A novel approach combining satellite and in situ observations to estimate the daytime variation of land surface temperatures for all sky conditions

Land surface temperature (LST) and its diurnal variability are key to understanding the land-atmosphere interactions, hydrological processes and climate change. However, at any given point in time approximately half of the Earth's surface is covered by clouds. This restricts the availability of LST through satellite remote sensing, which works best under clear skies. However, in situ observations continue to monitor atmospheric conditions beneath the clouds that could complement satellite measurements during cloudy conditions. The present study explores a novel approach to estimate hourly LST during the daylight hours using remotely sensed surface solar absorption and in situ observations of daily LST extremes (maximum and minimum) together with an adaptive non-linear fitting approach. A learning algorithm trained against in-situ measurements of LST extrema and diurnal cycle of surface solar absorption together with the associated linear correlation between the two parameters, is used to estimate an optimized set of parameters to approximate hourly LST for each day during the daylight hours between sunrise and sunset. Results show that the method captures the intra-day variability of LST very well under most sky conditions with rms errors below 1.5 K.