Fun stuff with your graph photon:
In cooler months (the least amount of deviation), there is a spread of about 2.5 standard deviations. If we take 2.5 standard deviations out, the chances of one not being the other are about 2.5% aka 97.5% confidence that one isn't equal to the other. I think the general cut offs are at 95% or 90% though...
My point? None really. Just applying stuff I rarely get to use to kind of explain that graph a bit better. Though I have to stress that from that data alone, all you can state is that one isn't the other. The original data could have been from icier years or the new ones an aberation in data.
__________________
|