Assessment of satellite-observed sea ice concentration and extent uncertainty estimates

Sea ice concentration and extent estimates are available from a variety of sources, primarily based on passive microwave imagery. These satellite-derived fields provide both a long history of sea ice conditions (since 1978) and near-real-time information. Thus they are valuable inputs to model simulations as validation, initialization, and assimilation. For models to best take advantage of the observed fields, uncertainty estimates are essential. However, until recently such uncertainty estimates were not routinely produced; concentration uncertainty was based on general ice regimes (e.g., cold winter, summer melt, thin ice) based on limited validation studies with high-resolution comparison imagery. Extent uncertainty has only been estimated in a heuristic manner or by comparison of different algorithms. Concentration uncertainty estimates have now been developed in a few products, but they have yet to be thoroughly evaluated. How consistent is the derived uncertainty with the true uncertainty at a given location? Here we investigate concentration uncertainty via comparison with high resolution Landsat-8 and MODIS imagery. There is no comparable total extent validation product, so we estimate uncertainty via comparison of various algorithms as well as attempt to estimate an internal relative uncertainty based on variations in input data and algorithm methodology.