Item

Abstract

The satellite-borne Moderate Resolution Imaging Spectroradiometer (MODIS) instrument can monitor atmospheric water vapor distribution based on near-infrared (NIR) measurements. However, the observation performance of MODIS-derived NIR water vapor data records is much poor under cloudy sky conditions. To date, no study has been reported to improve MODIS water vapor retrieval accuracy under cloudy conditions. Previous research improved MODIS water vapor observations only under clear sky conditions. In our work, we develop four water vapor retrieval algorithms for the first time to enhance the water vapor retrieval performance of MODIS under all weather conditions (including cloudy conditions), based on machine learning models and considering multiple dependence parameters that affect the water vapor retrieval performance. The result indicates that the quality of water vapor data from our algorithms is significantly higher than the official water vapor data from the MODIS, in terms of improved R², reduced root-mean-square error (RMSE), and reduced mean bias. The new MODIS-retrieved NIR water vapor estimates under all-weather conditions even have a better observation performance than the official MODIS NIR water vapor data under confident-clear condition, indicating the effectiveness of the new algorithms. Our research can reduce the RMSE of MODIS NIR all-weather water vapor measurements by over 50%. The enhanced satellite-observed MODIS water vapor data records could play a very important role in weather forecasting, climate monitoring, and many other applications.