Head in a Cloud

troposphere and stratosphere meet blogosphere

Head in a Cloud random header image

Quantifying Climate Change

January 26th, 2007 by Hasenkopf · 4 Comments

I was going to post about some recent cavity ring-down experiments, but unfortunately for CRD-enthusiasts (any maybe fortunately for everyone else), an article appearing earlier this month in GRL on quantifiably defining climate change as a single index and using this lone index to explore climate change through the end of this century caught my eye instead. The modeler in me (which doesn’t get out much since I don’t model) was intrigued by the logical and systematic way in which the authors (Michele Baettig et. al) defined a seemingly qualitative, all-inclusive concept such as climate change. The article, A climate change index: Where climate change may be most prominent in the 21st century, focuses on providing policy-makers with an information-packed, yet as least misleading as possible, index of climate change relative to current variability.

So how do you go about defining a climate change index (CCI)? It seems to me that *all* you do is determine the most important indicators of climate change, pick a period of time over which to project a Global Climate Model (GCM), compare future indicators to the ones in a current reference period, and then spatially define the region over which you are evaluating the indicators. Once you have an ensemble of indicators for a specific region, you can get a single index by averaging them in some manner you deem fair. Voila, your very own CCI.

The experts chose 9 indicators that can be described in four general groups: (1) changes in annual temperature, (2) changes in annual precipitation, (3) changes in extreme temperature events (ie warmest summer in 20 years), and (4) changes in extreme precipitation events (ie driest winter in 20 years). They looked at these changes over a 3.5 x 3.5 degree grid (the maximum grid spacing is about the size of Nepal and the average is roughly the size of Kuwait) over the entire globe. This is a small enough spatial scale to allow the climate change index of single countries to be resolved from the model, which is useful for policy-makers, for which the indexing system is designed. They ran three GCMs, with each data set using two future (years 2071-2100) scenarios of increased CO2 levels (based on IPCC SRES A2 and B2 scenarios) and a reference run from 1961-1990.

The indicators are evaluated with a “1 in 20 years criteria” relative to the reference run. What does this mean? For instance, if we were looking at the indicator group for change in annual temperature (which also happens to be an indicator and is therefore easy to use as an example) for, say Nepal, we’d look at what the highest annual mean temperature during a 20 year period in Nepal from our reference run was, and then see what the future scenario models predict. If the same extreme annual mean temperature occurs 5 times in 20 years in the predicted 2071-2100 period – or 4 more times than during the reference period, then the value for this indicator would be 4 for the region. To find the CCI, the authors found the multi-modal average of the indicators for both future scenarios and averaged them out to a single CCI per region. That’s the gist of the method they used; see the paper itself for more details and a critique on both the robustness and shortcomings of their method.

According to this methodology, here’s (linked to the JGR article) what the global climate appears to look like at the end of this century. To me, one of the most remarkable results is the increase in frequency of additional hottest years; what is now considered an annual mean temperature that tops any of the twenty years around it will be an average year at a given location on the globe by the end of the century. Looking at CCIs computed around the globe show the strongest climate changes at the poles and the equator. The authors point out that this method predicts more less-developed countries will have larger climate changes than developed ones.

Tags: climate · modeling

4 responses so far ↓

  • 1 peter lockhart // Jan 26, 2007 at 8:43 pm


    I agree, from a scientific perspective, it is not hard to see need for a grossed up view of the net effects (outputs) of climate change, and as suggested looks a sound approach.

    But we’re all in this together, it makes sense that the measures need also to take into account the input side, both natural and anthropogenic. In the case of our contribution measures must have consistency for individuals, across the topic and around the world.

    Lots of folks already recieve the annoying little graph which appears on the quarterly electricity bill, explaining how many tonnes of C02 their accomodations have produced, there needs be linkage and reporting at a community and also at a country level (example).

    It is not until people can see how their individual actions impact those around them, that the case for sustained community behaviour change can occur. Hopefully the politics and business community still follow the will of the people.

    Good post !


  • 2 Bill F // Jan 29, 2007 at 7:13 pm

    While I like the general idea of that kind of index, the actual output discussed in the post is worthless IMO. Are you really trying to say that we can predict what the weather of Kuwait will be 100 years from now? Or Nepal? Or the North Pole? I think what climate scientists need to do to convince the public to act is not to find a new means of graphically overstating the accuracy of their current predictive ability, but instead find a way to show people that their current predictive ability is actually valid for real life data. If you want people to believe that a model can predict what the temperature or rainfall patterns in Kuwait or Kansas will be in 2100, show them that the model can start in 1920 with a 40 year data set starting in 1880 and then accurately predict what the conditions will be in 1960. Then open up the code of the model and show that all of the equations match the current understanding of the physical processes they are intended to represent and aren’t simply fudge factors applied to make the model match the desired result. If a model can take real data from a period of time, and predict real data for another period of time that can be used to validate the model, then you can use it to convince people that they might want to do something about it. Until then, people who believe the model will be for action and people who distrust the models will be against action. Chaning the graphics doesn’t fix the underlying issue, which is lack of trust in the predictive capability of the model.

    FWIW, I am not saying any or all of the models currently out there are wrong or are incapable of doing what I said above. What I do know is that nobody has published a clear demonstration that their model can be initialized at some point in the past and accurately model the conditions at a given point in time beyond that starting point while simultaneously opening up their code and inputs to scrutiny to show that they aren’t simply fudging the solution.

  • 3 seand // Jan 30, 2007 at 4:38 pm

    Bill F,

    You’re argument that models haven’t been validated is quite simply not true. The point of this site, and this post, is not to get into a “Climate Wars” skirmish that a full response to your post would precipitate.

    Instead, as a start, I urge you to read the IPCC. They have a new assessment report coming out in the next couple of weeks, but even their last report should be sufficient to answer all of your questions about how models are validated. Another great resource is Real Climate.

    In short, everything you have suggested has been done. Not only has it been done, but it continues to be done as models are improved and refined. No doubt there are continuing model uncertainties (especially w.r.t. regional climate change), but at this point the spread of climate projections for the future is more due to uncertainty in future emissions than the uncertainty in climate sensitivity.

    The point of this article was to try to come up with a better framework for knowledge transfer from model output to inform policy. The efforts described in this paper represent one way by which this can be done. There are no doubt other methods, perhaps better for informing certain policy decisions.

    But by no means should this paper be seen a case of climate scientists trying to “convince the public to act”. Rather, this paper has the aim to provide interested policymakers with estimates of regional impacts of climate change in an easily digestible manner.

  • 4 Hasenkopf // Jan 31, 2007 at 11:07 am

    Michele Baettig was kind enough to write me an email regarding this post and specifically answer: “How do you plan to introduce this system of calculating a CCI to policymakers (since they are unlikely GRL readers)? Is this being published/presented in another forum?”

    Hi Christa,

    Thank you for your information about your blog. I read your article. I like it very much!

    Concerning your question about presenting the CCI to policy-makers: At the moment, numerous newspapers, magazines, radio, and TV worldwide have reported on the CCI. ETH Zurich has written a press release [read here]. However, about activities in/of a forum for policy-makers I do not know.

    Currently, I am finishing my PhD (and am a little short of time). But I should contact the Swiss forum for politicians – good idea. I will do that in the next days…

    Best regards,
    Michèle Bättig

Leave a Comment