Estimating spatiotemporally continuous snow water equivalent from intermittent satellite observations: an evaluation using synthetic data

Accurate estimates of snow water equivalent (SWE) based on remote sensing have been elusive, particularly in mountain areas. However, there now appears to be some potential for direct satellite-based SWE observations along ground tracks that only cover a portion of a spatial domain (e.g., watershed). Fortunately, spatiotemporally continuous meteorological and surface variables could be leveraged to infer SWE in the gaps between satellite ground tracks. Here, we evaluate statistical and machine learning (ML) approaches to performing track-to-area (TTA) transformations of SWE observations in California's upper Tuolumne River watershed using synthetic data. The synthetic SWE measurements are designed to mimic a potential future P-band Signals of Opportunity (P-SoOP) satellite mission with a (along-track) spatial resolution of about 500 m. We construct relationships between multiple meteorological and surface variables and synthetic SWE observations along observation tracks, and we then extend these relationships to unobserved areas between ground tracks to estimate SWE over the entire watershed. Domain-wide, SWE inferred on 1 April using two synthetic satellite tracks (similar to 4.5 % basin coverage) led to percent errors of basin-averaged SWE (PEBAS) of 24.5 %, 4.5 % and 6.3 % in an extremely dry water year (WY2015), a normal water year (WY2008) and an extraordinarily wet water year (WY2017), respectively. Assuming a 10 d overpass interval, percent errors of basin-averaged SWE during both snow accumulation and snowmelt seasons were mostly less than 10 %. We employ a feature sensitivity analysis to overcome the black-box nature of ML methods and increase the explainability of the ML results. Our feature sensitivity analysis shows that precipitation is the dominant variable controlling the TTA SWE estimation, followed by net long-wave radiation (NetLong). We find that a modest increase in the accuracy of SWE estimation occurs when more than two ground tracks are leveraged. The accuracy of 1 April SWE estimation is only modestly improved for track repeats more often than about 15 d.