Atmospheric effects on marginal ice zone sea ice concentrations from passive microwave algorithms

Accurate retrieval of sea ice concentrations in the Arctic is of importance for operational activities and for the detection of climatological changes in the ice cover. Passive microwave imagery from the Special Sensor Microwave/Imager (SSM/I) can effectively obtain useful ice concentration information. Over the past several years, several concentration algorithms have been developed using various methodologies and combinations of SSM/I channels. However, remote sensing of sea ice is limited by the complexity of the surface and uncertainties caused by atmospheric contamination.

These algorithms use the differing emissive properties of open water and various features of sea ice to detect ice and to discriminate among ice types. The selection of parameters for each algorithm requires a compromise because there is not enough information from the SSM/I channels to fully resolve all ice features of interest. For example, the NASA Team algorithm is useful in the central ice pack and can discriminate first-year ice from multiyear ice, but it significantly underestimates thin ice concentrations. The Cal/Val algorithm is adept at detecting thin ice, but yields little information on variability within the ice pack. These two algorithms, as well as several others, are compared in this study to determine strengths and weaknesses of each. Hybrid algorithms (combinations of two or more algorithms) are also tested. Hybrids compensate for weaknesses in one algorithm by including information from another algorithm that performs better in a given region.

This study focuses primarily on the marginal ice zone, where new ice growth occurs and thin ice dominates. These regions are crucial for operational analysis and forecasting and are also regions of high climatological variability. Unfortunately, ice concentration algorithms tend to perform poorly in these regions. Emission from the atmosphere, especially from thick clouds and precipitation, can contaminate the surface signal, resulting in errors in the algorithm. Many algorithms use simple threshold techniques to filter out weather effects. However, more recent studies indicate that more sophisticated treatments of the atmosphere - using remotely sensed atmospheric sounding data, meteorological station data, or radiative transfer models - can more effectively eliminate atmospheric contamination. Here, a variety of passive microwave sea ice concentration algorithms, including new hybrid algorithms, are tested in thin ice regions; the effects of atmospheric contamination on the algorithms are evaluated and methods to correct for it are examined. These comparisons elucidate differences in the algorithms and indicate situations where the algorithms perform poorly. The use of a combination of algorithms with better atmospheric information improves sea ice concentration retrievals in the marginal ice zone.