Guest post by Bob Tisdale
Longer Title: Do Multidecadal Changes In The Strength And Frequency Of El Niño and La Niña Events Cause Global Sea Surface Temperature Anomalies To Rise And Fall Over Multidecadal Periods?
############
UPDATE (November 19, 2010): I’ve added a clarification about the running total of scaled NINO3.4 SST anomalies and its implications. I changed a paragraph after Figure 13, and added a discussion under the heading of “What Does The Running Total Imply?”
############
OVERVIEW
This post presents evidence that multidecadal variations in the strength and frequency of El Niño and La Niña events are responsible for the multidecadal changes in Global Sea Surface Temperature (SST) anomalies. It compares running 31-year averages of NINO3.4 SST anomalies (a widely used proxy for the frequency and magnitude of ENSO events) to the 31-year changes in global sea surface temperature anomalies. Also presented is a video that animates the maps of the changes in Global Sea Surface Temperature anomalies over 31-year periods, (maps that are available through the GISS Map-Making web page). That is, the animation begins with the map of the changes in annual SST anomalies from1880 to 1910, and it is followed by maps of the changes from 1881 to 1911, from 1882 to 1912, etc., through 1979 to 2009. The animation of the maps shows two multidecadal periods, both containing what appears to be a persistent El Niño event, one in the early 1900s and one in the late 1900s to present, and between those two epochs, there appears to be a persistent La Niña event.
INTRODUCTION
A long-term (1880 to 2009) graph of Global Surface Temperature anomalies or Global Sea Surface Temperature (SST) anomalies (Figure 1) often initiates blog discussions about the causes of the visible 60-year cycle. The SST anomalies rise from early-1910s to the early-1940s, drop from the early 1940s to the mid-1970s, then rise from the mid-1970s to present. Natural variables like the Pacific Decadal Oscillation (PDO) and the Atlantic Multidecadal Oscillation (AMO) are cited as the causes for these variations.
http://i51.tinypic.com/119z1ux.jpg
Figure 1
Note: HADISST data was used for the long-term SST anomaly graphs in this post. The exception is the GISS SST data, which is a combination of HADISST data before the satellite era and Reynolds OI.v2 SST data from December 1981 to present.
THE PDO CANNOT BE THE CAUSE
The SST anomalies of the North Pacific region used to calculate the PDO are inversely related to the PDO over decadal periods. This was shown in the post An Inverse Relationship Between The PDO And North Pacific SST Anomaly Residuals. This means that the SST anomalies of the North Pacific contribute to the rise in global SST anomalies during decadal periods when the PDO is negative and suppress the rise in global SST anomalies when the PDO is positive. The PDO, therefore, cannot be the cause of the multidecadal rises and falls in global SST anomalies. That leaves the AMO or another variable.
MULTIDECADAL CHANGES IN GLOBAL SST ANOMALIES
If we subtract the annual global SST anomalies in 1880 from the value in 1910, the difference is the change in global SST anomalies over that 31-year span. Using this same simple calculation for the remaining years of the dataset provides a curve that exaggerates the variations in global SST anomalies. This dataset is identified as the “Running Change (31-Year) In Global SST Anomalies” in Figure 2. The data have been centered on the 16th year.
http://i55.tinypic.com/2cndnq1.jpg
Figure 2
Why 31 years? A span of 31 years was used because it is approximately one-half the apparent cycle in the datasets, and it should capture the maximum trough-to-peak and peak-to-trough changes that occur as part of the 60-year cycle. Using 31 years also allows the data to be centered on the 16th year, with 15 years before and after.
The curve of the “Running Change (31-Year) In Global SST Anomalies” is very similar to the curve of annual NINO3.4 SST anomalies that have been smoothed with a 31-year filter. Refer to Figure 3. (NINO3.4 SST anomalies are commonly used to illustrate the frequency and magnitude of El Niño and La Niña events. For readers new to the topic of El Niño and La Niña events, refer to the post An Introduction To ENSO, AMO, and PDO – Part 1.) Both datasets are centered on the 16th year. Considering how sparse the SST measurements are for the early source data, the match is actually remarkable at that time.
http://i55.tinypic.com/zmgv9l.jpg
Figure 3
Let’s take a closer look at that relationship. The purple curve represents the running 31-year average of annual NINO3.4 SST anomalies, and it shows that, for example, at its peak in 1926, the frequency and magnitude of the El Niño events from 1911 to 1941 were far greater than the frequency and magnitude of La Niña events. The blue curve, on the other hand, portrays the change in global SST anomalies based on a 31-year span, and it shows, at its peak in 1926 that global SST anomalies rose more from 1911 to 1941 than it did during the other 31-year periods in the early 20th century. Skip ahead a few decades to 1960. Both curves reached a low point about then. At 1960, the purple curve indicates the frequency and magnitude of La Niña events from 1945 to 1975 outweighed El Niño events. And over the same period of 1945 to 1975, annual global SST anomalies dropped the greatest amount. Afterwards, the frequency and magnitudes of El Niño events increased (and/or the frequency and magnitude of La Niña events decreased) and the multidecadal changes in global SST anomalies started to rise, eventually reaching their peak around 1991 (the period of 1976 to 2006).
Since Global SST anomalies respond to changes in NINO3.4 SST anomalies, this relationship implies that the strengths and frequencies of El Niño and La Niña events over multidecadal periods cause the multidecadal rises and falls in global sea surface temperatures. In other words, its shows that global sea surface temperatures rose from 1910 to the early 1940s and from the mid-1970s to present because El Niño events dominated ENSO during those periods, and it shows that global sea surface temperatures dropped from the early 1940s to the mid 1970s because La Niña events dominated ENSO.
This apparent relationship contradicts the opinion presented by some climate studies that ENSO is only noise, that ENSO is only responsible for the major year-to-year wiggles in the global SST anomaly curve. Refer back to Figure 1. Examples of these studies are Thompson et al (2009) “Identifying Signatures of Natural Climate Variability in Time Series of Global-Mean Surface Temperature: Methodology and Insights” and Trenberth et al (2002) “Evolution of El Nino–Southern Oscillation and global atmospheric surface temperatures”.
Link (with paywall) to Thompson et al (2009):
http://journals.ametsoc.org/doi/abs/10.1175/2009JCLI3089.1
Link to Trenberth et al (2002):
http://www.cgd.ucar.edu/cas/papers/2000JD000298.pdf
Keep in mind, when climate studies such as Thompson et al (2009) and Trenberth et al (2002)attempt to account for El Niño and La Niña events in the global surface temperature record they scale an ENSO proxy, like NINO3.4 SST anomalies, and subtract it from the Global dataset, removing the major wiggles. They then assume the difference, which is a smoother rising curve, is caused by anthropogenic greenhouse gases.
The relationship in Figure 3 (that the multidecadal variations in strength and frequency of ENSO events are responsible for the rises and falls in global sea surface temperature) also contradicts the basic premise behind the hypothesis of anthropogenic global warming, which assumes that the rise in global sea surface temperatures since 1975 could only be caused the increase in anthropogenic greenhouse gases.
The first question that comes to mind: shouldn’t a multidecadal rise in Sea Surface Temperatures require an increase in radiative forcing? The answer is no, and I’ll discuss this later in the post. Back to Figure 3.
Once more, the relationship in Figure 3 illustrates that multidecadal variations in the frequency and magnitude of El Niño and La Niña events cause the multidecadal changes in SST anomalies. But how do I verify that this is the case, and how do I illustrate it for those without science backgrounds? Again, for those who need to brush up on El Niño and La Nina events, refer to the post An Introduction To ENSO, AMO, and PDO – Part 1.
THE ANIMATION OF MULTIDECADAL CHANGES IN SST ANOMALIES
The Goddard Institute of Space Studies (GISS) Global Map-Making webpage allows users to create maps of global SST anomalies and maps of the changes in global SST anomalies (based on local linear trends) over user-specified time intervals. Figure 4 is a sample map of the changes in annual SST anomalies for the 31-year period from 1906 to 1936. In the upper right-hand corner is a value that represents the change in annual SST anomalies over that time span. GISS describes the value as, “Temperature change of a specified mean period over a specified time interval based on local linear trends.” And as far as I can tell, these local linear trends are weighted by latitude. I downloaded the GISS maps of the changes in annual global SST anomalies, starting with the interval of 1880 to 1910 and ending with the interval of 1979 to 2009, with the intent of animating the maps, but the data they presented was also helpful.
http://i56.tinypic.com/21ou8lg.jpg
Figure 4
Figure 5 shows the curve presented by the GISS Multidecadal (31-year span) Changes In Global SST anomalies for all those maps, with the data centered on the 16th year. Comparing it to the “Running Change (31-Year) In Global SST Anomalies” data previously calculated, Figure 6, illustrates the similarities between the two curves. The GISS data from the maps presents a much smoother curve.
http://i53.tinypic.com/14j50et.jpg
Figure 5
###############
http://i51.tinypic.com/9hq83s.jpg
Figure 6
And if we compare the curve of the GISS Multidecadal (31-year span) Changes In Global SST anomalies from those maps to the NINO3.4 SST anomalies smoothed with a 31-month filter, Figure 7, we can see that the multidecadal changes in Global SST anomalies lag the variations in strengths and magnitudes of ENSO events. The lag prior to 1920 appears excessive, but keep in mind that the early source SST measurements are very sparse. The fact that there are similarities in the curves in those early decades says much about the methods used by researchers to infill all of that missing data.
http://i54.tinypic.com/9gvyh0.jpg
Figure 7
THE VIDEO
The animations are presented in two formats in the YouTube video titled “Multidecadal Changes In Global SST Anomalies”. The first format is as presented by GISS, with the Pacific Ocean split at the dateline. That is, the maps are centered on the Atlantic. Refer back to Figure 4. The second format is with the maps rearranged so that the major ocean basins are complete. Those maps are centered on the Pacific. With the maps centered on the Pacific, the animation shows what appear to be two (noisy) multidecadal El Niño events separated by a multidecadal La Niña event.
As noted in the video, the long-term El Niño and La Niña events appear in the patterns, not necessarily along the central and eastern equatorial Pacific. For those not familiar with the SST anomaly patterns associated with ENSO, refer to Figure 8. It is Figure 8 from Trenberth et al (2002) “Evolution of El Nino–Southern Oscillation and global atmospheric surface temperatures”. Link to Trenberth et al (2002) was provided earlier.
Figure 8 shows where Sea Surface Temperatures warm and cool during the evolution (the negative lags) of an ENSO event, at the peak of an ENSO event (zero lag), and during the decay of ENSO events (the positive lags). The reds indicate areas that are positively correlated with ENSO events, and the blues are areas that are negatively correlated. That is, the red areas warm during an El Niño and the blues are the areas of that cool during an El Niño. During a La Niña event, the reds indicate areas that cool, and the blues indicate areas that warm.
http://i55.tinypic.com/z1f6o.jpg
Figure 8
And for those wondering why the ENSO events don’t always appear along the equatorial Pacific in the animated maps, keep in mind that the maps are showing the multidecadal changes in SST anomalies based on linear trends. The long-term linear trend of the equatorial Pacific SST anomalies are incredibly flat, meaning there is little trend. Refer to Figure 9, which shows the annual NINO3.4 SST anomalies and linear trend from 1900 to 2009.
http://i56.tinypic.com/2ag0u2u.jpg
Figure 9
#############################
http://www.youtube.com/watch?v=O_QopFYSyGE
Video 1
And here’s a link to a stand-alone version of the video. The only difference is that the following version includes a detailed introduction, discussion, and conclusion, which are presented in this post. It’s about 5 minutes longer.
http://www.youtube.com/watch?v=SMKA_uG3zK0
Link To Stand-Alone Version Of Video
DOES THE VIDEO AND DATA PRESENT MORE THAN MULTIDECADAL VARIABILITY IN GLOBAL SST ANOMALIES?
Yes. This has actually been stated a number of times, but the following explanation may be helpful.
One of the arguments presented during discussions of multidecadal variations in global SST anomalies is that the Atlantic Multidecadal Oscillation (AMO) is detrended and that it strengthens or counteracts the basic long-term rise in global SST anomalies. However, the data associated with the GISS maps used in the video are based on linear trends. And Figure 7 shows that the Global SST anomalies rose from 1910 to 1944 and from 1976 to 2009 because El Niño events dominated, and dropped from 1945 to 1975 because La Niña events dominated.
That is, the animation of the GISS maps and the data GISS provides with those maps show that the trends in global sea surface temperature are driven by the multidecadal variations in the strengths and magnitudes of El Niño and La Niña events. The “GISS Multidecadal (31-year span) Changes In Global SST anomaly” data peaked in 1931 at 0.39 deg C. Refer back to Figure 5. That is, from 1916 to 1946, global SST anomalies rose 0.39 deg C (based on local linear trends). That equals a linear trend of 0.13 deg C per decade. And the “GISS Multidecadal (31-year span) Changes In Global SST anomaly” data peaked in 1989 at 0.41 deg C, and that equals a trend of 0.137 deg C per decade from 1974 to 2004. Let’s look at the “Raw” Global SST anomaly data. The linear trends of the “Raw” Global SST Anomalies for the same periods, Figure 10, are approximately 0.12 deg C per decade. Again, the peaks in the “GISS Multidecadal (31-year span) Changes In Global SST anomaly” data represent the periods with the greatest linear trends, and, as shown in Figure 7, they lag the peaks of the multidecadal variations in NINO3.4 SST anomalies.
http://i56.tinypic.com/343r903.jpg
Figure 10
Note: The highest trend in the later epoch of the GISS-based “change data” is about 5% higher than the highest trend in the earlier warming period. And that’s not unreasonable considering the early period was so poorly sampled. Again, the similarities in trends between the two epochs speaks highly of the methods used by the researchers to infill the data
A NOTE ABOUT THE NORTH ATLANTIC
Oceanic processes such as Atlantic Meridional Overturning Circulation (AMOC) and Thermohaline Circulation (THC) are normally cited as the cause of the additional multidecadal variability of North Atlantic SST anomalies. This additional variability is presented in an index called the Atlantic Multidecadal Oscillation or AMO. The AMO data are simply North Atlantic SST anomalies that have been detrended. As discussed in the post An Introduction To ENSO, AMO, and PDO — Part 2, the NOAA Earth System Research Laboratory (ESRL) Atlantic Multidecadal Oscillation webpage refers readers to the Wikipedia Atlantic Multidecadal Oscillation webpage for further discussion. And Wikipedia’s description includes the statement, “While there is some support for this mode in models and in historical observations, controversy exists with regard to its amplitude…” The phrase “some support” does not project or instill a high level of confidence.
Early in this post we prepared a dataset that illustrated the “Running Change (31-Year) In Global SST Anomalies” by subtracting the annual SST anomalies of a given year from the SST anomalies 30 years later and repeating this each year for the term of 1880 to 2009. We can prepare the “Running Change (31-Year) In North Atlantic SST Anomalies” using the same simple method. Those two datasets (based on global and North Atlantic SST anomalies) are shown in Figure 11. The “Running Change (31-Year) In North Atlantic SST Anomalies” dataset appears simply to be an exaggerated version of the “Running Change (31-Year) In Global SST Anomalies”.
http://i52.tinypic.com/72yjwj.jpg
Figure 11
And comparing the “Running Change (31-Year) In North Atlantic SST Anomalies” to the NINO3.4 SST anomalies smoothed with a 31-year filter, Figure 12, shows that the NINO3.4 SST anomalies lead the multidecadal changes in North Atlantic SST anomalies.
http://i54.tinypic.com/besbvb.jpg
Figure 12
Putting Figures 11 and 12 into other words, the AMO appears to simply be the North Atlantic exaggerating the cumulative effects of the variations in the frequency and magnitude of ENSO. During epochs when El Niño events dominate, the SST anomalies of the North Atlantic rise more than the SST anomalies of the other ocean basins, and when La Niña events dominate, the North Atlantic SST anomalies drop more than the SST anomalies for the rest of the globe.
Why? The South Atlantic (not a typo) is the only ocean basin where heat is transported toward the equator (and into the North Atlantic). So warmer-than-normal surface waters in the South Atlantic created by the changes in atmospheric circulation during an El Niño should be transported northward into the North Atlantic (and vice versa for a La Niña). This effect seems to be visible in the animation of Atlantic SST anomalies from September 23, 2009 to November 3, 2010, Animation 1. (Note: By the start of the animation, September 2009, the 2009/10 El Niño was well underway.) Unfortunately, there is a seasonal component in those SST anomaly maps, and it’s difficult to determine whether the seasonal component is enhancing or inhibiting the appearance of northward migration of warm waters. Rephrased as a question, is the seasonal component in the SST anomalies creating (or detracting from) an illusion that makes it appear that the warm SST anomalies are migrating from the South Atlantic to the North Atlantic?
http://i55.tinypic.com/jzbdqe.jpg
Animation 1
The northward migration of warm waters from the South Atlantic to the North Atlantic also appears to be present in the following animation of the correlation of NINO3.4 SST anomalies with Atlantic SST anomalies at time lags that vary from 0 to 12 months, Animation 2. Again the correlation maps show areas that warm (red) or cool (blue) in response to an El Niño and the positive lags represent the number of months following the peak of the El Niño. Three month average NINO3.4 and Atlantic SST anomalies were used.
http://i52.tinypic.com/2gtai6d.jpg
Animation 2
Another reason the North Atlantic exaggerates the effects of ENSO is because the North Atlantic is open to the Arctic Ocean. El Niño events cause increases in seasonal Arctic sea ice melt during the following summer. It would also seem logical that El Niño events would increase the seasonal Greenland glacial melt as well. Refer again to Animation 2. Starting around the 9-month lag, positive correlations (warm waters during an El Niño) migrate south from the southern tip of Greenland, and starting around the 4-month lag from the Davis Strait, along the west coast of Greenland. Is that from glacial ice melt in Greenland and Arctic sea ice melt, with the melt caused by the El Niño? They’re correlated with NINO3.4 SST anomalies.
Regardless of the cause, in the North Atlantic, there are significant positive correlations with NINO3.4 SST anomalies 12 months after the peak of the ENSO event, and for at least 6 months after the ENSO event has ended. And this means that the El Niño event is responsible for the persistent warming (or cooling for a La Niña event) in the North Atlantic.
MYTH: EL NIÑO EVENTS ARE COUNTERACTED BY LA NIÑA EVENTS
One of the common misunderstandings about ENSO is that La Niña events are assumed to balance out the effects of El Niño events.
The fact: correlations between NINO3.4 SST anomalies and global sea surface temperatures are basically the same for El Niño and La Niña events; that is, El Niño and La Niña events have similar effects on regional sea surface temperatures; they are simply the opposite sign.
But that does not mean the effects of the El Niño event will be counteracted by the La Niña event that follows. First problem with that logic: La Niña events do not follow every El Niño event. That’s plainly visible in instrument temperature record. Refer to the Oceanic Niño Index (ONI) (ERSST.v3b) table. Also an El Niño event may be followed by a La Niña event that lasts for up to three years. And sometimes there are multiyear El Niño events, like the 1986/87/88 El Niño.
The easiest way the show that La Niña events do not counteract El Niño events is by creating a running total of annual NINO3.4 SST anomalies. If La Niña events counteracted El Niño events, a Running Total would return to zero with each El Niño-La Niña cycle. Refer to the Wikipedia webpage on Running total. The running total of NINO3.4 SST anomalies (to paraphrase the Wikipedia description) is the summation of NINO3.4 SST anomalies which is updated each year when the value of a new annual NINO3.4 SST anomaly is added to the sequence, simply by adding the annual value of the NINO3.4 SST anomaly to the running total each year. I’ve scaled the NINO3.4 SST anomalies by a factor of 0.06 before calculating the running total for the comparison graph in Figure 13.
http://i53.tinypic.com/29fcjl2.jpg
Figure 13
And what the Running Total shows is that El Niño and La Niña events do not tend to cancel out one another. There are periods (from 1910s to the 1940s and from the mid 1970s to present) when El Niño events dominated, and a period when La Niña events dominated (from the mid-1940s to the mid-1970s). And with the scaling factor, the running total does a good job of reproducing the global SST anomaly curve. Global temperature anomalies can also be reproduced using monthly NINO3.4 SST anomaly data. This was illustrated and discussed in detail in the post Reproducing Global Temperature Anomalies With Natural Forcings.
UPDATE– The original paragraph has been crossed out and the updated version follows.
Figure 13 implies that 6% of each El Niño and La Niña event remains within the global surface temperature record and that it is this cumulative effect of ENSO events that raises and lowers global Sea Surface Temperatures.
Figure 13 appears to imply that 6% of each El Niño and La Niña event remains within the global surface temperature record and that it is this cumulative effect of ENSO events that raises and lowers global Sea Surface Temperatures. Let’s examine that later in the post.
So that’s two ways, using sea surface temperature data, that the multidecadal rises and falls in global sea surface temperatures appear to be responses to the frequency and magnitude of El Niño and La Niña events.
HOW COULD THE OCEANS WARM WITHOUT AN INCREASE IN RADIATIVE FORCING?
Someone is bound to ask, how could the global Sea Surface Temperatures rise over multidecadal periods without an increase in radiative forcing? The answer is rather simple, but it requires a basic understanding of why and how, outside of the central and eastern tropical Pacific, sea surface temperatures rise and fall in response to ENSO events. Refer back to Figure 8, which includes the correlation maps from Trenberth et al (2002), and note that there are areas of the global oceans outside of the central and eastern equatorial Pacific that warm and cool in response to ENSO events. During an El Niño event, the warming outside of the eastern and central equatorial Pacific is greater than the cooling, and global SST anomalies rise.
But why do global SST anomalies rise outside of the eastern and central tropical Pacific during an El Niño event?
There are changes in atmospheric circulation associated with ENSO events, and these changes in atmospheric circulation cause changes in processes that impact surface temperatures. Let’s look at the tropical North Atlantic as an example. Tropical North Atlantic SST anomalies rise during an El Niño event because the trade winds there weaken and there is less evaporation. This is discussed in detail in the paper Wang (2005), “ENSO, Atlantic Climate Variability, And The Walker And Hadley Circulation.” Wang (2005) link:
http://www.aoml.noaa.gov/phod/docs/Wang_Hadley_Camera.pdf
Reworded, the reduction in trade wind strength due to the El Niño causes less evaporation, and since there is less evaporation, tropical North Atlantic sea surface temperatures rise. The weaker trade winds also draw less cool water from below the surface. So there are two effects that cause the Sea Surface Temperatures of the tropical North Atlantic to rise during El Niño events. And, of course, the opposite would hold true during La Niña events.
Again for example, during multidecadal periods when El Niño events dominate, the tropical North Atlantic trade winds would be on average weaker than “normal”, there would be less evaporation, less cool subsurface waters would be drawn to the surface, and tropical North Atlantic sea surface temperatures would rise. The western currents of the North Atlantic gyre would spin the warmer water northward. Some of the warm water would be subducted by Atlantic Meridional Overturning Circulation/Thermohaline Circulation, some would be carried by ocean currents into the Arctic Ocean where it would melt sea ice, and the remainder would be spun southward by the North Atlantic gyre toward the tropics so it could be warmed more by the effects of the slower-than-normal trade winds. Similar processes in the tropical South Atlantic also contribute to the warming of the North Atlantic, since ocean currents carry the warmer-than-normal surface waters from the South Atlantic to the North Atlantic.
Refer again to the correlation maps in Figure 8. Those are snapshots of monthly SST anomaly correlations. If those patterns were to persist for three decades due to a prolonged low-intensity El Niño event, global SST anomalies would rise. And the opposite would hold true for a prolonged La Niña event.
Let’s look at the average NINO3.4 SST anomalies during the three epochs of 1910 to 1944, 1945 to 1975, and 1976 to 2009. As shown in Figure 14, the average NINO3.4 SST anomalies were approximately +0.15 deg C from 1910 to 1944; then from 1945 to 1975, they were approximately -0.06 deg C; and from 1976 to 2009, the NINO3.4 SST anomalies were approximately 0.2 deg C. This is a very simple way to show that El Niño events dominated the two periods from 1910 to 1945 and from 1976 to 2009 and that La Niña events dominated from 1945 to 1975.
http://i56.tinypic.com/zxmsg8.jpg
Figure 14
Figure 15 compares annual Global SST anomalies to the average NINO3.4 SST anomalies for those three periods. Global SST anomalies rose from 1910 to 1944 because El Niño events dominated, and because the SST anomaly patterns (caused by the changes in atmospheric circulation) associated with El Niño events persisted. Because La Niña events dominated from 1945 to 1975, and because the SST anomaly patterns associated with La Niña events persisted, Global SST anomalies dropped. And Global SST anomalies rose again from 1976 to 2009 because El Niño events dominated, and because the SST anomaly patterns associated with El Niño events persisted.
http://i55.tinypic.com/33cwt4j.jpg
Figure 15
The fact that the rise in global Sea Surface Temperature anomalies since the early 1900s can be recreated without an increase in radiative forcing implies a number of things, one being that anthropogenic greenhouse gases do nothing more than cause a little more evaporation from the global oceans.
UPDATE – The following discussion (What Does The Running Total Imply?) has been added.
WHAT DOES THE RUNNING TOTAL IMPLY?
Earlier I wrote, Figure 13 [which was the comparison graph of global SST anomalies versus the running total of scaled NINO3.4 SST anomalies] appears to imply that 6% of each El Niño and La Niña event remains within the global surface temperature record and that it is this cumulative effect of ENSO events that raises and lowers global Sea Surface Temperatures. But is that really the case?
Keep in mind that the running total is a simple way to show the rise in global SST anomalies can be explained by the oceans integrating the effects of ENSO. It does not, of course, explain or encompass many interrelated ENSO-induced processes taking place in each of the ocean basins. Each El Niño and La Niña event is different and the global SST anomalies responses to them are different. For example, the South Atlantic SST anomalies remained relatively flat for almost 20 years, but then there was an unusual warming Of The South Atlantic during 2009/2010. Why? I have not found a paper that explains why South Atlantic SST anomalies can and do remain flat, let alone why there was the unusual rise. In this post, the gif animation of NINO3.4 SST anomaly correlation with North Atlantic SST anomalies, Animation 2, showed that the response of the North Atlantic can persist far longer than the El Niño or La Niña, but if I understand correctly, this type of analysis will emphasize the stronger events. What happens during lesser ENSO events? And there’s the East Indian and West Pacific Ocean. In January 1999, I began illustrating and discussing how the East Indian and West Pacific Oceans (60S-65N, 80E-180 or about 25% of the global ocean surface area) can and does warm in response to El Niño AND La Niña events. The first posts on this cumulative effect were Can El Nino Events Explain All of the Global Warming Since 1976? – Part 1, and Can El Nino Events Explain All of the Global Warming Since 1976? – Part 2. And the most recent post was La Niña Is Not The Opposite Of El Niño – The Videos. The Eastern Pacific Ocean is, of course, dominated by the ENSO signal along the equator. However, because of the North and South Pacific gyres, the East Pacific also influences and is influenced by the West Pacific, which can warm during El Niño and La Niña events. And there’s the Indian Ocean with its own internal variability, represented in part by the Indian Ocean Dipole (IOD). The decadal variability of the IOD has been found to enhance and suppress ENSO, and, one would assume, vice versa.
HOW MUCH OF THE RISE IN GLOBAL TEMPERATURES OVER THE 20TH CENTURY COULD BE EXPLAINED BY THE GLOBAL OCEANS INTEGRATING ENSO?
As shown in Figure 13 and as discussed in detail in the post Reproducing Global Temperature Anomalies With Natural Forcings, virtually all of the rise in global surface temperatures from the early 1900s to present times can be reproduced using NINO3.4 SST anomaly data. The scaled running total of NINO3.4 SST anomalies establishes the base curve and would represent the integration of ENSO outside of the eastern and central equatorial Pacific. Scaled NINO3.4 SST anomalies are overlaid on that curve to represent the direct effects of ENSO on the eastern and central equatorial Pacific. Add to that scaled monthly sunspot data to introduce the 0.1 deg C variations is surface temperature resulting from the solar cycle and add scaled monthly Stratospheric Aerosol Optical Depth data for dips and rebounds due to volcanic eruptions, and global surface temperature anomalies can be reproduced quite well. Refer to Figure 16, which is Figure 8 from the post Reproducing Global Temperature Anomalies With Natural Forcings.
http://i42.tinypic.com/2zqufzp.jpg
Figure 16
Basically, that was the entire point of this post. One of the mainstays of the anthropogenic global warming hypothesis is that there are no natural factors that could explain all of the global warming since 1975. But this post has shown that ALL of the rise in global sea surface temperatures since 1900 can be explained by the oceans integrating the effects of ENSO.
CLOSING
This post presented graphs and animations that showed Global SST anomalies rose and fell over the past 100 years in response to the dominant ENSO phase; that is, Global SST anomalies rose over multidecadal periods when and because El Niño events prevailed and they fell over multidecadal periods when and because La Niña events dominated. Basically, it showed that the oceans outside of the central and eastern tropical Pacific integrate the impacts of ENSO, and that it would only require the oceans to accumulate 6% of the annual ENSO signal (Figure13) in order to explain most of the rise in global SST anomalies since 1910. And the post provided an initial explanation as to why and how the global oceans could rise and fall without additional radiative forcings. It also showed that the Atlantic Multidecadal Oscillation (AMO) appears to be an exaggerated response to the dominant multidecadal phase of ENSO. Hopefully, it also dispelled the incorrect assumption that La Niña events tend to cancel out El Niño events.
SOURCES
The HADISST data used in this post is available through the KNMI Climate Explorer:
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere
The maps used in the video are available from the GISS map-making webpage:
http://data.giss.nasa.gov/gistemp/maps/
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
November 22, 2010 at 6:21 pm
The reason I referred to station records unadulterated by UHI is because
they provide the best available indication of multidecadal temperature
variations for the entire past century. (As you’re aware, pre-satellite
era SST data is not only very spotty, but fails to maintain consistent
datum levels due to inconsistent measurement methods.) The vetted station
records on average show the 1950-1979 period to be ~.15K below the 20th
century mean. The Nino3.4 index is no exception.
Getting a time-series well centered is essential for analytic work. It is
not a discretionary choice. Otherwise, one gets misleading estimates of
correlation functions, power spectra, etc. This is no less important in
integration over indefinite time. Only zero-mean functions yield
well-bounded integrals that do not increase or decrease idefinitely. What
you are doing by accepting Trenberth’s chosen–and clearly biased–“norm”
is introducing a positive trend into the cumulative sum of the otherwise
trendless record. That bogus trend is inversely proportional to the bias
of the “norm.”
Integration (cumulative summation) is a linear operator. Ipso facto, it is
incapable of producing multidecadal oscillations that are not there in the
record. The entire idea of higher-frequency components (the power spectrum
of the Nino3.4 record peaks at ~5.5a periods) producing multidecadal
oscillations upon integration is mathematically impossible. And the
striking thing is that cross-spectrum analysis with the (properly
recentered) HADCRUT3 series shows virtually no coherence at mutidecadal
periods and only marginal coherence at the Nino spectral peak. This is
also evident in your 31-yr smoothing comparison.
Why, then, does the cumulative sum of the miscentered Nino3.4 series so
strongly resemble HADCRUT3 visually over the available years? Can’t say
for sure, but I suspect that it’s a combination of a) strong coherence at
high frequencies b) bogus trend introduction and c) intra-record offsets in
both series, which are suspect on the basis of certain tests that I won’t
describe here. In other words, it’s a numerical oddity.
On the side-issue of cross-equatorial surface currents in the Atlantic, I read Lumpkin and Garzoli when they first published. There are two major weaknesses to their study: Eulerian inferences are made from Lagrangian (drifter) data and the effect of Amazonian discharge upon the dynamic topography along the NE facing coast of Brazil is not taken into account. Even so, if you read their paper carefully, you’ll find the Guyana Current questioned as a “true” current, with eddy ring shedding at the retroflection of the NBC into the ECC being offered as the tentative explanation. Inasmuch as both of these currents circulate primarily North Atlantic water in a narrow equatorial loop, this scarcely qualifies as the great conveyor of heat from SH to NH that you seem to suggest.
Thanks for the invite to continue the discussion tomorrow, but I have other commitments. Have a good Thanksgiving!
Cut and paste dropped Bob Tisdale’s name from the header of my last comment.
sky,
How can you say “producing multidecadal oscillations upon integration is mathematically impossible” when the integration is clearly showing multidecadal oscillations? (also when using Nino-data centered around zero)
lgl, I think sky is saying (data quality issues aside) that the low frequency component is present in the data. So in practical terms, integration is just helping to see it (for example if observers were blinded by the magnitude of interannual variations). In short, integration doesn’t create low-frequency oscillations that weren’t there in the first place. I again encourage everyone, including sky, to graph SOI (both raw & base-period-dependent anomaly) by month and to look at integrals by month (& by combination of months). It is very important to be aware of variations in the seasonal cycle. It appears to me that in their focus on the 1st moment, many folks are overlooking the second moment. It should not be assumed that a cycling average has random homoskedastic variance. Even worse than not stating assumptions, researchers are often not aware of their assumptions.
Bob, thanks for delivering posts that drive quality commentary.
sky: I’ll be addressing your comments in a follow-up post. Thanks.
Regards
Bob, l found time to read the 2 papers to which you linked here:
http://bobtisdale.blogspot.com/2010/11/guan-and-nigam-2008-and-2009.html
I remember a down-to-earth (even a bit red-neck) academic statistician once telling me that a new algorithm could secure a warm welcome if it was “cute”. I’m confident the EEOF thing would fit his “cute” criteria. While I’m a fan of factor analysis generally speaking, it is important to recognize that it runs straight into dead ends in many applications in nature (including terrestrial climate) due to (context-specific) untenable assumptions. Putting more bells & whistles on an algorithm (e.g. the upgrade from EOF to EEOF) cannot overcome obstacles stemming from fundamentally flawed (context-specific) assumptions. (Worse than not stating assumptions, many researchers are not even aware that they are making assumptions.)
Despite misapplication (to data for which assumptions are untenable), the EEOF algorithm yields worthwhile (but insufficient on its own) information and the articles (particularly the 2008 one [on the Pacific]) are stimulating. If I was running a graduate level university course, I would consider assigning the 2008 paper as required reading for an in-depth (2-3 hour) discussion.
Thanks for sharing the finds.
Bill Illis posted:
http://a.imageshack.us/img832/3174/newesthadsst3.png
These characters clearly do not pay attention to EOP (Earth orientation parameters) and the evolution of seasonal patterns in geomagnetic aa index. Worse than that, it’s like they don’t realize that clouds affect insolation – and that there is a relationship between wind & clouds. It is the responsibility of sensible people to find an efficient way to arrest the errant behavior of these vandals.
lgl says:
November 25, 2010 at 5:41 am
All linear operators are strictly frequency-preserving. An elementary example is the cosine function, which produces upon integration the sine function of EXACTLY the same frequency. That’s why I can say that integration of HIGH-FREQUENCY components of ENSO cannot produce MULTIDECADAL components. Only NONLINEAR operators can do that. Hope this clarifies matters for you.
sky says: “All linear operators are strictly frequency-preserving.”
And your assumption with respect to ENSO is?
Bob, I’m not making any dynamic assumptions whatsoever about what PHYSICALLY produces ENSO or how its effects may be redistributed around the globe. What I’m saying is that the strictly linear MATHEMATICAL operation of integration–continuous or discrete–cannot produce signal components at frequencies other than those already present in the signal. Incredible mathematics is hardly a good foundation for sound physics. Let’s leave it at that.
sky: in response to the relationship between forcing and temperature.
Given a forcing in units of Joules per second, acting on a mass with a heat capacity in units of Kelvin per Joule, the result of integration over time will be temperature in Kelvin.
“What I’m saying is that the strictly linear MATHEMATICAL operation of integration–continuous or discrete–cannot produce signal components at frequencies other than those already present in the signal.”
You should read Roe 2009. He says on pp106: “… variability at long periods can be the natural result of physical processes whose timescales are much shorter.” and again on pp108 “the vast majority of the variance in the PDO can be explained by simple integrative physics with a perhaps surprisingly short timescale.”
sky,
Thanks but it does not clarify matters. Running summation is low-pass filtering and the result clearly shows a low-frequency component. I don’t understand why you included the paragraph containing “it is incapable of producing multidecadal oscillations that are not there in the record” when there is a multidecadal oscillation in the record (other that to confuse of course).
sky is correct. It’s not the frequency content that changes with differentiation & integration, but rather the relative power (comparatively between timescales before & after differentiation &/or integration). [Also, note that sky has acknowledged the spatial dimension (“may be redistributed”), while restricting mathematical comment to the temporal dimension.]
I recommend that readers compare seasonal SOI integrals, noting in particular summer-winter divergences before 1959. This should shed some light on early 20th century differences between the integrals of NPI/ALPI & PDO. It may also help establish a framework for overcoming misconceptions about PDO vs. North Pacific SST.
Strongly suggested:
Carefully study & compare the following:
a) http://icecap.us/images/uploads/AMOTEMPS.jpg
b) Figure 10:
Carvalho, L.M.V.; Tsonis, A.A.; Jones, C.; Rocha, H.R.; & Polito, P.S. (2007). Anti-persistence in the global temperature anomaly field. Nonlinear Processes in Geophysics 14, 723-733.
http://www.uwm.edu/~aatsonis/npg-14-723-2007.pdf
http://www.icess.ucsb.edu/gem/papers/npg-14-723-2007.pdf
Anyone reproducing Bob’s 31-year-difference series should compare their results with the differenced series (plain monthly differences) (a) smoothed at 31 year bandwidth and (b) repeat 1 year smoothed (don’t hesitate to jack the repetition way up…)
This discussion has raised many items which have only peripherally received the attention they deserve. I look forward to Bob’s follow-up post.
David Stockwell linked to:
Roe, G. (2009). Feedbacks, timescales, and seeing red. Annual Review of Earth and Planetary Sciences 37, 93-115. doi: 10.1146/annurev.earth.061008.134734.
http://sheridan.geog.kent.edu/geog41066/7-Roe.pdf
‘Irreducible complexity’ slogans are too lazy. We don’t yet know enough about terrestrial climate for this approach to avoid killer problems with Simpson’s Paradox. The era of meaningful statistical inference will necessarily follow the era of careful data exploration.
David Stockwell:
You overlook the fact that Tisdale is integrating not the forcing, but its effect–the temperature. As for the frequency-preserving property of integration, that is a mathematical theorem. If Roe (2009) knows what he is talking about (always an open question in climate science), he will not make any claims in violation of it. In my earlier response to you, I already mentioned the impulse response function of linear systems that acts as a convolution kernel in the integral defining the response. Not all integration is the same! He may have in mind that sort of integration or NONlinear systems that indeed can create harmonics and subharmonics. Unless Roe can provide a concrete model, however, his words are just vague hand-waving.
lgl:
See additional comments above about frequency preservation by linear operators. The Nino3.4 index, of course, does have some multidecadal components (accounting for ~8% of its total variance) but they are incoherent with those of HADCRUT3 (where they account for ~75%), and no LINEAR operation will make them coherent. I don’t know how to make this crucial point any clearer. P.S. My time is much too valuable for me to ever waste it on generating confusion.
sky,
Wrong again, after 1900 they are coherent too. Both peak around 1940 and 2000 and trough around 1910 and 1975. The different variance is interesting but there is no rule saying the high-frequency components can not be more damped than the low-frequency on a global scale.
sky says: “You overlook the fact that Tisdale is integrating not the forcing, but its effect–the temperature. ”
I believe I know where our views differ. You’re looking at NINO3.4 SST anomalies as numbers. I view them as a proxy for a process. That process releases warm water from below the surface of the PWP, shifts it to the central and eastern equatorial Pacific, releases heat there through evaporation, which causes changes in atmospheric circulation, in turn causing SST outside of the tropical Pacific to vary. The process continues when a Rossby wave returns leftover warm water from the eastern to the western tropical Pacific during the subsequent La Nina……….
lgl says:
December 1, 2010 at 3:15 am
Cross-spectral coherence has little to do with when extreme values are achieved in a wide-band stochastic process such as ENSO. Having already demonstrated extraordinary difficulty in grasping signal analysis basics (you seem to think that low-pass filters CREATE low-frequency components) , it comes as no surprise that you totally misunderstand the concept of frequency-dependent coherence. I recommend Papoulis’ monograph “Signal Analysis” for getting a solid grasp. Good luck!
Bob Tisdale says:
December 1, 2010 at 3:43 pm
Bob, it is you who cumulatively summed NINO3.4 temperature anomalies (miscentered at that) and concluded on the basis of numerical similarity that they “cause” the global temperature variations seen in the HADCRUT3 anomalies. That scarcely conveys viewing ENSO as a physical process. All I did was point out the insurmountable logical lapses in drawing that conclusion on that basis.
sky
Cross-spectral coherence has little to do in this discussion. Cross-spectral coherence would mean SST is driven by ENSO alone and nobody ever claimed that of course.
The main point is the 60 years cycle in both ENSO and SST. SST detrended looks very similar to the integral of ENSO. Do you think that’s just a coincidence?
My “Running summation is low-pass filtering and the result clearly shows a low-frequency component … there is a multidecadal oscillation in the record”
can’t possibly be turned into “you seem to think that low-pass filters CREATE low-frequency components” If you don’t have an argument, leave it.
Bob Tisdale wrote (addressing sky), “I believe I know where our views differ. You’re looking at NINO3.4 SST anomalies as numbers. I view them as a proxy for a process.”
In following the exchange, that has been my instinct as well. To express it one way, Bob is treating SOI (an indicator of equatorial Pacific wind & cloud) as a proxy for the rate of change of global temperature. It’s no surprise that the integral of f'(x) equals f(x) plus a constant. Similarly, the integral of global atmospheric angular momentum will reveal the 1976 climate shift. I disagree strongly with sky’s decision to view ENSO as a largely stochastic process. However, sky is correct to point out that in a strictly temporal domain, integration does not create the low frequency pattern from high frequency components. How circulation aliases high frequency equatorial Pacific components onto nonstationary spatiotemporal modes elsewhere and what pattern variations emerge with variable aggregation criteria is another matter.
lgl:
The entire point of Tisdale’s presentation is to show a COHERENT relationship between the cumulative sum of Nino3.4 anomalies and global SST anomalies, not just the presence of UNRELATED multidecadal oscillations in both series. But that involves series with different units– in other words, a numerological/phenomenological rather than a physical comparison. Yet he does claim a causal relationship. That’s where I, who usually applauds his presentations, try to get him to consider the issue more rigorously. Cross-spectral coherence underlies the essence of his claims. See more below.
Paul Vaughn:
Nino3.4 is a localized temperature index, not a proxy for anything else. And your reservations about treating ENSO as stochastic processes are misplaced. Physical processes that involve turbulence or other deterministic chaos DO produce time-histories that are stochastic in their variability. A commonplace example is wind-driven ocean waves. Analyzing the records via spectral methods is eminently sensible in unraveling the complexities and–in particular–seeking out DETERMINISTIC physical relationships between pairs of series. That’s where cross-spectral coherence becomes absolutely crucial. Without strong coherence, no claimed physical relationship is credible.
As an empirical matter, the multidecadal components of Nino3.4 are INCOHERENT with those of HADCRUT3, which I use as an overall temperature index in cross-spectrum analysis. This tells us that, no matter how important ENSO is as a mechanism for redistributing the excess heat of the equatorial Pacific, it is NOT the mechanism that drives the multidecadal oscillations of global temperatures. Integration or cumulative summation cannot change that.
Sorry, gentlemen, I simply cannot take more time for this discussion.
Paul:
As a footnote, if ENSO were indicative of the differential of global temperature, then the cross-spectral phase (Nino3.4 – HADCRUT3) would be 90 degrees. In fact it is not, even at the 3-4yr periods where the coherence is strongest.
sky: “Sorry, gentlemen, I simply cannot take more time for this discussion.”
I’ll second that. Time to move on to additional ways to illustrate the cumulative effects of ENSO.
Bob,
can’t stop the fun now
sky
Thanks for your time, sorry you left but I have to add a few comments.
“Nino3.4 is a localized temperature index, not a proxy for anything else”
It’s localized but still probably a proxy for a global phenomena. The wind anomalies revers all over the globe when switching from Nino to Nina
http://www.crces.org/presentations/dmv_ipwp/images/figure7.gif from http://www.crces.org/presentations/dmv_ipwp/
and because this phenomena is not coherent around the globe, North-atlantic SST lags North-pacific SST about one year for instance (Bob can tell you all about other lags) , a cross-spectral coherence is impossible. In addition things like large volcanic eruptions are not visible in Nino3.4 but very visible in global SST.
“then the cross-spectral phase (Nino3.4 – HADCRUT3) would be 90 degrees. In fact it is not, even at the 3-4yr periods where the coherence is strongest.”
For the strong 3.6 yr period it is very close to 90 degrees (again, it can’t be expected to be exact) SST’ usually leads Nino3.4 about one year. http://virakkraft.com/SST-deriv-ENSO-1880-2010.png