Found in the Data is Beautiful subreddit, where there is some discussion.
I thought of sending this to a friend of mine who is a climate change skeptic, but I know his response will be that the changes reflect 1) instrumentation error and 2) normal cyclic variation. *sigh*