91ÃÛÌÒ¸ó

Spatial variability of snow depth and snow water equivalent (SWE) within a basin has long been recognized as a central challenge for predicting snowmelt runoff and managing water resources. Although spatial variability is widely studied, comparatively less attention has been given to the recurrence or repeatability of these patterns. Recent advances in high-resolution remote sensing (lidar, optical satellite, InSAR), increased computing capacity, and machine learning (ML) approaches have enabled new methods for characterizing and leveraging spatial snow patterns to estimate distributed snow depth and SWE. A crucial requirement for applying these methods, however, is the recurrence of these snow patterns. For a ML model to generalize from training datasets and perform reliably across years, the underlying spatial patterns between training and real-time snowpack must remain consistent. Yet, the drivers of year-to-year pattern repeatability remain poorly understood.Ìý

Here we present a brief review of recent studies addressing snow pattern temporal variability through time, and present new analyses of interannual repeatability using MODIS fSCA, airborne lidar surveys (ASO) observations, and SPIReS-MODIS-ParBal SWE. Over the western United States, repeatability of snow patterns is assessed (1) within basins under varying hydrometeorological conditions, and (2) regionally across basins due to climatological patterns. Initial analysis, using principal component analysis (PCA) on SPIReS-MODIS-ParBal SWE shows that many basins in the Sierra Nevada have high year-to-year repeatability, with the first principal component explaining over 85% of variability, while more northern and continental snowpacks show greater interannual variability (PC1 fractional variance ranges from 50-80%).