An Article On The Skill Of Seasonal Predictions of Arctic Sea By Ron Lindsay

There is an excellent article in the November 2010 issue of U.S. CLIVAR VARIATIONS by Ron Lindsay of the University of Washington titled

Seasonal Predictions of Arctic Sea Ice Coverage.

It is an informative article on the ability to produce  seasonal ice coverage predictions.

The article contains asks the questions

“What are the prospects for sea ice prediction on seasonal to decadal time scales? What seasons and regions show the most promise for accurate predictions?”

Excerpts from the author on the answers include

“On monthly to seasonal time scales, accurate weather predictions are challenging, and the initial ice concentration and thickness are most important.”

“The inherent predictability of Arctic sea ice on seasonal time scales was investigated by Holland et al. (2010). Running a series of ensemble experiments using the Community Climate System Model (CCSM) with identical initial ice conditions they determined that sea ice area exhibits predictability from January for the first summer and for winter conditions in the next year. Comparing experiments initialized with
different mean ice conditions indicates that ice area in a thicker sea ice regime generally exhibits higher predictability for a longer period of time. In a thinner sea ice regime, winter ice conditions provide little ice area predictive capability after approximately 1 year. In all regimes, ice thickness (as opposed to area) has high predictability for at least 2 years.”

“…the NCEP Climate Forecast System (CFS) is used to make ensemble forecasts of the global climate out nine months or more…. it has not been tested and improved for polar conditions and the simulated sea ice is not yet a good representation of the observed ice. CFS predictive skill in the Arctic is not great. More work needs to be done to know how to best initialize a global model with observed ice thickness data or ice thickness data from a high-resolution retrospective ice–ocean model.”

“Ice thickness observations suitable for evaluating model performance and eventually to initialize model forecasts are not yet readily available.”

“Because the ice is pushed by unpredictable winds, the prediction of regional ice area, extent, or thickness is much less skillful than for the basin-wide total extent, which is less sensitive to ice moving from one part of the basin to another. Yet for field operations a prediction for a particular place or region is much more useful than one for the entire basin. The prediction uncertainty principle applies here: the smaller the region the greater the uncertainty. It will be an additional challenge to develop skillful regional forecasts.”

The bottom line message from this article is that there is some prediction skill out to at least two years, but regional skill is still quite limited. For longer time periods, this indicates that skill will be even less.  This article further supports the conclusions in my posts

Dynamic Downscaling From Multi-Decadal Global Model Projections Does Not Add Spatial and Temporal Accuracy Of Value To The Impacts Community

Statistical Downscaling From Multi-Decadal Global Model Projections Does Not Add Spatial and Temporal Accuracy Of Value To The Impacts Community

The article by Ron Lindsay, however, is an excellent example of assessing predictability, as I discussed in the post

The Difference Between Prediction and Predictability – Recommendations For Research Funding Related to These Distinctly Different Concepts

Comments Off

Filed under Climate Models, Research Papers

Comments are closed.