The areal coverage of the Polar Oceans with sea ice, i.e. ice of marine origin, is described by the sea-ice concentration: the fraction of a known ocean surface area covered with sea ice. It is either provided in percent or as a fraction between 0 and 1. Sea ice area and extent, two well-known climate indicators, are derived from the sea-ice concentration. The most common way to compute the sea-ice concentration has been to utilize brightness temperature observations made by polar orbiting satellite microwave (1-100 GHz) radiometers, permitting a spatially almost complete, daylight and (mostly) cloud-cover independent coverage of both polar hemispheres every day. Depending on frequency and sensor, one single observation represents a spatial scale between 5 km and 70 km. Over time about 15-20 different sea-ice concentration retrieval algorithms have been developed and their results evaluated using different kinds of independent observations – mostly also from satellite sensors. On the way from the first algorithm to today’s suite of algorithms, many improvements made their way into the processing chains and gridded sea-ice concentration products. However, there are unsolved challenges and insufficiently specified uncertainties, for instance: 1) The accuracy of, basically, all sea-ice concentration retrieval algorithms is poor during summer because of the presence of melt water on top of the sea ice. It is still a challenge to retrieve sea-ice concentrations during summer with the same accuracy as during winter, and to evaluate these with independent, highest quality sea-ice concentration estimates that often are as difficult to obtain. 2) Elements already considered in sea-ice concentration retrieval uncertainties are sensor noise, tie point uncertainty, and uncertainties resulting from the gridding process. However, spatiotemporal correlations of the input data, and how these translate into correlations in the derived sea-ice concentration and the retrieval error are not well known. The same applies to how these errors and their correlations propagate into uncertainties in sea-ice area and extent (see contribution by Wernecke). 3) Most algorithms fold sea-ice concentrations below 0 % / above 100 % back into the range 0 % to 100 % even though the retrieval using tie points that represent these two end members of the sea-ice concentration results in a natural distribution of the retrieved sea-ice concentration values around these end members. Not considering this natural retrieval-induced variability creates biases in evaluation results at and reported accuracies for these particularly interesting end members of the sea-ice concentration distribution.
Topic : Theme 1: Cryosphere monitoring.
Reference : T1-E12
Back to the list of submissions
Previous submission · Next submission
Comments are only accessible to participants. If you are a participant, please log in to see them.