决策级融合技术在环渤海地区短期天气预报中的应用

Application of decision fusion to short-range weather forecasting in Bohai rim region

设计一种基于决策级融合的短期天气预报二级决策模型,并以线性回归法、极值剔除法、支持向量机法为初级决策方法,采用常规气象资料对该模型进行了系统实现和效果检验。结果表明,该预报模型的决策级融合结果取得了较高的平均历史拟合率和试报准确率,较初级决策方法平均提升约2%—3%;决策级融合的决策结果将具有更高的可信度,其有利于提升数理统计预报的应用水平,促进信息化气象保障系统的建设。

A short-range weather forecast decision-making model of di-level based on decision-level fusion is developed in this study. The statistical methods including linear regression, extremes elimination and support vector machine are selected as initial decision method. The model is implemented and verified by using conventional meteorological data. The result shows that the average fitting rate and accuracy rate from this model are improved by 2 to 3 percent, which is better than that from the initial decision method. The decision results of decision-level fusion will be more credible, which should be helpful to enhance the application level of physical-statistical weather forecasting and promote the construction of information meteorological support system.