The ability of regional climate models to capture the timing and magnitude of North American snow cover and snowpack is evaluated. Observational estimates of snow over North America are highly uncertain. Climate model fidelity is evaluated against the uncertainty in observations. The focus of this work is on evaluating the models used in the North American Regional Climate Change Assessment Program (NARCCAP).
Snow is a critically important part of Earth’s climate system, both for its physical properties, including its high albedo and low thermal conductivity, and its role in the hydrological cycle. Regional climate models are often used to downscale global climate projects and provide climate change information at a scale that is more useful for the impacts and adaptation communities. It is critical that snow cover and snow water equivalent (the amount of water stored on the surface as snow) are captured by these models. This paper provides a unique methodology and key metrics for evaluating snow in regional climate models at continental scales.
The magnitude, timing, and spatial extent of snow water equivalent (SWE) over North America are evaluated in the reanalysis-driven NARCCAP regional climate models. There is considerable uncertainty in observations of SWE over North America so an ensemble of SWE observations is created to evaluate how well the NARCCAP models perform relative to the observational uncertainty. While observational uncertainty is high, the model bias in SWE typically occurs outside of the observational range. The combined effects of biases in simulated temperature and precipitation and differences in land surface snow process parameterizations drive biases in the annual cycle of North American SWE in the RCMs. The drivers of SWE bias were found to vary regionally. Additional land-surface only simulations are needed as part of multi-model ensembles to fully assess the role that snow parameterizations play in driving snow biases.