Total column water vapor retrievals from the Geostationary Environment Monitoring Spectrometer (GEMS)

Kurzfassung

To enhance synergy with ultraviolet (UV)–visible satellite observations of water vapor and atmospheric pollutants in low Earth orbit (LEO), we develop a total column water vapor (TCWV) retrieval algorithm using spectral measurements from the Geostationary Environment Monitoring Spectrometer (GEMS). This work represents the first retrieval of TCWV from an UV–visible hyperspectral instrument operating in geostationary Earth orbit (GEO). The algorithm follows a two-step approach: (1) direct spectral fitting of measured radiance spectrum in the 435–467 nm spectral range, and (2) air mass factor (AMF) calculations. Using the first year of GEMS TCWV data within the study period (March 2021 to February 2023), pixel-by-pixel random uncertainties in the TCWV retrievals are quantitatively assessed. Retrieved TCWVs from GEMS under clear-sky conditions (effective cloud fraction (ECF) < 0.2) are validated against ground-based Aerosol Robotic Network (AERONET) measurements, ERA5 reanalysis data, and LEO satellite observations from the TROPOspheric Monitoring Instrument (TROPOMI) and the Special Sensor Microwave Imager/Sounder (SSMIS). The correlation coefficients (R) for all reference datasets range from 0.89 to 0.96, showing good agreement. However, the bias varies with reference dataset and surface type. Compared to ERA5, GEMS TCWV exhibits a mean bias error (MBE) of –3.00 kg m-2 over land and –7.71 kg m-2 over water, corresponding to normalized mean biases (NMBs) of –15.54% and –17.39%, respectively. Comparisons to TROPOMI data also show negative biases of –1.76 kg m-2 (–12.55%) over land and –6.77 kg m-2 (–16.20%) over water. The smallest bias is observed in comparison with AERONET data (–1.43 kg m-2, –5.52%), whereas a substantially larger negative bias is found relative to SSMIS observations (–14.96 kg m-2, –30.82%). Validation performance is generally reduced over water surfaces and during the summer season. A bell-shaped diurnal variation of MBE may arise from errors in AMF calculations under varying observation geometries in GEO. Furthermore, we assess the capability of GEMS data to capture diurnal variations in water vapor relative to ERA5 using monthly diurnal patterns over several regions. The results indicate generally reasonable diurnal behavior, while notable negative differences appear at late-afternoon scan times. Despite its limitations, the GEMS TCWV retrievals have the potential for use as a valuable observational dataset for climatological analyses, improving numerical weather prediction, and data-driven analyses (e.g., machine learning) of atmospheric composition.