Changing Climate: Report of the Carbon Dioxide Assessment Committee (1983)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Detection and Monitoring of 5 C02-Induced Climate Changes Gunter Weller, D. James Baker, Jr., W. Lawrence Gates, Michael C. MacCracken, Syukuro Manabe, and Thomas H. \fonder Haar 5.l SUMMARY Two questions are addressed in this chapter: (l) have we already detected a climatic change attributable to increasing CO2r and (2) what observations and analyses can most effectively enable us to detect such changes and to monitor their progress? The most clearly defined change expected from increasing atmospheric CO2 is a large-scale warming of the Earth's surface and lower atmo- sphere. A number of investigators have examined trends in globally or hemispherically averaged surface temperature for evidence of CO2- induced changes. Although differing in detail because of varying data sources and analysis methods, the records of large-scale average temperatures reconstructed by a number of investigators are in general agreement for the period of instrumental records, i.e., about the last l00 years. Northern hemisphere temperatures increased from the late nineteenth century to the l940s, decreased until the mid-l970s, and have apparently increased again in recent years. The mean temperature of the l970s was about 0.5°C warmer than that of the l880s. To the extent that one can judge from scanty data, southern hemisphere tem- peratures have increased more steadily by about the same amount. ^n view of the relatively large and inadequately explained fluctuations over the last century, we do not believe that the overall pattern of variations in hemispheric or global mean temperature or associated changes in other climatic variables yet confirms the occurrence of temperature changes attributable to increasing atmospheric CO? concentration. This chapter was commissioned by and reviewed by the Climate Research Committee of the Board on Atmospheric Sciences and Climate at the request of the Carbon Dioxide Assessment Committee. The contributions of Hugh W. Ellsaesser, Frederick M. Luther, Robert A. Schiffer, David E. Thompson, and Donald J. Wuebbles are also gratefully acknowledged. John S. Perry and Jesse H. Ausubel provided logistical, administrative, and editorial support. 292

293 Other factors than CO2—such as atmospheric turbidity and solar radiation—also influence climate. Attempts have been made to account for these influences on the temperature record, thereby making the sought-for CO2 signal stand out more clearly. Unfortunately, only indirect sources of historical data are available for the time pre- ceding the short period of instrumental records. Moreover, strato- spheric turbidity has been inferred primarily from volcanic activity, and solar radiance from such phenomena such as sunspots. The quan- titative reliability of these inferences is unknown. Despite these difficulties, a number of investigators, employing various combinations of data and methodology, have related the global or hemispheric mean temperature record with indices of turbidity and solar radiance and with estimates of the effect of increasing CO2. Although good agreement between modeled and observed variations has been obtained in some of these studies, it is clear that enormous uncertainties exist. When attempts are made to account for climatic influences of such other factors as volcanic and solar variations, an apparent temperature trend consistent with the trend in CQ2 concen- trations and simulations with climate models becomes more evident. However, uncertainties preclude acceptance of such analyses as more than suggestive. Nevertheless, the studies done to date have been most helpful in raising questions, suggesting relationships, and identifying gaps in data and observations. In essence, the problem of detection is to determine the existence and magnitude of a hypothesized CO2 effect against the background of climatic variability, which may be in part due to internal processes in the atmosphere and ocean and in part explainable in terms of fluctua- tions in external factors. A reasonable approach is to attempt to decompose the record of some climatic parameter, e.g., temperature, into a hypothesized "natural" value, a perturbation due to CO2, and a random component. The "natural" value may be taken as a constant long-run preindustrial mean or perhaps that mean corrected for variable factors such as volcanic and solar activity. The random component can probably be treated as "noise-like" but will have to differ substan- tially from both "white noise" and first-order autoregressive noise. It is clear that the magnitude of the derived CO2 signal will depend markedly on the hypothesis chosen for the underlying climatic trend and the change in CO2 assumed between the imperfectly known preindustrial value and the accurately measured current concentrations. The success achieved by several workers in explaining the temperature record in diverse ways demonstrates the availability of a number of sets of hypotheses that can fit the poorly defined historical data and estimated preindustrial concentrations. The available data on trends in globally or hemispherically averaged temperatures over the last century, together with estimates of CO2 changes over the period, do not preclude the possibility that slow climatic changes due to increasing atmospheric CO2 projections might already be under way. If the preindustrial COo concentration was near 300 ppm, the sensitivity of climate to CO-? (expressed as

294 projected temperature increase for a doubling of CO2 concentration) might be as large as suggested by the upper half of the range of the study of the CO9/Climate Review Panel (l982), i.e., up to perhaps 4.5°C; if the preindustrial C02 concentration was well below 300 ppm and if other forcing factors did not intervene, however, the sensitivity must be below about 3°C to avoid inconsistency with the available record. If, as expected, the CO2 signal gradually increases into the future, then the likelihood of perceiving it with an appropriate degree of statistical significance will increase. Given the inertia created by the ocean thermal capacity and the level of natural fluctuations, we expect that achieving statistical confirmation of the C02-induced contribution to global temperature changes so as to narrow substan- tially the range of acceptable model estimates may require an extended period. Improvements in climatic monitoring and modeling and in our historic data bases for changes in CO2, solar radiance, atmospheric turbidity, and other factors may, however, make it possible to account for climatic effects with less uncertainty and thus to detect a CO2 signal at an earlier time and with greater confidence. Improved monitoring of appropriate variables can be of great importance here by allowing improvement and more effective validation of models. A complicating factor of increasing importance will be the role of rising concentrations of greenhouse gases other than C02. While the role of these gases in altering climate may have been negligible up to the present, their significance is likely to grow, and their effects will be difficult to distinguish from those due to C02. A monitoring strategy should focus on parameters expected to respond strongly to changes in CO2 (and other greenhouse gases)* and on other factors that may influence climate. Candidate parameters may be identified, their variability estimated, and their time evolution predicted through climate model simulations. Through analysis of past data, continued monitoring, and a combination of careful statistical analysis and physical reasoning, the effects of CO2 may eventually be discerned. Monitoring parameters should include not only data on the CO2 forcing and the expected climate system responses but also data on other external factors that may influence climate and obscure CO2 influences. Climate modeling and monitoring studies already accom- plished provide considerable background for the selection of these parameters. Since fairly distinct climate changes are only expected to become evident over one or more decades, monitoring for both early detection and more rapid model improvement should be carried out for an *It will be difficult to distinguish between the climatic effects of C02 and those of other radiatively active trace gases. Their expected relative contributions to climatic change will have to be inferred from model calculations and precise monitoring of radiation fluxes.

295 extended period. Parameters may be selected for early emphasis on the basis of the following criteria: l. Sensitivity. How does the effect exerted on climate by the variables or the changes experienced by the variable on decadal time scales compare with that associated with corresponding changes in CO2? 2. Response characteristics. Are changes likely to be rapid enough to be detectable in a few decades? 3. Signal-to-noise ratio. Are the relevant changes sufficiently greater than the statistical variability to be measured accurately? 4. Past data base. Are data on the past behavior of the variable adequate for determining its natural variability? 5. Spatial coverage and resolution of required measurements. 6. Required frequency of measurements. 7. Feasibility of technical systems. Can we make the required measurements? Initial application of these criteria leads to the list of recommended variables for monitoring given below: Priority First Second Monitoring Causal Factors by Measuring Changes in C02 concentrations Volcanic aerosols Solar radiance "Greenhouse" gases other than CO2 Stratospheric and tropospheric ozone Monitoring Climatic Effects by Measuring Changes in Troposphere/surface temperatures (including sea temperatures) Stratospheric temperatures Radiation fluxes at the top of the atmosphere Precipitable water content (and clouds) Snow and sea-ice covers Polar ice-sheet mass balance Sea level In the above list, emphasis has been given to parameters that may contribute, either directly or through model improvement, to detection of CO2 effects at the earliest possible time. Over the long run it is important to build up a relatively complete data base of the pos- sible causes and effects of climate change and the characteristics of climate variability, not simply for detection but also to assist in research on and calibration of models of the climate system. Once we become convinced that climate changes are indeed under way, we will seek to predict their future evolution with increasing urgency and with increasing emphasis on parameters of societal importance (e.g., sea

296 level, rainfall). We should thus anticipate that a detection program will gradually evolve into a more comprehensive geophysical monitoring and prediction program. It should be emphasized that the strategy proposed here is a single tentative step in what must be an iterative process of measurement and study. In subsequent steps, we urge more thorough evaluation with greater attention, in particular, to the following: (a) Time: When could we have a record long enough to make a meaningful contribution to policy formulation? (b) Long-run values for model calibration and verification. (c) Cost. (d) Societal importance. Collection of the desired observations will require a healthy global observing system, of which satellites will be a major component. Satel- lites can provide or contribute to long-term global measurements of radiative fluxes, planetary albedo, snow/ice extent, ocean and atmo- spheric temperatures, atmospheric water content, polar ice sheet volume, aerosols, ozone, and trace atmospheric components; a well-designed and stable program of space-based environmental observation is essential if we are to monitor the state of our climate. Table 5.l summarizes requirements and technical systems for monitoring high-priority variables. We will also have to continue to improve our climate models in order to reduce the uncertainties in predictions of climate effects and to validate the models against observations (although we believe that climate models are at present sufficiently sound and detailed to enable us to identify a set of variables that could form the basis for an initial monitoring strategy). Also, statistical techniques for asses- sing the significance of observed changes will have to be improved so as to deal with the characteristics of the monitored variables. In the end, however, confidence that we have detected the effect of C02 will have to rest on a combination of both statistical testing and physical reasoning. Finally, we must recognize that despite our best efforts there will always remain room for differing interpretations of data. Within our own country, different investigators have reached quite different con- clusions from the same evidence. The detection issue is inherently global, and interest is growing throughout the world. It is only to be expected that investigators in different countries may also reach different conclusions. This diversity of judgments may lead to unnecessary confusion and division among nations. There is thus a clear need for an international focal point or clearinghouse for data, analyses, studies, and periodic assessments relevant to the climatic effects of increasing C02.

297 5.2 HAVE CO2-INDUCED SURFACE TEMPERATURE CHANGES ALREADY OCCURRED?* 5.2.l Introduction In l938, G. S. Callendar suggested that mankind's fossil fuel emissions were causing an increase in atmospheric CO2 concentration and that this, in turn, was leading to the climatic warming that had been detected early in this century. With new radiative calculations performed by Plass (l953, l956) and an improvement in understanding of oceanic chemistry due to Revelle and Seuss (l957), the quantitative basis for Callendar's suggestion became more acceptable. In January l96l, in one of his last papers, Callendar compiled global climatic data on temperature trends in order to assess the possible role of increasing CO2 concentrations (Figure 5.l). He found "...that the observed distribution of recent climatic trends over the earth is not incompatible with the CO2 hypothesis, and that in certain cases the latter [i.e., the CO2 hypothesis] can supply a reasonable explanation." The carefully measured tone of this endorsement of his own hypothesis resulted from its failure, in his mind, to explain three troubling features of the observed climatic trends. These features included the following: l. "...the rising temperature trend is very small in the south temperate zone as compared to that in the north." 2. "The tendency for precipitation to decrease in many warm regions, and remain sub-normal during the first three or four decades of this century (Kraus, l955)." 3. "...the big rise of temperature in most parts of the sub-arctic zone during the l920s and l930s...as compared to the changes in the lower latitudes." Callendar felt that some aspects of these features could be reconciled with the CO2 theory, by, for example, considering the large thermal inertia of the oceans, greatly delayed transport of fossil fuel derived CO2 to the southern hemisphere, radiatively induced stabilization of clouds, lower-latitude induced contraction of the polar vortex, and the consideration of other factors to explain short-term fluctuations.t *Some of the material in this section also appeared in MacCracken (l983). tThe thermal inertia of the oceans, and not delayed cross- equatorial transport of CO2, is now believed to be important in delaying the temperature response of the southern hemisphere (Climate Research Board, l979; CO2/Climate Review Panel, l982). Radiatively induced stabilization of clouds is not now recognized as the cause of geographical variations in CO2-induced rainfall changes; rather, the shifting circulation pattern, including contraction of the polar vortex, leads to these variations. Ice-albedo feedback and the presence of a polar temperature inversion are now believed to lead to the amplification of temperature changes in polar regions.

298 3 3 C \ CO 41 to -H to rji 4J 10 rH kl C .H 41 Qi 41 -H rH § § 41 1 -e ki 41 oi a • a 4J o 4J a -U 0 O id U H D kl CO tO C rH 4J 4J X. 41 CO rH -H o 41 • a oi to 0 w rg 0 rH 4J Oi-H C Oi • 10 C CO rH O C 41 U CO -H rH tO -H -H -H Techni X. C 41 4» V rH ki •O 0 4J 41 C rH 10 C CO 10 Jf 9 41 -O 10 J3 CO 0 O 4J Tl Q Q CO • 3 j 5 £ io -H I >E V C c C C 1O J3 -H .*H -H n rn ve k, ki u « 41 e t3 = a — 4J 4J 4J I! TJ 3 n rH . .H .H .H 4J 3 CO0 4J 1O31 C c C M o 3J3-U C -a 01 4J - OOQ3to 0QCO-H 0C o o o J3 x: -H rH ai § -H wCCO I I I 1 *1 S rH gp l-t 1041 41 idCOrH 41 5i 3BC klklklJ<£ 3 • C *t " tU f trS-H ooOkiS crvo-w »2 41 u W J4 M O 41-*-H10 1O -a 3 w e sc -a rH *j ui ion $ idW41414141414J.H fl rH n) 0C j£41O-Hidid10C>« it d • 4J -H K I rH 3 3 3 rH ki« 10 W 4J io c 4j -P 4j -M iu oiwoi e o kl O>rHO"rrcT— rH Ot04lki (Drg O • ••••••Sfl > \ T) -Q o: > 414jC4J-a-o'OO-H 4JCVC-H Uu 41-riroidioroao otnaOrH ol >rks Required CrHCNCO jQ 41 CCCO10 4109 rH~ Jt Jt Jf 9 & O-UO 3J3 rH0i41kiklklQtO rHto4l CO io-UU4JOOO— 41•^ 9 9 M 9 "i Requireeents a e e ro * s * -—4i JQ-HO4J -Uki Ono1ki4J-U4Joiki O1aoio C0 rH 4J.Q 414141tOC'r)41 H Q 41 ki Q Dl OrH Oi T3 W .H CGC0O0£ cn u -" • C OiH 41>irH 4JNrnQ. 10>iC-H E 4J3COid 4J4J^-HOklCO -U4J341-H'O O-H CC 0 CcCrH «-H CcCTji'0c -UtO 41-Hki 414l4lrH1HrHe 000 u 3 • 4J <tf O COatOfuf0c4l tt (B >D 4, 41 O -Oki D c -a .u 41oioi4J41oij: 41 oi «) i u o tuo kl5-rl klklklflO kluO 41 I oiUi-} a. a, PJ w H cuhia z 4J 41 z c c -H 0 kl ff O C* 41 41 CO -H .H kl kl ki t> ki C 41 41 10 0 C O 41 O X ^ -HOi ra ilC eg WtOCO cc a cio o-n iaQOO Ol0 41 O^5 -H 4J -HE-HB EX! COCO gO 4J3 •O 4J 4J « 4J U rH HJ gi 3J3 • • 4 • k i O Di C41 ki4l J2 .H u -H -H O 41 CO O C O -t-1 -H u 411a4l iiJ 4J O • NO lM10 kl4J rHj:id£ hi -H §kl41 O N 8 4JCO f04Jkl4J o>to O to-H 0to-H -H k l ui -H 1030COrj kirH u -H T1 u u-l u, uj 01 41 rH OJ 41 O T1 C .H O (110 3 T3 4JO«O -*JO-OUO.C-HQkl'H 4J 4J O CO rH i-Hg-«esJi!a oj1" idO-H o»aoo. CP E uj Xiau 01 o u -e oj go u-H ki kiOeO -H 3 O O -H 41 0) idQi£ a 41ki4l ki4J4J ki -Hrn£ ki 4 C9 tt • A uto a«41X 41^0 coco •OioQ, aiScOo" io ** oi E,£ a 4J4,4141 0.41 -nS *Olos*'8 o^0 ki Sjja. a CJ54JJ5 m C O kir-OaOi 41 41 4jOto O O'O-U'J*' >,3 Ou-tJ u-uJJrH^O >,i« K a 0 4J -iki ai H kirg a 3 • o1 -H CD -c w n u «-i 7i H ** a n> -U104irH4J idO kirHki x: Z z: U W H 1041 CO ff^O k i idjt10<ufO in 41 4J g V 0 <• 3 g 3 J3 3 k i 4J -H&l41 -H 41 u w co o •3 1 1 0 to C w •a 3 « u 1 kl £ kl -C 4J rH O 0. JJ PL, 41 < fj « a 2 • 2 • • EH 10 J3

299 II 3 r-4 M (O c 41 k i ia to S3 s 41 CJi oi 3 4J -H -H •H H 10 *rl S3 | I Ol rH rH 41 41 t1 4J 55 1 •H 10 EH CO ^ -c k i 41 tji C CO o 5 41 Sou 4J 41 O *4-( •H 0 C -H k i xs 4J 3 u a oi u n 'o ? g en u ai 3 |0 iM I Is • i u UH 10 4J O rH o •o "'S o> 8, g >| 41 C 13 -o -5 i -Hct o 4-i ra c 0 41 3 Oi •H 4J » M O **H O1 *j i! C C k i C 41 \ (0 (0 41 10 41 i! 4) « 3 in -H 4J a> O> k l 41 TJ -H >i rH 0 41 a1 k i C rH O( H rH (0 41 Ol U-l » 4J u • n •a a s|| J " 4l 10 10 » ^ 3 O O 1 3 •H 01 4J S!3 IrH a 2 l^rH1 •** ^ • s * ^ o c TJ t-j rH O 41 rH rH 4 J ^^ rH •U J3 41 O v4 4J •H 1g 41 41 uS c H1 1O rH > & C 4-1 0 g 41 41 a a 10 to ki 10 10 • i< M • 4J > 1o a oi co a -2 * (U fcl Oi Ql « CO a. — a. 4J •i • ; £ S rH j< y o IO 3 iu •H J •,- O JJ 41 a p 41 k i .U « 0 J3 §c t1 c T3 W 4> a 01 J3 8 s§ O 0 H H u C Q O u Qi-ri 0 1 1 H 10 41 3 a H 5 > S 0 O 4J £ k4 E -•H 4 % c 10 41 O< 41 41 41 4t 4) H( H o 41 u 10 4J 0 Oi 41 « CO 3 4J X. c ax: 41 4J Q* 41 u C 10 10 u •0 rH ^H OJ u k i c 41 41 c 5 ag o> ai 10 X a. • > -n • J3 41 g = £2" 8 •H o •" a u 0 CD m n ia rH 41 H 10 E-i Owe. > 10 o n ai O 10 rH ^j 41 c 10 0 rH 41 U C J CO n SB 0 O CO

300 North temperate zone Tropical South temperate zone 60° N to 50° S 1880 1900 I 1920 YEAR 1940 1960 Sub-arctic 60°-73° N North temperate zone Tropical South temperate zone 60° N to 50° S 1860 1960 FIGURE 5.l Reconstructions of changes in surface air temperature for different latitude regions by Callendar (l96l) applying (top) a 20-year moving average and (bottom) a 5-year smoothing to the data.

30l He did not, however, believe that volcanic eruptions or solar varia- tions could adequately explain the observed long-term climatic warming and the latitudinal pattern of the temperature change. It is interesting to contemplate how much more forceful Callendar's endorsement of his CO2 hypothesis would have been if he had had available the results of today's models. Although present climate model simulations indicate a sensitivity about half that found by Plass, they do project smaller temperature increases in the southern hemisphere (due presumably to ocean inertia and albedo effects), provide for regions of both increasing and decreasing precipitation, and indicate that there should indeed be a large polar amplification of the CO2-induced global temperature increase. Predictions and observa- tions would have seemed to be in agreement, and detection might have been claimed. Perhaps quite fortunately, the absence of present-day model results and the lag in theoretical understanding prevented a consensus from developing, however, for later in l96l, J. Murray Mitchell, Jr., presented his now classic paper on global temperature trends. This paper reanalyzed the data on temperatures of the previous hundred years and found that the previously claimed warming, when properly weighted by area, was not really as large as had been thought by Callendar (196l) and Willett (l950). This finding actually would have improved Callendar's agreement with today's model results. More importantly, however, Mitchell found that the climate during the l950s was, in fact, cooling (see Figure 5.2) at a time when the CO2 trend should have exerted a warming influence.* The reversal in temperature trends found by Mitchell set the stage for intensive efforts during the last 20 years to untangle the web of factors that influence the climate. The extensive monitoring, research, and analysis since then has taught us several lessons that should enable us to set a course for identifying the projected CO2-induced climate response. This chapter will review some of the diagnostic studies that have been performed and the problems that have arisen in the attempt to detect C(^-induced surface-air-temperature change. The chapter is not intended to serve as a comprehensive review of all studies analyzing the climate of the last l00 years but rather attempts to point out general problems and to use selected studies to illustrate how difficult these problems are to resolve and where we stand now with regard to identifying the climatic effects of increasing CO2 concentrations. 5.2.2 Requirements for Identifying CC^-Induced Climate Change To achieve widespread confidence in the projected climatic effects of increasing CO2 concentrations, both a consensus of climate model Concentrations were increasing relatively slowly, so very little warming should actually have been expected during this period. The cooling, however, is deserving of explanation.

302 0.8 0.7 0.6 0.5 ~ 0.4 u q^ $ 0.3 0.2 0.1 0.0 -0.1 I I 1860 1880 1900 1920 YEAR 1940 1960 FIGURE 5.2 Construction of changes in surface air temperature for different latitude bands by Mitchell (l96l). Average data are compiled for successive pentads starting with l870-l874. results and unambiguous identification in recent climatic records of the initial stages of the projected C02-induced warming are necessary. As expressed in recent reports of the National Research Council (Climate Research Board, l979; C02/Climate Review Panel, l982), the change in global average surface temperature (the most closely scrutinized model result) due to doubled C02 concentration is likely to be between l.5 and 4.5°C, with more physically comprehensive models giving results in the range of 2 to 3.5°C. While increases in the global average temper- ature of even l.5°C would lead to a historically unprecedented climatic situation, the range of estimated temperature (and other climate) changes remains wide and introduces considerable uncertainty into the detection debate. As the reasons for the spread in the estimates are becoming clear, we may expect that with further research the range will narrow. If we can achieve a consensus of model results, convincing identi- fication of the projected CO2 warming will increasingly depend on observational data, inspiring confidence and certainty sufficient to allow comparison with model results. This will require (l) that the necessary climatic data bases be accurate, comprehensive, and of suf- ficient length to allow application of appropriate statistical analyses and, in particular, identification of whether a change is or is not

303 occurring; (2) that data bases exist to evaluate the role of the factors that may have caused climate changes of comparable magnitude; (3) that the climatic effects of CO2 and other factors be estimated with suf- ficient detail and confidence that the projected induced changes can be identified amid the fluctuations, perhaps similarly patterned, caused by other known causal factors, unrecognized causes, and natural fluc- tuations (see MacCracken and Moses, l982). In the following subsections each of these points will be considered by example. 5.2.2.l Climatic Data Bases In the analysis of climatic change, and in particular in the search for evidence of CO2-induced effects, one is faced with the limitations in the availability and accuracy of climatic data bases (see Table 5.2, for example). As a result of these limitations, the primary indicator used in diagnostic studies of climate has been the change in surface air temperature during the last hundred years. (The problems illustrated and conclusions drawn in this section apply equally to analyses of variables other than temperature, such as sea ice extent and sea level rise.) This variable has been chosen because it is the only one for TABLE 5.2 Causes of Differences in Temperature Anomaly Data Sets (a) Differences Arising in Data Selection and Compilation Number of stations used in compiling average Methods for eliminating effects of unrepresentative stations (e.g., urban heat-island, station location changes) Relative distribution of stations over land and ocean Sources of data (ships, land sites, islands, for example) Treatment of stations with records starting in different years Differences or changes in observation times Unrepresentativeness of the sampling network Absolute temperature versus value of anomaly Methods of accounting for missing data Use of annual average versus monthly data Method of constructing daily average temperature. (b) Differences in Averaging Methods Time periods (range from annual to multidecadal) Spatial domains Climatic baselines and periods of record Time-averaging methods (running averages versus finite period averages) Spatial interpolation techniques (sum of representative sites, interpolation onto grid, weighting of stations, for example) Means of selecting normals and identifying trends

304 0.8 0.4 -0.4 -0.8 Jones etal. (1982) Vinnikovet al. (1980) , l. , , I , 1880 1895 1910 1925 1940 YEAR 1955 1970 1985 FIGURE 5.3 Comparison of the reconstructions of annual surface-air- temperature anomalies for the northern hemisphere from Jones et al. (l982) and Vinnikov et al. (l980). Figure from Clark (l982), but data for l98l added to Jones et al. (P. D. Jones, Climatic Research Unit, University of East Anglia, Norwich NR4 7TJ, England, personal communication). which a long record of measurements exists at many stations and because it is a convenient and straightforward, although not complete, measure of the climatic state. While .some investigators have attempted to develop data bases of global and polar temperature change, the most used indicator has been the change in northern hemisphere average surface air temperature. Comparison of the temperature records compiled by several inves- tigators shows that not only has temperature fluctuated but also that the temporal pattern of the temperature anomalies is not uniquely established (Figure 5.3). Estimates of surface-air-temperature anomalies using similar data sets and techniques agree well [e.g., Borzenkova et al. (l976), Vinnikov et al. (l980) and Jones et al. (l982)], although there remain differences in the details of the compilations that remain to be resolved (World Meteorological Organization, l982a). Hansen et al. (l98l) compiled data on the global temperature pattern and found a very different pattern of temperature change in the southern and northern hemispheres (Figure 5.4). Jones et al. (l982) show that the changes have different patterns by season. The differences in estimates of actual climatic change arise primarily for two reasons: the choice of data selection/compilation methods and the choice of averaging methods. Table 5.2(a) lists some of the problems that arise in data selection and compilation; Table 5.2(b) lists problems related to averaging techniques. Consider just a

305 Northern latitudes (90°N-23.6°N) Low latitudes (23.6 N-23.6°S) Southern latitudes (23.60S-90"S) 1880 1900 1920 1940 YEAR 1960 1980 FIGURE 5.4 Reconstruction of surface-air-temperature anomalies for various latitude bands by Hansen et al. (l98l).

306 few examples as illustrations of the problems. Mitchell (l96l) used Willett's (l950) data and used an area weighting method rather than averaging of representative stations; this reduced the estimated wintertime climatic change over the previous l00 years by a factor of 2. Yamamoto (l980) used both 30-year averaging (which might be appro- priate when considering the effect induced by the thermal lag of the ocean) and 5-year averaging; the time of maximum temperature and the start and pattern of the recent decline in temperature appear to be delayed about a decade when the longer averaging period is used. Paltridge and Woodruff (l98l) used ship data on sea surface temperature (SST) as an indicator of surface temperature in oceanic regions rather than the isolated island data used by most other investigators; they found the time of peak temperature delayed about 20 years and, quite surprisingly, a larger temperature change than shown by the land-based records. Their results, however, should be considered as only an initial examination of the SST data in that their averaging techniques for handling gaps in the record raise many questions. An additional complication in using available temperature records to estimate temperature changes induced by increasing CO2 concentrations is that the records do not extend back to before the time that CO2 concentrations began to rise. In addition, we cannot be sure that the climate prior to the start of large anthropogenic CO2 emissions was in equilibrium so that comparisons between early and late parts of the record may include biases due to trends, natural variations, or changes induced by non-CO2 factors. 5.2.2.2 Causal Factors The state of the climate system is determined by the interactions of a large number of processes and factors, some external to the system (e.g., solar radiance, aerosol and trace-gas concentration). A change in CO2 concentration is only one of the factors that may induce a change in the climatic state. Moreover, climate models show that, even when all external factors are held constant, there will still be substantial climate fluctuation due only to interactions between various processes having different time scales. The fluctuations will be even greater when variations in external factors are allowed. While there are several potential sources of climatic change, our discussion concentrates for the moment on CO2 concentration, volcanic aerosols, and solar variability, possible major influences that have been addressed by a number of investigators whose work deserves comment. In the future, other factors, such as changes in anthropogenic trace gases, may become influences comparable or greater in magnitude. Finally, we must recognize the possibility that some important factor may have been underestimated or remains unrecognized. To identify the climatic signal caused by increasing CO2 concen- trations among the background of fluctuations requires either (a) that we treat the changes of all other factors as noise and wait until the CO2-induced climatic change is large enough to cause a statistically significant change (assuming implicitly that no additional competing

307 influences have come into play) or (b) that we account for the relative roles of the possible factors that may induce changes comparable in magnitude to increasing C02 concentration during the period of interest, thereby reducing the amount of unexplained variation and allowing easier identification of the C02 effect. In the record of the last l00 years, the standard deviation of monthly global mean temperature ranges from 0.23 to 0.65°C depending on month (Jones et al., l982), while the change in annual global mean temperature is on the order of a half degree. Presumably the year-to-year variations were not due primarily to changes in C02 concentration. Since both observed variations and predicted changes are of similar magnitude, the latter approach of accounting for the role of as many causal factors as possible must be pursued if we are to have an early likelihood of identifying the projected C02 warming. To attribute climatic changes to causal factors requires adequate data bases for the changes in each causal factor, as well as for the climate state itself, over a period sufficiently long that changes attributable to the causal factors are comparable with the level of natural variability. Moreover, for carbon dioxide the record is not yet adequate. The mid-nineteenth-century baseline concentration is thought to have been between 250 and 290 ppm, and the time history of CO2 concentration between l850 and l950, when changes in the bio- sphere may have played a more important role than fossil fuel emissions, is not yet well defined (cf. Chapter 2). These uncertainties allow estimates of the change in CO2 concentration from l850 to l980 to range from as little as 50 ppm to as much as 90 ppm, which in turn converts to a factor of about 2 in the estimated mean value of the warming attributable to changes in CO2 concentration over this period if other factors remained constant. Combining the uncertainties caused by the range of possible change in CO2 concentration, the range in model estimates of the temperature change for a C02 doubling, and the time constant of a thermal lag in the ocean, and using a logarithmic approximation to relate the CO2 radiative effect to the temperature response, one finds that, assuming no net influence of other factors, the expected CO9-induced temperature change since l850 may range from a few tenths of a degree to more than one-and-a-half degrees; estimates in the lower part of this range appear more consistent with the climatic record (see Figure 5.5). The data bases used to estimate the role of volcanoes—presumed to be an important factor in creating fluctuations over at least the last l00 years—are quite uncertain. Table 5.3 lists some of the factors contributing to the disagreements between the various data sets that have been used to account for changes in volcanically injected strato- spheric aerosols. The different records show quite different relative magnitudes and temporal patterns of volcanic influences. Lamb's (l970) dust veil index, a measure of atmospheric aerosol content, is perhaps the most frequently used index, but various investigators modify it, often in ad hoc ways, thereby perhaps affecting later calculations (see Figure 5.6). Gilliland (l982) uses acidity as measured in a Greenland ice core as the basis for variations in stratospheric aerosol, suggest- ing that the volcanic chronology from this record should record the

308 2.5 2.0 1.5 IH <a 1.0 0.5 4.50 3.75 3.00 2.25 1.50 230 240 250 260 270 280 290 300 310 320 FIGURE 5.5 Relationship between C02 change, temperature change, and climate sensitivity assuming no other forcings. The abscissa represents a range of values for mid-nineteenth-century C02 concentration. The ordinate represents the increase (AT?) in global mean equilibrium surface temperature between l850 and the period l96l-l980. The response is calculated for a range of values of ATd, the change of global mean equilibrium temperature for a doubling of C02 concentration (assumed independent of initial CO2 concentration) , and assumes that the temperature range is logarithmically related to the change in CO2 concentration (Augustsson and Ramanathan, l977) . An ocean response time (mean thermal lag) of l5 years is used. The concentration of C02 was assumed in each case to increase linearly from the indicated value in l850 to 3l0 ppm in l950, and then linearly from l950 to 340 ppm in l980. Note that if the temperature increase from l850 to the interval l96l-l980 is taken to be 0.5°C, then for consistency, ATd may be as large as 4.5°C only if mid-nineteenth-century C02 concen- trations were about 300 ppm, whereas ATd may be as small as about l.5°C if mid-nineteenth-century CO2 concentrations were as low as 250 ppm. For ocean response times shorter than l5 years, the isolines slide upward. Varying the time of the start of the increase in C02 concentrations from l850 to l920 has little effect.

309 TABLE 5.3 Factors Contributing to Uncertainties in Creating a Volcanic Aerosol Data Base for Assessing the Climatic Effect of the Aerosol Reliance on variables not directly related to the radiative effect of volcanic aerosol (e.g., ejecta volume, volcanic explosivity, ice-core-derived precipitation acidity) Single or limited measurements extrapolated to global domain Lack of latitudinal resolution or pattern of aerosol distribution Assumptions about stratospheric lifetime of aerosols Lack of seasonal resolution of aerosol distribution Use of surface measurements to estimate changes in stratospheric aerosol (thereby introducing potential problems if trends exist in tropospheric aerosols) Averaging period of volcanic aerosol loading Definitions of major volcanoes and of the number of volcanoes considered Assumptions about size distribution of volcanic particles and of chemical composition of gaseous emissions, leading to ad hoc adjustments Assumptions regarding lifetime and distribution of aerosol due to season, latitude, and height of injection of volcanic dust and gases Circularity caused by estimation of dust loading from observations of subsequent temperature change high northern latitude volcanoes often cited as missing by other investigators—but strangely not evident in the core record. The recent Smithsonian compilation of volcanoes has also greatly expanded the number of volcanoes considered (Simkin et al., l98l). New listings, however, do not always reinforce confidence. For example, the recent compilation of an explosivity index by Newhall and Self (l982) attributed greater importance to the eruption of Mt. St. Helens than to Agung, whereas a sulfur-based stratospheric aerosol index would place Agung as being of much greater importance than Mt. St. Helens. In addition to estimates of dust amount, a few indices also rely on actual radiative measurements (Figure 5.7). The general pattern of the two types of data show broad qualitative similarity, but details of timing and magnitude are quite different. Even worse complications exist in generating data bases to evaluate the effect of changes in solar radiation, the third major factor often considered in these analyses. For this factor, the main problem is that there are several indicators that have been used as surrogates of the solar radiative flux reaching the Earth, but there is virtually no physical basis for deciding between these indicators, or even if any is relevant (i.e., is the solar constant constant?). Quite clearly, without adequate historical data bases on the causal factors, identification of the CO2 part of the climatic signal may not be possible. Therefore, examination and improvement of the data bases on potential causal factors must be an essential part of the early detection effort.

3l0 M 50 O 40 K k 30 20 * 10 -l 1 1 1 1 1 1——r (V) stratospheric loading by volcanic activity Adjusted to northern hemisphere - 120-yr average V 4.2x1 (olons 1850 1870 1890 1910 1930 1950 1970 UJ C DC O / Si CC I- 30 20 10 y o Krakatoa Bandai San, Ritter Island 4 Awu 4 Mont Pelee, Soufriere Santa Maria Shtyubelya Sopka Katmai 1880 1900 0 LU > 1920 YEAR 1940 1960 0 = 1 50 X UJ Q - 100 UJ i 150 200 250 J_ 1890 1910 1930 YEAR 1950 1970 FIGURE 5.6 Estimates of stratospheric aerosol loading by (top) Mitchell (l970) and (middle) Oliver (l976) based on use of volcanic index of Lamb (l970) and by (bottom) Bryson and Dittberner (l976) using volcano index of Hirschboeck (l980). The model of Oliver (l976) uses a residence time approach to calculate aerosol loading after injection. The curve of Bryson and Dittberner (l976) represents a l0-year running mean in arbitrary units.

3ll DC O H1 cc Ul 104 102 100 £* 98 01 < £ 5 H1 < Q- CC 96 1880 I I 1890 1900 1910 1920 YEAR 1930 1940 1950 1960 a. 1ll Q t O O K 0.12 0.10 0.08 0.06 0.04 0.02 1880 1900 1920 1940 1960 1980 YEAR FIGURE 5.7 Estimates of stratospheric aerosol loading based on surface measurements of downward direct solar radiation (top) from Budyko (l969) using data from cloudless days of each month with l0-year smoothing, and (bottom) from Bryson and Goodman (l980) using data from 42 stations between 20° and 65° N. The apparent inverse relationship occurs because direct radiation measures the clarity of the stratosphere whereas aerosol optical depth measures the lack of clarity. 5.2.2.3 Relating Causal Factors and Climatic Effects Even given precise records of how the climate has varied and the history of the important causal factors, a number of problems arise in attributing appropriate components of the climatic fluctuations of the last l00 years to the various causal factors, the remainder of the fluctuations being assumed to be natural (or, more properly, natural

3l2 TABLE 5.4 Limitations in Determining the Relationship between Changes in Causal Factors and Changes in Climate Imperfect or incomplete modeling of relevant climatic processes (e.g., with respect to the oceans and cryosphere) Imperfect model verification with respect to factors of interest (e.g., lack of sufficient test cases to verify treatment and time constants of long-term processes and of model performance as a whole) Limited areal extent of analysis (e.g., less than three-dimensional, less than global) Use of equilibrium rather than the transient perturbations in developing relationship between climate and causal factors Internal variability of climate models Limited length of model simulations Different times at which analysis starts (for example, the decade of the l880s was strongly volcanically affected; a stable baseline may not be available) Different temporal patterns of assumed climatic response functions plus as yet unexplained). For the attribution to be convincing, the change in the causal factor must be physically related to the climatic change in a quantitative way, which requires both theoretical under- standing of how the causal factor affects the climate and the ability to calculate the effect numerically. Since there is more than one causal factor, we must also be able to calculate the effects of inter- actions when more than one causal factor is acting. Table 5.4 lists some of the problems that arise in attempting to calculate the relationship. As an example of the type of problem that exists, consider the results of different approaches for treating the volcano-climate coupling used by different authors. Although Vinnikov and Groisman (l982) and Bryson (l980) both indicate that they use the same surface-air-temperature record and similar actinometric measures of stratospheric transmissivity, their different approaches to relating the causal factor to the induced climatic change lead them to very different conclusions about whether there has or has not been a temperature increase attributable to increasing CO2 concentrations. Rather than specifying a response function, Oliver (l976) and Mitchell (l983) chose generalized volcanic response functions that were then optimized in order to fit the temperature data. They also come to differing conclusions about the role of CO2 in the temperature record. Rather than employing such empirical and statistical approaches, Hansen et al. (l98l) used a one-dimensional radiative-convective model with a thermodynamically interactive ocean to simulate the climatic response to changes in the volcanic aerosol loading (and other factors). They then derived a quantitative relationship between C02 concentration and temperature to use in their analyses of the causes of climatic fluctuations. Such use of physically based relationships is an

3l3 important requirement for studies of this type. Comparison of their one-dimensional calculations with the different time histories of the northern and southern hemisphere temperatures, however, makes clear that there is a need to analyze the observed fluctuations with models having, at least, greater spatial detail. Such complications as arise in relating volcanic eruptions to the temperature response also arise in attempting to relate other causal factors to temperature change. In some cases, these complications can be resolved by just expanding the models to include omitted, but important, processes and domains or to treat the transient as opposed to the equilibrium response; in other cases, the complications can only be resolved by improving existing data bases (e.g., extending the temperature record back in time) or gathering new data (e.g., changes in Antarctic ice volume, if our climatic variable is sea level). As illustrations of the difficulty in untangling the recent climatic record, the next section considers several recent attempts to acquire the necessary climatic and causal factor data bases and then to relate these data bases to climatic changes. 5.2.3 Attempts to Identify COp-Induced Climate Change Mankind has always sought to relate climate fluctuations to causal factors. Certainly, the seasonal variations of solar position were widely recognized by early man as being related to the seasonal vari- ations in temperature, even if the reasons for the change in solar position were not understood. Geologists once feared that the Little Ice Age was caused by the cooling of the sun as its fuel sources ran down, until they found evidence for very cold periods thousands and then millions of years earlier. Since discovery of these ice ages, there have been many suggestions about the causes of climate change. Current views of many scientists, for example, are that the sun's energy output has actually been increasing over geological time, that vari- ations in the Earth's orbital parameters were major factors in the Pleistocene glacial cycles, and that the cold decades that characterized the Little Ice Age resulted from small fluctuations in solar output. Although changes in CO2 concentration were recognized as a possible factor in long-term climate change during the last century, the work of Callendar described earlier has served as the basis for numerous studies on that subject during the last 20 years (see Table 5.5) . These studies have used different data sets, different analysis techniques, con- sidered different causal factors and, perhaps not surprisingly, reached different conclusions. Because of these many differences, comparing the results is not straightforward. As one measure of the differences in findings, Table 5.6 presents the results of a representative set of investigators in terms of a ratio measuring the relative importance of each causal factor in causing a climatic change to the maximum variation of the climatic variable in the record that was used. In making these comparisons, it should be noted that in many of the studies there has been some effort to achieve an optimum fit to observations by choice of data base or by adjusting

3l4 C JJ <U (0 83 S1? k l C 41 •H M JQ oi O 1° U CO 41 UJ p « O -O O CO •w c 2 o •U CO O ffl u • ln 3 m io W Q1 9 | <c (u H EH U M Q * iM ,O —. I — O 3 0 1 -•H rH 4J 3 OI o> O O Oi C c x 41 Q -H -H U iM B rH — iM iIH u ja ta °*I S gs 10 3 O O - -H 4J O k i SO 2js 8 a u o Oi 41 41 o u > 4J 4j ' O 9 O 41 o Se -o ja V "* M 4J 4J > o o c c 41 -rl S c *9 S -H O 2 q « Oi •u ia M u j3 41 4j 0, < C § O J3 CJ CJ §x: Zi O tc -a o O u en c m 4J c oi 4J O 01 M Q >i— OQ *J 41 en O1 T3 4J 4J rH 4J 41 (0 CO ~- « 4J •H S i8 S O S 0 1 8. § Cn U rH 41 k i -H rH O, 41 4J C rH -U B O O iu 41 C > •o aO10i 301C tri c 0ujj.j= OiO-H V V > (M>a OQ O • ai e • !•> in • ri M CO 4J C 4J io OO in en • • in » crt oo I o o vo oo 00 r>l * 9 o •0- en O CO o o o p» en -g a U • •s 4J •H £ 11 (0 flj n 0 >iS k i 0 r» 01 >i-n >i •S>-g n n £1 a t* oi O 00 en oo in co roM »* o o in en — 0 o «3 I O O en en oo 00 rH u O -* (0 en o 4-1 O -J — >, Q CO 03 05

3l5 ai 5 E k i -a 10 I 41 a 10 n jSljl *l "S -H I 41 *' O J O 41 4t 41 "H . u a cn O C W C O -H 41 C -H -n C U1 3 irj 4J 10 di 3O O k i iM rH k i 3 0.O o< > M O i r* o1 a - S S • •4J u .Q 2 10 £ ft 3 S U trJ c tO 3 ug"° O -H HH »O rH S3S 41 u c e •a -H .H C rH 10 C 41 rH O 0> c Q -H •§ P C 0.91 1 -H iD e rH k i 41 10 4J o C r- 41 00 JJ rH <Q 1 4J C -H O rH U 10 41 JO O tsS 10 10 o ja cSV n U u o c O 85, •a n C 41 .C i0 o 4J 4J u Ul -H JJ O 0 S M IH O, •O rH 01 -H n) 01 Xi O C > 1O -H u -* S £ *> rn Q 10 4J -H O X cS'.'S e o > -H fn C u ai 41 -H C 01 Qt CO *M 10 O 'O CO 41 * » -H rH iW 3 -H in 4j a* i -a c O w rt) 4j 1O 41 O rH 41 o -rH 3 ja c 4-1 in rH 10 CO 41 a fi o -H u 41 O O i0 CM u -H ^ 4J o -H 4J ^ CO U 10 O -O rH 4j -- u_j S O •a rH Hi S id 4J O03 2 •J u u T3 fo 4j 5'. 0 41 u 41 Q.iM in n C 10 01 35 3 s -3*2 " "« S2 CO O 03 U §i iof n -** o1 • o CM A -H fff iTI Oi O • 41 • O CM O 0| I ? CM oi r* o4 CM r* r- 04 r o r- 04 «n CM r.- ' CM ' A ^ ^ "- -o 0% 0O 03 4! rH O 1 1 O O in co r* 01 m o r- ao Q 0O GO O O O 1 1 1 k i 0 e bl -a -a 00 -S 00 rH o1 rH r* H r- P^ o1 o1 O C9 rH 0 d 1B n TJ M «a n 10 > 10 0 c 41 a 41 n | a 4l 0 o i rH —t rH 0 rH o oo eo O 00 • rH o -. VD OJ O CM o i 9t 00 rH 00 1 rH O ^- CM o o 9i 91 JO O1 ^- 9i 9l rH P^ — 9i I O K I i 4J S, I 91 1 rH I- 1 °° O CO CM rH 9. -— O 0| I O 00 en § in

3l6 ? i - P 5 g >i « C ss 'O 10 •H OO |. 1 o rH 0 4J Hg H - P es « . 0 0 •O in oi 41 c-. 0 1 en -» .-. (O 41 rH 41 4-i aeoes solai l_l •H 'O • M c c ^ 'P 0) 5 J3 iH 41 0 -u tM Qj 10 CJ (0 TJ uons reuor C JJ 41 O 0i _ 41 •H -H 10 rH Cn 41 • o'S o o «H -8 S (O -O H en 10 -H g 4J VH 0 41 -H c u O t. 4j ja co oi •o i 41 rH —. c >H M 01 O * n iw 3 m a o o k i rH >i V4 * 'Si >i O •a co o> (O §u n O 41 O C rH 41 rH C u 10 -H •H • c o> 10 J3 to iU 41 rH JJ 10 O^ W •H C •H -rl •H (O 41 C *J 55 Y*. 1O *O *H 41 M >i JO rH 0 S u 41 •O •o rn O1 (0 C O 4) 41 10 41 3CCO UH 41 ri cu a u p I •H Qj X IB (B >. m u_i .gh-latitud Maeabe aed Ramaeathae 41 3 HBO oueaeiu mi: piriual St. sed oe Mae estimate o statistiua deteutioe . sed oe 30— J3 . nge for 12 differeet CO 2 fossil emissioe sc temper a tun oueaeiu lai estimate o: C°5 j o> c -H IH CM 4 J CM •HO* ull 10 > •P tO o to co M r- 8 •n 10 A S CO CO & 41 2 33 i H (N (0 p -o 1 ^•ss 41 > 1 — 'o 3 00 JJ -H M 41 UH J3 'H Oi eji iH 4J >i -H O 3_ o o f ' C c X 41 >o • H 4j c e O 41 O § 00 'a •* W **H 9 iM H rH ** WH s in Co i <H UH 3»l U) CN 0 ai g 00 *• H 10 H| O0 00 • *s D) u Z Is Q, « CM iH CM H g o 5 0 O 4J tD ! e 4J & •H 3 rH -0 O1 M 2 -5 3* O -— «. .-— * C -H __ . O O Oi iw « *J -*Q 01 CO CO ^_^ •H -H c c PN Cn CTt r-~ (O 4 J en • la TJ rH rH rH H 1* c ,— ^ rH Q1 qj S* 41 O 0 crv >. « as 41 rH oo co rH 1 CO •H -0 A •H CM CO CO rH "H 10 cn 00 O 41 a o rH rH 00 O 4 J r^ | 41 00 •H 41 Ot rH rH O M •o ja *J CO rH 00 «j ~* C B •H O t-fl ^* rH 00 00 CO tJ oi _. 3 4J C •H 4J 11 -85 CO 41 -H 01 O > O U3 - P O o o ~— • g in ' • co 41 JJ IM >o o si °H k i S § rH o u iM 41 (N M rH 41 4j 0. < 86 41 ID 0 o o S 9 • 9 O H oo 00 01 .— » ^_^ CM rH >~H (O ^ *»*,. rH — - <xi 41 00 fM ^^ ^^ en c o^ 00 f1 • rH O r— ) Ch S rH 1 — !-3 *^ i 5*™* A * C -0 41 -g rH *J 10 C "1 co 0 o (O r_| • u c c c *H ^t 00 -H rH ^ O 0 B 0 If O oi en •> •H OJ -H B ja 3 IS 0 1 g c U rH rH •H 15 3 -H •H 4j < £ S > X W s g° UH 41 C 41 *J o a. a M-H 41 £ ' *J -0 -P i o 1 3 2 ^ i ^ Jg rH : ca 4i o I rH 41 ! 0o ja u. cio >i a 41 x o J3 k i 3 >i C > 4J O -H 41 k i •H 4J .Q 41 «J co a o "* *H B rH ^J 4 J &s gs u 0 E 41 C -H rH 4J -H 4J 10 .S - a . . O O g O -H Oi S • S a 2 n «; jn -H jj w jj •H co f3 C k l > Tj 10 C0 1 CJ1 o • (0 (0 ^ -e ? c 4 J -a ra •i ' o 4JCJ fl c js HJ u O ;4-c M Lw8 u 10 U X 41 •H^gjOlWlMMH k i -O 10 > -P O VSgPVrHc -O-pcoO-H3cuO o E c jr u oi -H ^J § . o o1 c -P 1o JQ UH 4J es C T3 O 4J C 3 -O 3 -u 41 O* "* C. C .H 41 rH to 41 fl -H 01 o O -H -s 10 If] IH 41 co 10 .35 ; g1 ln 9 4 J C C O 10 o O - JC -H O Oi rH 41 rH 4JO4I C 10 C -H C k i J?< O O rH O cCin-Hcooiio 2OOfOpHcQ-CCD i0looa OJ3I i 01

3l7 O 10 4J C to cn Q 10 c k i J3 .H ii iii i • - i i i i § 10 O u rH 0 M C CO 4J O CO OL, + 1+1 rH 0, o1 « <0 C 41 ,2 T1 u S'S 5 •rt W 10 3 U 0 1 —. O 41 • CJO 41 0006 u 41 3 a 1O fit -H rHrH IS « 3 > >i >O ^* u C „ 3 10 41 D -H 3 0 3 CO 41 g. O 10 J3 -H 41 • Qi 4J rH il 1 - i 1 1 1 i 1 i i 1 oo k i B t? rH- § C 1 O o -o <0 41 0 >, O O eo O iS rH O k i 3 „ oi 25 o •a 0 C5 k i 3 u SO 41 JC T3 349 10 CO S I0 -ri >i CM >1 -H rH 11 o io « 1- 11 111 1 1 1 I 1 • 1 1 ft) a f a o -H 4J O 1 •flO * tO "^ 4l C 10 .si 41 n 3 -H 1O $ 10 4J 3 CO O O % 41 > 4J 1O 41 k i u O k i kl k l 3 C 41 >, f1 -O O 41 -H 41 ISPS 11 111 1 1 t 1 1 • 1 1 41 u 4J u < •O 1 cn o o to 0 C 0 « 41 CO 10 Ol HH '* o c n x: u o CO O ^ 0 rH •O 41 10 0 UH U * 41 j: CO J2 k i JJ 3 rH k l CO CN rH 3 10 10 3 k i .U rH 1 1 111 1 1 1 1 1 • 1 1 rH 4J c O CO O O • 41 «i u to o c o C 41 o 41 41 -H 10 Ol Oi JS iH -.H ifl e S s u •r-1 iV W 3 > 2 I I 3 n W 41 iW 4J £ C CO O -H OJ --! O £ 1 • 111 1 • •' 1 1 1 1 1 00 00 i •O 3 41 41 C Oi 10 c n O 00 e u a 10 10 3 -H 3 f0 JG 10 rH X 4J 41 CO U U 10 10 10 k i 41 -u A c o Io •i e u o to <y c « oi Q, -H tn 41 10 u 3 -c e o w i S 11 111 1 1 1 - • 1 1 1 0 0 4J S C 41 096 S>M 3 fe iM 10 0 1 -" O ^. U ^ 0 (U -U u o iJ -C O u CO 4J & 41 O a> O X K 3 k i Q, OJ TJ EH CO •* — rH rH 00 XT 1^ Q'O 1'llll'l -H 4J fi 4J iM Q 10 o "1 O ki 41 1 CU O CO C 3 10 to U O -H 1110 ~ . * tJ S oi a< ia o 3 O 4J S § -" c -H rH 3 0 S o j4 >O ^ 41 -H & 10 CO o cn ch <j\ o in ^y oo oo r- ^^ to c c . io e -H -H g 5 fM Jr. J3 C H 10 > 1 iH O OoH O o O o 3 O V -o 6 C 3 S 41 -H Ol rH JJ -U CT1 -H CO C X -H U MH CO 3 01 -H m a c Li iM e -H oi 10 U a Oj 10 41 41 •8 C > O -i-! W 41 10 -O Ol > * -H -H -U <y c k i O ^-t O o O v£OQOoOO 5 -H CO CO Mil .. s *^ u c 4J 10 U M + >i CTi .u o 41 J3 41 O -Q 41 O C 10 4J -x -* S •a . i« o 10 41 ^ C u -—* 1--t'-- ^ -^ ^ crv 00 0 •O rH -P 41 ^ o av av i --^ ^ -r> -H « O -U £ *H -* 01 3 >° ^* 4-* "i« ^^ ^^ VO CTv O to .^^ ^-* ^-^ 0O *"•* ^* &'13 I S3 UH «{ <yv r-»r*r- -^'.^j3-oveo rH sll^l^^l^^i .M jjj .H -U " ° § CO C 4J rH m 0 10 -H 41 rH 01 C rH in oj .H u x > io c o W 3 O •-" CO U u £ OC£oO5 4i0«"'§.H£-H'O -H -H — 1 w o 10 O 10 ^H -H 9 10 a. nil £ B rH 4J k i fQ -o 41 O 4» T3 3 « ia > «{ 4J 10l iw I0J3I CO 5 10 k i -r» k i k i rH Q O O fl -H -H -H

3l8 the strength of the relationship between variables, i.e., by model "tuning" or "curve fitting." The factors considered by the studies included in Table 5.6 include C02, volcanic (stratospheric) and tropospheric aerosol loading, and various suggested measures of the variation in solar radiation reaching the top of the atmosphere. The results indicate that, depending on investigator, each of these variables can separately account for any- where from nearly all to none of the observed climatic variations of the last l00 years. In addition, when the coupled effects of more than one factor are considered, there remains a significant range in the ratios. Closer examination of a few of these studies shows more clearly the uncertainties and discrepancies that are involved, although identifying the exact causes of these disagreements is beyond the scope of this report. In comparing the various studies, consideration should be given to how well their choice of causal factors is able to explain the major features of the climate of the last hundred years as indicated by almost all available data sets, including the warmth of the northern hemisphere from l920 to l950 and the relatively steady warming of the southern hemisphere. 5.2.3.l Carbon Dioxide as a Causal Factor If we consider every influence other than CO2 that affects the climate to be part of the natural climatic variability, we can attempt to identify the CO2 signal amid the natural fluctuations, or "noise," of the climate record. The level of noise that can, to first order, be estimated by examination of climatic records before C02 is thought to have had a substantial climatic effect. Past C02 concentrations have been frequently estimated by extrapolating back from current concentra- tions assuming that inputs from fossil fuel combustion have been the dominant source and that the fraction of each year's emission remaining airborne has been constant. If fossil fuels have been the primary factor perturbing atmospheric C02 concentrations, then the period prior to perhaps l930 can be used as a baseline. In this case, a reasonably long record is available, and the variability of annual average temperatures may be estimated with some reliability to be a few tenths of a degree. If, however, the biosphere has been an important net source of atmospheric C02 over the last l50 years, or parts thereof, then the C02-induced trend in temperature may not now allow selection of an adequate baseline from which to estimate natural climatic variability. Hansen et al. (l98l) calculated the effect of a CO2 increase on the global climate of the last l00 years, using a baseline CO2 con- centration of 293 ppm (which might be too high in l880 if biospheric sources have been important) and assuming that a doubling of C02 concentrations would lead to an equilibrium warming of 2.8°C. As shown in Figure 5.8, their model-computed response to the CO2 increase alone did not compare well with the global record. Interestingly,

3l9 0.4 0.2 -0.2 -0.4 Model (C02) —___ Observations 1880 1900 1920 1940 1960 1980 YEAR FIGURE 5.8 Comparison of observed global mean temperature anomalies of the last l00 years with anomalies predicted by a one-dimensional climate model assuming only that C02 concentrations are varying (Hansen, l980). however, although their result is in poor agreement with the northern hemisphere temperature record, it is in rather good agreement with the low-latitude and southern hemisphere temperature records (refer to Figure 5.4). These differences point to the need for conducting future analyses in more than one dimension; it may be, for example, that because of fewer volcanoes in the southern hemisphere and because of the thermal inertia of southern hemisphere oceans, the C02 warming may more easily be found in the southern rather than northern hemisphere. Madden and Ramanathan (l980) also attempted to identify the CO2 climatic effect, assuming that all other factors contributed to the climatic noise. They looked in the 50-60° N latitude band where the spatial coverage of the records is quite good and where equilibrium climate models predict that the temperature changes should be largest. Their results were negative (i.e., they found no statistically sig- nificant signal emerging from the noise), indicating, they suggested, that either the attempts to estimate present temperature changes from equilibrium climate models are inadequate or that the models are over- estimating the C02 warming. Wigley and Jones (l98l) also did not find evidence of a C02 signal when examining records of the northern

320 hemisphere surface temperature; they also did not attempt to reduce the noise by considering the possible climatic effects of other factors. The failure of these and earlier efforts to identify unequivocally a CO2 signal in the noisy global temperature record suggests that attempts should be made to take into account other causal factors in order to reduce the residual variance, and thus to make a hypothesized CO2 signal stand out more clearly. 5.2.3.2 Volcanic Aerosol as a Causal Factor Several authors have considered how to relate changes in volcanic injections of stratospheric aerosol to climatic fluctuations of the last l00 years. Model results of Oliver (l976), Robock (l978), and Bryson (l980) all suggest good agreement between observed temperature anomalies and volcanic forcing, finding no requirement for consideration of C02 or solar effects. The model and data used by Hansen et al. (l98l), however, found less satisfactory agreement (see Figure 5.9). The primary difficulty in achieving good agreement with just volcanic forcing is in explaining the temperature changes in the northern hemisphere, in particular the warming from the l920s to l930s and cooling from the l940s to l970s. The strength of these features varies depending on the temperature record used and the treatment of temper- atures over land and over ocean. Hansen et al. (l98l) indicate that the anomalous warm period in the northern hemisphere was about 0.4°C above a trend line through the rest of the record, whereas in Jones et al. (l982) and Budyko (l969) the anomaly is slightly less. Although most authors use the northern hemisphere temperature record as a basis for evaluation of the relationship between causal factors and climate changes, Hansen et al. (l98l) used a global record that somewhat reduces the intensity of the l920-l950 warming period that must be explained because of averaging in the rather steady warming trend in the southern hemisphere. Bryson (l980) believes the problem that most investigators have in explaining the cooling that occurred prior to the Agung eruption of l963 arises for two reasons. The first is the incompleteness of Lamb's (l970) volcanic record; in response to this criticism several volcanic eruptions in the l950s have been added to most volcanic chronologies. The second reason is the inadequacy of estimates of the radiative effect of volcanic injections by the traditional methods of estimating the amount of dust injected. The recent El Chichon volcano and analyses of the effect of the Agung eruption have emphasized that the injection of sulfur-bearing gases rather than total dust is likely the more appro- priate measure of the ultimate radiative effects. Although probably also affected by tropospheric aerosols, Bryson and Dittberner (l976) and Budyko (l969) urge use of actinometric data as a better measure of stratospheric turbidity. It is also difficult to reconcile the relative effects on average temperature in the northern and southern hemisphere based on volcanic records, especially during the l900-l9l0 period when (according to the temperature reconstruction of Hansen et al., l98l) the supposedly less

32l (T - T )1883 - 0.4, R = 1.0, C = 6.5, a = 0.162 Bezymyannaya and Agung at 10 1880 1960 1880 1900 1920 1940 1960 1980 YEAR FIGURE 5.9 Comparison of observed surface-air-temperature anomalies of the last l00 years with anomalies predicted by models that included only changes in stratospheric aerosol loading as calculated by (top) Oliver (l976) and (bottom) Hansen (l980). Oliver compares his results with the northern hemisphere temperature record and Hansen to the global temperature record. responsive southern hemisphere cooled more and prior to the northern hemisphere, even though listings of major volcanic activity (e.g., Mitchell, l970) indicate that all important eruptions were in the northern hemisphere. While the case for coupling between volcanic activity and climate is suggestive, the data bases and analysis methods still need much work in order to be able to account accurately for the climatic effects of volcanic injections with confidence. Analysis of the climatic fluctua- tions subsequent to the El Chichon eruption will be an important aspect of this effort.

322 5.2.3.3 Solar Variations as a Causal Factor The identification of cycles in various climatic data bases having the same periodicity as such indices of solar activity as sunspots has led to extensive efforts to attribute various climatic fluctuations to solar variations (Eddy, l977; Geophysics Study Committee, l982). The absence until very recently of both physical measurements to corroborate the value of the surrogate solar indices and of detailed explanations of the cause-effect relationships are important caveats to remember. Studies of sunspot-climate relationships have a long history. Of particular interest in recent years has been the apparent absence of sunspots (the Maunder minimum) noticed during the coldest periods in Europe in the l600s. In detailed analyses, however, Mass and Schneider (l977) and Robock (l978) found little statistical significance in the in-phase relationship between Wolf sunspot number and temperature. This has led to consideration of other surrogate indices of solar activity. Robock (l978) considered the suggestions of Eddy (l976) that the solar output is a function of the alternate 80- and l00-year cycles (the Gleissberg cycle) that seem to modulate the amplitude of the sunspot cycle. Robock did not find any statistically significant agreement using this index. Although Mitchell (l96l) had found some agreement with the observed temperature record up to l940 if he considered temperature changes due to both C(>2 and time-averaged sunspot number, these factors could not explain the later cooling that was observed. Eddy et al. (l982) suggest that the fractional areal coverage of the solar disk by sunspots, rather than actual sunspot number, is the appropriate index of solar variability. Hoyt (l979a,b) finds apparently good agreement between the northern hemisphere temperature record and the umbral/penumbral ratio, a possible measure of the convective energy transport in the sun's photosphere, and therefore of its time-varying radiant flux. The range of his results was calibrated to agree with the temperature record; note that he found that umbral/penumbral ratio and C02 forcing, and not vol- canic forcing, were sufficient to reproduce the variations in the observed temperature record. It would seem that, if valid, the solar effects should be similar in temperature records of both hemispheres except for the influence of differing amounts of ocean; the differences in the temperature records that exist would, it would seem, cast some doubt on this explanation. Hansen et al. (l98l) find that including the umbral/penumbral ratio improves their fit when also considering C02 effects (compare Figure 5.l0, top, with Figure 5.8). Gilliland (l98l) uses the variation in solar radius as an alternative measure of solar irradiance. This cycle is about 76 years in length and is negatively correlated with the Gleissberg cycle. There are indications that the Greenland ice core exhibits cycles with a length about the same as the solar radius and Gleissberg sunspot cycles. Broecker (l975) suggested that the cooling associated with the 80 year Greenland ice core cycle was counteracting the warming expected from increasing C02, leading to only slight change in present temperatures.

323 0.4 0.2 0 -0.2 -0.4 0.5 T 0 -0.5 Model • • • Observations 1880 1900 1920 1940 YEAR 1960 1980 FIGURE 5.l0 Comparison of the observed change in surface air temperature and model predictions of the change in temperature when considering the increase of C02 concentrations and changing solar radiance. (Top) Hansen (l980) uses umbral/penumbral ratio as an indicator of solar radiance and compares with global temperature change. (Bottom) Gilliland (l982) uses changes in solar radius as an indicator of solar radiance and compares with northern hemisphere temperature change.

324 This suggestion of a balancing of opposing effects of different causal factors is also evident in Gilliland's (l982) analysis, in which there could be no accommodation of CO2 warming were it not for the postu- lation of solar-induced cooling in the last few decades (Figure 5.l0, bottom). Given the lack of direct observational support for the suggested relationships between the surrogate solar variables and solar radia- tion, it seems premature to make quantitative estimates of the CO2 effects on the basis of hypothesized reduction in solar radiance. At the same time, one must recognize that solar variations may mask the C02 effect. It is also interesting that there is a similarity in character between the various surrogate indicators of solar activity (solar radius variations, Gleissberg cycle, and smoothed umbral/ penumbral ratio) and the actinometric and smoothed dust veil indices that are assumed to represent volcanic activity (e.g., see Siquig and Hoyt, l980). This may, in part, explain why different investigators using different data bases have reached similar conclusions. 5.2.3.4 Combinations of Causal Factors Because single factors have difficulty in explaining the records of climatic fluctuations, a combination of factors has often been used. Some authors (Hansen et al., l98l; Vinnikov and Groisman, l98l, l982; Mitchell, l983; Gilliland, l982) find that volcano and CO2 effects lead to reasonable fits to the data (Figure 5.ll), although it should be remembered that some other authors do not require C02 to achieve an equally good fit, depending on how volcanic effects are included. Vinnikov and Groisman (l98l) also indicate, for example, that use of the stratospheric aerosol series of Lamb (l970) and Mitchell (l970) instead of that of Budyko (l969) leads to a qualitatively different conclusion that there is no significant influence of C02 on climate change over the last l00 years. Thus, even among those considering just volcanic and CO2 effects, there is disagreement on the role of each factor. Hansen et al. (l98l) and Gilliland (l982) have achieved what appear to be excellent fits to the temperature record by considering C02, volcanic injections, and solar variations (Figure 5.l2). While the similar conclusions of these two studies may appear to be re-enforcing (see C02/Climate Review Panel, l982), it is disturbing that their data bases and relationships are quite dissimilar and in some instances contradictory. (a) The maximum variation of Hansen et al.'s global temperature record is about 0.5°C, whereas Gilliland's northern hemisphere temperature record has a range of about 0.9°C. (b) Hansen et al.'s volcanic record (from Lamb) has major peaks for Krakatoa (l883), Soufriere/Santa Maria (l902-l904), and Agung (l963). Gilliland's volcanic record appears to have major peaks for Askja (l875), unidentified (l885-l886), Katmai (l9l2), and Surtsey (l963-l965), most of which are high-latitude or local volcanoes.

325 0.4 0.2 o °ET 0 -0.2 -0.4 — — Observations Model l I 1880 1900 1920 1940 YEAR 1960 1980 0.5 O ° 0 -0.5 l I I I ! —— Observed (Borsenkova et al.) Fitted (Volcano + 1880 1890 1900 1910 1920 1930 1940 1950 YEAR 1960 1970 1980 FIGURE 5.ll Comparison of the observed change in surface air temperature and model predictions of the change in temperature when considering the increase of CO2 concentrations and changing stratospheric aerosol loading. (Top) Vinnikov and Groisman (l982) compare their results with data for the entire northern hemisphere from Vinnikov and Groisman (l98l); (bottom) Mitchell (l983) compares with data north of l7.5° N. (c) Hansen et al. show a CO2 effect beginning in the l880s, whereas Gilliland's CO2 effect does not start until l925. (The initial concentration of CO2 assumed by each of these investigators is also 5-l0% above recent estimates of CO2 concentrations in the last century.) (d) Hansen et al. use Hoyt's quite variable umbral/penumbral record as a measure of changes in solar radiance. Gilliland depends strongly on a smooth solar radius cycle of about 76 years and also includes solar variations having cycles of l2.4 and 22 years; phase and amplitude of the solar cycles were arbitrarily determined to provide the best fit with the temperature record.

326 0.4 0.2 ~ 0 o -0.2 -0.4 -0.6 - Model: sun + CO_ + volcanoes I I I . I 0.5 o -0.5 —— Model * • • Observations I I 1880 1900 1920 1940 YEAR 1960 1980 FIGURE 5.l2 Comparison of the observed change in surface air temperature and model predictions of the change in temperature when considering the increase of C02 concentrations and changes in solar irradiance and stratospheric aerosol loading. (Top) Hansen (l980); (bottom) Gilliland (l982).

327 These and other similar studies (Hansen et al., l98l; Vinnikov and Groisman, l982; Gilliland, l982; for example) acknowledge uncertainties in their presentation of data and their formulation of conclusions. They have been very helpful in raising questions, suggesting relation- ships, and identifying gaps in our data bases and observational approach, and we cannot preclude the chance that at least one may be correctly relating causal factors and temperature changes. However, contrasting causal components of the climate change and differences in data bases make it difficult to accept the results as reinforcement of the general hypothesis of the C02-induced climate shift. 5.2.4 Steps for Building Confidence Are results to date sufficient basis for declaring that a climate change due to increased CO2 has already occurred? Given the potential impor- tance of such a finding, we must require a high standard of agreement in attributing climatic effects to causal factors. Methodologies for determining the statistical significance of sug- gested climatic changes have recently been discussed by Madden and Ramanathan (l980), Wigley and Jones (l98l), Klein (l982), Epstein (l982), Hayashi (l982), Murphy and Katz (l982), Katz (l980, l98l), and an international group (World Meteorological Organization, l982a). Epstein, for example, decomposes the time series of observations of a climatic variable into three components: a "natural" climatic mean, a possible climatic change induced by some extrinsic factor such as CO2, and a random variability. The natural mean may be taken to be constant, or it may include the estimated influences of some forcing factors, e.g., solar or volcanic activity. Various hypotheses regarding the climatic change may now be stated and tested. The conclusions to date have depended crucially on the assumptions made regarding the underlying climatic trend, the increase of C02 between the mid-nineteenth century and recent times, and the expected CO2 influence. Once the hypothetical C02 signal is prescribed on the basis of estimated CO2 concentrations and model simulations, the remaining variance of the climatic record is partitioned between the fluctuations explained by other external factors and the unexplained fluctuations deemed to be "noise." Reduction in the unexplained variance relative to the variance explained by C02 is taken as supporting the hypothesis of C02 influence. It is difficult to draw unambiguous conclusions from studies of this type for the following reasons: l. The natural variability of global mean temperature is imper- fectly known because of the relatively short period of instrumental record. Moreover, the spectrum of observed climate variability exhibits considerable power at low frequencies. In other words, natural climate variations have been observed on time scales commensurate with the time scale of CO2 increase. At least in part, these may be attributable to external factors whose influence might, in principle, be removed. However, an unknown and possibly large amount of natural variability or

328 noise might well remain at low frequencies, making it difficult indeed to distinguish over short periods of record between slow trends induced by increasing C02 and equally slow trends that reflect the natural low-frequency variability of the climate system. For this reason, the hypothesis that the observed variability is entirely due to natural causes cannot be unequivocally rejected on the basis of the studies conducted so far. 2. The global mean temperature record has been reconstructed from a relatively short and geographically limited set of observations, pri- marily over land areas of the northern hemisphere. The period of record spans a period of great change and expansion in the human societies that make and record the observations. While investigators have conscien- tiously worked to remove or correct human influences such as urbaniza- tion on, for example, temperature records, it is difficult to assess their degree of success. The accuracy and representativeness of the data are thus open to question, although the broad trends are believed to be reliable. 3. Solar radiance variations have been estimated from surrogate observations, e.g., sunspots, umbra/penumbra ratios, and solar radius measurements, whose relationship to solar irradiance is by no means clear. Moreover, the relationship between solar variability and climate has been determined only empirically from a limited data base. 4. Atmospheric turbidity has been estimated from data on volcanic activity (e.g., individual eruptions or the "Dust Veil Index," or acidity measurements in ice cores) or from actinometric observations. None can be considered very reliable. As with solar variations, turbidity influences have been only empirically fitted to a limited record. 5. The history of atmospheric CO2 concentrations before l958 is poorly known, with estimates of mid-nineteenth-century concentrations ranging from 250 to 290 ppm. The expected temperature change is thus correspondingly uncertain. 6. Other neglected factors, e.g., surface albedo, aerosols, and—in recent years—anthropogenic trace gases, may have influenced the temperature record. 7. The choice of the appropriate signal of CO2-induced climate change is by no means clear; many results from models of different degrees of physical completeness are available (Schlesinger, l982, l983). Moreover, model results are available only for equilibrium conditions, while the real world is presumably exhibiting a transient response to increasing C02. The transient response will be greatly complicated by the thermal inertia of the ocean and the distribution of ocean and land on the Earth's surface. Thus, real-world transient responses might be quite different in their regional details from those that might be inferred from equilibrium models (see, Thompson and Schneider, l979; Schneider and Thompson, l98l; and Bryan et al., l982). 8. Such studies are prone to a familiar pitfall of statistical inference, namely, the testing of multiple hypotheses. As noted by Epstein (l982), "If enough different hypotheses are examined, then, by chance, it is likely that statistics supporting one of them will be found."

329 9. The "noise" accounted for by both external factors omitted from the analysis and unexplained sources is not independent from one time of observation to another, nor can it be reasonably modeled as a first- order Markov or autoregressive process. (If it could be so modeled, then our best prediction of next year's climate would only involve the present single year's data corrected for changes in the included external factors; in fact, however, averages over many preceding years prove to be the best climatological estimates.) Thus, statistical techniques that are more careful and sophisticated than those for white noise or first-order autoregression will be required to deal with the increase of fluctuation energy at lower frequencies. There is a methodological point that should be made here with regard to claims to have detected a CO2-induced warming based on a "success- ful" model fit to the climatological record (e.g., Hansen et al., l98l) as opposed to the simpler approach of identifying a warming signal rising above the "noise" of intrinsic climate variability (e.g., Madden and Ramanathan, l980). Tests of statistical significance are required in both approaches. However, these tests have usually been made only in the latter, partly because significance tests for model fits are not generally to be found in textbooks and, indeed, must be developed separately for each model type. Until such tests have been devised and carefully applied, the scientific community will remain skeptical of claims to have detected with statistical confidence a CO2-induced signal in a single parameter such as temperature. Notwithstanding methodological difficulties, earlier detection may be sought in two ways. First, we may attempt to improve the objectiv- ity and physical basis of the estimates of external influences on the past climatic record, thus constraining the range of plausible hypoth- eses that could be tested. Needs are the following: l. Better determination of the natural variability of temperature, particularly at low frequencies by extending the period of record back in time through use of proxy records and by distinguishing between ocean and land records. 2. Improvement in the accuracy and representativeness of the tem- perature record through incorporation of marine data and continued attention to influences such as urbanization. 3. Better data bases on possible changes in solar output and atmospheric aerosol loading. 4. Better reconstruction of the changes in CO2 concentration over the last hundred years. 5. Objective, physically based, and observationally validated relationships between solar variability, volcanic aerosols, other possible factors and climate. A second recently suggested approach is to attempt to isolate a pattern of changes specifically attributable to increasing CO2 concentrations (MacCracken and Moses, l982). This approach is discussed in the following sections of this chapter.

330 5.3 A STRATEGY FOR MONITORING C02-INDUCED CLIMATE CHANGE 5.3.l The "Fingerprinting" Concept Proposals for monitoring programs to detect the effects of increasing C02 date back to the SCEP (l970) and SMIC (l972) reports. Recently, participants at a Department of Energy sponsored workshop on First Detection of Carbon Dioxide Effects (Moses and MacCracken, l982) proposed a three-part framework for detection of CO2 effects involving: l. Identification of changes, 2. Identification of possible causative factors, 3. Isolation of the parts of the changes attributable to increasing C02. With respect to the last of these, a suggestion was made "... to develop a unique C02-specific 'fingerprint' for the C02 response involving a set of several parameters, distinctive from responses that would be caused by all other known influences, and to search for this correlated pattern of changes, not just for a change in one isolated parameter." (MacCracken and Moses, l982.) The concept of fingerprinting is based on the notion that a composite index based on multivariate statistical analysis of several parameters in space and time might enable us to attribute more positively climatic changes to increased CO2. Indeed, changes in some climatic elements might help to distinguish the effects of CO2 (or CO2 in combination with other radiatively active gases) from those due to variations of some of the other factors or external conditions that could influence climate. For example, one might anticipate that the pattern of tropospheric temperature changes caused by turbidity variations due to volcanic aerosols would differ from those caused by the more globally uniform variations of CO2. While the notion of an index that would unequivocally reveal the influence of CO2 on climate is indeed enticing, its practical application does not appear immediately feasible. The relationships between atmospheric C02 and climate variables must be deduced from model simulations. Simulations of the equilibrium response to highly elevated levels of C02 show considerable scatter in results (see Schlesinger, l982, l983), and simulations of the transient response to slowly changing CO2 have not yet been accomplished. Even if a plausible index could be deduced from models, its statistical char- acteristics in the appropriate frequency range would need to be assessed through study of past data and through model simulations. Appropriate multivariate statistical tests would then have to be designed and applied. These difficulties must be overcome before a scientifically rigorous monitoring and detection strategy based on this approach can be devised and applied to provide clear guidance to policymakers.

33l Nevertheless, the concept of a "CO2 fingerprint" provides useful guidance in the design of a geophysical monitoring program that will both provide data for research and help us to follow the course of climate. We thus suggest that monitoring programs designed to shed light on the effects of increasing CO2 should focus on the climatic variables whose patterns of change in space and time are indicated by models to respond most strongly to CO2 increases, together with those external factors that may also influence climate—particularly those external factors with greatest influence on the selected climatic vari- ables. The remainder of this chapter presents some specific suggestions based on this approach. 5.3.2 Considerations in Climate Monitoring 5.3.2.l Statistical Variability and Expectations of Change All climatic parameters are highly variable in space and time. Their variability or "noise" arises both from systematic physical processes whose effect could in principle be calculated and from the random fluctuations of a turbulent atmosphere. The characteristics of the variability, including its preferential occurrence in one frequency range or another, can in some cases be determined from observed data, especially for climatic parameters at the Earth's surface such as air temperature and precipitation. Direct estimates of variability may in other cases have to be supplemented from the data simulated by compre- hensive climate models (Manabe and Hahn, l98l). Estimation of the climatic change expected to result from increased CO2 depends on climate models that address the processes and scales of interest. Estimation of the changes to be expected from increased CO2 should therefore include their seasonal and geographical charac- teristics, their longer-term and arealy averaged properties, and their evolution in time as CO2 concentrations increase. Some preliminary information of this sort is already available; extended simulations are required to provide more complete information. A schedule of expected climatic changes is especially necessary. Most experiments that have been made to date have estimated the response of climate from the difference between two simulated equilibrium climates of a model employing normal and twice the normal concentrations of atmospheric CO2. From the results of such experiments the climatic changes expected at other (and generally lower) levels of CO2 are then found by simple interpolation, usually logarithmic. This approach overlooks the facts that all the elements of the climate system do not interact either at the same rate or in the same places, especially insofar as they involve the oceans, and that the CO2 concentration will not "wait" at any level for the climate to reach an equilibrium. In this connection, model experiments treating an exponential growth of CO2, and hence a linear growth of its effects, may be particularly helpful. The information necessary for a realistic schedule of changes, including their seasonal and geographical distribution, in a variety of climatic parameters that are expected to occur as the CO2 concentra-

332 tion reaches progressively higher levels, can be provided only by comprehensive general circulation model (GCM) simulations; such studies are important in the design of monitoring strategies. 5.3.2.2 Initial Selection of Parameters A program for monitoring and detection of C02 effects can be developed on the basis of our expectations of the effects of increasing C02, estimates of the accompanying climatic noise, and a schedule of the expected responses to increased CO2. Existing knowledge is adequate to formulate an initial monitoring strategy that can be revised as our understanding improves. The first step should be to identify the climatic parameters whose responses to increased CO2, individually or in combinations, are likely to be significant. Atmospheric model studies suggest that likely candidates are the tropospheric and surface air temperature (which should rise), sea temperature (which should rise), stratospheric temperature (which should fall), and atmospheric water vapor or specific humidity (which should increase). In addition, the downward flux of infrared radiation at the surface should increase, while at the top of the atmosphere the spectral distribution of out- going infrared radiation should shift with respect to the primary CO2 emission bands. In the ocean and cryosphere, the amount of snow and ice in polar regions should decrease over the long term and the global sea level should rise. In addition to large-scale responses, GCM climate simulations are expected to indicate the geographical and seasonal characteristics of such changes, and their evolution over time, all of which may provide additional indications. For example, the expected polar warming may occur principally in winter because of the increase of heat conduction through thinner sea ice, while soil moisture in midlatitudes may be reduced in summer (cf., Chapter 3). Other changes whose characteristics are yet to be definitively shown by climate model simulations may occur in the meridional circulation and hydrologic cycle. The statistical significance of changes in individual variables can be judged against appropriate measures of their natural variability. Assessment of the significance of changes in some composite index could be judged similarly. 5.3.2.3 Revision and Application of a Monitoring Strategy A multiple-effects monitoring strategy for C02-induced climatic change can be implemented even though all the information required for its rigorous design is not available. Using a combination of model simulations and observational data, at least partial information on the expected changes and levels of variability can be assembled. As further information becomes available from both new observations and new model simulations, additional variables can be included and a revised schedule or timetable of the effects expected over future years can be prepared.

333 The interpretation of the climatic record should also include an assessment of the possible effects of changes in other external factors in the climate system, apart from CO2 and apart from internally gen- erated noise. Such possibly competing factors include the large-scale injection of aerosols into the atmosphere by volcanic eruptions and possible variations of the solar constant on both interannual and decadal time scales. As with CO2, our best hope for a knowledge of the climatic effects of these events rests with the use of a hierarchy of models, including comprehensive GCMs. Some experiments have sug- gested, for example, that the annual average (equilibrium) tropospheric warming resulting from an increased solar constant resembles in some respects that found with increased CO2, although the transient response might differ in some respects (Schneider and Thompson, l98l). Since such factors have been used in conjunction with changes in CO2 in simple climate models to simulate the historical variations of the northern hemispheric surface temperature (Hansen et al., l98l), it is important that simulations be made with comprehensive GCMs in which the significance of any climatic changes due to changes in solar radiation can be determined and, if possible, differentiated from those due to other factors. This information will help us to interpret monitored climate records in terms of signals due to variations in incoming radiation, aerosol loading, or greenhouse gas concentrations and residual changes that may contain a signal attributable to increased CO2. 5.3.3 Candidate Parameters for Monitoring Previous sections have discussed the problems of identifying CO2- induced warming from present data and showed that there are weaknesses in the interpretations, caused by the inadequacy of the existing data sets, the methods of analysis used, and the underlying theory. Data employed, except in recent times, have rarely been collected with long-term monitoring of CO2 effects in mind, so that development of a coherent and integrated monitoring strategy becomes essential at this time. Monitoring of CO2 effects is a challenging task if the number of parameters to be monitored is to be kept low, in order to reduce costs, and if at the same time unambiguous cause-effect relationships are to be established. The following criteria were chosen to assess the suitability of a number of parameters for long-term monitoring: l. Sensitivity. How does the effect exerted on climate by the variable or the changes experienced by the variable on decadal time scales compare with that associated with corresponding changes in CO2? 2. Response characteristics. Are changes likely to be rapid enough to be detectable in a few decades? 3. Signal-to-noise ratio. How large are the relevant changes in relation to the variability due to measurement errors and causes not accounted for?

334 TABLE 5.7 Primary Parameters for Monitoring the Causes and Effects of Climate Change A. Primary parameters for monitoring the causes of climate change Carbon dioxide Stratospheric aerosols Solar radiance Other "greenhouse" gases and ozone Nitrous oxide (N2O) Methane (CH4) Chlorocarbons (CFCl3, CF2 Cl2, for example) Stratospheric ozone Tropospheric ozone B. Primary parameters for monitoring the effects of climate change Atmosphere Global temperature Mean surface air temperature Tropospheric temperature distribution Stratospheric temperature distribution Radiation Upward terrestrial and reflected solar radiation at the top of the atmosphere Cloud and water vapor Precipitable water content of the atmosphere Equivalent emission temperature (cloudiness) Cryosphere Sea ice cover Snow cover Ice cap mass balance changes Oceans Sea level Sea temperature 4. Past data base. Are data on the past behavior of the variable adequate for determining both a base level and its natural variability? 5. Spatial coverage and resolution of required measurements. 6. Required frequency of measurements. 7. Feasibility of technical systems. Can we make the required measurements? Cost is another important consideration, but meaningful assessments of technical feasibility and estimates of costs were beyond the capabilities of the present panel. As has been pointed out in other sections of this report, the detection of climate change is largely a signal-to-noise ratio problem, and it is therefore important to compile comprehensive data sets and to apply the best statistical techniques available to the data. The technical feasibility of making the measure-

335 merits and the associated cost, however, will put restrictions on any monitoring strategy that can be put into effect. However, a number of primary parameters for monitoring have been tentatively selected (Table 5.7), based on the above criteria. They are organized under the head- ings of "causes" of climatic change and "effects" of CO2. In the latter category, parameters to be monitored have been grouped under the headings of atmosphere, cryosphere, and oceans. This is not a comprehensive list of candidate parameters but includes only those that have high aggregate ratings in the seven criteria listed above. 5.3.3.l Causal Factors External factors that have been suggested as influencing the climate have time scales, i.e., periodicities or e-folding times, ranging from less than l to more than l0 years. Focus on the climatic effects of anthropogenic CO2 emissions allows us to limit our attention to those that have time scales of one to several hundred years. Within this time range, the external factors that may influence the climate include the composition of the atmosphere, volcanic activity, land surface modification, and solar variations. Table 5.8 lists trace gases and aerosols that affect atmospheric composition and that are influenced by man's activities. Table 5.9 lists the trace gases with potentially important radiative, climatic, or chemical effects. We further limit our consideration to external factors that might induce hemispheric or global temperature changes of magnitude 0.l K or greater over about a century and to a lesser extent according to feasibility of monitoring. According to these criteria, monitoring of the following factors is particularly important: l. Carbon dioxide 2. Solar radiance 3. Stratospheric aerosols 4. Other greenhouse gases* and ozone (a) Nitrous oxide (b) Methane (c) Chlorocarbons (d) Stratospheric ozone (e) Tropospheric ozone Although ozone is not strictly an external factor, its chemistry is affected by anthropogenic emissions of a number of gases (e.g., nitro- gen oxides and Chlorocarbons); by monitoring ozone directly, we avoid dependence on knowledge of the details of the chemistry. Nonvolcanic sulfur (OCS and SO2, for example) and volcanic emissions both contribute to stratospheric aerosol. Several other potential external influences have not been included at this stage. The possible effects of modifications in surface albedo, *This list will no doubt be revised as research continues. See Machta (this volume, Chapter 4, Section 4.3).

336 TABLE 5.8 Principal Anthropogenic Sources of Trace Gases and Aerosols Anthropogenic Source Comments Gas C02 00 Hydrocarbons C2H4, for example Chlorocarbons CH, N20 NO, NO2 Sulfur Compounds OCS, CS2 S02 Ozone Aerosols Fossil fuel combustion Internal combustion engines Internal combustion engines Refrigerants, solvents, propeHants Internal combustion engines, industry, change in land use Combustion, fertilizer manufacture Internal combustion engines, aircraft Fossil fuel conversion Possibly large biospheric component Those of concern entirely man-made Large component from biological activity Large natural component from biological activity High-flying aircraft are an upper-tropospheric and lower-stratospheric source Volcanoes are an intermittent source of sulfur Combustion Anthropogenic contribution is from chemical reaction of other trace gases Sulfate Silicate or car bon-conta in ing Conversion from S02 and other sulfur- bearing compounds Combustion, soil erosion Most important for stratospheric aerosols Diesel engines especially, closely tied to land use roughness, and surface thermal characteristics by desertification, urbanization, and extension of agriculture (e.g., Charney, l975; Sagan et al., l979; Potter et al., l975, l980) are not included because effects of such changes appear to be primarily regional and because of the difficulty of developing a data base. Tropospheric aerosols are not included because there are large regional variations in aerosol amount, composition, and characteristics, so that an accurate monitor- ing network would require many stations as well as detailed analysis of the samples. These neglected effects contribute, of course, to the

337 41 co W • *~- 4j • «•» ~ 14 c c to Qt Oi Oi 41 o o u O o O 4» g & •H 4J N to t0 o s i 4J kl c 0 41 .*J 4J •H CO Ul 41 41 41 41 41 oi 41 41 41 41 0i 41 S * CO CO CO CO CO (O UJ CO CO t0 10 41 41 a CO 10 41 S y t? e 41 oi to to to io to ro to « 1 s! 41 41 41 oi 41 41 41 41 M 1 c ° O y u y u y U u y u (W y 41 41 C C C 01 C as rH T3 41 CO Q Q H M M Q M M M 1 a> 10 41 S 41 y 3 10 jM c u UH c u i! O M 3 (0l CQ 41 n UH c 3 t, •H 4j o •oi 441 y •rH 4J 41 o1 41 41 41 41 0 41 41 UH co 4j 41 2 CO m 10 io CO CO CO CO t0 to (0 to en CO H ' 8 9 UH u (U ot 41 41 41 41 0 S JJ b] 41 J0 g JC 41 Qj >H kl kl kl kl kl 01 .H rH UH c ° y y y y c c c c O u 4J kl rH 10 H U EH M L-H] % to U M HUM M u 4j y •H u r^ •H 41 • UH 4^ JJ 41 0 CO JC UH y 3 a u 41 41 ki UH ,O •a y UH e 0 c •rH 41 •H 10 *J CO i0 rH ' to 41 0| B 10 0 0 ^ •H y hi •rH in rH -H e o • 4J H U E « C CO 10 C — «««-<w~ »*^* MfUHW^ 41 •H rH •H O O X 41 JS u O 13 •H .1 °l 0 j= y CO CO 41 4J ^* •* •» u ~.r- j f» c o a y io ^ II ^ rH • >n C1 rH n * * * 1 1 1 'o 5 41 0 O u t0 kl r^ 2 ° 00 . . rH | • • 1 ,o •H 41 4j UH 4J g * rH rH rH O O O 0 0 O O 0 0 Q| 3 4J t0 c kl Pi ro rH 1 rH rH rH rH rH 41 UH 10 10 3 41 p, X XX 1 (N ^^ ki y 4J y •— M XXX in k i O •._*; 4J -H kl W c ,HO X I 3 4J . C k1 o1 o in • « O CO O co oi rH fll 41 41 a O • • o o m H V? *•* 10 k i 41 3 10 H, U JC E u rH T* in 0 rH E co 3 c a M to 541 O CO t0 41 -H * 0 4J ?•! g • > 0 UH 10 rH CO « , S, "•* £ ,__| t0 10 13 41 c 4J 10 -~ s ^^ 41 CO U 10 5 B> • 41 CO 4J u c - 3 UH 41 o -P 3 •u £ oi 4J O 4J • 4j r* ,H y y t0 o •rl 4J • fe Q 10 io E 10 (0 ki C 4J •H m kl O Z o! ki u 3 41 UH |.2 M rH 4J kl II rH U M 4J — 41 rH .c cn ki 5 CO (•) 2, £8 4J (0 3 1-1 a TN T*l y ,3. x h 33 4J 1*^ ^D3 O U U 41 O O U Z ^r 41 01 10 S 9,0 ffi CO CN CN o 4J 41 c y .5 41 C CO u u u Z U «M O W O O O W ac1 o» o O u o> o y UH y N •rl 41 41 10 0 10 O ** ft 2 UH E co UH «0 rrj k i O 10 41 V4 41 u E oi 3 V I o C 41 O C 41 4I 3 j= -P u £ CO UH 41 N O *o *o CO 4J c o y 01 41 UH 41 41 4J ON co O 41 0 •H -H QJ C C CO rH UH UH T3 rH rH -H 10 (U TJ •o •H -H c c 10 3 -O CO y -o 0l -H 3 3 X 0 u 0 3 „ 41 CO ki 41 H •H U -H kl -H X 38 3 "a 41 41 CO CO O 10 ki y u Ot •^ -2 o1 u o g § J3 Q rH 'O 'O 2 co n § c o £ 41 JS r> •H CO •H CO in Chemical Caubon d Chlouoca U) Oi 41 41 Q Q C Hi 4J O «> O Nituous Cauboe in Hyduocau Sulfuu C Caubon Caubon Sulfuu S tuatosp wateu • JC C 2. Range il Assume inuuease £ Could 1 Effect o y a 41 41 9 co a x: u k l 0 p 4J £g 4J kl 41 .H W EH S Z

338 unexplained variance in the climatic record. Clearly, continuing research on the possible influences on climate and on the means for their measurement is essential. The following subsections provide more detailed discussion for those external parameters listed above and present monitoring requirements. In addition to recommending improved monitoring of some of the factors, in several cases it will be essential to continue the records of surrogate indicators that have been developed in order to be able to calibrate these methods and extend the data bases back in time. 5.3.3.l.l Carbon Dioxide Sensitivity. Numerical model studies indicate that doubling atmospheric carbon dioxide concentrations would increase global average temperatures (ATd) by 3 + l.5°C (Climate Research Board, l979; CO2/Climate Review Panel, l982) . Changes in radiative flux are approximately logarithmically related to CO2 concentration. Although water vapor and albedo feedback mechanisms are a function of global mean temperature, for a moderate warming, temperature change (AT) is nearly linearly proportional to radiative flux, so that where ATd is the expected temperature change for doubling, and [CO2]Q and [CO2] are the concentrations for the base period and the current period, respectively. Rate of Change. Since l958, the annual increase in the atmospheric concentration of CO2 has ranged between 0.6 and 2.2 ppmv (Machta, this volume, Chapter 3, Section 3.4). The concentration in l982 was slightly above 340 ppmv. The concentration before the industrial revolution probably was in the range 250-290 ppmv. Releases of CO2 to the atmosphere result from both fossil fuel emissions and changing land use. The mean growth rate of fossil fuel CO2 production over the period l86l-l980 was 3.5% per year (Elliott, l982). In l980 total CO2 emissions from fossil fuels are estimated to have been about 5.2 x l09 tons of carbon (Marland and Rotty, l982) . Uncertainty asso- ciated with calculations of fossil fuel CO2 emissions is estimated to be l0-l3.5%. CO2 emissions from cement manufacture may add an additional 2% to the estimate of fossil fuel emissions. The history of biospheric emissions is poorly known and controversial. Global net release of carbon to the atmosphere from the biosphere since l860 has been estimated to be as high as l80 x l09 tons of carbon. The release from biospheric sources in l980 was given as l.8 x l09 tons of carbon according to the FAO Production Yearbook, the most comprehensive and apparently detailed data source, and has been estimated to be as high as 4.7 x l0 tons of carbon according to other sources (see Woodwell, this volume, Chapter 3, Section 3.3) . Regrowth and stimulation of photosynthesis from increased atmospheric CO2 may have compensated for some of these emissions (see Machta, this volume, Chapter 3,

339 Section 3.4). Nordhaus and Yohe (this volume, Chapter 2, Section 2.l) project that the atmospheric CO2 concentration will most likely increase by an average of 0.3% per year up to the year 2000, reaching a concentration at that time of 367 ppmv. Signal-to-Noise Ratio. The CO2 concentration in a single sample can be measured to within 0.l ppm, but at most stations in the global net- work the accuracy is about l ppm. The seasonal cycle and annual increase of atmospheric C02 can be easily detected at remote locations where effects of local biospheric, anthropogenic, and geologic sources are avoidable. The atmosphere, averaged on a zonal and annual basis, appears to be within 4-6 ppm of being well mixed, and gradients of this magnitude can be explained by longer times for transport from oceanic or biologic sources and sinks (see Machta, this volume, Chapter 3, Section 3.4). Adequacy and Availability of Data Base. Continuous, accurate observa- tions of C02 began at Mauna Loa Observatory in l958. Additional research and measurements are needed to reconstruct past concentrations. Before l958, most measurements are of uncertain accuracy and may not be representative of the global average concentration. For purposes of detection of climate change, the present network provides adequate measures of global average concentrations, and data are readily available. Efforts are under way to estimate past C02 concentrations by, for example, analysis of solar spectral measurements, carbon isotopic ratios in tree rings, reconstructed pCO2 in former oceanic surface waters, and CO2 concentration in air bubbles trapped in ice. Efforts to reconcile the results of these various approaches, including the effects of local station bias and different temporal characteristics, are needed. Spatial Coverage and Resolution of Additional Measurements Required. The present global network provides adequate coverage and resolution of CO2 concentration for purposes of detection of climatic changes. Frequency of Measurements Required. For purposes of studying the climatic effect, the record of monthly average C02 concentration now being made is sufficient. Climate model studies of the climatic effects of increasing C02 concentrations are most unlikely to use more frequent data than annual average concentration. Feasibility and/or Existence of Technical Systems; Continuity. Mea- surements made by the present global network of surface stations are satisfactory for determining the trend of C02 concentrations. Although satellite measurement of atmospheric CO2 would provide improved global coverage, it is not evident that sufficient accuracy could be achieved to determine the interannual variations in the trend or in the range of the seasonal cycle.

340 5.3.3.l.2 Stratospheric Aerosols Sensitivity. Stratospheric aerosols affect climate through their net effects on the components of the Earth's radiation budget. The deli- cate balance that exists between the solar and thermal components of the radiation budget can be altered by the range of possible strato- spheric aerosol (or cirrus cloud) scattering and absorption properties such that either a net cooling or warming of the lower atmosphere may result. Radiative transfer calculations (e.g., Pollack et al., l976b) indicate that the sign and magnitude of the temperature change depend critically on the composition (refractive indices), size distribution, scattering phase function, and optical depth of the stratospheric aerosols. Surface-temperature changes associated with potential radiative perturbations of climate were calculated by various research groups. For example, Hansen et al. (l98l), using a simplified one- dimensional model, calculated that a persistent increase in strato- spheric (sulfuric acid) aerosol optical depth of 0.2 (representative of a large volcanic event) would produce a cooling of l.9°C in the Earth's surface temperature. In reality, however, effects over land and sea might be quite different. Natural volcanic eruptions throughout history have long been thought to constitute the primary source of aerosols in the stratosphere through the direct injection of large quantities of gases and ash. Although the Mount St. Helens eruption was powerful and injected large quantities of material into the stratosphere (~0.5 km ), its climatic impact was insignificant (Robock, l98l, l983). Climatic effects of volcanic eruptions, in general, depend on the chemical composition of the aerosols injected into the stratosphere and, in particular, on the amount of sulfur contained. The climate effects are not a simple function of the volume of solid material ejected or the magnitude of eruption. Rate of Change. Several empirical and statistical studies offer a history of volcanic impact on climate. A discussion of climatological evidence from past volcanoes is given in the recent report of the NASA Workshop on the Mount St. Helens Eruption of l980 (Newell and Deepak, l962). Lamb (l970) proposed an empirical dust veil index (DVI) in correlating atmospheric optical depth perturbations of the atmosphere with volcanic eruptions over the past century. Thus, the historical record of volcanic eruptions must account for composition factors in assessing climatic effects. This holds for future extrapolations. There is no strong basis for projecting aerosol perturbations or change of composition. No clear trend or pattern emerges from the historical record. Signal-to-Noise Ratio. Satellite measurements of peak background (pre-eruption) stratospheric infrared extinction (l.0 pm) are typically on the order of l0-4 km"l (McCormick et al., l98l). The corresponding optical depth between the tropopause and 30-km altitude is approximately l0"-*. Immediately following the Mount St. Helens eruption, local extinction values as high as l0"l km"l (optical

34l depths near l) were observed, although the average stratospheric optical depth was typically l0-3. Volcanic eruptions, which are believed to have had an impact on climate, have produced stratospheric optical depths of about T ~ l0"l for several years following the eruptions. Adequacy and Availability of Data Base. Volcanic eruptions have been related to climatic changes since a suggestion by Benjamin Franklin in l784 (cf. Pollack et al., l976a). Extended inventories of volcanic eruptions have been prepared by Lamb (l970) and Simkin et al. (l98l). Pollack et al. (l976a) have assembled and reviewed data on transmis- sivity. Actinometric data, which may also include the effects of tropospheric aerosols, have been assembled by Pivovarova (l968, l977) and Bryson and Goodman (l980). Hammer (l977) suggests that acidity in a Greenland ice core may be a measure of northern hemisphere volcanic activity. In recent years lidar data have become available, but time histories exist over only limited parts of the globe for about l0 years. Observations of stratospheric aerosol extinction have been made by the Stratospheric Aerosol Measurement II (SAM II) instrument aboard the Nimbus-7 research satellite (since October l978) and by the Strato- spheric Aerosol and Gas Experiment (SAGE) aboard the Applications Explorer Mission 2 (AEM-2) (February l979-March l982). A follow-on SAGE mission, SAGE II, is planned for launch in l984 in conjunction with the Earth Radiation Budget Experiment (ERBE). Aerosol information may also be inferred from operational Advanced Very High Resolution Radiometer (AVHRR) data from National Oceanic and Atmospheric Adminis- tration (NOAA) satellites. Stratospheric aerosol data products are routinely archived at the National Space Science Data Center (NSSDC) at the National Aeronautics and Space Administration's (NASA) Goddard Space Flight Center, Greenbelt, Maryland. Spatial Coverage and Resolution of Additional Measurements Required. Future high-latitude coverage beyond SAM II will be needed for inves- tigating the importance of polar stratospheric clouds (PSCs) on climate. SAGE covered the geographical regions between 72° N and 72° S latitude, while SAM II observes both polar regions. Planned SAGE II measurements will provide continuing aerosol extinction measurements at low and midlatitudes through the mid-l980s. No corresponding measurements are planned for the polar regions. Remote measurements must be accompanied by in situ measurements of aerosol radiation properties and size dis- tribution, as part of a coordinated ground-truth program. For example, lidars and aircraft flights are needed to calibrate the satellite mea- surements. In particular, additional lidars are needed in the southern hemisphere. Comprehensive data, including latitudinal variation, are needed in aerosol impact model studies. Frequency of Measurement Required. Measurements are needed to establish the monthly variation of stratospheric aerosols. Current and planned satellite observational programs are believed to be adequate for establishing the seasonal and interannual variability

342 of the present background stratosphere. Provision must be made to respond to major volcanic eruptions as they occur and to continue observations beyond the lifetime of current programs. Feasibility and/or Existence of Technical Systems; Continuity. Existing satellite systems supplemented by a few key lidar systems for calibra- tion purposes are adequate. 5.3.3.l.3 Solar Radiance Sensitivity. Solar radiation is the fundamental energy source that drives the motions of the atmosphere and oceans. The associated storage and transport of heat and mass then establish the Earth's temperature and climate on regional and global scales. Temporal variations in the sun's output have long been thought to be a major cause of recorded climate change. Although a great number of researchers have sought clear evidence in linking solar variations to specific weather and climate responses, the results have thus far been inconclusive. A recent report by the Geophysics Study Committee (l982) concludes that the role of the sun in producing global circulation, climatic zones, seasonal changes, and the recurrence of periods of glaciation is well recognized, but intrinsic solar variability is neither implied nor required to account for these phenomena. The simple changes in insolation responsible for these effects are predictable, far in advance, on the basis of known parameters of the orbit, figure, and motions of the Earth. Although numerous mechanisms exist that can cause climate change, it is important, nevertheless, to understand to what extent the basic solar forcing of the climate system is changing. Sensitivity studies reported by Hansen et al. (l98l) suggest a 0.5°C global average surface temperature rise associated with a hypothetical 0.3% increase in the solar constant. Systematic changes of only 0.5% per century could explain the entire range of past climate from tropical to ice-age conditions (Eddy, l977). Rate of Change. The integral of the broad spectrum of solar electro- magnetic radiation incident on the Earth has historically been termed the solar constant, as a consequence of the lack of evidence to the contrary. Until recently, detailed information on short-term natural variability in the solar constant was restricted mainly by the lack of adequate instrumentation and by the variability in atmospheric trans- mission, which limits the accuracy of surface-based observations of the sun. Since late l978, independent accurate measurements of solar radiance by two NASA research satellites have confirmed that the solar constant is indeed variable. Both the Solar Maximum Mission (SMM) and Nimbus 7 spacecraft carry advanced active cavity radiometers that overcome the basic deficiencies in prior measurements. SMM results reported by Willson et al. (l98l) yield a mean solar radiance at l AU of 1368 W m-2 with an absolute uncertainty of less than +0.5%. Several observed large decreases (dips) in radiance of up to 0.2%, lasting about l week,

343 are highly correlated with the development of sunspot groups whose below-average temperatures reduce the total output of solar radiation. Solar faculae, with above-average temperatures and, therefore, radia- tive output higher than normal, cause a much smaller variability appearing as a radiative excess. The facular effects correlate with irradiance peaks before and after some observed dips. A measured 0.05% per year downward trend in the 4-year record may be related to the general pattern of solar activity variation over an ll-year cycle. Signal-to-Noise Ratio. The solar radiance monitor carried aboard the SMM spacecraft consists of a combination of three independent elec- trically self-calibrated cavity pyrheliometers having nearly uniform wavelength sensitivity from the far-ultraviolet through the far- infrared regions. This sensor combination has provided daily obser- vations of the sun with an estimated measurement precision of less than 0.005% since launch. Such precision should be sufficient to detect solar variability of a magnitude able to affect climate. Adequacy and Availability of Data Base. Routine satellite measurements of solar radiance have been made since October l978. This measurement capability, if sustained, is believed adequate for meeting current climate requirements for broadband solar radiance monitoring. Spatial Coverage and Resolution of Additional Measurements Required. Current measurements should be continued to provide a data base encompassing at least one ll-year solar cycle. Frequency of Measurements Required. The wide range of natural variability found in the current data set supports the need for daily measurements over the next few years. Beyond then, less frequent (monthly?) measurements may be adequate. Feasibility and/or Existence of Technical Systems; Continuity. Although current technology can meet the scientific requirements for broadband solar radiance monitoring, no future satellite missions carrying these instruments have yet been approved. The proposed Upper Atmosphere Research Satellite (UARS) would have this capability. 5.3.3.l.4 Nitrous Oxide (N2O) Sensitivity. Donner and Ramanathan (l980) and Wang et al. (l976) calculated a surface warming of 0.3 K and 0.4 K, respectively, for a doubling of the N2O concentration. A doubling of ^o concentra- tions would also decrease the total ozone column by as much as l5% (World Meteorological Organization, l98l) through the catalytic reactions of N2O-produced nitrogen oxides with ozone; the change in O3 concentration would also have climatic effects. Rate of Change. Current tropospheric N2O concentrations are approxi- mately 300 ppbv. N2O levels appear to be approximately 0.8 ppbv higher in the northern hemisphere than in the southern hemisphere.

344 Atmospheric measurements (Weiss, l98l) indicate that tropospheric concentrations of nitrous oxide (N2O) have been increasing at a rate of approximately 0.2% per year since the first measurements were made in l963. Weiss suggests that a substantial fraction of this increase may be explained by combustion of fossil fuels. Nitrogen fertilizers may also be an important source of N2O. It would be desirable to find means for determining the pre-industrial N2O concentrations in the atmosphere. The N2O concentration decreases with altitude in the stratosphere. The primary sink for N2O is photodissociation in the stratosphere. The primary source of nitrogen oxides in the present stratosphere is the reaction of N2O with excited oxygen atoms. Signal-to-Noise Ratio. The N2O concentration at the surface can now be measured with a relative precision of less than 0.5% (Weiss, l98l) and to an absolute accuracy of 3%. Because of its long lifetime, N2O in the troposphere is well mixed, to within 5 ppbv, i.e., l.7% of its mean (World Meteorological Organization, l98l). This range is explain- able in terms of dynamics and variations in strong local sources and sinks (Levy et al., l979). Adequacy and Availability of Data Base. Measurements of N2O exist since l96l, with more extensive measurements beginning in l976. Several groups have undertaken programs to monitor tropospheric N2O (Weiss, l98l; Pierotti and Rasmussen, l977, l978). Weiss (l98l) reports measurements at three NOAA Geophysical Monitoring for Climatic Change (GMCC) sites (Mauna Loa, Point Barrow, and South Pole) plus ship data. Spatial Coverage and Resolution of Additional Measurements Required. The surface sources and sinks of N2<> are still not well understood. Additional research is needed, and continued surface monitoring should be supported. Frequency of Measurements Required. A monthly record at several sites in each hemisphere is desirable. Feasibility and/or Existence of Technical Systems; Continuity. Current ground-based observations should continue in order to monitor global concentrations, and satellite measurements of N2O are likely to become more available within the next decade. The latter may be better able to determine tropospheric and stratospheric variations in N2O. 5.3.3.l.5 Methane (CH4) Sensitivity. For a doubling of the present CH4 concentrations, Wang et al. (l976) estimated a surface warming of 0.2-0.4 K, depending on model assumptions. Dormer and Ramanathan (l980) and Lacis et al. (l98l) calculated a warming in surface temperature of 0.3 K for the same change in CH4 concentrations. Changes in methane concentration may also influence the global ozone distribution due to the reactivity of methane with hydroxyl radicals and other trace gases.

345 Rate of Change. Major sources of methane (CH4) are believed to be mining and production of fossil fuels; anaerobic fermentation of organic material due to microbial action in rice paddies, swamps and marshes, tropical rain forests, and tundra; and enteric fermentation in mammals and the activity of termites. The primary atmospheric sink is by reac- tion with hydroxyl radical (OH) in the troposphere and stratosphere. Methane degradation is an important source of atmospheric carbon monoxide. Water vapor and odd hydrogen species (OH, HO2) are impor- tant products of CH4 oxidation in the stratosphere. The atmospheric lifetime of methane is approximately 7 years. Current tropospheric concentrations are about l.7 ppmv. Methane concentrations have increased l-2% per year since careful observations intended to determine presence of a trend were started in l978 (Rasmussen and Khalil, l98l; see Machta, this volume, Chapter 4, Section 4.3). Signal-to-Noise Ratio. At this time it is difficult to assess the extent to which the recorded increase represents a short-term fluc- tuation in the methane cycle or a long-term trend. Historical studies of methane concentrations are desirable. Methane can be measured with a precision of 0.0l ppmv and an accuracy within a few percent. CH4 is reasonably well mixed in the troposphere but decreases in concentration with altitude in the stratosphere owing to reaction with OH and other radical species. More CH4 is found in the northern hemisphere than in the southern hemisphere. Adequacy and Availability of Data Base. Regular measurements of CH4 have been made for only a few years. Spatial Coverage and Resolution of Additional Measurements Required. A few sites in each hemisphere should be adequate, if well chosen. The existing programs may be sufficient, although additional measurements may be needed for a better understanding of surface sources and sinks of CH4 and variations with altitude. Frequency of Measurements Required. A monthly record at several sites in each hemisphere is desirable. Feasibility and/or Existence of Technical Systems; Continuity. Ground- based measurements may need to be expanded to improve monitoring of globally averaged tropospheric concentrations. Satellite measurements may become available within a decade. 5.3.3.l.6 Chlorocarbons (e.g., CFCl3, CF2Cl2) Sensitivity. A number of chlorocarbons and chlorofluorocarbons have strong IR bands (Ramanathan, l975; Wang et al., l976). Ramanathan (l975) estimated that increasing the concentrations of both CFCl-» and CF2Cl2 to 2 ppbv could raise the surface temperature by 0.9 K. For the same chlorocarbon abundances, Lacis et al. (l98l) calculated a surface temperature change of 0.65 K. Other chlorocarbons with known absorption features in the same region are CCl4, CHCl3, CH2Cl2,

346 and CH3Cl. Chlorocarbon emissions are also of concern because they dissociate in the stratosphere, and the resulting Cl0X species might significantly affect stratospheric ozone concentration through ozone- destroying catalytic reactions. Rate of Change. Large quantities of these industrially produced chemi- cals are made for a variety of uses, such as solvents, refrigerants, and spray-can propellants. The chlorocarbons of primary current concern are CFCl3 (also referred to as CFC-ll) and CFoCl2 (CFC-l2) because of large amounts produced, long atmospheric lifetimes, and increasing stratospheric and tropospheric concentrations. Concentrations of these two chlorocarbons continue to increase by approximately l0% per year, although rates of emission have not increased significantly since l976. The lifetime of CFCl3 and CF2Cl2 are approximately 60 years and l00 years, respectively. These species are well mixed in the tropo- sphere. Once in the stratosphere, these species eventually photodis- sociate or sometimes react with excited oxygen atoms. Some of the other chlorocarbons react with hydroxyl in the troposphere. These species have much shorter tropospheric lifetimes. Signal-to-Noise Ratio. CFCl3 and CF2Cl2 are well mixed in the troposphere with mixing ratios of approximately l90 and 320 pptv, respectively. Estimated precision for CFCl3 and CF2Cl2 measurements are 2-5% (World Meteorological Organization, l98l). Ten percent is the estimated accuracy for measurements of CFCl3 and CF2Cl2, with the limit of detection being about l pptv. Approxi- mately l0% larger concentrations of CFCl3 and CF2Cl2 exist in the northern hemisphere than in the southern hemisphere. Errors are larger for other chlorocarbons. Adequacy and Availability of Data Base. The present programs sponsored by the federal government and by the Chemical Manufacturers Association, if continued, are probably adequate to monitor the increase in the most important of these species. Additional data would be useful for such other chlorocarbons as CH3CCl3, CFCl2, CF2Cl, and CCl4. Spatial Coverage and Resolution of Additional Measurements Required. Because radiatively important chlorocarbons are long-lived species and relatively well mixed in the troposphere, measurements at only several sites in each hemisphere are required. This can be accomplished by assuring that the record from sites now sponsored for research purposes (or similarly located sites) is continued for monitoring purposes. Frequency of Measurements Required. A monthly record of the con- centration of a number of chlorocarbons at several sites in each hemisphere is desirable. Because of the accuracy required to detect changes in chlorocarbons with time, additional sites would have to be chosen carefully, so that long records can be developed. Feasibility and/or Existence of Technical Systems Required; Continuity. Present ground-based measurements are probably adequate for determina-

347 tion of global concentrations. Satellite measurements may come available within the next decade. 5.3.3.l.7 Stratospheric Ozone Sensitivity. Stratospheric ozone can affect climate through its influence on dynamic and radiative coupling mechanisms between the stratosphere and troposphere. The absorption of solar radiation by stratospheric ozone is primarily responsible for the increase in stratospheric temperatures with altitude and is thus linked to the dynamics of the stratosphere. Bates (l977) and Geller and Alpert (l980) examined the effect on tropospheric planetary waves from changes in the zonal mean wind and temperature structure in the stratosphere. Bates (l977) found that dramatic changes in the northward flux of sensible heat resulted from changes in stratospheric structure. Geller and Alpert (l980) found that the stratospheric structure had to be altered below about 35 km before any significant changes in the struc- ture of tropospheric planetary waves resulted. Many uncertainties remain in determining the effect of changes in stratospheric ozone on tropospheric dynamics. However, the modeling studies described above suggest that planetary wave coupling may link tropospheric weather and climate to changes in the stratosphere. A reduction in stratospheric ozone can modify the surface tem- perature through two competing radiative processes (solar and long wave). With less ozone, more solar radiation is transmitted through the stratosphere, thereby enhancing solar heating of the troposphere and Earth's surface. On the other hand, the reduced absorption of solar radiation in the stratosphere cools the stratosphere, thereby reducing long-wave emission from the stratosphere downward into the troposphere by all long-wave emitting species. The changes in solar and long-wave fluxes have opposing effects on surface temperature. Ramanathan et al. (l976) calculated a surface cooling of 0.l K for a l0% uniform reduction in stratospheric ozone. However, Ramanathan (l980) has shown that changes in the vertical ozone distribution can have a significant effect on surface temperature, even if the total ozone column does not change. Since the change in ozone concentration with altitude differs, depending on the source of this perturbation (e.g., CFCs, NOX), the change in surface temperature may differ depending on the perturbation even if the change in total ozone is the same (irrespective of the radiative effect of the perturbing species). Rate of Change. Several analyses have been made of measurements of total ozone from 37 Dobson stations. The studies by Bloomfield et al. (l98l), St. John et al. (l98l), and Reinsel (l98l) give the following 95% confidence intervals for global increase in total ozone in the l0-year period l970-l979: Bloomfield et al. (l.7 + 2.0)% St. John et al. (l.l + l.2)% Reinsel (0.49 + l.3)%

348 Total ozone variations differed significantly with geographical location of the instrument. Total ozone increased at some stations and decreased at others. Satellite measurements of ozone concentration in the upper strato- sphere between l970 and l979 indicated a reduction of up to 0.46%/yr at altitudes between 33 and 43 km (Heath and Schlesinger, l982). The satellite data are not consistent concerning whether ozone increased or decreased at higher altitudes. Recent analyses of surface-based Umkehr data are consistent with a decrease in upper stratospheric ozone of 0.3-0.4%/yr during the l970s (Reinsel et al., l983). Signal-to-Noise Ratio. There is a great deal of variation in the ozone record. Ozone concentrations fluctuate on a variety of spatial and temporal scales owing to natural causes; these fluctuations tend to mask possible systematic changes due to manmade perturbations. Observed ozone changes averaged over the northern hemisphere and the world suggest that a time interval of approximately l0 years is the shortest period that is meaningful for calculating ozone trends. Ozone varies with the ll-year solar cycle, and there is also a strong quasi-biennial oscillation (World Meteorological Organization, l98l) that affects trend analyses for shorter averaging times. Some variation is attributed to differences in instruments, in instrument operators, or in adjustment/calibration procedures. Those Dobson stations with the longest records (several decades) are gen- erally considered to be the most reliable. Unfortunately, most of the stations began operating after l957. The noise in the data contributes to the uncertainty in the trend analyses, which is considerable (as indicated above). Satellite records are relatively short, with the first ozone measurements beginning in mid-l970. Adequacy and Availability of Data Base. An extensive global network of stations measuring total ozone exists. However, most of the observing stations are on continents and in the northern hemisphere. Satellite measurements providing more uniform spatial sampling could eventually lead to better trend measurements. Also, since only one instrument is used, the present intercalibration errors between surface devices are eliminated. Recent modeling studies (e.g., Wuebbles et al., l983) suggest that total ozone measurements may not be a sensitive indicator of the impact of human activities on the global atmosphere. These studies suggest that monitoring of changes in the distribution of ozone with altitude should be more useful. However, major uncertainties exist in both the existing Umkehr network and satellite measurements. Only about l8 stations currently make regular Umkehr measurements; only 3 of these stations are in the southern hemisphere. In addition to calibration drift, the Umkehr method is also subject to errors from the effect of dust and aerosols and the inability to make measurement under cloudy conditions. As with the total ozone measurement, the satellite data are limited by the short duration of continuous and homogeneous global data coverage.

349 The ozone data are available from several central sources. The Dobson data are compiled and checked prior to public release by the Center for Ozone Data for the World located in Toronto, Ontario, Canada. The data are usually available about a year after they are taken by the Dobson stations. NASA satellite data are available from the National Space Science Data Center at the NASA Goddard Space Flight Center, Greenbelt, Maryland. Data from the NOAA satellites are not available from this center and must be obtained directly from NOAA. It usually takes more than a year for the satellite data to be processed, reviewed, and released to the Data Center. Spatial Coverage and Resolution of Additional Measurements Required. The ozone data based on Dobson and Umkehr stations have limited geo- graphic coverage. Most stations are located in continental areas at middle latitudes in the northern hemisphere. Consequently, it is difficult to estimate global total ozone when there are vast areas of the world (mostly in the southern hemisphere) where measurements are not taken. The existing total ozone and ozone distribution networks need to be enlarged, particularly by adding stations in the southern hemisphere. The spatial coverage on ozone from satellites is very good, but the satellite data are limited in temporal coverage. Global satellite measurements of ozone began in l970, and the longest single record of data is 7 years. Although a continuous ozone record is available owing to overlap of various satellite systems, there is a significant bias between systems that makes ozone trend analysis using satellite data difficult. Frequency of Measurement Required. Because of the large temporal and spatial variations of ozone, frequent measurements are needed. With a limited surface network, daily measurements of ozone should be taken (and usually are). The best stations take several measurements each day to monitor diurnal variations and to reduce the noise in the measurements. Satellite systems monitor stratospheric ozone continuously with the objective of covering the entire global area at least once each day. Solar UV instruments usually are in polar orbits and take measurements at the same local time at each geographic location. IR instruments are capable of taking day and night measurements. Because of the operating cycle of satellite systems, a satellite usually has gaps in its temporal and spatial coverage. Daily measurements are adequate for obtaining monthly-mean distribution of ozone. Feasibility and/or Existence of Technical Systems; Continuity. The Dobson instruments are considered to be excellent in quality and are often used as a reference for comparison with satellite measurements. NOAA, in cooperation with Environmental Protection Agency and the Chemical Manufacturers Association, is currently developing a new Umkehr network with improved instrumentation and extensive intercali- bration program. Improvements are being made in satellite instrumen-

350 tation, so the data have improved in spatial resolution and in consistency with other ozone data. Future NOAA satellites will carry operational ozone-measuring instruments. 5.3.3.l.8 Tropospheric Ozone Sensitivity. Although only 5-l0% of the total column amount of ozone is in the troposphere, a uniform percentage change in tropospheric ozone can have about the same effect on surface temperature as the same percentage change in stratospheric ozone. This results from the fact that owing to pressure broadening of the lines in the 9.6-um band of ozone, the total long-wave opacity of tropospheric ozone is nearly the same as that of stratospheric ozone. The solar effect of a change in tropospheric ozone is different than that of stratospheric ozone in that both the solar and long-wave effects are in the same direction in this case. An increase in tropospheric ozone increases solar absorption plus enhances the long-wave "greenhouse" effect of ozone; both effects tend to increase surface temperature. Fishman et al. (l979) calculated that a doubling of tropospheric ozone concentrations would lead to an increase in surface temperature of 0.9 K. Rate of Change. An analysis of ozonesonde data (World Meteorological Organization, l98l) shows that tropospheric ozone in the layer from 2 to 8 km increased in northern middle latitudes by about 7% during the last decade. An increase has also been observed in ozone in the layer between 8 and l6 km, but the magnitude is less than that for the 2- to 8-km layer. The cause of this increase is not known, but it is con- sistent with model calculations that predict an increase in upper tropospheric ozone due to increased NOX emissions from aircraft engines. An increase in tropospheric ozone is not observed at all latitudes nor at all stations in middle latitudes. Some model calcu- lations also suggest large photochemical production and destruction rates for ozone in the troposphere, implying night-day shifts in concentrations. Signal-to-Noise Ratio. The annual-mean values of the ozonesonde measurements in the 2- to 8-km layer have an uncertainty of +2-3%. Ozonesonde measurements in the troposphere have higher relative accuracy than in the stratosphere. As the ozonesonde reaches high altitudes, pump correction factors have to be applied to the measurements to account for reductions in pump efficiency. Adequacy and Availability of Data Base. Both Umkehr and ozonesonde data bases are limited in spatial coverage. Although Umkehr measure- ments are made of tropospheric ozone, the Umkehr technique is more accurate for ozone in the upper stratosphere. While ozonesonde measurements have high vertical resolution, there are only a small number (<20) of stations. Only one ozonesonde station is in the southern hemisphere. Many ozone measurements are taken near the ground (in the boundary layer) in urban areas to monitor pollution. To study

35l climatic effects, measurements ate needed of the ozone distribution through the troposphere. Spatial Coverage and Resolution of Additional Measurements Required. The ozonesonde data give detailed information about the vertical profiles of ozone through the troposphere. Because these balloon measurements are expensive, the number of launches from each station per year is limited. Budget cutbacks have led to a reduction in the number of operating stations. Since a long data record is needed for trend analysis, keeping or restoring the ozonesonde network to its peak level is important. Although adding additional stations would be desir- able, maintaining what has been operational is a reasonable objective during this time of limited resources. The Umkehr network provides an independent check on the ozonesonde measurements, and vice versa. Because Umkehr measurements are made with ground-based instruments, there is a lower cost per measurement, so the frequency of measurements is greater than that of the ozonesonde network. Frequency of Measurement Required. Monthly mean profiles of ozone are needed for trend analysis. Consequently, daily measurements are desirable. This is feasible for the Umkehr network, but costs prohibit ozonesonde measurement at this frequency. At least one ozonesonde measurement per week would be desirable at each location. Feasibility and/or Existence of Technical Systems; Continuity. The spatial and temporal coverage of the existing ozonesonde network is limited, and the network should be maintained and expanded, if pos- sible. New, highly automated, ground-based ozone monitoring instru- ments have been designed, allowing greater frequency of measurements and less dependence on .operator skill. Because the Dobson instruments are well established, there will be some reluctance to replace them with the new instruments until the quality and reliability of the instrument is fully demonstrated and adequate overlap calibrations are available and have been carefully studied. 5.3.3.2 Atmospheric Parameters The atmosphere is the part of the climate system with the least thermal inertia and, therefore, the part that can respond most quickly to the radiative effects of increasing CO2 concentrations. Climate models have focused most attention on the changes that increasing CO2 concentrations will induce in the atmosphere, and it is therefore quite appropriate to look first to the atmosphere for any evidence of CO2-induced climate changes. Since the greenhouse effect on the radiation balance of the atmo- sphere affects temperature directly, a first change to look for is in the temperature field. Global and hemispheric mean surface tempera- tures have been studied extensively in this regard—regional tempera- tures cannot be considered as sufficiently representative. Arctic and

352 Antarctic temperatures may respond more than the global mean, but unfortunately the variability of polar temperatures from year to year is also larger, so it is unclear whether the signal-to-noise ratio will be less there (Kelley and Jones, l98l). Similarly, at middle and high latitudes wintertime surface temperatures may show a slightly larger increase than summertime temperatures, but the variability of wintertime temperatures, or noise, is more than twice as large as that in summer. This suggests that summertime temperature response may have a larger signal-to-noise ratio. It will also be important to distinguish between temperatures over land and over the ocean. It has been argued that temperature averaged over the lower tropo- sphere, rather than surface temperature, may be more representative of changes in the radiation balance because of the greenhouse gases. Unfortunately, good upper-air coverage does not go back beyond about l950, and even now it is deficient over the oceans. So far, satellite indirect soundings have not been used to fill in the gaps in the climatological record. The situation is worse in the stratosphere where the global rocketsonde sounding system has been in operation for only a relatively short time and is on the point of being dissolved altogether. In view of the difference between tropospheric and strato- spheric temperature responses to increased CO2, both stratospheric and tropospheric temperatures should be monitored. Atmospheric fluxes of radiation should also be monitored. The monitoring of upward shortwave (reflected and scattered solar) radiation and emitted infrared radiation at the top of the atmosphere deserves special attention. Satellites are potentially powerful tools for monitoring these parameters. However, interpretation of most satellite measurements to the accuracy required to explain temperature changes of a few tenths of a degree appears difficult, as this requires that absolute changes in the radiation budget of the order of 0.2% over a period of years be detected. Detailed analysis of the spectrum of outgoing infrared radiation may, however, provide useful information. Clouds may also influence long-term climate variations through their radiative effects. To illustrate their potential importance, changes in global cloud cover of a few percent could mask the warming effect of doubled carbon dioxide. Unfortunately, it is difficult to determine with confidence the three-dimensional distribution of clouds because of their ill-defined boundaries and complicated configurations. Instead, one can monitor the equivalent blackbody temperature of the cloud ensembles by measuring their outgoing radiation in one or more atmospheric "windows." From this it should be possible to calculate not only the upward flux of thermal radiation but also the radiatively effective cloud height when the vertical distribution of temperature in the atmosphere is given. This, and knowledge of the planetary albedo, allows one to evaluate important aspects of the influence of the cloud cover on the radiation balance of the Earth-atmosphere system. Finally, the precipitable water content of the atmosphere will change with CO2 increases. Model calculations indicate that the atmospheric precipitable water will increase 5-l5% as the climate warms in response to doubled CO2 concentrations (Manabe and Stouffer, l980; Wetherald and Manabe, l98l). Observations of this parameter may, therefore, help

353 confirm whether the numerical models are properly simulating the role of water-vapor processes in contributing to climate change, although changes of precipitable water in the near future will, of course, be much less than for doubled C02. 5.3.3.2.l Global Mean Surface Air Temperature Sensitivity. When it is assumed that the C02 content of the atmo- sphere is doubled and statistical thermal equilibrium is achieved, the more realistic of climate modeling efforts predict a global surface warming of between 2 and 3.5°C, with greater increases at high lati- tudes; if one allows for more feedback mechanisms, a range of l.5 to 4.5°C is suggested (Climate Research Board, l979; CO^Climate Review Panel, l982). The results from climate models also suggest that the CO2-induced warming is approximately proportional to the increase of the logarithm of the C02 concentration in the atmosphere. Rate of Change. The large thermal inertia of the oceans will probably delay the response of the atmospheric temperature to an increase of the CO2 concentration in the atmosphere. Since the CO2-induced warming penetrates deeper into the ocean as time passes, the effective thermal inertia of the ocean depends on the time scale under consideration. Therefore, the length of the delay in the climatic response depends on the rate of the C02 increase and can change with time. By use of a simple model of the atmosphere-ocean system, Hoffert et al. (l980) recently estimated the temporal variation of the atmospheric tempera- ture in response to a predicted increase of the CO2 concentration. According to their results, the delay of response is approximately l0-20 years during the period of A.D. l980-2000. Transient responses may also vary zonally and regionally because of the ocean's thermal inertia (Schneider and Thompson, l98l; Thompson and Schneider, l982). Siqnal-to-Noise Ratio. It is expected that CO2-induced change of surface air temperature will be positive over most of the Earth's surface, whereas the trend of the natural temperature variation changes sign from one geographical location to another. Therefore, the signal-to-noise ratio of area-averaged surface air temperature should increase as the area for the averaging increases. For this reason, the global mean surface air temperature is one of the most promising quantities for the early detection of the C02 climate signal. According to the recent study of Jones et al. (l982), the standard deviation of the annually averaged area mean temperature over the entire northern hemisphere is about 0.86°C, while surface temperature anomalies for the northern hemisphere are typically +0.2-0.4°C (Clark et al., l982). Depending on assumptions of temperature sensitivity to increases in CO2 concentration and ocean thermal lag, an increase of atmospheric CO2 concentration to between 400 and 450 ppm should raise the global average surface air temperature distinctly beyond the expected natural fluctuations of surface temperature.

354 Adequacy and Availability of Past Data Base. Long, reliable, and representative temperature records are needed. Continuing analysis and estimates of error of the temperature record are called for; dissemina- tion of relevant records is generally good. Analyses of the temporal variation of the annually averaged global (or hemispheric) mean surface air temperature during the past l00 years have been made by many authors by use of instrumental data (see Section 5.2.2.l), with significantly differing results. Spatial Coverage and Resolution of Additional Measurements Required. Estimates of hemispheric (or global) mean surface air temperature are subject to large errors because of the sparseness of the observational network, particularly over the oceans. Improved observational coverage would not only improve the representativeness of hemispheric or global averages but would also permit monitoring of regional averages and differentiation between changes over ocean and over land. For example, model simulation (e.g., Manabe and Stouffer, l980) indicate larger C02-induced temperature changes in polar regions than in lower lati- tudes, and transient responses over the ocean may be expected to differ from those over land (Schneider and Thompson, l98l; Ramanathan, l98l). For these reasons, the observational network of surface air temperature measurements employed in monitoring and detection studies should be supplemented by determinations of sea-surface temperatures from satel- lite observations, despite the fact that sea-surface temperature is not identical with surface air temperature. Frequency of Measurements Required. One or preferably two observations per day will suffice. Feasibility. Satellite measurements are currently available. However, further improvement of the accuracy of sea-surface temperature deter- mination is required in order to improve estimates of surface air temperature. 5.3.3.2.2 Tropospheric Temperature Distribution Sensitivity. The C02-induced change of tropospheric temperature has been estimated by use of general circulation models of climate (e.g., Manabe and Wetherald, 1975; Manabe and Stouffer, l980). According to these studies, the warming of the surface layer of the atmosphere is expected to be particularly large in high latitudes (i.e., 2-3 times larger than the increase of global average temperature). The high- latitude warming decreases sharply with increasing altitude. In low latitudes, the CO2-induced warming is predicted to be somewhat smaller than the global average. The warming of the tropical troposphere will increase significantly with increasing altitude. Rate of Change. Since the troposphere is closely coupled to the Earth's surface through dry and moist convective transfer of heat, the response time of the tropospheric temperature is not very different from that of global surface air temperature discussed in the preceding

355 subsection. It is, however, probable that the response time of surface air temperature over continents is somewhat shorter than that over oceans because of the differences in effective thermal inertia between ocean and continent. The response time may also have latitudinal variation because the speed of the penetration of thermal anomaly into the ocean depends on latitude. Signal-to-Noise Ratio. The signal-to-noise ratio for the CO2-induced warming of zonal mean surface air temperature was recently estimated by Madden and Ramanathan (l980) and Wigley and Jones (l98l). The results from these studies suggest that the signal-to-noise ratio of the C02- induced warming of zonal mean surface air temperature is at a maximum in middle latitudes in summer when the magnitude of natural temperature fluctuation is relatively small. It is desirable to conduct similar analyses for the zonal mean temperature of the troposphere. Adequacy and Availability of Past Data Base. Tropospheric temperatures have been observed by radiosondes on a large scale since the Second World War. More recently, satellites have provided a dense coverage of remotely sensed soundings, although their accuracy is inadequate to determine trends. Angell and Korshover (l978a) estimated the temporal variation of the globally averaged temperature in the surface to l00-mbar layer for the period from l958 to l976. They used radiosonde data from 63 stations that are fairly evenly distributed. The recent compilation of upper-air data by Oort (l982) should also prove valuable for the determination of the past trend of tropospheric temperature. Spatial Coverage and Resolution of Additional Measurements. The cover- age over the oceans of radiosonde observations of atmospheric tempera- ture is sparse. To overcome this deficiency, the network of radiosonde observation should be augmented with satellite measurements of infrared or microwave radiances (to be converted to temperature through appro- priate inversion techniques). A spatial resolution of 250 km is desirable. Frequency of Measurements Required. Two observations per day are desirable. Feasibility and/or Existence of Technical Systems; Continuity. Currently, radiance data from satellites are routinely inverted to vertical temperature distributions of the atmosphere, which are used as input data for the daily numerical weather forecasts. These data might be used for the long-term monitoring of tropospheric temperature; careful assessment of the accuracy of current and proposed temperature determination from satellites should be made in the light of expected CO2-induced change of climate. 5.3.3.2.3 Stratospheric Temperature Distribution Sensitivity. Results from both radiative-convective models and general circulation models indicate that a cooling of the stratosphere could

356 result from an increase of the C02 concentration in the atmosphere. This cooling contrasts with the warming expected to occur in the tropo- sphere. Therefore, the simultaneous monitoring of temperature not only of the troposphere but also of the stratosphere should yield valuable information for distinguishing C02-induced climate change from climate change caused by other factors. According to Pels et al. (l980), cooling due to the doubling of the atmospheric C02 concentration would be about 7°C and ll°C at altitudes of 30 km and 45 km, respectively. Rate of Change. Since the thermal coupling between the stratosphere and troposphere is relatively weak, the response time of stratospheric temperature is not prolonged by the large thermal inertia of oceans as is the case with respect to the tropospheric temperature. According to results from time integration of a one-dimensional, radiative-convective model of the atmosphere, the response time (e-folding time) of the stratosphere is on the order of a few months (see, for example, Manabe and Strickler, l964). Signal-to-Noise Ratio. The standard deviation of natural temperature fluctuation in the stratosphere has a large seasonal variation, being small in summer and large in winter. In contrast, the CO2-induced change of stratospheric temperature estimated by model experiments has relatively small dependence on season. Therefore, it is expected that the signal-to-noise ratio of the C02-induced change of stratospheric temperature will be at a maximum in summer. In assessment of signal- to-noise ratio, it is necessary to know the standard deviations of the temperature fluctuation in the middle and upper stratosphere. Adequacy and Availability of Past Data Base. Recently, Angell and Korshover (l978b) estimated global temperature variation in the l00-30- mbar layer between l958 and l977. It is desirable to extend similar analyses to the l0-mbar level where the C02-induced cooling is expected to be substantial. In addition, the data from chopper radio- meter measurements from satellite and rocketsonde observations of temperature in the stratosphere and mesosphere should be analyzed in order to determine the recent trend of temperature in the middle atmosphere. Spatial Coverage and Resolution of Additional Measurements Required. It is recommended to monitor the future variation of temperature in the middle and upper stratosphere where the magnitude of the CO2-induced cooling is expected to be large. A spatial resolution of 500 km is desirable. This could be accomplished by observation of stratospheric temperature from satellite (e.g., by limb measurements), augmented by observations by radiosonde and rocketsonde. It is expected that these observations will yield a global distribution of stratospheric tempera- ture at various altitudes. Frequency of Measurements Required. One observation per day is adequate.

357 Feasibility and/or Existence of Technical Systems; Continuity. Current observation by radiosonde and satellite should be continued. In addi- tion, rocketsonde observation of stratospheric temperature should be continued at least at a few key locations in order to calibrate the temperature determined by inversion technique from radiance data obtained from satellites. Careful assessment of the accuracy of the determination of stratospheric temperature from satellites should be made. 5.3.3.2.4 Upward Terrestrial Radiation and Reflected Solar Radiation at the Top of the Atmosphere Sensitivity. Infrared emission to space and reflected solar energy from Earth depend primarily on clouds and surface characteristics. Models indicate that total precipitable water will increase, sea- surface temperature will increase, sea ice and snow cover will change, and stratospheric temperature will decrease as CO2 rises. Climate models cannot reliably tell us, at present, the impact of CO2 change on cloud amount, height, and type, although we know that key parameters for cloud formation, such as the water content of the atmosphere and the atmospheric lapse rate, will change. Rate of Change. Since most solar energy is absorbed in the oceans, the response time is linked to the thermal inertia of the oceans. Regional feedbacks, especially over continents, could be possible in terms of time periods as short as seasons or months. The secular rates of change of these two "top-of-the-atmosphere" parameters is likely to be very low. An annual cycle of the two radiation parameters has been observed on a global scale (Ellis et al., l978) and is under study. The CO2 impact on the amplitude and phase of these annual cycles should be studied as well. Signal-to-Noise Ratio. Natural variability of the emitted and reflected radiation on seasonal time scales has been estimated by Stephens et al. (l98l). In addition to the annual amplitudes, the tropical latitudes have a semiannual wave forced by Earth-sun geometry, and certain regions have quasi-random variations in emitted and reflected radiation likely due to cloud-cover changes. For the globe as a whole, the annual cycle of emitted IR radiation is +4 W/m2 and for albedo it is +l.5%. For regions of the Earth the natural "noise" is large, with albedo varia- tions of +l5% (absolute units) about a mean of 50% common in northern polar regions, for example. The CO2-induced signal is currently unknown but is likely to be small in the next decade or two, suggesting a low signal-to-noise ratio. Adequacy and Availability of Data Base. The existing record of Earth radiation budget measurements begins in the l960s with intermittent data from various satellite experiments until l975. Then, the Nimbus 6 and Nimbus 7 experiments provide a continuous data set to the present. If Nimbus 7 lasts two more years, it will overlap with the improved ERBE experiment planned for l984-l986. These data are adequate for the

358 C02-related task. However, it is urgent to ensure continued ERBE-type measurements beyond l986. Plans to do so have not yet been finalized. Spatial Coverage and Resolution of Additional Measurements Required. Satellite coverage is global. A few high-quality surface radiation stations, as in the U.S. Geophysical Monitoring for Climatic Change (GMCC) network, provide ground checks to complement the satellite observations. Frequency of Measurements Required. Daily satellite observations are desirable at as many local times as possible in order to account for diurnal effects (i.e., non-sun-synchronous satellite orbits). Feasibility and/or Existence of Technical Systems; Continuity. The Nimbus 6 and 7 experiments demonstrated high-precision proof of concept. Data are also being provided by the AVHRR instruments on NOAA opera- tional satellites. The ERBE will carry improved instrumentation and offer needed improvements in local time sampling and in-orbit calibra- tion. As noted above, it is important that ERBE-type measurements be planned now to continue beyond l986. 5.3.3.2.5 Precipitable Water Content of the Atmosphere Sensitivity. Model calculations predict that atmospheric precipitable water will increase 5-l5% as the climate warms in response to doubled CO2 concentrations (Manabe and Stouffer, l980; Wetherald and Manabe, 198l). The radiative effects resulting from this increase are a cen- tral aspect of the projected climatic response leading to the predicted warming due to increasing C02 concentrations. Observation of this parameter will, therefore, help to confirm whether the numerical models are properly simulating the role of water vapor processes in contri- buting to climatic change. Rate of Change. By the Clausius-Clapeyron equation, the fractional change in saturation water vapor mixing ratio (and therefore in precipitable water vapor if relative humidity remains constant, which is usually assumed) is approximately proportional to the fractional change in atmospheric temperature. Since most water vapor is present in the troposphere, the response time for the increase in water vapor should be about the same as for tropospheric temperature (except for the possible influence of year-to-year water storage in the land surface layers). If constant relative humidity is assumed, the response times would be identical. Signal-to-Noise Ratio. The intra-annual range in 5-year averages of mean monthly northern hemisphere specific humidity is from 2.l7 g/kg in February to 3.8l g/kg in August (Oort and Rasmusson, l97l). The inter- annual variation in annual average northern hemisphere specific humidity is about 0.05 g/kg {Oort, l982). Climate models project that a doubling of C02 concentrations would increase the specific humidity by about 0.3 g/kg (Manabe and Wetherald, l980), so that the effect should clearly

359 become evident. For a more complete analysis of signal to noise, infor- mation on the latitudinal distribution of the expected signal from models and noise from observations is needed. Adequacy and Availability of Past Data Base. The new l5-year global data base of radiosonde-derived atmospheric specific humidity by Oort offers the potential to develop important baseline data. Development of a global or hemispheric data base extending back before about l950 is doubtful, however, owing to the limited number of vertical profiles. Because the vertical soundings are taken on a fixed l2-hourly (or 24-hourly) basis only at radiosonde stations, adequate definition of the global integral may not be possible; however, the relatively long lifetime of atmospheric water (~l0 days) and recent evidence that the radiosonde network temperature record is in quite good agreement with the more extensive surface network may ameliorate these dif- ficulties and permit at least the detection of changes. Spatial Coverage and Resolution of Additional Measurements Required. A data set based on satellite measurements of atmospheric water vapor (whether direct or indirect) would be useful in order to provide adequate global coverage and to reduce variations introduced because of sampling errors in the present surface network. Prabhakara et al. (l982), for example, have developed a global map of precipitable water based on Nimbus 7 microwave measurements. Monthly and latitudinal averages should be developed since model results will probably indicate that the fractional increases in precipitable water will be a function of latitude. An alternative approach to direct or indirect measurement of water- vapor amount would be to measure the changes in the radiative flux expected to result from the projected changes in CO2 and, primarily, water-vapor amount. At the surface, downward infrared radiation is projected to increase l5-20 W/m when climate has reached equilibrium after a doubling of CO2 concentrations; this is about a 5% increase in total downward radiation. At the top of the atmosphere, total upward infrared radiation will likely change only about l% under similar conditions in response to small planetary albedo changes. To improve the signal-to-noise ratio, consideration of expected changes in the spectral distribution of the infrared radiation must be undertaken. Here the operational AVHRR on current NOAA satellites and other satel- lite programs may provide some useful information. Until detailed investigations are undertaken, however, the preferred option is to monitor atmospheric water vapor amount directly (from radiosondes and/or high-resolution spectral measurements on satellites). Frequency of Measurements Required. Measurements once or twice daily should be sufficient to develop monthly averages. Special care will have to be taken, however, so that account is taken of the presence and extent of clouds in the soundings. Feasibility and/or Existence of Technical Systems; Continuity. The continuation of the radiosonde network is assured for the purposes of

360 weather prediction. Surface-based microwave radiometry may provide additional data in the future (Skoog et al., l982). Improved coverage over the oceans, particularly in the southern hemisphere, would be helpful. Nimbus 7 and Seasat have demonstrated the feasibility of satellite water-vapor measurements over the ocean (Prabhakara et al., l982; Alishouse, l983). Research into satellite detection systems is currently under way and should be continued. Improvements will likely require either detailed spectral measurements by new satellites or clever computation of the ratios of broadband fluxes now being measured (e.g., Rosenkranz et al., l982). Compilation of the integral water vapor amounts used to initialize daily global weather forecasts should be started and compared with integrals based on the radiosonde network. Monitoring of precipitable water content will be a challenging task, but we believe that the problem should be addressed; an approach com- mencing with a feasibility study, evaluation of instrument errors, and similar questions is appropriate. 5.3.3.2.6 Equivalent Emission Temperature (Cloudiness) Sensitivity. One of the important factors that can control the long-term variation of climate is cloud cover. For example, an increase in cloud cover exerts two opposing influences on climate. On one hand, it tends to produce cooling of the Earth by reflecting a large fraction of insolation. On the other hand, it contributes to the warming of the Earth by reducing the outgoing terrestrial radiation at the top of the atmosphere. Small percentage changes in global cloudi- ness could thus accentuate or counteract a CO2-induced warming, especially at the regional scale. Unfortunately, it is difficult to determine with confidence the needed information on the three- dimensional distribution of cloud cover, because cloud cover often has ill-defined boundaries and complicated configuration. As an alter- native to direct monitoring of cloud cover, one can monitor the equivalent temperature for the upward window radiation. From equiva- lent emission temperature it is possible to compute both the upward flux of blackbody radiation from a cloud top and the effective cloud height when the vertical distribution of temperature in the atmosphere is given. By monitoring the long-term variation of both effective cloud height and planetary albedo, one can evaluate the net effect of the changes in cloud cover and optical properties of the Earth's surface. Rate of Change. Changes should occur in close association with CO2 and climate changes. Signal-to-Noise Ratio. Since the temperature change for doubled CO2 is expected to be about 3°C, one might estimate that global mean equivalent emission temperature should be measured to an accuracy of a few tenths of a degree. It has not been possible, however, to obtain a reliable estimate of the signal-to-noise ratio for the CO2-induced change of the equivalent emission temperature (or effective height of the source of upward window radiation) for the following reasons: (a)

36l In view of poor performance of current general circulation models of the atmosphere in simulating the global distribution of cloud cover and its seasonal variation, it is premature to trust the CO2-induced changes of cloud cover and effective emission temperature as determined by such models. (b) The analysis of the natural variability of effec- tive emission temperature for upward window radiation is not available. Availability of Past Data Base. Over a decade of full-disk geosta- tionary satellite visible images has been archived at the University of Wisconsin. Window radiance data from NOAA operational satellites are routinely archived by the National Environmental Satellite Service, which regularly publishes analyses including both images and tabulated data. Spatial and Temporal Resolution of Additional Measurements Required. The spatial and temporal resolution of future satellite observation should be determined in the light of careful assessment of sampling error involved in the time (or space) averaging of window radiance as obtained from a satellite. The past observation of this variable by a geosynchronous satellite provides ideal data for such an assessment. Frequency of Observations Required. The frequency of observations will be determined by the sampling error studies described above. Feasibility and/or Existence of Technical Systems; Continuity. This is an area for further study. Plans and feasibility studies should be made in conjunction with the recently initiated International Satellite Cloud Climatology Project (ISCCP) (World Meteorological Organization, l982b). 5.3.3.3 Cryospheric Parameters The features of the cryosphere include snow, sea ice, glaciers, ice sheets, permafrost, and river and lake ice. Perennial ice at present covers about 7% of the world oceans and ll% of the land surface, almost entirely in the polar regions. Seasonal snow and ice, however, occupy l5% of the Earth's surface in January, at a time when Antarctic sea ice is near its minimum extent, and 9% in July when there is almost no snow cover in the northern hemisphere. Table 5.l0 shows the global distribu- tion of ice and snow. The volume of water locked up in the Antarctic and Greenland ice sheets could potentially raise the world's mean sea level by 77 m. The general problems of possible snow and ice responses to a carbon dioxide-induced warming were reviewed by Barry (l978, l982) and Kukla (l982) in terms of the sensitivity of individual components of the cryosphere. The response of snow and ice covers to climatic change varies greatly in terms of time scale. Typical residence times of solid precipitation in the various reservoirs are approximately l0-l to l year for seasonal snow cover, l-l0 years for sea ice, and l03-l0^ years for ground ice and ice sheets. In each case certain

362 TABLE 5.l0 Distribution of Ice and Snow3. Volume Sea-Level Area (l06 km3 Equivalent (l06 km2) water) (m) Land Ice AntarcticaH,£ l2.2 25 70 Greenland l.8 2.7 7 Small ice caps and mountain glaciers 0.5 0.l2 0.3 Ground Ice (excluding Antarctica) Continuous 7.6 Discontinuous l7.3 l Sea Ice Arctic: max. l5 min. 5 Antarctic: max. 20 min. 2.5 Total Land Ice, Sea Ice, and Snow Jan: N. hemisphere 58 S. hemisphere l8 July: N. hemisphere l4 S. hemisphere 25 Global mean annual 59 ^SOURCE: Hollin and Barry (l979). ^Excludes peripheral, floating ice shelves (which do not affect sea level). —Roughly l0% of the Antarctic ice is in West Antarctica and 90% is in East Antarctica. phases of the seasonal regime are particularly critical for the occurrence of snow and ice and their response to climatic variations. Of primary importance are the times of seasonal temperature transition across the 0°C threshold (-l.8°C in the case of seawater). Other threshold effects that influence radiative and turbulent energy exchanges arise as a result of the large albedo differences between snow cover (about 0.80) and snow-free ground (0.l0-0.25) or between ice (0.65) and water (0.05-0.l0). Climate research indicates that doubling the C02 concentrations will lead to a significant reduction in the extent of snow cover and sea ice with perhaps, if the warming persists, melting and deteriora- tion of the major polar ice sheets. Sea-ice and snow-cover extent can be monitored routinely from satellites, and it has been shown that the mass balance of the large continental ice sheets of Antarctica and Greenland could also be monitored from satellites. These three cryo- spheric parameters are the most promising ones to monitor. Others, like smaller glaciers and river and lake ice, for example the dates of

363 freezeup of the latter, are in most cases probably too noisy to be used as good indicators of climate change. Several circulation models suggest that increased CO2 concentrations will lead to winter warming in the polar regions that is several times as large as in middle and low latitudes. Cryospheric conditions, particularly those responding rapidly to climatic change, may thus be excellent early indicators of CO2-induced effects. It is, nevertheless, important also to mention other, slower-responding phenomena, because of the environmental significance of potential changes that may be detected in them. 5.3.3.3.l Sea-Ice Cover Sensitivity. Model calculations by Manabe and Stouffer (l980) show sea-ice cover, both in areal extent and thickness, to be a sensitive indicator of climate change. They assume an increase of four times the present CO2 concentration and therefore project a very large warming in the polar region. This results in the sea-ice cover of the Arctic Ocean being reduced to a seasonal ice cover, which reforms in winter. Budd (l975) calculates from empirical data in Antarctica that an annual change of l°C in mean temperature corresponds to a 70-day variation in the duration of sea ice at the margin and a 2.5° latitude variation in maximum extent. Observations (Vinnikov et al., l98l) and paleoclimatic reconstructions using sediment data (Hays, l978) confirm the relation- ship between temperature and sea-ice extent. Despite the thicker ice in the Arctic (~3 m) compared with that in the Antarctic (~2 m), empirical and modeling results seem to indicate that the ice extent responds rapidly (on a seasonal time scale) to climate changes. Rate of Change. The natural variability of ice extent is large, and Zwally et al. (l983b) have not detected a systematic decrease of ice extent in Antarctica since l973. Data for the Arctic also so far do not indicate any clearcut effects due to a CO2 warming there in the past 25 years. Signal-to-Noise Ratio. Sea-ice extent is a noisy parameter when con- sidered over short time scales or small space scales, since it is determined by numerous environmental parameters and by both dynamic and thermodynamic processes (Pritchard, l980). In the Antarctic, where sea ice displays a wide seasonal variation in extent, recent studies (Zwally et al., l983b) have shown no systematic trend. Decreases between l973 and l980 were within l standard deviation of the long-term mean (Budd, l980) and have been followed by increases since l980. In the largely enclosed Arctic Ocean, variations in the ice extent are more limited on a seasonal basis, and ice thickness changes may be the first indicator of a climate change. Adequacy and Availability of Past Data Base. Accurate data on the extent of sea ice in both hemispheres are limited to the satellite records of the last two decades or so, although historical data in isolated instances, e.g., Iceland (Vilmundarson, l972) and northern Europe (Vinnikov et al., l98l) date back several centuries. Sediment

364 core data (Hays, l978) have extended this data base by many thousands of years. The satellite records are too short at present to determine definite trends, but continued monitoring over the next l0-l5 years should establish whether incipient or proposed trends are significant. Data on sea-ice extent are also not yet archived routinely in digital form. Spatial Coverage and Resolution of Additional Measurements Required. Satellite measurements allow routine integration of the areal extent of sea ice in both hemispheres. It would be useful also to have sea-ice thickness distributions in both polar regions, but these measurements are at present not feasible for satellites. Frequency of Measurements Required. Weekly averages for each hem- isphere, as determined at present, are adequate, from which monthly and annual means as well as maxima and minima can be derived. Feasibility and/or Existence of Technical Systems; Continuity. Tech- nical systems exist to carry out routine sea-ice monitoring from spacecraft. All-weather and night capability is essential, since both polar regions are dark for prolonged periods, and seasonal cloud systems are extensive. Microwave or radar systems are needed, but there is likely to be a hiatus in the launch of U.S. spacecraft with such systems that can perhaps be filled only by using European or Japanese satellites. 5.3.3.3.2 Snow Cover Sensitivity. For snow cover, the CO2 signal is more difficult to interpret than for sea ice, since the effects of CO2-induced warming on snowfall and snow cover will vary with latitude. In low and middle latitudes, where the occurrence of snow rather than rain is frequently marginal, warming will decrease the frequency of the snowfall and the duration of snow cover on the ground. In high latitudes, snowfall is limited by the frequency of cyclonic incursions and the moisture content of the air, and there is a tendency for warm winters to be snowy, as for example at Barrow, Alaska (Barry, l982). The year-to-year vari- ability of snow cover in the northern hemisphere is large, but global warming could eliminate the occurrence of snow completely in broad areas of low snowfall frequency (Dickson and Posey, l967), increase it at higher latitudes, but also possibly result in an overall increase in the length of the snow-free season in the higher latitudes due to warmer summer temperatures. Rate of Change. The generally thin snow cover of the Arctic requires little energy for melting and can therefore respond rapidly to changes in the energy balance triggered by CO2-induced warming. The duration of snow cover at high latitudes is determined primarily by summer tem- peratures, since the depth of snow is not highly variable from year to year (Barry, l982). Typically, a 30-40-cm snow cover in the Arctic disappears in about l0 days from the start of melting and requires about 2-3 kJ cm-2 (Weller and Holmgren, l974).

365 Signal-to-Noise Ratio. The signal-to-noise ratio of snow cover extent in the Arctic is likely to be high, because snowfall at high latitudes is highly variable in space and time. For example, one day's precipita- tion amount may be a large percentage of the total precipitation in some areas (Maxwell, l980). As discussed above, snow cover may either increase or decrease, depending on latitude, geographical location, and change in circulation patterns caused by CO2 effects. Adequacy and Availability of Past Data Base. Information on the dura- tion of snow cover and the last date of snow on the ground is available for Canada (Potter, l965) and in maps of probability for the northern hemisphere (Dickson and Posey, l967). Kukla (l98l) and Matson and Wiesnet (l98l) have recently compiled satellite information on monthly snow limits. Snow depth data are not available in convenient archives, although they are recorded in written synoptic weather reports, and selected mapping has been performed by the British Meteorological Office since l962 for Eurasia and since l97l for North America (Taylor, l980) . Spatial Coverage and Resolution of Additional Measurements Required. A regular program of mapping global snow cover extent and depth with a higher time and space resolution is needed for present purposes. A 50-km grid is required. Frequent satellite observations with a horizontal resolution of l-4 km would make this possible for snow-cover extent. Snow-depth data from synoptic weather reports currently have a coarse resolution, and refinement is desirable here as well. Frequency of Measurements Required. Snow maps, which at present are compiled weekly, are adequate for long-term monitoring studies, although studies of synoptic-scale interactions would require daily maps. Feasibility and/or Existence of Technical Systems; Continuity. Satel- lite systems like those currently in use are adequate for purposes of measuring snow cover extent, but at present snow depth cannot be measured from space. 5.3.3.3.3 Ice-Cap Mass Balance Changes Sensitivity. The effects of a warming on the Greenland and Antarctic ice sheets are likely to be complex (Bentley, l983; Revelle, this volume, Chapter 8). In the short-run CO2-induced climate changes could result in either positive or negative transient mass balance changes of the ice sheets, depending on regional shifts in temperature and precipitation. The potential of CO2-induced changes in the next few decades to initiate disintegration of the West Antarctic ice sheet is very small. On the other hand, Ambach (l980) states that a tempera- ture increase of only l.5°C will cause a decisive negative change in the mass balance of Greenland. Such mass balance changes will in turn slowly affect sea level. For a 3°C CO2-induced warming, Revelle (this volume, Chapter 8) calculates a 60-70-cm rise in global sea level, about half due to ablation of the Greenland and Antarctic ice

366 caps and about half due to increase in the specific volume of seawater resulting from an increase in temperature. We thus need to measure the mass balance of these ice sheets in order to understand long-term sea-level changes. Rate of Change. Because of the large mass and long residence time (about l03-l0b years) of the ice in the Greenland and Antarctic ice sheets, their responses to CO2-induced warming will be slow. Based on our present knowledge, it appears that a CO2-induced warming on the century time scale will have only minor consequences for ice sheets, but changes in their thickness may be detectable at intervals of 5-l0 years. Detectable shorter-term changes could include higher melting on ice shelves, changes in iceberg calving rates, or changes in surface gradient on ice shelves near ice rises. Ice shelves have shorter response times and provide a first indication of the state of health of the ice sheet. Signal-to-Noise Ratio. The signal-to-noise ratio is probably quite high, since the interiors of the large ice sheets are stable and relatively inactive. Noise may be increased by the limits of present measurement techniques. Adequacy and Availability of Past Data Base. The mass balance of the Greenland ice sheet is well known (Ambach, l980), and satellite measurements, using airborne radio echo sounding (Robin et al., l977) and radar altimetry have begun to give us similar data for Antarctica. Previous estimates of the mass balance changes of Antarctica are unreliable, varying from positive to negative values. Few data are available on the extent of melt features on ice shelves, on calving rates of iceberg, or on the response of ice shelves to climatic changes. Spatial Coverage and Resolution of Additional Measurements Required. Satelliteborne radar altimeter flights in a polar orbit are required for at least every l0° of longitude. The vertical resolution of height of the ice sheet should be +5 cm. Areas in which to look for changes are not only the elevation of the interior of ice sheets but also gradient changes near ice shelves and ice rises. Frequency of Measurements Required. Changes in the ice-sheet mass balance of Greenland and Antarctica should be monitored at 5-year intervals. Seasonal features, such as the extent of melt features on ice shelves, should be surveyed annually. Feasibility and/or Existence of Technical Systems; Continuity. The feasibility of making such measurements has been demonstrated by using the radar altimeter on the GEOS-3 satellite launched in l975 (Brooks et al., l978) and by studies based on Seasat data (Zwally et al., l983a) and simulations of possible future satellite systems (Zwally et al., l98l). No U.S. satellite currently carries either radar or laser altimeters, however, but plans for future satellites should include them.

367 5.3.3.4 Oceanic Parameters The oceans are key elements in the Earth's climate system. However, there are still major uncertainties in our knowledge of how the coupled ocean-atmosphere system works and, therefore, how it may change when CC12 is added to the atmosphere. The available, though sketchy, evidence points toward the fact that the ocean is most probably delay- ing the temperature signal of increasing C02 by mixing heat downward. It is clear from model studies, which up to now have treated the oceans quite simply (e.g., Gates et al., l98l; Manabe and Stouffer, l980), that the response of the atmosphere is paced by that of the ocean. Thompson and Schneider (l979) and Schneider and Thompson (l98l) discussed the question of the transient response of the atmosphere to C02 using models based on the thermal inertia of the upper layers of the oceans in combination with their interaction with deeper waters. Similar conclusions about the delay of a warming, i.e., as long as a few decades, were reached in reports of the National Academy of Sciences (Climate Research Board, l979; CO^Climate Review Panel, l982). Ramanathan (l98l) emphasizes the inadequacy of present coupled models for examining the transient response. Work of Fine et al. (l98l) on tritium penetration into the ocean suggested that the delay time may have been underestimated. Regional climate changes will also be strongly associated with climatic variation in the ocean (Schneider and Thompson, l98l; Bryan et al., l982; Thompson and Schneider, l982). In order to identify changes in the ocean due to CO2 warming, a long-term measurement program is required. Brewer (this volume, Chapter 3, Section 3.2) examines the changes in ocean chemistry to be expected from increasing CO2. Baker and Barnett (l982) describe the physical oceanographic variables that would be expected to respond. Of the parameters identified by these authors, which include sea level, sea temperatures, salinity, and ocean circulation patterns, the first two seem most appropriate for monitoring the possible effects of C02-induced warming. Changes in sea level, though not driven by thermal expansion alone, may be the best indicator of the global change in ocean temperature, because an observational network exists, at least in the northern hemisphere, and sea-level data are representative of integrated, rather than point, measurements. 5.3.3.4.l Sea Level Sensitivity. Sea level should be a sensitive indicator of C02 effects. A change of 0.l% of the global land ice cover will result in a sea-level change of about 5 cm (Flint, l97l), and increases in ocean temperature will presumably accompany increases in atmospheric tem- perature; a change of 0.5°C in the upper 200 m would increase sea level by roughly 2 cm (Baker and Barnett, l982). Given a 3°C atmospheric warming, Revelle (this volume, Chapter 8) estimates a rise in sea level about l00 years from now of at least 30 cm, resulting from ocean warm- ing, and a probable rise of between 60 and 70 cm, if ice ablation is included. Both of these effects will be global in nature, and it may be far easier to detect a signal that is coherent in all the oceans

368 than to identify one that is rather regional in character. Measurement problems exist, however, which are discussed below. Rate of Change. Rises in sea level in response to projected induced warming will be slow, but much more rapid than recent historic rates. Estimates of the rate of sea-level rise so far during this century range from l to 3 mm/year (see Revelle, this volume, Chapter 8) . Problems of tectonic movement and poor station distribution in terms of location, offshore current, and wind systems, for example, leave one uncertain as to the reality and meaning of these numbers. Signal-to-Noise Ratio. The relatively long time series of sea level provide opportunities to estimate signal-to-noise ratios and hence make detection of global changes more feasible (cf., Madden and Ramanathan, l980) . However, there are relatively few sea-level stations in the southern hemisphere that possess a long record, and, further, sea-level data in huge ocean regions must be reconstructed from limited hydro- graphic data These deficiencies will make detection of a truly global signal somewhat more difficult. Adequacy and Availability of Past Data Base. Long time series of sea-level data are available (e.g., Emery, l980; Revelle, this volume, Chapter 8) , but their interpretation is complicated by the problems listed above. Nevertheless, these studies suggest a coherent rising of sea level on scales of oceanic dimensions. The statistical significance of the changes and their relationship to CO2-induced warming are hard to estimate. Spatial Coverage and Resolution of Additional Measurements Required. It appears feasible to achieve spatial coverage sufficient to ameliorate some of the difficulties mentioned above, although problems in the interpretation of individual records will remain. Measurements are needed at all open ocean island locations; primarily lacking at present are islands in the Atlantic and Indian Oceans. Also needed are stations around the Antarctic continent. Global sea-level coverage will be available from the various altimetric satellites now being proposed for the late l980s. The estimated accuracy of these, ^l0 cm, will be too low for useful estimation of C02 effects in the next one or two decades; however, over the long term (i.e., a century) such satellite measurements will be helpful. Frequency of Measurements Required. Since most sea-level measurements are made for tidal prediction, the frequency of measurement is more than adequate for long-term sea-level change. It is critical to keep the measurement going for a long time (decades) . Feasibility and/or Existence of Technical Systems; Continuity. Tide- gauge measurements are simple and have been carried out for a long time. The existing technology that permits unattended operation for months to years with data recording on tape cassettes is entirely adequate for the purpose.

369 5.3.3.4.2 Sea Temperature Sensitivity. Most simulations of increased atmospheric CO2 show a substantial warming of surface air temperature over the globe. The effect is most pronounced in high latitudes. As with sea level, change in sea-surface temperature resulting from this air temperature increase should be global in nature. However, its magnitude will have a strong regional character (Baker and Barnett, l982), owing to regional varia- tion in vertical mixing and diffusion and the relative importance of different physical processes in the ocean heat budget; detection of a CO2-induced temperature signal in the oceans will thus be difficult. Certainly, the nature and magnitude of such a signal merits further study. Rate of Change. Response time of sea temperatures to CO2 effects could be relatively slow owing to the thermal inertia of the oceans. Data on sea-surface temperature (SST) for the Indian, North Atlantic, and North and Tropical Pacific Oceans show perhaps a l°C rise over the last 80 years in all oceans (Baker and Barnett, l982). The change may be smaller because of possible errors due to the gradual conversion from measurements by bucket thermometer to ship's injection thermom- eters; the latter consistently showing SSTs warmer by 0.4°C. Signal-to-Noise Ratio. Current data archives on the surface tempera- ture field of the world oceans are quite extensive and should make proper determinations of signal-to-noise ratios relatively simple. The historical record of surface temperature and subsurface temperature needs to be analyzed to determine the type of signal that could be detected over the background. Estimates of regional signal-to-noise ratios also need to be made; these have not been performed so far. The effects of mesoscale eddies add a large noise to subsurface ocean temperature that is generally not so prominent in the SST field. Adequacy and Availability of Past Data Base. Measurements of sea- surface temperature have been made in many regions of the world for the past l00 years from ships, islands, and coastal stations, although many of these data are crude and unrepresentative. More complete analyses of this record should be undertaken. Analyzed data covering the North Atlantic and Pacific are available back to about l948 (Walsh and Sater, l98l; Bunker, l980). Subsurface temperature fields are much less well known. The historical record of subsurface ocean temperature is unfor- tunately relatively short, and the data are subject to errors intro- duced by changes in the instrumentation (bathythermographs), which may introduce bias into the data sets. Spatial Coverage and Resolution of Additional Measurements Required. It is now possible to collect data on sea-surface temperatures for all oceanic regions of the world. Infrared satellite sensors have been used since the early l970s, and improved satellite analyses are now produced daily for all oceanic regions. Programs of direct measurement must, however, be maintained for their better accuracy, and new methods

370 (e.g., acoustic tomography) should be explored for measuring integrated ocean temperature. Frequency of Measurements Required. Daily analyses are available from satellite infrared sensors; their frequency is adequate. Feasibility and/or Existence of Technical Systems; Continuity. It is necessary to evaluate rigorously how well the sea-surface temperature signal can be extracted from the satellite microwave radiometer. The demonstrated accuracy of the spaceborne sea-surface temperature systems still remains inadequate to resolve the signals as observed by direct measurement. For studies of warming, accuracies of from 0.l to 0.5°C are needed, but this accuracy has not yet been achieved. Much of this uncertainty comes from surface foam, water vapor, and liquid water in the path of the sensor; the new multichannel sensors may show marked improvement. Acoustic tomography (Munk and Wunsch, l982) also offers promise of providing integrated ocean temperatures. 5.3.4 Conclusions and Recommendations 5.3.4.l Priority of Parameters to be Monitored To help determine the current and projected effects of increasing atmospheric CO2 on climate, we recommend further elaboration and, if sound, the development of a monitoring strategy in which many measures of the state of the climate systems are monitored and analyzed as an ensemble. If recent climatic trends are sustained, it seems likely that there will be an increasing number of claims to have distinguished a sig- nificant warming. The problem will then become increasingly one of attribution of cause and effect. The most promising means to achieve convincing attribution will be development of reliable records of several parameters in addition to temperature. Clearly, for technical or cost reasons adequate monitoring of some parameters will be much more readily achievable than others. To accomplish early attribution, initial emphasis should be given to these. While initial emphasis may be placed on parameters that may be monitored immediately, cheaply, or easily, it is important over the long run to build up a rather complete data base, not only for reasons of detection but also for research and for calibration of models of the climate system. It may take until 20l0 or 2020 to begin to have useful data bases on some of the parameters mentioned here, but they should not be neglected. Instead, we should anticipate that a monitoring program will gradually evolve into a program to verify and calibrate crucial aspects of model calculations, especially the numerous projected effects of increasing atmospheric CO2, for example, sea-level rise and changes in rainfall in midlatitudes. Based on this initial survey, we summarize our recommendations for monitoring in Table 5.ll.

37l TABLE 5.11 Priority in Monitoring Variables for Early Detection of 002 Effects Monitoring Monitoring Causal Factors by Climatic Effects by Priority Measuring Changes in Measuring Changes in First CO2 concentrations Troposphere/surface Volcanic aerosols temperatures (including Solar radiance sea temperatures) Stratospheric temperatures Radiation fluxes at the top of the atmosphere Precipitable water content (and clouds) Second "Greenhouse" gases Snow and sea-ice covers other than CO2 Polar ice-sheet mass balance Stratospheric and Sea level tropospheric ozone 5.3.4.2 Measurement Networks The key to a successful monitoring strategy is a global observation system. Satellites are a major component of such a system, and it is essential to be able to continue monitoring without interruption on a long-term basis the radiative fluxes, the planetary albedo, snow and ice extent, and sea-surface temperatures and to improve the spaceborne measurements of tropospheric and stratospheric temperatures, precip- itable water content of the atmosphere, mass balance of the polar ice sheets and sea level, as well as aerosols, ozone, and other atmospheric constituents. Many of the satellite measurements that are being made at present are difficult to calibrate. Of particular concern are the vertical soundings made from spacecraft and the lack of supporting surface-based or surface-launched profiling systems. A concern, for example, is the dismantling of the global rocketsonde system of stratospheric tempera- ture soundings. Some key stations should be retained to calibrate the instruments flown on satellites. Similarly, the only reliable method of characterizing stratospheric aerosols at present is by the deploy- ment of lidar systems. By adding one or two such systems in the southern hemisphere to the existing network, and through occasional aircraft flights to calibrate the satellite soundings, an adequate amount of data could be collected. Other parameters, for example ozone, are also inadequately measured at present. Total ozone values derived from Dobson spectrometer and satellite profile measurements are not enough in themselves, and the existing ozonesonde network must be maintained and augmented in data-sparse regions.

372 The southern hemisphere presents a particular problem in monitoring climatic changes that can only be solved by improving and perfecting the present satellite-based sounding techniques. Table 5.l summarizes requirements and technical systems for monitor- ing high-priority variables. 5.3.4.3 Modeling and Statistical Techniques The internal physical consistency and relative ease of diagnosis of simulated climatic data make the construction of realistic and compre- hensive models a prerequisite for the development of a successful fingerprinting strategy for the detection of CO2-induced climatic change. In addition, climate models are needed in order to determine the accuracy that is required in monitored climatic variables. Unfor- tunately, the C02-induced climatic changes calculated from the various current climate models continue to show substantial differences. In order to develop an effective monitoring strategy, it is essential that further intensive efforts be made to improve climate models by validat- ing them against the observed structure and behavior of the ocean- atmosphere system and to make effective use of model improvements. Another important element is the development of methods for the statistical identification of a C02-induced climate signal against the background of natural climatic fluctuations. Statistical tech- niques applied so far focus on the significance of the signal-to-noise ratio, assuming that the data at individual points may be modeled as a first-order autoregressive process. Further use should be made of significance tests that consider longer-term dependence in the climatic time series and that provide estimates of confidence limits. An essen- tial ingredient of a successful detection strategy will be the develop- ment of techniques that take into account not only the temporal corre- lation but also the spatial correlation that is characteristic of nearly all climatic variables and that lead to more careful and sophis- ticated statistical tests of a possible C02 climatic signal. Here again, model-simulated data can be used effectively to begin to develop and test the statistical procedures that will ultimately have to be applied to monitored observational data. However, purely statistical inferences will have to be buttressed to the greatest extent possible by physical reasoning. 5.3.4.4 Objective Evaluation of Evidence The differing interpretations of the effects of likely C02-induced climate changes, as arrived at by different authors (see Section 5.2), often using identical data sets, underline the need for objective evaluation of evidence. It is wise to anticipate the need at national and international levels for periodic efforts to evaluate evidence and arbitrate between divergent opinions, where necessary. Some centers that can perform such functions are already in existence as part of national and international programs. At these centers climatic indices

373 should be collected and compared, and evaluation and improvement of techniques for identification of climatic changes should be encouraged. REFERENCES Alishouse, J. C. (l983). Total precipitable water and rainfall determinations from SEASAT scanning multichannel microwave radiometer. J. Geophys. Res. 88:l929-l935. Ambach, W. (l980) . Anstieg der CO2-Konzentration in der Atmospha're und Klimaa'nderuug: Mfigliche Auswirkungen auf dem Grfinla'ndischen Eisschild. Wetter und Leben 32:l35-42. Angell, J. K., and J. Korshover (l978a). Global temperature variation; surface—l00 mb: an update into l977. Mon. Wea. Rev. l06:755-770. Angell, J. K., and J. Korshover (l978b). Estimate of global temperature variations in the l00-30 mb layer between l958 and l977. Mon. Wea. Rev. l06:l422-l432. Augustsson, T., and V. Ramanathan (l977). A radiative-convective model study of the CO2-climate problem. J. Atmos. Sci. 34:448-45l. Baker, Jr., D. J., and T. P. Barnett (l982). Possibilities of detecting CO2-induced effects: ocean physics. In U.S. Department of Energy (l982), pp. 30l-342. Baldwin, B., J. B. Pollack, A. Summers, O. B. Toon, C. Sagan, and W. Van Camp, (l976). Stratospheric aerosols and climatic change. Nature 263:55l-555. Barry, R. G. (l978). Cryospheric responses to a global temperature increase. In Carbon Dioxide, Climate and Society, J. Williams, ed. Pergamon, Oxford, pp. l69-l80. Barry, R. G. (l982). Snow and ice indicators of possible climatic effects of increasing atmospheric carbon dioxide. In U.S. Department of Energy (l982), pp. 207-236. Bates, J. R. (l977). Dynamics of stationary ultra-long waves in middle latitudes. Q. J. Roy. Meteorol. Soc. l03:397-430. Bentley, C. R., ed.(in press). CO2-induced Climate Change and the Dynamics of Antarctic Ice. Proceedings of AAAS Symposium, Toronto, January l98l, American Association for the Advancement of Science, Washington, D.C.. Bloomfield, P., M. L. Thompson, G. S. Watson, and S. Zeger (l98l). Frequency Domain Estimation of Trends in Stratospheric Ozone. Technical Report l82. Dept. of Statistics, Princeton U., Princeton, N.J. Borzenkova, I. I., K. Ya. Vinnikov, L. P. Spirina, and D. I. Stekhnovskii (l976). Variation of northern hemisphere air temperature from l88l to l975. Meteorol. Gidrol. 7:27-35. Broecker, W. S. (l975). Climatic change: are we on the brink of a pronounced global warming? Science l88:460-463. Brooks, R. L., W. J. Cambell, R. O. Ramseier, H. R. Stanley, and H. J. Zwally (l978). Ice sheet topography by satellite altimetry. Nature 274:539-43.

374 Bryan, K., F. G. Komro, S. Manabe, and M. J. Spelman (l982). Transient climate response to increasing atmospheric carbon dioxide. Science 2l5:56-58. Bryson, R. A. (l980). CO2, Aerosols, and Modeling Climate of the Recent Past. In U.S. Department of Energy (l980), pp. l39-l43. Bryson, R. A., and G. J. Dittberner (l976). A non-equilibrium model of hemispheric mean surface temperature. J Atmos. Sci. 33:2094-2l06. Bryson, R. A., and B. M. Goodman (l980). Volcanic activity and climatic changes. Science 207:l04l-l044. Bryson, R. A., and W. M. Wendland (l970). Climatic Effects of Atmospheric Pollution. Global Effects of Environmental Pollution, S. F. Singer, ed. Springer-Verlag, New York, pp. l30-l38. Budd, W. F. (l975). Antarctic sea ice variations from satellite sensing in relation to climate. J. Glaciol. 73:4l7-427. Budd, W. F. (l980). The importance of the antarctic region for studies of the atmospheric carbon dioxide concentration. In Carbon Dioxide and Climate: Australian Research, G. I. Pearman, ed. Australian Academy of Science, Canberra, pp. ll5-28. Budyko, M. I. (l969). The effect of solar radiation variations on the climate of the earth. Tellus 2l:6ll-6l9. Budyko, M. I. (l974). Climatic Changes. Hydrometeoizdat, Leningrad (47 pp.); transl., American Geophysical Union, l977. Budyko, M. I., and K. Ya. Vinnikov (l973). Modern climate variations. Meteorol. Gidrol. 9:3-l3. Bunker, A. F. (l980). Trends of variables and energy fluxes over the Atlantic Ocean from l948 to l972. Mon. Wea. Rev. l08:720-732. Callendar, G. S. (l938). The artificial production of carbon dioxide and its influence on temperatures. Q. J. Roy. Meteorol. Soc. .64:223-240. Callendar, G. S. (l96l). Temperature fluctuations and trends over the earth. Q. J. Roy. Meteorol. Soc. 87:l-l2. Charney, J. G. (l975). Dynamics of deserts and drought in the Sahel. Q. J. Meteorol. Soc. l0l:l93-202. Clark, W. C., ed. (l982). Carbon Dioxide Review: l982. Oxford U. Press, New York, 469 pp. Clark, W. C., K. H. Cook, G. Marland, A. M. Weinberg, R. M. Rotty, P. R. Bell, L. J. Allison, and C. L. Cooper (l982) . The carbon dioxide question: a perspective for l982. In Clark (l982). Climate Research Board (l979). Carbon Dioxide and Climate: A Scientific Assessment. National Research Council, National Academy of Sciences, Washington, D.C., 22 pp. CO2/Climate Review Panel (l982). Carbon Dioxide and Climate: A Second Assessment. National Research Council, National Academy Press, Washington, D.C., 72 pp. Dickson, R. R., and J. Posey (l967). Maps of snow-cover probability for the northern hemisphere. Mon. Wea. Rev. 95:347-53. Donner, L., and V. Ramanathan (l980). Methane and nitrous oxide: their effects on terrestrial climate. J. Atmos. Sci. 37:ll9-l24. Eddy, J. A. (l976). The Maunder minimum. Science l92:ll89-l202. Eddy, J. A. (l977). Climate and the changing sun. Climatic Change l:l73-l90.

375 Eddy, J. A., R. L. Gilliland, and D. V. Hoyt (l982). Changes in the solar constant and climatic events. Nature 300:689-693. Elliott, W. P. (l982) . A note on the historical industrial production of carbon dioxide. NOAA Air Resources Laboratories, Rockville, Md. Ellis, J. S., T. H. Vonder Haar, S. Levitus, and A. H. Oort (l978). The annual variation in the global heat balance of the earth. J. Geophys. Res. 83: l958-l962 . Emery, K. 0. (l980) . Relative sea level from tide gauge records. Proc. Nat. Acad. Sci. 77:6968-6972. Epstein, E. S. (l982). Detecting climate change. J. Appl. Meteorol. _ Pels, S. B., J. D. Mahlman, M. D. Schwarzkopf, and R. W. Sinclair (l980) . Stratospheric sensitivity to perturbations in ozone and carbon dioxide: radiative and dynamical response. J. Atmos. Sci. 37_:2265. Fine, R. A., J. L. Reid, and H. G. Ostlund (l98l). Circulation of tritium in the Pacific Ocean. J. Phys. Oceanog. ll:3. Fishman, J., V. Ramanathan, P. J. Crutzen, and S. C. Liu (l979). Tropospheric ozone and climate. Nature 282: 8l8-820. Flint, R. F. (l97l) . Glacial and Quarternary Geology. Wiley, New York. Gates, W. L., K. H. Cook, and M. E. Schlesinger (l98l). Preliminary analysis of experiments on the climatic effects of increased CO2 with an atmospheric general circulation model and a climatological ocean model. J. Geophys. Res. 86: 6385. Geller, M. A., and V. C. Alpert (l980). Planetary wave coupling between the troposphere and the middle atmosphere as a possible sun-weather mechanism. J. Atmos. Sci. 37: ll97-l2l5. Geophysics Study Committee (l982) . Solar Variability, Weather, and Climate. National Research Council, National Academy Press, Washington, D.C., l06 pp. Gilliland, R. L. (l98l) . Solar radius variations over the past 265 years. Astrophys. J. 248: ll44-ll55. Gilliland, R. L. (l982). Solar, volcanic, and C02 forcing of recent climatic changes. Climatic Change 4: lll-l3l. Hammer, C. U. (l977). Past volcanism revealed by Greenland ice sheet impurities. Nature 270: 482-486. Hansen, J. (l980) . CO2, solar variations, recent climate, and model predictions. In U.S. Department of Energy (l980) , pp. l44-l53. Hansen, J., D. Johnson, A. Lacis, S. Lebedeff, P. Lee, D. Rind, and G. Russell (l98l). Climate impact of increasing atmospheric carbon dioxide. Science 2l3: 957-966. Hayashi, Y. (l982). Confidence intervals of a climatic signal. J^ Atmos. Sci. 39:l895. Hays, J. D. (l978). A review of the late quarternary climatic history of Antarctic seas. In Antarctic Glacial History and World Paleoenvironments : August l7, l977, Birmingham, United Kingdom. F. M. van Zinderen Bakker, ed. A. A. Balkema, Rotterdam, The Netherlands, pp. 57-7l. Heath, D. F., and B. M. Schlesinger (l982). Secular and periodic variations in stratospheric ozone from satellite observations — l970-l979. NASA/Goddard Space Flight Center Preprint.

376 Hirschboeck, K. (l980). A new worldwide chronology of volcanic eruptions. Palaeogeogr. Palaeoclim. Palaeoecol. 29:223-24l. Hoffert, M. I., A. J. Callegari, and C.-T. Hsieh (l980). The role of deep sea heat storage in the secular response to climatic forcing. J. Geophys. Res. 85:6667. Hollin, J. T., and R. G. Barry (1979). Empirical and theoretical evidence concerning the response of the earth's ice and snow cover to a global temperature increase. Environ. Int. 2:437-444. Hoyt, D. V. (l979a). An empirical determination of the heating of the earth by the carbon dioxide greenhouse effect. Nature 282:388-390. Hoyt, D. V. (l979b). Variations in sunspot structure and climate. Climatic Change 2:79-92. Jones, P. D., T. M. L. Wigley, and P. M. Kelly (l982). Variations in surface air temperatures: Part l. Northern hemisphere, l88l-l980. Hon. Wea. Rev. ll0:59-70. Katz, R. W. (l980). Statistical evaluation of climate experiments with general circulation models: inferences about means. Climatic Research Institute Report l5. Oregon State U., Corvallis. Katz, R. W. (l982). Statistical evaluation of climate experiments with general circulation models: a parametric time series modeling approach. J. Atmos. Sci. 39:l446-l455. Kelley, P. M., and P. D. Jones (l98l). Annual temperatures in the Arctic, l88l-l98l. Climate Monitor l0..l22-l24. Klein, W. H. (l982). Detecting climate effects of increasing atmospheric carbon dioxide. In U.S. Department of Energy (l982) , pp. l75-l94. Kraus, E. B. (l955). Secular changes of tropical rainfall regimes. £. J. Roy. Meteorol. Soc. 8l:l08. Kukla, G. J. (l98l). Climatic role of snow covers. In Sea Level, Ice and Climatic Change, I. Allison, ed. International Assoc. of Hydrological Sciences Publ. l3l. IAHS, Wallingford, Oxfordshire, U.K., pp. 79-l07. Kukla, G. J. (l982). Carbon dioxide and polar climates. In U.S. Department of Energy (l982), pp. 237-288. Lacis, A., J. Hansen, P. Lee, T. Mitchell, and S. Lebedeff (l98l). Greenhouse effect of trace gases, l970-l980. Geophys. Res. Lett. 1:l035-l038. Lamb, H. H. (l970). Volcanic dust in the atmosphere with a chronology and assessment of its meteorological significance. Phil. Trans. Roy. Phil. Soc. London:425-533. Levy, H., J. O. Mahlman, and W. J. Moxim (l979). A preliminary report on the numerical simulation of the three-dimensional structure and variability of atmospheric N2O. Geophys. Res. Lett. 6:l55-l58. MacCracken, M. C. (l983). Have we detected CO2-induced climate change?—Problems and prospects. In U.S. Department of Energy (l983), Vol. 5, pp. 3-45. MacCracken, M. C., and H. Moses (l982). The first detection of carbon dioxide effects: workshop summary. Bull. Am. Meteorol. Soc. .63:ll64-ll78. Madden, R. A., and V. Ramanathan (l980). Detecting climate change due to increasing carbon dioxide. Science 209:763-68.

377 Manabe, S., and D. G. Hahn (l98l). Simulation of atmospheric vari- ability. Mon. Wea. Rev. l09:2260-2286. Manabe, S., and R. J. Stouffer (l980). Sensitivity of a global climate model to an increase of CO2 concentration in the atmosphere. J. Geophys. Res. 85:5529-5554. Manabe, S., and R. F. Strickler (l964). Thermal equilibrium of the atmosphere with a convective adjustment. J. Atmos. Sci. 2l:36l-385. Manabe, S., and R. T. Wetherald (l975). The effects of doubling the CO2 concentration on the climate of a general circulation model. J. Atmos. Sci. 32:3. Manabe, S., and R. T. Wetherald (l980). On the distribution of climate change resulting from an increase in CO2 content of the atmosphere. J. Atmos. Sci. 37:99. Marland, G., and R. Rotty (l982). Carbon dioxide emissions from fossil fuels, l950-l98l. Institute for Energy Analysis, Oak Ridge, Tenn. Mass, C., and S. H. Schneider (l977). Statistical evidence on the influence of sunspots and volcanic dust on long-term temperature records. J. Atmos. Sci. 34:l995-2004. Matson, M., and D. R. Wiesnet (l98l). New data base for climate studies. Nature 289:45l-456. Maxwell, J. B. (l980). The Climate of the Canadian Arctic Islands and Adjacent Waters. Environment Canada, Canadian Government Publishing Centre, Hull, Quebec, 53l pp. McCormick, M. P., W. P. Chu, L. R. McMaster, G. W. Grams, B. M. Herman, T. J. Pepin, P. B. Russell, and T. J. Swissler (l98l). SAM-II aerosol profile measurements, Poker Flat, Alaska, July l6-l9, l979. Geophys. Res. Lett. 8:3-4 Miles, M. K., and P. B. Gildersleeves (l978). A statistical study of the likely influence of some causative factors on the temperature changes since l665. Meteorol. Mag. l07:l93-204.. Mitchell, J. M., Jr. (l96l). Recent secular changes of global temperature. Ann. N.Y. Acad. Sci. 95:235-250. Mitchell, J. M., Jr. (l970). A preliminary evaluation of atmospheric pollution as a cause of the global temperature fluctuation of the past century. In Global Effects of Environmental Pollution, S. F. Singer, ed. Springer-Verlag, New York, pp. l39-l55. Mitchell, J. M., Jr. (l983). An empirical modeling assessment of volcanic and carbon dioxide effects on global scale temperature. American Meteorological Society, Second Conference on Climate Variations, New Orleans, Louisiana, January l0-l4, l983. Munk, W, and C. Wunsch (l982). Observing the ocean in the l990s. Phil. Trans. Roy. Soc. London Ser. A 307:439-464. Murphy, A. H., and R. W. Katz (l982). Statistical methodology for first detection of carbon dioxide effects in the atmosphere. In U.S. Department of Energy (l982), pp. l65-l74. Newell, R. E., and A. Deepak, eds. (l982). Mount St. Helens Eruptions of l980—Atmospheric Effects and Potential Climatic Impact. NASA report SP-458. National Aeronautics and Space Administration, Washington, D.C.

378 Newhall, C. G., and S. Self (l982). The volcanic explosivity index (VEI): an estimate of explosive magnitude for historical volcanism. J. Geophys. Res. 87:l23l-l238. Oliver, R. C. (l976). On the response of hemispheric mean temperature to stratospheric dust: an empirical approach. J. Appl. Meteorol. .l5:933-950. Oort, A. H. (l982). Global atmospheric circulation studies, l958-l973. NOAA Prof. Paper. U.S. Govt. Printing Office, Washington, D.C. (in preparation). Oort, A. H., and E. M. Rasmusson (l97l). Atmospheric circulation statistics. NOAA Prof. Paper 5. U.S. Govt. Printing Office, Washington, D.C. Paltridge, G., and S. Woodruff (l98l). Changes in global surface temperature from l880 to l977 derived from historical records of sea surface temperature. Mon. Wea. Rev. l09:2427-2434. Pierotti, D., and R. A. Rasmussen (l977). The atmospheric distribution of nitrous oxide. J. Geophys. Res. 37:5823-5832. Pierotti, D., and R. A. Rasmussen (l978). Inter-laboratory calibration of atmospheric nitrous oxide measurements. Geophys. Res. Lett. 5-: 353-355. Pivavarova, Z. I. (l968). The long-term variation of the intensity of solar radiation according to the observations of actinemetric stations [in Russian]. Glavnaya Geofiz. Obs. Trudy 233:l7-37. Pivavarova, Z. I. (l977). Radiation Characteristics of the Climate of the USSR (in Russian). Gidrometeoizdat, Leningrad. Plass, G. N. (l953). Some problems in atmospheric radiation. Proc. Toronto Meteorol. Conf.:53. Royal Meteorological Society, London. Plass, G. N. (l956). The carbon dioxide theory of climatic change. Tellus 8:l40-l54. Pollack, J. B., O. Toon, C. Sagan, A. Summers, B. Baldwin, and W. Van Camp, (l976a). Volcanic explosion and climatic change: a research assessment. J. Geophys. Res. 8l:l07l-l083. Pollack, J. B., O. B. Toon, A. Summers, B. Baldwin, C. Sagan, and W. Van Camp, (l976b). Stratospheric aerosols and climate change. Nature 263:5578 Potter, J. G. (l965). Snow cover. Climatological Studies 3. Meteorological Branch, Dept. of Transport, Toronto, Canada. Potter, G. L., H. W. Ellsaesser, M. C. MacCracken, and F. M. Luther (l975). Possible climatic impact of tropical deforestation. Nature 258:697-698. Potter, G. L., H. W. Ellsaesser, M. C. MacCracken, J. S. Ellis, and F. M. Luther (l980). Climate change due to anthropogenic surface albedo modification. In Interactions of Energy and Climate, W. Bach, J. Pankrath, and J. Williams, eds. Reidel, Dordrecht, pp. 3l7-326. Prabhakara, C., H. D. Chang, and A. T. C. Chang (l982). Remote sensing of precipitable water over the oceans from Nimbus 7 microwave measurements. J. Atmos. Sci. 59. Pritchard, R. S., ed. (l980). Sea Ice Processes and Models. Proceedings of the Arctic Ice Dynamics Joint Experiment, U. of Washington Press, Seattle, 474 pp.

379 Ramanathan, V. (l975). Greenhouse effect due to chlorofluorocarbons: climatic implications. Science l90:50. Ramanathan, V. (l980). Climatic effects of anthropogenic trace gases. In Interactions of Energy and Climate, W. Bach, J. Pankrath, and J. Williams, eds. Reidel, Boston, Mass., pp. 269-280. Ramanathan, V. (l98l). The role of ocean-atmosphere interactions in the CO2-climate problem. J. Atmos. Sci. 38:9l8. Ramanathan, V., L. B. Callis, and R. E. Boughner (l976). Sensitivity of surface temperature and atmospheric temperature to perturbations in the stratospheric concentration of ozone and nitrogen dioxide. J. Atmos. Sci. 33:l092-lll2. Rasmussen, R. A., and M. A. K. Khalil (l98l). Increase in the concentration of atmospheric methane. Atmos. Environ. l5;883. Reinsel, G. C. (l98l). Analysis of total ozone data for the detection of recent trends and the effects of nuclear testing during the l960s. Geophys. Res. Lett. 8:l227-l230. Reinsel, J. C., J. C. Tiso, A. J. Miller, C. L. Mateer, J. J. DeLuisi, and J. E. Frederick (l983). Analysis of upper stratospheric Umkehr ozone profile data for trends and the effects of stratospheric aerosols. American Geophysical Union Spring Meeting, Baltimore, Maryland, May 30-June 3, l983. Abstract in EOS 64:l99. Revelle, R., and H. E. Suess (l957). CO2 exchange between atmosphere and ocean, and the question of an increase of atmospheric CO2 during the past decades. Tellus 9:l8. Robin, G. de Q., D. J. Drewry, and D. T. Meldrum (l977). International studies of ice sheet and bedrock. Phil. Trans. Roy. Soc. London Ser. B 279:l85-96. Robock, A. (l978). Internally and externally caused climate change. J. Atmos. Sci. 35:llll-ll22. Robock, A. (l979a). The performance of a seasonal global climatic model. Report of the JOC Study Conference on Climate Models: Performance, Intercomparison, and Sensitivity Studies, Vol. 2. GARP Publ. 22. World Meteorological Organization, Geneva, Switzerland, pp. 766-802. Robock, A. (l979b). The Little Ice Age: the northern hemisphere average observations and model calculations. Science 206:l402-l404. Robock, A. (l98l). The Mount St. Helens volcanic eruption of l8 May l980: minimal climatic effect. Science 2l2:l383-l384. Robock, A. (l983). Ice and snow feedbacks and the latitudinal and seasonal distribution of climate sensitivity. J. Atmos. Sci. 10:986-997. Rosenkranz, P. W., M. J. Komichak, and D. H. Staelin (l982). A method for estimation of atmospheric water vapor profiles by microwave radiometry. J. Appl. Meteorol. 2l:l364. Sagan, C., O. B. Toon, J. B. Pollack (l979). Anthropogenic albedo changes and the earth's climate. Science 206:l363-l367. St. John, D. S., S. P. Bailey, W. H. Fellner, J. M. Minor, and R. D. Snee (l98l). Time series search for a trend in total ozone measurements. J. Geophys. Res. 86:7299-73ll. SCEP (l970). Man's Impact on the Global Environment, Study of Critical Environmental Problems. MIT Press, Cambridge, Mass.

380 Schlesinger, M. E. (l982). A Review of Climate Model Simulations of CO2-Induced Climatic Change. Report No. 4l, Climatic Research Institute, Oregon State U., Corvallis, l35 pp. Schlesinger, M. E. (l983). Simulating C02-induced climate change with mathematical climate models: capabilities, limitations, and prospects. In U.S. Department of Energy (l983), III.3-III.l40. Schneider, S. H., and S. L. Thompson (l98l). Atmospheric CO2 and climate: importance of the transient response. J. Geophys. Res. .86:3l35. Schonwiese, C. D. (l98l). Solar activity, volcanic dust, and temperature: statistical relationships since ll60 A.D. Arch. Meteorol. Geophys. Bioclimatol. Ser. A 30:l-22. Simkin, T., L. Seibert, L. McClelland, W. G. Nelson, D. Bridge, C. G. Newhall, and J. Latter (l98l). Volcanoes of the World. Hutchinson, Ross, New York. Siquig, R. A., and D. V. Hoyt (l980). Sunspot structure and the climate of the last one hundred years. In Ancient Sun: Proceedings of the Conference on the Ancient Sun: Fossil Record in the Earth, Moon, and Meteorites. Boulder, Colorado, Oct. l6-l9, l979. R. 0. Pepin, J. A. Eddy, and R. B. Merrill, eds. Pergamon, Elmsford, N.Y., pp. 63-67. Skoog, B. G., J. I. H. Askne, and G. Elgered (l982). Experimental determination of water vapor profiles from ground-based measurements at 2l.0 and 3l.4 GHz. J. Appl. Meteorol. 2l:394. SMIC (l97l). Inadvertent Climate Modification, Study of Man's Impact on Climate. MIT Press, Cambridge, Mass. Stephens, G., G. Campbell, and T. Vonder Haar (l98l). Earth radiation budgets. J. Geophys Res. 86:9739-9760. Taylor, R. A. H. (l980). Snow Survey of the Northern Hemisphere. Glaciological Data, Report GD-9. World Data Center-A for Glaciology, Boulder, Colo., pp. 75-76. Thompson, S. L., and S. H. Schneider (l979). A seasonal zonal energy balance climate model with an interactive lower layer. J. Geophys. Res. 84:240l-2404. Thompson, S. L., and S. H. Schneider (l982). Carbon dioxide and climate: the importance of realistic geography in estimating the transient temperature response. Science 2l7:l03l-l033. U.S. Department of Energy (l980). Proceedings of the Carbon Dioxide and Climate Research Conference. Washington, D.C., April 24-25, l980. Prepared by The Institute for Energy Analysis, Oak Ridge Associated Universities, L. E. Schmitt, ed., Report No. CONF-8004ll0, UC-ll, December l980, 287 pp. U.S. Department of Energy (l982). Proceedings of the Workshop on First Detection of Carbon Dioxide Effects. Harper's Ferry, West Virginia, June 8-l0, l98l. Harry Moses and Michael C. MacCracken, Coordinators. Prepared by The Institute for Energy Analysis, Oak Ridge Associated Universities, N. B. Beatty, ed., Report No. DOE/CONF-8l062l4, UC-ll, May l982, 546 pp. U.S. Department of Energy (l983). Proceedings: Carbon Dioxide Research Conference: Carbon Dioxide, Science and Consensus. September l9-23, l982. Berkeley Springs, West Virginia. Compiled

38l by The Institute for Energy Analysis, Oak Ridge Associated Universities. Report No. CONF-820970, 506 pp. Vilmundarson, T. (l972). Evaluation of historical sources of sea ice near Iceland. In Sea Ice, Proceedings of an International Conference at Reykjavik, Iceland, May l97l, pp. l59-l69. Vinnikov, K. Ya., and P. Ya. Groisman (l98l). The empirical analysis of CO2 influence on the modern changes of the mean annual Northern Hemisphere surface air temperature. Meteorol. Gidrol. ll:35-45. Vinnikov, K. Ya., and P. Ya. Groisman (l982). The empirical study of climate sensitivity. Atmos. Oceanic Phys. l8 (in press). Vinnikov, K. Ya., G. V. Gruza, V. F. Zakharov, A. A. Kirillov, N. P. Kovyneva, and E. Ya. Ran'kova (l980). Current climatic changes in the Northern Hemisphere. Meteorol. Gidrol. 6:5-l7. (English translation by Allerton Press: Sov. Meteorol. Hydrol. 6:l-l0.) Vinnikov, K. Ya., G. V. Gruza, V. F. Zakharov, A. A. Kirillov, N. P. Kovyneva, and E. Ya. Ran'kova (l98l). Modern changes in climate of the northern hemisphere. Meteorol. Gidrol. 6:5-l7. Walsh, J. E., and J. E. Sater (l98l). Monthly and seasonal variability in the ocean-ice-atmosphere systems of the North Pacific and North Atlantic. J. Geophys. Res. 86:7425-7445. Wang, W. C., Y. L. Yung, A. A. Lacis, T. Mo, and J. E. Hansen (l976). Greenhouse effects due to man-made perturbations of trace gases. Science l94:685. Weiss, R. F. (l98l). The temporal and spatial distribution of tropospheric nitrous oxide. J. Geophys. Res. 86:7l85-7l95. Weller, G., and B. Holmgren (l974). The microclimates of the Arctic tundra. J. Appl. Meteorol. l3:854-62. Wetherald, R. T., and S. Manabe (l98l). Influence of seasonal variation upon the sensitivity of a model climate. J. Geophys. Res. 86:ll94. Wigley, T. M. L., and P. D. Jones (l98l). Detecting C02-induced climatic change. Nature 292:205-208. Willett, H. C. (l950). Temperature trends of the past century. Centenary Proceedings l95, Royal Meteorological Society. Willett, H. C. (l974a). Recent statistical evidence in support of the predictive influence of solar-climate cycles. Mon. Wea. Rev. l^2:679-686. Willett, H. C. (l974b). Do recent climate fluctuations portend an imminent ice age? Geofis. Intern. l4:265-302. Willson, R. C., S. Gulkis, M. Janssen, H. S. Hudson, and G. A. Chapman (l98l). Observations of solar irradiance variability. Science 2ll:700-702. World Meteorological Organization (l98l). Joint WMO/ICSU/UNEP Meeting of Experts on the Assessment of the Role of CO9 on Climatic Variations and Their Impact (Villach, Austria, November l980) . World Climate Programme Report No. 3. World Meteorological Organization, Geneva, 35 pp. World Meteorological Organization (l982a). Report of the WMO(CAS)/JCS Meeting of Experts on Detection of Possible Climate Change (Moscow, October 3-6, l982). World Climate Programme Report No. 29. World Meteorological Organization, Geneva, 43 pp.

382 World Meteorological Organization (l982b). The International Satellite Cloud Climatology Project (ISCCP) Preliminary Implementation Plan. World Climate Programme Report No. 35. World Meteorological Organization, Geneva, 85 pp. Wuebbles, D. J., F. M. Luther, and J. E. Penner (l983). Effect of coupled anthropogenic perturbations on stratospheric ozone. J. Geophy. Res. 88:l444-l456. Yamamoto, R. (l980). Change of global climate during recent l00 years. Proceedings of the Technical Conference on Climate—Asia and Western Pacific, Guangzhou, China, l5-20 Dec. l980. World Meteorological Organization, Geneva, pp. 360-375. Zwally, H. J., R. H. Thomas, and R. A. Bindschadler (l98l). Ice-sheet Dynamics by Satellite Laser Altimetry. Technical Memorandum 82l28. National Aeronautics and Space Administration, Goddard Space Flight Center, Greenbelt, Md., May l98l, ll pp. Zwally, H. J., R. A. Bindschadler, A. C. Brenner, T. V. Martin, and R. H. Thomas (l983a). Surface elevation contours of Greenland and Antarctic ice sheets. J. Geophys. Res. 88:l589-l596. Zwally, H. J., C. L. Parkinson, and J. C. Comisco (l983b). Variability of antarctic sea ice and C02 change. Science 220:l005-l0l2.