While I was in the middle of looking at various climate datasets to analyze for various things, I stumbled upon something curious in the Northern Hemisphere Snowcover data: what appears to be a sudden shift towards less snow in the late eighties, preceded and followed by fairly stable conditions before and after. What’s curious about this is that I haven’t seen this identified by anyone previously. As we shall see, the strangeness does not stop there. But first, a little bit on how I came across this interesting feature of the data:
The NH snowcover data is unfortunately missing several months in the earlier part of the data set. My solution for dealing with this was to identify the missing months in separate time series, and estimate the missing values based on the same months in surrounding years. This doesn’t really have an influence on the identification of the step shift, as the missing months are mainly in the late sixties, and the apparent shift is in the late eighties.
The next step was to remove the seasonal effects by taking a twelve month moving average of the complete time series. This is the result, with the average line for 1967-2010, here:
I have split the series at the point at which values go from predominantly above average to predominantly below average.
Curiously, this step shift, while quite apparent when the data are averaged in this way. But except for the summer seasonal graph, which is not available for some reason, it is not clear to me that such a step shift can be seen in the seasonal plots. The winter actually shows an increasing trend, and the spring trend might possibly look step-ish but appears more continuous after.