ABSTRACT:

Snow reanalyses that combine process-based models and remote sensing observations of snow provide estimates of seasonal snow water equivalent (SWE) evolution that surpass the accuracies of traditional modeling approaches. However, snow reanalyses are only available over smaller subregions, and sometimes use computationally expensive modeling approaches. We investigate whether 1 km-resolution and daily SWE from a popular reanalysis could be learned by connecting only the most-trusted meteorological fields (multidecadal precipitation patterns and daily air temperature) and remotely sensed snow cover using a deep learning model. Relative to point observations of SWE evolution in the western United States, the lightweight deep learning model was able to reproduce the spatial and temporal evolution estimated by the snow reanalysis. Further, we found that the deep learning model could be trained in the western United States and then reused to estimate SWE evolution in the European Alps, demonstrating a high average coefficient of correlation (0.81) and low peak-SWE bias (< 1%) versus point estimates of SWE. SWE from the deep learning model also outperformed SWE estimates from physically based land surface simulations, capturing elevation-driven impacts on SWE spatial heterogeneity and interannual differences in seasonal SWE magnitudes important for water resources, climate regulation, and local ecology. This study demonstrates how deep learning approaches could be used to mine connections between daily SWE evolution, snow cover remote sensing, and limited meteorological information to generate and expand the geographical extent of fine-resolution historical snow estimates in complex terrain.

ABSTRACT:

Montane snowpack is a vital source of water in the Western United States. Here, we use a large-ensemble approach to evaluate the agreement across 124 snow water equivalent (SWE) projections with statistically downscaled forcing between end-of-century (2076 – 2095) and early 21st century (2106 – 2035) periods. Comparisons were performed on dates corresponding with the end of winter (15 April) and mid-spring snowmelt (15 May) in five western US domains. Using 1) the percent change to end-of-century SWE across different ensembles of snow projections, and 2) the shift between early 21st century and end-of-century SWE distributions for each snow projection, we identified relationships between projections that were consistent across each domain. In low to mid-elevations, end-of-century SWE decreases were 48% and larger on 15 April. These regions had projected changes to SWE that were both high-confidence and in relative agreement across projections. Despite this, the majority of 15 April SWE volume existed in higher elevations where the magnitude and direction (positive or negative) of SWE changes were most uncertain. The results of this study show that large-ensemble approaches can be used to measure coherence between snow projections and identify 1) the highest-confidence changes to future snow water resources, and 2) the locations and periods where and when improvements to snow projections would most benefit estimates of future snow water resources.

This resource provides the elevation and snow classifications pertaining to Pflug et al. (2024): Comparisons of montane snow water equivalent projections: Calculating total snow mass in regions with projection agreement and divergence in the Western US. Variables named 'SnowClass_0415' and 'SnowClass_0515' reference the snow class maps (labeled 1 - 6) for 15 April and 15 May, respectively.

ABSTRACT:

Snow distribution at wind-drift spatial scales ( 10 m) can be difficult to estimate due to modeling and observational constraints. Fortunately, the timing of snow disappearance is related to the distribution of snow water equivalent (SWE) throughout the spring snowmelt season. Here, we show that snow cover maps generated from PlanetScope’s constellation of Dove Satellites can resolve the 3 m date of snow disappearance across seven alpine domains in California and Colorado. Across a 5-year period (2019 – 2023), the average uncertainty in the date of snow disappearance, or the period of time between the last date of observed snow cover and first date of observed snow absence, was 3 days. Using a simple shortwave-based snowmelt model calibrated at nearby snow pillows, the PlanetScope date of snow disappearance could be used to reconstruct spring snow water equivalent (SWE). Relative to lidar SWE estimates, the SWE reconstruction had a spatial coefficient of correlation of 0.75, and SWE spatial variability that was biased by 9%, on average. SWE reconstruction biases were then improved to within 0.04 m, on average, by adjusting snowmelt rates using assumed distributions of SWE spatial heterogeneity and the evolution of fractional snow cover observed by PlanetScope. This study demonstrates the utility of fine-scale and high-frequency optical observations of snow cover, and the simple and annually repeatable connections between snow cover and spring snow water resources in regions with seasonal snowpack.

ABSTRACT:

Snow is a vital component of the global land surface energy and water budget. In this study, we investigate the how synthetic observations of snow water equivalent (SWE) representative of a synthetic aperture radar remote sensing platform could improve spatiotemporal estimates of snowpack. We use an Observation System Simulation Experiment, specifically investigating how much snow simulated using popular models and forcing data could be improved by assimilating synthetic observations of SWE. We focus this study across a 24°-by-37° domain in the Western United States and Canada, simulating snow at approximately 250 m resolution and hourly timesteps in water-year 2019. We perform two data assimilation experiments, including: 1) a simulation excluding synthetic observations in forests where canopies obstruct remote sensing retrievals, and 2) a simulation inferring snow distribution in forested grid cells using synthetic observations from nearby canopy-free grid cells. Results found that assimilating synthetic SWE observations improved average SWE biases at peak snowpack timing in shrub, grass, crop, bare-ground, and wetland land cover types from 14%, to within 1%. However, forested grid cells contained a disproportionate amount of SWE volume. In forests, SWE mean absolute errors at peak snowpack were 111 mm, and average SWE biases were on the order of 150%. Here, the data assimilation approach that estimated forest SWE using observations from the nearest canopy-free grid cells substantially improved these SWE biases (18%) and the SWE mean absolute error (27 mm). Data assimilation also improved estimates of the temporal evolution of both SWE, even in spring snowmelt periods when melting snow and high snow liquid water content block the synthetic SWE retrievals. In fact, in the Upper Colorado River basin, melt-season SWE biases were improved from 63% to within 1%, and the Nash Sutcliffe Efficiency of runoff improved from –2.59 to 0.22. These results demonstrate the value of a snow-focused globally relevant remote sensing platform, and data assimilation for improving the characterization of SWE and associated water availability.