Evaluation of the GOME Water Vapor Climatology 1995–2002

Abstract

An accurate knowledge of the 3‐D water vapor (WV) field is still limited, because of the limited capabilities of sensors in the past to cover the whole Earth’s surface and the lower part of the troposphere, as well as to measure over reasonably long time series. We show here water vapor total column retrieved from seven years of Global Ozone Monitoring Experiment (GOME) measurements collected from August 1995 until August 2002. Our aim is two‐fold: (1) to evaluate the accuracy and the limitations of the GOME water vapor total column and (2) to demonstrate its potential for climate studies. The column retrieval makes use of two innovative techniques operating in tandem, namely the University of Graz empirical air mass factor ratioing technique (IGAM) and the Spectral Structure Parameterization (SSP) retrieval method. The GOME instrument and its follow‐up instruments (SCIAMACHY and GOME‐2), using these algorithms, have the capability to cover nearly the whole globe in cloud‐free situations and collect robust WV total column information over more than 3 decades. In this work we evaluate the results for the first 7 years against independent in situ measurements from the operational WMO radiosonde network, against high spatial resolution water vapor columns from MERIS (the Medium Resolution Imaging Spectrometer on EnviSat) and also compare with ERA40 model results. The GOME water vapor total column exhibits a bias of less than 2.5% with an uncertainty of around 5 mm for collocated measurements against radiosonde and MERIS measurements. Spatial patterns and trends in the global distribution of WV total column fields from GOME against re‐analysis model results are well correlated with temperature in the tropics, and exhibit a lesser degree of correlation in the extra tropics. Cloud‐free total columns from GOME can be systematically lower by up to 5 mm in the sub‐tropics with respect to the all‐sky case. In contrast, the impact of the diurnal cycle on the monthly mean values is found to be very small.