Comparison of Univariate Assimilation using GRACE / GRACE-FO Retrievals, GPS Surface Deformation, and Leaf Area Index Retrievals in Enhancing Land Surface Model Performance

Land surface models (LSMs) are used to quantify land surface states and fluxes. However, LSMs often lack comprehensive representation of process physics and are inherently uncertain due to uncertain boundary conditions. One way to improve performance of LSMs is by conditioning model estimates on satellite observations and/or ground-based measurements using an ensemble Kalman filter framework.

In this study, the Noah-MP version 4.0.1 LSM without conditioning (a.k.a., Open Loop; OL) is compared against the same model with conditioning (a.k.a., Data Assimilation; DA). Three different DA experiments are conducted using the following datasets: 1) terrestrial water storage (TWS) anomalies from GRACE / GRACE-FO, 2) leaf area index (LAI) retrievals from MODIS, and 3) vertical deformation measurements from ground-based GPS stations. Modeled states and fluxes from the OL and DA are then compared to independent, ground-based measurements such as the U.S. SNOTEL network, the Canadian CanSWE product, the U.S. SCAN network for soil moisture, and the USGS measurement network river discharge. Statistical analyses, including bias, RMSE, and normalized information content (NIC), are computed to quantify the marginal improvements via each respective assimilation experiment. Results provide a basis to better understand the coupling between different state variables (i.e., snow mass, soil moisture, and groundwater) in order to better understand the sensitivity to changes in TWS associated with changes in the individual components of TWS. Experiments are conducted across different basins in North America with a particular focus on snow-dominated regions.