Seven Climate Models, Seven Different Answers

In a new report, scientists used seven different climate models to assess human induced land cover change (LCC) at regional and global scales. The first results from the LUCID (Land-Use and Climate, IDentification of robust impacts) intercomparison study by Pitman et al. show no agreement among the models. This study indicates that land cover change is “regionally significant, but it is not feasible to impose a common LCC across multiple models for the next IPCC assessment.” In other words, this important factor is missing from current models and scientists are at a loss as to how to add it.

The study, soon to be published in Geophysical Research Letters, performed the type of analysis that was recommended in the 2005 National Research Council report on how to improve climate model accuracy. The results were both as expected, LCC proved to be an important missing factor, and unexpected in that none of the seven models tested yielded the same results. This throws modelers into a quandary regarding how to proceed. Here is the abstract of the paper by A. J. Pitman et al.:

“Seven climate models were used to explore the biogeophysical impacts of human induced land cover change (LCC) at regional and global scales. The imposed LCC led to statistically significant decreases in the northern hemisphere summer latent heat flux in three models, and increases in three models. Five models simulated statistically significant cooling in summer in near-surface temperature over regions of LCC and one simulated warming. There were few significant changes in precipitation. Our results show no common remote impacts of LCC. The lack of consistency among the seven models was due to: 1) the implementation of LCC despite agreed maps of agricultural land, 2) the representation of crop phenology, 3) the parameterisation of albedo, and 4) the representation of evapotranspiration for different land cover types. This study highlights a dilemma: LCC is regionally significant, but it is not feasible to impose a common LCC across multiple models for the next IPCC assessment.”

One of the reasons that the researchers are concerned is that, though LCC has significant effect on models in the regions where the change occurs they failed to find a long distance connection that would influence climate globally. According to Roger Pielke Sr., Emeritus Professor of Atmospheric Science at Colorado State University and Senior Research Associate at the University of Colorado-Boulder in the Department of Atmospheric and Oceanic Sciences: “The finding of a "no common remote impacts of LCC" does not mean this teleconnection does not exist, since they also report that there is a "lack of consistency among the seven models." Thus, in addition to the other shortcomings that the authors list with respect to teleconnections, if there are significant real world teleconnections, but they are not spatially coherent among the models due to their lack of consistency, the analysis procedure they used will incorrectly conclude that there is no long range effect of LCC when there really is. This issue needs further exploration in order to remove this limitation in their excellent preliminary investigation.”

Prof. Roger Pielke Sr.In other words, because none of the models gave the same results on the long distance effects of changes in land cover, modelers are at a loss when trying to incorporate LCC into their GCM climate models. Previously, attempts to include LCC were limited to simple global or regional averages. Pielke (picture at right), who calls this “a very important study,” expressed concern that because the long distance effect could not be consistently identified modelers would conclude none existed and not attempt to correct their models.

It should also be noted that the simulations performed for this report only included the “biogeophysical” effects of LCC on climate. Changes in land use cannot help but affect other climatically important factors such as the exchange of greenhouse gases between the land and atmosphere, reactive trace gases and creation of mineral and biological aerosols. Figuring out how to include these LCC effects is only the beginning.

Pielke, who runs the Climate Science web site, has posted some very interesting conclusions regarding his study of climate change and climate modeling. One observation is that global warming is not equivalent to climate change. It is possible for humans to cause climate change without any global warming or cooling—changing forest into prairie, rerouting or damming rivers, city heat-islands affecting rainfall patterns and such.

The land use choices we make can cause significant differences in regional climate. Deforestation and agricultural expansion in the coming decades will alter the way the land interacts with the environment. Forests tend to absorb large amounts of CO2 and solar radiation, while cooling the air around them through transpiration. Agricultural cropland, on the other hand, reflects sunlight better than some types of forest but is not as effective at capturing carbon. In fact, if chemical fertilizers are used, cropland can release significant amounts of nitrous oxide, a greenhouse gas many times more potent than carbon dioxide.

Ecological activists would have us believe that any change caused by man is bad and unnatural, but all animals affect their environment, from ants to elephants. The global emissions of methane and carbon dioxide from termites alone are estimated at 19.7±1.5 and 3500±700 million tons per year, respectively. These emissions contribute approximately 4% and 2% to the total global fluxes of these gases. Elephants can have a major transforming effect on savanna lands by felling, stripping bark or uprooting trees. Humans have a bigger impact because we are currently the planet's dominant species.

According to Gordon Bonan, who works on climate modeling at the National Center for Atmospheric Research, current land use models “don’t account for diversity of crop types at all,” and assume a generic crop type for all cropland. In a Duke University research report, he emphasized how uncertain climate models can be, and how when more factors that are included, such as land cover change, irrigation, and the carbon cycle, more uncertainty is introduced. According to Bonan, there’s no clear way to model climate change that would result in any sort of consensus among the international scientific community.

Other notable observations made by Dr. Pielke regarding climate modeling include: “Global and regional climate models have not demonstrated skill at predicting regional and local climate change and variability on multi-decadal time scales.” Simply put, climate models do not seem to work farther into the future than a few years. James Hansen.The IPCC ran their model predictions out to 100 years. Also, “attempts to significantly influence regional and local-scale climate based on controlling CO2 emissions alone is an inadequate policy for this purpose,” meaning trying to stop global warming by restricting CO2 emissions will not work.

Perhaps the most fascinating report posted on the Climate Science web site is an analysis of ocean heating in terms of energy gain as predicted by NASA's James Hansen (shown at the right). Hansen's modeling team has reported extensively on their results and have not been reluctant to point out flaws and inadequacies in the GCM they use. In a 2005 paper, reported in Science, Hansen et al. concluded that the “Earth is now absorbing 0.85 ± Watts per meter squared more energy from the Sun than it is emitting to space.” This estimate was the result of a number of modeling runs.

As with all model predictions, this one comes with a great deal of uncertainty: “A caveat accompanying our analysis concerns the uncertainty in climate forcings,” said the report, “even if the net forcing is confirmed by continued measurement of ocean heat storage, there will remain much room for trade-offs among different forcings.” Which factor (forcing) is the most troublesome? “Aerosol direct and indirect forcings are the most uncertain.” Those pesky aerosols again (see Airborne Bacteria Discredit Climate Modeling Dogma and African Dust Heats Atlantic Tropics).

Hansen wrote Pielke with respect to their GISS model predictions that “our simulated 1993-2003 heat storage rate was 0.6 W/m2 in the upper 750 m of the ocean.” This value was used as a best estimate to calculate the heat change in Joules that should be expected in the upper ocean data from 2003 to the present. Pielke then compared the predictions derived from Hansen's work with observed best estimates of heating in the upper 700m of the ocean since 2003. His conclusions were as follows:

“Thus, according to the GISS model predictions, there should be approximately 5.88x1022 Joules more heat in the upper 700 meters of the global ocean at the end of 2008 than were present at the beginning of 2003.

For the observations to come into agreement with the GISS model prediction by the end of 2012, for example, there would have to be an accumulation 9.8x1022 Joules of heat over just the next four years. This requires a heating rate over the next 4 years into the upper 700 meters of the ocean of 2.45x1022 Joules per year, which corresponds to a radiative imbalance of ~1.50 Watts per square meter.

This rate of heating would have to be about 2 1/2 times higher than the 0.60 Watts per meter squared that Jim Hansen reported for the period 1993 to 2003.

While the time period for this discrepancy with the GISS model is relatively short, the question should be asked as to the number of years required to reject this model as having global warming predictive skill, if this large difference between the observations and the GISS model persists.”

Unsurprisingly, the GISS model predictions do not match up with reality, just as the seven LCC models don't agree with each other. Note that the GISS model, specifically modelE III, is the computer model the IPCC's predictions were based on in its last report (AR4). Given the rapid pace of new climate science discoveries, climate modelers find themselves with a burgeoning list of new factors that need inclusion in their models. Will they succeed in making the models reflect the real world? Probably not, if the experience of the LCC modelers is any indication.

Earth's climate system is amazingly complex and modeling is fraught with pitfalls and danger for even the most experienced computer scientists. No climate model predicted the current downturn in global temperatures, though many are now scrambling to predict possible decades of unchanging or cooling climate “within the general warming trend.” Still, climate science remains enthralled by its computerized playthings. I have to echo Professor Pielke's question, how many years of wrong results are necessary before we reject the IPCC reports and the models they are based on?

Be safe, enjoy the interglacial and stay skeptical.

These models cannot even

These models cannot even predict last week's weather.

And most leave out the Sun and about 900.000 other climate variables.

como jugar al poker

A few final thoughts

I thought it best to sum up the thread of discussion that this post generated, which can be read in detail below. Dr. Pitman, lead author for the LCC modeling paper, argues from the position that the IPCC's models are fundamentally correct and that the contributions of his team's work is to refine the models at the regional level. Further, intense land cover change is significant at only locally in the regions that it takes place. There are no couplings between these significant local effects and larger scale climate (continent scale or globally). He dismisses long distance linkages, what Dr. Pielke referred to as teleconections, as internal artifacts within the models themselves responsible for variability. He then calls this variability noise and claims that the inconsistency between the models examined is normal and lost in the expected noise. I will address each of these statements in turn.

First of all, I do not accept that the IPCC's models are fundamentally correct. This is because they consistently fail to accuratly reflect the real system they are supposed to model (i.e. Earth's climate). Perhaps the most detailed list of the problems with the IPCC's models, speciffically GISS modelE, was presented in a 2006 paper by Hansen et al, “Climate Simulations for 1880-2003 with GISS Model E.” A shorter version than the one analyzed in The Resilient Earth, chapter 14, is avalable but be aware that even this version is ~25 MB. If you check the section titled “Principal model deficiencies” you will find a littany of problems as identified by the modelers themselves. You do not have to accept my evaluation of the state of climate modeling, you can read the words of the modelers themselves.

Second, the fact that the impacts of intense LCC are greater in the region where they are taking place is unsurprising. What is surprising is that Dr. Pitman claims that these significant local impacts are in no way translated into effects at larger scales, continental or global. As I mentioned in my first reply to Pitman, there are any number of journal papers that dispute this. Simply Google on “land cover change impact” and click on the “scholarly articles” link at the top of the page. LCC impacts precipitation, atmospheric circulation, albedo, evaporation and myriad other factors that can and do affect climate on continental and wider scales. Dr. Pitman's not finding these linkages in his model does not negate their well documented existence in the real world.

Third, to claim that the variation between the models is significant enough to make selecting an LCC model for use within the greater GCM framework problematic but then saying that variability is lost in the noise of the model is contradictory. Ether the signal generated by LCC is identifiable and measurable against the background noise or it is masked, it cannot be both. Are this paper's results significant or are they too lost in the noise?

Finally, I would like to comment on a natural but sometimes regrettable tendency on the part of climate researchers and scientists in general to try to enhance the importance of their work. When submitting a paper for review and publication it is normal to present the work described as important, significant to the field. If the work isn't important why publish it or even do the study in the first place? Dr. Pitman was caught between trying to emphasize the importance of his work while diminishing its possible negative impact on the IPCC's climate modeling efforts, something he obviously supports. It is never possible to anticipate all of the interpretations that will be made of scientific data but that is part of how science works. My interpretation of this paper's results obviously struck a nerve with Dr. Pitman (as evidenced by the exchange of posts below).

A model that is missing significant factors cannot be considered correct, even if it gives the answers expected by the modelers (the “correct” answers). This is analogous to using the wrong equation but getting the correct answer to a test case and declaring the equation accurate as a result. If a model has been tuned to reproduce the past century's climate changes fairly accurately, but significant factors were not included, then the model's parameters must have been adjusted to compensate for the missing factors. This doesn't make the old model correct in any way, in fact it implies that the model is misconfigured. Modeling is so seductive because, with a bit of tweaking, a model can be coerced into reinforcing the modelers preconceived notions. A model is like a puppy, eager to please its owner, nature is a wild animal that doesn't give a damn what people want. You don't study puppies in order to understand lions, tigers and bears.

Seven Climate Models, Seven Different Answers

What the study actually shows (and as lead author I ashould know) is
that the global simulations from climate models as used by the IPCC
are not biased by any problems in implementing land cover change.

That is - our study increases our confidence in the warming modelled
by the IPCC.

Our results do not - even in the wildest imagining of dodgy skeptical
wishful thinking - suggest IPCC is wrong at continental scales and above
and I simply re-enforce the conclusions of hte IPCC that climate models
are reliable and robust tools for climate prediction'

Andy Pitman

Mr. Pitman, I laugh at thee

I have been a practicing computer scientist and software engineer for 30 years. I have developed various types of models, complex systems, artificial intelligence software and much more. I have been studying climate modeling for a few years now and have concluded emphatically that current climate modeling is nothing short of a joke. The climate modeling discipline, its participants, remind me of a room full of my past college freshmen students, carefully crafting a BASIC program to print 'hello world!', at which point they were certain they knew all there was to know about logical processing. The present climate modeling community is nothing more. They are not computer scientists, they have no understanding of logical systems, they have absolutely no clue how to develop software (at this point I will cite as example, the incredible climate modeling hack Gavin Schmidt) and they are in no position to conclude anything from that which they think they have built. I find it incredibly arrogant that climate modelers expect they must be invalidated in every way, when in fact, the models cannot validate themselves against any observation. Sir, the burden of proof lies with you! Incredibly stupid. With idiots like this, it is becoming ever more clear why the rest of the world is beginning to kick our butts in computing achievement.

In short, I simply laugh at thee.

Dodgy Thinking

Your study reported: seven models that product statistically significant results but are not in mutual agreement; the current models do not included these results; and the longer range affects of these significant local factors was not identifiable because because of the aforementioned lack of coherency among the models.

So let me see, the impact of LCC is by your own account significant, at least regionally, but no single model is correct under all conditions. By inference some models are more correct than others under different circumstances, making their inclusion into the IPCC GCM “problematic.” The fact that they report significant results and are not currently applied to the overall models says that the current IPCC models are missing this significant input and are therefore not accurate themselves (if they were accurate the LCC results would have to be insignificant).

Also, you state that you found “no common remote impacts of LCC,” but that does not imply that there are no remote impacts, simply that you were unable to identify them due to “lack of consistency among the seven models.” Because you can not find consistent affects at larger than regional scales it is safe to ignore your findings?

And this is without considering the impact that LCC can have on evaporation, albedo, and the creation of airborne particulates, both mineral and biological. Do you really think that current climate models are capturing the impact of deforestation and replanting as pastureland that is taking place in the Amazon river basin? Can you honestly state that this activity is having no impact on the rest of South America? Just a glance at the map below (taken from The Economist) would give even a layperson pause.

From “The Future of the Forest,” The Economist, Jun 11th 2009.

You claim that “Our results do not - even in the wildest imagining of dodgy skeptical wishful thinking - suggest IPCC is wrong at continental scales and above.” Johannes J. Feddema et al., would disagree with that statement with regard to the regional LCC in the Amazon. Specifically, in their December 2005 paper in Science, titled The Importance of Land-Cover Change in Simulating Future Climates:

“Adding the effects of changes in land cover to the A2 and B1 transient climate simulations described in the Special Report on Emissions Scenarios (SRES) by the Intergovernmental Panel on Climate Change leads to significantly different regional climates in 2100 as compared with climates resulting from atmospheric SRES forcings alone. Agricultural expansion in the A2 scenario results in significant additional warming over the Amazon and cooling of the upper air column and nearby oceans. These and other influences on the Hadley and monsoon circulations affect extratropical climates. Agricultural expansion in the mid-latitudes produces cooling and decreases in the mean daily temperature range over many areas.”

They found impacts at the regional scale plus impact on “extratropical climates” and over “many areas.” It seems that the IPCC's models don't get LCC correct and that there are predicted impacts on greater than regional scale. This paper is not alone in its conclusions, a quick literature search yielded dozens of journal papers that reached similar conclusions.

What dodgy set of semantic modifications and fuzzy logic allows you conclude that the IPCC models are not only accurate but that, in light of your study reporting significant but missing aspects of the problem being modeled, those models are actually strengthened? I find it highly amusing that climate change modelers constantly find “significant” new factors with which to improve their models yet steadfastly maintain that the old, unimproved models retain their accuracy—this is logically inconsistent and patent nonsense, sir.

There is misunderstandings

There is misunderstandings galore here.

What all these "new factors" show - the "new factors" that improve the models is that the projections by the IPCC models at large scales is consistent. That is, as you refine processes the large scale results stay functionally the same. So, no one can currently claim that the IPCC projections are "wrong" because they lack LCC. It is simply not demonstrable that there are large-scale uncertainties in the IPCC assessments resulting from LCC.

By "large-scale" I mean "at continental scales and above".

The results from our work points to problems in regional projections over areas of intense LCC. No suprise there, nothing inconsistent with IPCC. What our results also show is that these inconsistencies at REGIONAL scales do not affect anything remote (distant from) the individual regions.

This means:

(a) regional-scale land cover change is important - as stated in both the
3rd and 4th IPCC reports and our paper;
(b) the impact of regional-scale LCC is important, primarily locally to the LCC
(c) the impact of regional scale LCC does not affect areas away from the LCC.

The lack of consistency in the "teleconections" does not mean that the models are inconsistent necessarily. It is my view based on the science to date that what these "teleconnections" are is internal model variability. This is somthing that all models show and should show since the real climate system also shows natural variability. If it is internal model variability THERE SHOULD NOT BE consistency between the models - since its noise.

So, attempting to interpret "noise" between models and interpreting differences between noise generated by the models is simply not reasonable. Your state: "I find it highly amusing that climate change modelers constantly find “significant” new factors with which to improve their models yet steadfastly maintain that the old, unimproved models retain their accuracy—this is logically inconsistent and patent nonsense, sir".

Well - lets use an example. I bought a car from Ford in 1995. That car worked fine - I could drive it around no problem. I buy a new Ford in 2008. Its different - its got a new transmission, a new engine that is more fuel efficient and a new sound system that works great. My new Ford *does not* demonstrate that my old Ford was wrong - to assert that my old car was "useless" is [in your words "logically inconsistent and patent nonsense". My new car is better, but it is a refinement that might be a little more reliable, a little more refined but it does not make the old car wrong.

I do not want to debate this further - given the array of innacuracies on this web site I suspect I would be wasting my time.

Andy Pitman

Analogy with a car !

"My new car is better, but it is a refinement that might be a little more reliable, a little more refined but it does not make the old car wrong."
Andy Pitman's analogy illustrates well, even if he doesn't realize it, the problem with climate models. New cars are like "improved" models : they add new fancy features instead of solving fundemental problems, they are increasingly complex and the interactions between the new "features" become inextricable (ever had electronics problems ?), they cost a fortune to buy and maintain, their designers are convinced they are more reliable, the PR people take command and keep claiming it's "more reliable" often enough until an unsustantiated claim becomes Truth.

And what about reliability, after all ? It sucks, like anybody with a new car with plenty of added features well KNOWS. Accept Mr Pitman.

Jean Demesure

An Old Ford is not a Climate Model


Your analogy rings hollow—an old car, regardless of make, is not used to project the state of the world's highway system 20, 50 or 100 years hence. A car is not a climate model in purpose, use or intent and a new car supplanting an older one is not the same as upgrading a GCM, who's purpose is to divine the future. In computer modeling even slight errors can lead a complex and nonlinear system to report wildly different results unless that system is restrained from such excursions by artificial "reasonably" restraints imposed by the modelers. Of course, such restraints introduce the modeler's prejudices into the model and assure that the results are as expected.

You state that you don't wish to debate this further, fine. A petulant response is no substitute for reasoned discourse, but I understand your reticence. Many other scientists, myself included, will continue to dispute not the uncertainty but the inaccuracy of the IPCC climate models until it can be shown that they are accurate representations of empirically demonstrable physical processes which govern Earth's climate. Model away, to date climate models represent only a mystical convergence of software and wishful thinking, never rising to the level of acceptable scientific evidence.

An Old Ford is not a Climate Model

I wish the authors of the arrogant hyperbole that tries to state things in black and white could get a grip and some perspective! Perhaps they should all learn the Dunning-Kruger effect, but of course are unable to by virtue of the same effect. Sigh, see