Land surface temperature estimation to improve the assimilation of SEVIRI radiances over land

摘要

The aim of this work is to estimate the land surface temperature from satellite observations in order to improve the assimilation of surface-sensitive infrared observations from the Spinning Enhanced Visible and Infrared Imager (SEVIRI). To date, only a few SEVIRI observations are assimilated over land in limited area models; this is due to the still inappropriate description of the land surface in these systems (in both emissivity and surface temperature). In this paper, we demonstrate that the use of land emissivity climatologies at infrared wavelengths together with land surface temperature (LST) retrievals improve the assimilation of SEVIRI radiances against this operational configuration. Emissivity climatologies from the EUMETSAT Land Surface Analysis—Satellite Application Facilities (Land-SAF) were used, and LSTs were retrieved using one SEVIRI window channel and the radiative transfer equation. Retrieved LSTs were evaluated against independent observations/products. Some differences, due to instrumental specifications, were found when comparing SEVIRI and MODIS LSTs, but a good agreement was found between retrieved LSTs and the Land-SAF surface temperature product. The comparison of SEVIRI LSTs and LST analyses from ALADIN/France have pointed out warm (cold) biases during daytime (nighttime), which may be explained by an overall underestimation of the diurnal cycle by the model. The emissivity atlas combined with different LST retrievals were used to simulate SEVIRI radiances using the RTTOV radiative transfer model. A comparison was made between SEVIRI radiance simulations and observations. A significant improvement of the forward model statistics was noticed as well as an increase in the amount of data that could be potentially assimilated in ALADIN/France, compared to the operational setup. These developments were then tested in a context of data assimilation, thus enabling the use of more SEVIRI data over land. Two assimilation experiments were run over a 3 month period during summer 2009, one of which is representative of the operational model while the other differs by the assimilation of more SEVIRI data over land through a better representation of the emissivity and surface temperature. We show that the forecast impact is generally neutral to positive. In particular, SEVIRI data point to positive impact over southern Europe. SEVIRI data are also shown to improve the quality of analyses, particularly those of total column water vapor, and this is substantiated through comparisons with independent GPS measurements.