Global skin temperature is very important for the understanding of surface climate and for evaluating climate models. However, when the surface is obscured by clouds, this variable cannot be measured directly by using satellite thermal infrared channels. Methods for calculating skin temperature for a satellite cloudy pixel are important, yet little research on their relative merits has been done. The cloudy-pixel treatment presented here is a hybrid technique of neighboring-pixel and surface air temperature approaches. The neighboring-pixel approach (NP) for calculating skin temperature for a satellite cloudy pixel is described and tested against field experiments and climate model CCM3/BATS simulations. This approach is based on the surface energy balance with the soil heat flux being treated by a conventional force-restore method for bare soil and short vegetated surfaces, where ground heat flux is important. For other surfaces where soil heat flux is less important, for example, the fully vegetated forests in temperature and tropical regions, observed empirical relationships between solar radiative energy and skin temperature (i.e., ΔSn/ΔTs) are used. In addition, a surface air temperature (Ta) adjustment is developed from the Monin-Obukhov similarity theory to infer skin temperature from 2-m air temperature using a knowledge of wind speed, pressure, boundary layer stability, and other surface properties. This adjustment is useful wherever surface air temperatures are available to insure the skin temperature consistency during daytime and nighttime. Error analyses show that this cloudy-pixel treatment has an accuracy of 1¿--2¿ K at monthly mean pixel level resolution. This accuracy varies with season and vegetation type. Despite the uncertainty in this algorithm, this work can be practically used to calculate skin temperature for a cloudy pixel. ¿ 2000 American Geophysical Union