BOOKS ET AL. standing current plans for weather modifi ca- poses instead a middle path that seeks neither tion and geoengineering. Today’s planetary dominion over weather nor a diminishment , he argues, imagine themselves as of dangers that environmental problems pose technological pioneers and are unaware of for the 21st century. Acknowledging pre- decades, if not centuries, of similar specula- vious attempts at climate control can offer tions. More critically, technological solutions researchers and policy-makers some valuable for global warming have the potential to dis- historical lessons. Fixing the Sky provides an tract policy-makers and citizens from strate- essential foundation for understanding the gies based on changing regulatory regimes long and dubious scientific tradition from and people’s behavior. which plans for climate control hail. When it comes to addressing climate change, technological fixes remain attrac- References Bull. At. Sci. 22 tive to many conservative and libertarian 1. A. M. Weinberg, (10), 4 (1966). politicians and economists. Fleming pro- 10.1126/science.1201627

HISTORY AND PHILOSOPHY OF SCIENCE Computing the Climate and More Richard C. J. Somerville s a thought experiment, imagine is both a history of modern climate science an alternative history of our planet, and an analysis of the relation between that A one in which the digital electronic science and today’s concerns about global on January 14, 2011 had never been invented. Then ask warming. Climate contrarians often assert yourself how our ability to understand 21st- that computer of climate are century climate change would have been unreliable and that climate science should affected. Because fossil fuels were already instead deal with observations of the actual dominant in the precomputer era climate. Although their before World War II, it seems likely A Vast Machine primary motivation may that even in this alternative his- Computer Models, be opposition to possible tory they would power our global Climate Data, and policies, their stated con- . So human-caused the Politics of cern is usually a distrust www.sciencemag.org climate change would presumably Global Warming of mainstream climate now be well under way, much as it by Paul N. Edwards science. The contrarians is today. Burning coal, oil, and natu- MIT Press, Cambridge, MA, insist that this science ral gas would still produce a signifi - 2010. 546 pp. $32.95, £24.95. ought not to depend on cant increase in the amount of carbon ISBN 9780262013925. an analysis of virtual dioxide in the atmosphere, which climates produced by

would drive an altered greenhouse models, which are, after Downloaded from Science in the Age of effect and shift climates around the all, mere computer pro- Computer globe. In this hypothetical world grams. Edwards, how- without , however, climate by Eric B. Winsberg ever, makes the reader science would surely have achieved University of Chicago Press, understand that mod- only a relatively primitive level of Chicago, 2010. 164 pp. $66. els play a central role in understanding. Researchers would ISBN 9780226902029. Paper, producing nearly all the $24. ISBN 9780226902043. have neither the observational evi- climate observations dence to adequately document cli- that scientists use. For mate change nor the scientifi c tools to under- example, converting satellite measurements stand and predict it. In our actual world, that of atmospheric radiances into “observations” plentiful evidence and those powerful tools of is a complex task involving both rely heavily on computer models. models. In six short words, the central mes- A Vast Machine explores the ramifi cations sage of this book is “without models, there of this key insight. Paul N. Edwards (a his- are no data.” torian of science and at the Uni- Edwards traces the development of mod- versity of Michigan) has written a book that ern models of the climate system, research that branched off from numerical weather in the 1950s. Such weather pre- The reviewer is at Scripps Institution of Oceanography, Uni- versity of California, San Diego, CA 92093–0224, USA. diction relies on computer simulations that E-mail: [email protected] start with observations of present meteoro-

www.sciencemag.org SCIENCE VOL 331 14 JANUARY 2011 Published by AAAS BOOKS ET AL.

logical conditions and calculate the led rapidly to operational numer- detailed evolution of the state of ical weather forecasts in the the atmosphere (temperature, pres- mid-1950s, fi rst in Sweden and sure, winds, and humidity) for up the United States. Some scien- to about two weeks. This model- tists also foresaw a new role for centered process is the basis of the modeling. Von Neumann him- daily weather forecasts that appear self explicitly stated that com- on television, radio, and the Inter- puter simulations might “replace net and in newspapers. Present-day certain experimental procedures climate models are usually more in some selected parts of math- comprehensive physically. Typi- ematical .” Indeed, when cally, such models simulate a com- carrying out simulations rou- plex, interactive system compris- tinely in many fi elds of science, ing not only the atmosphere but researchers today often speak of also oceans, snow and ice, land “numerical experiments.” surfaces, and biogeochemical pro- Winsberg comes at this issue cesses. Compared with weather Computing power for early modeling. ENIAC fi lled a large room, con- from the perspective of a philos- forecast models, climate models sumed 140 kilowatts, and contained some 18,000 fragile vacuum tubes. opher. He reminds us that philos- also simulate far longer peri- ophy of science has historically ods, from months to millennia. cern “is as much about what philosophers always drawn its motivations and directions A Vast Machine does an especially good of science should learn in the age of simula- from the science of the day. He notes that job at recounting details of the historical evo- tion as it is about what philosophy can con- computer simulation now occupies a cen- lution of these models, without drowning tribute to our understanding of how the digi- trally important place in many fi elds of sci- the reader in jargon and, amazingly, without tal computer is transforming science.” He is ence. However, as he points out, such simu-

using any mathematics at all. Edwards has interested in issues such as the relationship lations typically are carried out within the on January 14, 2011 interviewed many of the pioneers in the fi eld between experiment and computer simula- constraints of existing fundamental theory, and has clearly explored the research litera- tion. He asks, for example, under what condi- rather than changing or revolutionizing the- ture extensively. His account will be readily tions should we expect a computer simulation ory. Winsberg suggests that philosophy of accessible to that legendary target, the general to be reliable? science, in these contemporary scientifi c cir- reader, a broadly educated person interested in Such questions, while of broad impor- cumstances, ought now to concern itself with science and technology, among other things. tance to science in general, are also highly the subject of simulating complex phenom- Such a reader will, I expect, be particularly relevant to climate modeling, and here it may ena within existing theory, as opposed to its interested in Edwards’s penetrating analysis be helpful to establish some historical con- traditional focus on the creation of novel sci- of the role of climate science and models in text. Computer simulations in science began entifi c theories. www.sciencemag.org the current political and policy discussion of after World War II and were at fi rst confi ned to In Science in the Age of Computer Simu- how best to meet the imposing challenges of and nuclear weapons research. lation, Winsberg explores this new direction confronting man-made climate change. John von Neumann, a towering fi gure in 20th- in depth, with extensive references to both The book will also fascinate members of century mathematics and mathematical phys- philosophical and scientifi c developments. the climate modeling research community. ics, understood immediately that these two He concludes, Although I have worked in this fi eld for more seemingly disparate fi elds are scientifi cally

than 40 years, I encountered many surpris- closely connected: both being centrally con- [W]hat we might call the ontological Downloaded from ing and fascinating nuggets—including the cerned with highly nonlinear fl uid dynamics. relationship between simulations and disclosure that one celebrated early climate And in both, carrying out controlled experi- experiments is quite complicated. Is it modeler never learned to program comput- ments and making measurements present true that simulations are, after all, a ers. Edwards’s coverage does have a few limi- great difficulties. The ENIAC (Electronic particular species of experiment? I have tations. He primarily emphasizes contribu- Numerical Integrator and Computer), the tried to argue against this claim, while at tions from scientists in the United States, as most important American computer during the same time insisting that the differences he acknowledges, and he focuses much more this period, was completed in late 1945 and between simulation and experiment are strongly on the atmosphere than on other initially used for hydrogen bomb calcula- more subtle than some of the critics of the important components of the climate system. tions. Primitive by modern standards, it had claim have suggested. Most important, Within these limitations, the author’s impres- a tiny memory and could carry out fewer than I have tried to argue that we should dis- sive scholarship and command of his material 400 typical multiplications per second. One connect questions about the identity of have produced a truly magisterial account. early numerical weather prediction actually simulations and experiments from ques- Science in the Age of Computer Simula- calculated not the meteorological conditions tions of the epistemic power of simulations. tion is a very different book, and it may pri- at Earth’s surface but atmospheric circulation marily intrigue a very different audience. In at an altitude of about 5500 meters. In 1950, Such provocative findings, and Wins- it, Eric Winsberg (a philosopher of science such a prediction on the ENIAC required berg’s exceptionally readable account of the at the University of South Florida) consid- some 24 hours of computing time to produce reasoning that led him to them, will interest ers the rise in the importance of computer a 24-hour forecast. many general readers as well as scientists and simulations, not only in climate research The crude but promising early computer philosophers of science.

but throughout science in general. His con- simulations astonished meteorologists and 10.1126/science.1200208 ARMY U.S. CREDIT:

150 14 JANUARY 2011 VOL 331 SCIENCE www.sciencemag.org Published by AAAS