Continental surface infrared emissivity strongly depends on the wavelength and on the type of the surface. Emissivity values as low as 0.7 may be observed around 8--10 ¿m or, at shorter wavelengths, around 4 ¿m, particularly over desert regions. Satellite observations are very sensitive to emissivity variations: At 11 ¿m an uncertainty as small as 2% may lead to a variation of up to 0.5K in brightness temperature. An accurate and coherent, i.e., simultaneous, determination of surface temperature and emissivity is essential to greatly improve the estimation of the longwave surface energy budget and, consequently, to improve the performance of surface-atmosphere interaction models. On the basis of a space differential approach and a nonlinear regression inference method, 4 years of NOAA 10 observations (July 1987 to June 1991) over northern Africa (5¿N--30¿N and 20¿W--60¿E) have been interpreted in terms of surface emissivity at three wavelengths, namely, 11.1 ¿m, 8.3 ¿m, and 4 ¿m, corresponding to atmospheric windows, and surface temperature, actually, skin temperature. Emissivity maps at a resolution of 1¿ ¿ 1¿ and 1 month reveal strong signatures of sand at 8.3 and 4.0 ¿m and of carbonates at 11.1 ¿m. Time series of zonal means may bring into evidence important seasonal variations as, for example, over regions of savannas: from 5% at 11.1 ¿m (peak to peak) to 15% at 4 ¿m. They are shown to be in phase with the precipitation and 1 month ahead of the Normalized Difference Vegetation Index time series. The mean uncertainty may be theoretically estimated on the order of less than 2% for the emissivity at 11.1 ¿m and of 1.6K for the surface temperature.