region C3. The prediction accuracy of the climatic suitability influence index generally is more than 80%. And the prediction accuracy of WOFOST in four regions all reaches 90.0%, except for certain instability and fluctuation. Through integrating these methods, the accuracy in each region and each period is significantly improved, which is generally above 95.0%, and the prediction result is stable. Therefore, the integrated prediction method could overcome shortcomings of the single forecast method, and it is more suitable for application.