Analysis of gap filling techniques for GRACE/GRACE-FO terrestrial water storage anomalies in Canada

Since its launch in early 2002, datasets from the Gravity Recovery and Climate Experiment (GRACE) and its 2018 follow-on mission (GRACE-FO) have become indispensable for monitoring terrestrial water storage. GRACE-derived terrestrial water storage anomalies (TWSA), especially when used in conjunction with other water budget datasets (i.e., precipitation, evapotranspiration, surface runoff), provide insight into groundwater and glacier mass fluctuations. With over 20 years of observations, long-term statistical analysis reveals water storage trends, however, the 11-month gap between the two missions must be filled. The goal of this presentation is to compare four gap filling methods over Canada, namely: extreme gradient boosting, artificial neural networks, an automated machine learning (ML) algorithm (AutoML), and projection onto convex sets (POCS). The GRACE mascon product (RL06M.MSCNv02) released by the Jet Propulsion Laboratory was used, and all data were bounded by GRACE availability (April 2002 - March 2022) at the time of the study. Reconstruction of TWSA data in Canada required consideration of glacier surface mass balance models derived from the GMAO Modern-Era Retrospective Analysis for Research and Applications, Version 2 reanalysis, and the Randolph Glacier Inventory. Nine sets of nation-wide hydrological and climatic parameters were used as predictors for the machine learning models: GRACE average seasonal signal, TWSA from the GLDAS Catchment Land Surface Model, precipitation, air temperature, glacier surface mass balance, ocean tides, the North Atlantic Oscillation, the Multivariate ENSO Index, and sea surface temperature. Results indicate that ML algorithms can fill the gap with mean normalized root mean square errors ranging from 6 – 13% and AutoML performs the best with 2.5 cm equivalent water height (EWH). The purely mathematical signal synthesis POCS method resulted in approximately 4.0 cm EWH for some basins. Filling the gap between missions allows for more comprehensive terrestrial water storage trend analysis, including basin-by-basin analysis, which is ongoing over Canada.