Data Assimilation of Satellite Soil Moisture Retrievals in the Land Surface Model SURFEX

Soil moisture plays an essential role in the land surface energy balance, ap- portioning the available energy into latent and sensible heat. As an impor- tant boundary for the atmosphere, accurate initialization of the land surface can help numerical weather prediction (NWP)-models increase their skill. The aim of this study is to identify the potential of integrating satellite de- rived soil moisture products in a land surface model over Norway. Offline simulations with the SURFace EXternalisée (SURFEX) land surface model are performed, forced by output from an operational NWP-model. Exper- iments cover the three summer months June, July, August (JJA) 2016, for a domain in south-eastern Norway. Level 3 data from the Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) satellites are, separately, integrated in soil moisture analyses using a simplified ex- tended Kalman filter (SEKF). In order to explicitly represent the retrieved soil moisture variable, the multilayer diffusion version of Interaction Sol- Biosphère-Atmosphère (ISBA) is used. Analysis is performed in the upper seven model soil layers to cover the root zone. The satellite data are found to be of questionable quality for the area of interest, with poor spatial coverage and noisy signal. Investigation of the linearized observation operator reveals that more weight is given to sur- face layers in dry conditions, and deep layers in wet conditions. Over all water is added to the surface layers, and removed below 10 cm. Layers below 40 cm have limited communication with the observed layer during a 6 hour assimilation window, and are suggested to be removed from the control vector. In comparison with observations of 2 meter temperature, simulations with data assimilation (DA) of soil moisture show limited im- provement over the control. However, the change in surface fluxes is more pronounced, mainly adjusting the Bowen ratio. Since the simulations are performed offline, the full potential of soil moisture increments are not be- lieved to be reached.