Seasonal snow cover and sea ice are integral to the cultural identity, history, and economy of northern nations like Canada. They also exert an enormous physical influence on the earth system, ranging from local interactions with winds and temperatures in the Arctic, to larger-scale interactions with weather systems and ocean circulation, to global-scale influences on the Earth’s energy balance.
In recent decades, dramatic changes in Canada’s snow cover and sea ice have been observed, and this has driven the need to better understand and predict these fields for the coming seasons, years and decades. To address this need, Canada, and the Canadian Sea Ice and Snow Evolution (CanSISE) network, in particular, has helped lead the global effort to better observe and model snow, sea ice and related climate processes. This study represents the first of three major deliverables from the CanSISE network and it evaluates the ability of Canada’s earth system model and climate prediction system to predict snow, sea ice and related climate parameters in the context of the development of new observational products. It seeks to more fully assess the simulation and prediction of seasonal snow cover and regional sea ice variability accompanied by a more complete characterization of observational uncertainty, model structural uncertainty and internal climate variability.
Methodology
We reviewed the current generation Canadian Seasonal to Interannual Prediction System (CanSIPS) and Canadian Earth-System Model 2 (CanESM2). CanSIPS and its component models provide Environment and Climate Change Canada’s current operational forecasts for climate variability on seasonal to interannual (several-month to multiple-year) timescales. CanSIPS combines a dynamical ocean model with atmosphere, land surface (including snow) and sea ice component models in a coupled framework in which all model components interact. It includes 1) a data assimilation system that estimates realistic initial states of the atmosphere, ocean, land, and sea ice to start the forecasts; 2) two separate coupled climate models and 3) diagnostic systems to analyze the output and generate useful forecasts for operational use.
CanESM2 includes interactive land and ocean carbon cycle components to project the future state of global temperature, circulation, carbon dioxide concentrations, etc. under the influence of external forcing. CanESM2 does not use flux adjustments that artificially constrain the climate system to be in a state of energy and water balance.
Our assessment of CanSIPS/CanESM2 is enhanced by two recent research products that are outputs from the CanSISE network: the observation-based Blended-5 snow water equivalent (SWE) dataset, and the CanESM2 Large Ensemble of simulations. The CanESM2 Large Ensemble afforded a more thorough assessment of the uncertainty connected to internal climate variability than was possible in past analysis efforts and has been used in several current and ongoing studies.
Outcomes
We first evaluate the climatological characteristics of CanESM2 temperature, precipitation and SWE (Fig.1). In winter and spring, the distribution of surface temperature over Canada is well reproduced, although a warm bias is evident in both seasons. In precipitation, the general spatial pattern and latitudinal gradients are captured in the model, but there is excessive precipitation over the Western Cordillera, sub-Arctic and Arctic in both seasons. This excessive precipitation contributes to a positive bias in SWE over much of Western Canada and the Canadian Sub-Arctic.
FIGURE 1: Comparison of simulated Canadian climate in CanESM2 with observations and other climate models for 1981-2005. The left column shows observed April -June mean temperature (top, HadCRUT4), precipitation (centre, CRU TS3.21) and snow water equivalent (bottom, Blended 5 data).
The observed SWE climatology, variability and trends are relatively non-robust compared to variables such as temperature, and for this reason, we evaluate the CanESM2 model simulation in the context of the spread across the five datasets that comprise the Blended-5 SWE product. CanESM2’s SWE is generally within the observational range for most of the year, with the notable exception of spring. This springtime positive bias is largest over Canada in particular, especially outside the Arctic. Our assessment is that, especially in midlatitudes, CanESM2 simulates excessive springtime snow associated with excessive wintertime precipitation building up throughout winter and into spring.
With regards to sea ice, we find that the summertime extent is biased low in CanESM2, particularly in the Canadian Arctic sector, where CanESM2 simulates less than half of the observed sea ice coverage in the Beaufort Sea-Arctic Ocean. The utility of regional sea ice analysis with this model is limited by the moderate spatial resolution of the model and its associated land-sea distribution, particularly in the network of small islands and channels in the Canadian Arctic Archipelago.
CanSISE-related research has examined the ability of CanSIPS both to realistically initialize SWE and to predict future SWE variations, and suggests potential for practical utilization of such forecasts. Two primary strengths of CanSIPS forecasts of SWE have been identified. The first is the demonstrated ability of CanSIPS to provide realistic initial values for SWE, combined with the natural tendency for SWE anomalies to persist throughout the snow season in regions where winter temperatures remain below freezing. The second is the ability of CanSIPS to predict climatic conditions such as temperature and precipitation anomalies three to six months ahead, which then directly influence snow accumulation and snow melt.
Conclusions
The CanESM2 earth system model can be classified as above-average in its simulation of temperature, precipitation, snow and sea ice in the Canadian Arctic, compared to other models of its generation. Certain issues have been identified and, accounting for the considerable disagreement among satellite-era observational datasets on the distribution of SWE, CanESM2 has too much springtime snow cover over the Canadian land mass, reflecting a broader positive bias in the Northern Hemisphere. It also exhibits retreat of springtime snow generally greater than observational estimates, after accounting for observational uncertainty and internal variability. Sea ice is biased low in the Canadian Arctic, which makes it difficult to assess the realism of long-term sea ice trends there. The strengths and weaknesses of the modelling system need to be understood as a practical trade-off: the Canadian models are relatively inexpensive computationally because of their moderate resolution, thus enabling their use in operational seasonal prediction and for generating large ensembles of multi-decadal simulations.
Many graduate students and postdoctoral researchers in our climate modelling group at Waterloo have contributed to the CanSISE network, which has delivered new and important scientific advances. A major contribution is the new observation-based global SWE product that has been adopted and widely used in the community for multiple purposes, such as to improve the initialization of seasonal forecasts for CanSIPS (another CanSISE contribution), to evaluate historical SWE trends, and to validate weather and climate models. These (among many others) are new capabilities that CanSISE has enabled, and they represent lasting contributions to the global climate and cryospheric communities.
Kushner, P.J., Mudry, L.R., Merryfield, W., Ambadan, J.T., Berg, A., Bichet, A., Brown, R., Derksen, C.P., Déry, S.J., Dirkson, A., Flato, G., Fletcher, C.G., Fyfe, J.C., Gillett, N., Haas, C., Howell, S., Laliberté, F., McCusker, K., Sigmond, M., Sospedra-Alfonso, R., Tandon,N.F., Thackeray, C., Tremblay, B., & Zwiers, F.W. (2018). Assessment of Snow, Sea Ice, and Related Climate Processes in Canada’s Earth-System Model and Climate Prediction System. The Cryosphere. 12, 1137-1156.
Contact: Christopher G. Fletcher, Geography and Environmental Management
For more information about WaterResearch, contact Amy Geddes.
