Land surface temperature plays an important role in land surface processes, and it is a key input to physically-based retrieval algorithms of important hydrological states and fluxes, such as soil moisture and evaporation. For this reason there are many independent estimates of land surface temperature from different sensors and satellites, often at multiple times during the day. Besides uncertainties and biases associated with each estimate, the very nature of the temperature at the very surface layer of the land makes it difficult to combine temperature estimates between different methods. The skin temperature is characterized by a strong diurnal cycle that is dependant in timing and amplitude on the exact sensing depth and thermal properties of the vegetation.

Addressing these structural differences and biases requires a continuous record of surface temperature. Such record would make it possible to estimate the temperature at different depths within the canopy/soil surface through established heat flow modeling techniques. Moreover, it would allow extracting temperature estimates at the exact overpass times of satellites that do not carry an on-board temperature sensor. Particularly for the Soil Moisture and Ocean Salinity satellite and the planned Soil Moisture Active Passive satellite this could potentially improve the quality of their soil moisture outputs.

This study investigates the use of temperature output from the Modern Era Retrospective-analysis for Research and Applications (MERRA) to serve as a backbone temperature set for the assimilation of satellite estimates of surface temperature. The satellite temperature estimates are derived from six satellites that carry a microwave Ka-band (37 GHz) radiometer. This study gives detailed new insights in the relative properties of different temperature data sets; numerical based estimates, and retrievals from passive microwave sensors and thermal infrared radiometers. It is shown that the satellite observations particularly have an advantage in terms of estimating the daily amplitude in temperature (as validated relative to in situ measurements).

Based on these analyses a framework for merging satellite temperature observations with the (hourly) MERRA skin temperature output is developed. In a pre-processing step the structural differences and biases are addressed. Then, an ensemble Kalman filter is parameterized to assimilate the observed amplitude with that of the MERRA skin temperature set, resulting in a reduction of error levels during the day.