Water-vapour retrieval from Meteosat 8/SEVIRI observations

This paper aims to propose operational algorithms to retrieve the total atmospheric water vapour content (W) using the Spinning Enhanced Visible and Infrared Imager (SEVIRI) on-board Meteosat 8. MODTRAN3.5 was used to obtain simulated data in the thermal infrared channels IR10.8 and IR12.0, in order to determine the numerical values of the coefficients of the algorithms. The algorithm proposed for land pixels takes into account the SEVIRI observation geometry and the radiometric temperatures obtained in the split-window channels at two different times during a day and requires a minimum difference of 10K in terms of temperature between the two situations. Comprehensive error analyses gave rms errors lower than 0.5gcm-2 when observations were taken between the nadir and 50. The algorithm is validated with in situ values, i.e. radiosondes and W measurements with a CIMEL CE318 sun photometer, both obtained from a field campaign, with rms validation errors of 0.2 and 0.7gcm-2, respectively. Additionally, six stations all over the SEVIRI field of view were selected to validate the algorithm from radiosondes data, providing an rms error of 0.4gcm-2. Concerning sea pixels, the linear atmosphere-surface temperature relation is adapted to SEVIRI and takes into account the sea-surface temperature, the atmospheric effective temperature, and the radiometric temperature in the IR10.8 channel. The total error obtained from this methodology has a value between 0.8 and 1.1gcm-2, and the validation is carried out using radiosonde data from four stations near the sea, providing rms errors lower than 0.6gcm-2.