Arctic sea ice signatures for passive microwave algorithms

Passive microwave satellite data can be used to estimate the concentrations of three surface types: open water, first-year ice, and multiyear ice. These estimates are made with an algorithm that requires knowing the microwave signatures of each of these three surface types. For the Scanning Multichannel Microwave Radiometer, the algorithm proposed by Cavalieri et al. (1984), known as the SMMR Team algorithm, is widely used. Two unsatisfying properties of this algorithm are (1) its estimates of multiyear ice concentrations are much lower than its estimates of summer ice concentrations and are also significantly lower than independent estimates of Arctic Ocean multiyear ice, and (2) it produces variations in estimated concentration that appear to be caused by spatial and temporal variations in the signatures. We examine two alternative sets of signatures. One set varies temporally but not spatially and is defined by the modes of major peaks in histograms of brightness temperature. The other set is defined as the overall means of the temporally varying signatures; we call these “new constant signatures.” An estimate of the average bias of the winter multiyear ice is derived and a corresponding correction applied to the new constant signatures. Both sets of new signatures have essentially the same precision: 0.05 (fractional area) for winter ice concentration, 0.10 for summer ice concentration, and 0.15 to 0.3 for winter multiyear ice concentration. These error estimates are upper bounds to the true precision, especially for multiyear ice for which some true concentration changes are included in the variability. The precision estimates include all spatial and temporal signature variability, and are compatible with estimates of sea ice emissivity variability given in the literature. Because the temporally varying signatures provide only a slightly better precision, we prefer the new constant signatures. Both sets of new signatures result in about 0.6 to 0.7 winter multiyear ice for the Arctic Ocean, substantially more than the 0.3 to 0.4 computed using the original Team algorithm signatures, and in the same range as several independent estimates of the multiyear ice, 0.54 to 0.75.