High-accuracy FY-3A/MERSI precipitable water vapour (PWV) data over the continental United States

Abstract

ABSTRACT Water vapour is an important atmospheric gas with a direct impact on climate and weather. Given the significant variations of water vapour distribution in the tridimensional and temporal space, it is of great significance to use remote sensing data to monitor it on a large scale. One of the main missions of the Medium Resolution Spectral Imager (MERSI) onboard FengYun (FY)-3A satellite is to monitor the distribution of precipitable water vapour (PWV). However, the official MERSI PWV data is not widely used due to the limitation of its accuracy. The primary objective of this study is to develop FY-3A/MERSI PWV data with higher accuracy. The study area of this paper is the continental United States (125°W–65°W, 25°–50°N), and the selected period is the year 2010. Three PWV retrieval models for three near-infrared (NIR) channels (0.905, 0.940 and 0.980 μm) of MERSI are constructed based on the matching results between MERSI data and SuomiNET PWV data. Compared with the existing semi-empirical PWV retrieval algorithms, the algorithm developed in this work has two main changes: (1) MERSI PWV retrieval models are constructed using data after removing outliers rather than all data. (2) The algorithm can identify abnormal retrieval results. The validation results based on PWV data derived from Aerosol Robotic NETwork (AERONET) indicate that these two changes both contribute positively to enhancing the accuracy of MERSI PWV data. The root mean square errors (RMSE) of MERSI PWV data developed using the three channels at 0.905, 0.940 and 0.980 μm are 0.28 cm, 0.22 cm and 0.24 cm, respectively. The relative errors (RE) of these PWV data are 0.14, 0.10 and 0.12, respectively. In comparison to the official MERSI PWV data, the RMSE of MERSI PWV data retrieved in this study is reduced by at least 44%, and RE is reduced by at least 42%.