Abstract:
Satellite retrievals of land surface temperature (LST) are available from a variety of polar orbiting and geostationary platforms. Assimilating such LST retrievals into a land surface model (that is either driven by observed meteorological forcing data or coupled to an atmospheric model) should improve estimates of land surface conditions. However, LST data from retrievals and models typically exhibit very different climatologies for a variety of reasons, including model trade-offs between numerical stability and computational cost, uncertainties in land surface emissivity, satellite look-angle, and other sensor characteristics.

We overcome the challenges facing LST assimilation through scaling and bias estimation approaches. In the scaling approach, the LST retrievals from each sensor are scaled to the model's LST climatology before they are assimilated into the land model. After assimilation, the merged LST product may be scaled back into the climatology of the LST retrievals if the application calls for it. Because of the strong seasonal and diurnal cycle of LST, scaling parameters must be derived separately for each 3-hour interval and for each month. The bias estimation approach dynamically estimates diurnally varying model bias parameters. This approach may be more appropriate as long as the satellite climatology of the LST retrievals is homogenenous across all the sensors that are utilized. In the presentation, we compare the different approaches by assimilating land surface temperature retrievals from the International Satellite Cloud Climatology Project (ISCCP) into the NASA Catchment land surface model with an ensemble-based data assimilation system developed at the NASA Global Modeling and Assimilation Office. The assimilation estimates are evaluated against in situ measurements of land surface temperature and surface turbulent fluxes from the Coordinated Energy and Water Cycle Observations Project (CEOP).