A Gradient Boosting Decision Tree Based Correction Model for AIRS Infrared Water Vapor Product
Jiafei Xu, Zhizhao Liu
Abstract
Abstract High‐quality precipitable water vapor (PWV) measurements have an essential role in climate change and weather prediction studies. The Atmospheric Infrared Sounder (AIRS) instrument provides an opportunity to measure PWV at infrared (IR) bands twice daily with nearly global coverage. However, AIRS IR PWV products are easily affected by the presence of clouds. We propose a Gradient Boosting Decision Tree (GBDT) based correction model (GBCorM) to enhance the accuracy of PWV products from AIRS IR observations in both clear‐sky and cloudy‐sky conditions. The GBCorM considers many dependence factors that are in association with the AIRS IR PWV's performance. The results show that the GBCorM greatly improves the all‐weather quality of AIRS IR PWV products, especially in dry atmospheric conditions. The GBCorM‐estimated PWV result in the presence of clouds shows an accuracy comparable with that of official AIRS IR PWV products in clear‐sky conditions, demonstrating the capability of the GBCorM model.