Advancing understanding of sea ice predictability with sea ice data assimilation in a fully-coupled model with improved region-scale metrics

Predictions of sea ice on subseasonal to interannual timescales has the potential to be of widespread value if they are skillful at the lead times and spatial scales needed by forecast users. Understanding sea ice predictability is needed for high-stakes decision-making, such as arises in shipping, accessing resources, and protecting Arctic communities. Current prediction efforts have focused mainly on predicting total northern hemisphere sea ice extent (SIE), termed pan-Arctic SIE. To succeed at predicting regional scales requires significant new effort in three key areas. First, data assimilation techniques must be advanced to accurately initialize sea ice and other components at proper spatial scales. Second, metrics are needed to quantify the skill at the relevant spatial scales and for patterns of interest. Identifying key metrics is motivated by the expectation that a forecast system can’t be improved without first developing adequate metrics for evaluating the features of importance. And third, effective statistical post processing methods are needed to correct for systematic biases in the resulting forecasts and to compute forecast probability.

We propose to investigate methods and develop the tools needed to address these three issues in building successful forecast systems. We propose to conduct our research in a well- studied, state-of-the-art sea ice component that is part of a global climate model. To turn this global model into a premier sea ice forecast system, we will work with Jeffrey Anderson and Nancy Collins and the NCAR Data Assimilation Research Testbed to implement DART to assimilate sea ice observations.

With this data assimilating forecast system, we will develop new evaluation metrics to investigate which observations are most essential among in situ measurements (including buoy and ship-based data) and remote sensing. We plan to investigate which regions are most predictable and what mechanisms (including mechanisms that involve coupling between ice, ocean and atmosphere) are responsible. Another important part of our project is to compare predictability in our system to others. We will undertake this with our links to the Sea Ice Outlook project and by providing our research on new metrics to evaluate regional patterns to other modeling centers for detailed intercomparisons. We have plans to collaborate directly with Rym Msadek and colleagues at GFDL to undertake a detailed comparison between the two premier U.S. global climate models, which have the most advanced sea ice components and high fidelity in the Arctic Ocean and atmosphere simulations. We also have discussed collaborating with Pablo Clemente-Colon, Chief Scientists at the National Ice Center, to better address sea ice forecast users needs in the metrics of local and regional-scale sea ice that we develop.

Our project has direct relevance to NOAA CVP Competition by exploring the value of assimilating sea ice observations, developing metrics that evaluate spatial distributions relevant to sea ice, and investigating mechanisms of regional sea ice variations. Our project is aligned with NOAAs goal of improving future operational predictions on time scales of a few months to decades. Our system will be capable of informing future data acquisition.