Recent changes in the extent, thickness, and properties of Arctic sea ice have captured attention and posed significant challenges to a wide range of stakeholders. There is a rising demand for sea ice prediction at seasonal-to-interannual timescales. Sea ice prediction is challenging in the context of climate prediction models. Relative to the NCEP Climate Forecast System version 1 (CFSv1), one of the most important developments in the CFSv2 is the incorporation of a sea ice model component. Our evaluations suggested that although the CFSv2 captures the observed seasonal cycle and trend of Arctic sea ice to some extent, large errors exist. The most significant biases are sea ice too thick with interannual variability that is too weak. A major cause of the bias is lack of observations of sea ice thickness over broad areas of the Arctic that would aid in the forecast procedures. Another potential cause of the bias is that assumptions of parameterizations of sea ice optical properties currently made in the sea ice model component of the CFSv2 are inadequate to accurately simulate radiative interactions among atmosphere, sea ice and ocean as Arctic sea ice entering a new regime of thinner and predominantly first-year ice.
This project targets to advance understanding of Arctic sea ice interactions, enhancing seasonal predictability and prediction of Arctic sea ice, and northern mid- and high-latitude winter climate associated with rapid changes of Arctic sea ice in the CFSv2. This serves as an important incremental step toward achieving improved operational prediction system. The proposed work not only enhances seasonal sea ice predictions for the existing operational CFSv2, but can also be applied to the development of the next generation of the NCEP Climate Forecast System, which will include various upgrades. The following targeted activities provide a framework for our project:
1) Assimilate the newly available satellite-based sea ice thickness in the Arctic using a local singular evolutive interpolated Kalman filter, which provides initial conditions for the CFSv2.
2) Incorporate a prognostic model of melt ponds in the sea ice model component of the CFSv2, which allows for changing pond conditions, with implications for the ice-albedo feedback.
3) Implement a more incremental modification of the existing radiative transfer scheme used in the sea ice model component of the CFSv2, and integrate it with the melt pond model.
4) Conduct hindcasts/forecasts with multi ensemble members for 2003-2015, and investigate impacts of the assimilation of observed sea ice thickness, improved sea ice optical parameterizations, and the use of the latest global forecast system on sea ice predictions.
5) Analyze overall skill in forecasting sea ice, the capability in capturing the observed intraseasonal-to-interannual variations of sea ice, and the predictability of sea ice and its relationship with the internal variability in the fully coupled forecast system.
6) Investigate impacts of improved sea ice predictions on the overall skill in forecasting winter climate (including extremes) over northern mid- and high-latitudes in the CFSv2.
This project directly addresses the FY15 CVP Arctic focus to “develop a capability to skillfully and reliably predict variations and changes in Arctic sea ice on time scales of a few months to decades to improve our predictive capability and address the need for environmental information for informed decision making.” The proposed work is highly relevant to the goal of this competition “improve future operation predictions”, and leverages scientific advances by the research community external to NOAA’s operational climate centers and seeks to test and evaluate the potential of experimental models and analysis for operational use.