Note: After writing a draft of this entry, I went to the Real Climate website (I should have thought of it first; thanks Fraser for the tip!) and they already have a post up on this topic. It looks like we essentially agree on this issue. I may be an astronomer, but I know data analysis! Anyway see links at Real Climate for more on this story. I have edited my entry a bot, but it's essentially what I drafted up originally.
Via Gia's blog, I have learned that some scientists are claiming that they have found that the warming trend in the last half of the 20th Century is wrong. If so, this is a major revision of climate change science. However, I don't think this is correct. Basically, the claim is that the numbers published by NASA scientist James Hansen had an error in them. When this is corrected, the warming trend in the last part of the 20th century disappears. But let's look closer at the numbers. The table below is from the Watt's Up With That blog, which discusses these claims. The first table lists the 10 hottest years on record, ordered using the new data. The columns are the year, the deviation from average temperature done the old way, and the deviation done the new way. Top 10 GISS U.S. Temperature deviation (deg C) in New Order 8/7/2007
Interestingly, the hottest year on record before was 1998 (1.24 degrees Celsius above average), but with the new data, it's now second to 1934. That's interesting! But hang on a sec. The difference is pretty small. 1998 went down by 0.01 degrees, and 1934 went up by 0.02. I would call that a wash... except that the error bars are not published where I could find them. Is an 0.01 change significant? What if the measurements are only good to 0.1 degrees? Then 1934 and 1998 are in a statistical dead heat (haha). Posting the numbers like this doesn't actually tell you what you need to know. Worse, some misleading claims have been made about them. For example, the Daily Tech site says:
YearOldNew
19341.231.25
19981.241.23
19211.121.15
20061.231.13
19311.081.08
19990.940.93
19530.910.90
19900.880.87
19380.850.86
19390.840.85
In fact, 5 of the 10 warmest years on record now all occur before World War II.
(Psssstt! Someone tell that guy that WWII started in 1939). But even if we take him on his word, what he said is still misleading! Look at the dates, and think about it. These data are culled from a table that goes back from 1880 to 2006, a 127 year interval. If there were no warming trend at all, you'd expect the hottest 10 years to be randomly distributed in that 127 range, or roughly 1 hottest year in the top ten every 12.7 years (call it 13). But look at the data! In the last 13 year bin alone, from 1994 - 2006, we see 3 of the Top Ten years (2006, 1998, 1999). Another was in the second to last bin (1990). Then there is nothing before that until we go back to 1953. The odds of this are pretty low, though not totally precluded by chance. If you look at the data, you'll see four of the years are from 1931 - 1939 (which to be fair is more than one bin wide), so the distribution is very lumpy indeed. It's not easy to produce a trend from this data. But having three of the top ten years in the last bin strikes me as being very interesting to say the least. In fact, if you get the actual data and plot it even using the new numbers, you get this:
The x-axis is year since 1880, and the y-axis is the temperature deviation from average. You see two big bumps, the one in the 1930s and the recent one since 1980 or so. Curious, I looked at the average deviation for 1930 to 1942, to encompass that bump, and got +0.44 degrees (if I go only to 1941 I get +0.47, which is a bit higher since 1942 was a cooler than average year, but not including it would be cherry-picking, so I left it in). For the 1994 to 2006 bin I get a deviation of +0.57. The average temperature in the second bump is 0.1 degrees higher than the first. In other words, it was warmer on average in the last 13 years than any time in recorded history. Look again at the numbers. If you look at the top ten years, even with the new method, only the year of 2001 goes away (replaced with 1939, see chart below). All the other years are fairly stable, though the numbers themselves have changed. In other words, old method or new, more record years have happened in the last decade or so of the record keeping than you'd expect.
Here's the old order of top 10 yearly temperatures.
YearOldNew
19981.241.23
19341.231.25
20061.231.13
19211.121.15
19311.081.08
19990.940.93
19530.910.90
20010.900.76
19900.880.87
19380.850.86
People are using this data to say that global warming is all wrong (for example, that bastion of rational thinking and science, Rush Limbaugh
). Let me be clear: from the data published, you cannot make that case. All you can say is that 1998 is not the hottest year on record, and that only four (instead of five) of the top ten hottest years on record have been since 1990. But if you look at the data (and I would love to see the error bars in this) you still see that the years since 1980 have been hotter on average than in the past 127 years. These data by themselves do not throw global warming into the trash heap. In fact, they can't: these are number for the United States only, and not the whole planet! And, to my eye, they in fact support a warming trend since the middle of the century. Note: I won't comment on the hockey-stick Mann graph since I don't know the details on how this affects it -- again, without errors bars there is no way to know. So, who's full of hot air?
