Wouldn't this graph be more informative if it also included a line
showing the season of maximum extent (over eg 1979-2000), ie to give a
clearer indication of the extent in a really cold recent winter.

As presently drawn, any small inflection in the direction of increased
extent looks like it's tending towards the upper bound of the extent
whereas in reality it's just tending towards - although certainly not
reaching, as yet - the mean. Or, put another way, if you're going to
include 2006-7 as a lower bound, surely it's logical also to include
an upper bound as well as the mean.