학술논문
A Machine Learning Algorithm for Retrieving Cloud Top Height With Passive Microwave Radiometry
Document Type
Periodical
Author
Source
IEEE Geoscience and Remote Sensing Letters IEEE Geosci. Remote Sensing Lett. Geoscience and Remote Sensing Letters, IEEE. 19:1-5 2022
Subject
Language
ISSN
1545-598X
1558-0571
1558-0571
Abstract
This study aims to retrieve cloud top height (CTH)—excluding cirrus—using passive microwave radiometer observations combined with humidity and temperature profiles. A machine-learning-based approach, combining neural network and gradient boosting methods, is used with Cloud Profiling Radar observations as input. The subsequently derived microwave CTH predictions show a mean average error of 2.1 km and a correlation index of 0.8. The algorithm is used to retrieve the CTH during Hurricane Maria and during a mid-latitude autumn storm. This new algorithm will allow to provide estimates of CTH, at world scale, for a 20-year period.