Assessing the impact of space and time resolution of brightness temperature simulation conditions in correcting SEVIRI SST over the Adriatic Sea

Majority of operational sea surface temperature (SST) infra-red (IR) products have small overall bias but exhibit higher absolute biases in specific regions. Previous studies [for example Tomazic et al., 2011] showed that over Adriatic Sea there is a positive summer bias sometimes exceeding 0.5 K. Methodology to decrease regional biases [Le Borgne et al., 2011], based on using atmospheric profiles, surface SST fields and radiative transfer model to simulate the non-linear split window SST (NLSST) algorithm error, was used to assess the impact of atmospheric profiles with different spatial resolution (ECMWF: 0.125 deg and ALADIN 2 km), different input surface SST fields (OSTIA 6 km and CNR UHR L4 1 km) and different time and space averaging criteria’s in deriving the algorithm correction. SST corrections derived for the NLSST Spinning Enhanced Visible and Infrared Imager (SEVIRI) algorithm for five months (July, October and December 2010 and March and July 2011) are validated with AATSR L2 SST fields and compared to already implemented regional correction procedure at Center de Meteorologie Spatiale (CMS) to assess the optimal combination of space and time averaging criteria’s and input fields.
Results show that the best improvement for all available months and for both day and night is obtained when using spatial averaging over the whole domain in combination with time averaging between the last 15 and last 31 days both for day and night time analysis. Using higher resolution ALADIN atmospheric profiles with OSTIA input SST fields didn’t improve the SST correction compared to using combination of coarser ECMWF atmospheric profiles and OSTIA input SST. Small improvement, based only on analysis for two months in 2011, is obtained when using both higher spatial resolution atmospheric profiles (ALADIN) and higher resolution input SST fields (CNR UHR L4 1 km). The best improvement obtained for spatial averaging over the whole domain (Adriatic Sea) suggests that the domain should be more extended (to Mediterranean Sea) to derive optimal spatial averaging, while conclusion of improvement obtained when using both the higher spatial resolution atmospheric profiles and input SST field need extension of analysis on other months in 2010.