The method used here to estimate sea ice age is one that involves tracking the sea ice from year-to-year using gridded ice motion vectors; this is called Lagrangian tracking. Using 12.5 km x 12.5 km EASE-Grid vector fields for 1979 onward, ice age can be estimated by treating each grid cell that contains ice as a discrete, independent Lagrangian parcel and tracking the parcels at weekly time steps. The process can be viewed as a set of stacked planes overlying the grid used, with each plane corresponding to an age category. Parcels move around on their respective planes, independent of parcels of other age categories, which in turn lie in their own planes. Then, to produce maps of ice age, the set of parcels for each weekly time increment is rasterized by assigning parcels to the 12.5 km x 12.5 km grid cell within which each parcel's position lies. In cases where parcels of different ages fall within a single grid cell, the age of the grid cell is assigned to the oldest parcel. Physically, this is based on the assumption that younger ice will deform more easily, such that more of the area within the grid cell would be covered by the older ice type. For example, if two parcels, one that represents first-year ice and one that represents third-year ice, both fall within the domain of a single grid cell, then the age of that cell will be assigned as third-year ice. If the passive microwave-derived ice concentration at the corresponding grid cell remains at least 15 percent throughout the melt season, then that parcel is assumed to have survived the summer, and the parcel's age is incremented by one year. It is important to note that while grid cells with less than 15 percent concentration are treated as open water in terms of the age product generation, the cells could still contain some ice.

The sea ice age in this data set is estimated using input gridded ice motion vector fields. Each grid cell that contains ice at a concentration of greater than 15 percent, as derived by the NASA Team Algorithm (Cavalieri et al. 1984), is treated as a discrete, independent Lagrangian parcel and is transported at weekly time steps (Maslanik et al. 2011). In this approach, the actual age of the ice is followed for each ice parcel, and a parcel is categorized as first-year ice, second-year ice, and so forth based on how many summer melt seasons the ice parcel survives (Tschudi et al. 2010). If the grid cell remains at or above 15 percent throughout the melt season, then that parcel is assumed to have survived the summer, and the age of the parcel is incremented by one year. The 15 percent value was used here because it captures greater detail within the marginal ice zone and is the more conservative approach in terms of assessing net loss of areas where some multiyear ice is present, where multiyear ice is ice that is second-year ice or older. If a grid cell contains ice of different ages, the oldest ice determines the grid cell's ice age value (Maslanik et al. 2011).

The basic procedure for estimating sea ice age is by advecting ice with the weekly sea ice motion vectors. This is done by first dividing the Arctic region into a 12.5 km grid. Each grid cell, treated like an independent Lagrangian parcel, is advected at each time step using the weekly mean velocity fields. If the ice survives from the minimum sea ice extent, typically in September, of one year to the minimum extent of the next year, then it is aged one year. This process is continued for each year of the data record (Fowler et al. 2004).

Processing Steps

• Input ice motion data, Polar Pathfinder Daily 25 km EASE-Grid Sea Ice Motion Vectors, Version 3, are converted to 12.5 km x 12.5 km EASE-Grid.
• The daily ice motion vectors are averaged into weekly ice motion estimates on the 12.5km EASE grid.
• Ice is tracked in yearly increments, where a year is the melt season which runs from one season's minimum Arctic ice extent (usually in September) to the next year's minimum.
• Each year's ice is tracked from year to year as a Lagrangian tracer parcel that starts at the center of each grid cell and moves according to the weekly mean ice velocity.
• If a parcel remains at 15 percent or more for a melt season, then it is aged one year. If a parcel travels to a grid cell that has less than 15 percent ice concentration, the tracer parcel is assumed to have melted away.
• The age of a grid cell is the age of the oldest tracer parcel that exists in the grid cell. 

Quality Assessment

A 15 percent ice concentration threshold was chosen since the intent is to be as conservative as possible regarding changes in areas where multiyear ice is present. For example, at the end of summer melt, a grid cell within the marginal ice zone might have a total passive microwave-derived concentration of 15 percent. Even though, upon freeze-up, 85 percent of the grid cell would consist of first-year ice, the age of that grid cell is assigned to the oldest ice that survived within that grid cell. Hence, the maps indicate the coverage of areas that contain at least some, 15 percent or more, multiyear ice but do not provide information on proportions of ice of different ages within individual grid cells.

In the past, sea-ice-cover studies have been based on passive microwave data computed from algorithms that employ multiple microwave frequencies to estimate total ice concentration and multiyear ice. Although these algorithms are good estimates of these parameters, they are hindered by secondary factors that change microwave emission in ways that adversely affect these estimates of ice concentration and type. The ice age data in this data set are calculated using an alternative technique of estimating ice age where ice ages are determined primarily based on transport calculated from ice velocities and are not affected by the types of error sources that are present in passive microwave ice concentration estimates. In instances where ice concentration is used, a conservative threshold is chosen to delineate ice extent. Thus, this technique is fundamentally independent of the factors affecting studies that rely on passive microwave estimates of ice concentration or ice age (Fowler et al. 2004).

Error Sources and Limitations

When age classes are aggregated into first-year and multiyear ice categories, the information is comparable to the passive and active microwave satellite-derived time series of first-year and multiyear ice analyzed by previous studies. Overall, this remote sensing-based age product is similar in nature and information content to the buoy-derived age fields produced by Rigor and Wallace (2004), but with greater spatial detail. The age estimates are restricted to open ocean areas only, where ice motion can be resolved in the microwave data. Note that this excludes the passages in the Canadian Archipelago. The cited values for ice coverage are therefore less than the actual amount of ice present in the Arctic.

Errors in the method of estimating sea ice age are dependent on the following ice motion errors:

• Resolution of the satellite sensor
• Geolocation and binning errors of each image pixel
• Atmospheric effects and temporal variability of the surface, especially during the summer months

Version History