Bias correction based on regime-dependent Cumulative Distribution Functions for soil moisture data assimilation in a Land Surface Model

We assimilate Advanced Microwave Scanning Radiometer-EOS (AMSR-E) observations of soil moisture into the Simulator for Hydrology and Energy Exchange at the Land Surface (SHEELS) land surface model using the Land Information System (LIS). LIS has the capability to perform Ensemble Kalman Filter data assimilation of satellite soil moisture observations. It can deal with possible observation bias (relative to the model) by correcting the Cumulative Distribution Function (CDF) of observed soil moistures to match the model distribution. Separate CDFs can in principal be derived independently for each grid point, but this approach would be subject to large sampling errors. Alternately, a single correction could be used for all observations, but this cannot remove biases that are dependent on time of day, type of vegetation/land use, type of soil, etc. We compare results of soil moisture data assimilation for 1) a single correction, 2) separate day/night corrections, and 3) landcover-dependent corrections; validating against standard surface observations and in situ soil moisture measurements.