Assessing the performance of GRACE and GRACE-FO data assimilation on regional water balance component estimates

Accurate estimation of water balance components (e.g., soil moisture, groundwater, evapotranspiration, runoff) is crucial for assessment of water resource availability and climate variability. Water balance components can be simulated by land surface models (LSMs), but their accuracy is commonly limited by land process representation, uncertainty in meteorological forcing, and model parameter calibration. Data assimilation (DA) can be used to combine various satellite observations with model simulation according to the relative errors, leading to improved hydrologic component estimates. A key component, terrestrial water storage (TWS), obtained from the Gravity Recovery And Climate Experiment (GRACE) satellite mission has been effectively used in various DA systems in the past decade. Recently, the GRACE Follow-On mission (GRACE-FO) extends the availability of satellite-derived TWS data to the present. This study exploits the data from both gravity missions by assimilating the derived TWS into the Noah LSM with multiphysics options (Noah-MP). Our research objective is to assess the impact of assimilating gravity information on the individual water balance components over the continental U.S. (CONUS). The GRACE/GRACE-FO DA scheme is developed as a part of the National Climate Assessment Land Data Assimilation System (NCA-LDAS) using the Land Information System (LIS) framework. To reflect the true GRACE error, the GRACE error variance-matrix obtained from GRACE and GRACE-FO products is first calibrated and used to derive the TWS error. The results are validated using measurements from ground observation networks including surface/sub-surface soil moisture, groundwater well, flux towers, stream gauges, etc. We find that application of GRACE/GRACE-FO DA has a noticeable impact on the deep storage component (e.g., groundwater) while showing only a slight improvement in the surface layer. The latter is explained by the insensitivity of GRACE and GRACE-FO data to the signal associated with the top soil component, which is strongly governed by high-frequency meteorological forcing. The surface component can be simultaneously improved by using the NCA-LDAS multivariate DA scheme, which includes DA of surface soil moisture and snow, and will also be discussed in this presentation.