江淮梅雨降水云顶光谱信息与降水强度关系研究

Relationship Between Cloud-Top Spectral Information and Precipitation Intensity of Meiyu Precipitation over the Yangtze-Huaihe Valley

云顶的光学和微观物理特征是基于卫星光谱定量反演降水强度的关键信息。基于1998—2007年江淮梅雨期内热带测雨卫星的光谱和雷达融合观测资料,建立了基于降水云顶光谱信息反演降水强度的随机森林算法模型,进一步探究了梅雨降水云顶微物理特征与降水强度变化的关系。研究表明,在随机森林模型反演降水强度的测试集中,预测降水强度与观测降水强度的相关系数R为0.67,均方根误差为4.06 mm/h,测试集具有较高的降水预测精度。随机森林模型中的云水含量(CWP,即云水质量浓度)在所有输入的云微物理变量重要性排序中位于前列。进一步分析表明:当CWP小于1.0 kg·m^(-3)时,江淮梅雨期降水主要以小雨等级为主,而当CWP大于1.5 kg·m^(-3)时,大雨和暴雨等级的降水概率明显增大;降水的云粒子有效半径(CER)大小主要位于10μm以上,且降水概率总体上随着CER的增大而单调递增;各等级降水概率随着云光学厚度(COT)的增大而不断增大,当COT大于120时,各级降水的概率显著增强,尤其是强降水。

The optical and microphysical features of the cloud top are key information for quantitative precipitation retrieval based on satellite spectra.With the spectral and radar observation data of the Tropical Rainfall Measuring Mission satellite collected during the Meiyu periods of the Yangtze-Huaihe Valley from 1998 to 2007,this paper built a random forest algorithm model that used the cloud-top spectral information of precipitating clouds to retrieve precipitation intensity.It also investigated the relationship between cloud-top microphysical characteristics and the intensity variation of Meiyu precipitation.The results show that in the precipitation retrieval testing set of the random forest model,the correlation coefficient R between predicted precipitation intensity and observed precipitation intensity is 0.67,and the root-mean-square error is 4.06 mm/h.This means the random forest model has high precipitation prediction accuracy.In the model,the cloud water path(CWP)ranks high in the importance hierarchy of all input variables.Further analysis shows that when the CWP is leas than 1.0 kg·m^(-3),precipitation at light rain level is dominant during the Meiyu period of the Yangtze-Huaihe Valley,while when the CWP is more than 1.5 kg·m^(-3),the probability of precipitation at heavy rain and torrential rain levels increases significantly.On the whole,the probability of precipitation at all levels increases monotonically with the increase of cloud effective radius(CER),which is mainly over 10μm.It also increases with the increase of cloud optical thickness(COT).When COT is more than 120,it increases significantly,especially the probability of heavy precipitation.