Saturday, April 14, 2018

April Temperature Variance

Fairbanks saw its first 50°F temperature of the spring on Wednesday, and this was right on schedule according to the long-term history.  Based on data since 1930, the median date is April 11, but the first 50°F has typically arrived a few days earlier in recent decades.

Here's a chart of average daily high temperatures in each April since 1930 (blue markers) along with the date of the first 50+°F temperature (red markers).  Note that I've excluded 3 instances when 50°F occurred prior to March (1943, 1981, and 2009); these were exceptional winter chinook events, and for today's analysis I'm interested in what happens during spring.


The month of April has seen pronounced long-term warming in Fairbanks, with the linear trend line showing more than 5°F of warming over 87 years.  It's no surprise, then, to see that the first 50°F has become slightly earlier - although there's only a weak trend in the date of first 50°F.  The trend in April mean high temperature is highly significant (p<0.01), but the first 50°F trend is not significant at all (p~0.20).

At first glance it's surprising that the first 50°F trend is so much less robust, but this is partly because the variance of 50°F dates is so large.  Over the 87 years, the first 50°F date has advanced about 4-5 days, but the standard deviation is about 3 times as large (14 days).  In contrast, the interannual standard deviation of mean April high temperatures is almost exactly the same as the 5-6°F of warming that has occurred, so the long-term change has produced a very significant shift in the frequency distribution of mean temperature.

If we consider the normal rate of warming at this time of year in Fairbanks, we find that 5-6°F of warming is equivalent to about 8 days of climatological warming.  Naively, then, we might have expected the first 50°F to have advanced about 8 days; but even that would not correspond to a statistically significant trend in the date.  The fact is that the variance of the first 50°F date is so large that sampling variability could plausibly produce trends that are nearly zero or trends that are much greater than expected.  And so we might say that dates of "first warm day" or "first cold day", at least in Fairbanks, are not particularly useful as indicators of long-term change.

1 comment: