The SMMR and SSM/I instruments are dual-polarized and multi-frequency microwave radiometers that sense emitted microwave radiation at an incidence angle of about 53° and an altitude of about 800 km. At microwave frequencies, the observed radiance varies linearly with the temperature of the emitting material, according to the Rayleigh-Jeans law, and is usually expressed in terms of brightness temperature. The constant of proportionality in the linear relationship is the emissivity, which provides information about the electrical and physical properties of the material itself. The observed brightness temperatures come primarily from the Earth's surface with some contributions from the atmosphere and outer space. Depending on frequency and polarization, the brightness temperature is affected by changing surface, subsurface, and atmospheric conditions. The multichannel capability allows for the discrimination of different surfaces and atmospheric effects and allows for the development of algorithms that provide a range of geophysical information about the surface.

Sea ice concentrations for this data set were derived from a revised Bootstrap algorithm which uses a different set of tie-points and adjustments. For algorithm background, please see the Enhanced Sea Ice Concentrations from Passive Microwave Data document (Comiso and Nishio 2008). For changes made to this Version 3 data set, see the Positive Trend in the Antarctic Sea Ice Cover and Associated Changes in Surface Temperature document (Comiso, et al. 2017).

Processing Steps

Interpolation

Data were interpolated in areas with missing pixels according to the following steps. First, data were spatially interpolated only for isolated empty pixels. An empty pixel was replaced by the average of four good surrounding pixels, or if four good pixels were not available, then a smaller number of pixels was selected. Second, a time interpolation was applied to the spatially interpolated map. Time interpolation was based on a weighting scheme; the closer the good data were in time, the higher the weighted value. For each empty pixel, the algorithm searched forward in time for a good pixel, and backward in time for a second good pixel. The algorithm determined how many days ahead and behind the two good pixels occurred and calculated the weight of each pixel. A weighted average of the two good pixels was then calculated, and the result was used for the empty pixel. During the SSM/I period, most temporal interpolations were conducted using only adjacent days. During the SMMR period, particularly in 1986 when larger gaps were present, temporal interpolations had separations of more days.

The staff at the NSIDC DAAC performed an evaluation between Version 2 and Version 3 data to verify any difference between the two versions. Here are their findings:

• The changes introduced in Version 3 have some effect on the total ice covered area and extent.  These differences have some seasonality and appear to be largely due to the use of dynamic open water tie points.
• The majority of pixels show no change in concentration value or less than one percent, although a small fraction have larger differences.  Roughly two percent of pixels show concentration changes of more than 10 percent.

Although most bad data are easily identifiable and removed with little ambiguity, human judgments must be made and applied consistently to the time series data. While efforts were made to minimize errors in judgment and inconsistencies, the resulting data set is not perfect. Nevertheless, the data provider feels that these maps are suitable for some types of research studies, such as basin-scale trend analysis, model inputs, and some regional studies. They are less useful in evaluating conditions on a specific day in a localized region.

NSIDC staff visually checked the entire set of data files, including file structure, comparisons to existing SMMR- and SSM/I-derived sea ice concentration grids, and information files and examination of data quality.

Errors and Limitations

Sensitivity analysis of potential errors associated with the retrieval of sea ice products were discussed previously. For example, see Comiso et al. 1997 and other references associated with the NASA Team Algorithm sea ice products. Sea ice concentration errors are difficult to quantify because sea ice is an evolving ice cover and does not have uniform physical and radiative characteristics throughout the ice pack. The emissivity of sea ice is known to change as the ice cover develops from grease ice to nilas, pancake to young ice, or first year ice to multiyear ice. For definitions of these types of ice, see the NSIDC glossary page. The line AD in the scatter plot that provides the sea ice tie-point in the Bootstrap Algorithm corresponds to 100 percent ice for the relatively thick ice types such as first year or multiyear ice. Areas with 100 percent of thinner ice types are often estimated to have ice concentrations as low as 80 percent due to lower emissivities (Comiso et al. 1992). In the summer, retrievals in the Arctic could have biases caused by meltponding effects (Comiso and Kwok 1996).

Inside the ice pack, observations from helicopter and aircraft flights typically show very high sea ice concentrations with some 5 to 10 percent of the leads covered by nilas or pancake ice. The frequency distribution of sea ice concentration in highly consolidated ice areas has a standard deviation of about 3 percent, which includes sea ice cover with concentrations of less than 100 percent. Overall, retrieval accuracy is estimated to be approximately 5 to 10 percent except in some unusual cases, such as the presence of a large fraction of thin ice or meltponding within the pixel and the presence of stormy weather conditions, especially near the ice edge. Limited comparative analysis with high resolution instruments and other measurements confirm this, but more extensive validation in a greater number of places over all of the seasons would provide more accurate error assessments.

For information describing the rough error/differences between each sensor (F8, F11, F13, F17) during the sensor overlap time period, refer to the following documentation: Comiso et al. 2008, Comiso et al. 2009, and Comiso et al. 2010.

Basic limitations arise from the sensor resolution, temporal coverage, and algorithm assumptions and characteristics. Users should review the information provided on fields of view, temporal sampling, and algorithm characteristics.

Particular care is needed in interpreting the sea ice concentrations during summer when melt is underway and in regions where new sea ice makes up a substantial part of the sea ice cover. As noted, some residual errors remain due to sensor differences and to weather effects along with mixing of ocean and land area within the sensor field of view.

No data coverage is available for regions poleward of 84.5° N for SMMR and 87.2° N for SSM/I or SSMIS due to the inclination of the orbits. SMMR data were acquired every other day, while SSM/I data were acquired daily. SSMIS data are also acquired daily.