The surface skin temperature is a key parameter at the land-atmosphere interface. An accurate description of its diurnal cycle would not only help estimate the energy exchanges at the interface, it would also enable an analysis of the global surface skin diurnal cycle and its variability within the last 20 years. This study is based on the 3-hourly surface skin temperature estimated by the International Satellite Cloud Climatology Project (ISCCP) from the infrared measurements collected by the polar and geostationary meteorological satellites. The diurnal cycle of surface skin temperature is analyzed almost globally (60N–60S snow-free areas), using a Principal Component Analysis. The first three components are identifyed as the amplitude, the phase, and the width (i.e., daytime duration) of the diurnal cycle and represent 97% of the variability. PCA is used to regularize estimates of the diurnal cycle at a higher time resolution. A new temporal interpolation algorithm, designed to work when only a few measurements of surface temperature are available, is developed based on the PCA representation and an iterative optimization algorithm. This method is very flexible: only temperature measurements are used (no ancillary data), no surface model constraints are used, and the time and number of measurements are not fixed. The performance of this interpolation algorithm is tested for various diurnal sampling configurations. In particular, the potential to use the satellite microwave observations to provide a full diurnal surface temperature cycle in cloudy conditions is investigated.