Societies and ecosystems within and downstream of mountains rely on seasonal snowmelt to satisfy their water demands. Anthropogenic climate change has reduced mountain snowpacks worldwide, altering snowmelt magnitude and timing. Here the global warming level leading to widespread and persistent mountain snowpack decline, termed low-to-no snow, is estimated for the world’s most latitudinally contiguous mountain range, the American Cordillera. We show that a combination of dynamical, thermodynamical and hypsometric factors results in an asymmetric emergence of low-to-no-snow conditions within the midlatitudes of the American Cordillera. Low-to-no-snow emergence occurs approximately 20 years earlier in the southern hemisphere, at a third of the local warming level, and coincides with runoff efficiency declines (8% average) in both dry and wet years. The prevention of a low-to-no-snow future in either hemisphere requires the level of global warming to be held to, at most, +2.5 °C.
Snowpack and snowmelt-driven extreme events (e.g., floods) have large societal consequences including infrastructure failures. However, it is not well understood how projected changes in the snow-related extremes differ across North America. Using dynamical downscaled regional climate model (RCM) simulations, we found that the magnitudes of extreme snow water equivalent, snowmelt, and runoff potential (RP; snowmelt plus precipitation) decrease by 72%, 73%, and 45%, respectively, over the continental U.S. and southern Canada but increase by up to 8%, 53%, and 41% in Alaska and northern Canada by the late 21st century. In California and the Pacific Northwest, there is a notable increase in extreme RP by 21% contrary to a decrease in snowmelt by 31% by the late century. These regions could be vulnerable to larger rain-on-snow floods in a warmer climate. Regions with a large variability among RCM ensembles are identified, which require further investigation to reduce the regional uncertainties.
Accurate prediction of snow water equivalent (SWE) can be valuable for water resource managers. Recently, deep learning methods such as long short-term memory (LSTM) have exhibited high accuracy in simulating hydrologic variables and can integrate lagged observations to improve prediction, but their benefits were not clear for SWE simulations. Here we tested an LSTM network with data integration (DI) for SWE in the western United States to integrate 30-day-lagged or 7-day-lagged observations of either SWE or satellite-observed snow cover fraction (SCF) to improve future predictions. SCF proved beneficial only for shallow-snow sites during snowmelt, while lagged SWE integration significantly improved prediction accuracy for both shallow- and deep-snow sites. The median Nash–Sutcliffe model efficiency coefficient (NSE) in temporal testing improved from 0.92 to 0.97 with 30-day-lagged SWE integration, and root-mean-square error (RMSE) and the difference between estimated and observed peak SWE values dmax were reduced by 41% and 57%, respectively. DI effectively mitigated accumulated model and forcing errors that would otherwise be persistent. Moreover, by applying DI to different observations (30-day-lagged, 7-day-lagged), we revealed the spatial distribution of errors with different persistent lengths. For example, integrating 30-day-lagged SWE was ineffective for ephemeral snow sites in the southwestern United States, but significantly reduced monthly-scale biases for regions with stable seasonal snowpack such as high-elevation sites in California. These biases are likely attributable to large interannual variability in snowfall or site-specific snow redistribution patterns that can accumulate to impactful levels over time for nonephemeral sites. These results set up benchmark levels and provide guidance for future model improvement strategies.
This study evaluates snow water equivalent (SWE) over North America in the reanalysis-driven NARCCAP regional climate model (RCM) experiments. Examination of SWE in these runs allows for the identification of bias due to RCM configuration, separate from inherited GCM bias. SWE from the models is compared to SWE from a new ensemble observational product to evaluate the RCMs’ ability to capture the magnitude, spatial distribution, duration, and timing of the snow season. This new dataset includes data from 14 different sources in five different types. Consideration of the associated uncertainty in observed SWE strongly influences the appearance of bias in RCM-generated SWE. Of the six NARCCAP RCMs, the version of MM5 run by Iowa State University (MM5I) is found to best represent SWE despite its use of the Noah land surface model. CRCM overestimates SWE because of cold temperature biases and surface temperature parameterization options, while RegCM3 (RCM3) does so because of excessive precipitation. HadRM3 (HRM3) underestimates SWE because of warm temperature biases, while in the version of WRF using the Grell scheme (WRFG) and ECPC-RSM (ECP2), the misrepresentation of snow in the Noah land surface model plays the dominant role in SWE bias, particularly in ECP2 where sublimation is too high.
The NARCCAP RCM–GCM ensemble is used to explore the uncertainty in midcentury projections of snow over North America that arise when multiple RCMs are used to downscale multiple GCMs. Various snow metrics are examined, including snow water equivalent (SWE), snow cover extent (SCE), snow cover duration (SCD), and the timing of the snow season. Simulated biases in baseline snow characteristics are found to be sensitive to the choice of RCM and less influenced by the driving GCM. By midcentury, domain-averaged SCE and SWE are projected to decrease in all months of the year. However, using multiple RCMs to downscale multiple GCMs inflates the uncertainty in future projections of both SCE and SWE, with projections of SWE being more uncertain. Spatially, the RCMs show winter SWE decreasing over most of North America, except north of the Arctic rim, where SWE is projected to increase. SCD is also projected to decrease with both a later start and earlier termination of the snow season. For all metrics considered, the magnitude of the climate change signal varies across the RCMs. The ensemble spread is large over the western United States, where the RCMs disagree on the sign of the change in SWE in some high-elevation regions. Future projections of snow (both magnitude and spatial patterns) are more similar between simulations performed with the same RCM than the simulations driven by the same GCM. This implies that climate change uncertainty is not sufficiently explored in experiments performed with a single RCM driven by multiple GCMs.