Articles | Volume 21, issue 11
Research article
28 Nov 2017
Research article |  | 28 Nov 2017

SMOS brightness temperature assimilation into the Community Land Model

Dominik Rains, Xujun Han, Hans Lievens, Carsten Montzka, and Niko E. C. Verhoest

Abstract. SMOS (Soil Moisture and Ocean Salinity mission) brightness temperatures at a single incident angle are assimilated into the Community Land Model (CLM) across Australia to improve soil moisture simulations. Therefore, the data assimilation system DasPy is coupled to the local ensemble transform Kalman filter (LETKF) as well as to the Community Microwave Emission Model (CMEM). Brightness temperature climatologies are precomputed to enable the assimilation of brightness temperature anomalies, making use of 6 years of SMOS data (2010–2015). Mean correlation R with in situ measurements increases moderately from 0.61 to 0.68 (11 %) for upper soil layers if the root zone is included in the updates. A reduced improvement of 5 % is achieved if the assimilation is restricted to the upper soil layers. Root-zone simulations improve by 7 % when updating both the top layers and root zone, and by 4 % when only updating the top layers. Mean increments and increment standard deviations are compared for the experiments. The long-term assimilation impact is analysed by looking at a set of quantiles computed for soil moisture at each grid cell. Within hydrological monitoring systems, extreme dry or wet conditions are often defined via their relative occurrence, adding great importance to assimilation-induced quantile changes. Although still being limited now, longer L-band radiometer time series will become available and make model output improved by assimilating such data that are more usable for extreme event statistics.

Short summary
We have assimilated 6 years of satellite-observed passive microwave data into a state-of-the-art land surface model to improve surface soil moisture as well as root-zone soil moisture simulations. Long-term assimilation effects/biases are identified, and they are especially dependent on model perturbations, applied to simulate model uncertainty. The implications are put into context of using such assimilation-improved data for classifying extremes within hydrological monitoring systems.