Climate Feedback

Global Warming and Forecasts of Climate Change

Posted by Olive Heffernan on behalf of Kevin Trenberth

Given that human induced climate change is with us, a looming challenge is to predict just what the climate will be. To date, there are no such predictions although the projections given by the Intergovernmental Panel on Climate Change (IPCC) are often treated as such. The distinction is important. A paper presented at the International Forecasting Symposium in New York City in late June 2007 by J. Scott Armstrong and K. C. Green is highly critical of IPCC procedures and “forecasts” for not being based on “evidence based” procedures as outlined in an earlier 2001 book of his. It is true that IPCC does not refer to Armstrong’s work as it has dubious relevance.

In fact IPCC does not do forecasts, as explained in my earlier post. The IPCC instead proffers “what if” projections of future climate that correspond to certain emissions scenarios. Armstrong has evidently read only chapter 8 of the IPCC Working Group I report and has therefore overlooked the fact that the other chapters address many of the things he is critical of.

In particular there is clear evidence (“warming is unequivocal”) that climate is changing in ways consistent with the climate forcings. Also, the projections are for all aspects of climate, not just global mean temperature. It has been said that “All models are wrong, but some are useful”. The Armstrong forecast of no change is not useful when the system is already changing in ways consistent with human influences on the composition of the atmosphere. Nonetheless, improvements in forecasting procedures are always welcome.

Bob Carter, a climate change doubter in Australia, has written a distortion of all this in the Courier Mail, issuing various attack against the science of climate change. Andrew Ash has written a rebuttal of these comments.

Another key point is that unlike forecasts based on past experience, weather forecasts are based on numerical weather prediction models and rigorous procedures, not empirical methods, although the latter are used to provide added value (e.g., based on known biases in the model). My own presentation at the same conference provides a description of weather and climate prediction.


The same atmospheric models are the atmospheric component of climate models and they are well tested and evaluated, although in climate models lower resolution is used.

The authors should recognize that IPCC does not make forecasts but rather makes projections to guide policy and decision makers. If those changes are considered undesirable, it can create efforts to change that outcome. Such mitigation is already happening in the U.S. Congress, in many states, and internationally under the Kyoto Protocol. Hence the projection will not be correct as actions are being taken to make it so. As such it is not a forecast of what will actually happen.

Kevin E. Trenberth

Head of Climate Analysis

NCAR

Comments

  1. Report this comment

    Mark Bahner said:

    Dr. Trenberth writes, “In fact IPCC does not do forecasts, as explained in my earlier post. The IPCC instead proffers ‘what if’ projections of future climate that correspond to certain emissions scenarios.”

    OK, I agree. The IPCC does not do “forecasts” or “predictions.” They only do “projections.” (Which, not coincidently, even scientists often assume to be the same thing!)

    But let’s make the meaning of this difference perfectly clear for all the readers of Nature’s Climate Feedback blog. Here are eight assertions regarding the “projections” in the IPCC’s Fourth Assessment Report (AR4). They concern the “projected” temperature increase in degrees Celsius at 2090-2099 relative to 1980-1999.

    I’d like you—or Dr. Heffernan, or anyone else at Nature—to please label each assertion as “true,” “false,” or “don’t know,” based on the material in AR4:

    1) The IPCC thinks there is more than a 0.1 percent chance of warming more than 6.4 deg C.

    2) The IPCC thinks there is more than a 10 percent chance of warming more than 6.4 deg C.

    3) The IPCC thinks there is more than a 50 percent chance of warming more than 6.4 deg C.

    4) The IPCC thinks there is more than a 99 percent chance of warming more than 6.4 deg C.

    5) The IPCC thinks there is more than a 0.1 percent chance of warming less 1.1 deg C.

    6) The IPCC thinks there is more than a 10 percent chance of warming less than 1.1 deg C.

    7) The IPCC thinks there is more than a 50 percent chance of warming less than 1.1 deg C.

    8) The IPCC thinks there is more than a 99 percent chance of warming less than 1.1 deg C.

  2. Report this comment

    Bruce Bernott said:

    The comment #1 posted by Mark Bahner is excellent – please answer his 8 questions, so we casual readers can get a little clearer understanding of this debate.

    “Given that human induced climate change is with us” – what percent of the last 100 years’ temperature change is human-induced? 100%, 25%, 5%?

    What has caused the global warming on Mars?

  3. Report this comment

    JamesG said:

    The use of the term “doubter” is much nicer than “denier”. Can everyone start to use this term so we can start to converge on some things?

  4. Report this comment

    Jos de Laat said:

    I noted the following remark:

    “It is true that IPCC does not refer to Armstrong’s work as it has dubious relevance.”

    Could anyone please explain why it is of dubious relevance?

    My personal impression was quite the opposite: it hits the nail right on the head. The debate about projections vs predictions is non-relevant to me. There would only be a difference between them if one would define a prediction as a projection with a proven level of skill and thus a well quantified uncertainty. In weather we can speak about forecasts because of evaluation of afterwards. However, IPCC projections are not called predictions exactly because of a lack of (proven) skill. Which is an important part of the Armstrong paper, and points to a serious issue: if we can’t estimate the reliability of projections, does it have much use beyond the science, apart from providing stories that may, or may not occur? I would argue not, because then a climate model becomes, as is stated by Armstrong, merely a numerical expression of how we as scientists believe climate is working, an expert judgment. And this is where the Armstrong’s work adds value.: expert judgment does not provide a solid basis for doing predictions.

    So one would like to know the reliability of the IPCC projections. But how? What are the requirements for a climate model to be trusted? I wouldn’t know, since there is only one climate, one realization of the complex interactions within the system (unlike weather forecasts). I don’t think currently there exist sufficient long term observational datasets to constrain climate models, which is highlighted by various recent papers in science journals that all emphasize this dire need for longer datasets.

    And that’s is a worrying observation, considering the amount of attention climate has gained and the impact of IPCC on politics and society.

  5. Report this comment

    guthrie said:

    Those genuinely interested in why Armstrongs work is irrelevant will be interested in this post at realclimate:

    http://www.realclimate.org/index.php/archives/2007/07/green-and-armstrongs-scientific-forecast

    As for the Martian warming, that is due to albedo changes brought about by dust storms, and nothing to do with the sun.

    AS for reliability of climate models, I suggest you go and read up on realclimate.org, the IPCC reports and any other websites, as well as Spencer Wearts book on global warming, which is available online.

    The quick answer regarding reliability is “IPPC projections are pretty reliable”, but actually explaining it would take pages of material.

  6. Report this comment

    Luke Silburn said:

    Bruce@3 “What has caused the global warming on Mars?”

    There has been no global warming observed on Mars. The phenomenon you are referring to is a regional warming trend that has been inferred by observations of ice cover changes at the southern pole since 2001 (ie over three martian years).

    Given how much seasonal and inter-annual variability there is in the Martian climate, three years of proxy observation covering a rather limited locality is nowhere near enough data if you want to draw any meaningful conclusions about possible global trends.

    Regards

    Luke

  7. Report this comment

    MIchael Tobis said:

    AR4 and the consensus it represents doesn’t address Bahner’s questions in the first posting explicitly because they don’t have unique answers as specified.

    It is necessary to specify a time frame and a forcing trajectory.

    It will be very surprising if the equilibrium sensitivity to CO2 doubling, all else equal, turns out to be outside the range 1.5-4.5 C, but that isn’t a prediction what will happen.

    That is, after a doubling of CO2 equivalent and waiting a few decades for the ocean to catch up, leaving aside land ice and carbon feedbacks, the world will warm about 3 C +/- 1.5. A quadrupling leaves 6.5 C well within range, while an very rapid limit on net emissions leaves 1.1 C feasibly within range.

  8. Report this comment

    Mark Bahner said:

    AR4 and the consensus it represents doesn’t address Bahner’s questions in the first posting explicitly because they don’t have unique answers as specified.

    In other words, the answer to every single one of the 8 assertions is “Don’t know,” correct?

    So one could read the summary of temperature projections in Table SPM-3 of the IPCC Fourth Assessment Report, and the answer to assertions #4 and #8:

    4) The IPCC thinks there is more than a 99 percent chance of warming more than 6.4 deg C, and

    8) The IPCC thinks there is more than a 99 percent chance of warming less than 1.1 deg C…

    …would be “don’t know.”

    Why hasn’t Olive Heffernan or someone else simply written that in the past two weeks? Why is it so hard to state a simple truth? Isn’t science all about finding the truth?

  9. Report this comment

    Ron Cram said:

    Trenberth writes “It is true that IPCC does not refer to Armstrong’s work as it has dubious relevance.” Not true. Armstrong’s work is very relevant as he performed an audit of the procedures used by the IPCC to come up with their projections. He found a number of key criteria that should have been considered were not.

    People do not like to be audited, especially when mistakes are found. This situation is analogous to the Hockey Stick controversy and the claim by Michael Mann that the work of McIntyre and McKitrick was not relevant. Mann, who now admits not being a statistician, tried to innovate statistical methods without checking with any statisticians. When statisticians (Wegman Report) looked into the debate, they agreed with McIntyre and McKitrick.

    Trenberth also writes that Armstrong and Green are “highly critical of IPCC procedures and ‘forecasts’ for not being based on ‘evidence based’ procedures as outlined in an earlier 2001 book of his.” This makes it sound as though Armstrong is the only academic who is involved in scientific forecasting and the only one who has been critical of the IPCC. Not true. Armstrong is one of the most respected academics in the field of forecasting. He is one of the founders of the Journal of Forecasting (founded in 1982) and the International Journal of Forecasting, See http://www.forecasters.org/ijf/

    The fact Trenberth and other climatologists are making projections without checking with the forecasters to see if they are doing it right is just sloppy work.

    Several other academics have been critical of IPCC projections. I am certain Trenberth is aware of the criticisms of Roger A. Pielke. I am not certain if Trenberth is aware of criticisms by Hendrik Tennekes, Antonino Zichichi and Orrin Pilkey. One would expect Trenberth to show a little more humility when IPCC projections are being criticized by leading scientists in different fields and for different reasons.

  10. Report this comment

    bi said:

    Since Green and Armstrong don’t seem to be showing any humility to the thousands of scientists in the IPCC, and they couldn’t be bothered to check with the IPCC to make sure they got their climate science right… why should Trenberth extend this courtesy to them?

  11. Report this comment

    Robert Fildes said:

    Dr Trenbarth’s interesting comments a year ago to the International Symposium on Forecasting has formed part of my preparation for my own, in Nice at ISF2008. I’m most struck by his reluctance to acknowledge that there has been at least 40 years of forecasting research, much longer if weather forecasting is included, and there are lessons to be drawn from this earlier research, relevant to climate prediction and forecasting and the question of global warming. Nobel prizes have been awarded in the area, many books written and successful academic journals established (Fildes 2006), with many substantive conclusions drawn (Armstrong (2006) – which of them are directly relevant to IPCC Report 4? Green and Armstrong’s (2007) article makes a number of severe criticisms of the IPCC and it was disagreement with their conclusions that led me also to decide to examine the report.

    Having examined recent data from a forecasting perspective, I broadly accept that we’re likely to experience global warming over the next 20 years. So where’s the problem? – the problem is that the writers of the report seem totally unaware that many of their claims are not adequately substantiated. If these researchers had spent even a short time looking at the forecasting research literature they would know that econometricians have for years adequately distinguished between a prediction (i.e a conditional forecast) and an unconditional forecast. Moreover, the limitations of the large climate simulation models in terms of identifiability, unconditional forecasting and control have been discussed for many years (see e.g. Shackley et al, (1998); Young and Parkinson (2002); Young (2006)). So why don’t the IPCC scientists address these limitations? The IPCC report makes a number of conditional forecasts. As G&A point out the word forecast is used as often as prediction in the report – and in forecasting research, when the distinction is necessary, researchers are careful to distinguish the conditional from the unconditional. Similarly the issue of within-sample fitting of models has long been established as a dangerous practice leading to overestimates of forecasting accuracy. Dr Trenbarth’s statements attempt to suggest that this does not take place – but a reading of the IPCC report makes it clear that it does, and that judgment, another core aspect of forecasting research is also used in calibration. What we therefore observe in the usually graphical output from the models is an illusion of accuracy, unvalidated by out-of-sample tests. The final important point is a core element of G&A’s critique – the lack of the use of benchmark comparisons. Again, we’ve learnt that this is a crucial aspect of learning how to improve models and their accuracy. A challenge – I’ve searched through the climate modelling literature. Where are the benchmark comparisons?

    Within the scientific simulation literature there has been no definitive source discussion simulation model validation, despite it being a well-known trap for modellers that have led authors such as Pidd (2003) to try to clarify the tests that a valuable model should pass. It seems to me that the climate modellers have fallen into the trap of ignoring other areas of research in the vain belief that their approach will lead to the knowledge we all need. A bigger computer and more data will not (necessarily) solve the problem (and I write this as member of the UK’s high performance computing strategy committee). Time for the climate modellers to read more widely and recognize other disciplines have valuable perspective that should lead to improved models, forecasts and policy and for funding bodies to take a more multi-disciplinary perspective.

    My presentation together with Peter Young’s will be available shortly on the ISF2008 web site.

    Armstrong, J.S. (2006). Findings from evidence-based forecasting: Methods for reducing forecast error. International Journal of Forecasting, 22, 583-598.

    Fildes, R. (2006). The forecasting journals and their contribution to forecasting research: Citation analysis and expert opinion. International Journal of Forecasting, 22, 415-432.

    Green, K.C. & Armstrong, J.S. (2007). Global warming: Forecasts by sciientists versus scientific forecasts. Energy & Environment, 18, 997-1021.

    Pidd, M. (2003). Tools for thinking, Chichester, UK: Wiley.

    Shackley, S., Young, P., Parkinson, S., & Wynne, B. (1998). Uncertainty, complexity and concepts of good science in climate change modelling: Are gcms the best tools? Climatic Change, 38, 159-205.

    Young, P.C. (2006). The data-based mechanistic approach to the modelling, forecasting and control of environmental systems. Annual Reviews in Control, 30, 169-182.

    Young, P.C. & Parkinson, S. (2002). Simplicity out of complexity, In M. B. Beck (ed.), Environmental foresight and models: A manifesto ((pp. 251-294; Oxford: Elsevier).

Comments are closed.