Land surface temperature retrieval from METOP-AVHRR3 data using a Split-Window algorithm

In this paper we provide a Split-Window (SW) algorithm which combines bands 4 (10.8 mu m) and 5 (12 aem) to retrieve Land Surface Temperature (LST) from Advanced Very High Resolution Radiometer/3 (AVHRR/3) onboard the MetOp-A platform. The coefficients involved in the SW algorithm have been calculated from a complete simulated dataset, which was obtained using MODTRAN4 radiative transfer code, atmospheric profiles extracted from the Thermodynamic Initial Guess Retrieval (TIGR) database and emissivity spectra of different natural surfaces extracted from the ASTER spectral library. A sensitivity analysis was carried out according to classical error theory by assuming an error on the surface emissivity of 1%, a Noise Delta Equivalent Temperature (N Delta ET) of 0.1 K and an error of 0.5 g/cm(2) for the total atmospheric water vapour content. In this case, the total error on LST retrieval is 1.7 K. LST retrievals from AVHRR/3 data using the proposed algorithm were compared to standard LST products such as the ones generated by the MODIS and LSA SAF (SEVIRI) projects over different forest areas. Results show a mean difference (AVHRR/3 minus product) of 0.7K for SEVIRI and 0.2 K for MODIS, with Root Mean Square Errors bellow 1.5 K in both cases.