After Hansen’s 20 year anniversary talk last week, I read some comments debating the question, Is it warmer than June ‘88?. It appears different people understand the question differently, and so give different answers both at other blogs and in comments on my previous post.
During the course of the conversation, various other questions were raised. These include:
- How are Hansen’s prediction/projections tracking observations? and
- Is Dave Stephen’s graph at Climate Skeptic showing a relatively poor comparison between satellite data and Hansen’s projections fair?
So, today, I’ll show some graphs comparing Hansen’s A, B & C scenarios to data from NOAA NCDC, HadCrut, and GISS Land/Ocean. The comparisons will be based on “the ordinary eyeball test”, with little discussion of statistics.
So, now, on the the meat of the post which will include a comparison to what a naive person might have predicted in 1988.
What did trend analysis suggest in 1988?
Before showing graphs and data comparisons, I think it’s important to note that a question like “How are Hansen’s prediction/projections tracking observations?” is a bit vague. Ideally, when asking how a prediction is tracking observations, one wants to know how the prediction is tracking compared to some other prediction. So, in that regard, I’m going to explain and show a naive forecast someone might have suggested in June 1988 when Hansen made his projection.
Here is a graph comparing the linear regression fit to data from 1900 to May 1988, the month before Jame’s Hansen gave his speech.
Click for larger.
Figure 1: Blue indicates monthly observations of global surface temperature anomalies based the average of NOAA/NCDC, GISS Land/Ocean, HadCrut. Pink indicates leading 12 monthly average surface anomaly. The solid black line is the result of an ordinary least squares fit to the data from 1900-May 1988; dashed lines are the 95% uncertainty bands for a forecast, with uncertainties corrected for red noise, residuals are assumed normal.
What might one have expected in 1988?
In some sense, figure 1, showing a trend of 0.6C/century represents a “base”; with or without a GCM, one might expect recent past history to be a somewhat plausible predictor for future behavior. (One might, of course, end up entirely wrong.)
They might also expect monthly future temperatures would fall largely inside 95% confidence bands shown, and that there would be some obvious structure in the temperature variations. That is: there are distinct periods where temperatures are low and others where temperatures are high.
In my opinion, this would have been the projection for a “good model” to beat.
How do Hansen’s predictions compare to data?
So, having prepped you with that graph, I will now super-impose Hansen’s data on figure 1. In my comparison, I put all data on identical baselines (in so far as possible.) The measured data are set to a 1951-1980 baseline. Hansen data is set to a 1958-1980 baseline. (I can adjust this and make another graph if anyone would like to see that.)
Here is the graph:
Click for larger.Figure 2: Predictions for annual average global surface temperature and observations from June 1988-now added to figure 1. A vertical dashed black line indicates June 1988. Comparisons after 1988 are indisputably forecasts.
- Scenario A: Continued growth rate in emissions at 1.5% / year.
- Scenario B: Emissions frozen at 1988 rates.
- Scenario C: Drastic reductions in emissions in 1990.
(Update: Details on forcings here.)
I can say a few things based on the eyeball test:
- It’s quite clear Scenario A has been over predicting warming for several years. No one disputes this anymore. Moreover, the 1990 IPCC projection of 3.0 C/century which may have been influenced by papers written circa 1988 has been largely abandoned.
- Currently, Scenario B over predicts the trends that occurred. But, given weather variability, a super-El Nino might get us back on the Scenario B track.
- Scenario C is currently over predicting the warming that occurred by a small amount.
- A naive person using linear extrapolation would have under predicted warming. There have been distinct excursions. So, the eyeball test suggests warming has been a bit faster than one might have expected based on the 1900-June 1988 empirical trend.
- If we compare the graphs to Climate Skeptic’s, we see we get more or less the same results using land based or satellite data. So, as far as I can tell, his graph seems to fall within the range of “fair comparison”.
Any thoughts?
P.S. if you click the graphs, you can see larger versions. 🙂
Updates/Requests
Nick Stokes asked for a graph that focuses in on the predictive period. This is figure 2, but focusing on 1985-2010. The hot pink is annual average data, and so most comparable to the Hansen predictions. However, Hansen predicted Jan-Jan averages, and the pink curve is annual averages for each possible month. (That is, Jan-Jan, Feb-Feb, March-March etc.) The average is selected such that the jan-jan averages should line up.)
Lucia,
First, a suggestion. Your second plot is hard to read. The period of interest is since 1980 – is it necessary to plot back to 1900?
The predictions were for annual averages – shouldn’t you compare them with annual averages rather than monthly?
On the scenarios – they were designed so you could decide which one applied, in the future, based on emissions over the period, rather than by reference to observed temperatures. None applied perfectly, but I believe on that basis B is the closest.
Other graphs of this kind that I’ve seen had troubles with different baselines. GISS is OK, and satellite data is difficult. You might like to mention what you have done there.
Nice graphs. Are the temperature trends you use to create your baseline forecast identical to the datasets that were published in 1988, or are you using the most recent re-re-re-adjusted datasets? It will probably make a difference on what the least-squares forecast comes out at.
Nick Stokes,
“GISS is ok”
Have you been to the Climate Audit site at all?
Nick– I can show starting from any year at all. The advantage of since 1900 is to put in context. The advantange of since 1958, is it shows Hansen’s entire projection. The advantage of since 1988 is it shows predictions only. I can slap up all three tomorrow!
The pink curve is annual– but as it would be computed in any particular month. The difficulty is Hansen spoke in June. People are likely to be curious about June-June annuals just as well as Jan-Jan annuals. So, I decided to show the curve.
Paminator- I’m using the most recent data. I’d have to go to the way back machine to figure out what GISS reported as “the” temperature back in 1988! I’m not sure if the projections would look better or worse using historic data.
The choice of baselines makes a bit of difference also.
Dave,
I mean GISS is OK in baseline terms. It uses 1951-1980, as Hansen did. Others use more recent (warmer) periods, so the anomalies are lower. For proper comparison, this has to be corrected (and often isn’t). Satellite data is harder, because there isn’t much of an overlap to get a good correction.
Nick– That’s right. The satellite data does not exist during the baseline period originally selected by Hansen.
There are difficulties associated with picking “fair” baselines. In his paper, Hansen used 1951-1980 for experimental data and “perpetual 1957” for his computer runs. (He ran with the forcings for 1957 for the equivalent of 100 years.)
I thought about rebaselining everything to 1958-1980 for that reason. But, at this point we all know that our choice of baseline can make the projections looks slightly worse or slightly better. If people want to see a particular choice, I’m happy to show it.
Lucia,
Do I take it then that you have plotted the other data with respect to their original (varying) baselines? If so, it seems to me that GISS is the only fair comparison.
Nick– I rebaselined all experimental data to 1951-1980. Everything ends up on to of each other.
The difficulty is that Scenarios A,B,C start in 1958. So, those are base-lined to 1958-1980. I could rebaseline everything to 1958-1980. (I suspect it would make Hansen’s projections look worse– but I’m not sure.) Hansen originally baselined A,B,C to “perpetual 1957”– meaning that in the original paper, the scenarios and the data are on slightly different baselines. Mine are, in principle, closer. But I can’t make up A,B,C data for 1951-1957 inclusive.
Hansen’s method is the correct method. He did the 100 year runs to get rid of annual variability. So choosing a baseline different from the one he uses will skew the results.
In other words, Hansen’s choice of baseline is a prediction of the model itself.
My beef is still with the comparison of apples and oranges. Scenario’s A, B, and C are based on the physics of Hansen’s models not his guestimate based on projections of past temps. If Hansen’s models were showing any predictive power the temps would be tracking somewhere around A and B. The C scenario is for drastic reductions in the increase of CO2 growth, which is the only one close to the actual data, but it is the one whose premise is farthest from the actual data (increasing CO2 rates). Would you not agree that from that standpoint Hansen’s model is failing? If the runs were based on changes in other variables or the slope of the scenarios at least mirrored the slope of the real data, I would say there would be some room to argue that the model was a reasonable analog of reality. I’m not arguing that temperatures are not rising nor even that CO2 is not playing a role, just that Hansen’s model is not validated by real data and is either totally wrong or missing major data or parameters.
Boris–
1) When Gavin looked at this, he rebaselined.
2) It makes no particular sense to compare predictions with a baseline of 1950 and to data with a baseline of 1951-1980.
Why Hansen did this is a mystery. Why the reviewers let him is an even greater mystery.
That said, using Hansen’s “perpetual 1957” baseline for the ABC scenarios makes Hansen’s projections look slightly worse than shown in my graphs, though you’d have to have sharp eyes to tell. I used Hansen’s baselines choice for the data.
So, if you’d like me to make a new graph using Hansen’s exact choices, I’d be glad to. I looked at a number of possible choices that made sense. The one I picked is more favorable to Hansen’s projections than any others I tried.
That’s my understanding. Emissions are generally thought to be tracking between A&B. I’m just showing how the data look plotted together.
BarryW,
No, all the scenarios involved increasing GHG. A was exponential, B linear, C capped in 2000. C and B postulated a volcano in 1995, about equivalent to Pinatubo which turned up in 1991. Gavin’s graph of the scenarios and the actual forcing is here. I’d say it turned out about B-.
Lucia,
I didn’t understand how you could rebaseline the satellite data to 1951-80, since the data starts much later?
I assume Hansen used the 1957 baseline because he didn’t then have forcing data from 1920, which he would need to do a converged simulation from 1951-80. Anyway, he himself compared that with the GISS (or its ancestor) up to 1988, so it seems the fair thing to do is to continue that – in effect, to continue the graph that he published. If you want to bring in the other data, it seems to me that the right thing to do to equate baselines is to line that up with GISS, by shifting the baseline of each anomaly (CRU, RSS, UAH etc) to have the same mean, say, in the forecast period 1988-2007 (or any other substantial common period).
We should be noting here that in 2006 Hansen himself published a review of the 1988 forecast, with updated graphs. And here is the link to his original 1988 paper.
Lucia,
I’ve put here a plot which is as near possible, in my view, to a comparison of Hansen’s predictions vs observation. It uses just annual GISS anomaly data, and plots them without adjustment against the scenario simulations, as Hansen did in 1988. I’ve emphasised scenario B, in red, the GISS stations index in black and the Land/Sea index in green. The scenario simulation data is from here and the GISS data from here. The agreement between the stations record and scenario B, in particular, is pretty good.
Nick–
Why do you say this looks like good prediction?
Your graph and my graph show pretty much the same thing, with mine showing about 6 months more worth of monthly data — all scenarioss over-predict the land/ocean data on your graph just as much as on mine. The most recent 6 months make the overprediction appear worse.
The agreement is such that if there is a super El Nino in the next few months, the predictions might look ok in future. But for now, they are low.
Also, why do you add the land-only station data to the graph? The Hanson simulations provided the equivalent of Land/Ocean not land only. He could have computed land only, but he didn’t.
Had Hansen’ graph shown land-only predictions, I’d have compared those to the land-only GISS data. All models predict greater rises over land. Presumably Hansen’s did also. Naturally, his land/ocean data which predict too much warming don’t look quite so bad if we compare to land only data.
My graph graph doesn’t show any satellite data. NOAA HadCrut and GISS are all land and all have data back to 1900. That’s why I picked them.
This is precisely what I did– I put all observations on the same baseline. I picked the baseline Hansen used for GISS Land/Ocean. I only used Land/Ocean versions of data since that’s what’s required for the apples/apples comparison.
I could rebaseline to 1988-2007, but the difficulty is then obviously everything matches now by definition. To see disagreement, we need to scan back to the past. Also, 30 years is good to avoid having disagreement be due to weird shifts due to weather noise.
I don’t know why Hansen picked 1957. I have no issues with that choice, it’s fine. It just makes picking a baseline a bit of a pain in the neck. Ordinarily, I’d prefer baselines that are absolutely the same, but also to make the same choices Hansen made. Unfortunately, his original figure has two baselines. For the data, he used 1951-1980. For the simulations, he used the control run of 1957, so I can’t rebaseline to 1951-1980.
I suspect Hansen didn’t go back to 1920 mostly because he was running the whole thing in the background on a creaky old computer, and doing an extra 30 years meant results might take an extra year. Given how computers were turning over at the time, the thing might have been scheduled to be replaced before he was done!
If I rebaseline agains, my thoughts for rebaselineing were:
1957-1987. That’s 30 common years. I think it makes little difference. I’m not adding the satellites because, by definition, the only fair 30 year baselines makes comparisons nearly impossible!
Lucia/anybody
Something that’s been niggling me for some months now, How much of the pre-late-20th-C (in this case pre-Hansen) warming is considered AGW and how much is just background warming?
I seem to remember reading somewhere that up until the late-20th-C it’s pretty well impossible to see the impact of C02 on the temperature record?
Gary
BarryW–
The confusing thing is the language. You are wording the growth rate discussion the way Hansen did– by dicussing the increase in the growth rate.
The annual rate of emissions were already growing in 1988. In scenario B, Hansen assumed emmissions would continue to grow at the rate they had been growing. So, the slope of the line in scenario B is equal to the slope in 1988 (or so). You can see that in the figure Nick linked. (I inserted the html to make it show.)
Scenarios A assumed the emissions rate would grow even faster and grow faster and faster each year. The growth rate in emmissions makes the curve veer up in the figure Nick inserted.
Scenario C caps eventually sets the emissions rate constant.
The words Hansen himself used are in the figure on his caption. I’m going to be lazy and use climate-skeptics figure instead of cutting and pasting a new one.
and now you all know why moshpit objects to the anomaly method.
report the data in actual C.
Then, if you like, do whatever anomaly dance you like or whatever period you like.
BUT, report the actual temp, without doing your anomaly dance. then go amuse yourself
and confuse others by picking different periods
Steven–
I also would prefer real temperatures. But, those aren’t available either from GISS or for the scenarios.
I suspect the reason real temperatures aren’t used is most models have a cold bias. Some have a warm bias. If real temperature were used, people would see the scatter.
From what I understand, the bias is on the order of 5%-10% of the 33K explained by the ordinary, undisputed greenhouse gas effect. One can decree this “good” or “bad” depending on perception. It’s small compared to 33K but large compared to the increases projected over the century.
The one thing I will observe, Hansen’s three cases were not of different possible sensitivities. My understanding from reading his testimony is that they are all using the same sensitivity, with three different CO2 forecasts. Case 1 is an accelerating CO2 output, case 2 is a flat output, and case 3 is some sort of reduced output. In this sense only really case 1 or case 2 is relevant, since our current world Co2 production is probably between these two cases. Case 3 in yellow is not even relevant. Its presence in your chart implies that Hansen might have been open to cases with lower sensitivities, which was not the case
Coyote– Hansen used letters. So what you call 1 he called A etc. You are correct the model sensitivity is the same for all cases; the predictions are for differnet levels of GHG.
Nick inserted this graph from real climate which suggest forcing rose according to “B”.
I have difficutlies with this graph because Gavin rebaslined the forcings to 1984. They had diverged before that. If there is heating “in the pipeline” rebaselining the forcings is a rather odd thing to do. But the graph does show that relative to 1984, the forcing increased more along the lines of “B”. I don’t about agreement prior to 1984. The runs begin in 1957.
Perhaps the header “What would did trend analysis suggest in 1988?” should be fixed.
Thanks Kriek.
Lucia,
The reason is that it is the data Hansen himself used for comparison in 1988. He used the temperature measurements assembled according to Hansen and Lebedeff 1987. Possibly he would have used land/ocean if they were available, but maybe not. He is predicting global air temperatures, and land/ocean includes water temperatures. Neither index exactly matches what he is predicting.
I think stations data matches the scenario B prediction well. It was under in 2004, 2006 and 2007, over in 2002 and 2004 and very close in 2003.
Apologies for my misunderstanding about satellite data – I assumed, without reading properly, that your combined figure included them. Your baselining sounds good, then.
I thought that too. But I can see the point of anomalies. Temperature varies in the first few metres above the surface. Instrumental measures try to standardise that, but especially with satellites, it won’t be the same standard. So instead of a discrepancy in years, there’ll be a discrepancy in location.
Nick,
I think you’re being a little too generous. The real test will come in the next few years when Hansen’s Scenario B diverges sharply from Scenario C. Your graph doesn’t show that, but a few months ago at CA, I posted an updated version of Hansen’s own graph from 2005, which was used by his supporters to show he had gotten things “right”.
Sure, GISS station data has tracked close to Scenario B, but it’s tracked close to Scenario C too, and is now even closer to C than it is to B. In fact, except for 1998, which we are now told is so unusual that only a blithering idiot would try to compare it to anything, the years in which observations exceeded B are also years in which B and C were almost identical in Hansen’s projection due to climate variability built into the model.
The next 2-3 years will indeed be quite interesting.
http://img230.imageshack.us/img230/2682/hansen2007jpegwn6.jpg
(How do I get this to show as an image and not just a link?)
Answer— just wrap it in the html for an image. If the image shack doesn’t block display it will. Usually, I just add the html.
I think Coyote’s point was that you can’t use any of the temperature data to ‘choose’ between A, B and C. You have to look at the CO2 release data. Where we’ve exceeded the output assumed for the ‘A’ case IIRC.
IOW: The actual temperature data better matches a ‘B-‘ scenario, perhaps. But the CO2 releases in Europe and the US continued the ‘steady increase’ pattern, and a large chunk of the rest of the world (India + China) are in exponential growth. That would seem to be an “A++” scenario on Hansen’s chart.
There isn’t a prediction “CO2 increases wildly, but CO2 forcing isn’t quite that strong,” basically because that still isn’t Hansen’s position.
Nick– I understand your reasoning now. Did you put the met data on the 1951-1980 baseline? (Maybe it’s already on that?) I’ll check the crut and NOAA data too.
The sea temperatures are an issue when comparing. There is that pesky ‘jet-inlet to bucket transition” issue causing the kerfuffle over the record suspect after WWII. Even though the correction for recent temperatures is expected to be small, it could affect the baselining. If the temperaure from the 50-60s are corrected up, the 80s-now will drop relative to the baseline.
As for agreement– I generally agree with you that hte Hansen predictions don’t look bad compared to met data. I just don’t think that’s the right test. The question of whether there can be any right test is open!
Lucia,
Yes, the Met station series is baseline 1951-80 (stated in the heading of the dataset) I believe it is a continuation of the H&L series that Hansen quoted in 1988. I didn’t modify any numbers. I believe my graph should be of the same data as that of John M.
I would expect that the scenarios do start to diverge in 1984 (when Hansen started this calc). Before then, there is no basis for scenarios, the history is known. I don’t think there is any numerical significance in saying the forcings are zero in 1984; he’s just saying that he knows the state there, and the scenario figures are the changes.
Alan B – did you see the forcings plot above? Gavin says that, from measurement, emissions did follow scenario B pretty much. Is there contrary data?
John M – I didn’t see your point about scen B and C diverging. It looks like we’ll be following B for a while, so that’s the only one that counts. The prediction is been on the high side for the last two years, but is not too far away.
Nick–
(Click for l arger)
This is a plot of the forcings Gavin linked to. I’ve set the “zero” in 1958. As I said, I’ve never understood why he rebaselined to 1984.
To me, the graph looks like real forcings did “C”. Yet, I know Gavin, Roger Pielke etc. all say between B and A.
Lucia,
I think I see what is going on with the scenarios. I thought they should all start in 1984, because they make provision for what you don’t know about decisions in the future. The past should be assumed known. Looking back to Hansen’s paper (1988), that’s almost true. He assumes the main gases behave the same in each scenario up to 1984, But, as a sensitivity test, he included some other effects in scenario A, prior to 1984. This covered some things that he couldn’t estimate accurately, like ozone. In Gavin’s RC post, he notes this in a parenthesis, but, as you say, seems to have matched A to B&C in 1984. The pre-1984 divergence is reflected in the numerical data which Gavin linked to. It’s just an issue of how the forcings are illustrated; the post-1984 forcings aren’t used for any calculations in the plots.
I think, for various reasons, scenario A is not very useful. And I think the plot you’ve just shown is of the forcings currently used (GISS E). They are lower than those used by Hansen pre-1984. However, what counts pre-1984 is what he used, so I think what is needed is a composite forcing – Hansens’s pre 1984, and GISS E post 1984. This is probably why Gavin are normalised them also to zero at 1984. That brings them up to about B level.
Lucia,
I got spam filtered again, and my comment vanished. Here’s what the message said
“Submitted comment 3795 to Akismet.com as spam (Akismet said it’s ham)
Sorry, but your comment has been flagged by the spam filter running on this blog: this might be an error, in which case all apologies. Your comment will be presented to the blog admin who will be able to restore it immediately.
You may want to contact the blog admin via e-mail to notify him.”
Regards,
Nick Stokes
Hi, your paper is interesting, but would be more convincing if proof read properly… See caps and [ ]
IF Dave Stephen’s graph at Climate Skeptic showing a relatively poor comparison between satellite data and Hansen’s projections was fair
Before showing graphs and data comparisons, I think it’s important TO NOTE that a question like “How are Hansen’s prediction/projections tracking observations?†is a bit vague.
Here is a graph comparing the linear regression fit to data from 1900 to May 1988, the month before JameS Hansen gave his speech.
So, having prepped you with that graph, I will now superimpose Hansen’s data on figure
(I can adjust this and make another graph iF anyone would like to see that.)
It’s quite clear Scenario A has been over[-]predicting warming for several years. No one disputes this any LONGER. Moreover, the 1990 IPCC projection of 3.0 C/century which may have been influenced by papers written circa 1988 has been largely abandoned.
If we compare the graphs to Climate Skeptic’s, we see we get more or less the same results using land[-]based or satellite data. So, as far as I can tell, his graph seems to fall within the range of “fair comparisonâ€.
Nick– I don’t know why SpamKarma hates you! I’ve been trying to white list your name etc.
Nick, the data I plotted may well be what is currently used in GISS Model E, and not measured forcings. However, my plot is exactly the data in the file you referred previously as “actual forcings”, using these words:
So, if those aren’t actual forcings in my plot, they also aren’t actual forcings in Gavin’s plot you included in comments above.
On the issue of rebaselining.
Of course the post-1984 portion of the line in purple or green in my figure aren’t used in Hansen 1988. Neither is the pre-1984 portion of that line.
Supposedly, Hansen used the scenario A, B and C data Gavin provided in the file and indicated as scenarios A,B and C in the legend in my figure and Gavin’s. The purple and green in my figure are supposed to compare what was used to what is “real”. However, it’s not entirely clear the A,B, C data in Gavin’s file are exactly what Hansen used , because if we read the Hansen paper, he included volcanos in A,B and C.
I could cut and past Hansen’s graphics, but they aren’t in KW/m^2, they are in temperature, having been scaled by the models sensitivity. What we would see is that Hansen’s grapic’s don’t precisely match the curves Gavin provides. (I don’t think they were meant to. I think Gavin meant to show GHG increases only.)
With regard to scenario A: You say “scenario A is not very useful”.
If you mean it’s turned out clearly incorrect, I agree. It’s totally wrong.
But the topic of this article is “How did Hansens predictions do?”
In that regard, the results of all scenarios need to be mentioned.
Scenario A results were in the paper; they were presented to Congress. So, clearly, Hansen felt these results should be shown to the public and legislators at the time. When the paper was discussed in Congress, and he didn’t suggest that case as unrealistic or useless. So, in 1988 it was represented as useful as some sort of prediction or projections. Part of the answer to how they did is: Scenario A was wayyyyy off. (For whatever reason.)
That said, the published paper does suggest “B” as the most probable trajectory. So, at least with regard to what Hansen included in the peer reviewed paper, he endorsed “B”. This means the fairest test of the claims in the peer reviewed article are comparisons to “B”.
(We can’t really test what was in Hansen’s inner mind back in 1988. We all know Hansen often makes stronger claims in public than in peer reviewed articles now. Scenario A got a lot of play in the public back in 1988, so mentioning it was totally wrong is far from useless.)
I’m not sure I buy your theory of why Gavin rebaselined. If we are trying to figure out which of the forcings applied to a run was closest to reality, we need to compare the forcings in a run to reality that happened later.
If we consider phenomenology, to compare forcings applied to relatity requires us to see forcing from 1958- now. We know this because incorrect forcing in say, 1960, affects the temperature in 1961, and later 1962 and so on. Since the climate has a finite time constant, and in particular models seem to think the response time is on the order of 20-30 years, the forcing from 1958-1988 sets the initial conditions for what happens later. If the bulk response time of the climate is 30 years, we can’t expect a “wrong” past forcing to be ‘forgotten’ in less than roughly, 30 years. (And saying climate has response times on all scales doesn’t help any.)
So, I really can’t see any rational reason for forcings to 1984 when comparing, particularly as, on Gavin’s chart, Scenario A visibly deviates from B and C around 1965. That’s why A was already hotter than B and C in 1984.
If the reason for rebaselining the forcings is that no one (including Gavin) knows the forcings Hansen used, then figuring out which case best matched the forcings is impossible. In that case, it’s best to simply admit that we can’t know which forcing was closest. If so, all we can do is compare the temperature profiles to what happened. In that case, A look terrible. B and C might look ok if you accept the idea that it makes sense to compare global temperatures to land only data. Otherwise, if we accept the land/ocean data, they all look high.
I hope you don’t mind an atmospheric layman dropping in and looking at this on an empirical basis.
I used my own eyeball technique to compare real CO2 with Lucia’s real forcing graph. I took the forcing graph and overlaid the NOAA CO2 concentration graph that I found at Anthony Watts’ blog. After stretching the abscissa to match the time scale, I scaled the CO2 concentration to match the Real Forcing w/o volcanoes from 1958 to 1988 (Dr. Hansen’s data, I presume).
From that fit, I believe the CO2 concentration plot represents the form of Hansen’s correlation to the real 1958-88 forcing data. This may be the cause and effect that led Hansen to conclude “… there is only 1 percent chance of an accidental warming of this magnitude, the agreement with the expected greenhouse effect is of considerable significance.â€
If this CO2-forcing correlation does actually hold, then the CO2 plot should be a reliable predictor of real forcing without volcanoes after 1988. The CO2 plot does match the real forcing through the early 1990’s, then it diverges from reality, becoming an unreliable predictor by 1995. I think rather than “a greenhouse effect … of considerable significance,†this comparison shows that the “1 percent chance of accidental warming†was what actually happened. Or, despite the outlandish probabilities, someone (Jim Hansen) does win the lottery.
I also note that the CO2 plot adjusted to the early-period real forcing does provide a fantastic fit to Hansen’s Scenario B through the end of the CO2 data. This is where I believe the AGW debate is today – i.e., Who are you gonna believe, Dr. Jim Hansen (NASA Phd!) or your lying ordinary eyeballs?
Are you still allowed to post this kind of stuff? Jim Hansen is apparently God and free speech re his projections may be blasphemy.
see http://energybulletin.net/node/45651
Lucia,
Since there is concern about GISS’s scientific objectivity why can you not over lay UAH, Hadley and RSS satellite data on your graph. You go back to 1900 for one of your graphs, it would be just as valid to include the satellite data.
As it stands you are using Hansen’s oft “corrected” data to validate Hansen’s predictions seems like a fairly glaring lack of impartiality.
Bill
Bill–
I don’t show the satellites on this test because the main difficulty when doing “ordinary eyeball” test is matching baselines. To minimize the “shift” effect of any possible cherry picking of baseline periods, it’s best to pick a rather long baseline. Thirty years is fairly conventional. But, the satellites have only been operating 30 years, to shift the satellites properly, I would need to set the average for 1979-2009 equal to the Hansen scenarios.
Anyway, climate skeptic already compared to UAH. I’m not sure how he baselined, but the scenarios don’t look good against the satellites. The only way to make them look good is to stick to land only measurements, measured using land based methods. You need to omit measurements over the ocean.
Nick Stokes thinks using the land only measuremetns is fine, I’m dubious about it. In a few years, we should have sufficient spread between all scenarios to know for sure.
Nick,
Lucia has sort of alluded to this already, but I’m not sure why you say this:
“I didn’t see your point about scen B and C diverging. It looks like we’ll be following B for a while, so that’s the only one that counts. The prediction is been on the high side for the last two years, but is not too far away.”
If the GHG between B and C are used in the model, and the model doesn’t distiguish between the two until late this decade, how can the model be tested? Why do you say that only B matters?
And for what it’s worth, we’re now closer to C than to B, even using Hansen’s preferred data. What if I were to say, “it looks like we’ll be following C for a while, so that’s the only one that counts”?
John,
Scenario C capped CO2 emissions in 2000. Scenario A has exponential growth, which we haven’t followed. (That’s why I said A wasn’t useful, as well as its inclusion of other drivers). I said that we’re following B, because that is the one with linear growth in CO2 emission. I’m not talking about which one the temperatures are following – that’s the test. But the stations GISS data follows the B scenario projection fairly well.
Nick/Lucia
The second chart describes the scenarios as:
1. Scenario A: Continued growth rate in emissions at 1.5% / year.
2. Scenario B: Emissions frozen at 1988 rates.
3. Scenario C: Drastic reductions in emissions in 1990.
That doesn’t match Nick’s description:
“Scenario C capped CO2 emissions in 2000. Scenario A has exponential growth…B, …is the one with linear growth in CO2 emission.”
What’s up? And, what is the actual emissions growth rate?
Larry,
The graphs shown above describe the situation, probably better than I did. Scenario B has linearly increasing forcing; cumulative emissions rise, but the annual rate of emission is constant. Scenario C caps forcing in 2000; the rate of emission drops drastically. Lucia’s caption is correct, interpreting “emission” as rate of emission (which I agree is probably normal usage).
Again, I would refer to the graphs – decide yourself which emission scenario best describes what has actually occurred.
BTW, in suggesting that GISS met station index is perhaps the one to compare, I’m not suggesting that it is the best. I’m only saying that it is probably closest measure to what Hansen was predicting, since it was probably the only index available at the time, and was the one he used in his plot.
Nick
Thanks, but haven’t annual emissions increased over the interval? I.e., doesn’t B underestimate actual emissions?
Larry,
Yes, the rate of CO2 emission has increased. The graph plots total forcing (including other gases), and suggests that the nett effect is about linear (constant rate). Again, I’m just interpreting the graph – I have no special information. I would certainly say that B is closer than A or C.
It would be nice if Scenario’s A, B and C were spelled out a little more clearly. While I think Nick and Lucia are right that Scenario A calls for an increasing emissions growth rate (not just increasing emissions), Scenario B calls for a linear emmissions growth rate,, etc., I don’t think Hansen’s usage was either the normal usage or very clear at all. Regardless, as I understand it, CO2 emissions have been experiencing an increasing growth rate, maybe not as high as Scenario A, but definitely not linear as assumed in Scenario B. However, I think other GHGs, such as methane, may have actually had a somewhat decreasing growth rate. So maybe everything balances out to more or less Scenario B.
I think what would be useful would be to see a graph of the actual GHG forcings over the last 20 years, not just the actual total forcings. Based on actual GHG forcings, my guess is that we should be comparing Hansen’s predictions to something slightly greater than Scenario B due to the increasing growth rate in CO2.
Does anyone know whether emissions referred to in the graphs refer to emissions by humans or increased emissions into the atmosphere as a whole? It seems(per a brief glance at Wiki) that atmospheric CO2 levels have increased by ~ 10% in the last 20 yrs. This is obviously less, than the increase of 1.015^20 predicted in scenario A and would put it about a third of the way(in Co2 increase) btw scenario B and A.
OTOH, if emissions refer to human activities only, actual Co2 increases must pretty closely match(if not exceed) Hansen’s predicted levels for A.
Cheers, 🙂
All:
There were a sufficient number of questions that I went through Hansen’s paper and posted what he actually said about the forcings. The disucssion is:
What Forcings Did Hansen Use?
There are points of ambiguity. If anyone can fill them in, that would be great!
I continue to enjoy immensely the ascendency of prediction to a status superior to fact. Arguing over whose future fantasy is correct is like two fleas arguing over who owns the dog they are riding on. Let’s try a few facts on for size, just to see how they fit. We live today in the Holocene Epoch, the last 11,500 years since we melted our way out of the Wisconsin ice age, which, incidentally, is the time in which all of human civilization has occurred. This is the sixth, and possibly the longest interglacial of the past six interglacials, dating back to the Mid Pleistocene Transition, when we went from a 41k year ice age/interglacial couple to the present 100k year ones (of which there have been seven so far). During the previous interglacial (called the Penultimate Deglaciation), the one in which Homo sapiens first appears in the fossil record, sea levels were only a paltry 12.5 to 16 meters above present (could this have been due to the simultaneous discovery of beans and salsa? or was it just campfires form a few million protohumans?). In the interglacial before that (two back), sea level highstands only made it to just 11.5 to 22.5 meters above present (we weren’t here yet so this is a special kind of whodunit). Wander three interglacials back and things start to look kind of staid with sea level highstands reaching between 10.5 to 23 meters above present. But that fourth interglacial back is real special, with sea level highstands recorded in the 25.5 to 29 meter above present level (Ironshore Formation, Grand Cayman, believed to be vertically stable for the past 500k years).
Recent sediment studies conducted by the USGS in the LA basin found 16 major climate change events in the Pleistocene, each measuring on average 120 meters in sea level shift, about 90 meters below present to 30 meters above. Ice core data from Vostok and Dome Concordia tells us these happen on the most regular clock in all of geology, nice regular, dramatic and unavoidable climate shifts, with temperature increases leading CO2 increases by an average 1,300 years on the major shifts, and temperature decreases leading CO2 decreases by an average 2,700 years. Meaning, of course, that just like us, CO2 has been a spectator at these major and abrupt climate shifts, and not an agent provocateur. That doesn’t mean CO2 can’t cause a climate change, only that it never has before. So if it takes an industrial revolution and 6 plus billion H. sapiens to cause a 2 foot (IPCC) or 20 foot (the Gorical) rise in sea level, what does it take to cause the 400 footers that are so regular we set our geologic clocks by them?
But not to cast too much doubt on the prowess of prediction, I do indeed have a favorite. Recently I read that the Arctic ice cap is expected to be completely melted away by 2070. This was a shocker for not all that long ago I read that sedimentologists working the Arctic have recently speculated that the trigger event that seems to send us abruptly into the next ice age is the complete melting away of the arctic ice cap…………..
Now if we are going to somehow engineer a reversal of a 2 or 20 foot rise in sea level against a backdrop of natural rise many times (possibly orders of magnitude) greater, doesn’t this strike you as something like an ant crawling up an elephant’s leg with rape on its mind? And what about population, it is predicted to be 10 billion by 2050. And how about rainforests, which were predicted in 2001 to be gone in just a few decades, a prediction in dire straits of coming true due to the advent of biofuels and the doubling in Amazon rainforest devastion this year alone.
Facts, pesky little things aren’t they?
I have 2 questions:
1) Why did you pick a different scale for the X and Y axis from Dave Stephen’s graph? This makes the eye-ball comparison more difficult.
2) Can you extend the graph from the start of the Holocene period to now? Without out using smoothing on the base data and adding a beginning to end average line?
Ron–
1) I used the baseline Hansen used.
2) No. I can’t extend the line from the start of the Holocene to now. This is a comparison of Hansen’s projections to observations. Neither Hansen’s projections nor the data go back to the start of the Holocene.
If I am looking at the charts right am I to belive that heat tempetures have started on a down turn in the last few months to year. if that is the case doesn’t the blow most peoples theory of the earth heating faster and faster due to all the CO2 we are releasing into the atmosphere. or am I not getting this correct