date: Sun, 7 Dec 2003 07:49:37 -0800 (PST) from: Stephen H Schneider subject: Dessai-Hume review (fwd) to: Mike Hulme Hi Mike, hope all is well. Haven't heard back about the APril review and my very close schedule--any decisions? I forward to you--spoke to Suraje already about it--my review of your excellent paper on uncertainties, but of course a few mostly narcissistic nit-picks. hOpe it is useful. Cheers, Steve PS pls forward to Suraje, I'v misplaced my address book ------ Stephen H. Schneider, Professor Dept. of Biological Sciences Stanford University Stanford, CA 94305-5020 U.S.A. Tel: (650)725-9978 Fax: (650)725-4387 shs@stanford.edu ---------- Forwarded message ---------- Date: Sun, 7 Dec 2003 07:40:24 -0800 (PST) From: Stephen H Schneider To: Climate Policy Subject: Dessai-Hume review HI Ray, sorry to take so long with this, but I finally read it on the plane to COP9-just discussed my minor complaints with Suraje, so he knows who I am-I nearly always self-confess, as I encourage most Climatic Change reviewers to do, but of course I do not insist. In short, this is an excellent review, brings lots of literature in-some of which even I who am at the center of this uncertainties battle-didn't know, so it will be a clearly valuable entry for the Climate Policy readership and beyond. It lays out the paradigmatic differences among groups fairly, and tries to be neutral in laying out pros and cons. Some in certain schools will think that wimpy, but it is the best summary I've seen of the state of the art, so my hat off to Suraje and Mike for a fair and balanced piece. It could be shorter and still make it's main points, but then some of the excellent scholarship would be lost so I vote to publish it about how it is now. Of course, I have a few nit-picks, mostly narcissistic, which I'll list below. Other than that I think it should be provisionally accepted right now subject to a final version that deals with my minor comments and other reviewers comment--presuming you get some of those too. P10-analogs discussion. While literature is cited about analogs to past adaptation, the authors need to warn the readers that global change forcing may be unique and no-analog impacts seem likely, so analogs, either to paleoclimatic states or adaptations are just the backdrop against which we calibrate our understanding of how the system works, but not necessarily analogs to the unique and transient changes now evolving. Also on this page, in the middle, the Pielke and Sarewitz little polemical sentence is quoted suggesting irrelevancy of probabilities for "climate adaptation policy". This is a speciesist prejudice-only humans count. For plants and animals, for which adaptation is much less likely, but systems would be damaged, for humans to decide how much they worry about this possibility, relative to other calls on our scarce resources, probabilities are essential, not irrelevant. The likelihood of 2 versus 5 degrees is the difference between some species lost and a mass extinction event. Also, in a sentence below the meaningless word "accurate" is given. As Moss and I complained two thousand times in the TAR, words like accurate, definitive, certain etc are meaningless rhetoric if not defined versus a quantitative scale of subjective probabilities, since one analyst's "accuracy" might be a 95% chance of something being true, where another's is a 5% chance because they adhere to precaution rather than proof. Just unpack this a bit with caveats along the lines I call for here. P14; The worry that uncertainty may increase with more research is a certainty, in fact lots of literature--including later in this paper--show how climate sensitivity has grown with research. Of course it will narrow as nature continues to perform the warming experiment, but no need to be tentative-some things will grow less certain, others more as research progresses, depending on the maturity of the field at the point of the research increase and to some extent on luck. More complex systems more likely to have uncertainty grow at first with more research than simple well-constrained systems. P15 The point that neither I nor Naki/Arnulf explicitly mentioned reflexivity" is a bit unfair for two reasons. (1) We were debating in a narrow column-SRES scenarios/storylines which were self-constructed to be "policy independent". Now they can criticize rightly SRES for thinking such a thing is meaningful, but we kept our debate in those citations to those issues mainly for the one-point-at-a-time principle. (2) The second reason is in my rebuttal to Naki/Arnulf a year later in Climatic Change (2002) that Suraje/Mike do cite, I explicitly address this as in the quoted section below (see especially the caps), though I don't use the word "reflexive" but rather feedback, but it means the same (quote on page 445 of my Editorial): Moreover, Grübler and Nakicenovic (2001) also argue that probabilities in natural science are different from those in social science, since we can perform frequency experiments in the former, whereas in the latter we must make judgments. Grübler and Nakicenovic say that in an interdisciplinary scientific assessment, the concept of probabilities as used in natural sciences should not be imposed on the social sciences. Probability in the natural sciences is a statistical approach relying on repeated experiments and frequencies of measured outcomes, in which the system to be analysed can be viewed as a 'black box'. Scenarios describing possible future developments in society, economy, technology, policy and so on, are radically different. First, there are no independent observations and no repeated experiments: the future is unknown, and each future is 'path-dependent': that is, it results from a large series of conditionalities ('what if. . . then' assumptions) that need to be followed through in constructing internally consistent scenarios. Socio-economic variables and their alternative future development paths cannot be combined at will and are not freely interchangeable because of their inter-dependencies. However, natural scientific projections for the future still require judgments, as no frequency experiments can be made before the fact. We must still assume that our assumptions which govern the structural design of our systems models will hold in the future, often for values of dependent variables that are outside of the range of past experience. Moreover, there are conditionalities in natural science as well, and the solutions are, like Grübler and Nakicenovic rightly assert for social systems, 'path dependent' for natural systems as well as social systems. Therefore, I believe there is no in principle difference between natural and social sciences in this regard, since both require feedback mechanisms and contain path dependent systems. However, I agree there is one aspect in which social systems are harder to predict than natural systems. Although in both social and natural systems interactions among subsystems can cause alterations over time, IN THE CASE OF SOCIAL SYSTEMS, CHANGING BELIEFS AND ATTITUDES, THEMSELVES PARTIALLY DRIVEN BY INFORMATION ABOUT HOW THE SYSTEM IS EVOLVING, CAN LEAD TO MODIFICATIONS OF POLICY CHOICES. While the latter property of social systems is different in kind from natural system predictions, to me both natural and social systems models involve the necessity to model feedback processes, and thus are very similar. In essence, we need a systems model that explicitly deals with the many subcomponents that we believe will influence the evolving emergent properties of a complex socio-natural system, and that when social sciences are included, the system becomes more complex in detail, but not necessarily in principle. For us simply to redefine the classical definition of risk to consequences alone, because subjective probabilistic analysis is fraught with deep uncertainties, is in essence to offer no advice to the policy community as to how it should order its investments in alternative actions, for without probabilities it is very difficult to engage in risk management. And if we in the scientific assessment business do not offer some explicit notions of the likelihood of projected events, then the users of our products - policy analysts and policy makers - must guess what we think these likelihood estimates are. That is hardly preferable in my view to a carefully worded set of subjective probabilistic estimates in which our (often low) confidence in such estimates accompanies any likelihood statements. . p16-be careful about nobody does reflexive modeling assertions. What about the whole integrated assessment cabal with agent-based decision making responding to evolving climate and mitigation costs. Nordhaus' DICE is the most famous example. I have been personally critical of the assumptions he and other neo-classical economists use in their current models, but in principle they are modeling human reactions to evolving climate and imposing policy changes that feed back on the climate and society. In fact I've said one gets emergent properties of coupled socio-natural systems in the pages of Climate Policy-particularly when abrupt changes are included. See: Mastrandrea, M. and S.H. Schneider, 2001: Integrated Assessment of Abrupt Climatic Changes. Climate Policy, 1, 433-449. So, to be sure feedback-reflexivity-is a major obstacle as asserted, but because it is hard doesn't mean there haven't been some heroic-even if weak-attempts and that many more will and should be forthcoming. Just tell the story straight. P18-Myles Allen has already started, not about to as said at bottom. Might also note that part of the model-data inter-comparison test will reveal model errors, part will reveal errors in the forcings used to drive the model simulations and some error will be in the instrumental data themselves. Thus independent tests-like looking for climate signals in plants and animals--also needed. See, e.g.: Root, T.L., J.T. Price, K.R. Hall, S.H. Schneider, C. Rosenzweig, and A. Pounds, 2003: Fingerprints of Global Warming on Wild Animals and Plants. Nature, 421, 57-60. P24. I think the Clark /Pulwarty quote is itself misleading, since it is missing an essential requirement (in the Moss/Schneider guidance paper to IPCC on uncertainties), which is all probabilistic info-via pdfs, presumably, should also contain a measure of subjective confidence in the pdf itself. So I fully agree we should not wait for perfect information via a single pdf, but we can offer pdfs AND confidence assessments of them in the meanwhile, as better than offering no pdfs at all. Just because some who do not understand probabilities will also not understand probabilistic formulations for problems other than that of climate policy-how about medical or military policy. We cannot refuse to do probabilistic information because of ignorance outside of us, when that is the most honest assessment of the state of the art. What is called for in my view is expert popularization using gambling, health and insurance metaphors to make probabilistic formulations clearer to non-specialists, not abandonment of the most honest descriptors of the state of the art. Most scientists are obscure and lousy popularizers I admit, but correct the problem right, not by suppressing pdfs and subjective confidence estimation-that is my view and I don't expect the authors to necessarily agree with it but I do expect they will raise these issues explicitly in their text and give their views. P25, 1st paragraph-anthropocentrism again. P 26 Statement "Human reflexive uncertainty is unquantifiable in probabilistic terms" is certainly wrong-it has been done in the economics/integrated assessment literature for a dozen years already. Now, is it very credible?--that is another thing. Some predictions-like production will respond to price signals--probably pretty robust, whereas others-how will future generations see the intrinsic value of a songbird-much tougher to have even medium confidence in. But ALL are quantifiable via various techniques: modeling, CV or decisional analytic elicitations. That is where the confidence assessment part comes in, for some such predictions will carry very low confidence and that must be said explicitly-but not all will and thus don't over generalize or miss the distinction between the possibility of quantification per se and its relative credibility-two different things that should be explicitly separated in the text. OK That's my nit-pick list. I look forward to seeing this in Climate Policy soon. OK Ray, LET ME KNOW IT CAME OK. i'LL ATTACH A WORD VERSION OF THE LETTER IF tHAT IS USEFUL TO YOU. Cheers, Steve ------ Stephen H. Schneider, Professor Dept. of Biological Sciences Stanford University Stanford, CA 94305-5020 U.S.A. Tel: (650)725-9978 Fax: (650)725-4387 shs@stanford.edu Attachment Converted: "c:\eudora\attach\Dessai-HumeReview.doc"