I’ve moved to WordPress. This post can now be found at Comments On Tamino’s AMO Post#####################
UPDATE (3-16-11): Refer to the update at the end of the post.
Tamino’s AMO post is a response to my post Removing The Effects of Natural Variables - Multiple Linear Regression-Based or “Eyeballed” Scaling Factors (hereinafter referred to as the “Removing” post). Tamino took exception to my inclusion of the AMO as one of the datasets used to explain the rise in GISS Land-Ocean Temperature Index (60S-60N) during the satellite era. Please read Tamino’s AMO post before continuing.
My “Removing” post, as discussed in its opening paragraph, was the second in a series follow-ups to the earlier post Can Most Of The Rise In The Satellite-Era Surface Temperatures Be Explained Without Anthropogenic Greenhouse Gases? (hereinafter referred to as the “Can Most” post). The first follow-up was Notes On Polar Amplification.
And for those new to the Atlantic Multidecadal Oscillation (AMO) refer to the post An Introduction To ENSO, AMO, and PDO -- Part 2.
THE REAL CLIMATE DESCRIPTION OF THE AMO
Tamino wrote in his post, “Bob Tisdale (and others) simply can’t wrap their brains around the fact that global warming is the cause, not the effect, of much of the changes in N.Atl SST anomaly. Therefore global warming is the cause, not the effect, of much of the variation in the AMO.”
My AMO posts typically include the RealClimate description of the Atlantic Multidecadal Oscillation (“AMO”), but I failed to include it in “Removing” post. RealClimate states, “A multidecadal (50-80 year timescale) pattern of North Atlantic ocean-atmosphere variability whose existence has been argued for based on statistical analyses of observational and proxy climate data, and coupled Atmosphere-Ocean General Circulation Model (“AOGCM”) simulations. This pattern is believed to describe some of the observed early 20th century (1920s-1930s) high-latitude Northern Hemisphere warming and some, but not all, of the high-latitude warming observed in the late 20th century. The term was introduced in a summary by Kerr (2000) of a study by Delworth and Mann (2000).”
Tamino’s opinion contradicts the opinions of his associates at RealClimate, or at least the opinion of the author of the RealClimate AMO webpage. RealClimate describes the AMO as being responsible for some, but not all, of the warming, but Tamino states it’s the other way around, that the global warming signal is the cause of the AMO variability.
Tamino’s RealClimate associates must be among “the others” who “simply can’t wrap their brains around the fact that global warming is the cause, not the effect, of much of the changes in N.Atl SST anomaly.”
A NOTE ABOUT THE SST DATASET USED IN THIS POST
GISS uses two SST anomaly datasets in its Land-Ocean Temperature (LOTI) product: HADISST from January 1880 to November 1981 and Reynolds OI.v2 from December 1981 to present. There is little difference between the HADISST and Reynolds OI.v2 data for the North Atlantic during the satellite era, as shown in Figure 1. So my use of HADISST data in the short-term will not influence the results of this post.
However, there is a significant difference between the long-term Kaplan North Atlantic SST data used by the ESRL (and Tamino) and the HADISST data used by GISS. Refer to Figure 2. Keep in mind my use of the ESRL data was only for the AMO index in the short term, not the long-term SST data used by Tamino. (Note: I confirmed via email that the ESRL uses the coordinates of 0-70N and 80W-0 for its AMO data.)
And the difference does impact Tamino’s post. He uses the wrong North Atlantic SST anomaly dataset when he subtracts global temperatures from it. That is, assuming Tamino did not switch to the HADISST version of the North Atlantic, he biased the results in his last graph by the difference in the trends of the HADISST data (used by GISS) and the Kaplan data (used by ESRL) shown in Figure 2.
ON THE NONLINEARITY OF THE WARMING SIGNAL
The natural multidecadal variability of the North Atlantic SST anomalies is significantly greater than that of the Global (90S-90N) SST anomalies. This is very apparent if we compare detrended North Atlantic SST anomalies (AMO) to detrended Global SST data, Figure 3. The data have been smoothed with a 121-month running-average filter.
Tamino opens his post with a discussion of the how the AMO is calculated by detrending North Atlantic SST anomalies, and he notes that the Wikipedia definition warns about the nonlinearity of the actual warming signal. The nonlinearity of the detrended global SST signal is shown clearly in my Figure 5 above. Based on his presentation, Tamino concludes, “Variations in the forced signal do leak into the AMO definition.”
Let’s compare the short-term linear trends of the North Atlantic SST anomalies to the trends of the other ocean basins. This is a general discussion of the AMO, so I’ve left in the Arctic and Southern Ocean data. Keep in mind that my “Removing” and “Can Most” posts only dealt with the period starting in 1982, which is the satellite era for SST data. As shown in the spaghetti graph, Figure 4, the SST anomaly linear trend of the North Atlantic is significantly higher than all other SST basins. The linear trend of the Arctic Ocean SST anomalies comes in second, in part because those two datasets overlap and due to the influence of the North Atlantic on the Arctic Ocean. Regardless, the North Atlantic linear trend is almost twice that of the Arctic Ocean. The North Atlantic trend is more than 3 times higher than the trends of the North Pacific and Indian Oceans and more than 5 times higher than the trends of the South Atlantic and South Pacific. And of course, the Southern Ocean linear trend is negative. (Note: The impact of the Southern Ocean cooling is so substantial that the trend is basically flat for all HADISST anomaly data south of 40S, or about 35% of the global oceans, since 1982.)
This difference in linear trends can also be seen in the comparison of North Atlantic SST anomalies and the SST anomalies for the rest of the world. To determine the rest-of-the-world data (identified as “Global Without No Atlantic” in Figure 5), I approximated the North Atlantic surface area as a percentage of the global oceans. The Atlantic represents approximately 30% of the surface area of the global oceans. I assumed the North Atlantic made up half of that, or 15%, before scaling the North Atlantic data and subtracting it from the global data for Figure 5. The linear trend of the North Atlantic SST anomalies is more than 5 times greater than the average of the other ocean basins.
In fact, the contribution of the North Atlantic is so great, without it, the global trend drops by 45%, Figure 6.
Tamino did not suggest how to account for the global warming signal in his AMO post, unless the last graph in which he subtracts global GISS LAND-Ocean Temperature Index data from North Atlantic SEA Surface Temperature data is his recommendation. But he did make a suggestion on his earlier How Fast is Earth Warming? thread. He wrote in response to a January 23, 2011 at 4:42 pm comment, “It might be interesting to correlate AMO to short-term global temperature fluctuations, if AMO is detrended nonlinearly, or if only the modern era (1975 to present) is detrended separately. But then: the denialists' claim disappears.”
To account for the nonlinear signal, Trenberth and Shea (2006) proposed subtracting the global (60S-60N) SST data from the North Atlantic in “Atlantic hurricanes and natural variability in 2005”. But the North Atlantic represents a major portion (almost 50%) of the recent rise in global SST anomalies (90S-90N) since 1982, Figure 6. Therefore, Trenberth and Shea are suggesting the subtraction of a dataset with a strong North Atlantic signal from the North Atlantic SST data itself. Why not subtract the SST anomalies of the rest of the world from the North Atlantic? It’s the additional variability of the North Atlantic, above and beyond the rest of the world, that’s of interest, not a signal that’s been suppressed by itself.
The reason that method hasn’t been suggested becomes obvious when one compares that dataset to the AMO data based on detrended North Atlantic SST anomalies. Refer to Figure 7. (The “Rest of the World” data is calculated the same as the “Global Without North Atlantic” from Figures 5 & 6.) Note how the curves mimic one another from 1905 to the early 1980s. They diverge from time to time, but the curves are similar. But note how VERY similar the two curves are after 1982. That’s the period of the AMO data used in my “Removing” post.
Let’s look at the satellite-era portion (1982 to present) of those two datasets, Figure 8. The trends are basically the same, and the year-to-year variability of the two signals mimic one another with small divergences and lags. Based on Figure 10, the “Variations in the forced signal do leak into the AMO definition,” as Tamino notes, but they have had little impact on the results of my “Removing” post.
THE DIFFERENCE BETWEEN THE KAPLAN AND HADISST NORTH ATLANTIC SST ANOMALIES
The Kaplan and HADISST versions of the North Atlantic SST anomalies were illustrated together in Figure 2. There was a significant difference in their linear trends. For Figure 9, I subtracted the HADISST version of the North Atlantic SST anomalies from the Kaplan SST anomalies used by ESRL (and Tamino for his last graph). Note the similarities between Figure 9 and Tamino’s final graph in his AMO post.
TAMINO’S FINAL COMPARISONS
Tamino’s post included a comparison graph of Global (90S-90N) GISS LOTI and the North Atlantic SST anomalies he created from the data on the ESRL AMO webpage. The last illustration was a graph of the difference. While I can’t find fault is his not knowing there was a shift in the Kaplan North Atlantic SST data, I can find fault in his using the wrong SST dataset. GISS does not use Kaplan SST.
There is little difference between the HADISST and Reynolds OI.v2 versions of the North Atlantic SST data, as shown in Figure 1. To assure the following comparisons were correct, for the following graphs I spliced those two North Atlantic SST anomaly datasets using the method described by GISS in Step 4 on their current analysis webpage. Had Tamino used the HADISST/Reynolds OI.v2-based GISS SST anomalies for the North Atlantic in his comparison, Figure 10, the difference between it and the Global GISS LOTI data would have maintained the appearance of the AMO.
And had Tamino detrended both datasets and smoothed them with 121-month filters, Figure 11, he would have noted that the multidecadal variability of the North Atlantic far exceeds that of the Global GISS LOFTI data—even with the additional land surface temperature variability in the LOTI data—even with the exaggeration of polar amplification in the LOTI data—even with the bias caused by GISS’s deletion of polar sea surface temperature data in the LOTI data.
I’ll respond to his comments about “eyeballing” in another post.
With the exception of the ESRL North Atlantic SST data (linked numerous times in the post), all data are available through the KNMI Climate Explorer:
In this post, one method used to determine the AMO was to calculate it as the difference between the North Atlantic and the Rest of the World SST anomalies. I had used 15% as the scaling factor of the North Atlantic. The 15% was 1/2 of the 30% represented by the surface area of the Atlantic (North and South) when compared to the rest of the world, BUT this excluded the Arctic and Southern Oceans. That is, using the surface areas (Source Wikipedia) of the Atlantic (106.4 million sq. km), Indian (73.5 million sq. km) and Pacific (165.2 million sq. km) Oceans, the Atlantic represents about 30% of the surface area of those ocean basins. Half of that is obviously 15%. Excluding the Arctic and Southern Oceans seemed appropriate for the ballpark number since GISS deletes most of the Southern and Arctic Oceans, and the original post in the series was Can Most Of The Rise In The Satellite-Era Surface Temperatures Be Explained Without Anthropogenic Greenhouse Gases? However, if the Southern and Arctic Oceans are included, the North Atlantic surface area ballpark percentage would drop to around 12%. If you include only the surface area of the coordinates used in the post, that scaling factor for the North Atlantic would drop to around 11.5%.
I will clarify this in another post and illustrate the minimal effect this has on the AMO when the AMO is calculated as the difference between the North Atlantic and Rest of the World SST anomalies.
Ultimately, this has no impact on the conclusion of this post, which was that Tamino had used the wrong SST dataset in his post. To illustrate that fact, I subtracted the Global Land+Ocean Temperature anomalies from the North Atlantic SST anomalies, not the AMO.
I give you high marks for the time and effort that you put into this post. With what we know today, you've pretty much covered all the bases and anyone with a open mind would agree I believe. You are, however, dealing with an individual with a closed mind I'm afraid and for that problem there is no logic or solution
Pascvaks, I understand.
You may note a slight difference in this version of the post, Figure 9. I pulled the post and corrected the error. I then reposted it with a different url. But I saved your comment.
Thanks Bob for your comprehensive analysis!
I have been reading your blog for a while but haven't been a regular commented. I like your SST analysis and graphs, and they help a lot to understand the dynamics of the oceans.
However, there are some points of your presentations which I remain skeptical, I'm sure you dont mind:
The explanation of KOE and SPCZ-anomalies is a bit of a question mark. Like is the East-Indian Western Pacific ('EIWP' from now on) SST anomalies. How do you know, and how have you addressed that it is not the GHG's which help those regions to stay warm after El Nino? I am propably not the first one who asks this I think. Some calculations & graphs with radiative forcing included would help in this regard.
Maybe those anomalies should be just detrended?
What I also would like to see EIWP minus Global SST (without EIWP) look like, in the same way you did for AMO in this post. And perhaps you could do the same for KOE and SPCZ anomalies too?
Even though these small (or large as some might think) questions I think you are definitely on the right track doing these analysis. At least it beats Tamino 10-0. Keep up the good work!
juakola: I'll try to address your concerns in future posts, but the largest problem, as far as I can see, about the hypothesis of AGW is the impact of downward longwave radiation (DLR), that is, infrared radiation from greenhouse gases, on the oceans. DLR can only penetrate the top few millimeters of the ocean surface, but downward shortwave radiation (visible light" penetrates to a couple of hundred meters, with the intensity weakening with depth.
And most of the DLR is lost from the surface due to eveporation.
Anonymous: Your first link set off my virus software so I deleted your comment.
Is there any trend in the blue curve in fig. 7?
juakola said: "Is there any trend in the blue curve in fig. 7?"
A very small positive trend. Based on the unsmoothed data, the linear trend of the North Atlantic Minus the Rest of the World SST anomalies is 0.06 deg C/Century.
So in contradiction to Tamino's claims, detrending the NA SST is doing actually the opposite - leaking some AMO signal to other signals.
Am I correct?
juakola said: "So in contradiction to Tamino's claims, detrending the NA SST is doing actually the opposite - leaking some AMO signal to other signals."
Is this a continuation of the previous question and answer? If so, I'm not sure how you came to that conclusion. Please explain.
Figures 7 & 8 show that there is little difference if you create the AMO by detrending the data or if you create it by subtracting the rest of the world SST anomalies from the North Atlantic.
Post a Comment