基于星载微波亮温数据的西北太平洋热带气旋生成预测方法

Tropical Cyclone Genesis Prediction based on Spaceborne Microwave Brightness Temperature Observations in the Northwest Pacific Ocean

从热带云团中识别有可能生成热带气旋的云团,对台风灾害的早期预警至关重要。基于星载微波辐射计亮温数据和决策树方法,开展了西北太平洋热带气旋生成过程识别研究,构建了热带气旋生成预测模型。样本数据来自2005~2009年间每年5~10月台风季获取的SSM/I微波辐射计观测数据,共包含326个热带气旋样本和2112个未生成气旋的热带扰动样本。基于该数据集计算得到了35个亮温特征参数,包括25个统计量参数和10个描述像元空间分布的景观格局参数。然后,将数据集分为训练集(2007~2009年)和测试集(2005~2006年),在训练集上模型的总体精度为85.27%,热带气旋正确预报率为90.91%,错误预报率为20%;上述3项指标在测试集上的结果中分别为79.47%、80%和20.79%;表明利用微波亮温数据构建的热带气旋生成模型能有效预测热带气旋的发展。

Forecasting and identifying the clouds that may develop into Tropical Cyclones(TCs)from tropical clouds are essential for early warning disasters.In this study,spaceborne microwave brightness temperature observations are used to build a new model to predict TC genesis in the Northwest Pacific.A sample dataset consists of 326 developing TC samples,and 2112 non-developing tropical disturbance samples is established.They are collected by DMSP satellite during the typhoon season(May to October)from 2005 to 2009.Based on this dataset,35 brightness temperature characteristic parameters were calculated including 25 statistical parameters and 10 landscape pattern parameters.Landscape pattern parameters are included to describe the spatial distribution of brightness temperature near TC center.The dataset is then divided into a training set(2007~2009)and a test set(2005~2006).For the training set,the model’s overall accuracy is 85.27%,the hit rate and false alarm rate of tropical cyclones are 90.91%and 20%,respectively.For the test set,the above three parameters are 79.47%,80%,and 20.79%,respectively.Therefore,the tropical cyclone generation model based on microwave brightness temperature data can effectively predict tropical cyclones’development.