One of the statistical tools that the Climate Prediction Center uses for monthly and longer period forecasting is called the "Optimum Climate Normals", which is simply how temperatures and precipitation have averaged in the past 10 years (15 for precipitation) compared to normal. This is potentially important because most longer term forecasts are made in reference to some "normal".
Here's a plot Fairbanks 1981-2010 normal monthly temperatures along with how the past 10 years (Mar 2003-Feb 2012) compare. Due to the very short period (10 years), I've included the decadel mean and median, which in some cases (e.g. January and October) are substantially different; this is an artificant that one extreme value can have on the mean in such a small sample.
How does this look if we standardize the differences (multiplied by 10 so scale is the same as the first chart), so that we account for the fact that temperatures have much wider ranges in winter than summer:
So now we see that the decline in January temperatures is not so spectacular (but March still is), but the recent warmth in May and August through October really stands out. This is a nice illustration of the power of standarization when the there are big differences in the variance in the data through the year.
Thanks for the interesting analysis. So, colder winters and warmer spring and autumn recently, with a notably sharper transition to deep winter from October to November.ReplyDelete
One wonders about the causes... perhaps the diminished Arctic sea ice is related to the autumn warmth and the negative PDO phase to the winter cold. Would be interesting to examine maps of the monthly changes to see how localized or extensive the differences are. My guess is, more autumn warmth in the north, more winter cold in the interior, west and south.
I believe you spot on for causes. I'm working on extending this analysis for other locations this weekend. Thus far, it looks like what you would expect with the decline in sea ice and a negative PDO.
Rick, a couple of years ago a ran an analysis of monthly temperature anamolies against various telleconnection indices. Interestingly, the month with the strongest correlations was February. Anecdotally, it seems like long term blocking patterns have set up in January and March in most of the prior 5 years. Coincidentally, the patterns have shifted with the changing of the calendar - which has the practical effect of magnifying the disparities.ReplyDelete
Some of the March "dip" may simply be "reversion to the mean", as 1981-2010 has at least five of the warmest Marches of record in the sample.Delete