Assessment of Snow, Sea Ice, and Related Climate Processes in Canada's Earth System Models and Climate Prediction Systems

This paper represents a focus activity of the Canadian Sea Ice and Snow Evolution (CanSISE) Network, which is a multi-year research project on seasonal snow cover, sea ice, and related climate processes involving a network collaboration between Canadian university and Environment and Climate Change Canada (ECCC) scientists.

We here assess the capability of the Canadian Seasonal to Interannual Prediction System (CanSIPS) and the Canadian Earth System Model 2 (CanESM2) to simulate and predict snow and sea ice from seasonal to multi-decadal timescales, with a focus on the Canadian sector. This analysis takes place in the context of observational uncertainty, internal climate variability, and available output from international Earth System Models. The report highlights the application of multi-source observational datasets and model simulation sets created in the collaboration between the CanSISE Network and ECCC.

 

It is found that the quality of the CanESM2 simulation of snow-related climate parameters, such as cold-region temperature and precipitation, lies well within the range of currently available international models. Accounting for the considerable disagreement among satellite-era observational datasets on the distribution of snow water equivalent, CanESM2 has too much snow cover and an unrealistic spatial distribution of SWE in the spring over the Canadian land mass and has too much springtime snow over the Northern Hemisphere as a whole. CanESM2 exhibits springtime retreat of snow in the satellite era that is generally greater than observational estimates. Sea ice is biased low in the Canadian arctic and subarctic, and the amount of snow on floating wintertime sea ice is probably also biased low, although observational estimates of snow on sea ice are uncertain. The report discusses tradeoffs in having a model system that is sufficiently computationally inexpensive to afford operational seasonal prediction and multiple realizations under different forcings, but of high enough resolution to capture key geographical features for simulating snow and sea ice. Improvements in climate prediction systems like CanSIPS relies not just on simulation quality but on being able to take advantage of novel observational constraints and being able to transfer research to an operational setting. Research results from CanSISE suggest potential improvements in seasonal forecasting practice using CanSIPS, including the impact of accurate initialization of the state of snow and frozen soil, properly accounting for observational uncertainty in forecast verification, and operational implementation of sea ice thickness initialization using statistical predictors available in real time.