Arguments
Global warming and the El Niño Southern Oscillation
What the science says...
The El Nino Southern Oscillation shows close correlation to global temperatures over the short term. However, it is unable to explain the long term warming trend over the past few decades.
It's El Niño
"Three Australasian researchers have shown that natural forces are the dominant influence on climate, in a study just published in the highly-regarded Journal of Geophysical Research. According to this study little or none of the late 20th century global warming and cooling can be attributed to human activity. The close relationship between ENSO and global temperature, as described in the paper, leaves little room for any warming driven by human carbon dioxide emissions. The available data indicate that future global temperatures will continue to change primarily in response to ENSO cycling, volcanic activity and solar changes." (Climate Depot)
The paper claiming a link between global warming and the El Niño Southern Oscillation (ENSO) is Influence of the Southern Oscillation on tropospheric temperature (McLean 2009). What does the paper find? According to one of it's authors, Bob Carter,
"The close relationship between ENSO and global temperature, as described in the paper, leaves little room for any warming driven by human carbon dioxide emissions."
In other words, they claim that any global warming over the past few decades can be explained by El Niño activity.
How do they arrive at this conclusion? They begin by comparing satellite measurements of tropospheric temperature to El Niño activity. Figure 1 plots a 12 month running average of Global Tropospheric Temperature Anomaly (GTTA, the light grey line) and the Southern Oscillation Index (SOI, the black line).
Figure 1: Twelve-month running means of SOI (dark line) and MSU GTTA (light line) for the period 1980 to 2006 with major periods of volcanic activity indicated (McLean 2009).
The Southern Oscillation Index shows no long term trend (hence the term Oscillation) while the temperature record shows a long term warming trend. Consequently, they find only a weak correlation between temperature and SOI. Next, they compare derivative values of SOI and GTTA. This is done by subtracting the 12 month running average from the same average 1 year later. They do this to "remove the noise" from the data. They fail to mention it also removes any linear trend, which is obvious from just a few steps of basic arithmetic. It is also visually apparent when comparing the SOI derivative to the GTTA derivative in Figure 2:
Figure 2: Derivatives of SOI (dark line) and MSU GTTA (light line) for the period 1981–2007 after removing periods of volcanic influence (McLean 2009).
The linear warming trend has been removed from the temperature record, resulting in a close correlation between the filtered temperature and SOI. The implications from this analysis should be readily apparent. El Niño has a strong short term effect on global temperature but cannot explain the long term trend. In fact, this is a point made repeatedly on this website (eg - here and here).
This view is confirmed in other analyses. An examination of the temperature record from 1880 to 2007 finds internal variability such as El Nino has relatively small impact on the long term trend (Hoerling 2008). Instead, they find long term trends in sea surface temperatures are driven predominantly by the planet's energy imbalance.
There have been various attempts to filter out the ENSO signal from the temperature record. We've examined one such paper by Fawcett 2007 when addressing the global warming stopped in 1998 argument. Similarly, Thompson 2008 filters out the ENSO signal from the temperature record. What remains is a warming trend with less variability:
Figure 3: Surface air temperature records with ENSO signal removed. HadCRUT corrections by Thompson 2008, GISTEMP corrections by Real Climate.
Foster and Rahmstorf (2011) used a multiple linear regression approach to filter out the effects of volcanic and solar activity and ENSO. They found that ENSO, as measured through the the Multivariate ENSO Index (MEI), had a slight cooling effect of about -0.014 to -0.023°C per decade in the surface and lower troposphere temperatures, respectively from 1979 through 2010 (Table 1, Figure 4). This corresponds to 0.045 to 0.074°C cooling from ENSO since 1979, respectively. The results are essentially unchanged when using SOI as opposed to MEI.
Table 1: Trends in °C/decade of the signal components due to MEI, AOD and TSI in the regression of global temperature, for each of the five temperature records from 1979 to 2010.
Figure 4: Influence of exogenous factors on global temperature for GISS (blue) and RSS data (red). (a) MEI; (b) AOD; (c) TSI.
Like Foster and Rahmstorf, Lean and Rind (2008) performed a multiple linear regression on the temperature data, and found that although ENSO is responsible for approximately 12% of the observed global warming from 1955 to 2005, it actually had a small net cooling effect from 1979 to 2005. Overall, from 1889 to 2005, ENSO can only explain approximately 2.3% of the observed global warming.
Ultimately, all the data analysis shouldn't distract us from the physical reality of what is happening to our climate. Over the past 4 decades, oceans all over the globe have been accumulating heat (Levitus 2008; Nuccitelli et al. 2012, Figure 5). The El Niño Southern Oscillation is an internal phenomenon where heat is exchanged between the atmosphere and ocean and cannot explain an overall buildup of global ocean heat. This points to an energy imbalance responsible for the long term trend (Wong 2005).
Figure 5: Land, atmosphere, and ice heating (red), 0-700 meter OHC increase (light blue), 700-2,000 meter OHC increase (dark blue). From Nuccitelli et al. (2012),
Data analysis, physical observations and basic arithmetic all show ENSO cannot explain the long term warming trend over the past few decades. Hence the irony in Bob Carter's conclusion "The close relationship between ENSO and global temperature leaves little room for any warming driven by human carbon dioxide emissions". What his paper actually proves is once you remove any long term warming trend from the temperature record, it leaves little room for any warming.
Intermediate rebuttal written by dana1981
Last updated on 10 July 2015 by pattimer. View Archives
The differences between that graph and the first in your post are that I have used weekly rather than monthly data, and that I have rescaled by the ration of the Standard Deviations of the Data, ie, by multiplying the Nino 3.4 anomaly by 0.1224305012. Had I scaled the global SST to match Nino 3.4 as done in the first graph, the factor would have been 8.1678992563 rather than 10. In other words, that graph inflates the Global SST anomaly values by 22%. I also detrended the Nino 3.4 anomaly, and it is not clear that Tisdale has done so in the first graph. I hope that helps.
It is simply the weekly SST anomaly for various regions of interest to this discussion. It is presented without rescaling, rebaselining, or lagging or any other device that might conceal the vast differences of variability between the two regions. So presented it is immediately apparent just how variable the Nino 3.4 region is compared to other regions.
You might think that that is just because it is a small area, and the larger areas contain sub-regions with contrasting trends that reduce the mean variability. However, one of those regions, (Indian Ocean Tropical, in orange) is the region of the Indian Ocean from 50 to 100 degrees East, and from 5 degrees North to 5 degrees South. That is, it occupies the same area, and the same latitude as the Nino 3.4 region, yet it has only 32.3% of the variability of the Nino 3.4 region. Clearly the Nino 3.4 region (and presumably other directly ENSO effected regions) are exceptionally variable relative to other regions of the world's oceans.
Another way of saying that is that the region has a very high noise to signal ratio when it comes to detecting medium term trends. The same is true of the East Pacific, which is dominated by ENSO variations.
We all know that when it comes to detecting long term trends, we want to focus on the signal, not the noise. If you do not, you will find yourself "going down the up escalator, and all sorts of other improbable contortions. That is not reason to avoid studying the Eastern Pacific SST, but it is reason to take care that you are not fooling yourself with noise to avoid looking at the signal.
You can take several steps to avoid fooling yourself in that regard. You can check out statistical significance:
This is a bit crude, but as the table shows, the increase in regional temperatures plus two standard deviations for Nino3.4 (1.52 C)and the East Pacific (0.4 C) comfortably exceeds the increase in Global SST anomaly (0.26 C). A proper analysis of the error of trend calculations taking into account auto-correlation and multi-decadal influences such as the PDO is likely to find (I suspect) that they do not differ significantly from the model predictions for the East Pacific of 0.42 and 0.44 C (as quoted by Tisdale).
Be that as it may, the important point here is that Tisdale's hypothesis does not even pass the most rudimentary test in showing that the limited warming in the East Pacific is unexpected given global warming.
An even more rudimentary test is to check that your analysis is robust with regard to endpoints, and hence is not just a function of short term fluctuations. In this case, we that ENSO dominates the region and is the source of the large fluctuations in temperature. Looking at the ENSO 3.4 index, we see that Tisdale's start point just happens to precede the second strongest El Nino event since 1950, and precedes a period note worthy for its lack of La Ninas; but that the end of the period is noteworthy for weak El Ninos and several strong La Ninas:
This pattern fully accounts for the negative trend in the Nino 3.4 anomaly, and hence the flat trend for the Eastern Pacific. Is it, however, a consequence of the start point of the data?
As it turns out, yes it is. Just preceding the start point was another period of strong La Ninas. Including that period in the data results in a warming East Pacific:
It turns out that not only is the lack of warming in the East Pacific not statistically significant, its existence is an artifact of the choice of start and endpoints in the data.
I assume this is not the result of conscious cherry picking. Tisdale's chosen data set just happens to start in 1981. But he should have been aware of the possibility that what he thought he saw was just an artifact of incomplete data, and checked longer time periods using other data. He was negligent at least.
This leaves him in an awkward position. Tisdale has provided no physical basis for his theory. He is, therefore, recommending it to us soley on the basis of the statistically unusual nature of the data. But it turns out that the data is not statistically unusual after all. That leaves his theory with nothing to recommend it until he can provide us with a physical mechanism behind his theory.
Yet if you fit a trend line it yields a statistically significant result.
Now my question to you is, if you just eyeball the graph and claim that there is an increase till 2005 and then decrease from there, is it a valid claim?
If you claim that there is a step, you at the bare minimum have to run some statistical testing as a sanity check. If your assertion passes a statistical test, we can then proceed with caution. If does not, there is no reason for it to be taken seriously.
############################
Now that I think about it a bit more, I think there is a much bigger problem.
Your basis for shifting the global SSTA at those two points is based on your arguement that "extra heat piled up globally after an El Niño and during the transition to the first following La Niña", which you supported with the animation, that supposedly shows the "extra heat globally".
The problem is this: since you are plotting a scaled global temperature anomalies and a regional temperature anomalies, you can't even interpret the difference between the two as a temperature difference because one of the quantities is scaled, let alone extra heat.
For those wanting technical details, I adjusted each interval between Kayell's stated adjustment points to have a common mean. Transitions where spaced over five weeks with an equal change in each week to avoid introducing very large adjustments in a single week. Both the Nino3.4 and Global anomalies have been divided by their Standard Deviation. The rescaling makes the magnitude of variation consistent for comparison without pretending global fluctuations are much larger in degrees C than they actually are.
Contrary to Sphaerica @74, although this method of detrending is arbitrary, it does not give a noticeably inferior fit to simple linear detrending:
Indeed, if you look at the residuals - the difference between the scaled Nino 3.4 anomaly and the scaled Global anomaly in the two graphs, it is difficult to tell them apart:
Of particular interest here is the yellow line, which shows the difference in the residuals between the linearly detrended and Kayell adjusted graphs. It shows a clear pattern of rising gradually with the linear detrending, with three abrupt shifts down from the adjustments following Kayell's method. The important thing to note is that the difference never exceeds (approximately) 0.5 Standard Deviations. That means the difference between the two methods is not statistically significant.
That simple fact is devastating to Kayell's argument.
Kayell (and Tisdale) argue that the data show step changes in the SST data, yet it is well known that the combination of a linear trend, noise and natural cycles can give the appearance of step changes where none exist:
Consequently, when we see what may naively be interpreted as a step change, we need evidence that we are not simply fooling ourselves. That evidence, for Kayell (and Tisdale) is the ability to "detrend" the global SST anomaly by introducing three major shifts rather than by simply linear detrending. But as it turns out, there is not statistical difference between the two methods. Therefore, the ability to "detrend" by introducing step changes is not significant evidence.
Indeed, it is worse than that. We can test the "fit" between the Nino3.4 and Global anomalies for the two different methods of "detrending". We can take the correlation, for example, although there are reasons to think it is not a suitable method for the comparison. Alternatively, we can take the Root Mean Squared Deviation:
That's all the difference there is between the two methods of "detrending". The linear detrend scores (just) better in terms of Root Mean Square Deviation, but the Kayell Adjustment shows a whopping 0.0075 better correlation between Nino 3.4 and adjusted Global SST anomalies. And on that massive 0.0075 difference rests the entire claim of a step change in sea surface temperatures following the 97/98 El Nino.
Colour me unimpressed.
Of course, this is not a case where we are comparing the predictions of two physical theories with distinct predictions. Climate science clearly predicts an approximately linear trend in SST over the period of interest, as modulated by ocean cycles and noise. There is a substantial, well worked out physical theory predicting the result. What is more, there are known factors likely to lead to the appearance of a step change in the data. The Pinatubo eruption in the 1990s, and a weakening solar cycle coupled with a switch from El Ninos to La Ninas in the 2000s are, when coupled with global warming, perfectly adequate explanations for that appearance.
In stark contrast, neither Kayell nor Tisdale have provided any physical theory predicting a step change. Consequently their only justification for believing one exists is statistical, ie, the 0.0075 point superior correlation when using the Kayell Adjustment compared to a linear detrending. So once again we find that Kayell and Tisdale have no evidence to support their claims, and no theory to justify making them.
Source: 2012: Record Arctic Sea Ice Melt, Multiple Extremes and High Temperatures, WMO Press Release No 966, Nov 28, 2012
Climate Myth...