Rapid warming of Lake Superior
Now, here's the kicker: when you look at the date of the start of the stratified season as a function of that year's average ice cover (the top row in this figure; each dot represents a different year from 1979-2005), you see that the more ice cover you have, the later the start of the stratified season. In the same vein, the more ice you have over the winter, the cooler the following summer is going to be (the bottom row). By delaying the start of the stratified season, ice essentially shrinks the amount of time that the lake has to heat up in the spring and summer. In fact, winter ice cover is the primary indicator of how warm the lake will be the following summer- in other words, if you have a lot of ice, but a warm summer, the lake will still be cool that year.
Here's a figure that summarizes these relationships:
Here I've plotted the water temperature (averaged carefully between the three stations) and the regional air temperature on the top graph, and ice cover on the bottom. You can see that in warm air temperature years (i.e. 1983, 1998, or 2005) we get warm water; in low ice years (1983, 1987, 1995, 1998-2000, 2002) we get warm temperatures- if both are true (i.e. 1983, 1998) we get really warm temperatures. In cool years (i.e. 1992-1993) we get cool temperatures, and in high-ice years (i.e. 1979, 1994, 1996, 2003) we get cool water temperatures. It doesn't work perfectly- this is a massively complicated system. But it explains what sets water temperatures to first order.
To conclude, summer temperatures in Lake Superior (and Huron and Michigan, by the way) are increasing due to two separate but related trends: summer air temperatures are increasing, and winter ice cover is decreasing. Both of these effects add to produce the observed response of around a degree C per decade increase in Lake Superior water temperatures.
As it Happens, CBC (the first piece in the second half)
Here and Now, on Wisconsin Public TV (go to about 7:40)