Enhancement of NCA-LDAS Version 3 Through Multisensor, Multivariate Data Assimilation

The National Climate Assessment-Land Data Assimilation System (NCA-LDAS) is an enabling tool for development, evaluation, and dissemination of hydrological indicators in support of the NCA. The primary motivation is to improve the characterization of regional-scale terrestrial energy and water budgets as well as monitoring and prediction of climate-relevant water availability indicators, including droughts and floods, through the assimilation of satellite observations. NCA-LDAS, developed within the NASA Land Information System (LIS), assimilates multi-sensor observations from 1979 to present into a state-of-the-art surface model (LSM). Former NCA-LDAS (Versions 1 and 2) only included assimilation of soil moisture and snow using Noah 3.3 LSM. Version 3 now employs Noah-MP 3.6 with dynamic vegetation and a groundwater store, together with a more comprehensive suite of satellite data including soil moisture (from ESA-CCI, AMSR-E, ASCAT, SMOS, SMAP), snow depth (from SMMR, SSM/I), snow cover area (from MODIS), leaf area index (from GLASS), and terrestrial water storage (from GRACE, GRACE-FO). The multivariate data assimilation (DA) scheme employs a 1-D ensemble Kalman smoother (EnKS), which is capable of assimilating monthly-averaged (e.g., GRACE) and instantaneous observations simultaneously, based on the time window of the data record. The DA scheme allows the surface components to update toward the high-frequency observation (e.g., soil moisture, snow) while the root zone and groundwater components are constrained by, e.g., LAI, GRACE. Quality control is conducted throughout the implementation of multivariate DA. The climatology corrections required in satellite soil moisture and GRACE observations are analyzed to minimize bias that is commonly caused by assimilation of multiple observations. Results are evaluated using ground truth networks including surface/root zone soil moisture, groundwater wells, flux towers and stream gauges, and other independent data products.