Signal to Noise

We keep hearing that the "Science is Settled", that the earth is warming, and has warmed by a degree or so over the last hundred years. This is usually presented in a breathless zomg-we're-all-gonna-die report of impending Thermageddon.

OK, whatever.

Willis Eschenbach (seen before here at Borepatch) started looking at the way the data is presented, and wondered why it was shown as year-to-year. Why not the months of the year? Does it tell the story a different way if you do it that way?

As it turns out, it does indeed:

NORDKLIM is a climate database compiled by the governments of the Scandinavian region, and included in the major global data sets (GHCN, etc). Each decade during the 20th Century has a different color line, showing the average temperature by month for each month of the year. What can we see from this display?

Well, we see that the differences are much more pronounced in the colder months (January is month 1, February is month 2), and much less pronounced in other months. So the "annual average" mostly shows changes from warmer winters, not warmer summers. Look at months 6, 7, and 8 - the warmest decade was 1930-1939. January and February were much warmer in 1990-1999.

OK, so what? Well, the "Urban Heat Island"* is known to be much more pronounced during colder months than warmer months. Eschenbach explains:

The NORDKLIM notes say:

Especially one should notice that stations represent local conditions, which may have been effected e.g. by urbanisation

This effect is known to be greater in winter than in summer. In a study done in Barrow, Alaska, for example, there is a 4.5°C difference in the UHI effect between January and July. The winter to summer difference in the UHI in Fairbanks, Alaska is estimated to be 1.2°C.

Let's put that in perspective: the entire warming for the lower 48 states between 1940 and 1999 (as reported by the GHCN data set) is 0.6°C. The annual error due to UHI seems to be between tw0 and seven times that amount.

Let me restate that in engineering terms: the noise is between two and seven times the signal that we're trying to detect. Anyone who's tried to tune in a staticy AM radio broadcast of your favorite baseball team knows what that means - you can't tell what's happening in the game.

* The tendency of cities to be warmer than the surrounding countryside, due to paved roads, parking lots, and building absorbing the sun's heat. Energy use in the city is also a major contributor, especially in the winter. Houses heated to 70° radiate a lot of heat when it's 0° outside. Rural areas don't have much of either of these (roads or houses), and so don't suffer from an UHI effect.