It's worth following up on a couple of points from last week's post.
First, to extend the comments on what happened in January, the Arctic as a whole did not see the widespread and unusual warmth that affected Alaska, although conditions were highly variable across the basin:
The European side was even warmer than the Alaskan sector, but much of inland Arctic Siberia was colder than normal. I say "inland" because the most northerly Russian sites around 90°E were much warmer relative to normal, illustrating the moderating influence of the ocean despite its ice cover.
The relationship between ENSO and winter temperatures is not as simple for the Arctic as a whole as it is for Alaska, because there are strong regional differences across the Arctic. The Arctic-wide temperature trend since 1980 is similar for both El Niño and La Niña, judging from the 32 sites I've been using for realtime monitoring:
Notice that of the three extremely warm winters of 2015-16 through 2017-18, one was an El Niño winter, one a La Niña, and the other one was neutral.
The neutral-ENSO trend is slightly less than the El Niño and La Niña trends, which is similar to the result I showed last time for Alaska. But let's discuss that result a bit more. It turns out that the neutral-ENSO temperature trend is very sensitive to the ENSO classification method, i.e. which index we use to determine whether a winter is El Niño, La Niña, or neutral. Last week I used the Multivariate ENSO Index, which is a comprehensive measure of ENSO's expression in several different variables. This was the result for Fairbanks:
In contrast, here's the analysis using the Southern Oscillation Index to define the occurrence of El Niño and La Niña. The SOI is a simple but venerable index based solely on pressure differences between Darwin (Australia) and Tahiti.
What a difference this makes to the neutral-ENSO trend! It now shows no significant discrepancy from the El Niño and La Niña trends, and all three are similar. The trends are also quite similar if we use data back to 1930 (the SOI is available back into the 19th century):
How could the ENSO classification method make such a large difference? Part of the answer is that these linear trend estimates are inherently uncertain (as I noted last time) because of the small sample size in each category, and the large year-to-year variance. And in particular, the classification of neutral-ENSO years is subject to considerable uncertainty, because the phase estimate can differ in either direction. In contrast, the stronger El Niño and La Niña episodes are unambiguously classified with more confidence.
As an example, it seems that three very warm winters in the first half of the MEI series were classified as neutral ENSO winters according to the MEI, and all three had a significantly positive PDO phase: 1969-70, 1976-77, and 1980-81. The warmth in these earlier years contributes to the lack of warming trend in the ENSO neutral category:
But according to the SOI, all three winters were marginally El Niño winters, and so they are excluded from the neutral-ENSO trend calculation.
Similar differences are found for the statewide Alaska temperatures, and remarkably the situation is now reversed with the neutral-ENSO trend, which is now the steepest of the three:
Another interesting point here is how much more shallow the long-term trend is when including the data prior to 1950. Starting the trend in 1950, as I did for the MEI classification, produces very steep trends because of the cold (negative PDO) era of the 1950s and 1960s.
In conclusion, there's obviously very large uncertainty around the neutral-ENSO trend, but the SOI classification supports the consistency in Alaska's warming trend for both La Niña and El Niño winters.