Variational Method for Assimilating GOES-Retrieved LSTs into SVATS Model for Soil Moisture Initialization

Soil moisture is a key variable in mesoscale model initialization, particularly when the distribution of soil moisture is heterogeneous and organized. Boundary layer circulations can quickly respond to organized gradients at particular spatial scales and therefore in conducting predictions from a heterogeneous surface initial state, it is generally important to characterize the broad features in near surface wetness. In this study we describe a method in which GOES satellite retrievals of diurnally varying land surface temperature (LST) over the southwest ARM-CART/GCIP measuring domain, are used for data assimilation purposes in a SVATS model in order to produce optimal soil moisture initialization. This study is part of a larger research effort to test the impact of improved soil moisture initialization on a non-hydrostatic mesoscale model, run at high resolution over a semi-arid landscape where soil moisture discontinuities are often large. For this presentation, a description of the assimilation methodology for both 3D-VAR and 4D-VAR constructs and special properties of the SVATS adjoint model are addressed.