Assessment of an ESA CCI Soil Moisture data assimilation framework

We present a comprehensive assessment of a land surface data assimilation system, in which microwave-based satellite retrievals of surface soil moisture from the combined active-passive ESA CCI Soil Moisture product are assimilated into the Noah-MP model, using a one-dimensional Ensemble Kalman Filter (EnKF) within the NASA Land Information System (LIS). This data assimilation system produces consistent estimates of surface and root-zone soil moisture, as well as all other geophysical variables, over the European continent from January 2002 to December 2019.

The aim of this study is twofold. Firstly, we explore the impact of design choices and forcing inputs on the skill of the data assimilation system, specifically: (1) the magnitude of observation errors, (2) the bias correction method, i.e., climatological or seasonal CDF matching, and (3) the choice of the meteorological reanalysis dataset used to drive the land surface model. For the latter, we compare the results obtained by forcing the Noah-MP model with the NASA Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2) with those obtained by forcing the model with the European Center for Medium-Range Weather Forecasts (ECMWF) Reanalysis version 5 (ERA5). Secondly, we explore how the data assimilation skill is related to the quality of the satellite retrievals and environmental factors such as land cover, soil texture, and climate.

For both objectives listed above, the skill of the data assimilation system is evaluated by comparing the surface and root-zone soil moisture estimates with in situ observations. Furthermore, we evaluate the behavior of internal diagnostics derived from the data assimilation innovations and increments.

The results display the inevitable trade-off in choosing the observation error magnitude: a smaller observation error will cause the data assimilation to perform worse than an open loop run at some sites, whereas a larger observation error will reduce the skill at well-performing sites. We also show that the bias correction method and the choice of meteorological forcing both have a clear effect on the data assimilation diagnostics, but a negligible impact on the skill of the system that is observed over in situ reference sites. Finally, we show that the skill improvement by the data assimilation framework is strongly related to the quality of the satellite soil moisture retrievals.        

Acknowledgments: this work is part of the ESA CCI+ Soil Moisture CCN1 Scientific Evolution project and the FWO-FWF CONSOLIDATION project.