Guest Post by Ira Glickstein
According to the latest from NASA GISS (Goddard Institute for Space Studies), 2010 is shaping up to be “the warmest of 131 years”, based on global data from January through November. They compare it to 2005 “2nd warmest of 131 years” and 1998 “5th warmest of 131 years”.
We won’t know until the December data is in. Even then, given the level of noise in the base data and the wiggle room in the analysis, each of which is about the same magnitude as the Global Warming they are trying to quantify, we may not know for several years. If ever. GISS seems to analyze the data for decades, if necessary, to get the right answer.
A case in point is the still ongoing race between 1934 and 1998 to be the hottest for US annual mean temperature, the subject of one of the emails released in January of this year by NASA GISS in response to a FOIA (Freedom of Information Act) request. The 2007 message from Dr. Makiko Sato to Dr. James Hansen traces the fascinating story of that hot competition. See the January WUWT and my contemporary graphic that was picked up by several websites at that time.
[My new graphic, shown here, reproduces Sato’s email text, including all seven data sets, some or all of which were posted to her website. Click image for a larger version.]
The Great Hot 1934 vs 1998 Race
1) Sato’s first report, dated July 1999, shows 1934 with an impressive lead of over half a degree (0.541ºC to be exact) above 1998.
Keep in mind that this is US-only data, gathered and analyzed by Americans. Therefore, there is no possibility of fudging by the CRU (Climategate Research Unit) at East Anglia, England, or bogus data from Russia, China, or some third-world country. (If there is any error, it was due to home-grown error-ists :^)
Also note that total Global Warming, over the past 131 years, has been, according to the IPCC, GISS and CRU, in the range of 0.7ºC to 0.8ºC. So, if 1934 was more than 0.5ºC warmer than 1998, that is quite a significant percentage of the total.
At the time of this analysis, July 1999, the 1998 data had been in hand for more than half a year. Nearly all of it was from the same reporting stations as previous years, so any adjustments for relocated stations or those impacted by nearby development would be minor. The 1934 data had been in hand for, well, 65 years (eligible to collect Social Security :^) so it had, presumably, been fully analyzed.
Based on this July 1999 analysis, if I was a betting man, I would have put my money on 1934 as a sure thing. However, that was not to be, as Sato’s email recounts.
Why? Well, given steadily rising CO2 levels, and the high warming sensitivity of virtually all climate models to CO2, it would have been, let us say inconvenient, for 1998 to have been bested by a hot golden oldie from over 60 years previous! Kind of like your great grandpa beating you in a foot race.
2) The year 2000 was a bad one for 1934. November 2000 analysis seems to have put it on a downhill ski slope that cooled it by nearly a fifth of a degree (-0.186ºC to be precise). On the other hand, it was a very good year for 1998, which, seemingly put on a ski lift, managed to warm up by nearly a quarter of a degree (+0.233ºC). That confirms the Theory of Conservation of Mass and Energy. In other words, if someone in your neighborhood goes on a diet and loses weight, someone else is bound to gain it.
OK, now the hot race is getting interesting, with 1998 only about an eighth of a degree (0.122ºC) behind 1934. I’m still rooting for 1934. How about you?
3) Further analysis in January 2001 confirmed the downward trend for 1934 (lost an additional 26th of a degree) and the upward movement of 1998 (gained an additional 21th of a degree), tightening the hot race to a 28th of a degree (0.036ºC).
Good news! 1934 is still in the lead, but not by much!
4) Sato’s analysis and reporting on the great 1934 vs 1998 race seems to have taken a hiatus between 2001 and 2006. When the cat’s away, the mice will play, and 1998 did exactly that. The January 2006 analysis has 1998 unexpectedly tumbling, losing over a quarter of a degree (-0.269ºC), and restoring 1934‘s lead to nearly a third of a degree (0.305ºC). Sato notes in her email “This is questionable, I may have kept some data which I was checking.” Absolutely, let us question the data! Question, question, question … until we get the right answer.
5) Time for another ski lift! January 2007 analysis boosts 1998 by nearly a third of a degree (+0.312ºC) and drops 1934 a tiny bit (-0.008ºC), putting 1998 in the lead by a bit (0.015ºC). Sato comments “This is only time we had 1998 warmer than 1934, but one [on?] web for 7 months.”
6) and 7) March and August 2007 analysis shows tiny adjustments. However, in what seems to be a photo finish, 1934 sneaks ahead of 1998, being warmer by a tiny amount (0.023ºC). So, hooray! 1934 wins and 1998 is second.
OOPS, the hot race continued after the FOIA email! I checked the tabular data at GISS Contiguous 48 U.S. Surface Air Temperature Anomaly (C) today and, guess what? Since the Sato FOIA email discussed above, GISS has continued their taxpayer-funded work on both 1998 and 1934. The Annual Mean for 1998 has increased to 1.32ºC, a gain of a bit over an 11th of a degree (+0.094ºC), while poor old 1934 has been beaten down to 1.2ºC., a loss of about a 20th of a degree (-0.049ºC). So, sad to say, 1934 has lost the hot race by about an eighth of a degree (0.12ºC). Tough loss for the old-timer.
Analysis of the Analysis
What does this all mean? Is this evidence of wrongdoing? Incompetence? Not necessarily. During my long career as a system engineer I dealt with several brilliant analysts, all absolutely honest and far more competent than me in statistical processes. Yet, they sometimes produced troubling estimates, often due to poor assumptions.
In one case, prior to the availability of GPS, I needed a performance estimate for a Doppler-Inertial navigation system. They computed a number about 20% to 30% worse than I expected. In those days, I was a bit of a hot head, so I stormed over and shouted at them. A day later I had a revised estimate, 20% to 30% better than I had expected. My conclusion? It was my fault entirely. I had shouted too loudly! So, I went back and sweetly asked them to try again. This time they came in near my expectations and that was the value we promised to our customer.
Why had they been off? Well, as you may know, an inertial system is very stable, but it drifts back and forth on an 84 minute cycle (the period of a pendulum the length of the radius of the Earth). A Doppler radar does not drift, but it is noisy and may give erroneous results over smooth surfaces such as water and grass. The analysts had designed a Kalman filter that modeled the error characteristics to achieve a net result that was considerably better than either the inertial or the Doppler alone. To estimate performance they needed to assume the operating conditions, including how well the inertial system had been initialized prior to take off, and the terrain conditions for the Doppler. Change assumptions, change the results.
Is 2010 going to be declared warmest global annual by GISS after the December data comes in? I would not bet against that. As we have seen, they keep questioning and analyzing the data until they get the right answers. But, whatever they declare, should we believe it? What do you think?
Figuring out the warmest US annual is a lot simpler. Although I (and probably you) think 1934 was warmer than 1998, it seems someone at GISS, who knows how to shout loudly, does not think so. These things happen and, as I revealed above, I myself have been guilty of shouting at analysts. But, I corrected my error, and I was not asking all the governments of the world to wreck their economies on the basis of the results.