SCOPE 29 - The Greenhouse Effect, Climatic Change, and Ecosystem, Chapter 6, Empirical Climate Studies

In order to understand, and eventually predict, how increasing atmospheric CO₂ concentration might alter the Earth's climate, both modelling studies and empirical analyses of observational data are required. The current state of the modelling art has been reviewed by Dickinson in Chapter 5. Here we describe a number of empirical approaches to the CO₂ problem. Before beginning, however, we note that empirical studies do not yet distinguish the effects of CO₂ from those of other radiatively active trace gases, such as methane (CH₄), nitrous oxide (N₂O), ozone (O₃) and chlorofluorocarbons (CFCs). These gases are therefore considered together under the broad heading 'greenhouse gases. Since the different gases do have spatially distinct effects on the atmosphere's radiative balance, distinguishing their separate influences is theoretically possible, but this is a problem which has practical difficulties well beyond those applying to the characterization of their combined effects on climate.

Empirical studies of the greenhouse gas problem may be grouped under three broad headings: analysis of past climatic change, production of warm-world scenarios, and the detection of the effects of these radiatively active gases on climate. In Section 6.2 we review past climatic changes and fluctuations. Although the bulk of this section deals with the past 100 years (since this is the period for which we have the most abundant and highest quality data, and it is the period over which most anthropogenic changes have occurred), longer time scale climatic change is also considered in order to place the more recent changes in a proper perspective. Section 6.3 covers the scenario approach to the estimation of future climatic conditions. In Chapter 5, Dickinson has shown that present climate GCMs cannot yet predict the regional details of future climates, largely because of uncertainties in the way these models parameterize (i.e. simplify) many important climate processes, and the fact that they model the oceans so crudely. Because of these model limitations a number of authors have used the past climate record to construct 'scenarios' for the future, i.e. geographically detailed empirical projections of future possibilities. In Section 6.4 we consider the question: has the climatic effect of greenhouse gases been detected in the observational data? Detection of greenhouse gas effects is the ultimate test of model results and so is a research area of considerable importance, one that requires a careful and statistically rigorous approach. Finally, in Section 6.5, we summarize the detection issue and present a number of recommendations for future research.

6.2 PAST CLIMATIC CHANGE

The record of past climate is relevant to the study of anthropogenic CO₂ and trace gas effects for four reasons. First, it provides information on the natural variability of the climate system which allows us to evaluate possible future changes in a broader context. A quantification of this natural variability, which constitutes the background `noise' level above which anthropogenic effects must be identified, is central to the detection problem. Second, in trying to understand the causes of past climatic change, we obtain insights into the way the climate system responds to external forcing and into the system's complex internal interactions and feedback mechanisms. Third, climatic data provide an important source of information for testing climate models, both in the control mode (i.e. in simulating present climatic conditions) and in the perturbed mode (i.e. in simulating past conditions for some assumed change in external forcing). Fourth, some aspects of past climatic change are undoubtedly related to past changes in atmospheric CO₂ level, so the paleoclimatic record provides direct information on the effects of CO₂ forcing.

On the very longest time scales, changes in CO₂ concentration are thought to be one of the primary controls on climatic variability. When the Earth first formed, the Sun was much fainter than today, only about 75% as bright, yet we know that the early Earth was not ice-covered. This 'Faint Sun Paradox' has been explained by the much higher CO₂ concentration that existed then, with an enhanced greenhouse effect counteracting the lower solar output (Owen et al., 1979; Roxburgh, 1980). Life on Earth owes its existence to CO₂, both because of the greenhouse effect and, perhaps, because the chemistry of the oceans in a high-CO₂ world provided a more favourable environment for the chemical precursors of life (Wigley and Brimblecombe, 1981). Since those early days, billions of years ago, it appears that life has had a major effect on atmospheric CO₂ concentration, with photosynthesizing plants slowly reducing CO₂ levels as the Sun's irradiance increased, maintaining approximate homeostasis for the planet. Superimposed on this very long time scale decreasing CO₂ trend, there have been noticeable fluctuations (revealed by analysis of the geological sediment record; see Berner et al., 1983, and Schneider and Londer, 1984, pp. 2417) which almost certainly had major impacts on the climate. For example, the warmth of the Cretaceous period some 100 million years ago, given the vastly different positions of the continents in this era, can most easily be explained by an elevated CO₂ level (Barron and Washington, 1984; Lloyd, 1984). Budyko (1982) has considered the role of CO₂ as a climatogenic factor on these long time scales in considerable detail.

On the 100,000 year time scale, the Earth has experienced a fairly regular series of cold glacial periods which become most apparent in the climate record roughly 1.7 million years ago. Fluctuations are clearly evident prior to 1.7 x 10⁶ BP, but an increase in the amplitude of these fluctuations seems to have begun at that time in association with the appearance of seasonally permanent Arctic ice (Shackleton et al., 1984). These glacial/interglacial cycles occur approximately every 100,000 years. While the causes are not yet known in detail, the primary forcing factor appears to be changes in the Earth's radiation budget due to orbital parameter effects (Milankovitch, 1941; CLIMAP, 1976, 1984; Hays et al., 1976). The most convincing evidence for this is that the main orbital parameter periodicities (at approximately 100,000, 40,000, 23,000 and 19,000 years, see Berger, 1977) are also observed in the proxy climate record from deep sea sediments (albeit with different relative strengths).

In spite of this strong evidence, independent climate models (i.e. those with few empirically tuned parameters) have, to date, been unable to reproduce the observed equilibrium climate using orbital forcing alone (see North et al., 1983, for a recent review). Important feedback mechanisms may be missing in these models and changing CO₂ levels may be a significant factor (Hansen et al., 1984; Pisias and Shackleton, 1984; Kerr, 1984). There is direct evidence to support this possibility, and large glacial/interglacial time scale changes in CO₂ concentration have been measured using fossil CO₂ in bubbles from ice cores (Delmas et al., 1980; Neftel et al., 1982) and inferred from carbon isotope data from deep-sea sediments (Shackleton et al., 1983). Atmospheric CO₂ concentration was around 200 ppmv at the peak of the last glaciation (2015,000 years ago). This low level was probably a response to orbitally forced climate changes in the oceans (Broecker and Takahashi, 1984; Knox and McElroy, 1984; Sarmiento and Toggweiler, 1984; Siegenthaler and Wenk, 1984) and changing CO₂ concentration may constitute an important positive feedback process which amplifies the orbital effects. Oceanically mediated, natural CO₂ changes may operate to amplify climatic change no matter what the forcing mechanism.

The last interglacial/glacial cycle can be considered to have begun around 120,000 BP. The period 120,000 to 80,000 BP (known in Europe as the Eemian) was a time of general warmth, similar to today. This was interrupted by two cold intervals which began and ended relatively abruptly (i.e. over a period of order 100 to 1,000 years) and which lasted for only a few thousand years (see, for example, Shackleton and Opdyke, 1973; Woillard, 1978, 1979). Earlier warm periods tended to last for less than the 40,000 years of the last interglacial, but their durations varied widely. The last interglacial is particularly interesting because, for at least some of this period, mean sea level may have been considerably (around 6 m) above present-day levels indicating a substantially lower ice volume than today. This higher sea level is thought by some to be due to partial melting of the West Antarctic ice sheet (Mercer, 1978). This is an important observation since such an event has also been suggested as a possible consequence of the warming induced by an increased concentration of greenhouse gases (see Robin, Chapter 7, for further details).

The period 70,000 to 15,000 BP was globally cool, with global mean temperature probably ranging between 2 and 5 °C below today's level (Gates, 1976a,b; Peterson et al., 1979; Kutzbach and Guetter, 1984; Hansen et al., 1984). Maximum coldness occurred around 20,000 to 15,000 BP, associated with maximum extents of continental ice masses in the Northern Hemisphere over North America and Eurasia, and maximum extent of sea ice around Antarctica. At this time, mean sea level was roughly 100 m below today's level due to the mass of water locked up in the continental ice sheets.

Although orbital changes are the most likely primary cause of these past glaciations, it is of interest to note that the orbital insolation signature 18,000 years ago was quite similar to that prevailing today. The cold conditions at 18,000 BP are presumed to have developed from earlier orbital changes, and were maintained partly by the large Northern Hemisphere ice masses, by prevailing sea surface temperatures, and by reduced CO₂ levels. The relative importance of these factors is still under study (see, for example, Hansen et al., 1984, and Manabe and Broccoli, 1984). These studies suggest that CO₂ played a relatively minor role in maintaining the glacial equilibrium state, but CO₂ changes may well have been more important in initiating glacial/interglacial climatic changes (Knox and McElroy, 1984; Pisias and Shackleton, 1984). The 70,000 to 15,000 BP interval, while generally much colder than today, was, nevertheless, punctuated by a number of rapid warming and cooling events. Of particular relevance to the CO₂ issue are a series of fluctuations between about 33,000 and 22,000 BP, when ice core data show that virtually simultaneous, rapid changes in both climate and CO₂ level occurred (Stauffer et al., 1984).

The period 15,000 to 10,000 BP is known as the late-glacial period. The recovery from full-glacial conditions was geographically and temporally complex. Warming appears to have begun earlier in the Southern Hemisphere. In the Northern Hemisphere, particularly around the eastern side of the North Atlantic, very rapid warming occurred around 13,000 BP. Warming in western Europe was so rapid that conditions were almost as warm at 12,700 BP as today. This relatively warm period prevailed until around 11,000 BP when a sharp and severe cooling episode occurred. Warm conditions returned with a rapid warming around 10,000 BP. Over a period of only a few centuries, winter temperatures warmed by around 15 °C and summer temperatures by 8 °C, mirroring the ~13,000 BP warming event (Atkinson et al., 1985). These rapid changes in European climate appear to have occurred in phase with similar changes in Greenland (based on the oxygen isotope record from the Dye 3 ice core; see Siegenthaler et al., 1984, for details) and may relate to changes in the influx of cold fresh water into the North Atlantic from the melting continental ice sheets (Mercer, 1969; Ruddiman and McIntyre, 1981) and changes in the rate of formation of deep water in the North Atlantic (Broecker et al., 1985). There is ice core evidence that changes in atmospheric CO₂ concentration may have occurred at about the same time as these rapid late-glacial climatic changes (Oeschger et al., 1984).

Although these rapid changes in climate were probably not global in extent (Mercer, 1969; Broecker et al., 1985), the possible contemporaneous CO₂ changes and the link with changes in the production rate of North Atlantic Deep Water (NADW), make them highly relevant to the anthropogenic CO₂ issue. Because NADW represents a sink of water at ~2 °C in a region where the balancing influx of water has a considerably higher temperature, the net effect of NADW formation is to provide a major heat source for the atmosphere. Lower NADW formation rate should, therefore, result in atmospheric cooling (for further details see Hoffert et al., 1985, and Watts, 1985a,b). This same ocean circulation change should also, on the basis of carbon cycle considerations, result in a reduction in atmospheric CO₂ concentration. Broecker et al. (1985) hypothesize that NADW formation rate changes are the main mechanism for the observed rapid late-glacial climatic fluctuations. Similar NADW changes may well occur today (indeed, there is observational evidence for such changes). Although the magnitude of the changes possible today would probably be less than in the late-glacial, they could be of considerable importance in modifying any anthropogenic greenhouse gas-induced changes in climate (Broecker et al., 1985). NADW effects are discussed further in Section 6.4.4.

By 10,000 BP, the globe as a whole had warmed to approximately its present-day mean temperature, but large ice masses still existed in the Northern Hemisphere. In Scandinavia, ice retreat, which had begun at around 13,000 BP, was particularly rapid from 109,000 BP and ice had virtually disappeared by 8,000 BP. On the North American continent, the ice retreat chronology was initially similar to this, but a substantial amount of ice was left at 8,000 BP. Between approximately 8,000 and 7,500 BP, the North American ice sheet became two separate ice sheets. At 6,000 BP, only three small ice vestiges were left in northern Canada and these soon disappeared. Ice retreat in Eurasia and North America, however, was nowhere smooth, and details of retreat timing are still sketchy (Andrews and Barry, 1978). It is likely that the decay was not simply by melting, but also by mechanical processes such as surging and rapid movement along ice streams (Denton and Hughes, 1981).

The period 10,000 BP to the present is known as the Holocene. The early Holocene (9,000-6,000 BP) was a time of possibly global and certainly regional warmth compared with today (Webb and Wigley, 1985). For example, at 6,500 BP, in spite of the existence of small residual ice masses, forests were present significantly further north in parts of North America compared with their limits today. In addition, many of the world's desert regions were substantially wetter than today, with fewer sand dunes, more vegetation and higher lake levels (Sarnthein, 1978; Street and Grove, 1979; Kutzbach and Street-Perrott, 1985).

This early Holocene warm period has been used as an analogue for a CO₂-induced warmer world (see Section 6.3.2). The cause of this warmth appears to be, at least in part, orbital variations (Kutzbach, 1981; Kutzbach and Otto-Bliesner, 1982; Kutzbach and Guetter, 1984; Kutzbach and Street-Perrott, 1985). At 9,000 and 6,000 BP, the spatial and seasonal distribution of solar radiation input was quite different from that prevailing today. In the Northern Hemisphere, July insolation at 9,000 BP was 8% above the present-day level (+6% at 6,000 BP), while Northern Hemisphere January insolation values were below those prevailing today by roughly the same amount. There is some (debatable) evidence of CO₂ levels higher than the nineteenth century 'pre-industrial' level (Neftel et al., 1982).

While the whole of the Holocene was a time of warmth relative to the previous 6070,000 years, global climate was far from static during this period. Many warming and cooling episodes occurred on the 100-year timescale. Because these events are only observable through indirect or proxy evidence, which is both local and often poorly dated, we do not know how global mean climate changed, or even if global mean temperature has changed since, say, 6,000 BP. Regionally, however, substantial changes have occurred, as evidenced by vegetation changes, glacial fluctuations, isotopic data from ice cores, and many other proxy climate indicators.

Over the last 2,000 years the most widespread changes observed in the climate record are those associated with the Medieval Warm Epoch (approximately 800 to 1200 AD) and the Little Ice Age (approximately 1400 to 1800 AD). The former may well be restricted to the North Atlantic Basin region, and, even over this large area, changes in different locations show large differences in timing on century and shorter time scales (Williams and Wigley, 1983). The Little Ice Age, however, seems to have been more nearly global, but, again, substantial differences can be seen when time series from different regions are compared. The cause of the Little Ice Age is not known. Changes in solar activity associated with the period of minimum sunspot activity known as the Maunder Minimum have been suggested (Eddy, 1976a,b), but the evidence is equivocal. Another possible causal factor is a change in the frequency of explosive volcanic activity. Some support for this hypothesis is available from ice cores (Hammer et al., 1980; Mosley-Thompson and Thompson, 1982), but even this evidence is open to question. There is also ice core evidence that atmospheric CO₂ levels may have been depressed during the Little Ice Age (Raynaud and Barnola, 1985), and atmospheric methane concentration was apparently much less at this time than during the twentieth century, perhaps as low as 700 ppbv compared with a 1980 level of around 1650 ppbv (Craig and Chou, 1982).

Suggestions of CO₂ effects on climate can be found on almost all time scales in the past climatic record. The evidence, however, is insufficient to quantify the CO₂ influence. It is only for the past century or so, the era of instrumental meteorology, that we have sufficient and suitable data to be able to make meaningful quantitative analyses. The potential effects of CO₂ and other trace gases embrace the whole climate system; temperatures, precipitation, cloudiness, winds, and so on. Here we consider recent changes in the most well documented weather variables, temperature and precipitation, in order to provide a background to the discussion of the detection issue given in Section 6.4.

It is only since about 1850 AD that we can begin to make detailed quantitative statements about past climatic change. However, the confidence that can be placed in such statements depends critically on the particular variable considered, on the size of the area covered, and on location. Precipitation, temperature and surface (or mean sea level) pressure data extend back into the nineteenth century and earlier, but spatial coverage is poor in the early years of these records (Lamb, 1977; Bradley and Jones, 1985). Even today, spatial coverage is relatively poor over the oceans, particularly in the southeastern Pacific and in high southern latitudes. This deficiency may eventually be overcome using satellite data. Plentiful upper air data, essential in obtaining a three-dimensional picture of climate, exist only since the late 1940s.

In describing past temperature or precipitation changes it is of greatest value to consider large scale area averages. Local or small scale regional changes may be of relatively minor consequence in understanding the climate system as a whole. Moreover, in the present context, the uncertainty in the expected signal due to CO₂ and other trace gases increases as the spatial scale decreases. For these reasons, we concentrate here on large scale spatial averages. Greatest attention will also be given to temperature, since this is the best documented climate variable and since the signal for temperature is always likely to be better established than that for other variables.

Although this had been attempted earlier by Köppen (1873, 1914), the first reasonably reliable estimates of large scale area average temperatures were those made by Callendar (1961), Mitchell (1961, 1963) and Budyko (1969). These and other works have been reviewed by a number of authors (e.g. Jones et al., 1982; Wallén, 1984; Ellsaesser et al., 1985; Wigley et al., 1985). The best documented records are those of Jones et al. (1982, 1986a,b; see also Kelly et al., 1982, and Raper et al., 1984). At least four other groups have recently published important work in this area. For convenience, these five groups will be referred to below as Jones (Jones et al., 1982, 1986a,b,c; Kelly et al., 1982; Raper et al., 1983, 1984), Vinnikov (Borzenkova et al., 1976; Vinnikov et al., 1980), Hansen (Hansen et al., 1981), Yamamoto (Yamamoto, 1981; Yamamoto and Hoshiai, 1979, 1980) and Folland (Folland and Kates, 1984; Folland et al., 1984a,b). The work of Folland and colleagues and of Jones et al. (1986c) differs from the others in using marine (i.e. ship-based) data. The other analyses are based on land station data with sparse coverage over ocean areas. Only Jones et al. (1986c) have produced averages which incorporate both land and marine data.

Large scale averages for the land masses are based on individual station observations. Such records may be influenced by various sources of inhomogeneity. Variations may be caused by non-climatic factors such as site changes, changes in the type of instrument and/or its exposure, changes in observation times and/or the method of producing daily and monthly averages, or local anthropogenic effects such as urban warming (see Mitchell, 1953, and Bradley and Jones, 1985). Provided the data are carefully screened for possible in homogeneities, these problems do not appear to have a noticeable effect on large scale averages (Jones et al., 1986a). The reliability of these averages will be discussed further below.

Three of the four land-based studies give extremely similar results, a necessary consequence of their reliance on the same basic data sources. Minor differences exist due to small differences in the area of coverage, the data sources and the different methods of producing area averages. In the case of Yamamoto, averages were based on the incorrect assumption that no changes had occurred in regions with no data, so their results show much smaller amplitude changes and are not directly comparable with the others. Of the analyses of land-based data, only Hansen claims an estimate of global mean temperature. Since data coverage in the Southern Hemisphere is extremely poor (approximately 80% is ocean), this claim is difficult to justify.

The Northern Hemisphere land-based records of Jones, Vinnikov and Hansen are shown in Figure 6.1. The three curves correlate highly on annual and longer time scales and show changes of similar amplitude. The general trends of warming to around 1940, cooling to the mid-1960s, and warming since around 1970, are common to all analyses. When individual seasons are considered, similar trends can be seen, but with inter-seasonal phase differences of a decade or more (see Jones et al., 1982, and Wigley et al., 1985). The curves in Figure 6.1 have often been used as indicators of global mean changes. While a true global average can only be obtained by including data from the ocean areas and all of the Southern Hemisphere, recent comparisons of land and ocean data and of Southern and Northern Hemisphere data show remarkable parallels (Wigley et al., 1985; Jones et al., 1986b; see also Figure 6.2).

A number of workers have attempted to study ocean, or ocean-plus-land temperature changes (Folland and co-workers; Barnett, 1978, 1984; Chen, 1982; Paltridge and Woodruff, 1981; Fletcher, 1984; Jones et al., 1986c). There are, however, great difficulties in obtaining a strictly homogeneous time series of ship-based data due to changes in instrumentation and problems related to instrument exposure (reviewed by Barnett, 1984, 1985a, and by Folland et al., 1984a). Either sea surface temperatures (SSTs) or shipbased (marine) air temperatures (MATS) may be used; but both types of data require adjustments to produce a homogeneous record. SST and MAT data are highly correlated.

Figure 6.1 Comparison of three estimates of Northern Hemisphere temperature changes. The top curve is from Jones et al. (1982) and the middle curve is from Vinnikov et al. (1980). The lower curve is Hansen et al.'s (1981) data for the whole hemisphere (personal communication, J. E. Hansen). In Hansen et al. (1981), only the average for 23.690° N is given, and the data are smoothed with a 5-year running mean. The curves given here show both annual mean values and values smoothed with a 10-year Gaussian filter (padded at the ends to cover the whole period of record)

In spite of these instrumentation problems, the marine data appear to be reliable at least back to the early twentieth century. The MAT data correlate highly with the land-based data (Jones et al., 1986a,b,c). Such high correlations are expected on physical grounds (Wigley et al., 1985). The fact that they occur between independently derived data sets and are similar in both hemispheres is evidence supporting the reliability of the data.

Figure 6.2 Northern and Southern Hemisphere annual mean surface and tropospheric air temperature changes. Top curve (NHT): land-based record from Jones et al. (1986a), a revision and extension of the data given by Jones et al. (1982). Second and fourth curves (NHNMAT and SHNMAT); Folland et al.'s night-time marine air temperature data (data from C. K. Folland and D. E. Parker, personal communication). Third and fifth curves (NH(850300) and SH(850300)); Angell and Korshover's 850300 mb column mean temperatures (data from J. K. Angell, personal communication). Smooth curves are 10-year Gaussian filtered values. Note that the NMAT values in the early 1940s may be too high due to measurement uncertainties (see Folland et al., 1984a)

In Figure 6.2 we compare Folland et al.'s night-time marine air temperature (NMAT) records for both hemispheres with the Jones et al. (1986a) land-based Northern Hemisphere record. Similarities between the hemispheres on decadal and longer time scales are clearly evident. Both hemispheres show general warming from the late nineteenth century to around 1940 and a decline to the mid-1960s. Since the mid-1960s, the globe as a whole appears to have warmed substantially. However, this warming trend seems to have been delayed in the Northern Hemisphere (see Figure 6.2 and also Jones et al., 1986c). The reason for this difference between the hemispheric trends is not yet known (but see Section 6.4.4).

Prior to about 1900, marine data coverage was generally less extensive, although it was still good in some periods. Because of the greater spatial and temporal autocorrelation of marine temperatures, data requirements are less than over land. Nevertheless, the Northern Hemisphere land data correlate less well with Folland et al.'s marine data in the nineteenth century than in the twentieth century. A visual comparison of the land-based and marine-based (NMAT) Northern Hemisphere time series in Figure 6.2 shows that, as one moves back in time, the curves begin to diverge around the beginning of the twentieth century, although they show similar decadal time scale fluctuations back to around 1875. Similar discrepancies exist between the Southern Hemisphere land and marine data (see Jones et al., 1986b). In view of the overall similarity for the twentieth century data, we must suspect that either the land record or marine record is unreliable prior to about 1900. Lowering the NMAT data by about 0.3 °C would largely remove the discrepancy. This issue is discussed in more detail by Jones et al. (1986a,b,c).

If one accepts the reliability of the land data in the nineteenth century, then, for Folland et al.'s long time scale changes to be correct, there would have to be quite large differences between the temperature trends over land and sea in the nineteenth century, but not in the twentieth century. Although this seems unlikely, the possibility of differences in land and marine temperature trends requires further comment, since it has also been suggested by the work of Paltridge and Woodruff (1981). These authors showed an apparent lag of twentieth century temperature changes over the oceans behind those over the land by approximately 20 years. The reality of this lag is highly suspect because Paltridge and Woodruff used uncorrected marine data. Use of the more carefully homogenized twentieth century data fails to reveal any noticeable lag between SST data and either MAT data or Northern Hemisphere land-based data.

Jones et al. (1986c) argue that the land-based data are intrinsically more reliable than the marine data and that the best way to correct the marine data is by reference to the land-based data. In making such corrections they have produced hemispheric-mean MAT and SST time series that differ noticeably from those of Folland et al. on time scales greater than decadal, especially prior to 1900. Their global mean temperature curve is shown in Figure 6.3. It should be noted that there is still some dispute over the pre-1900 temperature trends because of the differences between the marine time series of Jones et al. (1986c) and Folland et al (1984a,b). These differences have yet to be fully resolved.

Figure 6.3 Global mean annual surface temperature changes from Jones et al. (1986c). Smooth curve shows 10-year Gaussian filtered values.

By 'upper air', we mean the whole of the atmosphere, although useful observations are only available for the troposphere and low to middle stratosphere. World-wide, three-dimensional observations of the atmosphere began only in the late 1940s.

Upper air temperatures may be obtained either directly from temperature observations at the mandatory constant-pressure observing levels, or indirectly from thickness data (i.e. the height difference between constant pressure levels, which is proportional to the column mean temperature). Both types of data, especially the former, are affected by changes in instruments and changes in correction procedures, both of which may produce spurious non-climatic trends and variations (Bradley and Jones, 1985). Large scale area averages of these data may be based either on station records or on data interpolated on to a regular grid. Additional inhomogeneities in the latter data may arise through changes in the gridding analysis procedure.

Many analyses of upper air temperatures have used gridded thickness data. These analyses (reviewed in Wigley et al., 1985) show little evidence of any long-term temperature trend. In some records, the main event is a rapid drop in tropospheric temperatures around 1963/64. This has been associated with the eruption of the volcano Mt. Agung by a number of workers. However, the evidence is equivocal (see Parker, 1985) and the drop in temperature may be partly due to changes in analysis procedures (see below).

In this regard, Parker (1980) has exposed a major problem with grid-point thickness data. Values determined by different meteorological agencies are in considerable disagreement, particularly over the oceans and in the subtropics. The differences, which represent a basic data uncertainty, are of similar magnitude to inferred temperature fluctuations based on the individual agency analyses. They can be attributed partly to instrumental problems and partly to changes in the methods of analysis. Grid-point analyses were originally produced manually and are currently produced by interpolation techniques that use model-generated error statistics. Recent data are, therefore, to some degree model dependent (see, for example, Leith, 1984). Because of these problems with grid-point data, we concentrate here on the results of station data analyses.

Extensive analyses of station data based on a 63-station network distributed fairly evenly over the globe have been published by Angell and Korshover (1977; 1978a,b; 1983a,b). Their analyses probably avoid most of the problems associated with grid-point data pointed out by Parker (1980), but there is some doubt about the representativeness of global (or large area) averages because of the limited station network (Oort, 1978). Furthermore, with a limited station network, spurious trends may arise if significant changes occur in the long wave pattern of the general circulation (see Parker, 1981).

Angell and Korshover's hemispheric mean values for the 850300 mb layer are shown in Figure 6.2, where they can be compared with the surface data discussed above. The correlations between the surface and these middle tropospheric air temperature changes are striking (r = 0.80 with Northern Hemisphere NMAT and r = 0.81 with Southern Hemisphere NMAT), especially given the fairly limited upper air station network.

Angell and Korshover also document stratospheric temperature changes. Figure 6.4 shows the global mean change for the 10050 mb layer, with their 850300 mb global mean shown for comparison. None of their results up to around 20 km (50 mb) show any strong trends. Above 20 km there are strong cooling trends, but data coverage is restricted to only part of the Northern Hemisphere and even these data may have severe instrumentation problems (Wigley et al., 1985). The lack of clear trends in lower stratospheric temperatures may be partly the result of the shortness of the record (less than 30 years) and the quite large decadal time scale variability (see also Figures 6.2 and 6.4). Even if a CO₂-induced trend did exist, it would be obscured by this medium time scale noise because of the shortness of the record.

Figure 6.4 Global mean temperature changes for 850300 mb (bottom) and 10050 mb (top) from Angell and Korshover (data from J. K. Angell, personal communication)

Model results suggest that increasing CO₂ may cause an increase in the intensity of the global hydrologic cycle and, therefore, a slight increase in global mean precipitation. Estimates of the global mean change in precipitation due to a doubling of CO₂ vary widely, from +3% (Manabe and Stouffer, 1980) to + 11 % (Rind and Lebedeff, 1984). Even the lower value may be an overestimate because of deficiencies in the way current climate models model the oceans, a point noted by Stone (1984). Regional changes are expected to differ noticeably from the global mean. Although the details of these future changes are unknown, there is strong evidence for increases in precipitation in high latitudes (see Chapter 5). There is also evidence for possible increases in mid-latitude dryness (Manabe et al., 1981), but these changes are not necessarily associated with reductions in precipitation.

Little work has been carried out on temporal changes in large area averages of precipitation for a number of reasons. The spatial variability of precipitation is considerably more than for temperature and it is generally thought that a much denser observational network is required in order to establish area averages. In addition, precipitation is subject to more measurement difficulties than temperature. Precipitation time series often show inhomogeneities; for a review of the problems see Rodda (1969) and Barnett (1985b). Precipitation also tends to be highly variable from year-to-year. Global mean estimates cannot be made because of the sparseness of data coverage over the oceans and because of difficulties in using ship-based precipitation data. These latter problems, in the context of estimating long-term mean precipitation values, have been discussed by a number of authors (Dorman and Bourke, 1979, 1981; Reed and Elliott, 1977, 1979; Reed, 1979; Elliott and Reed, 1984). No work has been published on changes in precipitation over the oceans. There have, however, been a number of important regional studies for continental areas (e.g. Diaz, 1981; Diaz and Fulbright, 1981; Fleer, 1981; Nicholson, 1980, 1981; Tabony, 1981).

Although precipitation data show high spatial variability, inter-annual precipitation variations do show large scale (order 1,000 km or more) coherent patterns which account for significant amounts of overall precipitation variance (Barnett, 1985b). The most spatially comprehensive studies of precipitation (i.e. those taking a global or, at least, hemispheric view) have been the works of Corona (1978, 1979), Gruza and Apasova (1981; see also Angell and Gruza, 1984) and Barnett (1985b). These studies have shown that large area average precipitation data exhibit considerable medium time scale variability (order 10 years) together with a few significant long time scale trends (order 100 years). For example, Barnett (1985b) notes long-term trends of increasing precipitation over Europe/western Asia and over India, and a long-term decreasing trend over Africa. Barnett's analysis is only preliminary, however, and is based on rather limited numbers of stations with little quality control.

Barnett concludes that there is no evidence of any overall (i.e. global mean) trend and that, if any global mean trend were to result from increasing CO₂, it would almost certainly fail to reach statistical significance due to the high inter-annual and inter-decadal variability of precipitation data. The lack of any overall trend is supported by the work of Gruza and Apasova (1981), although these authors note a slight increase in Northern Hemisphere continental precipitation in January and a slight decrease in July over the past century. These studies of hemispheric-scale precipitation trends, however, have deficiencies, the major problem being that the data used have not been critically examined for homogeneity.

6.3 CLIMATE SCENARIOS

In order to assess the effects of changes in climate on Man's activities, we require detailed regionally and seasonally specific simulations of future anthropogenic climatic change. Such information may be obtained either by modelling the future climate with appropriate general circulation climate models (see Chapter 5) or by using past climate data to provide analogues for the future. Both approaches have limitations. At their present stage of development, climate models cannot produce reliable predictions of climatic change at the regional and seasonal level. Similarly, because we do not yet fully understand the causes of past changes in climate, analogues for the future based on the past cannot be considered as reliable predictions. Nevertheless, both methods do produce internally consistent representations of climatic conditions that could reasonably be expected to occur in a highgreenhouse-gas world. Such simulations are referred to as scenarios. Climate model results are discussed in detail in Chapter 5. In this section, we consider scenarios based on past climate data, either data from the distant past (paleoclimatic analogues) or data from the twentieth century (instrumentally based scenarios).

The value of these analogue scenarios is based partly on the argument that past climate patterns can be used as future climate analogues, even if the causes leading to these past patterns are unknown. The underlying assumption is that, for similar atmospheric boundary conditions (as represented by the oceans, the land surface and the cryosphere), the general circulation of the lower atmosphere responds in a similar way to different forcing mechanisms (Wigley et al., 1980). This assumption can be tested using both model results and empirical data (see Section 6.3.4).

Pittock and Salinger (1982) have distinguished three different approaches to analogue-based scenario development. The first is to use a suitably defined ensemble of warm years from the recent instrumental record and to compare this, either with the long-term mean, or with a similarly defined cold-year ensemble (Wigley et al., 1980; Williams, 1980; Namias, 1980). The second is to use regional reconstructions of paleoclimate during past warm periods. Flohn (1977) has suggested a number of possible paleo climatic analogue periods: the Medieval Warm Epoch (c. 8001200 AD); the time of maximum early Holocene warmth (c. 6,0009,000 BP, referred to variously as the Altithermal, the Hypsithermal or the post-glacial climatic optimum); and the last (Eemian) interglacial around 100,000 BP. A third alternative is to use atmospheric dynamical arguments together with a knowledge of empirical climate relationships and correlations to develop an educated guess (Bryson, 1974; Flohn, 1979; Pittock and Salinger, 1982). Of these possibilities, scenarios based on instrumental data have been most extensively developed.

A number of papers have been written about the early Holocene analogue, notably by Kellogg (1977, 1978), Kellogg and Schware (1981) and Butzer (1980). The early Holocene is generally thought to have been a globally warm period, but this could be open to doubt. Conditions were up to 2 °C warmer than today in many regions of the globe, but the global mean temperature cannot be reliably estimated because of incomplete data coverage. Until recently (see Webb and Wigley, 1985), there has been an insufficient amount of well-dated, reliably interpreted paleoclimatic data from this period for much confidence to be placed in derived analogues. Scenarios that have been produced have generally used the available data uncritically, particularly with regard to dating uncertainties. Furthermore, the seasonal and spatial distribution of incoming solar radiation in the early Holocene differed radically from today (see Section 6.2.3) and the boundary conditions for the atmospheric circulation were also substantially different, particularly in North America where the decaying Laurentide ice sheet did not finally disappear until around 6,000 BP. Clearly, with such large spatial differences in solar forcing and with large residual ice masses, the early Holocene is hard to justify as a suitable analogue for a warm, high-greenhouse-gas world.

The other two possible paleoclimatic warm-world analogues mentioned above (the Medieval Warm Epoch and the Eemian) have not been extensively discussed in the literature, largely because of incomplete data coverage. Williams and Wigley (1983) have shown that the former period was not a time of spatially uniform warmth even in the Northern Hemisphere. Warm conditions (relative to today) were largely confined to the North Atlantic Basin region and, even in this region, warm intervals were interspersed by shorter time scale (<100 yr) changes in climate which show little large scale spatial coherence (see Section 6.2.4).

Flohn (1980, 1981) has considered the Eemian as a warm-world analogue, and the recently published CLIMAP (1984) data may allow more detailed interpretations of this period. The main interest in this period stems from the evidence of higher sea levels that probably prevailed at this time. The period may, therefore, provide insight into the possibility of partial melting of the Antarctic ice mass, an event which has been suggested as a possible extreme future consequence of global warming (Mercer, 1978).

More distant analogues for a future warmer world have been considered by Flohn (1980, 1981) in trying to assess the consequences of complete melting of the Arctic ice. The possibility that atmospheric CO₂ concentration increases might cause the Arctic ice to disappear, at least seasonally, was first suggested by Budyko (1962, 1969). Complete disappearance is now thought to be unlikely for any anticipated future CO₂ level, but the possibility cannot be ruled out entirely because all studies of the problem have been based on over-simplified models (see Chapter 5, and also Semtner, 1984).

The basic method is to select a set of warm years and a set of cold years and to produce spatially detailed composites of the differences in pressure, temperature and precipitation between the two sets. (Some workers have used the long-term mean as a baseline rather than the mean of a set of cold years.) In this way, scenarios can be produced down to the monthly time scale (although published work has considered only seasonal or annual scenarios).

There are a number of different factors which should be considered when selecting years for scenario construction, imposed by the need to simulate, as closely as possible, conditions in a warmer world (see Palutikof et al., 1984). Scenarios based on composites of isolated individual warm and/or cold years, the method used by Wigley et al. (1980), Williams (1980), Namias (1980) and Pittock and Salinger (1982), are not compatible with the slow evolution of anthropogenic climatic change. A better way to simulate the effects of the gradual atmospheric greenhouse gas increase is to use blocks of warm and cold years rather than individual extreme years. Studies by Jäger and Kellogg (1983) and Palutikof et al. (1984) show that scenarios obtained by compositing data from isolated individual years do differ from those based on groups of consecutive years.

Even when groups of consecutive years are considered, an appropriate data-set must be chosen from which to select the years. Scenarios should ideally be based on differences between past warm and cold periods, where warm and cold apply to the global average surface air temperature. In the absence of global mean data, the years used to construct most published scenarios have been based on Northern Hemisphere land data, either the hemispheric or high-latitude, annual or winter mean values. The various possibilities have been discussed and compared by Palutikof et al. (1984).

An alternative approach to instrumental scenario construction has been employed by Budyko et al. (1978), Groisman (1981) and Vinnikov and Kovyneva (1983), based partly on earlier work by Drozdov (1966, 1974) and on the empirical modelling results of Vinnikov and Groisman (1979) and Groisman (1979). The method used is to derive linear relationships between local, seasonally specific climate data and the Northern Hemisphere surface air temperature (17.587.5° N) record of Borzenkova et al. (1976) and Vinnikov et al. (1980). This method may be expressed as

where C_i is the year-i value of a chosen local climate variable (e.g. temperature or precipitation), T_i is the Northern Hemisphere temperature averaged over the 12 preceding months, a and b are regression coefficients and e_i is an error term. Having determined a and b, the change in the chosen climate variable can be estimated for any given Northern Hemisphere temperature change, and regionally and seasonally specific scenarios can be developed.

The results of this approach depend on the period over which the regression equations are developed, and Vinnikov, Groisman and colleagues' use of as long a period as possible is clearly advisable. However, for the case where C_i is local temperature, Brinkmann (1979) and Jones and Kelly (1983) have shown that the correlation coefficient between C_i and T_i (and, hence, the regression coefficient, a, in equ. (1)) varies with time. This temporal instability casts doubt on the validity of equ. (1) as a forecasting tool.

The most detailed scenarios produced to date are those of Lough et al. (1983) and Palutikof et al. (1984). These workers place greatest confidence on scenarios which exploit the early twentieth century warming. Warm and cold year composites were calculated using the warmest and coldest twentyyear periods from the Northern Hemisphere surface air temperature record of Jones et al. (1982), viz. 193453 and 190120. (Note that there was also a warming trend in the Southern Hemisphere over this period, see Figure 6.2.) Lough et al. (1983) present regional scenarios for Europe and discuss the implications of these for energy demand and for agriculture. Their most surprising result is that winter temperatures over a substantial part of Europe were colder and showed greater inter-annual variability during the warm period, probably as a result of increased blocking (Figure 6.5). Rainfall patterns showed overall decreases in spring and summer, and increases in autumn and winter.

Further details of these European scenarios are given by Palutikof et al. (1984), along with similarly derived scenarios for North America. The North American scenarios for temperature and pressure exhibit much less interseasonal contrast than is the case for Europe. Temperatures are shown to be generally higher and less variable throughout the year in a warm world, although there is a band of cooler conditions which runs across the continent between about 50° and 60°N in all seasons. Most of the continent south of 50°N shows considerable warming, especially in summer (Figure 6.6). Precipitation patterns are complex with substantial areas of increase and decrease (Figure 6.6).

Figure 6.5 Winter pressure (upper), temperature (middle) and precipitation (lower) scenarios for Europe showing higher pressure in the north (pressure changes in mb) and a band of cooler conditions across central Europe. Precipitation changes show an irregular pattern (shown in multiples of the local standard deviation, s). The values shown here are differences between the 193453 and 190120 mean values and correspond to a warming of the Northern Hemisphere of about 0.5 °C. (From Palutikof et al., 1984)

Figure 6.6 Summer pressure (upper), temperature (middle) and precipitation (lower) scenarios for North America. Pressure (changes in mb) is generally lower over the continent. The main part of the continent shows substantial warming; but a band of cooling exists between 50 and 60^° N. The precipitation changes are shown in multiples of the local standard deviation (s). The dots on the lower diagram show the locations of precipitation stations used. The values shown here are differences between the 193453 and 190120 mean values and correspond to a warming of the Northern Hemisphere of about 0.5 °C. (From Palutikof et al., 1984)

A disadvantage of the instrumental scenario method is that the scenarios are based on relatively small temperature changes compared with those expected to result from future increases in atmospheric CO₂ and other trace gas concentrations. Furthermore, a CO₂-doubling is likely to produce substantial changes in the oceanic and cryospheric boundary conditions, changes which may be well beyond the range experienced in the twentieth century. Thus, instrumental scenarios can only be taken as indicative of conditions during the early phase of a warming, changes which are expected to take place by the early decades of the twentyfirst century.

As noted above, instrumentally based scenarios have been justified on the grounds that, no matter what the cause of a change in climate, the spatial patterns of surface changes will be approximately the same. There is evidence both for and against this assumption. Some of this material also relates to the relevance of equilibrium GCM results as scenarios for a future changed climate.

The equilibrium modelling studies of Manabe and Wetherald (1980) and Hansen et al. (1984) provide the most direct supporting evidence. They applied two different forcing mechanisms to their general circulation models: an increased carbon dioxide concentration and an increased value of the solar constant. In both studies, the latitudinal character of the near-surface response of the model climate was similar for both types of forcing.

Similarities between equilibrium modelling results and the patterns of recent surface air temperature changes provide empirical support for instrumentally based climate scenarios. In the Northern Hemisphere, the latitudinal and seasonal patterns of the early twentieth century warming are remarkably similar to the patterns predicted by the equilibrium response general circulation model study of Manabe and Stouffer (1979, 1980) for a quadrupling of CO₂ concentration (see Wigley and Jones, 1981, their Figure 6.3). If this similarity were a result of early twentieth century CO₂ forcing then it may indicate that equilibrium models can give realistic results, at least when one considers seasonal data and zonal averages covering around 30° of latitude. If CO₂ change were not the dominant forcing mechanism then this similarity supports the idea that the latitudinal patterns of climatic change are largely independent of the type of forcing.

Evidence against the value of instrumentally based scenarios can be found in the observational record. Work by Jones and Kelly (1983) indicates that the regional details of a scenario may depend strongly on the time period(s) used for scenario construction. These authors compared the patterns of annual mean temperature change during the early twentieth century warming with the patterns of change during two other periods, the cooling from 1940 to the mid-1960s and the warming that occurred subsequently. They found noticeable differences between these periods. These differences arise mainly from longitudinal changes in regions of warming and cooling which in turn reflect changes in meridional heat transport associated with changes in the stationary and transient eddies of the general circulation.

Contrary evidence also comes from modelling studies of the transient response of the climate system. Most model studies of CO₂ effects, including the work cited above, have considered only the steady-state response to a specified step-function increase in atmospheric CO₂ concentration. In reality, CO₂ (and trace gas) levels are changing continually and, because of spatial variations in the response time (Hansen et al., 1984), the spatial patterns of the transient response to steadily increasing CO₂ may differ from those for the steady-state response to a step function CO₂ change. This was first pointed out by Schneider and Thompson (1981). Bryan et al. (1982, 1984) and Spelman and Manabe (1984) have considered this possibility in greater detail. They used coupled ocean-atmosphere general circulation models to show that the latitudinal signature of the transient response was similar to the steady-state response, provided that the rate of change of CO₂ was not too rapid. Thompson and Schneider (1982), however, in a response to the Bryan et al. (1982) paper, still attest that the general spatial character of the transient response (i.e. including longitudinal detail) might differ noticeably from the steady-state response.

These results are doubly important because they determine, not only the validity of instrumental climate scenarios, but also the relevance of equilibrium GCM results as warm-world scenarios. If, as transient response studies suggest, the patterns of climatic change for time-varying forcing differ from those for the equilibrium response, then the above conclusions regarding the effects of different forcing mechanisms based on equilibrium model studies may simply not be relevant to the case of a continual increase in greenhouse gas concentrations. Furthermore, equilibrium GCM experiments may, themselves, produce information which is largely irrelevant as far as the geographical detail of future anthropogenic climatic change is concerned. Of course, equilibrium model studies can provide valuable physical insights into the models themselves and hence, indirectly, into the possible future behaviour of the climate system at the regional level.

In summary then, the scenario approach (both analogue-based and equilibrium-model-based) is neither convincingly supported nor contradicted by available theoretical evidence or observational data. It is pertinent to recall, however, that scenarios are not meant to be predictions of future climate; rather they are meant to be internally consistent pictures of a plausible future climate, a basis for other workers to evaluate the possible impacts of climatic change on Man and society.

6.4 DETECTION OF CLIMATIC CHANGE

A doubling of atmospheric CO₂ concentration is expected to raise the global annual mean surface air temperature by 1.5 to 5.5 °C (see Chapter 5). As indicated by this range of values, there is considerable uncertainty about the magnitude of the CO₂ warming effect. Detection of such a warming and other associated greenhouse-gas-induced changes of climate in the observational record has, therefore, become a high priority issue (MacCracken and Moses, 1982; Kellogg and Bojkov, 1982). Detection would not only test model predictions, but would, ideally, also allow one to place reliable, empirically determined bounds on the magnitude of the greenhouse effect.

As noted earlier, the concentrations of other radiatively active trace gases are also increasing. These increases will add to the CO₂ effect and recent estimates suggest that, over the past few decades and in the future, the trace gas contribution to global warming may be of similar magnitude to the CO₂ contribution (Lacis et al., 1981; WMO, 1982; Ramanathan et al., 1985; Wigley, 1985; see also Chapter 4). It is virtually impossible to distinguish the climatic effects of CO₂ from those of other trace gases on the basis of observational data alone. As a consequence, statements about the detection of CO₂ effects should be considered as applying to the detection of the combined effects of increases in CO₂ and trace gases.

The detection of these effects on climate requires, first, some idea of what we are seeking to detect. The simplest indicator of large scale climatic change is global mean surface air temperature. Most model experiments have estimated the equilibrium temperature change corresponding to a doubling of the atmospheric CO₂ concentration (viz. T_2x). This is a good indicator of the sensitivity of the climate system to CO₂ changes, but it does not immediately tell us what changes to expect at any given time (viz. T(t)). To find T(t), we need to study the transient response of the climate system to time-dependent forcing. To date, only very simple models have been applied to this problem. Such models can, however, give considerable insight into the 'signal' that we seek to detect, and, in particular, can give us some idea of the magnitude of the important damping effect of oceanic thermal inertia.

One of the simplest models that can be used is a box-diffusion energy balance model (Siegenthaler and Oeschger, 1984; Hansen et al., 1984; Wigley and Schlesinger, 1985). More sophisticated variants employ an advective term to account for deep water production (e.g. Hoffert et al., 1980; Harvey and Schneider, 1985a,b; Watts, 1985a,b; see also the review by Hoffert and Flannery, 1985). Here, to illustrate certain key points, we will use the boxdiffusion energy balance model of Wigley and Schlesinger (1985). For this model, the main uncertainties in T(t) arise through the diffusivity term, which parameterizes mixing processes below the oceanic mixed layer, and the climate sensitivity, determined by T_2x.

In Figure 6.7, we show T(t) for one diffusivity value (2 cm² sec^-1) and for T_2x = 1.5, 3.0 and 4.5 ^°C. We have used the ice-core-based CO₂ history from 1765 to 1958 recently published by Neftel et al. (1985, their Figure 6.1), the accurately measured Mauna Loa values for 1958 to 1984, and the values employed by Wigley (1985) for the period to 2050. For the trace gases (CH₄, N₂O and the various CFCs), we have used measured values for recent decades and the future projections employed by Wigley (1985); see Figure caption for further details. Ozone changes have not been considered, although these may well be important (Ramanathan et al., 1985).

Figure 6.7 therefore, shows estimates of past and future changes in global mean surface air temperature. The biggest uncertainty arises through uncertainties in T_2x, but a considerable degree of additional uncertainty exists because of uncertainties in oceanic mixing processes and because of the simplicity of the model. Nevertheless, Figure 6.7 is sufficiently realistic to illustrate a number of important points. First, there is a change in the modelled warming rate around 1960, largely due to the influence of trace gases which began to increase around this time. Between 1960 and 1985, CO₂ and trace gases contributed almost equally to the warming. Second, the warming at any given time is substantially less than the instantaneous equilibrium warming because of the damping effect of oceanic thermal inertia (around 50-75% depending on how the ocean is modelled). Third, the warming becomes less sensitive to T_2x variations for higher T_2x (note the larger separation between the 1.5 ^°C and 3.0 °C lines than between the 3.0 °C and 4.5 °C lines). Finally, the expected warming between 1900 (the earliest date for which a reasonable estimate of global mean surface air temperature can be made; see Section 6.2.5) and 1985 is between about 0.5 and 0.9 °C (roughly 0.2 °C of this is the trace gas contribution). A wider range of possible values is obtained if one accounts for the uncertainty in oceanic mixing effects. This particular element of uncertainty cannot be reliably quantified because current ocean models are inadequate, but some idea can be gained by varying the diffusivity in the box-diffusion model used here. Such calculations suggest that the range should be increased by +0.2 °C to 0.3-1.1 ^°C. The high end of this range corresponds to high climate sensitivity (T_2x~ 4.5 ^°C) and slow ocean mixing, while the low end corresponds to low sensitivity (T_2x~l.5 ^°C) and rapid ocean mixing.

Figure 6.7 Past and future global mean surface temperature changes due to atmospheric CO₂ and trace gas concentration changes based on the transient response model of Wigley and Schlesinger (1985) for three different values of the equilibrium CO₂-doubling temperature change (T_2x = 1.5, 3.0 and 4.5 °C). CO₂ values from Neftel et al. (1985) to 1958 and Mauna Loa data to 1984. Projections to 2050 based on C = 260.1 + 9.9 exp(( Y - 1850)/63), which gives concentrations of 338 ppmv for Y(year) = 1980, 367 ppmv for Y = 2000 and 497 ppmv for Y = 2050. Trace gas concentrations are measured values to 1980 and linear projections to 2050 based on estimates given by W.M.O. (1982), Hansen et al. (1984) and Ramanathan et al. (1985) (see Wigley, 1985, for details). The two transverse dotted lines show when the modelled global mean temperature change since 1900 reaches 0.5 °C and 1.0 °C

It is useful to view the detection problem in terms of the signal-to-noise ratio concept. In broad terms, one can claim to have detected a change in climate once the signal has risen appreciably above the background noise level. The distinction between 'signal' and `noise', however, depends on the particular application. For convenience, we will consider surface air temperature as the detection variable. The signal here is the slow warming shown schematically in Figure 6.7 that is expected to result from the gradual increases in CO₂ and other trace gases. All other aspects of temperature variability can be considered as noise. Contributions to this noise occur on all time scales, and components of the noise can be divided into climatic and non-climatic effects. We consider the latter first.

Non-climatic noise in individual site records may arise from changes in instrumentation, changes in times of measurement or methods of calculating averages, or from local anthropogenic effects such as urban warming. These noise elements, which we can refer to as site inhomogeneity noise, can be minimized by careful examination and inter-comparisons of site records and the subsequent production of homogeneous time series. Non-climatic noise may also arise in area averages if these averages are based on a small, or unrepresentative, or changing number of site records. This noise component can be referred to as the spatial sampling noise. The problem of spatial representativeness has been considered in some detail by Oort (1978) using general circulation model output data and has also been discussed by Wigley et al. (1985), but no systematic attempt has yet been made to estimate the magnitude of this noise component in the various data series considered in Section 6.2.5. Studies by Jones et al. (1985) of the representativeness of their land-based Northern Hemisphere data indicate that, for these data, this particular noise component is small, at least for the period after the late 1870s.

Climatic noise arises in a number of different ways. We will discuss noise according to time scale, but note that this is a purely artificial separation since even the shortest time scale processes may have noticeable longer time scale effects. One contribution to climatic noise comes from weather noise, the uncertainty in a time-averaged quantity resulting from unpredictable day-to-day weather variability. Note that, in some applications (for example, in studies of predictability on the monthly to seasonal time scale), the terms weather noise and climate noise have been used interchangeably. Here, climate noise has a more general meaning. Most studies of weather noise have been concerned with its fractional contribution to monthly or seasonal time scale variability at individual sites (e.g. Madden, 1976; Madden and Shea, 1978). Stefanick (1981) has shown theoretically that the weather noise component of a spatially and temporally averaged climate parameter reduces as the averaging period and/or the averaging area increases. Weather noise may contribute appreciably to the inter-annual variability of the climate both directly and through the modulating effect of^-the ocean, along the lines described by Hasselmann (1976).

Many climate data time series show considerable inter-annual variability over and above the weather noise (see, for example, the weather noise publications cited above; and Weare, 1979). This derives from poorly understood short time scale climate processes, including quasi-biennial oscillations (e.g. Trenberth and Shin, 1984) and fluctuations arising from, or linked with, sea surface temperature anomalies (the Southern Oscillation/El Niño phenomenon is an important example of the latter). Present understanding of the causes of such inter-annual climatic variability is limited, and its study is an important part of the World Climate Research Programme (see, for example, J.S.C., 1984).

The most important noise component, so far as detection of CO₂ and other trace gas effects is concerned, is that which is manifest on time scales longer than the inter-annual time scale. All temperature time series show significant variability on decadal and longer time scales (see Figures 6.1-6.3), i.e. the time scales on which the main effects of increasing CO₂ and trace gases should occur. Such fluctuations can be most troublesome in attempts to detect the CO₂ influence. Unless they can be factored out in a statistically and physically convincing way, variables used in detection studies must have records spanning many decades in order to minimize the effects of decadal time scale noise.

Two statistical detection strategies may be distinguished: first, detecting a significant change in the mean value of a given climate variable or detecting a significant trend in a variable over a specified time interval; and second, the use of regression techniques to associate a statistically significant part of past variations of a variable with past variations in atmospheric CO₂ and trace gas concentrations.

The first method has been employed by Madden and Ramanathan (1980) and by Wigley and Jones (1981, 1982). From a purely statistical viewpoint, detecting a trend is closely related to detecting a change in the mean, although the appropriate t-test is slightly more efficient in the former case (in that less data are required to achieve a significant result). Since CO₂ and other trace gas increases should cause a continual increase in lower tropospheric temperature (with superimposed decadal and shorter time scale natural fluctuations), searching for a significant trend may be more appropriate than searching for a change in the mean. Work published to date, however, has used the latter approach.

It is clear from signal-to-noise ratio arguments that detection will be hastened if the noise level can be reduced. This can be achieved using some form of regression analysis with CO₂ variations and other climate forcing factors as predictors. The form of the regression equation may be guided by model results. By identifying a proportion of past climate fluctuations with factors other than increasing CO₂, and removing these effects, this method effectively reduces the noise level and increases the signal-to-noise ratio. A number of authors have used this approach in works which were not necessarily aimed at detection of CO₂ effects per se, but towards explaining recent changes in global climate (see Weller et al., 1983, for a comprehensive summary of such work). The method, however, requires an accurate record of past variations in the various possible forcing factors (which we do not have), and is fraught with statistical difficulties which will be discussed below.

All studies to date have considered only temperature data. The analyses of Madden and Ramanathan (1980) and Wigley and Jones (1981, 1982) effectively assumed that all observed variations are noise and that this noise is random with a simple first-order autoregressive structure. The observed noise level was taken to be the square root of the variance of the sampling distribution of the mean. For data with no serial correlation, this is given by S(N) =where S(N) is the standard deviation of the mean of N observations and is the standard deviation of the individual observations. However, since the temperature data are autocorrelated, due allowance must be made for this, with S(N) appropriately inflated. Madden and Ramanathan achieved this using a frequency-domain approach, while Wigley and Jones used a time-domain approach.

The assumption that the noise has a simple autoregressive structure is an oversimplification. Although there may well be an autoregressive component (associated with the weather noise discussed earlier), it is difficult to separate this stochastic element from medium to long time scale trends that are probably deterministic in nature. Nevertheless, these analyses provide extremely useful insight into the detection problem.

Madden and Ramanathan were primarily concerned with identifying a significant signal in the observational temperature record around 60° N. They concluded that the signal cannot yet be detected. Wigley and Jones were concerned with choosing_, the right combination of season and latitude band to maximize the ratio of theoretical signal to observed noise. They found the highest ratio for summer, mid-latitude temperatures; but, given the uncertainties in both signal and noise levels, all seasons and latitude bands are similar and the best detection variable is probably mean annual temperature averaged over as large an area as possible. Madden and Ramanathan also concluded that summer data were better for detection. Wigley and Jones used data from Manabe and Stouffer's (1980) GCM experiment to define the signal. In this model, the summer signal is much less than in other seasons, yet summer still yields the highest signal-to-noise ratios because observed summer noise levels are so small. Like Madden and Ramanathan, Wigley and Jones also concluded that the CO₂ signal cannot yet be detected in the available data.

Bell (1982) has considered an interesting modification of this approach. Instead of using a simple area average, he shows that the signal-to-noise ratio can be increased by appropriate weighting, a result previously demonstrated (in a different context) by Hasselmann (1979). The weighting factors, however, depend on the spatial character of both the expected signal and the observed noise; neither of which is well defined (especially the former).

The noise level can be reduced by relating part of the past variations to specific forcing factors, removing these effects, and considering only that which remains. This approach has been used indirectly by many authors in attempts to explain past temperature changes (Bryson and Dittberner, 1976; Miles and Gildersleeves, 1977, 1978; Hoyt, 1979a; Robock, 1979; Bryson and Goodman, 1980; Hansen et al., 1981; Gilliland, 1982; Mitchell, 1983; Gilliland and Schneider, 1984). These studies range from the purely empirical through to strongly model-oriented work. Most of them conclude that the past record is consistent with theoretical estimates of the magnitude of CO₂induced warming, but none provides statistically convincing and conclusive evidence that increasing CO₂ is the cause of the warming that has already occurred. Gilliland (1982) and Gilliland and Schneider (1984), in particular, stress the lack of statistical significance in their results. Vinnikov and Groisman (1981, 1982) have also used this approach. Their work is more directly concerned with detecting CO₂ effects and they state quite categorically that this exercise has been successful, not only in detecting CO₂-induced global warming, but even in identifying the spatial and seasonal patterns of warming. These conclusions are not statistically convincing.

To illustrate the statistical problems involved, let us consider a rather simple direct regression approach which encapsulates the main difficulties. Suppose surface air temperature in year i can be expressed in the form

where V, S and C are past variations in volcanic activity, solar output and atmospheric CO₂ level (none of the analyses cited above considered trace gases), A_i is an autoregressive term (i.e. A_i = f (T_i_ ₁))and e_i is a residual error term. Fitting this equation to observed temperatures produces estimates of a, b and c with associated confidence limits. If the limits on the CO₂ coefficient, c, exclude zero, this might be taken as proof of a significant CO₂ effect. It is, however, an extremely difficult task to assign confidence limits to these regression coefficients, and standard formulae are inappropriate for two reasons.

First, the values of the forcing functions V, S and C are uncertain. (So too is the response variable, T; it must involve some spatial sampling noise.) Forcing function uncertainties are manifest in the different functions used by different authors. For example, the volcanic forcing functions used by Vinnikov and Groisman (1981, 1982), Hansen et al. (1981) and Gilliland (1982) are quite different. Vinnikov and Groisman used Pivovarova's (1977) actinometric data on atmospheric transmissivity (P), Hansen et al. used Lamb's (1970) Dust Veil Index (L), and Gilliland used the ice core acidity record of Hammer et al. (1980) (H). The correlations between these three volcanic forcing functions are small: r_PL =

0.46, r_PH =

0.32 and r_LH = 0.41. While these correlations are all statistically significant at the 5% level, the time series clearly have only a small amount of variance in common (<22%).

The solar and CO₂ forcing functions used by these authors are also quite different. Gilliland used the 80-year cycle in solar diameter (see Parkinson et al., 1980; Gilliland, 1980, 1981; and Section 5.2 in Dickinson), tuning the phase and amplitude of the implied solar irradiance cycle to best fit the observed temperature data. Hansen et al. (1981) used the umbral

penumbral ratio of Hoyt (1979a,b) as an indicator of solar irradiance changes (an indicator that has since been largely discredited by satellite data; see Eddy et al., 1982). Vinnikov and Groisman have no solar term, although a solar effect may be implicit in their transmissivity data. Finally, the assumed CO₂ forcing functions differ noticeably from author to author and trace gases have not been considered.

It is obviously impossible for a single record of past climate to be correctly explained by three different sets of forcing function records; yet all studies claim to explain a large fraction of the temperature variance. The only way this can occur is for all results to be subject to considerable statistical uncertainty; i.e. to have wide confidence bands for the regression coefficients and for the total explained variance.

The second reason why confidence levels are difficult to assign is purely statistical. Because the response and predictor variables show medium to long (> 10 yr) time scale trends, a good fit could be obtained if some of these trends happened to match by chance. For example, the umbral

penumbral ratio solar index used by Hansen et al. has rising and falling trends either side of (approximately) 1940 which are similar to those in the temperature record. This is sufficient to ensure a good correlation between T_i and S_i. However, since the strength of this correlation depends mainly on one factor, the similar turning point around 1940, the link is almost certainly not statistically significant. Building an arbitrary and variable time lag into any relationship, as in Gilliland's analysis, introduces the statistical problem of multiplicity (namely, that if enough experiments are performed, different phases and amplitudes in this case, a statistically significant result will eventually arise by chance). This further reduces the statistical significance of any results.

The central statistical problem in these analyses is that the number of degrees of freedom that should be used in testing significance cannot easily be determined. In addition to the reasons for this already alluded to, further reductions in the number of degrees of freedom may arise through autocorrelation in the data and from multicollinearity. The effect of autocorrelation in both the response and predictor data is difficult to account for, and Bartlett's method, for example, (see Quenouille, 1952) cannot be applied because the autocorrelation does not arise solely through an inherent (and stable) autoregressive process. Multicollinearity (i.e. inter-correlated predictor variables) is an issue because, on decadal and longer time scales, all three forcing variables may show similar low frequency fluctuations. Apart from making it impossible to assign a causal role to any particular variable in such circumstances, correlations between the predictors can produce unstable regression results, which can further undermine estimates of statistical significance.

In addition to their failure to properly address these problems, in none of the studies cited above is the transient response of the system (i.e. the effects of oceanic thermal inertia) adequately modelled. Thus, most of the noise reduction studies cited above are deficient because of their neglect of one or more of the factors mentioned here. Other less detailed evaluations and reviews of the detection problem (e.g. Kellogg and Bojkov, 1982) have been similarly deficient.

Given these uncertainties, the noise reduction method at present can only give supporting evidence in the detection problem. The possibility of removing part of the low-frequency variance by ascribing it confidently (and with known confidence) to other causes is an important consideration. However, because our understanding of other climate forcing mechanisms is probably at an equal or lower level than our understanding of the effects of CO₂ and other trace gases, a statistically rigorous application of this method is extremely difficult.

At a U.S. Department of Energy workshop on the first detection of CO₂ effects (Moses and MacCracken, 1982; MacCracken and Moses, 1982), it was suggested that detection would be facilitated by developing 'a unique CO₂-specific "fingerprint" for the CO₂ response involving a set of 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, 1982, p. 1172). The main purpose of the fingerprint method is to aid in the attribution aspect of detection; the association of a statistically identified change or set of changes specifically with CO₂ (or greenhouse gas) forcing. As an illustration, we might consider both tropospheric and stratospheric temperatures as a simple two-element fingerprint. Since a solar irradiance increase should warm both the troposphere and stratosphere, while a greenhouse gas increase should cause stratospheric cooling, this would be sufficient to distinguish between solar and greenhouse gas forcing, although it would not allow greenhouse gas effects to be readily isolated from, for example, the effects of changing stratospheric aerosol concentration. The statistical aspects of this simple type of fingerprint have been explored fully by Epstein (1982).

Such 'limited discrimination' markers can be expected to be useful before complete fingerprints can serve us directly. However, even limited discrimination fingerprints are beset by practical problems. Difficulties in detecting the effects of changing CO₂ and other trace gas concentrations using a single variable must also apply to both limited discrimination and complete fingerprint methods. The uncertainty inherent in both may be as great as in their most ill-defined part.

Numerous variables have been suggested as detection parameter candidates and, therefore, as candidates for part of either a discriminator or a fingerprint. Some of these are listed in Table 6.1 (see Section 6.5). Others that have been proposed include mean values of daily temperature range, temperatures near the seasonal snow-ice boundary, regional precipitation, and so on. Problems with using these and other suggested discriminator or fingerprint candidates arise because, in most cases, the expected signal is not well-defined and/or our knowledge of the natural variability is meagre. Very few variables are currently suitable as fingerprint components.

In order to be able to distinguish greenhouse gas effects from other possible causes of climatic change, we require a much better understanding of the natural variability of the whole climate system and its responses to forcing. In particular, since a useful discriminator or fingerprint probably requires a knowledge of the greenhouse gas signal at the regional spatial scale (because of the paucity of global-scale data), we need to know more about the spatial details of climatic change. Both discriminator and complex fingerprint methods, therefore, can be viewed at present only as goals that modellers and data analysts should strive towards. It is likely that relatively sophisticated multivariate statistical tests will be required to implement fingerprint detection strategies and, even though no useful results have yet been obtained, such techniques are currently being developed (e.g. Hasselmann, 1979; Bell, 1982; Epstein, 1982). A continuation of this effort is essential.

If model results are correct, we should already have experienced a substantial global mean surface warming due to the increases in the concentrations of CO₂ and other trace gases, although this warming may well be partly or wholly obscured by natural climate fluctuations. In Section 6.2.5, we showed that the globe has warmed noticeably since the late nineteenth century, by about 0.5 °C, with quite large amplitude decadal and shorter time scale fluctuations superimposed on this overall warming trend. In Section 6.4.1, we used a simple model to show that, between 1900 and 1985, CO₂ and trace gas changes should have warmed the globe by 0.3

1.1 °C. Can we, therefore, claim to have detected the anthropogenic greenhouse effect in a statistically rigorous way? The short answer to this question is `no'; but, before explaining why, let us first look more closely at the uncertainties in model predictions.

The predicted warming due to greenhouse gas increases since pre

industrial times depends on their initial concentrations, the size of the signal for a CO₂ doubling, and the damping effect of oceanic thermal inertia. The pre-industrial CO₂ concentration is not known with any certainty and the recently accepted range of 260

280p pmv (W.M.O., 1983; Wigley, 1983; Siegenthaler, 1984) has, yet more recently, been revised upward by some of the newest ice core data (Neftel et al., 1985, see also Chapter 3). There is also considerable uncertainty in the equilibrium temperature changedue to a CO₂ doubling. We consider 1.5 to 5.5 °C to be a more realistic estimate of the range of uncertainty than the Carbon Dioxide Assessment Committee's 1.5-4.5 °C range (N.R.C., 1983); see Chapter 5. Additional uncertainties accrue from uncertainties in the oceanic thermal inertia effect. It is the sum of all these uncertainties that leads to the range of T(1900

1985) values, 0.3

1.1 ^°C, given in Section 6.4.1.

Two further comments are pertinent. This range was based on T_2x values of 1.5

4.5 °C, so a slight upward revision of the upper limit would be required to cover the 5.5 °C upper limit for T_2x recommended in Chapter 5. An upward revision of the whole range would be required if the Neftel et al. (1985) CO₂ values are too high. An overestimate by 15 ppmv would raise the T (1900

1985) range by 0.1

0.2 ^°C. In this regard, it is worth recalling some details of Neftel et al.'s data. These show a smooth upward trend in CO₂ concentration from around 280 ppmv in the mid-eighteenth century, through 285 ppmv in 1850 and 297 ppmv in 1900. The 1850 value is 15 ppmv higher than the mid-point of the previously accepted 260-280 ppmv range.

If model results suggest a global warming of 0.3

1.1 °C since 1900, and if the observational data show a warming of 0.5 °C, are these results compatible? We note first that the 0.5 ^°C observed warming is also subject to some uncertainty due to the various problems discussed in Section 6.2.5. The post1900 warming could be anywhere in the range 0.3

0.7 °C due to statistical uncertainties in estimating the trend, uncertainties in data quality, and gaps in coverage. It is clear that the observational and model-predicted ranges overlap so they must be judged as compatible (or, at least, not inconsistent). Since observations are in the low end of the model range, this implies, either that the climate sensitivity (as quantified by T_2x) is in the low end of the range 1.5

5.5 °C, or that ocean mixing processes are relatively rapid, or both. If the Neftel et al. CO₂ values are too high, then (as noted by Wigley and Schlesinger., 1985) this would reduce the implied climate sensitivity.

On the other hand, the observed warming since 1900 may well include substantial natural variations, and it may be entirely wrong to assume, as has been done above, that the 0.3

0.7 °C observed warming is largely a greenhouse gas effect. Since the mid-nineteenth century there have been marked decadal and longer time scale temperature fluctuations in both directions, warming and cooling. For example, between 1940 and 1965 the

Northern Hemisphere cooled by approximately 0.3 °C, while the Southern Hemisphere showed little overall change (see Figures 6.1 and 6.2). These changes are contrary to the predicted warming effect.

Explaining the 1940-1965 cooling is clearly important, but it is unlikely that we will ever be able to do so convincingly, simply because the appropriate data are not available. If model estimates of T_2x are correct, then this cooling must be due either to some external factor other than the greenhouse gases, or to a major internal climatic oscillation (possibly associated with a change in ocean circulation). Changes in stratospheric volcanic aerosol loading and/or changes in solar irradiance have been suggested as causal factors by the noise reduction studies described in Section 6.4.3. However, proof of these contentions is virtually impossible because the aerosol and solar forcings are so poorly known.

The contrast between the temperature trends in the Southern and Northern Hemispheres (see Figure 6.2) is clearly an important factor that must be considered. While this does not necessarily eliminate the aerosol and solar forcing hypotheses, it does suggest that changes in some hemispherically specific factor (such as the rate of formation of North Atlantic Deep Water) could be involved. Evidence that NADW changes might have affected global climate in the past has been presented in Section 6.2.3. We also know that important changes have occurred in North Atlantic water masses in recent years (Brewer et al., 1983; Swift, 1984; Roemmich and Wunsch, 1984; Bennett et al., 1985), and that these changes are compatible with significant changes in NADW formation rate. Furthermore, simple modelling studies (Hoffert et al., 1985; Watts, 1985a,b) and heat budget calculations (Broecker et al., 1985) indicate that NADW formation rate changes are a physically realistic forcing mechanism. Further analyses are required to test this hypothesis.

To summarize, three points should be noted. First, if the post-1900 global warming is assumed to be due to changes in atmospheric CO₂ and other trace gas concentrations, then the observed data are compatible with model predictions, possibly favouring the lower half of the T_2x range of 1.55

5 ^°C. However, uncertainties in modelling the transient response of the climate system are such that higher T₂_x values cannot be ruled out. Second, because the observational record shows large variations in global mean temperature that we cannot explain, we cannot yet claim to have unequivocally detected the signal due to increases of CO₂ and other trace gases.

A third and final point should be stressed. Because of oceanic thermal inertia effects, the present global mean temperature may well be quite far removed from equilibrium (Hansen et al., 1984; Wigley and Schlesinger, 1985). Thus, even if greenhouse gas increases could be halted today, it is possible that the globe would warm substantially over coming decades as the system tended towards a new equilibrium with the prevailing greenhouse gas level. If T_2x were in the upper half of the range 1.5

5.5 °C, this residual warming could currently exceed 0.5 °C. Its value is highly sensitive to uncertainties in T_2x (Wigley, 1985; Wigley and Schlesinger, 1985) and to the way the oceans are modelled.

6.5 MONITORING REQUIREMENTS

For monitoring in general, there is a basic need for spatial averaging in defining variables suitable for detecting CO₂ and other trace gas effects. There are two reasons for this. First, much of the variability at a point is specific to that location and is not coherent with variations on larger spatial scales (thus, spatial averaging reduces the noise level). Second, the reliability of available model projections is generally less for smaller spatial scales (i.e. the signal is best defined for the largest spatial scales).

For a particular climate variable to be useful for detection, we require the available data to be accurate (i.e. minimum site inhomogeneity noise), representative (i.e. minimum spatial sampling noise) and of sufficient length to allow for appropriate statistical tests to be applied and to ensure that climatic noise levels are well-defined on time scales relevant to possible greenhouse gas effects. We also require the expected signal to be at least qualitatively well-defined. Potential detection variables can be selected on the basis of, and rated in terms of these criteria (see Table 6.1). For a more extensive discussion of possible detection variables see Weller et al. (1983).

In addition to monitoring climate variables which respond to changes of CO₂ and other trace gases, it is important to monitor the greenhouse gas changes themselves, and, where possible, other climate forcing factors. A knowledge of these is essential for future noise reduction studies. Table 6.1, therefore, also includes important climate forcing factors.

Definition of the noise level is primarily dependent on the amount of data available; i.e. on the length of the available observational record. At present, only surface temperature, precipitation and pressure have sufficiently long records to be able to establish their noise levels reliably, but continued monitoring of other variables will rapidly enhance their value in detection studies.

For confidence in the expected signal, variables such as global mean annual surface air temperature and annual mean lower stratosphere temperature have reasonably well-defined signals, but variables that depend on the regional or seasonal details of climate or on poorly understood physical links within the climate system have, at present, poorly defined signals. Further developments in climate models should improve this situation.

Table 6.1 also gives an estimate of the accuracy required for particular variables to be useful in detection studies. In some cases, this is within present capabilities, in others it is far beyond present capabilities. Accuracy levels have been assigned on the basis of the expected magnitude of the signal between the years 1980 and 2000, on the assumption that the year 2000 is a detection threshold. In order to account for transient response effects we have based values for all climate variables on the global mean temperature change calculations illustrated in Figure 6.7. These show that T (1980

2000) is between one fifth and one seventh of the equilibrium CO₂-doubling temperature change. We have used a factor of one-sixth for all variables. We have further assumed that the accuracy required to detect such a change is ±20% of this figure.

Table 6.1 Variables which could or shouldbe monitored to detect the effects of atmospheric greenhouse gas concentration increases on climate

Reliable

Confidence

Required

Variable

record

in signal

monitoring

Priority

length

accuracy

Climate variables

Surface air

~100 yr

high²

±0.1 °C

high

temperature¹ (land)

Surface air

~100 yr

high²

±0.1 °C

high

temperature¹

(marine)

Sea surface

~ 100 yr

high²

±0.1 °C

high

temperature¹

Tropospheric

~30 yr

high²

±0.1 °C

high

temperature¹

Lower Stratospheric¹

~30 yr

high²

±0.1 ^°C

high³

temperature

(to ~50 mb)

Mid/upper

⁴

high²

±0.3 °C

high³

stratospheric¹

temperature

Precipitation¹

⁵

moderate²

±0.2%

low

Sea-ice extent

~^.20

80 yr(N.H.)

low

±0.2^°6

moderate⁷

10-20 yr(S.H.)

Global mean sea level

100 yr⁸

moderate

± l cm

moderate

Ocean circulation^9.10

¹¹

low

¹²

high¹³

Global cloud cover

low

¹⁴

moderate

Forcing factors

Atmospheric CO₂

~100yr¹⁵

¹⁶

high

and trace gas

concentrations

Solar irradiance

~10 yr¹⁷

±0.1%

high

Stratospheric aerosol

¹⁸

¹⁹

moderate²⁰

concentration

Tropospheric aerosol

²¹

¹⁹

low

concentration

	Footnotes for Table 6.1
¹	Large-scale area averages of seasonal and/or annual means.
²	Less for smaller spatial scales and for seasonal compared with annual means.
³	Particularly important in fingerprint studies in distinguishing greenhouse gas effects from those of other forcing
	factors.
⁴	Present spatial coverage limited, and data have possible instrumental errors.
⁵	Reliable hemispheric-scale averages not yet available, but suitable data exist.
⁶	Shift in longitudinally-averaged latitude of ice boundary.
⁷	Lower priority as a detection variable, but an important boundary condition for models (with snow cover
	and surface albedo).
⁸	Subject to some uncertainty (see Barnett, 1983a.b).
⁹	Both vertical and horizontal, and including circulation-change indicators such as temperature and salinity.
¹⁰	Can be considered as a forcing factor in some instances.
¹¹	No spatially comprehensive data set yet available, but numerous short time series and regional studies exist.
¹²	Strongly dependent on the particular ocean variable.
¹³	Low priority as a detection variable, but high priority as a modulator of past and future climatic change.
¹⁴	Difficult to estimate due to the complex ways that clouds affect climate, but, based solely on albedo-change,
	accuracy required is < ± 1%.
¹⁵	Data prior to 1958 less reliable.
¹⁶	Present accuracy adequate.
¹⁷	Homogeneous records available for only part of this period.
¹⁸	Considerable differences exist between various data sets.
¹⁹	Not estimated.
²⁰	Reduced priority due to difficulties in obtaining globally representative values.
²¹	Only available for a few locations whose representativeness is uncertain.

For the forcing factors listed in Table 6.1, accuracy levels were estimated on the basis of the change required to produce a global mean temperature change of ±0.1 ^°C. Unfortunately, this is only possible for solar and greenhouse gas forcing because our understanding of the decadal and longer time scale climatic effects of volcanic or tropospheric aerosols is inadequate to make reasonable estimates. In both cases, however, monitoring requirements are almost certainly beyond our current measuring capabilities.

It should be noted that the figures given in Table 6.1 are only a guide, based on rather simple arguments. However, given the uncertainty in estimates of the various aspects of the signal, a more sophisticated method for defining accuracy figures (e.g. incorporating information about noise levels and statistical significance) does not seem warranted at this stage.

Table 6.1 also lists subjective evaluations of priorities for each detection variable. These are based on data length and signal confidence, and on required monitoring accuracy relative to what is currently possible. Priorities for the forcing parameters are based on their relative importance as forcing factors and on the current feasibility of adequate monitoring.

6.6 CONCLUSIONS

ACKNOWLEDGEMENTS

Jim Hansen (Goddard Institute for Space Studies) and David Parker and Chris Folland (United Kingdom Meteorological Office) kindly provided unpublished data. A large number of persons provided important feedback on earlier versions of the manuscript. Comments from Eero Holopainen (Department of Meteorology, University of Helsinki), Syukuru Manabe (Geophysical Fluid Dynamics Laboratory), Roland Madden (National Center for Atmospheric Research) and David Parker were particularly valuable. Other useful comments were received from T. Asai, A. L. Berger, H. W. Ellsaesser, H. L. Ferguson, R. M. Gifford, G. S. Golitsyn, J. Goudriaan, J. A. Laurmann, A. S. Monin, O. Preining, G. de Q. Robin, J. W. Tukey, J. D. Woods and Du-zheng Ye. Much of the authors' work reported here was funded by the U.S. Department of Energy, Carbon Dioxide Research Division, under contract numbers DE-AC02-79EV 10098 and DE-AC02-81 EV 10738.

6.7 REFERENCES

Andrews, J. T., and Barry, R. G. (1978). Glacial inception and disintegration during the last glaciation. Annual Review of Earth and Planetary Sciences, 6, 205

228.

Angell, J. K., and Gruza, G. V. (1984). Climate variability as estimated from atmospheric observations. In Houghton, J. T. (Ed.) The Global Climate: 25

36. Cambridge University Press.

Angell, J. K., and Korshover, J. (1977). Estimate of the global change in temperature, surface to 100 mb, between 1958 and 1975. Monthly Weather Review, 105, 375

385.

Angell, J. K., and Korshover, J. (1978a). Global temperature variation, surface

100mb: an update into 1977. Monthly Weather Review, 106, 755

770.

Angell, J. K., and Korshover, J. (1978b). Estimate of global temperature variations in the 100

30mb layer between 1958 and 1977. Monthly Weather Review, 106, 1422

1432.

Angell, J. K., and Korshover, J. (1983a). Global temperature variations in the troposphere and stratosphere, 1958

1982. Monthly Weather Review, 111, 901

921.

Angell, J. K., and Korshover, J. (1983b). Comparison of stratospheric warmings following Agung and Chichón. Monthly Weather Review, 111. 2129

2135.

Atkinson, T. C., Briffa, K. R., Coope, G. R., Joachim, M. J., and Perry, D. W. (1985). Climatic calibration of coleopteran data. In Berglund, B. D. (Ed.) Handbook of Holocene Palaeoecology and Palaeohydrology, 851

858. John Wiley and Sons Ltd.

Barnett, T. P. (1978). Estimating variability of surface air temperature in the Northern Hemisphere. Monthly Weather Review, 106, 1353

1367.

Barnett, T. P. (1983a). Recent changes in sea level and their possible causes. Climatic Change, 5, 15

38.

Barnett, T. P. (1983b). Long-term changes in dynamic height. Journal of Geophysical Research, 88, 9547

9552.

Barnett, T. P. (1984). Long-term trends in surface temperature over the oceans. Monthly Weather Review, 112, 303

312.

Barnett, T. P. (1985a). On long-term climate changes in observed physical properties of the oceans. In MacCracken, M. C., and Luther, F. M. (Eds). Detecting the Climatic Effects of Increasing Carbon Dioxide, U.S. Dept. of Energy, Carbon Dioxide Research Division, Washington D.C.

Barnett, T. P. (1985b). Long-term changes in precipitation patterns. In MacCracken, M. C., and Luther, F. M. (Eds). Detecting the Climatic Effects of Increasing Carbon Dioxide, 149

162. U.S. Dept. of Energy, Carbon Dioxide Research Division, Washington D.C.

Barron, E. J., and Washington, W. M. (1984). The role of geographic variables in explaining paleoclimates: Results from Cretaceous climate model sensitivity studies. Journal of Geophysical Research, 89, 1267

1279.

Bell, T. L. (1982). Optimal weighting of data to detect climatic change: application to the carbon dioxide problem. Journal of Geophysical Research, 87,11161

11170.

Bennett, T., Broecker, W., and Hansen, J., Eds. (1985). North Atlantic Deep Water Formation, NASA Conference Publication 2367. NASA Scientific and Technical Information Branch: 66 pages.

Berger, A. (1977). Support for the astronomical theory of climatic change. Nature, 268, 44

45.

Berner, R. A., Lasaga, A. C., and Garrels, R. M. (1983). The carbonate-silicate geochemical cycle and its effect on atmospheric carbon dioxide over the last 100 million years. American Journal of Science, 283, 641

683.

Borzenkova, I. I., Vinnikov, K. Ya., Spirina, L. P., and Stekhnovskii, D. I. (1976). Variation of Northern Hemisphere air temperature from 1881 to 1975. Meteorologiya i Gidrologiya 1976, no. 7, 27

35 (in Russian).

Bradley, R. S., and Jones, P. D. (1985). Data bases for isolating the effects of the increasing carbon dioxide concentration. In MacCracken, M. C., and Luther, F. M. (Eds). Detecting the Climatic Effects of Increasing Carbon Dioxide, 29

53. U.S. Dept. of Energy, Carbon Dioxide Research Division, Washington D.C.

Brewer, P. G., Broecker, W. S.. Jenkins, W. J., Rhines, P. B., Rooth, C. G., Swift. J. H., Takahashi, T., and Williams. R. T. (1983). A climatic freshening of the deep North Atlantic north of 50' N over the past 20 years. Science, 222, 1237

1239.

Brinkmann, W. A. R. (1979). Associations between temperature trends. Annals of the Association of American Geographers, 69, 250

261.

Broecker, W. S., and Takahashi, T. (1984). Is there a tie between atmospheric CO₂ content and ocean circulation? In Hansen, J. E., and Takahashi, T. (Eds.) Climate Processes and Climate Sensitivity (Maurice Ewing Series, No. 5): 314

326. American Geophysical Union, Washington, D.C.

Broecker, W. S., Peteet, D. M., and Rind, D. (1985). Does the ocean-atmosphere system have more than one stable mode of operation? Nature, 315, 21

26.

Bryan, K., Komro, F. G., Manabe, S., and Spelman, M. J. (1982). Transient climate response to increasing atmospheric carbon dioxide. Science, 215, 56

58.

Bryan, K., Komro, F. G., and Rooth, C. (1984). The ocean's transient response to global surface temperature anomalies. In Hansen, J. E., and Takahashi, T. (Eds.) Climate Processes and Climate Sensitivity (Maurice Ewing Series, No. 5): 29

38. American Geophysical Union, Washington, D.C.

Bryson, R. A. (1974). A perspective on climatic change. Science, 184, 753

760.

Bryson, R. A., and Dittberner, G. J. (1976). A non-equilibrium model of hemispheric mean temperature. Journal of the Atmospheric Sciences, 33, 2094

2106.

Bryson, R. A., and Goodman, B. M. (1980). Volcanic activity and climatic change. Science, 207, 1041

1044.

Budyko, M. I. (1962). Polar ice and climate. Izvestia Ak. Nauk. SSR, Ser. Geog. 1962, no. 6, 3

10 (in Russian).

Budyko, M. I. (1969). The effect of solar radiation variations on the climate of the Earth. Tellus, 21, 611

619.

Budyko, M. I. (1982). The Earth's Climate: Past and Future. Academic Press, New York: 307 pages.

Budyko, M. I., Vinnikov, K. Ya., Drozdov, O. A., and Efimova, N. A. (1978). The forthcoming climatic change. Izvestia Ak. Nauk. SSR, Ser. Geog. 1978, no. 6, 5

20 (in Russian).

Butzer, K. W. (1980). Adaptation to global environmental change. Professional Geographer, 32, 269

278.

Callendar, G. S. (1961). Temperature fluctuations and trends over the Earth. Quarterly Journal of the Royal Meteorological Society, 87, 1

12.

Chen, R. S. (1982). Combined Land/sea Surface Air Temperature Trends, 19491972. MSc dissertation, Massachusetts Institute of Technology.

CLIMAP Project Members, (1976). The surface of the ice-age Earth. Science, 191, 1131

1144.

CLIMAP Project Members, (1984). The last interglacial ocean. Quaternary Research, 21,123

224.

Corona, T. J. (1978). The Interannual Variability of Northern Hemisphere Precipitation. Environmental Research Paper No. 16, Colorado State University, Fort Collins: 27 pages.

Corona, T. J. (1979). Further Investigations of the Interannual Variability of Northern Hemisphere Continental Precipitation. Environmental Research Paper No. 20, Colorado State University, Fort Collins: 20 pages.

Craig, H., and Chou, C. C. (1982). Methane: The record in polar ice cores. Geophysical Research Letters, 9, 1221

1224.

Delmas, R., Ascencio, J.-M., and Legrand, M. (1980). Polar ice evidence that atmospheric CO₂ 20,000 yr BP was 50% of present. Nature, 284, 155

157.

Denton, G. H., and Hughes, T. Eds. (1981). The Last Great Ice Sheets. Wiley-Interscience. New York.

Diaz, H. F. (1981). Eigenvector analysis of seasonal temperature, precipitation and synoptic scales system frequency over the contiguous United States, Part II: spring, summer, fall and annual. Monthly Weather Review, 109, 1285

1304.

Diaz, H. F., and Fulbright, D. C. (1981). Eigenvector analysis of seasonal temperature, precipitation and synoptic scales system frequency over the contiguous United States, Part 1: winter. Monthly Weather Review, 109, 1267

1284.

Dorman, C. E., and Bourke, R. H. (1979). Precipitation over the Pacific Ocean, 30°S to 60°N. Monthly Weather Review, 107, 896

910.

Dorman, C. E., and Bourke, R. H. (1981). Precipitation over the Atlantic Ocean, 30°S to 70°N. Monthly Weather Review, 109, 554

563.

Drozdov, O. A. (1966). On the precipitation changes in the Northern Hemisphere when the temperature of the polar basin changes. Trudy GGO 1966, no. 198, 3

16 (in Russian).

Drozdov, O. A. (1974). Anthropogenic impact on the hydrological cycle. Trudy GGO 1974, no. 316, 83

103 (in Russian).

Eddy, J. A. (1976b). Climate and the changing Sun. Climatic Change, 1, 173

190.

Eddy, J. A., Gilliland, R. L., and Hoyt, D. V. (1982). Changes in the solar constant and climatic effects. Nature, 300, 689

693.

Elliott, W. P., and Reed, R. K. (1984). A climatological estimate of precipitation for the world ocean. Journal of Climate and Applied Meteorology, 23, 434

439.

Ellsaesser, H. W., MacCracken, M. C., and Walton, J. J. (1985). Global climatic trends as revealed by the recorded data. Reviews of Geophysics and Space Physics (in press).

Epstein, E. S. (1982). Detecting climate change. Journal of Applied Meteorology, 21, 1172

1182.

Fleer, H. (1981). Large-scale tropical rainfall anomalies. Bonner Meteorologische Abhandlungen No. 26: 114 pages.

Fletcher, J. O. (1984). Clues from sea-surface records. Nature, 310, p. 630.

Flohn, H. (1977). Climate and energy: a scenario to a 21st century problem. Climatic Change. 1, 5

20.

Flohn, H. (1979). A scenario of possible future climates

natural and man-made. Proceedings of the World Climate Conference Geneva, WMO-No. 537, 243

266.

Flohn, H. (1980). Possible Climatic Consequences of a Man-Made Global Warming, HASA publication RR-80-30. International Institute for Applied Systems Analysis, Laxenburg, Austria.

Flohn, H. (1981). Major Climatic Events Associated with a Prolonged CO₂-Induced Warming, ORAU/IEA-81-8(M). Institute for Energy Analysis, Oak Ridge Associated Universities, Oak Ridge, 80 pages.

Folland, C. K., and Kates, F. E. (1984). Changes in decadally averaged sea surface temperature over the world, 1861

1980. In Berger, A., Imbrie, J., Hayes, J., Kukla, G., and Saltzman, B. (Eds.) Milankovitch and Climate, Part 2, 721

727. D. Reidel, Dordrecht.

Folland, C. K., Parker, D. E., and Kates, F. E. (1984a). Worldwide marine temperature fluctuations, 1856

1981. Nature, 310, 670

673.

Folland, C. K., Parker, D. E., and Newman, M. (1984b). Worldwide marine temperature variations on the season-to-century time scale. Proceedings of the Ninth Annual Climate Diagnostics Workshop, U.S. Dept. of Commerce, NOAA, 70

85.

Gates, W. L. (1976a). Modeling the ice-age climate. Science, 191, 1131

1144.

Gates, W. L. (1976b). The numerical simulation of ice-age climate with a global general circulation model. Journal of the Atmospheric Sciences, 33, 1844

1873.

Gilliland, R. L. (1981). Solar radius variations over the past 265 years. Astrophysical Journal, 248, 1144

1155.

Gilliland, R. L. (1982). Solar, volcanic and CO₂ forcing of recent climatic change. Climatic Change, 4, 111

131.

Gilliland, R. L., and Schneider, S. H. (1984). Volcanic, CO₂ and solar forcing of Northern and Southern Hemisphere surface air temperatures. Nature, 310, 38

41.

Groisman, P. Ya. (1979). Algorithm of estimation of linear structural relation between macroclimatic parameters. Trudy GGI 1979, no. 257, 76

80 (in Russian).

Groisman, P. Ya. (1981). The empirical estimates of the relationship between the processes of global warming and cooling and the moisture regime over the territory of the USSR. flzvestia Ak. Nauk. SSR, Ser. Geog. 1981, no. 5, 86

95 (in Russian).

Gruza, G. V., and Apasova, Ye. G. (1981). Climatic variability of the Northern Hemisphere precipitation amounts. Meteorologiya i Gidrologiya, 1981, No. 5, 516 (in Russian).

Hammer, C. U., Clausen, H. B., and Dansgaard, W. (1980). Greenland icesheet evidence of post-glacial volcanism and its climatic impact. Nature, 288, 230

235.

Hansen, J. E., Johnson, D., Lacis, A., Lebedeff, S., Lee, P., Rind, D., and Russell, G. (1981). Climatic impact of increasing atmospheric carbon dioxide. Science, 213, 957

966.

Hansen, J. E., Lacis, A., Rind, D., Russell, G., Stone, P., Fung, I., Ruedy, R., and Lerner, J. (1984). Climate sensitivity: analysis of feedback mechanisms. In Hansen, J. E., and Takahashi, T. (Eds.) Climate Processes and Climate Sensitivity (Maurice Ewing Series, No. 5): 130

163. American Geophysical Union, Washington, D.C.

Harvey, L. D. D., and Schneider, S. H. (1985a). Transient climate response to external forcing on 10⁰

10⁴ year time scales. 1: Experiments with globally averaged, coupled atmosphere and ocean models. Journal of Geophysical Research, 90, 2191

2206.

Harvey, L. D. D., and Schneider, S. H. (1985b). Transient climate response to external forcing on 10⁰-10⁴ year time scales. 2: Sensitivity experiments with a seasonal, hemispherically averaged, coupled atmosphere, land and ocean energy balance model. Journal of Geophysical Research, 90, 2207

2222.

Hasselmann, K. (1976). Stochastic climate models, part 1, theory. Tellus, 28, 473485.

Hasselmann, K. (1979). On the signal-to-noise problem in atmospheric response studies. In Shaw, D. B. (Ed.) Meteorology over the Tropical Oceans 251

259. Royal Meteorological Society.

Hays, J., Imbrie, J., and Shackleton, N. (1976). Variations in the Earth's orbit: pacemaker of the ice ages. Science, 194, 1121

1132.

Hoffert, M. I., Callegari, A. J., and Hsieh, C. T. (1980). The role of deep sea heat storage in the secular response to climatic forcing. Journal of Geophysical Research, 85, 6667

6679.

Hoffert, M. I., Gaffin, S., Wang. Z. Y., Hsieh, C. T., and Volk, T. (1985). Interannual climate oscillations in past and future temperature records. Extended Summaries, Third Conference on Climate Variations and Symposium on Contemporary Climate 1850

2100, American Meteorological Society, 118

119.

Hoffert, M. I., and Flannery, B. P. (1985). Model projections of time-dependent response to increasing carbon dioxide. In U.S. Dept. of Energy State of the Art Report on the Climatic Effects of Increasing CO₂ (in press).

Hollin, J. T. (1980). Climate and sea level in isotope stage 5: an East Antarctic ice surge at ~95,000 BP? Nature, 283, 629

633.

Hoyt, D. V. (1979a). An empirical determination of the heating of the Earth by the carbon dioxide greenhouse effect. Nature, 282, 388

390.

Hoyt, D. V. (1979b). Variations in sunspot structure and climate. Climatic Change, 2, 79

92.

Imbrie, J., and Imbrie, J. Z. (1980). Modeling the climatic response to orbital variations. Science, 207, 943

953.

J. S. C. (1984). Scientific Plan for the World Climate Research Programme. WCRP Publ. Series No. 2, WMO/TD-No. 6.

Jäger, J., and Kellogg, W. W. (1983). Anomalies in temperature and rainfall during warm Arctic seasons. Climatic Change, 5, 39

60.

Jones, P. D., and Kelly, P. M. (1983). The spatial and temporal characteristics of Northern Hemisphere surface air temperature variations. Journal of Climatology, 3, 243

252.

Jones, P. D., Raper, S. C. B., Bradley, R. S., Diaz, H. F., Kelly, P. M., and Wigley, T. M. L. (1986a). Northern Hemisphere surface air temperature variations 1851

1984. Journal of Climate and Applied Meteorology, 25, 161

179.

Jones, P. D., Raper, S. C. B., and Wigley, T. M. L. (1986b). Southern Hemisphere surface air temperature variations 1851

1984. Journal of Climate and Applied Meteorology (in press).

Jones, P. D., Wigley, T. M. L., and Kelly, P. M. (1982). Variations in surface air temperature: Part 1, Northern Hemisphere 1881

1980. Monthly Weather Review, 110, 59

70.

Jones, P. D., Wigley, T. M. L., and Wright, P. B. (1986c). Global temperature variations. Nature (in press).

Kellogg, W. W. (1977). Effects of human activities on climate. WMO Technical Note 156, WMO No. 486, World Meteorological Organization, Geneva.

Kellogg, W. W. (1978). Global influences of mankind on the climate. In Gribbin, J. (Ed.) Climatic Change: 205

227. Cambridge University Press.

Kellogg, W. W., and Bojkov, R. D. Eds. (1982). Report of JSC/CAS Meeting of Experts on Detection of Possible Climate Change. WCP-29. World Meteorological Organization, Geneva.

Kellogg, W. W., and Schware, R. (1981). Climate Change and Society. Consequences of Increasing Atmospheric Carbon Dioxide. Westview Press, Boulder, Colorado: 178 pages.

Kelly, P. M., Jones, P. D., Sear, C. B., Cherry, B. S. G., and Tavakol, R. K. (1982). Variations in surface air temperatures: Part 2. Arctic regions 1881

1890. Monthly Weather Review, 110, 71

83.

Kerr, R. A. (1984). Carbon dioxide and the control of the ice ages. Science, 223, 1053

1054.

Knox, F., and McElroy, M. B. (1984). Changes in atmospheric CO₂: influence of the marine biota at high latitude. Journal of Geophysical Research, 89, 4629

4637.

Köppen, W. (1873). Uber Mehrjährige Perioden der Witterung, insbesondere Über die 11-jährige Periode der Temperatur. (Concerning multi-year climatic periods, in particular l1-year periods of temperature.) Zeitschrift der Österreichischen Gesellschaft für Meteorologie, 8, 241

248: 257

267.

Köppen, W. (1914). Lufttemperaturen, Sonnenflecken and Vulkanausbrüche. (Air temperatures, sunspots, and volcanic eruptions.) Meteorologische Zeitschrift, 31, 305

328.

Kukla, G., Berger, A., Lotti, R., and Brown, J. (1981). Orbital signature of interglacials. Nature. 290, 295

300.

Kutzbach, J. E. (1981). Monsoon climate of the early Holocene: climatic experiment using the earth's orbital parameters for 9000 years ago. Science, 214, 59

61.

Kutzbach, J. E., and Otto-Bliesner, B. L. (1982). The sensitivity of the African-Asian monsoonal climate to orbital parameter changes for 9000 years B.P. in a low-resolution general circulation model. Journal of Atmospheric Science, 39, 1177

1188.

Kutzbach, J. E., and Guetter, P. J. (1984). Sensitivity of late-glacial and Holocene climates to the combined effects of orbital parameter changes and lower boundary condition changes: 'snapshot' simulations with a general circulation model for 18000, 9000 and 6000 years ago. Annals of Glaciology, 5, 85

87.

Kutzbach, J. E., and Street-Perrott, F. A. (1985). Milankovitch forcing of fluctuations in the level of tropical lakes from 18 to 0 kyr B.P. Nature, 317, 130

134.

Lacis, A., Hansen, J., Lee, P., Mitchell, T., and Lebedeff, S. (1981). Greenhouse effect of trace gases, 1970

1980. Geophysical Research Letters, 8, 1035

1038.

Lamb, H. H. (1970). Volcanic dust in the atmosphere; with a chronology and assessment of its meteorological significance. Philosophical Transactions, Royal Society, London, Series A, 266, 425

533. (See also the updates given by Lamb in Climate Monitor, 6, 57

67, 1977 and Climate Monitor, 12, 79

90, 1983.)

Leith, C. E. (1984). Global climate research. In Houghton, J. T. (Ed.), The Global Climate: 13

24. Cambridge University Press.

Lloyd, C. R. (1984). Pre-Pleistocene paleoclimates: the geological and paleontological evidence; modeling strategies, boundary conditions, and some preliminary results. In Saltzman, B. (Ed.), Advances in Geophysics (Vol. 26): 35

140. Academic Press.

Lough, J. M., Wigley, T. M. L., and Palutikof, J. P. (1983). Climate and climate impact scenarios for Europe in a warmer world. Journal of Climate and Applied Meteorology, 22, 1673

1684.

MacCracken, M. C., and Moses, H. (1982). The first detection of carbon dioxide effects: workshop summary, 8

10 June 1981, Harpers Ferry, W. Va. Bulletin of the American Meteorological Society, 63, 1164

1178.

Madden, R. A. (1976). Estimates of the natural variability of time averaged sea-level pressure. Monthly Weather Review, 104, 942

952.

Madden, R. A., and Ramanathan, V. (1980). Detecting climate change due to increasing carbon dioxide. Science, 209, 763

768.

Madden, R. A., and Shea, D. J. (1978). Estimates of the natural variability of time averaged temperatures over the United States. Monthly Weather Review, 106, 1695

1703.

Manabe, S., and Broccoli, A. J. (1984). Ice-Age climate and continental ice sheets: some experiments with a general circulation model. Annals of Glaciology, 5, 100

105.

Manabe, S., and Stouffer, R. J. (1979). A CO₂-climate sensitivity study with a mathematical model of the global climate. Nature, 282, 491

493.

Manabe, S., and Stouffer, R. J. (1980). Sensitivity of a global climate model to an increase of CO₂ concentration in the atmosphere. Journal of Geophysical Research, 85, 5529

5554.

Manabe, S., and Wetherald, R. T. (1980). On the distribution of climate change resulting from an increase in CO₂ content of the atmosphere. Journal of the Atmospheric Sciences, 37, 99

118.

Manabe, S., Wetherald, R. T., and Stouffer, R. J. (1981). Summer dryness due to an increase of atmospheric CO₂ concentration. Climatic Change, 3, 347

386.

Mercer, J. H. (1969). The Allerod oscillation: A European climatic anomaly? Arctic and Alpine Research, 1, 227

234.

Mercer, J. H. (1978). West Antarctic ice sheet and CO₂ greenhouse effect: a threat of disaster. Nature, 271, 321

325.

Milankovitch, M. M. (1941). Canon of Insolation and the Ice Age Problem. Königlich Serbische Akademie, Beograd. English translation by the Israel Program for Scientific Translations, published for the United States Department of Commerce and the National Science Foundation, Washington, D.C.

Miles, M. K., and Gildersleeves, P. B. (1977). A statistical study of the likely causative factors in the climatic fluctuations of the last 100 years. Meteorological Magazine, 106, 314

322.

Miles, M. K., and Gildersleeves, P. B. (1978). Volcanic dust and changes in Northern Hemisphere temperature. Nature, 271, 735

736.

Mitchell, J. M. Jr. (1953). On the causes of instrumentally observed temperature trends. Journal of Meteorology, 10, 244

261.

Mitchell, J. M. Jr. (1961). Recent secular changes of global temperature. Annals of the New York Academy of Science, 95, 235

250.

Mitchell, J. M. Jr. (1963). On the world-wide patterns of secular temperature change. Changes of Climate, Arid Zone Research XX, UNESCO, Paris, 161

181.

Mitchell, J. M. Jr. (1983). Empirical modeling of effects of solar variability, volcanic events and carbon dioxide on global-scale average temperature since AD 1880. Weather and Climate Responses to Solar Variations. McCormac, B. M. (Ed.): Colorado Associated University Press, 265

273.

Moses, H., and MacCracken, M. C., Coordinators (1982). Proceedings of the Workshop on First Detection of Carbon Dioxide Effects. DOE Report CONF-8106214, U.S. Dept. of Energy Carbon Dioxide Research and Assessment Program, Washington D.C.: 546 pages.

Mosley-Thompson, E., and Thompson, L. G. (1982). Nine centuries of microparticle deposition at the South Pole. Quaternary Research, 17, 1

13.

Namias, J. (1980). Some concomitant regional anomalies associated with hemispherically averaged temperature variations. Journal of Geophysical Research, 85, 1580

1590.

Neftel, A., Oeschger, H., Schwander, J., Stauffer, B., and Zumbrunn, R. (1982). Ice core sample measurements give atmospheric CO₂ content during the past 40,000 years. Nature, 295, 220

223.

Neftel, A., Moor, E., Oeschger, H., and Stauffer, B. (1985). Evidence from polar ice cores for the increase in atmospheric CO₂ in the past two centuries. Nature, 315, 45

47.

Nicholson, S. E. (1980). The nature of rainfall fluctuations in sub-tropical West Africa. Monthly Weather Review, 108, 473

487.

Nicholson, S. E. (1981). Rainfall and atmospheric circulation during drought periods and wetter years in West Africa. Monthly Weather Review, 109, 2191

2208.

North, G. R., Mengel, J. G., and Short, D. A. (1983). Simple energy balance model resolving the seasons and the continents: application to the astronomical theory of the ice ages. Journal of Geophysical Research, 88, 6576

6586.

N.R.C. (Carbon Dioxide Assessment Committee, Board on Atmospheric Sciences and Climate, National Research Council) (1983). Changing Climate. National Academy Press, Washington D.C.: 496 pages.

Oeschger, H., Beer, J., Siegenthaler, U., and Stauffer, B. (1984). Late glacial climate history from ice cores. In Hansen, J. E., and Takahashi, T. (Eds.) Climate Processes andClimate Sensitivity (Maurice Ewing Series, No. 5): 299

306. American Geophysical Union, Washington, D.C.

Oort, A. H. (1978). Adequacy of the rawinsonde network for global circulation studies tested through numerical model output. Monthly Weather Review, 106, 174

195.

Owen, T., Cess, R. D., and Ramanathan, V. (1979). Enhanced CO₂ greenhouse to compensate for reduced solar luminosity on early Earth. Nature, 277, 640

642.

Paltridge, G., and Woodruff, S. (1981). Changes in global surface temperature from 1880 to 1977 derived from historical records of sea surface temperature. Monthly Weather Review, 109, 2427

2434.

Palutikof, J. P., Wigley, T. M. L., and Lough, J. M. (1984). Seasonal Scenarios for Europe and North America in a High-CO₂, Warmer World. U.S. Dept. of Energy, Carbon Dioxide Research Division, Technical Report TR012: 70 pages.

Parker, D. E. (1980). Climatic change or analysts' artifice-a study of grid-point upper-air data. Meteorological Magazine, 109, 129

152.

Parker, D. E. (1981). The effects of inadequate sampling and of circulation pattern on real and apparent zonal mean temperature. Meteorological Magazine, 110, 200

204.

Parker, D. E. (1985). The influence of the Southern Oscillation and volcanic eruptions on temperature in the tropical troposphere. Journal of Climatology, 5, 273

282.

Parkinson, J. H., Morrison, L. V., and Stephenson, F. R. (1980). The constancy of the solar diameter over the past 250 years. Nature, 288, 548

551.

Peterson, G. M., Webb, T., Kutzbach, J. E., van der Hammen, T., Wijmstra, T. A., and Street, F. A. (1979). The continental record of environmental conditions at 18,000 yr BP: an initial evaluation. Quaternary Research, 12, 47

82.

Pisias. N. G., and Shackleton, N. J. (1984). Modelling the global climate response to orbital forcing and atmospheric carbon dioxide change. Nature, 310, 757

759.

Pittock, A. B., and Salinger, J. M. (1982). Towards regional scenarios for a CO₂ warmed Earth. Climatic Change, 4, 23

40.

Pivovarova, Z. I. (1977). Radiation characteristics of Climate in the USSR. Leningrad, Gidrometeoizdat, (in Russian).

Ouenouille, M. H. (1952). Associated Measurements. Butterworth Scientific: 242 pages.

Ramanathan, V., Cicerone, R. J., Singh, H. B., and Kiehl, J. T. (1985). Trace gas trends and their potential role in climate change. Journal of Geophysical Research, 90, 5547

5566.

Raper, S. C. B., Wigley, T. M. L., Jones, P. D., Kelly, P. M., Mayes, P. R., and Limbert, D. W. S. (1983). Recent temperature changes in the Arctic and Antarctic. Nature, 306, 458

459.

Raper, S. C. B., Wigley, T. M. L., Mayes, P. R., Jones, P. D., and Salinger, M. J. (1984). Variations in surface air temperatures: Part 3. The Antarctic, 1957

1982. Monthly Weather Review, 112, 1341

1353.

Raynaud, D., and Barnola, J. M. (1985). An Antarctic ice core reveals atmospheric CO₂ variations over the past few centuries. Nature, 315, 309

311.

Reed, R. K. (1979). On the relationship between the amount and frequency of precipitation over the ocean. Journal of Applied Meteorology, 18, 692

696.

Reed, R. K., and Elliott, W. P. (1977). The comparison of oceanic precipitation as measured by gauge and assessed from weather reports. Journal of Applied Meteorology, 16, 983

986.

Reed, R. K., and Elliott, W. P. (1979). New precipitation maps for the North Atlantic and North Pacific Oceans. Journal of Geophysical Research, 84, 7839

7846.

Rind, D., and Lebedeff, S. (1984). Potential Climatic Impacts of Increasing Atmospheric CO₂ with Emphasis on Water Availability and Hydrology in the United States. U.S. Environmental Protection Agency Report.

Robock, A. (1979). The `Little Ice Age': Northern Hemisphere average observations and model calculations. Science, 206, 1402

1404.

Rodda, J. C. (1969). Rainfall measurement problems. IASH, Proceedings of the Bern Association, 78, 215

231.

Roemmich, D., and Wunsch, C. (1984). Apparent changes in the climatic state of the deep North Atlantic Ocean. Nature, 307, 447

450.

Roxburgh, I. W. (1980). Long term variations in the solar constant. In Sun and Climate, CNES/SNRS/DGRST Conference Proceedings: 261

268.

Ruddiman, W. F., and McIntyre, A. (1981). North Atlantic ocean during the last deglaciation. Palaeogeography, Palaeoclimatology, Palaeoecology, 35, 145

214.

Sarmiento, J. L., and Toggweiler, J. R. (1984). A new model for the role of the oceans in determining atmospheric PCO₂. Nature, 308, 621

624.

Sarnthein, M. (1978). Sand deserts during glacial maximum and climatic optimum. Nature, 272, 43

45.

Schneider, S. H., and Londer, R. (1984). The Coevolution of Climate and Life. Sierra Club Books, San Francisco: 563 pages.

Schneider, S. H., and Thompson, S. L. (1981). Atmospheric CO₂ and climate: importance of the transient response. Journal of Geophysical Research, 86, 3135

3147.

Semtner, A. J. (1984). On modeling the seasonal thermodynamic cycle of sea ice in studies of climatic change. Climatic Change, 6, 127

137.

Shackleton, N. J., Hall, M. A., Line, J., and Can Shuxi, (1983). Carbon isotope data in core V19

30 confirms reduced carbon dioxide of the ice-age atmosphere. Nature, 306, 319

322.

Shackleton, N. J., Backman, J., Zimmerman, H., Kent, D.V., Hall, M. A., Roberts, D. G., Schnitker, D., Baldauf, J. G., Desprairies, A., Homrighausen, R., Huddlestun, P., Keene, J. B., Kaltenback, A. J., Krumsiek, K. A. O., Morton, A. C., Murray, J. W., and Westberg-Smith, J. (1984). Oxygen isotope calibration of the onset of ice-rafting and history of glaciation in the North Atlantic region. Nature, 307, 620

623.

Shackleton, N. J., and Opdyke, N. D. (1973). Oxygen isotope and paleomagnetic stratigraphy of equatorial Pacific core V28-238: Oxygen isotope temperatures and ice volumes on a 10⁵ year and 10⁶ year scale. Quaternary Research, 3, 39

55.

Siegenthaler, U. (1984). 19th century measurements of atmospheric CO₂-a comment. Climatic Change, 6, 409

411.

Siegenthaler, U., Eicher, U., and Oeschger, H. (1984). Lake sediments as continental ¹⁸Orecords from the glacial/post-glacial transition. Annals of Glaciology, 5, 149152.

Siegenthaler, U., and Oeschger, H. (1984). Transient temperature changes due to increasing CO₂ using simple models. Annals of Glaciology, 5, 153

159.

Siegenthaler, U., and Wenk, Th. (1984). Rapid atmospheric CO₂ variations and ocean circulation. Nature, 308, 624

626.

Spelman, M. J., and Manabe, S. (1984). Influence of oceanic heat transport upon the sensitivity of a model climate. Journal of Geophysical Research, 89, 571

586.

Stauffer, B., Hofer, H., Oeschger, H., Schwander, J., and Siegenthaler, U. (1984). Antarctic CO₂ concentration during the last glaciation. Annals of Glaciology, 5, 160

164.

Stefanick, M. (1981). Space and time scales of atmospheric variability. Journal of the Atmospheric Sciences, 38, 988

1002.

Stone, P. H. (1984). Feedbacks between dynamical heat fluxes and temperature structure in the atmosphere. In Hansen, J. E., and Takahashi, T. (Eds.) Climate Processes and Climate Sensitivity (Maurice Ewing Series, No. 5): 6

17. American Geophysical Union, Washington, D.C.

Street, F. A., and Grove, A. T. (1979). Global maps of lake-level fluctuations since 30,000 yr BP. Quaternary Research, 12, 83

118.

Swift, J. H. (1984). A recent (Q-S shift in the deep water of the northern North Atlantic. In Hansen, J. E., and Takahashi, T. (Eds.), Climate Processes and Climate Sensitivity (Maurice Ewing Series, No. 5): 39

47. American Geophysical Union, Washington, D.C.

Tabony, R. C. (1981). A principal component and spectral analysis of European rainfall. Journal of Climatology, 1, 283

294.

Trenberth, K. E., and Shin, W.-T. K. (1984). Quasi-biennial fluctuations in sea level pressures over the Northern Hemisphere. Monthly Weather Review, 112, 761

777.

Thompson, S. L., and Schneider, S. H. (1982). Carbon dioxide and climate: the important of realistic geography in estimating the transient temperature response. Science, 217, 1031

1033.

Vinnikov, K. Ya., and Groisman, P. Ya. (1979). Empirical model of modern climatic change. Meteorologiya i Gidrologiya 1979, no. 3, 26

36 (in Russian).

Vinnikov, K. Ya., and Groisman, P. Ya. (1981). The empirical analysis of CO₂ influence on the modern changes of the mean annual Northern Hemisphere surface air temperature (in Russian). Meteorologiya i Gidrologiya 1981, no. 11, 30

43.

Vinnikov, K. Ya., and Groisman, P. Ya. (1982). The empirical study of climate sensitivity. Isvestiya AS USSR, Atmospheric and Oceanic Physics, 18(11), 1159

1169 (in Russian).

Vinnikov, K. Ya., Gruza, G. V., Zakharov, V. F., Kirillov, A. A., Kovyneva, N. P., and Ran'kova, E. Ya. (1980). Current climatic changes in the Northern Hemisphere. Meteorologiya i Gidrologiya 1980, no. 6, 5

17 (in Russian).

Vinnikov, K. Ya., and Kovyneva, N. P. (1983). Global warming: the distribution of climatic change. Meteorologiya i Gidrologiya 1983, no. 5, 10

19 (in Russian).

Wallén C. C. (1984). Present century climate fluctuations in the Northern Hemisphere and examples of their impact. WCP-87, WMO/TD-No. 9, World Meteorological Organization, Geneva.

Watts, R. G. (1985a). Climatic transients caused by variations in the thermohaline circulation. Extended Summaries, Third Conference on Climate Variations and Symposium on Contemporary Climate 18502100, American Meteorological Society, 83

84.

Watts, R. G. (1985b). Global climate variation due to fluctuations in the rate of deep water formation. Journal of Geophysical Research, 90, 8067

8070.

Weare, B. C. (1979). Temperature statistics of short-term climatic change. Monthly Weather Review, 107, 172

180.

Webb, T., III and Wigley, T. M. L. (1985). What past climates can tell us about a warmer world. In MacCracken, M. C., and Luther, F. M. (Eds). The Potential Climatic Effects of Increasing Carbon Dioxide, 237

257. U.S. Dept. of Energy, Carbon Dioxide Research Division, Washington D.C.

Weller, G., Baker, J. D., Gates, W. L., MacCracken, M. C., Manabe, S., and von der Haar, T. (1983). Detecting and monitoring of CO₂-induced climate changes. In Changing Climate. Report of the Carbon Dioxide Assessment Committee, National Research Council: 292

382. National Academy Press, Washington D.C.

Wigley, T. M. L. (1983). The pre-industrial carbon dioxide level. Climatic Change, 5, 315

320.

Wigley, T. M. L. (1985). Carbon dioxide, trace gases and global warming. Climate Monitor, 13, 133

148.

Wigley, T. M. L., Angell, J. K., and Jones, P. D. (1985). Analysis of the temperature record. In MacCracken, M. C., and Luther, F. M. (Eds). Detecting the Climatic Effects of Increasing Carbon Dioxide, 55

90. U.S. Dept. of Energy, Carbon Dioxide Research Division, Washington D.C.

Wigley, T. M. L., and Brimblecombe, P. (1981). Carbon dioxide, ammonia and the origin of life. Nature, 291, 213

215.

Wigley, T. M. L., and Jones, P. D. (1981). Detecting CO₂-induced climatic change. Nature, 292, 205

208.

Wigley, T. M. L., and Jones, P. D. (1982). Signal-to-noise ratios for surface air temperature and the detection of CO₂-induced climatic change. In Moses, H., and MacCracken, M. C. (Eds.), Proceedings of the Workshop on First Detection of Carbon Dioxide Effects. DOE Report CONF-8106214, U.S. Dept. of Energy Carbon Dioxide Research and Assessment Program, Washington D.C.: 143

158.

Wigley, T. M. L., Jones, P. D., and Kelly, P. M. (1980). Scenarios for a warm, high-CO₂ world. Nature, 283, 17

21.

Wigley, T. M. L., and Schlesinger, M. E. (1985). Analytical solution for the effect of increasing CO₂ on global mean temperature. Nature, 315, 649

652.

Williams, J. (1980). Anomalies in temperature and rainfall during warm Arctic seasons as a guide to the formulation of climate scenarios. Climatic Change, 2, 249

266.

Williams, L. D., and Wigley, T. M. L. (1983). A comparison of evidence for late Holocene summer temperature variations in the Northern Hemisphere. Quaternary Research, 20, 286

307.

Woillard, G. M. (1978). Grande Pile peat bog: a continuous pollen record for the last 140,000 years. Quaternary Research, 9, 1

21.

Woillard, G. M. (1979). Abrupt end of the last interglacial s.s. in north-east France. Nature, 281, 558

562.

W.M.O. (1982). WMO Global Ozone Research and Monitoring Project, Report No. 14 of the Meeting of Experts on Potential Climatic Effects of Ozone and Other Minor Trace Gases, Boulder, Colorado (13

17 Sept. 1982). World Meteorological Organization, Geneva: 35 pages.

W.M.O. (1983). Report of the WMO(CAS) Meeting of Experts on the CO₂ Concentrations from Pre-industrial Times to I.G.Y., WCP-53. Bojkov, R. (Ed.), World Meteorological Organization, Geneva.

Yamamoto, R. (1981). Change of global climate during recent 100 years. Proceedings of the Technical Conference on ClimateAsia and Western Pacific, 15

20 Dec. 1980. Guangzhou, China. WMO Report No. 578. World Meteorological Organization, Geneva.

Yamamoto, R., and Hoshiai, M. (1979). Recent change of the Northern Hemisphere mean surface air temperature estimated by optimum interpolation. Monthly Weather Review, 107, 1239

1244.

Yamamoto, R., and Hoshiai, M. (1980). Fluctuations of the Northern Hemisphere mean surface air temperature during recent 100 years. Journal of the Meteorological Society of Japan, 58, 187

193.

	6.1 INTRODUCTION
	6.2 PAST CLIMATIC CHANGE
		6.2.1 Relevance to the Greenhouse Gases Issue
		6.2.3 The 10³10⁵ Year Time Scale6.2.2 CO₂ Effects on the 10⁶10⁹ Year Time Scale
		6.2.4 The interval 6000 BP to 1850 AD
		6.2.5 Recent Climatic Change: 1850 AD to Present
	6.3 CLIMATE SCENARIOS
		6.3.1 Introduction
		6.3.2 Potential Analogues
		6.3.3 Instrumental Scenario Construction
		6.3.4 The Relevance of Scenarios
	6.4 DETECTION OF CLIMATIC CHANGE
		6.4.1 Background
		6.4.2 The Signal-to-noise Ratio Concept
		6.4.3 Statistical Strategies
		6.4.4 A Simple Analysis of the Recent Surface Air Temperature Record
	6.5 MONITORING REQUIREMENTS
	6.6 CONCLUSIONS
	6.7 REFERENCESACKNOWLEDGEMENTS

6	Empirical Climate Studies
	Warm World Scenarios and the Detection of Climatic Change Induced by Radiatively Active Gases
	T. M. L. WIGLEY, P. D. JONES and P. M. KELLY