Here’s another post in my on going series to illustrate that variability in calculated trends from the collection of models used by the IPCC AR4 is strongly influenced by the variability across climate models. The variability does does not simply represent “the impact of the uncorrelated stochastic variability (weather!) in the models that is associated with interannual and interdecadal modes in the models”, as was suggested by gavin of Real Climate. In today’s post, show how the variability in simulated trends for ocean heat content in hindcasts are noticeably affected by
- The choice of individual climate model and
- The forcing scenario selected by the modeling group.
One might speculate that choice of climate model and forcing scenario would also noticably affect trends for forecasts, suggesting that the variations in calculated trends across the model projections. If so, then the idea that such variability is a solely manifestation of “(weather!)” seems dubious.
Evidence that differences in trends is not solely due to “(weather!)”
Today, Dominguez, C.M. et al. 2008 appeared in Nature. The authors reported improved estimates of the ocean heat content and sea level rise, based on observations and compared these empirically based estimates to hindcasts from the GCM simulations runs used by the IPCC in the AR4. Comparisons were separately presented for a) models that included volcanic forcing in the hindcasts and b) models that did not include volcanic forcings. The results for ocean heat content (OHC) were encapsulated in figures 2a and 2c shown below:
Figure 1: Comparisons of observed and simulated ocean heat content (OHC) for the upper 700 m of the ocean. Figure “a” compares observations to IPCC hindcasts to that ignored volcanic eruptions; figure “c” compares hindcasts with volcanic eruptions. Colored traces indicate individual models. Black indicates the 3 year running averages of observations; model results are annual averages. The yellow at the bottom indicates optical depth due to volcanic eruptions; the brown curve indicates a 3 year running mean of optical depths.
Original from Dominguez, C.M. et al. 2008. I squished figure to permit easier comparison of the relative magnitude of trends in OHC over the half decade.
Many things could be said about the graph– but I want to focus on the idea that the variability of trends predicted by models is a manifestation of “(weather!).
The choice of model affects the trend
First, examine the graph on the left which shows a range of results ofr OHC from the subset of AR4 climate models that neglected volcanic activity during the 50 year period. Black indicates the estimate of the ocean heat content (OHC)
Note the large variability in predicted rises in OHC over 50 years: The ratio of the highest to lowest increases are at least a factor of 2. This is quite a large amount of variability in the heat content added to the ocean, and this is in a hindcast.
Next note this variability is dominated by the choice of model. The results for FOAL are all clustered near the top; the results for CSIRO are near the bottom. If you squint you can notice something else: some models do have larger amounts of internal variabilty (aka “model-weather noise”) — as exhibited by more “wiggliness” about smooth behavior. My ability to differentiate between light blue green and light blue is a bit challenged– but it appears the blue-green representing GISS-AOM is “wigglier” than other models.
So, at least with regard to ocean heat content, that model may exhibit a larger amount of “model-weather noise”.
Including volcanic eruptions affects the trend
Examine the graph on the right: these are models that included the effect of volcanos. First, scan quickly to the right, and notice these projections the average in the hindcast increase in OHC is much lower than in simulations that neglected the volcanic eruptions.
Now, once again notice the range in variability between highest and lowest appear differ by at least a factor of two, and that’s putting in kindly, as I’m pretending I’m not seeing the “wild hair” model that hindcast a drop in the ocean heat content (OHC).
Once again notice: Results for separate runs from individual models are bunched together. So, we see the dominant factor in the variability in trends is the choice of model.
But maybe that’s only because this is 50 years?
Let’s zoom in on a 10 year periods:
Exploded view of variability in OHC over 10 year periods for models without volcanic eruptions.
I don’t know how this looks to you. To me, it appears even at 10 years, trends from individual simulations are clustered by model. Weather does matter more at 10 years, but “(weather!)” could hardly be said to be the dominant factor in driving the variability of trends over 10 year.
Wrap up!
So, we see with regard to Ocean Heat Content, the variability across models and in different forcing histories greatly influences the variation in 10 year trends.
That much of the variability appears to be due to variability across models and variations in forcing history is not surprising. It is expected on physical grounds, understanding of the impact of different parameterizations in model, which, in some sense, makes each “model” a different “planet”. It is also what we saw in the results for global mean surface temperature. In that case the variability in model projections of 8 year trends during a volcano-free period, which Gavin attributed to “(weather!)” was seen to exceed the variability for similar 8 year trends in the thermometer record for the honest to goodness earth. (And the thermometer record included volcanic eruptions, measurement error and non-linear variations in forcing due to GHGs in addition to variability due to true-earth weather.)
There are other speculations one could make based on these figures. But, I’ll cut it short now and invite speculation in comments. After all: What are blogs for?
Lucia,
I’m not familiar with the source document, so am not aware what the smooth brown line running through the volcano forcing represents. For that matter, I don’t know what the yellow shaded curve represents. Can you explain either one?
Earl–
The yellow shading are supposed to be the measurements of the optical thickness associated with volcano eruptions. The brown curve is the three year running average of the yellow shading. If you squint, you’ll see the word “Pinatubo”, Agung and El Chicon. (That means Fuego and another one are left off.)
The grey shading is estimated uncertainty of measurements during the time period.
Lucia – it’s OHC, not OCH!
This all contradicts my convenient earlier belief that the volcanoes wouldn’t affect longer-term temperature trends because they have an effect for only a couple of years. From the pictures it looks like even though the direct forcing effect is gone quickly, the impact on ocean heat content is still significant after at least 15 years: even in 1980, all the curves are at least 5×10^22 J lower from the Agung volcano in the early 1960’s. Sounds like more grounds for a very large long-term component in the climate response.
It seems a bit odd to rely on “new techniques to assess ocean temperatures to a depth of 700 meters (2,300 feet) from 1961 to 2003,” when the ARGO float programme was only just beginning at that time, and its first results seemed to cast doubt on the reliability of previous XBT ocean temperature measures.
Also the paper “shows that thermal warming contributed to a 0.53 millimeter-per-year rise in sea levels rather than the 0.32 mm rise reported by the IPCC”. Well excuse me if I am underestimating the terrible danger we are in, but that is the difference between a rise of just over 1.26 inches per century and the new figure of 2.08 inches per century.
Thanks Arthur. (I hate acronymns! Oh well.. I should learn to proof read.)
Yes. It appears that according to this study, the volcano eruptions have a fairly long term effect on the oceans. I want to see if I can get the time series of data and make a two lump lumpy with communication between. Sure it might be worthless, but it’s still fun.
Patrick–
I’m waiting to see how this is received as reliable way to estimate the sea level rise. Also, in my view, with all these corrections on the data, we sort of have to watch and wait to see how the models validate as new data come in.
Unlike in other fields where data are measured and honestly frozen, the climate data seems to be reprocessed relatively frequently. While I think every scientists tries to process data without regard to what it will later say about the models, it’s really difficult for human beings to ignore information about how some new massaging technique affects agreement between data and theories they believe. So, the problem in climate science is a circular loop is created.
I don’t know how this can be avoided given the reality that we can’t go back to 1950 and remeasure 1950. However, it’s still a difficulty.
Note how ModelE overestimates the response to volcanic forcing
Lucia,
Thanks for the info. I’m just trying to grasp the relevance of a three-year average of the optical thickness. Is that what was fed into the models? Are the models becoming prescient and initiating cooling 18 months prior to eruption? I don’t know, it just seems irrelevant. Or worse.
Earl– I think the optical thickness is just shown as an explanatory indicator. Those models that use volcanic forcing use the instantaneous optical thickness.
Steve– Yes. Model E seems to over estimate response to volcanic forcing. On the other hand, it’s a lot better than the model with the gold circle which predicts cooling over 50 years. Overall, other than the excess wigglyness, Model E and two others appear to be the only ones that are within a factor of 2 of the correct value.
Since the IPCC method is to assume the average of all models is somehow “best”, that would suggest that on this metric, their mean-model hindcast would be off by quite a bit, even now that the data are corrected.
That said, the mean model isn’t actually shown, so I’m trying to average mentally drawing from data in two separate graphs.
I am unfortunately also unfamiliar with the original paper, but at least by “eye-balling” I find it striking that a fairly strong drop in OHC PREDATES each of the tabulated volcano events. Is this just a matter of sloppy graphing, me being cross-eyed, or does anyone have other suggestions?
Cassanders
In Cod we trust
Cassander– As seems the habit in climatology, the time series are smoothed. In this case, they are smoothed with a 3 year running average. That’s probably why they show the smoothed eruptions too– so you can eyeball what smoothing does.
I added some text to the captions to describe the meaning of the brown curve, the averaging etc.
It appear the observations are three year running means; the model predictions are annual averages. This makes illustrated comparison a bit “apples” to “oranges”. It may be the over-response of Model E to volcanic eruptions is an illusion due to the averaging of the observations. Or, maybe it still over-responds after averaging. Who knows?
Where is a good place to track sea-level rise. I recently saw that Joe Romm said it was accelerating, and a NYC friend of mine noted “I’ve never seen the Hudson so high before and I’ve lived here for years.” When I suggested the tide was in (and it was), as the Hudson is tidal almost to Albany, we had an argument about how I was a denier and some such nonsense but it was all in good fun. At any rate I think there’s a perception that the seas have gone up in terms of feet lately, which we know isn’t true. However I don’t know how much they’ve gone up since say, 2000. Does anyone keep good track of that other then the IPCC?
Terry–
One of the amazing things about these blogviations is the tendency of some to respond to claims that either are or are not true with “You must be a denier!”
Obviously, either
a) It is true the Hudson is tidal almost to Albany and the tide is in and the combination is sufficient to explain the rise or
b) It’s either not tidal, the tide is not in or that combination is not sufficient to explain the rise.
Both are facts that could be checked. But instead the response is to accuse you of being a denier! You may or may not be one– but if (a) is true then the people marveling over the Hudson and seeing it as proof of AGW are wrongly interpreting the information. Clearly, at least some would rather respond with an ad hominem than risk testing whether thier “proof” holds up.
As for how much it’s gone up since 2000… I don’t know. Joe at Icecap ran an article on sea level rise. and showed this:
and this
Both show noticable rises since 2003. Very recently, there are brief downturns. These would need to persist for us to suggest any upward trend is broken. Note that joe says:
(Click either image for larger.)
While I seem to be back in the “Earth is Warming” camp the psychology around what’s going on is just fascinating to me and I also think it’ll hinder any action that may need to be taken.
Thanks for the images from ICECAP. We should have better data on sea level soon as I believe a new satellite designed track sea-level so was launched recently.
A freely available paper that might be of interest:
http://ocean.mit.edu/~cwunsch/papersonline/Wunschetal_jclimate_2007_published.pdf
Thanks Mike!
Thanks Lucia for all you do.
Could you put together a simple sea level +/- chart and a simple temperature chart +/- here? Agree on a baseline and stick to it.
Perhaps you could put together a data set for sea level like the one you have put together for measured tempurature. We can place bets and such on the monthly data!
EJ — I was thinking of putting together sea level rise. The AR4 didn’t have a specific prediction, but there is a very brief flattening. So, it’s worth putting together a “horserace” type plot.