Pitfalls and perils of machine learning-based passive microwave brightness temperature data assimilation over terrestrial snow in High Mountain Asia

This study explores the use of a support vector machine (SVM) as the observation operator within a passive microwave brightness temperature data assimilation framework (herein SVM-DA) to enhance the characterization of snow water equivalent (SWE) over High Mountain Asia (HMA). A series of synthetic twin experiments are conducted within the NASA Land Information System (LIS) at a number of locations across HMA. Two potential pitfalls of using the SVM-DA framework are observed for a typical deep, wet snowpack. First, variations in snow liquid water content dominate the brightness temperature spectral difference (ΔT_B) signal associated with emission from a wet snowpack, which can lead to abrupt changes in SWE during the analysis update. Second, the ensemble of SVM-based predictions can collapse (i.e., yield a near-zero standard deviation across the ensemble) when the prior estimates of snow are outside the range of snow inputs used during the SVM training procedure. Such a scenario can lead to the presence of spurious error correlations between SWE and ΔT_B, and as a consequence, can result in a degraded SWE estimate as a result of the analysis update. The SVM-DA framework is effective at improving SWE estimates (~70% reduction in RMSE relative to the Open Loop) for SWE depths less than 200 mm during dry snowpack conditions. The SVM-DA framework also improves SWE estimates in deep, wet snow (~45% reduction in RMSE) when snow liquid water is well estimated by the land surface model, but can lead to model degradation when snow liquid water estimates diverge from values used during SVM training. Such pitfalls can be largely mitigated by applying rule-based approaches. For example, restricting the SWE update when the standard deviation of the predicted ΔT_B is greater than 0.05 K helps prevent the occurrence of filter divergence. Similarly, adding a thin layer (i.e., 5 mm) of SWE when the synthetic ΔT_B is larger than 5 K can improve SVM-DA performance in the presence of a precipitation dry bias. If these pitfalls can be systematically avoided, the SVM-DA framework holds promise for consistently improving the posterior model estimate of snow mass across regional and continental scales.