I regularly show climate trends over 50 years because it’s a round number, it’s a long enough time to allow for robust statistics and it encompasses most of a typical adult human lifetime’s memory. Obviously, other choices are perfectly reasonable.
Occasionally something unexpected shows up in these climate change maps. A good example is in the August temperature trend for Alaska and vicinity. While much of southern Alaska, northwest Canada and all the ocean areas show warming, a large portion of northwest and western Alaska, which have some of the largest annual temperature increases, show a decrease in average August temperature since the early 1970s
This kind of “unexpected result” is an invitation to dig deeper. For time series data, an obvious place to start is to inspect the data underlying the analysis. Here I’ll focus on the area that shows the largest decrease, basically the inland North Slope and Brooks Range (outlined in Fig. 1) from the the Dalton Highway west toward but not to the Chukchi Sea coast. Although this looks like a rather small area, at almost 89,000 km² it’s about the same land areas as the state of Maine. Figure 2 shows the August average temperature in this region each of the past 50 years and shows us immediately why there is a downward temperature trend: four of the five warmest Augusts in the past 50 years occurred between 1974 and 1979. We immediately ask if this is reasonable. Could this be an artifact of ERA5 reanalysis ? It turns out that the answer is “yes, this is reasonable.” While the inland North Slope and Brooks Range have hardly any place-specific climate observations, we are fortunate that the climate site at Umiat (69.4N, 152.1W), in the northeast portion of the area, was reestablished in the fall of 19751 and confirms that these were indeed very warm months.
However, if we extend the time series back into the 1950s, as in Fig. 3, we see that the 1970s were quite unusual and in this longer time perspective there no significant change in August average temperature. However, a feature that does stand in this longer time span is the relatively high year-to-year variability in August average temperatures from the late 1960s into the early 1980s (coldest August 1969, warmest August 1977), but there’s been very limited variability in the past decade.
Modern statistical methods allow for us to model temporal change (if any) in variability as well as the average. In fig. 4 we show the simultaneous estimates of both the average (mean) and the variability (standard deviation) since 1957 for the inland North Slope August average temperature. Again we see that the mean has hardly changed but the variability has decreased by more than 50 percent (5.3F in 1957 to 2.7F in 2022). Why would this be? One obvious (but untested) possibility would be the fact August precipitation is also trending upward, potentially limiting the range of temperatures.
Technical details:
Technical details: Temperature and precipitation analysis spatial maps from ERA5 reanalysis data from ECMWF/Copernicus.
The time varying mean and standard deviation shown in figure 4 generated from a Gaussian regression from the R package gamlss.
Umiat was built by the US Navy in 1945 as an aircraft resupply stop. Weather and climate observations were made by the Navy, the US Weather Bureau and the US Air Force between 1945 and 1954. Weather observations were made irregularly between 1954 and 1975, when regularly scheduled observations resumed (and continued until 2001).