Long-term Goals: The overarching goals of this project are to understand the role of sea ice-albedo feedback on sea ice predictability, to improve how sea ice-albedo is modeled and how sea ice predictions are initialized, and then to evaluate how these improvements influence inherent sea ice predictability. Objectives: The sources of errors in a model forecast are from initial conditions and the model itself. Both can be evaluated with observations and potentially improved. We will use observations and field studies to improve how sea ice-albedo is modeled as much as possible. We will use methods to quantify feedback in models, and thereby directly relate feedback to predictability. We will use initial conditions from the model itself in idealized, perfect model studies, and from other models with data assimilation. Soon the modeling system we use will have its own sea ice data assimilation scheme (it has data assimilation in the atmosphere and ocean already) and we can investigate how model improvements influence the initialization procedure as well.