A procedure is presented that allows using informa-tion from the MODerate resolution Imaging Spectroradiometer (MODIS) sensor to improve the quality of emissivity maps for the Meteosat Second Generation/Spinning Enhanced Visible and In-frared Imager (SEVIRI) currently in use as input to a generalized split window (SW) algorithm for land-surface temperature (LST) retrievals in the operational chain of the Satellite Application Facility on Land Surface Analysis (LSA SAF). Information from MODIS is incorporated by means of linear regression models expressing emissivity in SEVIRI thermal-infrared channels as a linear combination of emissivities in MODIS bands. The linear models are applied to the MODIS emissivity product MOD11C3, and a comparison is performed with the operational LSA-SAF product. Special attention is devoted to the semiarid and arid regions of North Africa where emissivity is highly variable. When compared with the new emissivity maps, the LSA-SAF product displays more uniform emissivity values over these regions, lead-ing to higher retrievals for all channels (bias around 0.03) except for IR3.9 (bias from −0.05 to −0.08). The root-mean-square error (RMSE) varies from 0.06 to 0.09 (0.02 to 0.03) for IR3.9 (IR10.8 and IR12.0) and is about 0.06 for IR8.7. The impact on LST is assessed by comparing the retrievals from a SW algorithm using as input the following: 1) the SEVIRI emissivity LSA-SAF product and 2) SEVIRI emissivity maps from MOD11C3. The uncertainty in the LSA-SAF emissivity product results into LST values with bias ranging from −0.4 to −1.0 K and RMSE around 1.6 K. The new emissivity maps based on MODIS data may be an alternative to the standard LSA-SAF emissivity product over semiarid and arid areas, which cover 26% of the land surfaces within the SEVIRI full disk. Index Terms—Emissivity, land-surface temperature (LST), linear regression, MSG/SEVIRI, TERRA/MODerate resolution Imaging Spectroradiometer (MODIS).