This calculation uses the ice motion vectors and the assumption that the age of a pixel is the area weighted average of the source pixels plus the time step. Basically I drew a map of the source pixels and offset the pixel area by the minus the motion vector. The source box will overlap a group of 4 pixels. The overlap area was used as the weight.
This wouild work if one knew the ice age at the start January 1, 1979. The spin up took about a decade for the computation to forget the initial condition. To make the first decade more realistic, I took the minimum age for the second decade as used that as the initial value in 1979 and repeated the computation.
For open water the ase was zero.
The ice age maps should be treated as qualitative.
One sees expected features like old ice north of Canada and younger ice in the 2010's than in 1980 to 2000.
G.G. Campbell NSIDC 6/2015 to 6/2019.
An alternative computation using the idea that sea ice is either new ice: that is ice for this freeze up season or old ice: more than 1 year is provided by Mark Tshudi, Walter Meier, Scott Stewart, Charles Fowler and Jim Maslanik at https://nsidc.org/data/nsidc-0611