Dr. Andreas Sterl graciously accepted my invitation to post a guest weblog in response to the Climate Science weblog on his paper (see).
His reply and guest weblog are posted below, along with my response
E-MAIL FROM DR. STERL
Dear Dr. Pielke,
Thank you very much for your interest in our work and for your invitation to reply to your weblog. It is only now, after my vacation, that I have the opportunity to reply, and I hope that it is not too late. Please find my reply below.
With kind regards,
Andreas Sterl
GUEST WEBLOG BY DR. ANDREAS STERL OF KNMI
In our research we used a state-of-the-art global climate model (not a regional one, as the weblog suggests) and forced it by time varying greenhouse gas concentrations (including CO_2, CH_4, N_2O,several CFCs, and sulphur) according to the SRES A1b scenario. We analyse extreme 2m-temperatures, expressed as 100-year return values, both for the present climate and for the future one. The main result (which, by the way, is not really new) is that extreme temperatures rise faster than the mean temperatures. We compare present-day values between model and observations, and use the difference to correct future values and present those corrected values. They appear to far exceed 40 deg C (104 deg F) in large part of Europe and North America, and reach 50 deg C (122 deg F) in parts of the Middle East and northern India. It is obvious that such temperatures would pose serious health and economic problems.
I cannot see why this should be an “abuse of the scientific process” or “an embarrassment to the scientific community”. The scientific community obviously likes the paper - the reviews were quite positive, and GRL even highlighted it (http://www.agu.org/journals/scripts/highlight.php?pid=2008GL034071) - and does not seem to be embarrassed. Anyhow, the scientific process requires a thorough description of what has been done, so that everyone can reproduce the results. In this sense our work is purely scientific: Take the same model, run it with the same forcing, apply the same analysis, and you will get the same results.
Of course you can argue that reproducibility is not enough: you should also start from basic and recognized principles. For instance, it is of no use to employ a climate model which does not include the absorption of long-wave radiation by CO_2 to make inferences of the climate at the end of this century, or to assume an emission scenario in which the CO_2 emission drops to zero in 2009 and stays there for the rest of the century. A paper written using such a model or such an emission scenario should clearly be judged as “unscientific”. However, where is the limit? We used the SRES A1b scenario, and while everyone would agree that it is more realistic than the above-mentioned zero-emission scenario, we can disagree a lot whether A2 (say) would be even more realistic and therefore be preferable. The fact is simply that we do not know. A1b is plausible, not violating basic principles (as does the zero-emission scenario), but the same is true for A2, so why not use that? The choice is purely subjective, and every modeller knows that. He (she) also knows that using A2 instead of A1b will give different results - at least quantitatively. Qualitatively, the results are often identical. In our case, for instance, the result that extreme temperatures grow faster than mean temperatures would certainly survive, simply because there is a good physical reason why this should be case. The same reasoning applies to the model used. Had we employed a different climate model the numeric results would certainly have been different, but the qualitative result would be the same (as is shown in, e.g., Kharin et al., 2007, J. Clim. 20, 1419-1444).
So, yes, there are a “modelling uncertainty” and a “scenario uncertainty”, and therefore we do not sell our results as “forecasts” as is claimed in Pielke’s weblog (looking before shooting might help sometimes). Our results are to be read as “under an A1b emission scenario, using the ECHAM5/MPI-OM climate model and assuming a bias correction employing present-day ERA-40 values, the 100-year return temperature in Phoenix, Az (say) at the end of this century will be 49.1 deg C (120.4 deg F)”. Of course, using a different model, a different emission scenario, a different correction would give a different number.
We know that the model we used is not perfect - no model is. It fails to reproduce present-day extremes well - if we regard ERA-40 values as a good proxy for them. So an additional conclusion from our paper is that the particular aspect of extreme temperatures needs improvement in the models. What is unscientific about that? You have a tool, use it, observe that it is not perfect, and start trying to improve it. This evolutionary process is called scientific progress and is the heart of science.
To conclude,
* our paper obeys the rules of scientific conduct,
* extreme temperatures rise faster than average temperatures,
* they might reach problematic levels in the course of this century,
* the precise numbers depend, inter alia, on emission scenario and
model used,
* models need improvement in the very important aspect of extreme
temperatures, and
* one should read and (try to) understand a paper before criticising
it.
RESPONSE FROM ROGER A. PIELKE SR.
Dear Dr. Sterl
Thank you for your reply and guest weblog! This exchange of viewpoints will significantly benefit the scientific process. I will set up for your guest weblog to post tomorrow.
My comments regarding the publication of unvalidated forecasts remain, however. We know less about how the climate responds to human and natual climate forcings than represented in the global model simulations you report on. I have no issue with presenting such results as sensitivity studies, but that is not how they were presented in your paper, nor how they are interpreted by those in the policy and impacts communities.
I very much appreciate that you have engaged in the discussion of this issue, and readers of Climate Science can form their own conclusions about the merit of your and my perspectives. Climate Science will be blogging further in the coming weeks on the scientific value of publishing climate model predictions which are not yet verified.
I look forward to continued discussion on this subject.
With Best Regards
Roger