Rob Painting of SkepticalScience recently published the post A Looming Climate Shift: Will Ocean Heat Come Back to Haunt Us? Many of you will find his post entertaining.
But this post is not about the SkepticalScience post. This post discusses the recent paper Rob Painting cited: Meehl et al (2013) “Externally forced and internally generated decadal climate variability associated with the Interdecadal Pacific Oscillation”. The abstract is here, and a preprint copy of Meehl et al (2013) is here. You’ll note that Kevin Trenberth is one of the authors. Meehl et al (2013) is an update of Meehl et al (2011) Model-based evidence of deep-ocean heat uptake during surface-temperature hiatus periods.
Since Meehl et al (2013) refer to the Interdecadal Pacific Oscillation, we’ll also discuss it—what it is and what it is not.
Last, this post includes a model-data comparison of the trends in satellite-era sea surface temperatures for the Pacific versus the outputs (ensemble mean and individual ensemble members) of the CCSM4 model used by Meehl et al (2013).
MEEHL ET AL (2013)
Meehl et al (2013) is basically a climate model-based study of the effects of El Niño and La Niña events on surface temperatures and ocean heat content. Like a couple of other recent studies, Meehl et al (2013) are suggesting that, during periods when La Niñas dominate, surface temperatures do not warm, but ocean heat content continues to warm at depths below 700 meters. Meehl et al (2013) are also suggesting that global surface temperatures warm during periods when El Niños dominate. I was not able to find any suggestion in Meehl et al (2013) that the heat driven below the depths of 700 meters would somehow come back to haunt us as some alarmists suggest nonsensically. Meehl et al (2013) are simply trying to explain why surface temperatures warm during some decadal periods and cool (or remain flat) during others—and the mechanism they propose is the El Niño-Southern Oscillation, but they’re using the Interdecadal Pacific Oscillation (IPO) as a proxy.
The problem: Meehl et al (2013) are using a climate model (NCAR CCSM4) that doesn’t properly simulate the coupled ocean-atmosphere processes that drive El Niños and La Niñas. As far as I know, there are no climate models that properly simulate El Niño and La Niña processes. We’ve discussed this numerous times, and as a reference on those occasions, I’ve presented Guilyardi et al (2009) Understanding El Niño in Ocean-Atmosphere General Circulation Models: progress and challenges. Guilyardi et al (2009) is a detailed overview of the problems climate models have in their attempts at simulating ENSO, and it cites more than 100 other papers. I introduced Guilyardi et al (2009) in my post here. Later in this post, we’ll also present one of the very obvious failures that occur because models can’t simulate El Niño and La Niña processes.
The differences between the modeled Interdecadal Pacific Oscillation spatial patterns in sea surface temperatures from Meehl et al (2013) and as determined from Pacific sea surface temperature data can be seen in my Figure 1. The three left-hand maps in my Figure 1 are Figure 5 from Meehl et al (2013). The caption for Meehl et al (2013) reads:
The first EOF of 13 year low pass filtered SSTs for the Pacific Ocean domain shown from a 300 year period of the unforced CCSM4 control run (the model’s IPO); b) first EOF of 13 year low pass filtered SSTs for the ensemble mean RCP4.5 simulations from CCSM4 indicating the dominant forced response pattern; c) same as (b) except for second EOF.
So cell a is the dominant sea surface temperature pattern in the Pacific when the models aren’t forced by greenhouse gases, and cell b is the dominant sea surface temperature pattern in the future Pacific when the models are forced by greenhouse gases. To the right are the observed sea surface temperature patterns for the Pacific, using the HADISST dataset, from Meehl and Hu (2006). Neither the unforced or forced spatial patterns of the models are the same as the observed spatial pattern in sea surface temperatures.
In fact, the unforced pattern in cell a of Meehl et al (2013), the upper left hand map in my Figure 1, resembles an El Niño Modoki pattern as presented by JAMSTEC. Figure 2 presents the El Niño and El Niño Modoki EOF maps from JAMSTEC for the period of 1979 to 2004. See their El Niño Modoki webpage here. So the most dominant pattern in the unforced sea surface temperatures of the models resembles the second most dominant pattern in the observations as presented by JAMSTEC.
It’s curious that the dominant pattern of the NCAR CCSM4 climate models used by Meehl et al (2013) is not the one created by East Pacific El Niño events. Why? East Pacific El Niños are what caused the sea surface temperatures of the Atlantic, Indian and West Pacific Oceans to warm in upward shifts. The El Niño events of 1987/88 and 1997/98, which caused the surges in global temperatures in Trenberth’s “big jumps” graph, were East Pacific El Niños, not El Niño Modoki events.
[Trenberth’s “big jumps” graph was discussed here. And El Niño Modoki events were discussed here and here. You’ll note in the latter post that the 1986/87/88 El Niño starts as an El Niño Modoki but ends as an East Pacific El Niño. That shift in El Niño type suggests that a secondary downwelling Rossby wave (a second burst of warm water from the Pacific Warm Pool) drove the El Niño farther to the east and created a more powerful El Niño during 1987, after the initial El Niño season of 1986/87.]
I also find it curious that Meehl et al (2013) presented the Interdecadal Pacific Oscillation-related spatial patterns for the control runs and for the future forced runs, but not for a hindcast period. If the models performed well at simulating the past, one would assume those maps would also have been presented.
And as noted in the post Open Letter to the Royal Meteorological Society Regarding Dr. Trenberth’s Article “Has Global Warming Stalled?”, even if La Niña events are causing the additional warming below the depths of 700 meters as suggested by the Meehl et al (2013) models, it’s sunlight, not infrared radiation from manmade greenhouse gases, that create that additional warm water in the tropical Pacific during La Niñas. Kevin Trenberth understands that very well. In fact, it was one of the topics discussed in Trenberth et al (2002) Evolution of El Nino–Southern Oscillation and global atmospheric surface temperatures. There, he and others wrote (my boldface):
The negative feedback between SST and surface fluxes can be interpreted as showing the importance of the discharge of heat during El Niño events and of the recharge of heat during La Niña events. Relatively clear skies in the central and eastern tropical Pacific allow solar radiation to enter the ocean, apparently offsetting the below normal SSTs, but the heat is carried away by Ekman drift, ocean currents, and adjustments through ocean Rossby and Kelvin waves, and the heat is stored in the western Pacific tropics. This is not simply a rearrangement of the ocean heat, but also a restoration of heat in the ocean.
In that post, I also presented a comparison graph of downward shortwave radiation (sunlight) versus downward longwave (infrared) radiation for the period of 1979 to present, along the equatorial Pacific (5S-5N, 120E-80W), based on the NCEP-DOE Reanalysis-2. See Figure 3. The downward longwave radiation (red curve) serves as a reasonable proxy for the El Niño and La Niña events. Notice that downward longwave (infrared) radiation decreases during La Niña events. That of course means its variations do not support the model-based supposition that infrared radiation is creating the additional warm water that is being forced below depths of 700 meters. In the real world, it’s downward shortwave radiation (sunlight) that increases during La Niña events.
The following section is for those concerned that I’m discussing El Niño and La Niña events while Meehl et al (2013) are discussing the positive and negative phases of the Interdecadal Pacific Oscillation.
INTERDECADAL PACIFIC OSCILLATION
Like the Pacific Decadal Oscillation dataset, the Interdecadal Pacific Oscillation (IPO) is an abstract form of the sea surface temperature data of the Pacific Ocean. Seemingly, there are as many definitions of the Interdecadal Pacific Oscillation as there are papers about it. Since this post is about Meehl et al (2013), we’ll used the latitudes they presented for the IPO in their Figure 5, which were 40S-60N, and use the longitudes of 120E-80W to minimize the influence of other ocean basins. Afterwards I’ll compare the “official” IPO index data to the IPO data I’ve created. There’s little difference in the timing of the changes from positive to negative IPO modes.
While the Pacific Decadal Oscillation data is derived from the sea surface temperatures of the extratropical North Pacific, for this presentation the Interdecadal Pacific Oscillation is determined from the data for most of the Pacific basin (40S-60N, 120E-80W). In the Pacific Decadal Oscillation data, global surface sea temperatures are subtracted from the North Pacific data. This does not make sense for an examination of Pacific data. The global data obviously includes the other ocean basins. Subtracting the data from the other ocean basins would leave us with a difference that no longer represents the Pacific Ocean. It leaves a difference influenced by the Atlantic, Indian, Arctic and Southern Oceans. So I’ve simply had the Pacific data detrended before the statistical analysis.
Empirical Orthogonal Function (EOF)/Principal Component (PC) analysis is used to determine the dominant pattern in the Pacific sea surface temperature anomalies. That is, for this discussion, the Interdecadal Pacific Oscillation represents the dominant spatial pattern of the sea surface temperature anomalies in the Pacific basin. See Figure 4. The map was created at the KNMI Climate Explorer. It uses HADISST data and is based on the period of 1900 to 2012.
Since El Niño and La Niña events are the dominant mode of natural variability in the Pacific (and the globe), the Interdecadal Pacific Oscillation basically presents an East Pacific El Niño-caused spatial pattern as shown above in Figure 4.
The data for the Interdecadal Pacific Oscillation represents how strongly the spatial pattern in the Pacific at a given time portrays (resembles) the El Niño-like spatial pattern. That’s why the year-to-year variations in the Interdecadal Pacific Oscillation data mimic NINO3.4 sea surface temperature anomalies, as shown in Figure 5. (NINO3.4 sea surface temperature anomalies are a commonly used index for the strength, frequency and duration of El Niño and La Niña events.) For two climate-related datasets, they agree quite strongly, with a correlation coefficient of 0.93.
But it’s also easy to see in my Figure 5 that there are subtle differences in the annual variations. We can highlight these minor differences by smoothing both datasets with121-month running-average filters. See Figure 6. Even with the decadal smoothing, the variations agree quite well. The additional variability of the Interdecadal Pacific Oscillation data is likely caused by the effects of changes in wind patterns (and the interdependent changes in sea level pressure) outside of the tropical Pacific that are not related to El Niño and La Niña processes. Changes in the direction and speeds of winds also impact the spatial pattern of sea surface temperatures.
Of course, some of the differences could also be caused by the uncertainties of the data. Keep in mind, sea surface temperature data are reconstructions—that is, we’re not examining raw sea surface temperature data—and numerous corrections/adjustments have been made that can impact these results.
Note: The IPO is presented, again depending on the paper, using a number of different low pass filters. I’ve used a 121-month running mean because it’s easy to understand and easy to duplicate, for those wishing to confirm the results.
And as shown in Figure 7, if we change the ENSO index from NINO3.4 sea surface temperature anomalies (5S-5N, 170W-120W) to the sea surface temperature anomalies for the Cold Tongue Region (6S-6N, 180-90W), we get slightly different results on the timing of the changeovers from positive to negative IPO. So the results are also dependent on the ENSO index.
Bottom line, as the comparisons show in Figures 6 and 7, for the most part, during periods when El Niño events dominate, the Interdecadal Pacific Oscillation data are positive and when La Niñas dominate, the Interdecadal Pacific Oscillation data are negative. It was those minor differences that interested researchers, because they influenced rainfall patterns in Australia.
As you’ll note in Figures 5 through 7, there are no units in those graphs. That is, they do not represent temperature in units of deg C. The reason: the two datasets have been standardized, divided by their standard deviations, so that they’re easier to compare. I’ve included Figure 8 as a reference. It compares NINO3.4 sea surface temperature anomalies to the 1st Principal Component of the detrended sea surface temperatures of the Pacific (Interdecadal Pacific Oscillation data) without the standardization. Based on their standard deviations, the variations in the NINO3.4 data are about 8 times stronger than the Interdecadal Pacific Oscillation data.
As noted earlier, there is an Interdecadal Pacific Oscillation dataset available on the internet. The version updated in 2008 is here. It’s based on the 1999 paper by Power et al Inter-decadal Modulation of the Impact of ENSO on Australia. Unfortunately, it ends in 2008.
To provide an up-to-date version, I used HADISST sea surface temperature data for the IPO in Figure 5 through 8. And as you’ll note, Meehl and Hu (2006) also used HADISST, so it’s a commonly used sea surface temperature dataset for these types of analyses. The problem: the HADISST data is infilled using EOF analysis. So the IPO data based on HADISST data also reflects, in part, the infilling method.
Ideally, I would have used HADSST3 data because it is not infilled. But that creates another problem because the HADSST3 data is not spatially complete—there are grids without data or gaps in the data—and this is especially true in the South Pacific. This presents a problem when performing EOF/PC analyses at the KNMI Climate Explorer. The default setting for the analysis requires that certain percentage of the grids contain data. For the months that don’t reach that threshold, the KNMI Climate Explorer leaves a blank, no output. That leaves numerous gaps in the Interdecadal Pacific Oscillation data during the 1940s and little data prior to the 1920s.
There are subtle differences between the Interdecadal Pacific Oscillation dataset I created and the “official” Interdecadal Pacific Oscillation data. I’ve compared them in Figure 9 for your information. Regardless, when El Niños dominate, the IPO is generally positive and when La Niñas dominate, the IPO is generally negative. And as discussed earlier, the differences are likely caused by the other factors that influence the ENSO-related spatial patterns in the Pacific.
And of course, as shown in Figure 10, the yearly variations in the “official” Interdecadal Pacific Oscillation data mimic the variations in NINO3.4 sea surface temperature anomalies—our ENSO index.
The document that contains the “official” Interdecadal Pacific Oscillation data includes the following statement:
The physical nature of the IPO is under investigation; it is still not clear, despite the above studies, to what extent the IPO is really independent of ENSO red noise and especially of SST variations near a decadal time scale where some physical processes have been identified.
Shakun and Shaman (2009) Tropical origins of North and South Pacific decadal variability examined the data in the North and South Pacific using EOF analyses. The conclusions of Shakun and Shaman (2009) reads (Note PDV stands for Pacific Decadal Variability):
Deriving a Southern Hemisphere equivalent of the PDO index shows that the spatial signature of the PDO can be well explained by the leading mode of SST variability for the South Pacific. Thus, PDV appears to be a basin-wide phenomenon most likely driven from the tropics. Moreover, while it was already known PDV north of the equator could be adequately modeled as a reddened response to ENSO, our results indicate this is true to an even greater extent in the South Pacific.
To sum up this section, the Interdecadal Pacific Oscillation is simply another El Niño- and La Niña-related index. When smoothed with decadal (or longer) filters, it portrays the decadal and multidecadal variations in the dominance of El Niño and La Niña events. The variations in the Interdecadal Pacific Oscillation index and an ENSO index (NINO3.4 sea surface temperature anomalies) agree quite well on annual and decadal time scales.
Unfortunately, many persons misunderstand what the Interdecadal Pacific Oscillation represents.
WHAT THE IPO DOES NOT REPRESENT
While the Interdecadal Pacific Oscillation data is derived from the sea surface temperature anomaly data of the Pacific Ocean, it does not represent the sea surface temperature anomalies there. The Interdecadal Pacific Oscillation data also does not represent the multidecadal variations in the sea surface temperatures of the Pacific Ocean. To confirm that, Figure 11 compares the “official” Interdecadal Pacific Oscillation data to detrended sea surface temperature anomalies of the Pacific Ocean (40S-60N, 120E-80W). Note again that the data in Figure 11 have been standardized, which greatly exaggerates the variations in both.
MODEL-DATA COMPARISON – SATELLITE-ERA PACIFIC OCEAN SEA SURFACE TEMPERATURE TRENDS
The sea surface temperatures of the Pacific Ocean warm in a specific fashion. El Niño events release naturally created warm water from below the surface of the tropical Pacific and ocean currents carry the leftovers toward the poles (and into the tropical Indian Ocean). During multidecadal periods when La Niña events dominate, less warm water is released to the surface by El Niños and less warm water is carried toward mid-latitudes at the surface in their wake. On the other hand, during multidecadal periods when El Niño events are dominant, more warm water than normal is released by El Niños to the surface and carried by ocean currents toward mid-latitudes. This can be seen by looking at the warming trends in sea surface temperatures for the Pacific Ocean on a latitudinal (zonal mean) basis.
El Niño events dominated the satellite era of sea surface temperature data, and because the Pacific has complete coverage during that period, we’ll use satellite-era sea surface temperature data in the following model-data presentation. The Reynolds OI.v2 data includes data from satellites, ship inlets and buoys (fixed and floating).
Let’s first discuss the type of graph being used in the comparisons. See Figure 12. The (vertical) y-axis is scaled in deg C per decade, and it indicates the warming rates, based on the linear trends, for the term of the satellite-era Reynolds OI.v2 sea surface temperature dataset. The (horizontal) x-axis is latitude. The units range from the South Pole (-90 or 90S) on the left all the way to the North Pole (90 or 90N) on the right. “0” latitude is the equator. The data ends at Antarctica and at the Bering Strait.
Figure 12 presents the trends of the sea surface temperature data for its term, which starts in November 1981 and runs to present, May 2013. As shown, there is little to no warming for much of the tropical Pacific—from 10S to about 20N. In fact, much of it cooled slightly over the past 31+ years. The high latitudes of the South Pacific (south of 50S) and into the Southern Ocean also show cooling. We can also see the aftereffects of the El Niño domination during this period. The mid-latitudes of both hemispheres show warming. Again, that’s caused by the leftover warm water from the El Niños is being carried by ocean currents to the mid-latitudes. That residual warm water is carried to regions called the South Pacific Convergence Zone in the Southwest Pacific and the Kuroshio-Oyashio Extension (KOE) in the Northwest and Central North Pacific.
In Figure 13, I’ve added the ensemble mean (the average of the 6 model runs) of the NCAR CCSM4 models stored in the CMIP5 archive. (The NCAR CCSM4 was the climate model employed by Meehl et al (2013) for their study. And CMIP5 is the archive being used by the IPCC for its upcoming 5th Assessment Report.) Those model outputs are available to the public in easy-to-use formats through the KNMI Climate Explorer.
As shown, the warming rates of the average of the model outputs show no resemblance to how the Pacific actually warmed. That indicates a number of things: First, the models aren’t simulating El Niño and La Niña events properly. If the models were simulating them properly, there would be no warming of sea surface temperatures along the equator. Second, the models are not transporting leftover warm water (from the El Niños) from the tropics to the mid-latitudes properly. That’s especially true in the Southern Hemisphere, where the modeled warming rates at mid-latitudes are comparable to those in the tropics. That severely undermines the conclusions reached by Meehl et al (2013). If the models show no skill at the transport of warm water from the tropics to the mid-latitudes on the surface, we have no reason to believe their results at depths below 700 meters.
Additionally, the models basically show how the Pacific sea surface temperatures would have warmed IF (big if) they were warmed by manmade greenhouse gases. Phrased another way, it indicates that modelers are still trying to force the warming of the oceans with greenhouse gases, while the data indicate that greenhouse gases were not involved in the warming of the oceans—sea surface temperatures or ocean heat content. If the natural warming of the global oceans is new to you, refer to my illustrated essay “The Manmade Global Warming Challenge” [42MB].
Animation 1 includes comparison graphs of the data to the six individual ensemble members and to the ensemble mean. Not one of the ensemble members captures the warming pattern correctly. Ensemble Member 2 (EM-2) shows some similarities, but the warming rates are way too high for most of the Pacific basin. And based on how poorly the other members simulate the warming rates, climate modelers are still a long way from being able to provide any information about the warming of the oceans—a long way away. And that’s after decades of modeling efforts. (You may need to click start the animation.)
Climate scientists are scrambling to explain why surface temperatures have not warmed since the 1997/98 El Niño. Meehl et al (2013) is yet another example. Fundamentally, Meehl et al (2013) is climate model-aided speculation. It is a failed attempt to explain how and why global surface temperatures cool (or remain flat) when La Niña events dominate and to explain that the warming continues but it occurs out of sight and out of mind below the depths of 700 meters.
Curiously, while Meehl et al (2013) depends on El Niño and La Niña processes, they never use the terms El Niño or La Niña in the paper. They elected to present the effects of ENSO on the Pacific basin in an abstract form, the Interdecadal Pacific Oscillation. The problem with that: If the climate models show no skill at being able to simulate the processes of ENSO on short-term bases, there’s no reason to believe they perform any better on decadal or multidecadal bases.
Another curiosity: Meehl et al (2013) presented the outputs of 5 ensemble members of the CCSM4 model using RCP4.5 forcings. But the CMIP5 archive contains 6 ensemble members with those forcings. Why did they exclude one? We illustrated how poorly all six ensemble members simulate the warming rates in the Pacific on a zonal mean basis, so why they excluded one is somewhat of a mystery.
Climate modelers still cannot simulate the coupled ocean-atmosphere processes associated with El Niño and La Niña events. And they are still trying to force El Niño and La Niña processes with longwave radiation from greenhouse gases, while in reality the processes are fueled by sunlight. One of the authors of Meehl et al (2013), Kevin Trenberth, actually “wrote the book” on ENSO, and he understands the processes well. Yet somehow he has lent his name to a paper that presents an alternative to that reality. That is, Meehl et al suggest that manmade global warming continues but it’s being driven to depths below 700 meters by La Niñas. But the warm water created during La Niñas results from an increase in sunlight, not infrared radiation.
All of these factors dictate that studies such as Meehl et al (2013) have no value. Meehl et al (2013) employed fatally flawed models in an attempt to determine how and why the warming of surface temperatures has stalled and they employed those same fatally flawed climate models in an effort to determine where Kevin Trenberth’s missing heat might have gone—which assumes the missing heat even exists.
Come to think of it, we have not found a variable that climate models simulate properly. That is, in recent months we’ve also illustrated and discussed that the climate models stored in the CMIP5 archive for the upcoming 5th Assessment Report (AR5) cannot simulate observed:
So the value of climate models in any research effort is questionable.