A.B. Lovins, “Mighty Mice,” Nuclear Engineering International, pp. 44–48, Dec. 2005, www.rmi.org/images/PDFs/Energy/E05-15_MightyMice.pdf, summarizing ref. 2. A World Nuclear Association critique and my response are at www.neimagazine.com/comments.asp?sc=2033302.
 “Nuclear Power: Economics and Climate-Protection Potential,” RMI Publ. #E05-14, 6 Jan. 2006, www.rmi.org/images/PDFs/Energy/E05-14_NukePwrEcon.pdf.
 A.B. Lovins, I. Sheikh, & A.M. Markevich, “Forget Nuclear,” RMI Newsletter, Apr. 2008, www.rmi.org/sitepages/pid467.php, summarizing refs. 4 and 5.
 idem, “Nuclear Power: Climate Fix or Folly?,” RMI Publ. #E09-01, Dec. 2008,
www.rmi.org/images/PDFs/Energy/E09-01_NuclPwrClimFixFolly1i09.pdf, updating and expanding ref. 3.
 A.B. Lovins & I. Sheikh, “The Nuclear Illusion,” draft-18 preprint posted by permission May 2008 at www.rmi.org/images/PDFs/Energy/E08-01_AmbioNucIllusion.pdf, draft-20 revision to be published in early 2010 in Ambio (Royal Swedish Academy of Sciences).
 A simple introduction is at http://en.wikipedia.org/wiki/Concentrating_solar_power. In spring 2008, J. Romm’s assessment found a practical potential to scale up and mass-produce 50–100+ GW/y of concentrating solar power indefinitely (www.salon.com/news/feature/2008/04/14/solar_electric_thermal/print.html) at a busbar cost Sandia National Laboratory estimated in 2008 at ~8–10¢/kWh once 3 GW has been made. The current order pipeline, with scores of projects (by some counts ~180 contemplated in just Spain and the U.S.), is a substantial multiple of 3 GW and may exceed 40 GW. Some innovators also believe costs around or below 6¢/kWh are coming into view. CSP capacity coming online in 2009 appears competitive with new nuclear capacity. Of course, large-scale deployment in deserts would require dry cooling due to water scarcity -as is similarly or more true for nuclear or coal plants.
 All these and other micropower data, documented to standard industry sources, are posted at RMI’s longstanding database: see www.rmi.org/sitepages/pid256.php, Publ. #E05-04. The 2008 renewable data will be posted shortly, and the latest cogeneration data in late 2009. The 2008 capacity factor of the global installed base is ~66% for all micropower, ~83% for non-biomass cogeneration, ~60% collectively for geothermal/small hydro/biomass/ waste, ~40% and rising for all distributed renewables, ~0.26 for wind, ≥0.17 for PV, and 80% for nuclear power.
 For example, untapped U.S. industrial cogeneration potential is at least comparable to U.S. nuclear capacity and output: O. Bailey and E. Worrell, “Clean Energy Technologies: A Preliminary Inventory of the Potential for Electricity Generation,” LBNL-57451, Apr. 2005, http://repositories.cdlib.org/lbnl/LBNL-57451/. See also P. Lemar Jr., "The potential impact of policies to promote combined heat and power in U.S. industry," En. Pol. 29(14): 1243–1254 (Nov. 2001). Cogeneration potential in buildings is unmeasured but very large, and is not confined to large buildings; e.g., Honda has sold over 100,000 home cogeneration systems, and VW has just entered the market with LichtBlick, which plans a German 2-GW “virtual decentralized power plant” to firm renewable power. Old but still useful estimates of European industrial and building CHP potential are in F. Krause et al., Fossil Generation: The Cost and Potential of Low-Carbon Resources Options in Western Europe, IPSEP, 1994, http://files.me.com/jgkoomey/c49xzn.
 Noted gas expert R.A. Hefner’s The Grand Energy Transition (Sept. 2009 rev. edn., Wiley, Sept. 2009) notes that simply dispatching existing U.S. combined-cycle gas-fired plants before coal-fired plants would displace about onethird of all U.S. coal-fired electricity, lowering CO2 emissions by several hundred million tonnes a year, without building any new capacity. This would increase operating costs by ~2¢/kWh -many times less than substituting new nuclear plants (refs. 4–5).
 Nobody claims that efficiency can “generate power,” but rather that it displaces the need to generate part of the power currently needed to do a given task. “Negawatts” are functionally equivalent, not identical, to megawatts.
 S. Doig et al., “Assessing the Electric Productivity Gap and the U.S. Efficiency Opportunity,” RMI, 2009, ert.rmi.org/research/cgu.html.
 McKinsey Global Energy and Materials, Unlocking Energy Efficiency in the U.S. Economy, July 2009, www.mckinsey.com/clientservices/ccsl/. An authoritative U.S. government study also shows encouraging potential: M. Brown et al., “Scenarios for a Clean Energy Future,” En. Pol. 29(14):1179–1196 (Nov. 2001), LBNL-48031.
 COMPETITEK, The State of the Art series, 1986–92, RMI, 6 vols., 2,509 pp., 5,135 notes. Condensed versions were republished by E SOURCE as the Technology Atlas series,, www.esource.com/public/products/prosp_atlas.asp.
 A. Fickett, C. Gellings, & A.B. Lovins, “Efficient Use of Electricity,” Sci. Amer. 263(3):64–74 (1990).
 A.B. Lovins, “Least-Cost Climate Stabilization,” Ann. Rev. En. Envt. 16:433–531 (1991), citing ORNL/CON-312.
 A.B. Lovins, “Energy End-Use Efficiency,” RMI Publ. #E05-16, 2005 white paper commissioned by S. Chu for InterAcademy Council (~90 National Academies), www.rmi.org/images/PDFs/Energy/E05-16_EnergyEndUseEff.pdf; “Advanced Energy Efficiency,” Stanford Engineering School lectures, spring 2007, www.rmi.org/stanford.
 This emerges clearly from e.g. McKinsey’s January 2009 analysis of how to abate global greenhouse-gas emissions by ~70% at an average cost of just 4 Euro per ton of CO2: www.mckinsey.com/clientservice/ccsi/. (That analysis, however, doesn’t yet include integrative design (ref. 16).)
 Conversely, saving electricity can take four orders of magnitude less capital—three in intensity and one in velocity—than supplying more electricity, turning the capital-hungry power sector into a net exporter of capital to fund other development needs: A.J. Gadgil, A.H. Rosenfeld, D. Aresteh, & E. Ward, “Advanced Lighting and Window Technologies for Reducing Electricity Consumption and Peak Demand: Overseas Manufacturing and Marketing Opportunities,” LBL-30890 Revised, Procs. IEA/ENEL Conf. Adv. Technols. El. Demand-Side Mgt. 3:6-135–6-152 (Sorrento, 2–5 Apr. 1991), LBNL; www.rmi.org/images/PDFs/Energy/E91-23_NegawattRevolution.pdf.
 In utility operators’ parlance, “baseload” actually refers to resources with the lowest operating cost, so they are dispatched whenever available. This definition embraces essentially all efficiency and renewables, since their operating cost is below even that of nuclear plants. Economic (“merit-order”) dispatch next uses nuclear, then coal, then gas-fired plants, in order of their increasing operating cost. Utility resource planners use “baseload” to refer to resources of lowest total cost -information that guides acquisition rather than operation. “Baseload” is also often but erroneously applied by laypeople to the big thermal plants that traditionally produce relatively steady output.
 Some loads are actually steady; others only appear so because of the way they’re aggregated with other loads.
 Jim Harding, who led strategic planning for Seattle City Light, says it has no “baseload” resources in the quoted definition's sense; its assets’ system capacity factor is around 25%, comparable to a mediocre wind turbine’s. Yet retail electricity prices are relatively low and the system is highly reliable. If the 'demand' definition was right, this would be impossible.
 North American Electric Reliability Corporation availability reports, www.nerc.com/page.php?cid=4|43|47.
 IAEA, “Lifetime Unplanned Capability Loss Factor,” www.iaea.org/programmes/a2/index.html, accessed 7 Sept. 2009. The lost output varied from 1.3% in South Korea to 22.9% in Pakistan; the U.S. figure was 7.1%, France
7.6%. The global average in 2008 was 5.3%.
 Michael Eckhart (former strategic planning head of GE’s Power Systems sector) makes the intriguing point that a simple-cycle combustion turbine has a ~97% probability of coming online within 30 minutes of coldstart, while Danish utility operators have demonstrated the ability to predict wind force with 98% accuracy within a 30-minute
window. So which resource is more reliable and which is more intermittent?
 C.L. Archer & M.Z. Jacobson, “Supplying Baseload Power and Reducing Transmission Requirements by Interconnecting Wind Farms,” J. Appl. Meteorol. & Climatol. 46(11):1701–1717 (2007).
 See RMI publications "Intermittent Renewables in the Next Generation Utility," PowerGen-RE, Feb. 2008, www.rmi.org/images/PDFs/Energy/RMI_PowerGen_090924.pdf; "Spatial and Temporal Interactions of Wind and Solar in the Next Generation Utility,” Solar 2008, May 2008, www.rmi.org/images/PDFs/Energy/Solar_2008_in_NGU_090924.pdf; "Spatial and Temporal Interactions of Wind and Solar in the Next Generation Utility: Expanded Analysis,” Windpower 2008, June 2008. http://www.rmi.org/images/PDFs/Energy/RMI_Windpower_NGU_090924.pdf.
 The U.S. fleet’s lifetime average rose to 78.7% through 2008, vs. 77.1% globally and 76.9% for France: www.iaea.org/programmes/a2/index.html. The 2008 global average was 80.0%, the lowest value since 1999 (www.iaea.org/programmes/a2/index.html). Assuming 90% for the average new plant seems a stretch.
 Author’s analysis from U.S. Nuclear Regulatory Commission data posted at www.nrc.gov/reading-rm/doccollections/event-status/reactor-status/2003/i... and www.nrc.gov/info-finder/reactor/.
 This is often done by hydropower (like BPA’s 0.3¢/kWh firming rate), but demand-response “virtual peakers” are comparably cheap and can be very large: FERC has found up to 188 GW of U.S. demand-response potential, the resource may well be even larger, and of the 10 GW just bid into the PJM pool’s auction, 7 GW cleared the market.
 More Fight, Less Fuel, Feb. 2008, www.acq.osd.mil/dsb/reports/2008-02-ESTF.pdf.
 M. Bolinger & R. Wiser (LBNL), 2008 Wind Technologies Market Report, July 2009, p. 49, http://eetd.lbl.gov/ea/ems/re-pubs.html. See also www.awea.org/pubs/factsheets/Backup_Power.pdf, which illustrates the tiny amount of net variability -on a one-hour timescale, just ~1–2% of the renewable capacity- that large additions of variable renewables would impose on various U.S. power systems, and how any extra fuel burned by the resulting reserve capacity would be about a thousandth of the fuel that those renewables displace.
 A useful summary is the European Wind Energy Association’s March 2009 study Integrating Wind; see also EWEA’s The Economics of Wind Energy (www.ewea.org) and Bolinger & Wiser, ref. 31. See also Small Is Profitable, ref. 41, and citations in ref. 5.
 “Wind Energy Myths,” DOE/GO-102005-2137, May 2005, www.nrel.gov/docs/fy05osti/37657.pdf, item 5. Backup and storage are functionally equivalent for purposes of this discussion.
 Summarized and referenced at www.claverton-energy.com/wind-energy-variability-new-reports.html.
 G. Czisch & G. Giebel, “Realisable Scenarios for a Future Electricity Supply based 100% on Renewable Energies,” Risø-R-1608(EN), www.risoe.dk/rispubl/reports/ris-r-1608_186-195.pdf; G. Giebel, N.G. Mortensen, & G. Czisch, “Effects of Large-Scale Distribution of Wind Energy In and Around Europe,” www.iset.unikassel.de/abt/w3-w/projekte/Risoe200305.pdf.
 E.g., M.Z. Jacobson & M.A. Delucchi, “A Path to Sustainable Energy by 2030,” Sci. Am., Nov. 2009, pp. 58–65; on PVs, V. Fthankis, J.E. Mason, & K. Zweibel, En. Pol. 37: 387–399 (2009).
 This is a hoary myth. Around the 1970s and early 1980s, before the issue was well analyzed, many people assumed a limit of 5–10%, then 15%, then 20%, then 25%, then 30%…but all such limits have dissolved on closer scrutiny. For example, the West Danish system operator reports that as he gained experience with windpower, he became confidently able to manage nearly five times more of it than he had thought possible 7–8 years earlier; he
was just learning to treat fluctuating windpower the same way he’d always treated fluctuating electricity demand (EWEA, “Wind Power Technology: Operation, Commercial Developments, Wind Projects, and Distribution,” ~2004, www.ewea.org/documents/factsheet_technology2.pdf, p. 10).
 Bolinger & Wiser, ref. 31, pp. 37–39.
 Europe has ≥38 GW of hydroelectric pumped storage, the U.S. ≥20 GW, with much more being built. The U.S. has demonstrated compressed-air storage in solution-mined salt caverns, and the economics look promising: S. Succar & R.H. Williams, “Compressed Air Energy Storage: Theory, Practice, and Applications for Wind Power,” Apr. 2008, www.princeton.edu/~cmi/research/Capture/Papers/SuccarWilliams_PEI_CAES_2.... Demand response (influencing when customers use electricity) provides cheap and abundant “virtual peakers” to firm variable renewables. The coming electrification of light vehicles will add large and lucrative opportunities for distributed storage (move.rmi.org/innovation-workshop-category/smart-garage.html). And in practically any utility system, the simplest method of integrating variable renewables is just to dispatch them when they’re available, ramping down costlier fueled plants. This requires no new technology—only running plants differently—and this is widely done in Europe. U.S. operators are already developing the tools: see, e.g., NERC, “Accommodating High
Levels of Variable Generation,” 16 Apr. 2009, www..nerc.com/files/IVGTF_Report_041609.pdf.
 The main exception is that since nuclear plants are best and safest run steadily, some regulators, e.g., in California, allow them to be dispatched instead of cheaper-to-run renewables, so the nuclear plants needn’t ramp down their output when renewables are abundant: their inflexibility makes it hard to ramp their output up and down rapidly or economically. Such favoritism sometimes causes available windpower to be “spilled” (lost). The resulting economic penalty improperly falls on wind, not on nuclear, operators, helping the latter to suppress fair competition without compensation. Some key Midwest utilities simply refuse to buy cheap and available windpower in order to protect their profits from old coal and nuclear plants; so far, state regulators have condoned this anticompetitive practice.
 T. Dinwoodie (SunPower Corp., Systems, Founder and CTO), “Price Cross-Over of Photovoltaics vs. Traditional Generation,” 2008. In 2008, the National Renewable Energy Laboratory expected 2010 U.S. PV power to cost 13–18¢/kWh residential, 9–12¢ commercial, and 10–15¢ for utility power; NREL’s targets for 2015, respectively 8–10¢, 6–8¢, and 5–7¢, now look likely to be achieved sooner. In contrast, NREL says the current market price ranges for retail grid power are about 6–17¢, 5–15¢, and 4–8¢ respectively. I calculate the delivered cost of power from a new nuclear plant at ~15–22¢ or higher (2007 $; see refs. 5–6). This comparison omits many hidden economic benefits of PVs and other distributed renewables that collectively increase their economic value by often about
tenfold: A.B. Lovins, Small Is Profitable, 2002, www.smallisprofitable.org.
 T. Dinwoodie (SunPower), personal communication, 1 Oct. 2009. Two-axis trackers produce more but cost more.
 See e.g. www.solarbuzz.com/Marketbuzz2009-intro.htm.
 C.L. Archer & M.Z. Jacobson, “Evaluation of global windpower,” www.stanford.edu/group/efmh/winds/global_winds.html. Class ≥3 sites (≥6.9 m/s), normally competitive with new coal power at zero carbon price, could yield ~72 TW at 80-m hub height. Contrary to the widespread impression that the best lower-49-states wind areas are only in the Great Plains, the East Coast, and certain West Coast sites, the data show that the Great Lakes wind resource, conveniently near upper Midwest load centers, is also Class 6±1. (It needs marine cables and engineering plus ice protection, but is much closer than Dakotas windpower.) The underlying data are in J. Geophys. Res. 110 (2005), D12110, doi:10.1029/2004JD005462, www.stanford.edu/group/efmh/winds/2004jd005462.pdf. The global windpower potential will become far larger even just on land if tethered high-altitude wind-turbine R&D projects succeed.
 M.Z. Jacobson, “Review of solutions to global warming, air pollution, and energy security,” En. & Envtl. Sci. 2:148–173 (2009), www.stanford.edu/group/efmh/jacobson/PDF%20files/ReviewSolGW09.pdf.
 World Energy Council, www.worldenergy.org/publications/survey_of_energy_resources_2007/solar/7....
Variation within the continental U.S. is smaller: Buffalo gets only one-fourth less and Arizona one-fourth more annual sunlight than Kansas City—less than regional differences in conventional energy prices (ref. 72). For detailed U.S. solar resource data, see http://rredc.nrel.gov/solar/pubs/redbook/.
 USDOE and Electric Power Research Institute, Renewable Energy Technology Characterizations, TR-109496, 1997, www.nrel.gov/docs/gen/fy98/24496.pdf, at p. 4-19. See also ref. 36.
 A cautionary note: land-use analyses assess land transformation (m2) -land altered from a reference state- or land occupation (m2-y) -the product of area occupied times duration of occupancy- for various energy outputs or capacities. The results can be hard to interpret if durations are long, effects are partly irreversible, or impacts are incommensurable. For example, the facilities and activities on a nuclear or coal system’s land are often more permanent and damaging than windpower or solar installations, which can readily be removed altogether. Most metrics used here are, or are converted to, occupancy (simple land areas) to reduce the risk of unit confusion.
 Ref. 47, p. 161. By international norms, the minimum buffer zone is 200 ha or 0.77 mi2: GEN IV International Forum, Cost Estimating Guidelines for Generation IV Nuclear Energy Systems, Ref. 3.03b, 29 Sep. 2006, http://nuclear,inl.gov/deliverables/docs/emwgguidelines_ref3.03b.pdf. We don’t count here the ~10-mile radius typical of the Emergency Planning Zone in which no public activities are permitted.
 J.G. Delene, K.A. Williams, & B.H. Shapiro, “Nuclear Energy Cost Data Base,” DOE/NE-0095 (1988), cited in ref. 57. H.C. Kim & V. Fthenakis, both of Brookhaven National Laboratory, give a similar figure of 52 m2/GWh or, for our nominal 1-GW plant, 6.3 mi2: “The Fuel Cycles of Electricity Generation: A Comparison of Land Use,” Mater. Res. Soc. Symp. Proc. Vol. 1041, 1041-R05-03 (2008). Their ref. 57 expands this analysis to include the full nuclear fuel cycle.
 D.V. Spitzley & G.A. Keoleian, “Life Cycle Environmental and Economic Assessment of Willow Biomass Electricity: A Comparison with Other Renewable and Non-Renewable Sources,” Rpt. #CSS04-05R, 2004, Center for Sustainable Systems, University of Michigan (Ann Arbor), cite at p. 57 some 2000 DOE data (www.eia.doe.gov/cneaf/nuclear/page/umtra/title1map.html) showing that 18 U.S. decommissioned uranium mines and mills disturbed an average of 0.025 ha/tU3O8 for 15 years. However, those 18 operations ran from the 1940s to 1970, and during 1948–70, the average U.S. ore milled contained 0.453% U3O8 (author’s analysis from USEIA, Uranium Industry Annual 1992, DOE/EIA-0478(92), http://tonto.eia.doe.gov/FTPROOT/nuclear/047892.pdf, p. 37). Through the mid-1980s, the modern ore grade reflecting most of the U.S. resource base averaged ~0.1% U3O8 (G.M. Mudd & M. Diesendorf, “Sustainability of Uranium Mining and Milling: Toward Quantifying Resources and Eco-Efficiency,” Environ. Sci. Technol. 42:2624–2630 (2008), Fig. 1). Assuming, probably conservatively, a constant stripping ratio over the decades, the historical land-use of ~0.025 ha/tU3O8 should therefore be adjusted to a modern U.S. value ~4.5x higher, or ~0.112 ha/tU3O8. According to www.wise-uranium.org/nfcm.html, a modern EPR-class reactor (4.0% enrichment, 45 GWd/t burnup, 0.9 capacity factor, 0.36 thermal efficiency) uses ~219 tU3O8/y on standard assumptions, or 8,769 tU3O8/40 y -hence a lifetime total of 986 ha, or 3.8 mi2, for the nominal 1-GW plant. (That figure would be comparable at Australian ore grades; higher at South African; and lower for Canadian, especially for two extraordinarily high-grade but short-lived deposits: see E.A. Schneider & W.C. Sailor, “Long-Term Uranium Supply Estimates,” Nucl. Technol. 162:379–387 (2008).) Ref. 57 is in excellent agreement at 3.66 mi2. As a cross-check of reasonableness, at a nominal 0.1% ore grade and 91.5% recovery, the modern 1-GW nuclear plant’s uranium consumption over 40 y will produce roughly 8.94 million short tons of mill tailings. The tailings piles at 26 uranium mills reported at p. 7 of EIA’s 1992 Uranium Industry Annual averaged 46,327 short ton tailings per acre (24 ft thick), committing 193 acres or 0.30 mi2 for the 1-GW plant’s tailings; at the modern 0.1% ore grade this would be ~1.35 mi2. Adding the mine area and waste rock disposal (a typical stripping ratio is ~5, and it swells when removed, so it can’t all go back in the excavated area) obtains reasonable agreement.
 The traditional U.S. method of enrichment (coal-fired gas diffusion, 0.3% tails assay) would use during the 1-GW plant’s 40-year life ~10 TWh to power separative work of ~4.3 million SWU. According to Spitzley & Keoleian, average U.S. pulverized-coal-fired electricity averages a land commitment of 580 ha-y/TWh, so we must add another ~5,800 ha-y or 22 mi2-y to power the enrichment -less with centrifugal enrichment or with less landintensive electricity sources. Such a reduced modern estimate, from ref. 57, is presented below.
 The Yucca Mountain high-level waste repository, according to D. Bodansky’s data cited by Spitzley & Keoleian (ref. 51), commits 6.2 km2 x (40 y x 23 t spent fuel/y / 70,000 t facility capacity); but those authors failed to notice that this counts only the facility’s direct footprint. Dr. Bodansky omitted its permanently withdrawn, DOE-controlled exclusion zone of ~600 km2 (232 mi2, 150,000 acres; see Final EIS, pp. 4-5 and 2-79), thus understating its
land-use by 97 x as ~0.08 rather than the correct ~7.7 km2 for the nominal 1-GW plant. (That plant’s lifetime spentfuel output of ~920 t represents 1.3% or 1.5% of Yucca Mountain’s 63,000 tHM or ~21 PWh of authorized capacity.) Kim & Fthenakis (ref. 50) derive 29 m2/GWh, or 3.5 mi2 for our nominal 1-GW plant.
 I have not found reliable data, other than old DOE data in Fig. 1, on the minor land-uses for uranium conversion, enrichment, or fuel fabrication facilities including exclusion zones, nor for any land commitment for cooling water.
 That is, (7 + 3.8 + 0.55 + 3) / 0.33 = 14.35, which is 43 x Cravens’s 0.33. As a cross-check, using slightly different global-average nuclear data, Jacobson (ref. 45) uses the Spitzley & Keoleian data to calculate a land commitment of ~20.5 km2/847 MW reactor at 85.9% capacity factor, or 25.4 km2 using our assumptions here but excluding enrichment fuel and the Yucca Mountain exclusion zone. That’s 9.8 mi2 (29 x Cravens’s number), or, adjusted to 0.1%U ore, 16.1 mi2 or 48 x Cravens’s claim. Another paper using the Spitzley & Keoleian data (R.I. McDonald et al., “Energy Sprawl or Energy Efficiency: Climate Policy Impacts on Natural Habitat for the United States of America,” PLoSONE, 2009, www.plosone.org/article/info:doi/10.1371/journal.pone.0006802#pone.00068...), expresses its nuclear land-use as 1.9–2.8 km2/TWh/y, or 5.8–8.5 mi2 for our nominal 1-GW plant, but shows no derivation, and I have not been able to reproduce its results from its stated sources.
 W. Häfele et al., Energy in a Finite World, International Institute for Applied Systems Analysis (Laxenburg), 1977, & Ballinger (Cambridge MA), 1981, Vol. 1, p. 286, found that the total area disturbed by the LWR system is ~0.7 mi2 for fixed facilities, plus ~0.5 mi2/y for the fuel cycle using 0.203%U ore, which would be ~1 mi2/y at the modern U.S. norm of 0.1%U ore. (I’ve adjusted the IIASA figures for the 14% lower uranium use per TWh in today’s EPRs and for 90% nuclear capacity factor.) This implies ~41 mi2 for the 1-GW nuclear plant over its 40-y lifetime, which is 2.9 times my conservative estimate or 123 x Cravens’s claim.
 V. Fthenakis & H.C. Kim, Renewable and Sustainable Energy Reviews 13:1465–1474 (2009), Fig. 1, assuming 50% underground and 50% openpit mining, 70% centrifuge and 30% gas-diffusion enrichment, and apparently counting all terms except disposal sites for low- and medium-level wastes, which neither they nor I can quantify from available data. Erroneously in my view, though, they count windpower area spread across, not occupied.
 Ausubel’s charming essay “Renewable and nuclear heresies,” Intl. J. Nuclear Governance, Economy & Ecology 1 (3):229 (2007), claims energy sources that use material amounts of land are not green because some Greens think human land-use shouldn’t increase. Its untransparent but clearly flawed analysis has been heavily criticized privately
and publicly, e.g. www.newscientist.com/blog/environment/2007/07/just-how-much-land-does-so....
 According to the European Wind Energy Association’s 2009 treatise The Economics of Wind Energy, ref. 36, p. 48. The American Wind Energy Association at
www.awea.org/faq/wwt_environment.html#How%20much%20land%20is%20needed%20... gives the older and more conservative figure “5% or less”, and notes that the land the turbines spread across can decrease by up to 30 x on a hilly ridgeline (from 60 to 2 nominal acres/peak MW), though some such sites may require maintained roads, taking back some of the turbine-spread land savings. In a 23 Sept. 2009 online Wall Street Journal letter, AWEA gives a 2–5% range and states that “for America to generate 20% of its electricity from wind, the amount of land actually used is about half the size of Anchorage, Alaska, or less than half the amount currently used for coal mining today.” DOE / EPRI’s 1997 data (ref. 47), reflecting early California practice before turbine layout was well understood, mentions 5–10%. J.G. McGowen & S.R. Connors’ thorough
“Windpower: A Turn of the Century Review,” Ann. Rev. En. Envt. 25: 147–197 (2000), at p. 166, give 3–5% for U.S. windfarms in the 1990s, but find 1% typical of U.K. and 1–3% of continental European practice, with “farm land… cultivated up to the base of the tower, and when access is needed for heavy equipment, temporary roads are placed over tilled soil.” I consider 1–2% typical of modern practice where land is valued enough to use attentively.
 Wind turbines on flat ground are typically spaced 5–10 diameters apart (e.g., in an array designed at 4 x 7 diameters) so they don’t unduly disturb each other’s windflow. (Spacing over water or on ridges is often much closer.) A typical modern wind turbine with its infrastructure has a nominal footprint of ~1/4 to 1/2 acre for roads, installation, and transformers (NREL, Power Technologies Energy Data Book, Wind Farm Area Calculator,
www.nrel.gov/analysis/power_databook/calc_wind.php) and has a peak capacity ~2–5 megawatts, hence an average capacity ~0.6–2 megawatts. That’s 0.2–2 mi2 of actual equipment and infrastructure footprint to match a 1-GW nuclear plant’s annual output. As a more rigorous cross-check, a nominal 1.5-MW, 77-m-diameter, 80-m-hub-height
turbine in a Class ≥3 wind site would nominally be sited 6 turbines per km2 (ref. 45, p. 17), so 667 of them would match the peak output and (at 35% wind vs. 90% nuclear capacity factor) 1,715 would match the annual output of a 1-GW nuclear plant. Including roads, 1,715 turbines would physically occupy a nominal 1–2% (EWEA, ref. 59) of the area they spread across, which is 1,715/6 = 286 km2 or 110 mi2. That 1–2% occupied area is thus 2.9–5.7 km2 or 1–2 mi2. Even in probably the highest official land-use estimate, which generously assumes about a thousand times the minimal physical footprint, the Bush Administration’s 20% Wind Energy by 2030, at pp. 110–111, found that 305 GW of U.S. windpower could disturb ~1,000–2,500 km2 of land, or 1.3–3.2 mi2/installed GW, or at 35% capacity factor, 3.3–8.1 mi2/1-GW-reactor-equivalent -still 37–90 times lower than Ausubel’s claim of 298 mi2.
 Ref. 45.
 With each 5-MW turbine at 35% capacity factor producing 1.75 average MW, 514 turbines would produce 900 average MW to match the 1-GW nuclear plant. Each turbine has a direct footprint (foundation and tower) of ~20 m2, so 514 turbines directly occupy ~20 x 514 = 10,280 m2 or ~0.004 mi2. We round up to 0.005 to allow for transformers; the cables are always underground. This footprint is normal for flat open sites not needing permanent roads.
 In an average U.S. site, PVs spreading across 15 mi2, but not actually using much or most of it, would produce the same annual grid electricity as a 1-GW nuclear plant from flat horizontal solar cells like the 19.3%-efficient Model 315 in SunPower’s current catalog (that firm’s prototypes in May 2008 also achieved 23.4%, heading for market ~2010). The math is simple. The U.S. receives annual-average, 24/7/365 sunlight of 1,800 kWh/m2y (one-fifth of
full equatorial sea-level noon irradiance), so a 19.3%-efficient module captures an average of 347 kWh/m2y or 40 average WDC/m2. AC output is nominally ~23% lower due to practical losses (dirt, fill fraction, wiring and conversion losses, mismatch, system availability, heat: http://rredc.nrel.gov/solar/codes_algs/PVWATTS/system.html), yielding 31 average WAC/m2. Now derate generously by another 25%, to 23.1 average WAC/m2, to allow ample access space for maintenance (possibly shared with other uses). Thus horizontal flat PVs spread across 3/4 of 900,000,000/23.1 = 39 million m2 or 15 mi2 will produce 900 average MWAC in an average U.S. site. Tracking collectors could reduce the module area by ~25–36%, or southwestern Nevada siting by ~22%, or both; simply tilting up the panels at the local latitude saves ~16%, but some space is still needed between the panels for access, so for simplicity and conservatism I’ve used the horizontal model in this illustration. NREL (ref. 66) found that the most efficient packing of tilted 15%-efficient PV modules can spread across 10 km2/GWp, or 17.4 mi2 to match the annual output of our nominal 1-GW nuclear plant; at our 19.3% efficiency that would be 13.5 mi2. In excellent agreement, CTO Tom Dinwoodie (personal communication, 2 Oct. 2009) confirms that in a typical U.S. site,
SunPower’s land-efficient one-axis/backtracking T0 tracker typically yields 0.3 capacity factor at 0.4 ground cover ratio (the ratio of panel area to total land area), so a nuclear-matching PV farm at 20% module efficiency and 80% DC/AC efficiency would spread across 17.8 mi2 (or 5.9 if it matched the nuclear plant in capacity rather than in energy). Also consistent with these figures, J.A. Turner (NREL), Science 285: 687–689 (30 July 1999), showed that 10%-efficient PVs occupying half of a 100 x100-mile square in Nevada could produce all 1997 annual U.S. electricity. But the phrase “occupying half of” is conservative: PVs normally get mounted not on the ground but well above it, leaving the space between ground mounts available for other uses such as grazing. (The moving shade can reportedly benefit both grass and sheep.) Mounting poles punched into the ground can make actual land-use a very small fraction of the total site areas calculated here, and livestock graze right up to the poles. Two-axis trackers, though typically less cost-effective than one-axis, have an even smaller footprint because they’re PVs-on-a-pole, analogous
to wind turbines. For comparison, concentrating solar thermal power systems spread across roughly one-third more area than PVs for the same annual (but firm) output, and require cooling, though this can use dry towers. Other revealing land-use comparisons are at www.sourcewatch.org/index.php?title=Concentrating_solar_power_land_use.
 Ref. 45, which conservatively projects that 30% of long-term PV capacity will be roof-mounted.
 According to Lawrence Berkeley National Lab’s world-class roof expert Dr. Hashem Akbari (www.climatechange.ca.gov/events/2008_conference/presentations/2008-09-09...), the world’s dense cities occupy 1% of the earth’s land area, or ~1.5 trillion m2. About one-fourth of that, or 0.38 million km2, is roofs. So ignoring all parking structures, and all smaller cities’ or non-urban roofs, and assuming that just one-fourth of the big-city roof area has suitable orientation, pitch, shading, and freedom from obstructions, PVs just on the world’s urban roofs could produce ~106 PWh/y, or 5.8 x global 2005 electricity use. (This assumes the same 75% module derating factor as before, and global-average horizontal surface irradiance of 170 W/m2 (WEC, ref. 46, but most big cities are at relatively low latitudes with more sun.) Large land areas now occupied by old landfills, or overwater, could also be covered with PVs without displacing any useful activity.
 NREL, “PV FAQs: How much land will PV need to supply our electricity?,” DOE/GO-102004-1835 (2004), www.nrel.gov/docs/fy04osti/35097.pdf, italics in original.
 Vestas, “Life cycle assessment of offshore and onshore sited wind power plants based on Vestas V90-3.0 MW turbines,” Vestas Wind Systems A/S, 2006,
www.vestas.com/Files/Filer/EN/Sustainability/LCA/LCAV90_juni_2006.pdf, assuming 105-m hub height onshore. See also www.vestas.com/en/about-vestas/principles/sustainability/wind-turbines-a...(
 See e.g., Ref. 57’s citations 27, 34, and 35.
 E.g., Kim & Fthenakis, ref. 50, Fig. 3. Ref. 63 states that using U.S. average solar irradiance (1800 kWh/m2y) and a 30-y assumed life, the indirect land-use for PV balance-of-system is 7.5 m2/GWh, plus for the installed PV array itself, 18.4, 18, and 15 m2/GWh for multi-, mono-, and ribbon-Si. Scaled to 900 average MW for 40 y, these would correspond respectively to 0.9, 2.2, 2.2, and 1.8 mi2. For comparison, that paper calculates 30–60-y direct land-use as 164–463 m2/GWh with optimal tilt but ~10% efficiency. These direct land-uses correspond to 20–56 mi2/900 average MW -higher than my ~10 because the paper assumes half my empirical array efficiency and uses layouts with severalfold less dense packing (id.; Ref. 47, p. 4-30). Their analysis confirms that PVs produce about two-fifths more electricity per unit of land (over 30 y at 13% efficiency and average U.S. irradiance) than typical U.S. coalfired power plants do.
 Many durable trends, not counted in ref. 1–5’s “snapshot” analyses of current and recent market costs, all favor efficiency and renewables. These include: side-benefits of efficiency often worth 1–2 orders of magnitude more than the saved energy; distributed benefits (Small Is Profitable, ref. 45) often worth about an order of magnitude in value; technical and economic synergies of efficiency/renewables and renewables/renewables integration; generally decreasing cost and construction time for efficiency and micropower (but increasing for central plants); generally rising fuel-price volatility and supply risk; increasing climatic and environmental costs and consequences of central plants; financial risk aversion; greater competition in power generation; and more transparent decisionmaking.
 As of 1 August 2009, there were 52 reactors under construction -compared with 120 at the end of 1987 or with 233 at the ordering peak in 1979. But, from those 52:
* 13 have been “under construction” for over 20 years
* 24 have no officially planned start date
* half are late, often substantially
* 36 (over two-thirds) are in just four countries -China, India, Russia, and South Korea- none of which use competitive markets to choose whether or which power plants are built, and none of which is very transparent about construction status or decision process. (see M. Schneider et al., The World Nuclear Industry Status Report 2009, German Federal Ministry of Environment, Nature Conservation and Reactor Safety, 27 Aug. 2009), www.bmu.de/english/nuclear_safety/downloads/doc/44832.php
 See M. Schneider, “Nuclear Power in France: Beyond the Myth” (Dec. 2008, firstname.lastname@example.org), and “What France got wrong,” Nucl. .Eng. Intl., Aug. 2009, p. 42. Financial stress is evident from past bailouts of parts of the nuclear complex and from Areva’s overextension today. The nuclear system is so overbuilt, and so reliant on very peaky electric space-heating loads, that by February 2009 the gap between minimum and maximum daily loads was 61 GW, requiring >40 reactors to load-follow.
 A. Grubler, “An assessment of the costs of the French nuclear PWR program 1970–2000,” Interim Report IR-09-036, International Institute for Applied Systems Analysis (Laxenburg, Austria), 6 Oct. 2009, www.iiasa.ac.at/Admin/PUB/Documents/IR-09-036.pdf.
 Even on favorable assumptions, nuclear officially fell off its always-cheaper-than-gas-combined-cycle throne as early as 1997 (ref. 73, p. 14). It still lacks any honest official comparison with micropower and efficiency.
 K. Bradsher, www.nytimes.com/2009/07/03/business/energy-environment/03renew.html, “Green Power Takes Root in the Chinese Desert,” N.Y. Times, 2 July 2009,. The Global Wind Energy Council’s 2009 Outlook foresees 352–1,193 GW of global windpower in 2020 producing 864–2,600 TWh/y; the latter equals nuclear’s output today.
 RMI periodically updates its documented database of global micropower data from industrial and governmental sources at www.rmi.org/sitepages/pid256.php#E05-04. This graph is from the current update-in-progress. A comparable independent database of distributed renewables, not including cogeneration, is at www.ren21.net.
 E. Chen & L. Hornby, “China official warns on ‘too fast’ nuclear plans,” 27 Sep. 2009,
 M.B. McElroy, X. Lu, C.P. Nielsen, & Y. Wang, “Potential for Wind-Generated Electricity in China,” Science 325: 1378–1380 (11 Sep. 2009), www.sciencemag.org/cgi/content/abstract/325/5946/1378. The turbines analyzed are smaller (1.5 MW), shorter (80 m), and less efficient and well sited (~20% average capacity factor) than modern Western ones, leaving considerable room for improvement without sacrificing China’s speed and cost advantages.
 S.W. Hadley & W. Short, “Electricity sector analysis in the clean energy futures study,” En. Pol. 29(14): 1285–1298 (Nov. 2001).