I understand some anonymous commenters at other blogs are under the impression the graph below, created by JohnV, and I discussed here long ago, proves that the recent 0 C/century (or less by some reporting agencies) 92 month trend is not inconsistent with a mean trend of 2C/century:
See those big temperature swings?
Those who think those swings “just happen”, or might be caused by El Nino or La Nina have probably never compared the big dips to the periods when major stratospheric volcanic eruptions occurred. These were discussed a while back and shown below:
You can scan up and down and begin to notice that, yes, as we have all heard, major stratospheric volcanic eruptions affect the global mean surface temperature. In fact, part of the proof that GCM’s are able to predict some variations in GMST arising from variations in forcings is that they do predict these the dip in temperature after eruption like Pinatubo.
The phenomenology is understood: After a major stratospheric volcanic eruption, the stratosphere fills with aerosols. Radiative forcing decreases, and the global surface temperature drops for a while. Not surprisingly, 7 year trends with their start dates roughly 7 years or less before the eruption tend to dip.
Eventually, the aerosols start to clear. Radiative forcing increases; global surface temperatures rise. Not surprisingly, 7 year trends with their end dates roughly 7 years after the eruption tend to rise.
Oddly, though these dips are smoothed out when a 22 year average is applied, it is the case if we were to a GCM model that accounts for the eruption many, many times, the dip around the volcanic eruption would not average out.
To show that the major excursions in 7 year trends in JohnV’s graph really are due to volcanic eruptions, I’ve reproduced it, but I changed 7 years to 92 months, and added the dates some major volcano eruptions. As a bonus, I’ve also indicated the know “jet-inlet-bucket transition” year– 1945. That dip is thought to be due to measurement error.
Look closely now. Focus on the period from about1916-1944. During that period, the stratospheric aerosol loading was not oscillating widely. Also, the method of measuring sea surfaces is thought to be well characterized. (In contrast the plunge in 1945 is thought due to errors.)
Now, notice only one small excursion outside the ±2C/century band during the period not affected by volcanic eruptions of the “jet-inlet bucket” error period in 1945? (And that excursion is from a 22 year average, not an ensemble average over more than 50 model runs. So, the ±2C/century bands aren’t entirely equivalent.)
Of course there is variability in 92 month trends even when stratospheric aerosols aren’t being hurled into the sky. But the variability is smaller than at other times. This is important when assessing whether a the current dip is remarkable.
When I say the current dip is inconsistent with ±2C/century, I say that, fully aware that this flat line trend is not caused by a volcanic eruption in, say January 2008. Had a stratospheric volcano erupted and hurled as much dust into the stratosphere as Pinatubo, then the current flat to negative trend in GMST would not be remarkable.
The reality is this dip is the only one that cannot be explained by the great “jet-inlet to bucket error incident around 1945” or an volcanic eruption known to have thrown large amounts of aerosols into the stratosphere. (And for what it’s worth, the obvious relationship between temperature variations and volcanic eruptions is also why I think it inappropriate to estimate ARMA(1,1) parameters for the current period based on GMST observations since 1975. Much of the variability during that period was due to El Chicon, Fuego and Pinatubo. Pretending that level of variability would be expected after 2001 shows a wanton disregard for phenomenology.)
what about the period approx. 1935 to 1940…seems to be a downward (perhaps not quite as precipitous as the current) trend prior to the inlet bucket temp issue (or am I mistaken here in the period affected by said issue?)
Thanks, great stuff and your point seems to makes sense RE “wanton disregard for phenomenology”.
SteveG–
On this graph, we are seeing trends– those are derivatives of temperature plotted. The key question with regard to “the argument” about hypothesis testing is: Do the seven year trends stick far out from the 22 year trends.
So, on this graph, the relevant thing is the “pop over” around 1935.
The temperatures did dip after 1935. The bucket-jet inlet error seemed to be in the direction of accentuating an ongoing dip. Had it been the other direction, we wouldn’t see that “plunge” that causes the 7 year trend to penetrate the 22 year trend around 1948. (The combination of the eruption of Hekla, the error and natural weather causes may have colluded to make this plunge particularly dramatic! )
That drop in temperature is partly real. The land record shows a drop in the 40’s also, and that’s not affected by the “bucket-jet inlet”. However, the land record doesn’t have quite the drama. That dip doesn’t stand out as huge relative to other ups and downs.
Since I would like to estimate variability during a non-volcanic eruption era, it’s unfortunate that happened.
Lucia, an excellent post as usual. Its interesting how this observation of the temperature data appears to be ignored by a lot of those who argue that 7 year dips in temperature series are not unusual. To me it speaks buckets.
I bet that if another El Chicon or Pinatubo, or heaven help both, where to occur tomorrow, and temperatures were to inevitably fall, they would be the first to yell “volcano”.
Oh, but its OK to estimate ARMA (1,1) parameters for the current period based on temperature observations in periods which included volcanos.
Now while I’m the 1st to admit that 7-8 years is not enough to call a trend change, it should however be a something of a flag to current official (read IPPC) thinking that perhaps, just perhaps, something is seriously & significantly amiss with the current models & theories of how earth’s climate works.
Lucia, unless you are willing to present a graph where you can show temperatures and corrections to wipe out volcanic eruption influences, you are merely stating correlations. Correlation doesn’t imply causation, and while you can imply the causation that volcanoes do dip the temperature, it’s not evident how much and if that will change Tamino’s conclusions, or not. Until I see those questions answered, I can’t see how anyone can point to your blog’s analysis for a “debunking” of Tamino’s point.
Luis Dias-
1) The argument that volcanoes cause the dips is rather well accepted. Modelers include the forcing in their runs because the causal effect is well accepted. However, here is a graph showing a hindcast representing the average of 5 runs of model E. You can see Pinatubo seems to “cause” a drop of between 0.2C and 0.5C, and the “deviation” from linear would have a substantial autocorrelation:
It is impossible for this to have no effect on derived ARMA(1,1) parameters.
2) Debunking? I haven’t suggested this post debunks Tamino. This post is motivated not by Tamino, but by people flashing around JohnV’s graph as “proof” about the expected variability of 7 year trends now.
At the end of the post– in comments at Tamino’s and in Tuesday’s post, I said I question Tamino’s analytical choice on physical grounds. So, the flip side to my not debunking Tamino is that Tamino has not debunked the idea that the current flat trend is unusual. The reason he has not is he picked a period with large volcanic eruptions, and it’s well accepted the eruptions make a difference.
It’s not like I’m complaining he didn’t account for cosmic rays when doing his ARMA fit.
Of course I agree with you that the magnitude of the effect should be quantified. I’ve begun to quantify the difference. I re did Tamino’s numbers using a “volcano light” period from Dec. 1913-1944. That was the post just before this one. 🙂
Here’s the link:
http://rankexploits.com/musings/2008/what-happens-if-we-assume-weather-noise-is-arma11/
My time period includes that rise after little dip in Model E projectoins in the mid 1910s. Model E associates that with Kusdah.
I didn’t compare and contrast the my uncertainty bounds with Tamino’s yet. Tamino ran GISS and I wanted to check HadCrut first. I also wanted to introduce my readers to the issue, and focus on describing how we get the magnitude of the parameters. Some of my readers asked my opinion on that method.
FWIW, I have run the numbers. For GISS, the variability of 7 year trends based on an ARMA(1,1) fit during Tamino’s period is substantially larger than during the non-volcano period. I’d tell you the numbers now, except I need to drink coffee, check the files etc. So, you’ll get the numbers when I organize that and double check them. (I do recall the difference is a factor of about 1/3rd or so).
In fact, the modelers specifically explain that if a volcano erupts, we shouldn’t expect the IPCC models to stay on track. This is discussed as a known cause and effect relationship.
But currently, the only way to show the empirical record suggest 7 year trends are highly and inexplicably variable is to claim the “weather noise” GMST is some sort of constant process unaffected by the volcano eruptions.
Lucia,
Somewhat O/T but I’ve been playing around a bit with AMO, PDO and GMST as a result of a thread on another blog. I am not well versed in Stats methods but I’ve noticed a fairly strong co-relation between these indexes, especially the AMO. I know co-relation does NOT imply causation (could say the say the same about CO2) but it is significant. Have you done any analysis with that?
Fred,
I haven’t done anything with that. However, the AMO, PDO are basically observations of weather patterns. In some sense, they are correlations.
Unlike volcanic eruptions, are inherent to the weather system. The cause of the AMO, PDO and other correlations is one of the big question. The phenomenological explanation of volcanic eruption GMST correlation is understood.
The AMO and PDO are a measure of SST patterns in each respective ocean. Since SST is a major component of GMST datasets, the co-relation between GMST and the indexes should be and is obvious. However, I am wondering if anyone has done an analysis of the sensitivity of the GMST datasets to these indexes. I assume there has.
FYI. I have posted my musings regarding the above here: http://weblog.xanga.com/fnieuwenhuis/674919203/amo-pdo-and-the-gmst.html
John Cristy has just published a paper (for the tropics only and using the UAH satellite data) which adjusts for ENSO and volcano impacts.
The temperature trend for CO2/GHG impact only is 0.070C per decade (35% of the magic 0.2C per decade)
http://arxiv.org/ftp/arxiv/papers/0809/0809.0581.pdf
AGW is not just predicated on a causal connection between CO2 levels and temperature but also that the connection is inherently catastrophic, as in tipping points and runnaway; your work, lucia, about which you are refreshingly modest, shows the connection is miniscule; a conclusion which is verified by the latest Christy and Douglass paper as linked by Bill Illis above.
cohenite-
I wouldn’t sat what I look at shows the link is minicules. But it points to the link may be lower rather than higher. I’ll read the christy paper.
Just some quick Lamb DVI reference numbers for a few of the volcanic eruptions listed on the graph:
1947 – Hekla DVI = 20
1956 – Bezymjannaja DVI = 10
1963 – Agung DVI = 800
1974 – Fuego DVI = 200
1982 – El Chichon DVI = 800
1991 – Pinatubo DVI = 1000 (Robock)
Presuming a linear relationship between DVI and temperature, the listed Pinatubo, El Chichon, and Agung eruptions should be noticeable. Hekla and Bezymjannaja should be no-see-ums.
Regards.
Bob–
GISS doesn’t include Heckla and Bezymjannaja in their forcing files. My main reason for stopping in 1944 is the buckets issue. Also, it’s hard to tell, but the squares on the graph only go around the volcanos listed at the top. The eruptions listed at the bottom are for references.
That said, if I wanted “pure” data, I’d wish for a run with none of these volcanoes on it and also when human forcing effects are more or less monotonic or at least varying slowly. Unfortunately, the darn buckets thing happened right smack in the middle of that!
Lucia:
As I commented over at Tamino’s, if the purpose is to check whether the data of the last 10 years are consistent with a longer term (i.e. 30 year) linear trend, then the source of the variability (internal/external, ENSO/volcanoes) is irrelevant, as long as the variability and influence of perturbations in the 30 year reference period is comparable to that in the 10 year period of investigation.
I don’t deny the large effect of volcanoes at all, but ENSO also has a significant effect on the temperature swings:
Strong El Nino: 1997/8, 1991/2, 1982/3, 1972/3
Strong La Nina: 1988/9, 1975/6, 1973/74, 1955/6
(http://ggweather.com/enso/years.htm)
Notably, the 1998 super El Nino, and perhaps the weak (?) La Nina of 2007/8, are quite important in shaping the apparent trend of the last decade. Disregarding the ENSO influence from your analysis therefore amounts to allowing some applejuice into your mixture.
It could perhaps be argued that the effect of ENSO is more evenly distributed in time than large volcanoes are, and that therefore ENSO does not need to be treated the same way (ie excluded from the analysis period). Is that indeed your reasoning regarding ENSO?
Bart Verheggen
Of course El Nino has a strong effect. But ENSO and volcanos together is more than ENSO alone.
I don’t disregard ENSO. ENSO exists during all periods. Unlike volcano eruptions, it isn’t something that “stops”. If we could create an earth with no volcanic eruptions, we would still have things like La Nina and El Nino. We had La Ninas and El Ninos’ during the period I used to estimate parameters. We’ve also had some since 2000. So, I picked a period when ENSO was acting, but volcanoes were fairly calm. I applied that to a period when ENSO was active and volcanoes are calm. So, like to like.
Tamino applies a period with ENSO and volcanos to a period with ENSO and no volcanos.
The difficulty for both of us is that it would be advantages to have an even longer period.
lucia,
If super El Ninos and large hurricanes were to lead to a lot of excess heat being lost to space would this not reflect a NON-linear AR(1) process? Is the current flatline (starting in 1998 or 2001) a product of such a global heating “hiccup”, effectively putting the earth back in radiative balance (assuming it was actually out) by removing all the much-vaunted and mysteriously-absent “heating in the pipe”? If so, then the GHG warming trend might resume any time now. Relevant to thread as follows. Under a period of heavy volcanic activity you might not observe such nonlinear effects; they might appear linear AR(1). Maybe a nonlinear heating hiccup effect is something the GCMs do not simulate very well? And that’s why they fail to predict a flatline (?)
Just a thought.
92 months? Don’t you need the # of months to be divisible by 12, either 84 or 96, if you want to compare like months to like months? Is this what accounts for the difference in mean levels between Figs 1 and 3? What am I missing?
Bender– The anomaly method supposedly takes care of the systematic variation over 1 year cycle. But I don’t compare individual months to individual months– I just treat new data as “incoming”.
The graph is not affected much by using 7 year or 8 year trends.
As for the cause of the flat line- the only reason one need search for a “cause” if if we admit it exists and is odd. But in anycase, El Nino’s is part of internal variability of the system rather than due to a change in the optical properties of the atmosphere. So, while each individual event might have some predictive ability for what follows, the models are supposed to account for the effects of “El Nino” like things — at least on average.
The difference between fig1 and fig 3 is fig 3 shows when the eruptions occurred. If you were to look carefully, the bottom of the deep spikes all coincide with major eruptions. (That is– except the ones during the bucket-jet inlet period.)
But if your stream is 92 months long you’re starting and ending in different seasons and that will introduce a systematic trend bias. 92/12 = 7 years, 8 months.
Maybe I’m gullible, but to me that 2001-2008 flat line exists and is odd and needs explaining. Especially when you consider other GCM model failures: lack of tropospheric warming, lack of OHC rise. If it was just one thing, ok. But all 3??!! Somethin’s up.
Bender, I thought anomolies take care of all that seasonal stuff….
Ah,”the anomaly method” – I got you. I wasn’t sure what you meant by that, it went over my head.
But that was a minor side point. What do you think of nonlinear AR model (vs. linear), and what do you think of the nonlinear “hiccup hypothesis?” That 2001-08 flatline bothers me. (And my hunch is it will, annoyingly, persist through 2009.) Are “alarmists” and “inactivists” (I can’t believe it’s come to this) arguing at cross purposes because they’re not understanding this “anamolous trend” for what it is – a “scheduled” post-El Nino warming pause?
Bender–
Trend bias in which direction?
The purpose of the anomaly method is to eliminate any trend biase due to the annual cycle. Admittedly, it may not work… but that’s the purpose is to eliminate the effect of the annual cycle.
Oh, Raven, that was you, not lucia. I had assumed these graphs were based on raw temperatures, not anomalies. I wasn’t following their development. I was just going by the Figure captions presented here.
Maybe call it the “punctuated AGW” hypothesis. During a period of low volcanic activity you get super El Ninos that move a lot of heat from tropics to pole and off into space. Ergo nothing left in the pipe.
R. Lindzen:
http://arxiv.org/ftp/arxiv/papers/0809/0809.3762.pdf
“In 2001, I published a paper (Lindzen, Chou and Hou) that used geostationary satellite data to suggest the existence of a strong negative feedback that we referred to as the Iris Effect. The gist of the feedback is that upper level stratiform clouds in the tropics arise by detrainment from cumulonimbus towers, that the radiative impact of the stratiform clouds is primarily in the infrared where they serve as powerful greenhouse components, and that the extent of the detrainment decreases markedly with increased surface temperature. The negative feedback resulted from the fact that the greenhouse warming due to the stratiform clouds diminished as the surface temperature increased, and increased as the surface temperature decreased – thus resisting the changes in surface temperature. The impact of the observed effect was sufficient to greatly reduce the model sensitivities to increasing CO2, and it was, moreover, shown that models failed to display the observed cloud behavior.”
Were 1999-2001 years of unusual planetary heat loss? Episodic heat expulsion leading to punctuation (and overall reduction) in the GHG warming trend? Maybe the “3C” sensitivity (with + fdbks) is bid down to 2 or lower when iris-like negative feedbacsk are included?
Iris effect would suggest a NON-linear AR process, might account for 2001-08 flatline.
Bender–
Do you mean a scheduled “unusual Super El Nino pause”? Who knows if such things exist? I suspect that if we did simplistic radiation heat loss calculations estimating heat loss during the El Nino using T^4, it would not be anywhere near enough to “cause” a huge heat expulsion and enough heat loss to create an 8 year flat line. So, testing that sort of speculative theory will require a detailed description of the mechanism and some supporting calculations.
Oddly enough, the “heat in the pipe” expression does not actually describe excess heat stored in the pipeline.
What it describes is this:
a) We’ve turned up the heat (forcings.)
b) The earth/oceans/atmosphere have thermal mass.
c) So, the earth/ocean/atmosphere temperature are not in equilibrium, the are too cold.
d) The earth /ocean/atmosphere is destined to rise until it is in equilibrium with the forcings.
So, the heat in the pipeline is the heat that is being applied now and which will continue to be applied. So, it’s the future forcing not excess heat stored in the earth. The earth can’t, by definition, “throw off” future forcing during the 1998 El Nino.
I suspect if we were to look at models, even at 1998 temperatures, the balance of incoming heat to out going heat at 1998 temperatures was supposed to warm the planet. Otherwise, if 1998 heat is sufficient to throw off more heat than enters the planet, 1998 actually represents the upper bound of warming at current levels of GHG’s, and there is no “heat in the pipeline!”
Re (#5541)
Hi Lucia, that’s an interesting (but different?) way of describing “heat in the pipe”. I might be wrong, but I had thought that before, when I’d heard the “heat in the pipeline” argument, it was simply something like, ‘since the ocean’s upper mixed layer has a higher heat capacity than the atmosphere, even if the enhanced GHG forcing were to freeze at current levels, atmospheric temperatures would still continue to rise as the upper mixed layer cools as they(ocean/atmo) continue toward a new equilibrium’ type thing. So perhaps this is where bender was coming from? The ‘heat stored’ version?
Well, I came back and figured I should add a reference to appease the gods 😉
http://www.columbia.edu/~jeh1/2005/Imbalance_20050415.pdf
I had never heard “in the pipe” explained that way either, Mike N. It leads to a peculiar realization. The larger the difference between the predicted temperature trend and the observed temperature trend (which is now quite large!) the stronger the “evidence” for heating in the pipe. Read that again. The larger the discrepancy the stronger the evidence that there’s lots more warming to come.
But tell me if this is not circular logic. The models that are used to project the trend are the same ones used to infer that the earth is out of radiative balance, such that there must be something “in the pipe”?
If the positive feedbacks that we hear so much about are whittled down by some of Lindzen’s negative feedbacks (iris effect), that might take quite a bit of that heat “out of the pipe” by restoring the perception of radiative balance that is inferred from the models (now revised to include an iris effect).
Yes, Mike N, that reference describes what had been loaded in my head, not knowing any better. But as I understand it, there is now some data/thoughts that the ocean is NOT heating as much as they thought when that article was written.
Does this add or subtract from the heat “in the pipe”? According to Hansen’s definition it subtracts (it was supposed to be there, but it’s just not, and we don’t know why). According to lucia’s it adds (it isn’t there yet, so it must be coming).
I readily admit that I find this stuff confusing. No flames, please. I am no authority. I am not worthy.
Crap – didn’t make the 10 minute time limit! Ignore the previous! Delete if possible!
………………………………………………………………………………….
Yes, Mike N, that reference describes what had been loaded in my head, not knowing any better. But reading it more closely now I see it is consistent with what lucia said.
That was supposed to be “the smoking gun” – the heat that was “in the pipe” was starting to “come out” in the form of a warmer ocean.
But as I understand it, there is now some data/thoughts that the ocean is NOT heating as much as they thought when that article was written. Does the lower rate of OHC increase now mean that there was not as much “in the pipe” as we thought?
Not sure what to make of this, as the topic has not been clarified very well yet at some of the more authoritative sounding web sites.
But it seems to me that Lindzen’s negative feedback processes, if they were included in the GCMs, could bid down what was formerly presumed to be in the pipe. Not to play “inactivist”, but this would have some implications for the “alarmist” movement.
Hansen quotes from Mike N’s document:
1. Direct significance of the energy imbalance found by the authors in their study.
“This is the ‘smoking gun’ that we have been looking for (with regard to identifying the human role in causing global warming).â€
“There can no longer be genuine doubt that human-made gases are the dominant cause of observed warming. The imbalance occurs at a time when natural forcings, e.g., changes of solar irradiance and volcanic aerosols, are small and while human-made greenhouse gases increased at
a rate consistent with the imbalance.â€
“The deduced planetary energy imbalance also helps us quantify additional climate change that is coming down the pike.â€
—————————————————————————-
Any chance of a re-assessment, before it cools much more?
Lucia,
I think you covered the question of – in the pipeline – very well in your blog back here.
http://rankexploits.com/musings/2008/linear-thinking-can-we-fit-a-line-to-figure-out-climate-sensitivity/
I think you can say that current ocean temperatures reflect past heat gains or losses. Anything in the pipeline is to be found in the difference between current energy stored and the energy that will be stored in the future depending on a predicted continuation of the excess heat gain.
What is in the pipeline is simply a prediction of how much of the sun’s future energy will make its way into the ocean.
Basically, I am agreeing with you 100%. 🙂
Jorge–
That image is the idea.
It actually matches part of what Mike N says here:
Because the heat capacity is higher, it takes longer to gain temperature after heat is applied.
The following bit has a mixup toward the end:
Change to “as the upper mixed layer also heats, but ceases to act as a heat sink.”
The picture idea is that during early periods of heating, the atmosphere is warmer than the ocean. So, during that period, there will be net transfer of heat from the atmosphere to the ocean. But, eventually the ocean will heat up too, and it won’t act as a heat sink for the atmosphere.
On this:
So perhaps this is where bender was coming from? The ‘heat stored’ version?
The difficulty is that in the cartoon version of “heat in the pipeline” cold is stored in the ocean! The heat is the heat applied.
This is sort of true. If you assume you have predicted the “Sensitivity” to forcing correctly, but you find the temperature has not risen at the rate predicted, then the fault ought to be in the “time constant”. Heat in the pipeline is a metaphor to explain the time lag between when you start heating and when the temperature rises.
There are loads of engineering problems that work this way. I’m trying to think of a good analogy though. This is not too good, but it’s not wretched:
When you defrost a turkey on the counter, the surface temperature approaches the room temperature first. Meanwhile the giblets inside might still be frozen solid. During this period of time, if you were monitoring the turkey surface, you’d find it wasn’t at room temperature yet. Even though it’s temperaure has risen a lot, it’s still a little below room temperature until such time as the entire turkey is defrosted.
The temperature rise “yet to come” is “heat in the pipeline”!
Lucia,
when you refer to the anomaly method are you:
1.) taking a base period of some years of data, averaging over all months of that period and adjusting the full data set by that amount?
or
2.) taking each month of the year of the base period (say january) and creating an average for that and adjusting all of the other januarys of the full period then repeating for each other month of the year? In other words one unique adjustment for each month of the year.
I always assumed the first for the normal anomaly adjustment from what I’ve seen described for the global temp data, but that wouldn’t adjust for any seasonal variation. It would just adjust the whole data set by an offset. Since the global data is an average over the whole globe January or June should be nearly the same (although I’ve wondered about that given the ratios of ocean to land and orbit).
Here’s a comment from the GISS file on conversion back to deg from anomaly values
BarryW–
GISS, Hadley and NOAA report the data after they apply their interpretation of the anomaly method.
I am under the impression they do (2) when reporting monthly data. They do (1) when applying annual averages– as described above. But the note is telling you they don’t do (1) for the monthly or quarterly data. For that data, you need to get the average for each month. That’s difficult to locate, but it exists. (I think it’s possible to find for Hadley, but I don’t remember where to get it.)
Lucia,
I interpret the single time constant models mainly as ocean heat capacity and ocean temperature with some mechanism for losing heat as the temperature rises. The slow initial response to a ramp input of heat energy, eventually followed by a constant lagging ramp in temperature, is exactly what is expected of this model.
Trying to talk about surface temperatures as distinct from ocean temperatures is quite difficult with the simple model. In reality you would have to include a boundary layer in the model to separate the ocean heat capacity from the atmospheric heat capacity.
Then you have to decide where the additional CO2 forcing is to be applied. Should it be applied to the ocean capacity via SW radiation or do we think it is applied to the air part which then communicates via the boundary layer to the ocean?
If we choose the latter we have another interesting time constant based on the ocean heat capacity and the conductivity/resistivity of the boundary layer.
I could see a situation where air temperatures rise rapidly in response to extra forcing and the ocean temperature lags way behind. However on that model the ocean and the air would share the total energy content when/if the forcing stopped and the final temperature would be close to the ocean’s simply because of the huge disparity in heat capacity.
The way I see it is that you can make whatever predictions you like about huge temperature increases and blame the non-observability on the lags.
So what is observed may be consistent with small forcings and short lags or large forcings and long lags. No matter how you look at it, the in the pipeline stuff is not to be found on earth but in the eye of a climate modeller!
It seems you need a smaller turkey or a bigger microwave. 😉
Jorge–
You are correct. With a single time constant model, it’s not really possible to differentiate between the temperature of the atmosphere and ocean.
The only idea I’m trying to convey is that “in the pipeline” is related to the idea of a lagged response. The heat isn’t hidden somewhere in the earth. It’s yet to be gained.
Re: #5549,
” they(ocean/atmo) continue toward a new equilibrium’ type thing.
Change to “as the upper mixed layer also heats, but ceases to act as a heat sink.—
Lucia, good spot, thanks. All the analogies and simplified versions (as well as the complicated versions 😉 get confusing to me. I mean, in reality, the oceans are almost always (with the exception of wind) losing energy to the atmosphere, right? So the ‘heat sink’ bit, while somehwat apt (since it’s representing an enhanced GHG SW forcing warming the skin layer, reducing the temp. gradient and slowing the oceans’ rate of cooling to the atmo; at least I believe that’s the RC take) is weird to me, I guess that’s where my ‘cooling’ came from. I don’t know if that makes sense but I’m in a rush to go to a baseball game.
Lucia,
I looked at some Hadley documentation and it seems as if the actual station data is converted to an anomaly for that station which is then gridded with the other station anomalies which is then converted into a hemispheric anomaly. To get to the global temperature:
So I’m not sure that you can ever get back to an actual temperature mean for any of the Hadley data at least, unless they’ve calculated it for the base period somewhere.
So based on your answer to bender the only way adding a month to the data set would affect the trend line is if the trend for a particular month or season of the year had a different trend than the yearly trend?
Lucia wrote:
“The heat isn’t hidden somewhere in the earth. It’s yet to be gained”
No.
A forcing pulse is suddenly transformed in Joules via heat (temperature) increase in the atmosphere, a minor part, or via the lack of cooling of the ocean, which has the major and dominant role.
Regarding the heat in the pipe, I invite all you (and Bender) to read the today’s post on Climate Science:
http://climatesci.org/2008/09/25/misconception-and-oversimplification-of-the-concept-of-global-warming-by-v-ramanthan-and-y-feng/
Thanks, Paolo. I have seen these questions asked at RC and never answered very fully or satisfactorily. Sometimes dodged completely. Pielke Sr’s post makes it clear that I’m probably not alone in my confusion!
I would like to thank lucia for this apolitical forum (and for all her work), and thoughtful people like Paolo who choose to reply to ill-posed questions from strangers.
In a reply this week at RC, gavin finally admitted “The heat going into the ocean is not going to be ‘released to the atmosphere’ any time soon – it is instead part of what will be the higher OHC in a warmer world” – in other words, the time-lag warming-in-the-pipeline argument was never real to start with – there is certainly no 30 year lags in the climate system.
Let’s look at “real” lags in the real climate system. The summer and winter solitice provide a good example.
Temperatures on the surface and in the atmosphere peak about 30 days after the maximum solar energy is received from the Sun at the solstice. The same is true for the winter solstice in that surface temperatures bottom out about 30 days after the solstice.
Sea surface temperature peak and bottom out about 75 days after the solstice.
Therefore, one would have to say the real climate has a lag potential of 30 days to 75 days – not 30 years.
Hmmm… well, I guess I need to go read RC?
What is the lag attributed to? I can understand sub surface water temperatures lagging surface, but I’m not sure i grasp the mechanism for surface temps. Downward propagation?
I would think that there would be an analogy to a child “pumping” a swing to go higher. Would increasing temps (heat) cause a greater lag since it would take more time to propagate?
BarryW– Do you mean you don’t graph the mechanisms for surface temps lagging?
Even air has a heat capacity. Also, things aren’t perfectly mixed.
It’s difficult to come up with simplifying analogies that explain one every possible question about lag all at once.
Part of the lag has to do with heat capacity. Part has to do with where the heat is applied, and how far it has to travel to arrive at something that needs to be heated.
But, matter that is closest to the source of the heating will heat up first. (Just as metal at the bottom of a boiling pot is hotter than the water. The water near the hot surface heats up before the rest of the water.)
But, then the heat will travel elsewhere. So, eventually the heat gets to the more distant parts.
If the planet is warming, the deep ocean will heat last. But while it’s cold, ti also acts as a heat since to the upper water. Heat will propagate from warmer water which is near the source of heating. So, yes, that’s downward propagation.
If the ocean where solid instead of liquid, I suspect the lag for heating and cooling would be the same. However, since there is convection, there may be differences in how the lag manifests itself for heating and cooling. (Hot water rises, cold sinks. Salt modified things.)
Sorry, I should have referenced Bill’s post relative to the solstice. I was wondering if that was a well understood phenomena that I had just missed. I understand what you’re saying, it just seems important to understand it causally. If the surface temps and SST’s lag the solstice the heat has to be somewhere other than those places prior to them heating up, true? The input (sun) is declining, but the heat is still rising. That has to be buffered somewhere until it’s transfered to the surface air or ocean surface. Daily surface temps lag noon by about 4 hours I think, which is attributed to land heat absorption/release. So if I measured daily in-ground temps I would assume they would lead the air temps. That I could buy as an explanation for the solstice phenomena for surface air temps. Not sure how to explain the SST lag by the same mechanism, since the temperature being measured is part of the storage device itself (equivalent to measuring ground temperature in the surface air temp scenario). How does it get hotter after you turn the heat off? Can’t get warmer from colder water beneath. Transfer from the air above?
Barry,
The sun!
I’m not kidding. The issue is that even though the sun reached it’s peak heat at the summer solstice, the earth’s temperatures was not yet in equilibrium. So, on June 21 the soil in my yard near chicago was gaining heat because the earth was colder than it would be if the planet got “stuck” so that Chicago got june 21 amounts of heat addition for ever and ever and every.
In fact, on June 21 the soil temperature in Chicago is colder than it would be if the sun got “stuck” at june 22, 23, 24 rates of heat addition and so on.
On June 22, the amount of heat coming from the sun is less than on June 21. BUT, the earth’s temperature is still cooler than it would be if the planet got “stuck” in the June 22 configuration. So, on June 22, the earth still gains heat. This continues. The earth keeps warming and the heat addition from the sun keeps dropping. Eventually, there is a point where the earth starts to cool– but that cross over is not on the day of the solstice.
For simple systems, there is an mathematical formula for this!
I think daily air temps lead in ground temps on average.
As you see the heat isn’t turned off. It’s turned down. But it’s still hot enough to heat the earth/ocean/air because the earth/ocean/air is still colder than it would be at equilibrium.
We already know the surface lags the solstice by 30 days and the sea suface lags the solstice by 75 days. I’m sure there is a simple rational explanation for this so we don’t need to go into it.
The more important question is how can global warming be lagged 30 years? What possible simple rational explanation could justify a lag wherein warming-in-the-pipeline from increased GHGs could be effectively be double in the future from what has occurred to date?
So far, the surface has warmed by 0.7C but the global warming models say we should have seen 1.5C of warming already given that (nearly) half of the increased forcing from increased CO2 should have already occurred. Half of the 3.0C increased temperature as a result of doubled GHGs should have already occurred.
The warmers answer to this discrepancy is that it is being absorbed (or is hiding) in the deep ocean. Many of the warmer’s followers accept this explanation without reservation. Many sceptics also accept this explanation since it sounds quite reasonable. After all, the deep oceans should warm more slowly.
But this explanation is not logical or rational. The surface still needs to increase at the rate the models predict or they are wrong.
If the deep ocean can absorb half of the increase so far, it will also absorb the next half of the next tranche of warming to come as well. The models are still off by a factor of two. Maybe the oceans can absorb more of the warming than the models predict. Maybe the models are wrong.
But the warmers answer to the discrepancy is illogical since the oceans cannot “increase” the temperatures at the surface if they are lagging the surface temperature. They are colder than the surface and hence cannot increase the surface temperature. If anything, there is only 30 days to 75 days of warming to come. Global warming is supposed to operate at the speed of light given we are talking about photons of the IR radiation here.
Bill Illis–
You are partly right and partly wrong. You can’t just translate the lag to the annual cycle to the lag for long term forcing.
But yes, if the models don’t predict the surface warming, then they fail to predict it. The issue of thermal lag would be an explanation for what they might get wrong. But speculating that about the reason why they might be wrong doesn’t magically transform incorrectness into correctness. (It may help modelers figure out which parameterizations are off, or which other factors are wrong. But that only means future models might improve. It doesn’t mean they current ones aren’t wrong.)
Bill Illis:
“But the warmers answer to the discrepancy is illogical since the oceans cannot “increase†the temperatures at the surface if they are lagging the surface temperature. They are colder than the surface and hence cannot increase the surface temperature”
I can’t think of “warmers” saying things like those.
No “warmer” thinks that it is the atmosphere to warm the ocean.
The process is as it follows:
ocean gets its energy from the sun radiation and each photon is immediately transformed in thermal energy.
Then the ocean has two ways to lose its heat, or via the infrared window, that directly connects the earth surface to the outer space, or giving it (sensible and latent heat) to the atmosphere which, also after some processes (green house effect and meteorology), radiates back this energy to the space.
The proner the atmosphere to radiate back energy, the bigger the transfer of heat from the ocean to the atmosphere. If the atmosphere is less incline to radiate back, the ocean could retain a bit more sun radiation and gets warmer.
If you take into account energy (heat) and not surface air temperature, the climate system has no lag and this endless discussion will stop at once. The issue will be: how energy will distribute inside the climate system?
Paolo M,
I agree that additional heat energy is supposed to be going into the ocean but the only way we can detect this is by measuring the rise in ocean temperature. The lag that Lucia was explaining in her other post was the lag between heat input and resulting temperature rise. Lucia was talking about a generic heat balance model and such a model does certainly show a lag between heat input and temperature response.
The shape of the temperature response is very dependent on how the heat input is varying. If there is a step response in the heat input, the temperature rise will be exponential, eventually settling at the new equilibrium temperatue. If the heat input is rising continuously, a ramp, the initial temperature rise is tiny but eventually rises to produce a delayed temperature ramp.
If the heat input is varying like a sine wave, the output temperature will also be a sine wave but the amplitude will be dependent on the frequency and will also lag the input by up to 90º. This is why variations with a daily period are much smaller than annual or decadal periods for the same variation in heat input.
So these lags are very real, even if we restrict the discussion to ocean heat content and ocean temperatures. The problem is that the ocean is not one large stirred isothermal body and so it is very hard to measure changes in ocean temperature and even harder to know what changes have occurred to the ocean heat content.
Like a lot of things to do with climate we find the uncertainty in real world measurements is very nearly the same as the whole effect claimed by modellers. This is exactly the problem that Lucia is battling against with predictions of surface temperature versus the observations. Lucia is at least trying to use standard statistical methods to find what range of temperature trends is compatible with the measured temperatures.
To me, this is a more interesting question than the one about whether climate models have a sufficiently large spread in trend predictions to cover anthing likely to occur in the real world.
Jorge,
if we restrict our discussion to ocean (and the climate system in general), there is no lag.
Let’s consider a constant albedo and sun: the input energy is constant, so there is no additional energy going into the ocean.
If, instead, the ocean is cooling at a lower rate for whatever reason, the higher heat content manifests itself through an higher water temperature and this immediately. You cannot have an increase in internal energy without a change in temperature.
If you can know the thermal state of the ocean, you know its temperature and the state of the climate system. Nowdays we are close to a situation in which we know the heat content of the upper ocean via the Argo system. And through Argo we know the instantaneous state of global warming or cooling to a good degree of accuracy.
Of course, I know that those processes within the atmosphere, the meteorological responses and feedbacks, make difficult to understand how the energy will distribute.
Paolo M,
You say:
I see it this way. If the ocean starts to lose heat at a slower rate it means the heat content will start to increase at that rate. The temperature will follow the rising heat content depending on the thermal capacity.
As the temperature rises it will increase the rate of heat loss until it restores a temperature and heat content equilibrium. Both the heat content and temperature will show an exponential rise to that new equilibrium. There is no way the heat content and temperature can instantly take on the values that correspond to the final equilibrium.
Whether the final equilibrium is ever reached depends on the cause of the original reduction in heat loss staying in place until that time.
So I agree with you that the heat content and temperature must march in step but the response to the change in the rate of heat input takes time to fully emerge.
Like you I am hopeful that the Argos system will give us a better measurement of the ocean’s heat content. If we know the change in Joules over some time period we will then be able to translate that to a positive or negative heat flow in terms of Watts per m2.
One problem of modelling is trying to find a level of complexity that can represent earth processes without losing the understanding that we can get from simple models with well known behaviour. With the simple models we are at least clear about the assumptions we are making. If the models get too complicated we may not fully realise how the additional assumptions affect the outcome. That is why complex models have to be tested against future predictions and not just their ability to replicate past climates.
Jorge,
just to be pedantic:
…heat content and temperature must march in step but we hadn’t the tool to measure the change in rate of heat content which was hidden by every day processes.
But today that tool is available and we have a way to assess model expectations, look at:
http://climatesci.org/2007/04/04/a-litmus-test-for-global-warming-a-much-overdue-requirement/
Quite many volcanic eruptions if they make the seven year trend to saw up and down like that…
Or could it be that a seven year trend is all over the place actually, and is a too short interval to deduct the longer term trend.
Gravityloss–
Examine the period in the “volcano eruption gap”. It doesn’t saw up and down anywhere near as much!
Practically everywhere the 7 year trend is a lot of the time +1 or -1 degrees per century above or below the 22 year trend, and quite often even + or – 3 degrees. Be it volcanic or volcano-free periods…
Your idea of proving or disproving great things from 7 years of data when the results you get are widely different if you take 8 or 6 years should be a big warning sign that maybe 7 years isn’t very much.
Gravityloss– What’s the relevance of ±1 C/century to anything in this post.
Also, where in the world do you get the idea the temperature is quite often ±3 C/century off during non-volcanic periods. That is simply incorrect. During non volcanic periods, the 7 year trends are rarely off more than 2C/century from the 22 year trend.
gravityloss,
1. What makes you think the analysis is highly sensitive to the choice of 7 years? (She didn’t choose it, you know. John V did.)
2. How is it lucia’s fault that the volcano-free period is short?
Shoot the messenger, why don’t you?
bender–
As it happens, the variability of 6 year trends is smaller during “no volcano” periods than “volcano” periods. This result is, as they say “robust” to choice of number of years. Unfortunately, the data record is short. I wish the “no volcano” period were longer–but heck, I wish the whole thermometer record was longer.
With the seven years I was referring to the whole tenet of the blog, falsifying IPCC’s 2 C per century from observing a 7 year’s trend.
That’s why John V was interested in how the 7 year trend behaves. It’s all over the place. It’s not a reliable indicator for the 22 year trend, with volcanoes or not. With volcanoes it’s even less reliable. Without them it’s still not reliable.
Now Lucia takes one single point of the red curve and falsifies a longer term trend based on it.
The whole plus or minus one year is related to that. If you exclude 1998 from the analysis, you get a very different result. That should be a warning sign that the method is not very robust if with very little change in assumptions you get a very different result.
But all this has been hashed before and I don’t expect you to budge any, I just post this for the possible curious readers. I probably won’t comment more.
gravityloss,
Thanks for the clarification. I think you are mistaking one particular analysis for the “whole tenet of the blog”. In fact I am quite skeptical that you have read the blog. No one here is saying the current GMST flatline refutes AGW theory. In fact I believe lucia is on record as suspecting otherwise. That doesn’t mean her work doesn’t have value. Are you sure you’re not taking your cues from some other source? My advice is to read the blog before you go generalizing thus.
gravityloss–
It appears based on your comment, that you are rebutting notions in your head which have nothing to do with things I post.
1) The seven year trend period analyzed in February did not include 1998. I add months as we accumulate data; the now 92 month period also does not include 1998. Count back; they start in January. So, my results are unaffected by the inclusion or exclusion of 1998.
2) I don’t know what you mean by taking a single point. I use a time series since 2001 to test projections in the AR4. The projections are supposed to apply going forward, not predict the past. So, I test what is the “future” based on when the clock starts for freezing models, SRES etc.
4) I don’t falsify a longer term trend. There has been no 2C/century long term trend. The projected value of 2C/century is exceeds the long term trend and is projected to apply now, not in the past. (The trend was lower in the past.)
5) You might do well to quantify things you claim. For example, what do you mean by trends are “all over the place” or “reliable indicator”. Hadcrut is currently more than 3C/century below the predicted 2C/century right now. I’ve said deviations of this magnitude should happen roughly 5% of the time if we analysize based on and ARMA(1,1) model. They should happen even less often if you just determine the fraction of time it happened during the “no volcano” period shown above. That means deviations of this sort are unlikely.
Thanks for visiting. 🙂
Your *method* is very sensitive to the inclusion of 1998. Your method is not robust. That is what I am talking about. That is just one demonstration of its fault.
I don’t bother evaluating it numerically, as 1) it’s blindingly obvious and 2) somebody else has probably done it already. Everyone can go and calculate trends here:
http://www.woodfortrees.org/plot/hadcrut3vgl/from:1999/trend/plot/hadcrut3vgl/from:1998/trend
And see the difference of a 8 year vs 9 year trend (or 9 vs 10, I don’t know how the inclusion paradigm works, and it doesn’t matter in this case).
I don’t have the time to read everything you write, which is a lot, but people cite this blog for their claims (different variants of “IPCC 2C per century projections falsified”) and then that leads me to come here and read a bit. I genuinely try to show you what’s wrong in your methods and how they are leading people astray.
The 2000-2100 trend might be +2 C, or it might be not, but we can not deduce it from the trend of 2000-2008.
Gravityloss–
My analysis does not include 1998. Period.
I have no idea what you think my method is. However based on simply showing the trends vary, it appears you quite likely, have either not read what I have posted or do not understand the nature of statistical inference.
You seem to be applying an ill-defined term– “not robust” to critcize examining a trend without looking at it’s associated uncertainty intervals. So, my response to you is: a) that is not the method I use b) you fail to define robust and c) you don’t seem to state what argument you believe you are refuting.
You now tell me you don’t have time to read my blog, and you don’t have time to run the numbers to see that claims you have made are simply incorrect.
If you don’t have time to read my blog, that’s fine. If you don’t have time to do calculations thats’ fine also. If you don’t understand the nature of statistical inference, I have no problem with that.
But visiting this blog, rebutting arguments I have not made, and posting bald assertions you cannot back up because you don’t have time to run the numbers or run statistical analysis of any sort whatsoever is pointless.
gravityloss is shadow-boxing. But that’s ok. Anyone with strongly held beliefs does it.
Bender–
I agree that’s what she appears to be doing. 🙂
Still, I”ve read those odd “rebuttals” to things I haven’t said elsewhere.