Guest post by Bob Tisdale
GLOBAL AND PACIFIC OCEAN SEA SURFACE TEMPERATURE
OVERVIEW
This series of posts examines the differences between multi-model mean of the IPCC 20C3M (Hindcast)/SRES A1B (Projection) data and the Reynolds OI.v2 Sea Surface Temperature (SST) anomalies, a satellite-based SST dataset. In addition to times-series graphs, the linear trends of the Sea Surface Temperature anomalies are presented on a latitudinal (Zonal Mean) basis.
Part 1 looks at the Global SST anomalies and the SST anomalies of the Pacific Ocean.
The post includes discussions of the El Niño-Southern Oscillation (ENSO). If ENSO is new to you and if you’re interested in learning more about it, refer to the post An Introduction To ENSO, AMO, and PDO – Part 1.
INTRODUCTION
Climate change-related blogs present comparisons of IPCC Model Mean hindcast and projection data to show how well the models have performed versus the observed data. Real Climate has done this on an annual basis for the last few years. Refer to Updates to model-data comparisons and 2010 updates to model-data comparisons for examples. And Lucia, at her blog The Blackboard, compares the models to the observations with her monthly updates per dataset. Refer to RSS: Drop from 0.052C to -0.026C. Global temperature data, along with other datasets, are presented in those posts.
But as we know, the globe does not warm uniformly. Land surface temperatures vary at different rates than sea surface temperature. And some parts of the globe are warming faster than others. In fact, since the start of the Reynolds OI.v2 SST dataset, there have been decreases in SST anomalies for many portions of the global oceans. For this reason, I’ve elected to also examine the model mean hindcast/projections for Sea Surface Temperature data on an ocean-basin basis and on a zonal-mean (average temperature per 5-degree latitude band) basis.
Note: Kevin Trenberth provided a good overview of the IPCC models in Nature’s Climate Feedback: Predictions of climate post. It is a worthwhile read, even as a refresher.
This post is not about that post at Climate Feedback or its author so please refrain from any comments along those lines.
GLOBAL AND PACIFIC TIME SERIES GRAPHS
Let’s look at the time-series graphs first.
Figure 1 compares the Global (90S-90N) Reynolds OI.v2 SST anomaly data from January 1982, the start of that satellite-based dataset, and the model mean for the IPCC 20C3M (Hindcast)/SRES A1B (Projection) TOS (Sea Surface Temperature) data. The obvious difference is, the models do not show the yearly variations caused by the El Niño-Southern Oscillation (ENSO). This is to be expected based on the Trenberth post in Nature. And if I’m looking at this correctly, ensemble members for each model are averaged and so are the models in turn to provide the multi-model mean. With all of that averaging, the multi-model mean for ENSO signals would be greatly smoothed.
There is a significant difference in the linear trends. The linear trend for the models is about 50% higher than the observed trend for Global SST anomalies.
http://i55.tinypic.com/x3y4xs.jpg
Figure 1
If you were to click on the Monthly scenario runs link to the KNMI Climate Explorer and scroll down to the top of the table, you’d note that both SST and TOS are shown. “TOS” is identified as “(Sea surface temperature) in K {time mean}” in the descriptions, while “SST” is identified as Sea Surface Temperature with some modifications. I believe the difference is how they treat sea ice. But just in case you’re concerned those definitions are somehow responsible for the difference between the observed SST data and the models in Figure 1, I’ve eliminated the polar data in Figure 2. It compares the observed and modeled Global SST data from 60S-65N. The model trend is still about 50% higher than the observations.
http://i56.tinypic.com/8ww51e.jpg
Figure 2
The linear trend of the satellite-based SST observations for the North Pacific, Figure 3, are about half that of the models.
http://i52.tinypic.com/zjxxxy.jpg
Figure 3
The disparity is greater in the South Pacific. Refer to Figure 4. There the linear trend of the models is almost 2.5 times higher than the trend of the SST data.
http://i55.tinypic.com/34q1sh0.jpg
Figure 4
Dividing the Pacific data into their East and West components is very revealing. As shown in Figure 5, the linear trends for the SST observations and models in the West Pacific are nearly identical.
http://i55.tinypic.com/zn91sp.jpg
Figure 5
That means the major difference between the observations and the models in the Pacific exists in the East Pacific. Refer to Figure 6. The linear trend of the IPCC Hindcast/Projection is more than 6 times higher than the nearly flat Sea Surface Temperature observations.
http://i53.tinypic.com/2h4xrwh.jpg
Figure 6
SST ANOMALY TREND COMPARISONS ON ZONAL MEAN BASES
Zonal Mean graphs offer a different perspective. As you’ll note in Figure 7, the y-axis is temperature, same as the time-series graph. But the x-axis is latitude. The zonal mean data in the post are based on the average SST anomalies for the latitude bands of 80S to 75S, then 75S-70S, etc., from pole to pole. And the graphs present the linear trends of the SST data for those latitude bands in Deg C/Decade. The data starts in January 1982 and ends in February 2011. As we can see in Figure 7, the trends of the models are higher at the equator than they are at mid-latitudes. The trends then drop off to near zero at high latitudes.
http://i53.tinypic.com/219uhhx.jpg
Figure 7
The model data basically follows this pattern in all ocean basins. Refer to Figure 8, which compares the trends for the zonal mean SST anomalies for the Atlantic, Indian, and Pacific Oceans. Note that I’ve further divided the Indian and Pacific Oceans into their respective east and west portions. There are some differences, primarily at the mid latitudes of both hemispheres, but in general the same overall patterns exist in all basins.
http://i56.tinypic.com/t4wpys.jpg
Figure 8
-HOWEVER-
That’s not how the global SST anomalies have risen.
Figure 9 compares linear trends of the Global SST observations and IPCC hindcasts/projections on a zonal mean basis. In the Southern Hemisphere, the SST observations are, for the most part, cooling south of 50S. The Southern Hemisphere trends then peak in the mid latitudes before dropping significantly in the tropics. The trends of the observations then rise again from the tropics to the high latitudes of the Northern Hemisphere. Poleward of the peak near 60N, the observations drop quickly to near zero. The sea surface temperature models appear to have no basis in reality.
http://i53.tinypic.com/wjt82o.jpg
Figure 9
Recall in Figure 8 how, on a latitudinal basis, the trends of modeled SST anomalies were similar for all ocean basins. Figure 10 presents the same comparison with the Reynolds OI.v2 SST data. As shown, the ocean basin SST anomalies have warmed very differently since 1982. The North Atlantic has very high trends at high latitudes. This should be a function of the Atlantic Multidecadal Oscillation (AMO) which is not visible in the IPCC model data. (That will be presented in part 2 of this post.) Note the significant negative trend in the Eastern Tropical Pacific.
http://i53.tinypic.com/24zf4f9.jpg
Figure 10
Let’s compare the observed and the modeled trends (on a zonal mean basis) for the West Pacific (125E-180). Refer to Figure 11. The models miss the significant drop in Southwest Pacific SST anomalies, south of 45S. The model mean trends are similar to the observations from there until the mid latitudes of the Northwest Pacific. But the models also miss the significant warming of the Northwest Pacific between 30N-45N, the location of the Kuroshio-Oyashio Extension (KOE). I discussed the processes that cause the rise in the SST anomalies for the KOE in The ENSO-Related Variations In Kuroshio-Oyashio Extension (KOE) SST Anomalies And Their Impact On Northern Hemisphere Temperatures. While the models may not reproduce the recent ENSO variability, should we expect to see evidence of ENSO-related processes outside of the eastern equatorial Pacific?
http://i54.tinypic.com/165950.jpg
Figure 11
And should we expect to see evidence of the ENSO process in the eastern Pacific zonal mean data? That question relates to the process of ENSO, not the timing or magnitude of the modeled ENSO events. As illustrated in Figure 12, Eastern tropical Pacific SST anomalies have dropped since the start of the Reynolds OI.v2 SST anomaly data, with the greatest cooling along the equator. The models, on the other hand, show a high trend, one that is comparable to those of the other ocean basins (Refer back to Figure 8).
http://i53.tinypic.com/kb6kk2.jpg
Figure 12
Keep in mind that the eastern equatorial Pacific is only a temporary home of the warm water associated with El Niño events. At other times, the eastern equatorial Pacific is one of the largest areas of upwelling in the global oceans. Again, this temporary warming in the eastern Pacific happens during El Niño events, when warm water sloshes east from the surface and below the surface of the Pacific Warm Pool. An El Niño does not “exhaust” all of the warm water, and what remains has to go somewhere at the end of the El Niño. Where does it go? It goes back to the western Pacific during the La Niña. Some of the warm water helps to recharge the Pacific Warm Pool for the next El Niño. The rest remains on the surface and finds its way, most noticeably, to the Kuroshio-Oyashio Extension (KOE). For these reasons, we would expect the West Pacific trends to be considerably higher than the East Pacific in the tropics and mid latitudes of the Northern Hemisphere. We see this in the observations, Figure 13.
http://i54.tinypic.com/zjgcg1.jpg
Figure 13
But we do not see it in the model mean data, Figure 14. In fact, the opposite happens in the models. In them, the Eastern tropical Pacific rises faster than those in the Western tropical Pacific.
http://i53.tinypic.com/25at6yw.jpg
Figure 14
BUT THE SST DATA STARTS WITH AN EL NIÑO AND ENDS DURING A LA NIÑA
The Reynolds OI.v2 SST dataset starts just before the significant 1982/83 El Niño and ends during the peak of the 2010/2011 La Niña. For those who are concerned that starting the comparisons on an El Niño and ending them on a La Niña has biased this presentation, I’ve redone a few of the zonal mean comparison graphs, model versus observations. I’ve ended the data in December 2009, at the peak of the 2009/10 El Niño, in the trend comparisons of the zonal mean data for the East and West Pacific, Figures 15 and 16, and for the global data, Figure 17. The data now runs from El Niño to El Niño. It does not change the results to any significant extent.
http://i56.tinypic.com/15g6v78.jpg
Figure 15
#############################
http://i51.tinypic.com/5otws1.jpg
Figure 16
#############################
http://i53.tinypic.com/14l36o5.jpg
Figure 17
CLOSING
In part 2, we’ll examine the Indian and Atlantic Oceans.
As illustrated in this post, the IPCC modeled SST data have significantly higher trends than the satellite-era SST observations for Global and Pacific Oceans.
Also illustrated, there are few similarities in the zonal mean comparisons for those datasets. In other words, the models appear to have little basis in reality, at least since 1982.
SOURCE
The data presented in this post is available through the KNMI Climate Explorer:
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere
A nice compilation of the available data…however, I’d like to suggest that the meaningful discrepancy between the models and data is not the difference in linear trend, but the much larger fluctuations in the “real” data.
The fluctuations in the actual data contradict the over-simplistic notion that “CO2 is the singular control knob that determines our climate” and also indicate how little of the Earth’s natural variation is really understood. Without this understanding, we have no idea how much of the warming signal is CO2 vs. natural, and correspondingly the climate models are basically meaningless.
“…the models appear to have little basis in reality…”
There you go. Climate “science” in a nutshell…
In Burrito says: “…I’d like to suggest that the meaningful discrepancy between the models and data is not the difference in linear trend, but the much larger fluctuations in the ‘real’ data.”
Keep in mind the timing of ENSO events in the models vary per ensemble member for each model that attempts to model it. The ensemble members are averaged for each model. Then the models are averaged for the model-mean data, so the ENSO variability in the models is going to be suppressed.
However, it appears based on the model mean that the typical model treated ENSO as noise (not a process) back when they compiled these. It also appears the modelers missed the constant upwelling of waters in the eastern equatorial Pacific (except during El Nino events), which is why there is no trend in NINO3.4 SST anomalies since 1900.
I’ll address multidecadal variability in part 2.
Regards
It staggers the imagination to consider that modelers would be willing to present their results in public. I would not let my high schooler present this material in a science fair. What have modelers been doing all these years? Why are they still funded?
Model means won’t have ENSO signals. Individual realizations are where you would have to look for ENSO.
Think of it this way Bob. A given model run might have an ENSO signal arise in
year X. ENSO isnt programmed in, it emerges. Another run of that model may have
ENSO emerge at year X+2. A third at year X -3. Another model may have it emerge
at year X+10, X+8.. and so on for 20 models and 50 runs. So the mean of the
models will just smooth it all away. The models cant match the time of ENSOs but
the test would be do they get the frequency and amplitude correct. For that you
have to look at individual realizations.
Having spent a (small/short) bit of time in my past career attempting to develop statistical models of complex systems, why am I not surprised at these results. They are typical of trying to model a system which you do not understand.
Thank you for a clear and logical presentation of the data. And, I would have to agree with you in that “the models appear to have little basis in reality, at least since 1982”.
Looks like the modelers need to get bigger, faster, better computers ’cause it can’t be the models surely /sarc
Very nice presentation, Bob. Easy enough for just about anyone to follow, even me ;o)
In Burrito beat me to it; it really sticks out like a sore thumb how conservative the models are compared to observations, let alone the difference in trend lines. And then there are just too many instances where the model zigs and the observations zag. If they zigged and zagged together, then I’d say the models were doing something right, but they don’t. It seems like it’s almost an accident that both the models and the observations are producing positive trends.
I wonder how the models would plot against car sales of Honda? About the same, I’d bet.
steven mosher at April 11, 2011 at 8:52 am: Thanks for the clarification and the confirmation of what I had written in the post. I had prepared a model versus data comparison of NINO3.4 SST anomalies for this post, but decided against including it.
I’ll likely include one in a post comparing the models to HADISST for a long-term comparison. There the discussion of ENSO would be more appropriate, because the HADISST NINO3.4 SST anomalies have no trend since 1900, while the long-term (and short-term) NINO3.4 trend in the models is similar to the rest of the tropics.
steven mosher says:
April 11, 2011 at 8:52 am
“Model means won’t have ENSO signals. Individual realizations are where you would have to look for ENSO…
So the mean of the models will just smooth it all away. The models cant match the time of ENSOs but the test would be do they get the frequency and amplitude correct. For that you have to look at individual realizations.”
You say that like it is a good thing. To me, it screams “not ready for prime time.” All the models that I work with are used to make decisions; that is, each model run yields definitive information that can serve as the basis for one or another important decision. Can you specify some decisions that can be made on the basis of results from these model runs?
Keith Battye says:
April 11, 2011 at 9:23 am
Looks like the modelers need to get bigger, faster, better computers ’cause it can’t be the models surely /sarc
Nah… they can compensate by making bigger fudge factors: “… Kuroshio found the rate of energy dissipation within the boundary layer to be enhanced by 10 to 20 times …” http://www.sciencemag.org/content/early/2011/03/14/science.1201515.abstract
“This series of posts examines the differences between multi-model mean of the IPCC 20C3M (Hindcast)/SRES A1B (Projection) data and the Reynolds OI.v2 Sea Surface Temperature (SST) anomalies, a satellite-based SST dataset. In addition to times-series graphs, the linear trends of the Sea Surface Temperature anomalies are presented on a latitudinal (Zonal Mean) basis.”
Bob, hope you don’t think the models you see created for IPCC are to represent the real reality, they each have their own reality. ENSO is just one of their many tools.
As mosher said here: “Think of it this way Bob. A given model run might have an ENSO signal arise in year X. ENSO isnt programmed in, it emerges. Another run of that model may have ENSO emerge at year X+2. A third at year X -3. Another model may have it emerge at year X+10, X+8.. and so on for 20 models and 50 runs. So the mean of the models will just smooth it all away. The models cant match the time of ENSOs but
the test would be do they get the frequency and amplitude correct. For that you have to look at individual realizations.”, models can create the ENSO event any year so with enough models, all events disappear (except for the upward trend that is).
/sarc
Thanks, Bob, for an excellent and informative post. Yet more confirmation that the models can’t even hind-cast, so predicting future outcomes is not even on the map!
Your charts also show the futility of trying to use linear trends to get a signal by averaging and smearing the detail, when the system is non-linear and driven by spatio-temporal chaos. In such circumstances it is easy to cherry-pick the trend you want to see just by choosing the time period which gives the best result. Cargo-cult science at its best!!!
Reminds me of the new rotary combustion concept, nice models and cartoons! I’m sure all that is needed is a few hundred million or a few billion or so to get these things to work.
What are the merits of modelling, again? To refute real data? /sarc
wayne says: “models can create the ENSO event any year so with enough models, all events disappear (except for the upward trend that is).”
Except there is no upward trend in the SST of the eastern equatorial Pacific.
Averaging the trend over the full 30 years makes the models look much better than they are (or less hopeless). The trend shows an obvious discontinuity at 1998. If you split the trend into 1980 to 1998 and 1998 to 2010, the models are too low for the first period and totally wrong for the second period. The models predict aproximately the same trend for both periods, but the measurements show no increase in the second period.
Bob Tisdale says:
April 11, 2011 at 12:59 pm
wayne says: “models can create the ENSO event any year so with enough models, all events disappear (except for the upward trend that is).”
Except there is no upward trend in the SST of the eastern equatorial Pacific.
Eastern Pacific? Oh, my, Bob. You’re right. Must be in the algorithms.
—> More grants needed here boys! We’ve got one flattening out! <—
🙂
If I were building a model I would want to be able to determine if it could accurately hindcast. As such I would have as input real event dates that the models could use instead of randomly selecting the dates. I would include volcanoes and ENSO events at a minimum. Would be interesting to see what they got with this kind of information fed into them.
Speaking of ENSO, these guys are abysmally ignorant when they treat ENSO as noise. It is not noise but a real physical oscillation of ocean water from shore to shore that has a period of four to five years. Even those who have written about it like Hansen and Trenberth are ignorant of this fact. It shows when Hansen talks of a “permanent El Nino condition” that supposedly existed in the Pliocene. That is a classic oxymoron because the El Nino is part of an oscillation which cannot be made to stand still. An El Nino is physically a mass of warm water from the Indo-Pacific Warm Pool that crosses the ocean from near New Guinea to South America as a Kelvin wave. It follows the equatorial countercurrent and right smack in the middle of that countercurrent sits Nino 3.4 and watches all those El Ninos go by. The lag time between Nino 3.4 signal and classic El Nino symptoms is due to the fact Nino 3.4 sees it before it has hit the coast. As it hits the coast its warm water splashes ashore, spreads out north and south, and warms the air. Warm air rises, stops the trade winds (not the other way around as you have been told), mixes with global circulation, and is carried across the continent by prevailing winds. That is when our temperature goes up. But any wave that runs ashore must also retreat. As the El Nino retreats water level behind it drops by half a meter or more, cool water from below wells up to fill this gap, and a La Nina has started. As much as the El Nino raised global temperature the La Nina will lower it by an equal amount. This temperature oscillation has an amplitude of about half a degree and is visible in all temperature curves if some idiot did not wipe it out with a running mean. This massive periodic heat exchange between the ocean and the atmosphere is very precise as satellite records from the eighties and nineties demonstrate. IPCC so far has been ignorant of this. The dynamical theory I described is mine and you can find out more about it by reading “What Warming?” available from Amazon.
Thanks Bob.
If one goes way out into the future with the climate model projections, we see it just a formula based on the projected GHGs (no matter how much Steve Mosher explains that results spontaneously erupt out of the models).
In the future, they are not models, they are formulae.
The monthly Multi-model mean under A1B from 1900 to 2100 (note the formulae predict an increase in the seasonal cycle over time – haven’t heard that one before but there it is right in the numbers).
http://img811.imageshack.us/img811/4146/ar4ensemblemeanvshadcru.png
Note, I’ve added an update to the cross post at my blog:
http://bobtisdale.wordpress.com/2011/04/10/part-1-%e2%80%93-satellite-era-sea-surface-temperature-versus-ipcc-hindcastprojections/
The update reads:
I wasn’t happy with the “El Niño to El Niño” years selected under the heading of “BUT THE SST DATA STARTS WITH AN EL NIÑO AND ENDS DURING A LA NIÑA”, the heading directly above. The 1982 El Niño was significantly stronger than the 2009/2010 El Niño. This is very apparent if we look at annual NINO3.4 SST anomalies, Figure 18. The annual data, however, does allow us to determine the two ENSO neutral years that are closest to the beginning and end of the Reynolds OI.v2 SST dataset. They are 1986, with a NINO3.4 SST anomaly of +0.111 deg C, and 2005, at +0.114 deg C. Can’t get much closer than that, but the use of those years would shorten the time span of the data considerably.
http://i52.tinypic.com/16ow20.jpg
Figure 18
Figure 19 shows the monthly NINO3.4 SST anomalies over that period.
http://i52.tinypic.com/w9k66g.jpg
Figure 19
So I plotted the zonal-mean comparisons (model versus data) for the East Pacific, West Pacific and Global data once again, Figures 20, 21, and 22. The curves have changed somewhat, but the differences are still extreme.
http://i51.tinypic.com/20qjn08.jpg
Figure 20
#############################
http://i53.tinypic.com/6o3q0g.jpg
Figure 21
#############################
http://i53.tinypic.com/2ebfm6w.jpg
Figure 22
Bob (and Steve Mosher)
The models are hindcast/projections. When does the hindcast finish and the projection start? Is the hindcast part being ‘trained’ by real world data?
HR: Sorry. The hindcast was for the 20th Century. It ran from 1900 to 2000. And I don’t believe there’s any “training”. Aren’t the models driven by the forcings?
Arno Arrak says:
April 11, 2011 at 5:34 pm
“Speaking of ENSO, these guys are abysmally ignorant when they treat ENSO as noise. It is not noise but a real physical oscillation of ocean water from shore to shore that has a period of four to five years. Even those who have written about it like Hansen and Trenberth are ignorant of this fact. It shows when Hansen talks of a “permanent El Nino condition” that supposedly existed in the Pliocene. That is a classic oxymoron because the El Nino is part of an oscillation which cannot be made to stand still. An El Nino is physically a mass of warm water from the Indo-Pacific Warm Pool that crosses the ocean from near New Guinea to South America as a Kelvin wave.”
Love this post. It goes right to the heart of the matter. ENSO is physical. And no one will really understand ENSO until they have a set of physical hypotheses that describe the regularities that make it up. Ah, there is the rub. Warmista just do not do physical hypotheses. One begins to suspect that they are in climate science, so-called, because they could not do physical hypotheses in some genuine science.
wayne says:
April 11, 2011 at 11:05 am
“Bob, hope you don’t think the models you see created for IPCC are to represent the real reality, they each have their own reality. ENSO is just one of their many tools.”
Wouldn’t it be more hip to say that each has its own way of expression or manifestation or…Maybe they are caught in a flashback to the Sixties.
In Burrito says: “…I’d like to suggest that the meaningful discrepancy between the models and data is not the difference in linear trend, but the much larger fluctuations in the ‘real’ data.”
If you look at Figure 1 for example, the natural data shows a much larger deviation in the signal than does the model data. This is the really significant difference. Natural variability is much greater than the models are predicting.
Yes, it is significant that the models predict greater warming than observed. But even more significant is that they are predicting too little variability. The climate models are under estimating natural climate variability. SO, when they see natural variability, they don’t recognize it as being natural, because it doesn’t mathc the models.
In other words – the scientists are asssuming the models are “true nature”, and “observed nature” is un-natural.
fred;
Yes, Mother Un-Nature is a beech!
/grin
Arno Arrak says: “It follows the equatorial countercurrent and right smack in the middle of that countercurrent sits Nino 3.4 and watches all those El Ninos go by.”
Your description assumes an Eastern Pacific El Niño. Many El Niño events are Central Pacific El Niño, also known as El Niño Modoki.
You wrote, “As it hits the coast its warm water splashes ashore, spreads out north and south, and warms the air.”
Not if it’s a Central Pacific El Niño. Also, most of the heat loss from the tropical Pacific Ocean during an El Niño is through evaporation. The warming of the atmosphere takes place when that moisture then condenses and turns to rain. While that’s going on, the change in location of convection in tropical Pacific (from the Pacific Warm Pool to the central or eastern equatorial Pacific) alters “normal” atmospheric circulation patterns, and sea surface temperatures outside of the tropical Pacific warm due to those changes in atmospheric circulation. The rise in sea surface temperature, in turn, causes the rise in sea air temperature. All of those “changes” take a few months to make their way eastward around the globe, and that’s what is responsible for much of the lag.
You wrote, “Warm air rises, stops the trade winds (not the other way around as you have been told), mixes with global circulation, and is carried across the continent by prevailing winds.”
During the ENSO-neutral phase before the El Niño, the trade winds in the western tropical Pacific hold the warm water in place in the Pacific Warm Pool. Keep in mind that the sea surface height during the ENSO-neutral (and La Nina) phase is measurably higher in the western tropical Pacific than it is in the east due to those trade winds. The trade winds west of the dateline have to relax first in order for gravity to carry the warm water to the east, the result of which is the El Niño.
You wrote, “As much as the El Nino raised global temperature the La Nina will lower it by an equal amount.”
That is a myth. First, there are few La Niña events that match the strength of the El Niño before it. (An exception was the 2007/08 La Niña. It was stronger than the 2006/07 El Niño.)
Second, I have found no other instances during the satellite era when global temperatures drop proportionately during a La Niña. Look at the 1997/98 El Niño. As NINO3.4 SST anomalies drop from their El Niño peak to the La Niña trough, do global temperatures follow the NINO3.4 signal completely? No. If you can illustrate this for the transition from the 1997/98 El Niño to the 1998/99 La Nina, please do.
Third, the reason global temperatures don’t drop completely is because a La Niña is not the opposite of an El Niño. You described how the warm water travels along the Pacific equatorial countercurrent during the evolution of an El Niño. But as I wrote in the post, the El Niño does not “exhaust” all of the warm water that had traveled east. When the El Niño is done, where does all of the leftover warm water go, Arno?
You wrote, “IPCC so far has been ignorant of this.”
Many of the papers referenced by the IPCC also perpetuate the myth of “La Nina will lower it [global temperature] by an equal amount”. You’re just following their lead.
You concluded with, “The dynamical theory I described is mine and you can find out more about it by reading ‘What Warming?’ available from Amazon.”
An ad? I replied to an ad? Now I know how Ralphie felt in “A Christmas Story” with the decoder ring.
Theo Goodwin says: “Love this post. It goes right to the heart of the matter. ENSO is physical. And no one will really understand ENSO until they have a set of physical hypotheses that describe the regularities that make it up.”
Unfortunately, much of Arno’s ENSO hypothesis is wrong. See my comment above this.
HR
Bob Tisdale says:
April 11, 2011 at 6:59 pm
HR: Sorry. The hindcast was for the 20th Century. It ran from 1900 to 2000. And I don’t believe there’s any “training”. Aren’t the models driven by the forcings?
########
I will try to explain how the models work and why its difficult to compare them against observations over a 30 year period.
The models are driven by forcings. They are not trained on data. The first thing that happens is the model is Spun up till it reaches an equillibrium state. So, for example, you set the forcings to the values they have in 1900 and then you run the model till it reaches equillibrium. That is, for example the temperature is holding steady.
Think about what has to be happening in the ocean for this to be the case. No up and down cycles. Its steady state.
Now, ask yourself, was the REAL ocean in 1900 in this state? Nope. The only thing we can do is have the forcings set to what we THINK their values were in 1900.
Then, the model is run forward and year after year forcings change and the climate evolves only certain lines. Ocean cycles arise or emerge as time marches forward.
Their timing in the model cant match the timing in the real world. The only think you can hope to match is the average amplitude and the frequency. So you get one realization of climate from one model run. It has ups and downs that correspond to the emergence of ocean cycles and a general trend upward as a result of forcings.
The timing of volcanic events can be accurate. But the timing of ocean cycles can only match those in the real world by chance. Models typically have 1-5 runs and there are abou 20 or so models. Now average all of them together and the variability due to ocean cycles will just go away, IFF you average over long enough periods, say at least twice the average cycle length of the longest cycle. what will be left is a trend line of warming due to forcings.
To capture actual cycles, you would have to initialize the model in a non steady state configuration. That would mean initializing the ocean and the whole atmosphere to the real state of the world. The science is not at that state yet. It cant do what you testing it for. Wasnt designed to do what you are testing it for.
During the next big el nino the real trend will come back to the model trend lines.
AND the cycles in the smoothed (average) ensemble mean will still “miss” the actual ups and downs, but it will capture the long term trend. Thats because the long term trend is determined by EXTERNAL forcing and not by internal oscilation.
steven mosher says: “I will try to explain how the models work and why its difficult to compare them against observations over a 30 year period.”
Again, thanks for the clarification, but…
I used a satellite-based SST dataset (Reynolds OI.v2) for this post because it has the least amount of infilling. That limits the time period to the last (less than) 30 years. It would have been much easier to present only the time-series graphs, but I included the zonal-mean plots to show that the models do not properly represent the distribution of the warming (and cooling in the real world) in the tropics or mid latitudes or high latitudes.
I could just as easily have started this series of posts with HADISST, a long-term SST dataset, and started the comparisons in 1900. Referring to the zonal means plots, the models still do not represent reality on a latitudinal basis in the West Pacific…
http://i53.tinypic.com/4lou4k.jpg
…or in the East Pacific:
http://i51.tinypic.com/219ri2f.jpg
The models do not represent reality over a 30-year period or a 110-year period.
Bob Tisdale says:
April 12, 2011 at 3:56 am
“Unfortunately, much of Arno’s ENSO hypothesis is wrong. See my comment above this.”
I was making the point that El Nina is a physical phenomenon, not a creature solely of statistics or models. Do you agree? If so, are the Warmista treating is as something other than a physical phenomenon?
steven mosher says:
April 12, 2011 at 9:51 am
“I will try to explain how the models work and why its difficult to compare them against observations over a 30 year period.”
“Then, the model is run forward and year after year forcings change and the climate evolves only certain lines. Ocean cycles arise or emerge as time marches forward.
Their timing in the model cant match the timing in the real world…”
“To capture actual cycles, you would have to initialize the model in a non steady state configuration. That would mean initializing the ocean and the whole atmosphere to the real state of the world. The science is not at that state yet. It cant do what you testing it for. Wasnt designed to do what you are testing it for.”
If you are willing to say the things you say above, why are you not willing to say that climate science, at least as it is found in your models, is in its infancy and cannot be used to make predictions of global warming, climate change, or whatever it is you are calling it at the moment? To make the point crystal clear, why are you not willing to say that it must mature before it can be used as a guide for policy makers?
Theo Goodwin says: “I was making the point that El Nina is a physical phenomenon, not a creature solely of statistics or models. Do you agree?”
ENSO is a process, yes.
You continued, “If so, are the Warmista treating is as something other than a physical phenomenon?”
There are a number of papers that describe how well (or poorly, depending on your viewpoint) ENSO is treated in the models used for the IPCC AR4. Here’s a link to one, as a starting point for you:
http://www.knmi.nl/publications/fulltexts/guilyardi_al_bams09.pdf
As I understand, they are improving the models, but I have not gone through each model to determine if they represent ENSO processes as I have documented in a number of posts. Since they are outdated, I’m not sure what I’d learn from such an exersize.
Bob Tisdale says:
April 12, 2011 at 5:06 pm
Thanks much!
Theo:
“If you are willing to say the things you say above, why are you not willing to say that climate science, at least as it is found in your models, is in its infancy and cannot be used to make predictions of global warming, climate change, or whatever it is you are calling it at the moment? To make the point crystal clear, why are you not willing to say that it must mature before it can be used as a guide for policy makers?”
You make several confused points.
1 why are you not willing to say that climate science, at least as it is found in your models, is in its infancy and cannot be used to make predictions of global warming, climate change, or whatever it is you are calling it at the moment?
A. it’s not in it’s infancy. We’ve known for quite some time that Adding C02 to the atmosphere will warm the planet, NOT cool it. The question has always been how much and what are the feedbacks. It was in its infancy over 100 years ago.
B. It can be used to make predictions. the question is Predictions of WHAT and
how accurate are those predictions. So, if you ask me can it predict how warm it will
be in 50 years or 100 years, Of course it can. It does in fact make predictions.
The question is how good are those predictions and how much trust should
we put in them. Suppose you live 10 miles from me. Suppose I tell you to
drive your car over to my house. can I model how long it will take you? Sure,
its easy. T = D/R. Since I know its surface streets between me and you, I’ll
predict that it will take you more than 10 minutes to get here, and probably less
than 1 hour. For some purposes that is enough accuracy.
“To make the point crystal clear, why are you not willing to say that it must mature before it can be used as a guide for policy makers?””
Very simple. See the example above. A model does not have to capture EVRYTHING about a process to be useful to making decisions. So I use my little model to predict that it will take you 30 minutes +-20 minutes to get here. I Know that there are stoplights along the way, but my model ignores them. It ignores how long you would
spend at them, ignores your speeding up and your slowing down. It ignores the wind, it ignores many things. Now if you compare what my model predicts for your travel every second, you’ll see that it misses a whole bunch of detail. But, in the end, I predict 30 minutes and it takes you 25.
Do I care that it missed by 5 minutes? Nope, I just needed to know if you could get here faster than 1 hour. So,
the REAL question is what kind of accuracy do we need to make decisions. That’s the first question you need to ask. I can build a model of a missile hitting a plane. That model can be super detailed, but in the end I just need to know, does it destroy the plane. I can do that with a simple model or a very complex detailed physics model where every bit of metal is simulated. Again, the important question is not about
the ability of the model to determine which bit of metal cuts which fuel line leading to
an explosion .5 seconds after impact, the important question is what are we interested in measuring and how well do we need to measure it to answer the specific question we are asking.
If I were to criticize models like GCMs it would be for THIS feature, which is really a problem in the PROCESS of using models.
If you believe that you have said something other than nonsense then I simply do not know what to say.
Part 2 of this post is available here:
http://bobtisdale.wordpress.com/2011/04/19/492/
or the WUWT cross post here:
http://wattsupwiththat.com/2011/04/19/satellite-era-sea-surface-temperature-versus-ipcc-hindcastprojections-part-2/
Awesome post. There’s a lot of good information right here, however I would like to let you know something – I’m running Fedora with the circulating beta of Firefox, and the design of your web site is kind of bizarre for me. I could understand the content, however the navigation does not work so great.