首页|青藏高原夏季FY-4A卫星对流初生产品的分类识别

青藏高原夏季FY-4A卫星对流初生产品的分类识别

扫码查看
为了解并提升风云四号卫星A星(FY-4A)对流初生(Convective Initiation,CI)产品对青藏高原夏季降水的指示意义,基于FY-4A CI产品及全球降水测量计划(Global Precipitation Measurement,GPM)降水数据,根据青藏高原地区2020-2022年6-8月FY-4A CI产品识别出的CI样本与1h后实际观测降水的对应关系,将CI样本划分为无降水CI、弱降水CI和强降水CI三类,并结合大气对流参数与地理位置等信息,利用决策树和随机森林两种机器学习算法建立CI类别识别模型并检验,结果表明:青藏高原地区对流初生后1h内的降水情况存在明显区域差异,其西北部无降水比例高而东南部降水的比例高;利用抬升指数、云水总量、垂直风切变、中低层湿度、云底高度、零度层高度等大气对流参数信息,能较好区分青藏高原CI出现后是否有降水及降水的强弱;随机森林识别模型结果对于CI类别的识别效果优于决策树识别模型结果,利用随机森林识别模型可以更有效地对青藏高原夏季CI按照降水强度的分类进行识别.
Classification and identification of FY-4A convective initiation products in summer on the Qinghai-Tibet Plateau
This study aims to understand and enhance the indicative significance of convective initiation(CI)products from Fengyun-4A(FY-4A)satellite for summer precipitation over the Qinghai-Tibet Plateau.Using the convective initiation(CI)products of FY-4A and pre-cipitation data from the Global Precipitation Measurement Program(GPM),and based on the correspondence between the CI samples identi-fied by the FY-4A CI product and the actual observed precipitation one hour after identifying the CI in the Qinghai-Tibet Plateau region from June to August 2022,three categories of the CI samples,including no precipitation CI,weak precipitation CI,and strong precipitation CI,were divided.Then a CI class recognition model was established and the model performance testing was conducted by combining atmo-spheric convective parameters and geographic location information,two machine learning methods,decision tree and random forest.The re-sults show that there are significant regional differences in the precipitation situation within one hour after the occurrence of CI in the Qing-hai Tibet Plateau region,with a higher proportion of no precipitation in the northwest region and a higher proportion of precipitation in the southeast region.By utilizing atmospheric convective parameters such as lift index,total cloud water,wind shear,middle and low level hu-midity,cloud bottom height,zero degree layer height and so on,it is possible to better distinguish whether there is precipitation and the strength of precipitation after the appearance of CI in the Qinghai Tibet Plateau.The random forests identify model have better performance for CI classification than decision tree,and the use of random forests identify model can more effectively classify summer CI on the Qinghai Tibet Plateau according to precipitation intensity.

FY-4Aconvective initiationTibet Plateaudecision treerandom forest

姜晓飞、章丽娜、张昕、姚爽

展开 >

中国气象局气象干部培训学院,北京 100081

国家气象信息中心,北京 100081

FY-4A 对流初生 青藏高原 决策树 随机森林

中国气象局风云卫星应用先行计划项目国家自然科学基金重点项目第二次青藏高原综合科学考察研究项目中国气象局干部培训学院重点项目

FY-APP-20220102420306112019QZKK01052023CMATCZDIAN06

2024

暴雨灾害
中国气象局武汉暴雨研究所

暴雨灾害

CSTPCD
影响因子:1.533
ISSN:1004-9045
年,卷(期):2024.43(2)
  • 29