Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

Climate Variability & Predictability (CVP) logo

Improving Initialization of Arctic Sea Ice in NCEPΓÇÖs Climate Forecast System for Advancing Long-Range Predictions

Prediction and predictability of Arctic sea ice on different time scales has received increasing attention recently. While “perfect-model” studies have shown that Arctic sea ice extent is predictable out to eight months or longer, diagnoses of the forecasts from the current dynamical operational climate models show that the useful prediction skill for interannual sea ice anomalies is lost beyond the first 2-3 months. Similarly, the analysis of the outlook collected by the NOAA SEARCH (Study of Environmental Arctic Change) indicates that predicting the variability of the September Arctic sea ice is difficult even from July initial conditions, and further, there exists a substantial spread among the forecasts from different prediction systems. Understanding the causes of the forecast errors and the discrepancy between the potential predictability and actual skill, and enhancing the skill of prediction of Arctic sea ice to at par with predictability estimates is a highly desirable goal to improve the scope and reliability of NOAA’s sea ice operational prediction capabilities.

Improvements in the skill of long-range forecasts for Arctic sea ice can stem from many sources. One such source is the correct initialization of sea ice thickness (SIT). Despite its perceived importance, however, the influence of the observed information in initial SIT on the prediction is not well incorporated in the current generation of operational forecast systems. As an example, the initial SIT in the National Centers for Environmental Prediction (NCEP) Climate Forecast System version 2 (CFSv2) is from the Climate Forecast System Reanalysis (CFSR) which contains substantial SIT errors. The erroneous initial SIT used in CFSv2 is an important factor limiting its sea ice prediction skill.

The main objectives of this proposal are to i) investigate the contribution of initial SIT to the prediction of Arctic sea ice in the NOAA’s NCEP seasonal climate prediction system, and ii) improve its prediction skill by improving the SIT initialization. In addition, we will also analyze how the influence of initial SIT on prediction skill relates to the reduction of forecast model’s systematic bias caused by uncertainties in the atmosphere-ice-ocean interactions. We will accomplish these objectives with the following activities:
1) Historical forecast experiments with an alternative initial SIT from the well calibrated Pan-arctic Ice/Ocean Modeling and Assimilation System (PIOMAS) to analyze improvements in sea ice prediction skill, and how the skill can be affected by forecast model’s systematic bias;
2) Development of an improved initialization by adopting the approach of the PIOMAS in the CFSv2 ice/ocean component model to explore optimal sea ice parameterizations, and implement an approach similar to PIOMAS for improving SIT initialization in the fully coupled CFSv2; and
3) Forecast experiments with the new initial conditions from 2) to demonstrate the advantage of using an improved SIT initialization that is consistent with the forecast model for an improved sea ice prediction.

The proposed research will lead to a better sea ice forecasts to meet the requirement for the prediction of seasonal sea ice melting and freezing by operational institutions such as the NWS field offices in Alaska. The initialization procedure developed herein will also contribute to an improved sea ice prediction for NOAA’s participation in the national SEARCH sea ice outlook and in the international collaboration for the Polar Prediction Project (PPP).

This project is based on the framework of the current operational CFSv2. The sea ice component in CFSv2 will continue to be used in the updated Modular Ocean Model version 5 (MOM5) which is expected to be the oceanic component for the next generation of the Climate Forecast System (CFSv3). Accordingly, the proposed research will not only result in an improved prediction using the current CFS framework but will also contribute to the continued improvement of CFS. The proposed research will also improve our understanding of sea ice prediction skill and predictability associated with initial SIT, which has been an active area of scientific research among the climate community.

Scroll to Top