REVIEW ARTICLE

For comparison, the 2100 values for the IPCC90 scenarios are also shown in Fig. 2.

The most important points to, note are the substantially reduced forcing in each best-guess case relative to the corre sponding base case; the wide range of forcing possibilities, all corresponding to the same basic 'existing policies' assumption; and the reduced range of uncertainty in the best guess compared with the base case, largely because the high forcing cases have their larger CO2 effects offset more by aerosol forcing. The reduced forcing compared with what would have been obtained using the IPCC90 methodology arises primarily from our inclusion of CO2 fertilization feedback and sulphate aerosol forcing. As can be seen from Table 1, the aerosol effect depends on the scenario: IS92b is similar to IS92a, scenarios e and f have larger negative aerosol forcing contributions, whereas c and d have small positive contributions. For the best-guess case, total forcing under the c and d scenarios is below the (policydriven) IPCC90 B scenario. None of the IS92 scenarios, however, approaches the very low forcing projections of IPCC90 scenarios C and D.

2000

+A(~)

2020 2040 2060 2080 2100
Year

FIG. 4 Radiative forcing, global-mean temperature and global-mean sea level projections for emissions scenario IS92a. For forcing, the base case follows IPCC90 methodology (curve 1). Curve II includes CO2 fertilization feedback in calculating concentration trends. Curve III, slightly below II, additionally accounts for ozone-depletion feedback. L, M and H refer to low, middle and high aerosol forcing possibilities. Temperature and sea-level results are shown only for the base (dashed curves) and best-guess case forcings (full curves). Low, middle and high estimates are given (denoted L, M, H). For temperature, these are obtained using AT2, -1.5, 2.5 and 4.5 °C with other model parameters kept at their best-guess values. For sea level, corresponding thermal expansion contributions are combined with ice-melt values based on low, middle and high model parameter values. For comparison, the equivalent 2100 low, middle and high estimates of temperature and sea level change for the IPCC90 Business as Usual scenario (SA90) are shown in the right (denoted A(L), A(M) and A(H)), with SA90 forcing shown by A in the top panel.

298

Temperature changes to 1990

Temperature projections for the six IS92 scenarios were made using our upwelling-diffusion energy-balance climate model3, as used in IPCC90, Before describing the projections beyond 1990, however, we consider changes to 1990, as these affect the range of possible sulphate aerosol forcings and have implications for the climate sensitivity.

Climate models of the type used here incorporate very simplified representations of the physical processes involved. Nevertheless, they seem to simulate well the responses to external forcing produced by more complex models. They account for the lag effect of oceanic thermal inertia, and they allow one to assess the sensitivity of the results to model uncertainties, in particular to uncertainties in the climate sensitivity. The present model additionally differentiates between the hemispheres and so permits the consideration of forcing changes specific to one hemisphere, as is necessary if we are to account for the effects of sulphate aerosol forcing.

The most important model parameter is the climate sensitivity which is specified by the equilibrium global-mean temperature change for a CO, doubling. Largely on the basis of general circulation model experiments, this temperature change, AT2、, is thought to lie in the range 1.5-4.5 °C, with a 'best-guess' value of 2.5 °C (ref. 1). The other parameters affecting the model's response are the mixed-layer depth (h), the oceanic vertical diffusivity (K), the upwelling rate (w) and the temperature change of high-latitude sinking water relative to the global-mean change (#). Sensitivity to these parameters is less than to AT2x (ref. 35). We assume h=90 m, K = 1 cm2 s', w = 4 m yr1 and

0.2. Slightly different values were used in IPCC90. The set of values used here is justified more fully in ref. 36. The greatest difference from IPCC is in our choice of # (IPCC901 used π = 1). Our choice of a lower value is supported by Schlesinger and Jiang. Smaller produces larger surface warming and smaller oceanic thermal expansion. The net effect on sea level change is a small reduction.

Figure 3 shows the radiative forcing to 1990 and the corresponding modelled temperature changes, both for the global mean and for the Northern Hemisphere. Modelled temperature changes are shown only for AT2, 2.5 °C. For comparison, observed temperature changes, as recently updated by IPCC, are also shown.

If ozone-depletion feedback and best-guess sulphate aerosol effects are included, then the total forcing change to 1990 is reduced markedly (by ~1 W m2 or 40%) from the IPCC90 estimate". This brings the predictions of temperature change using AT2x=2.5 °C into closer agreement with observations. In the absence of these two factors, the modelled changes with AT2 = 2.5 °C are appreciably larger than the observed changes; the value of AT2x giving the best fit to the 1880-1990 warming is only 1.5°C (compared with 1.4 °C in ref. 36, which used slightly different temperature data). For the case with ozonedepletion feedback and best-guess aerosol forcing, the modelled changes using AT2 = 2.5 °C are noticeably less than the observed changes. The required sensitivity to obtain the best fit to the observations is AT2, 3.4 °C. If only best-guess aerosol forcing is included (without ozone-depletion feedback), our best-guess empirical value for the sensitivity is 3.0 °C (compare with ref. 40 for similar assumed forcing). The inclusion of the aerosol forcing flattens the model-projected warming trend over 1950-70, especially in the Northern Hemisphere, in qualitative agreement with observations.

Figure 3 also indicates the possible range of values for the negative aerosol forcing. Our largest negative values, which are at the smaller end of the range given by Charlson et al.10, lead to an overall negative forcing trend over 1860-1980 in the Northern Hemisphere. As a consequence, the modelled temperature change for the Northern Hemisphere over this period is much less than observed, and, more important, there is a marked difference in warming between the Northern and

NATURE VOL 357 28 MAY 1992

Southern hemispheres which is not evident in the observations. It is possible that differential warming has been masked by natural variability, but this seems unlikely. On empirical grounds, therefore, and in accord with refs 29 and 40 and three-dimensional modelling results", our choice for the range of likely aerosol forcing is more defensible than the values suggested in ref. 10 which were based on a simpler onedimensional analysis.

Future temperature and sea level

Projections of temperature change are obtained by running the climate model with the various past forcing possibilities com. bined with consistent projected forcings to 2100. The number of model runs to be made is determined by the product of the number of future emissions scenarios (6), the number of forcing projections for each scenario (12, see below) and the number of sets of model parameters (3, see below). For each scenario, we have considered forcing projections with and without CO2 fertilization feedback, with and without ozone-depletion feedback, and for low, middle and high aerosol forcing. This set of 12 possibilities gives an idea of the uncertainties in future forcing under any particular emissions scenario, although it is not a comprehensive analysis of uncertainty. For the climate model, we have carried out simulations for different climate sensitivity values, AT2 = 1.5, 2.5 and 4.5 °C. As noted above, the sensitivity of the results to other model parameters is less than to AT2. (see ref. 35) so these have been kept at their best-guess values. Only a selection of results is presented here.

The climate model also produces results for global-mean thermal expansion of the oceans, a principal component of future sea-level rise. To obtain total sea-level rise, meltwater contributions are added from separate, highly aggregated models representing small glaciers, the Greenland ice sheet and Antarctica, each driven by the global-mean temperature projection. The ice-melt models are similar to those used by IPCC90 (which we developed in conjunction with Warrick and Oerlemans"), with improvements as described in refs 35 and 42. Ice-melt uncertainties are accounted for by using a range of values for ice-melt model parameters 35.42.

Detailed results for IS92a are given in Fig. 4. This is the scenario most nearly equivalent to SA90. The upper panel shows the sequential reductions in forcing that arise from including a CO2 fertilization effect (I to II), accounting for ozonedepletion feedback (II to III) and accounting for the negative forcing due to sulphate aerosols (III to L, M or H). Including CO, fertilization to balance the contemporary carbon budget reduces the 1990-2100 forcing change by 11% (compared with

REVIEW ARTICLE

a maximum of 22% for IS92c and a minimum of 7% for IS92e). The effect of including stratospheric-ozone-depletion feedback on post-1990 forcing is negligible (in all scenarios), whereas aerosols reduce the IS92a forcing by a further 6-19% relative to the base case. Overall, the best-guess forcing for IS92a is -19% below the base case in the year 2100 (see Table 1 for results for other scenarios).

Low, middle and high temperature and sea-level projections for IS92a are shown in the lower panels of Fig. 4. For temperature, the warming over 1990-2100 is between 1.7 °C and 3.8 °C for the best-guess forcing case (compared with a warming range of 2.1 °C to 5.0 °C for base-case forcing). The corresponding sea-level rise lies between 15 and 90 cm for best-guess forcing and between 22 and 115 cm for base-case forcing. Temperature and sea-level change projections based on best-guess forcing are 20-24% and 21-30%, respectively, below those for base-case forcing.

The most direct comparison that can be made with IPCC90 is between IS92a and the SA90 results. The latter results are shown (for the year 2100 only) on the right-hand side of Fig. 4. IS92a and SA90 are comparable only in a limited sense, and differences arise because of differing emissions projections, revised methods for calculating the implied concentration and radiative forcing changes, changes in the assumed best-guess climate model parameters, and improvements in the ice-melt models. For temperature changes over 1990-2100, the IPCC90 best guess for SA90 (based on AT,, = 2.5 °C) was 3.3 °C. Our corresponding best guess for IS92a is 2.5 °C. For sea level, the equivalent projected changes are: SA90, 66 cm; IS92a, 48 cm. These are substantial reductions, but the projected changes in global-mean temperature and sea level are still very large. For comparison, the warming corresponds to a rate roughly five times that observed over the past century, and the sea level rise is at a rate roughly four times that estimated for the past century. The uncertainty ranges for these projections are large, and they are higher in the present results than in IPCC90.

Figure 5 summarizes temperature and sea-level projections for all scenarios for the overall best-guess cases (AT2x = 2.5 °C, best-guess ice-melt model parameters, and best-guess forcing). IPCC90 best-guess cases are also shown (2100 values). For the IS92 scenarios, the differences in temperature and sea level projections up to 2050 are fairly small, less than the forcing differences between scenarios (see Fig. 2). For the next 60 years or so, future temperature and sea-level changes are, therefore, relatively insensitive to the uncertainties in future emissions. This means that the 1592a results given in Fig. 4 can be taken as representative of all the emissions scenarios until about 2050.

REVIEW ARTICLE

These projections are subject to other large uncertainties, particularly uncertainties in the climate sensitivity (Fig. 4). Reducing the error bounds on the climate sensitivity should therefore have high priority in greenhouse research.

Conclusions

These concentration, radiative forcing, temperature and sealevel projections are meant to complement the scientific update2 of the 1990 IPCC report'. We have addressed the new scientific issues raised in this update, the effects of CO, fertilization on CO, concentration projections, the possible negative feedback on halocarbon radiative forcing due to stratospheric ozone depletion, and the negative forcing due to sulphate aerosols. Largely because of these factors (specifically, the first and third), future projections of global-mean temperature and sea level change are reduced markedly (by 20-30%) below the equivalent projections made using the IPCC90 methodology (compare, for example, base-case and best-guess estimates in Fig. 4). But our projections, by including the effects of sulphate aerosols, introduce an additional complication: a reversed and enhanced differential in the rates of warming between the hemispheres. This could have an important influence on the climate system beyond the direct greenhouse effects currently considered by general circulation climate models.

The following important results have emerged. First, the effects so far of ozone-depletion feedback and sulphate aerosol forcing bring the best-guess empirical estimate of the climate sensitivity (AT2x = 3.4 °C) into closer agreement with the range of estimates based on general circulation climate model results. If both of these factors are ignored, the best-guess empirical sensitivity is only 1.5 °C. The improved agreement here is encouraging, but it should not be over-interpreted. If other uncertainties are taken into account, then one finds a very wide range of AT2x values to be compatible with observed temperature changes.

This point reinforces our second result, namely that uncertainty in future greenhouse-gas-related temperature and sealevel changes is strongly affected by the climate sensitivity. Reducing uncertainties in this parameter should have high

priority, although the best strategy for this is far from clear. The main reason for our conclusion here is that projections of temperature and sea-level changes out to 2050 have only a small range for any given AT2, value, but differ widely according to the chosen value of AT. The similarity between scenarios should not be interpreted as implying a general insensitivity to future emissions trajectories. Rather, it emphasizes the very long-term nature of the greenhouse problem. A more salutory result is the fact that, beyond the middle of next century, projections for the various scenarios diverge rapidly. Inertia in population growth and socio-economic systems would make it difficult to shift from one scenario to another.

Climate sensitivity is not the only uncertainty that needs to be reduced. In the scenarios in which SO, emissions increase markedly (a, b, e and f), an important additional uncertainty arises through not being able to quantify the effect of aerosols to better than around 50%. In scenario IS92a, for example, this translates to an uncertainty of ±6% in total forcing, +8% in temperature change and ±9% in sea-level change (over 19902100). Uncertainties arising from our inadequate understanding of the global carbon cycle, judged from a comparison of results with and without CO, fertilization feedback, are of comparable magnitude.

The projections presented here show only the anthropogenic component of future change, and natural variability will be superimposed on this. Natural variability includes both internally generated variability and the effects of external forcing agents", the latter including short-term factors like the recent volcanic eruption of Mount Pinatubo. The expected anthropogenic rates of change, however, are much greater than the corresponding natural rates of change over intervals of 30 years or longer, becoming increasingly so as the interval increases. The reduced rates of warming and sea-level rise that we have obtained compared with IPCC90 are still four to five times those that have occurred over the past century. Such changes are certain to present a considerable challenge for humanity.

ט

T. M. L. Wigley and S. C. B. Raper are at the Climatic Research Unit, School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, UK.

1. Houghton, J. T... Jenkins, G. J. & Ephraums. J. J. (eds) Climate Change The IPCC Scientific Assessment (Cambridge Univ. Press, 1990).

2. Houghton, J. T. Callander, B. A. & Varney, S. K. (eds) Climate Change 1992 The Supplementary Report to the IPCC Scientific Assessment (Cambridge Univ. Press, 1992).

3. Pepper, W. J. et al. Emission Scenarios for the IPCC-an update Background Documentation on Assumptions. Methodology and Results US Environmental Protection Agency. Washington DC 1992).

4. Leggett, J. A., Pepper, W. J. & Swart, R. J. in Climate Change 1997 The Supplementary Report to the IPCC Scientific Assessment (eds Houghton, 1 T. Callendar, B. A. & Varney. SK) 69-95 (Cambridge Univ Press, 1992)

5. Mitchell, J. F. B. & Gregory. J. M. in Climate Change 1992 The Supplementary Report to the IPCC Scientific Assessment (eds Houghton, J. T. Callendar, B. A. & Varney SK )171 175 (Cambridge Univ. Press, 1992).

6. Ramaswamy. V.. Schwarzkopf, M. D. & Shine, K. P. Nature 355, M10 R12 (1997)

7. Isaksen, I., Ramaswamy. V. Rode, H & Wigley, TML in Climate Change 1997. The Supplementary Report to the IPCC Scientific Assessment (eds Houghton, J. 1. Callendar, 8 A & Varney, SK) 47-67 (Cambridge Univ Press. 1992)

8. Charlson, R. J. Langner. J. Roche. H. Leovy. C. B & Warren, S G Tellus AB 43, 152-163 (1991) 9 Wigley. T M. L. Nature 349, 503-506 (1991)

10. Charlson, R. J. et al. Science 255, 423-430 (19921

11 Nordhaus, W. D. & Yone, G. W in Changing Climate 87-153 (National Academy Press. Washington DC, 1983).

12. Edmonds, J. A., Reilly, J. M., Gardner, R. H. & Breckert, A. Uncertainty in Future Global Energy Use and Fossil Fuel CO, Emissions 1975 to 2075, TRO36 (US Dept of Energy. Carbon Dioxide Research Division, Washington DC, 1986).

13. Wigley, TM L. Holt, T. & Raper, S C B STUGE (an Interactive Greenhouse Model) User's manual (Climatic Research Unit, Norwich, 1991).

14. Shine, K. P. Derwent, R. G. Wuebbles. D. J. & Morcrette. J-J. in Climate Change. The IPCC Scientific Assessment (eds Houghton, J.T. Jenkins, G. J. & Ephraums. J. J) 41-68 (Cambridge Univ Press. 1990)

15. Harvey. L. D. D Clim Change 15, 343-381 (1989).

16. Wigley, T. M. L. Global biogeochem Cycles 5, 373-382 (1991)

17. Maier Reimer. E & Hasselmanın, K. Clim Dynam 2, 63-90 (1987)

18 Wigley. T ML. Tellus (in the press).

19. Watson, R. T. Roche. H.. Oeschger, H. & Siegenthaler, U. in Climate Change The IPCC Scientific Assessment leds Houghton, J. T., Jenkins, G. J. & Ephraums. J. J) 1-40 (Cambridge Univ Press. 1990)

20. Watson, R. T. Meira Filho. L. G. Sanhueza. E & Janetos. A in Climate Change 1992 The Supplementary Report to the IPCC Scientific Assessment leds Houghton, J. 1. Callendar, 8 A & Varney. S K) 25-46 (Cambridge Univ Press 1997)

21. Enting 1. & Mansbridge Tellus B43, 156-170 (1991).

22. Born, M, Door, H. & Levin, I. Tellus 842, 1-8 (1990)

300

23. Osborn, T. J. & Wigley. T. M. L. Clim Dynam (submitted).

24 Watson, R. T. & Albritton, D L. Scientific Assessment of Ozone Depletion: 1991 (World Meterological Organization, Geneva, in the press).

25 Vaghjani, G & Ravishankara. A Nature 350, 406-409 (1991).

26 Leleveld, J & Crutzen, P. Nature 355, 339-342 (1992).

27 Lal M & Holl. T Clim Res 1, 85-95 (1991).

28 Ramanathan, V., Cicerone, R. J. Singh. H. B & Kichl, J. T 1 geophys. Res. 90, 5547-55GG (1985)

29 Wigley, T. M. L Nature 339, 356-367 (1989)

30 Charlson, R. J. Langner, J & Roche. H. Nature 348, 22 (1990).

31 Twomey S. A. Peipgrass, M & Wolfe. TL Tellus B 36, 356-366 (1984).

32 Charlson R. J. Lovelock, J. E. Andraca. M. O & Warren, S. G. Nature 326, 655-661 (1987) 33 Atrecht B. A. Scirnice 245, 1777 1230 (1989)

34 Wigley, 1. ML & Raper. 5 CB Nature 330, 177-131 (1987).

35 Wigley. T. ML & Raper, SCB in Climate and Sea Level Change Observations, Projections and Implications (eds Warrick, R. A. Barrow, E M. & Wigley, T. M. L.) (Cambridge Univ. Press, in the press).

36. Wigley T M. L. & Raper, S C B in Climate Change Science. Impacts and Policy (eds Jager. J & Ferguson, M. L) 231-242 (Cambridge Univ Press, 1991)

37. Bretherton, F. P... Bryan, K. & Woods, J. D. in Climate Change The IPCC Scientific Assessment leds Houghton, J. T. Jenkins, G J & Ephraums. J. J) 173-193 (Cambridge Univ. Press, 1990). 38 Schlesinger, M. E & hang, X Nature 350, 219-221 (1991).

39. Folland. C K et al in Climate Change 1992 The Supplementary Report to the IPCC Scientific Assessment (eds Houghton, J. T., Callendar, B. A. & Varney, S. K.) 135-170 (Cambridge Univ. Press, 1992).

40. Schlesinger, M. E., Jiang X & Charlson, R. J. in Proc Conf. on Global Change: Its Mitigation through Improved Production and Use of Energy (eds Allred, J. C. & Nichols, A. S.) (American Institute of Physics, in the press).

41. Warrick RA & Oerlemans. H. in Climate Change, The IPCC Scientific Assessment (eds Houghton, JT. Jenkins. G J & Ephraums, JJ) 257-281 (Cambridge Univ Press, 1990)

42. Raper, S. C B. Wigley, T M. L. & Warrick, R. A in Proc SCOPE Workshop on Rising Sea Level and Subsiding Coastal Areas (ed Milliman, J. D) (Wiley, Chichester, in the press)

43 Hansen, JE, Rossow, W & Fung. I. Issues Sci. Technol 7, 62-69 (1990).

44 Wipley, T. ML & Raper. S C B Nature 344, 324 327 (1990)

45 Hansen, JE & LOCIS, A A Nature 346, 713-719 (1990).

46 Moller, D Atmos Envir 18, 19-27 (1984)

47. Smith, F. 8. Q. JR met Soc 117, 657-683 (1991).

ACKNOWLEDGEMENTS. This work was supported by grants from the US Department of Energy. Atmospheric and Climate Research Divison, the Commission of the European Communities (DGXH). and the UK Department of the Environment Comments by KP Shine and P M Kelly helped to improve the manuscript Computer-generated diagrams were produced by M. Salmon.

NATURE VOL 357 28 MAY 1992

TAKE A GOOD LOOK AT THE GRAPH ON THIS page, reproduced from a report that appears on page 698. It's giving climatologists goose bumps. The curves show that when the interval between peaks in sunspot abundance began shortening at the end of the past century, the Northern Hemisphere began to warm. When the sunspot cycle stopped shortening and began lengthening around 1940, the temperature peaked and began falling. And when the solar cycle length started shortening again in the 1960s, temperature turned around too. The close association of these two curves is the most striking correlation ever found between climate and small variations in solar activity-and the strongest suggestion ever of a causal link.

"If it's correct," says atmospheric scientist Keith Shine of the University of Reading, "we have to change our view of climate fundamentally. It's an incredible correlation; it would imply that almost nothing else (beside solar variation] is important in the climate system." Even greenhouse warming would have played little role in the 0.5°C warming of the last century-which is not to say that it couldn't be important soon. Temperature

researchers will assume just that, given the long, checkered history of the search for sun-climate relations.

"We have seen a number of these that haven't worked out," says John Eddy of the University Corporation for Atmospheric Research in Boulder, Colorado, who has studied possible sun-climate connections on time scales of centuries. "We've been fooled so many times that we should be careful." One memorable example he cites is a stack of 680-million-year-old sediments that to believers and even some skeptics seemed to record climate variations on a timetable uncannily similar to various sunspot cycles (Science, 3 September 1982, p. 917). But then the bubble burst: It turned out that the layers had recorded nothing more than the cycles of the tides (Science, 18 November 1988, p. 1012).

In searching for their new sun-weather link, Eigil Friis-Christensen and Knud Lassen of the Danish Meteorological Institute faced the same handicaps as their predecessors: lack of good records of either climate or the activity of the sun. The researchers used the best available record of global temperature, but it covers only the land surface of the North

0.3

652

ern Hemisphere and goes back only to about 1870. The risk in using such a short record is the greater likelihood that, just by chance, it will seem to vary in concert with some change on the sun.

The two researchers also lacked any direct indicator of the sun's total en

crgy output, the most likely driving force of any sun-rc

lated climate change. Changes in solar irradiance, after all, have been accurately monitored for only the last 10 years. So, like others in search of a sun-climate link, FriisChristensen and Lassen had to use a stand

in. Previous workers have tried to gauge the sun's activity by such measures as the height of the peaks in the sunspot cycle, the fre quency of auroras, and the variation in the amount of carbon-14 produced in the upper atmosphere and trapped in tree rings. FriisChristensen and Lassen opted for the vary

ing length of the sunspot cycle simply because it worked out and because it seems to track long-term variations in the solar wind-a direct, though feeble, part of the sun's energy output.

But it's far from clear that sunspot-cycle length actually changes in concert with any solar variation strong enough to drive climate change. And the fact that it is only the latest in a long line of solar-output proxies that have been tried doesn't inspire confidence. After so many rolls of the dice, some climatologists complain, a lucky nuntber might finally have come up. "The more parameters you look at," notes climatologist David Parker of the Hadley Center for Cli mate Prediction and Research in Bracknell, England, "the more likely you are to find one that fits the global temperature curve [by chance."

If such skepticism weren't enough, FriisChristensen and Lassen have another problem to overcome in gaining acceptance for their correlation: Climate researchers can't explain how known solar variability might affect climate in the first place. The variations in solar output traced in recent years are too small to account for the temperature changes of the past century, and so far solar physicists haven't come up with any mechanism for larger swings in output. Some researchers have talked about the possibility that the upper atmosphere somehow amplifies the effects of the known changes in irradiance or other solar variations, but the search for such an atmospheric amplifier has been going on for decades without success.

After voicing all their doubts about the report, though, some researchers admit that they are mightily impressed by the close intertwining of the two curves. They have a correlation coefficient of 0.95, probably the highest ever found in this sort of work. Even Parker says he is impressed, though cautious. And Eddy goes further: "The fit is so good," he says, that "the burden of proof that something's wrong almost rests with any detractors." The usual qualifications must still apply, he says, but, "I'll still make a bet they're on to something."

Even so, it could be a long time before anyone proves it. Decades more of irradiance measurements would be required to show that the sun's output can vary enough on long time scales to have caused the halfdegree warming of the past century. Until then, researchers are stuck in limbo. Metco

SCIENCE, VOL. 254