 Section 10 of Understanding Climatic Change This is a LibreVox recording. All LibreVox recordings are in the public domain. For more information or to volunteer, please visit LibreVox.org, recording by Anna Simmel. Understanding Climatic Change, a program for action by the US Committee for the Global Atmospheric Research Program. Section 10, Appendix A, Survey of Past Climates, Part 2 Geographic Patterns of Climatic Change While the chronology of certain features of climatic change may be revealed by the analysis of instrumental and paleoclimatic data at individual sites, the geographic pattern of these changes is an equally important characteristic. From what we know of the behaviour of the present atmosphere, it would be remarkable if there were not a definite spatial structure to the variations on all climatic times scales. The search for these patterns requires synoptic data for the various climatic elements and this is presently available only from the records of modern observations and from a few marine proxy sources. Structure Revealed by Observational Data The task of describing the spatial and temporal structure of climatic variations from the observations of the instrumental era is far from complete. Most studies have therefore focused primarily on local or regional climatic changes. Lambe and Johnson, 1961-1966, have made comprehensive analyses of inter-hemispheric and inter-hemispheric climatic indices and the statistical structure of these circulation variations has been studied by Willett, 1967, Wagner, 1971, Eudin, 1967, Breyer, 1968 and Kutzbach, 1970. Such analyses, especially of hemispheric pressure data, reveal that the year-to-year and decade-to-decade variations have a spatial structure that may be associated with amplitude and phase changes of the long planetary waves in the atmosphere. The essentially two-dimensional character of climate is masked in studies of zonally-average parameters, although these may be useful for other purposes. An example of the importance of both zonal and nonzonal spatial variability of the atmospheric circulation is provided by the first eigenvector pattern or empirical orthogonal function of hemispheric pressure for January shown in figure A16 as well as by the patterns of pressure, temperature and rainfall variability shown in figures A17, A18 and A19. These data suggest an association between the changes in the monthly average intensity and position of the Elysian and Icelandic lows. For example, during the first two decades of this century there has been a tendency for decreased intensity and westward extension of the Elysian low, coupled with an increased intensity and north-eastward shift of the Icelandic low. LAM 1966 and NAMIAS 1970 have described important regional changes in temperature and precipitation associated with these circulation changes. The opposite tendency has prevailed since the mid-1950s and LAM 1966, Winstonley 1973 and Bryson 1974 have described the possible relationships between the changing mid-latitude circulation patterns of the 1960s, the equatorial shift of the subtropical highs and the increasing frequency of droughts along the southern fringes of the monsoon lands of the northern hemisphere, see figure A8. These changes appear to reflect an equatorward extension of the westerly wave regime and a contraction of the Hadley circulation, although much further analysis is clearly required to confirm such a conjecture. Interhemispheric relationships of climatic indices have been and remain less amenable to study because of the general lack of observations from the southern hemisphere. Observations are sufficient however to show that the circulation in the southern hemisphere is somewhat stronger and steadier than that in the northern hemisphere. Whether this results in the southern hemisphere circulation leading that in the northern hemisphere, or whether variable features in the equatorial circulation influence both hemispheres similarly is not presently known. Bjergnes 1969 B, Fletcher 1969, LAM 1969, NAMIAS 1972 A 1972 B. It is likely that interhemispheric relationships of one sort or another are important for the understanding of climatic variations and that our ability to describe them will require the availability of much more comprehensive data than now exist from the southern hemisphere, the equatorial region and the oceanic and polar regions of the northern hemisphere. The present accumulation of upper-air data, especially in the northern hemisphere since the early 1950s however, has permitted a beginning of the study of the three-dimensional spatial and temporal variability of the general circulation. A foundation of basic statistics is provided by calculations of the means and variances of standard meteorological variables. See for example, Crutcher and Miserve 1970, Tulliard et al. 1969, and by atlases of energy budgets, Vettico 1963. The covariance structure of circulation patterns at 700 millibars in the northern hemisphere is treated by O'Connor 1969 and other aspects of the tropospheric circulation have been considered by Gommel 1963 and while 1972. The most comprehensive analysis of atmospheric circulation statistics however, is that based on the period 1958 to 1963 as undertaken by Ord and Rasmuson 1971. While this work documents the monthly, seasonal and annual variations of many features of the observed general circulation, in the northern hemisphere, it does not directly address many of the variables of primary climatic interest. Using the same data set however, Star and Ord 1973 have reported an unmistakable downward trend of the mean air temperature in the northern hemisphere of 0.6 degrees Celsius over the five-year interval shown in figure 820. Diagnostic studies of this type represent great investments of time and effort but are essential steps to work the monitoring of climate and an assessment of the mechanisms of climatic variation. A complete description of climatic changes from instrumental records must also include studies of the momentum and energy budgets of the atmosphere and oceans and their variability with time over many years and decades. While this must remain largely a task for the future, several efforts have established the existence of significant inter-annual variations in the atmosphere. Guru Granthal 1965 have discussed the inter-annual variations of available potential energy and Kang and Sung 1969 have described the fluctuations of the atmospheric kinetic energy budget. As noted previously, the inter-annual variations of pole-wood angular momentum and energy fluxes have been studied comprehensively by Ord and Rasmussen 1971. A measure of this variability is shown in figure A21 and amounts to about 30% of the mean transports. The unique global potential of satellite-based measurements has been exploited by van der Haag and Suomi 1971 who have summarized satellite measurements of planetary albedo and of the planetary radiation budget for the 5 years 1962 to 1966. They found large inter-annual variations in the zonally averaged equator-to-pole gradient of the net radiation as shown in figure A22. This forcing function can now be monitored routinely by meteorological satellites and opens the door to more detailed studies of atmospheric energetics than heretofore possible, Winston 1969. Van der Haag and Ord 1973 have combined satellite measurements of the Earth's radiation budget with atmospheric energy transport calculations to produce a new estimate of the pole-wood energy transport by the northern hemisphere oceans. They find that the oceanic heat transport averages about 40% of the total in the 0-70°N latitude band and accounts for more than half at many latitudes. Another example of the use of satellite-derived measurements of climatic indices is given by Kukla and Kukla 1974. Their measurements of the inter-annual changes in the area of snow and ice cover in the northern hemisphere are shown in figure A23 and reveal year-to-year fluctuations by the order of a few percent. Note, however, the relatively large change during 1971 and the subsequent maintenance of extensive snow and ice coverage and an associated increase of the reflected solar radiation. Time variations of the surface energy budget on a global scale are not available from direct observations and must be inferred from the conventional measurements of temperature, humidity, cloudiness, wind and radiation. Fletcher 1969 has drawn attention to the variations in the energy budget of polar regions as a function of variable sea ice conditions, while Sawyer 1964 has noted the possible role of fluctuations in the surface energy budget as a cause of inter-annual variations of the general circulation itself. A number of observational studies of large-scale interaction between the ocean and the atmosphere have illustrated the complexity and importance of this mechanism. See, for example, Whale 1968 and Lamb and Redcliff 1972. Bjergnes 1969 B has considered the response of the North Pacific westerlies to anomalies of equatorial sea surface temperature and variations in the Hadley circulation, while Namius 1969 and 1972 B has described positive feedback relationships between large-scale patterns of ocean surface temperature in mid-latitudes and the circulation of the overlying atmosphere. Such modes of atmosphere-ocean coupling may be important parts of the climatic fluctuations and must be given further study. In summary, we may say that observational data at the Earth's surface show that during the period 1900 to 1940 the Northern Hemisphere as a whole warmed, although some areas, mainly the Atlantic sector of the Arctic and Northern Siberia, warmed far more than the global average. Some areas became colder and others showed little measurable change until 1963. In the time since 1940 an overall cooling has occurred, but is again characterized by a geographical structure. Cooling since 1958 has occurred in the subtropical arid regions and in the Arctic, star and oared, 1973. There is also some evidence that the Northern Hemisphere oceans are cooling, Namius 1972 B, although the oceanic database necessary to confirm this has not yet been assembled. Structure revealed by paleoclimatography. Most of the work done to date on climatic change beyond the timeframe encompassed by meteorological observations represents a study of time series taken at specific sites. This lack of synoptic data on the longer-range climatic changes is a serious handicap to the portrayal and understanding of the mechanisms involved. In order to underscore these points and to encourage further research we present here examples of the few proxy data that have been assembled to reveal a spatial structure of climatic change. Distribution of ice sheets. The continental margins of the Northern Hemisphere ice sheets at their maximum extension during the last million years are clearly marked by the debris deposits in terminal morands while the extent of sea ice is recorded by features preserved in marine sediments. Figure A24 shows the distribution of maximum ice cover and Table A2 gives statistics of the areas of the individual continental ice sheets. In North America the ice extended as far south as 40 degrees north and spanned the entire width of the continent while in Europe the ice sheet extended only to about 50 degrees north. Note that large regions in eastern Siberia were unglaciated. Sea surface temperature patterns. The north-south migration of polar waters in the North Atlantic in response to major cycles of glaciation is shown in Figure A25. During glacial maxima these waters were found as far south as 40 degrees north well beyond the present extent of polar waters. A synoptic analysis of the ocean surface temperatures of 18,000 years ago at about the time of the last glacial maximum is shown in Figure A26. These temperature estimates have been derived by multivariate statistical techniques applied to planktonic organisms as preserved in about 100 deep seacoals in scattered locations across the North Atlantic. McIntyre et al. 1974. The most striking feature of this glacial age map is the extensive southward displacement of the 10 to 14 degrees Celsius water while the warmer water was found in nearly its present position. In parts of the Sargasso Sea the glacial age ocean was slightly warmer than it is today. Because the atmosphere receives much of its heat from the sea such estimates of sea surface temperature are likely to be important in developing a satisfactory reconstruction of past climates and it is therefore important to consider their reliability. Berger 1971 for example has suggested that carbonate dissolution on the seabed may distort the taxonomic composition to the fossil fauna on which such paleotemperature estimates are based. Kipp 1974 on the other hand shows that when the statistical transfer functions are calibrated on materials that incorporate the dissolution effects an unbiased estimate of such parameters as the sea surface temperature can be obtained. The temperature reconstruction in Figure A26B is based on the statistics of the forum minifural fauna distribution and encompasses 91% of the variance of the data McIntyre 1974. The 80% confidence interval of each of the cores is about 1.8 degrees Celsius, Kipp 1974. Shackleton and Obdike 1973 using a revised isotopic method based on the difference between oxygen 18 values in benthic and planktonic species have provided an independent confirmation of the sea surface temperature estimates of Embry et al. 1973 for a portion of the glacial age Caribbean. Other reconstructions of paleo-ocean surface temperatures have been based on data from radiolaria, coccoliths and forum minifera and although some discrepancies are revealed where independent data are available the derived ocean temperatures show considerable spatial coherence McIntyre 1974 such estimates of past sea surface temperature will prove useful in climatic simulations with numerical general circulation models C appendix B Patterns of vegetation change Figure A27 illustrates the use of fossil pollen data to record changes in vegetation determining the deglaciation of eastern North America during the interval 11,000 to 9,000 years ago. At the beginning of this time pine species occupied sites in the southeastern Appalachians but as the ice retreated the pine moved farther north and west to colonize newly uncovered areas. A relatively complete chronology of the retreat of the Laurentide ice sheet itself is given by radiocarbon dating Bryson et al. 1969 Patterns of aridity For only four desert areas in the world do we have enough information to plot aridity as a function of time and even in these areas the record extends back only a few tens of thousands of years. As shown in figure 828 the data suggests a degree of synchronyity between the two African regions and the Great Basin while the records from the Middle East are quite different None of the data from closed basin lakes show significant correlation with the glacial record and we are clearly a long way from understanding the response of arid regions to glacial cycles. More generally insufficient research has been devoted to the role of desert regions in the processes responsible for the climate of the earth. Patterns of tree rain growth Changes of thickness of the growth rings added by trees each year reflect environmental change in a complex way. By appropriate calibration such data may be made to furnish significant climatic information for the past several hundred to several thousand years. Studies of many tree ring series over a wide geographic area can moreover provide accurately dated synoptic evidence of regional climatic patterns. Fritz 1965 Fritz et al. 1971 have demonstrated the feasibility of reconstructing the anomalies of sea level pressure and temperature from the spatial patterns of tree growth over western North America. Examples of such synoptic maps based on average decadal growth are given in figure 829. Although such reconstructions show considerable variation in the year-to-year climatic states the inferred variations in the intensity of Icelandic and Aeluthian lows, for example, are similar to those described in the modern record Kutzbach 1970 The development of an expanded network of tree ring sites could significantly broaden our knowledge of the patterns of climatic fluctuations over the past several centuries. Summary of the climatic record In this survey of past climates the characteristic time and spatial structures of climatic variations have been discussed as though there were sufficient data to document large regions of the globe. This is true only for the more recent parts of the instrumental period as there are large gaps in the presently available historical and proxy climatic records. With these limitations in mind it is nevertheless useful to summarize the general characteristics of the climatic record. 1. The last post-glacial thermal maximum was reached about 6,000 years ago and climates since then have undergone a gradual cooling. This trend has been interrupted by three shorter periods of more marked cooling similar to the so-called little ice age of AD 1430 to 1850 each followed by a temperature recovery. The well-documented warming trend of global climate beginning in the 1880s and continuing until the 1940s is a continuation of the warming trend that terminated the little ice age. Since the 1940s, mean temperatures have declined and are now nearly halfway back to the 1880 levels. 2. Climatic changes during the past 20,000 years are as severe as any that occurred during the past million years. At the last glacial maximum extensive areas of the northern hemisphere were covered with continental ice sheets, sea level dropped about 85 meters and sea surface temperatures in the North Atlantic fell by as much as 10 degrees Celsius. At northern mid-lethitude sites not far from the glacial margins locations now occupied by major cities and extensive agricultural activity air temperatures fell markedly, drastic changes occurred in the precipitation patterns and wholesale migrations of animal and plant communities took place. 3. The present interglacial interval which has now lasted for about 10,000 years represents a climatic regime that is relatively rare during the past million years most of which has been occupied by colder glacial regimes. Only during about 8% of the past 700,000 years has the Earth experienced climates as warm as or warmer than the present. 4. The penultimate interglacial age began about 125,000 years ago and lasted for approximately 10,000 years. Similar interglacial ages each lasting 10,000 plus or minus 2,000 years and each followed by a glacial maximum have occurred on the average every 100,000 years during at least the past half million years. During this period fluctuations of the northern hemisphere ice sheets caused sea level variations of the order of 100 meters. In contrast, the east Antarctic ice sheet has apparently varied little since reaching its present size about 5 million years ago while the west Antarctic ice sheet appears to have been disintegrating 5. About 65 million years ago global climates were substantially warmer than today and subsequent changes may be viewed as part of a very long period cooling trend. For even earlier times the proxy climatic evidence becomes increasingly fragmentary. The best documented records suggest two previous extensive glacations occurring about 300 million and 600 million years ago. Future climate some inferences from past behavior. The overall picture of past climatic changes described in this survey suggests the existence of a hierarchy of fluctuations that stand out above the white noise or random fluctuations presumed to exist on all timescales. In addition to the dominant period of about 100,000 years there are apparent quasi-periodic fluctuations with timescales of about 2,500 years and shorter period fluctuations on the order of 100 to 400 years. Each of these explains progressively less of the total variance but may nevertheless be climatically significant. No periodic component of climatic change on the order of decades has yet been clearly established although significant excursions of climate are observed to occur in anomalous groups of years. In view of the limited resolving power of most climatic indicators especially those for the relatively remote geological past it is difficult to establish whether the apparent fluctuations are quasi-periodic or whether they more nearly represent what are basically random Markovian red noise variations. In the case of the longer period variations of 100,000 year and 20,000 year periods there is circumstantial evidence to suggest that these may have been induced in some manner by the secular variations of the Earth's orbital elements which are known to alter the seasonal and latitudinal distribution of solar radiation received at the top of the atmosphere. In other cases the observed variations have yet to be convincingly related to any external climatic control. The mere existence of such variations does not necessarily mean that changes in the external boundary conditions are involved however. The internal dynamics of the climatic system itself may well be the origin of some of these features. Whether forced or not climatic behaviour of this type deserves careful study as the conclusions reached there directly upon the problem of inferring the future climate. The prediction of climate is clearly an enormously complex problem although we have no useful skill in predicting weather beyond a few weeks into the future. We have a compelling need to predict the climate for years, decades and even centuries ahead. Not only do we have to take into account the complex year-to-year changes possibly induced by the internal dynamics of the climatic system and the likely continuation of the yet unexplained quasi-periodic and episodic fluctuations of the last few thousand years discussed above but also the changes induced by possibly even less predictable factors such as the aerosols added to the atmosphere by volcanic eruptions and by man himself. Mitchell, 1973 A, 1973 B These questions lie at the heart of the problem of climatic variation and are given consideration elsewhere in this report. In the face of these uncertainties any projection of the future climate carries a great risk. Nevertheless, we may speculate about the possible cause of global climate in the decades and centuries immediately ahead by making certain assumptions about the character of the major fluctuations noted in the climatic record. In the following paragraphs we attempt to draw together these considerations into an overall assessment of the probable direction and magnitude of present-day climatic change taking into account the risk of a major future change associated with a seemingly inevitable onset of the next glacial period. Potential contribution of sinusoidal fluctuations of various timescales to the rate of change of present-day climate. Estimates of the amplitudes of all the principal climatic fluctuations identified in this report are listed in table A3 where they have been made consistent with the data presented in figure A2 and are expressed in terms of the total range of temperature between maxima and minima. On the assumption that all of these fluctuations can be approximated by quasi-periodic sine waves, the ratio of the amplitude A to the period P of each fluctuation becomes proportional to the maximum contribution of that fluctuation to the rate of change of climate. By considering also the phase of each fluctuation as inferred from the paleoclimatic record the contribution of each fluctuation to the present-day rate of change can be estimated C table A3. Estimation of the phase of each sinusoidal fluctuation indicated by the estimated dates of the last temperature maximum in table A3 permits an assessment of the sine and magnitude of the contribution of each fluctuation to the total rate of change of globally average temperature in the present epoch. The sum of these individual contributions minus 0.015°C per year agrees reasonably well with the observed rate of change of minus 0.01°C per year during the past two decades as determined from analyses of surface climatological data by Rytan 1971 and by Budiko 1969. It should be noted that this trend is dominated by the shortest fluctuations and especially by the fluctuations of the order of 100 years C figure A6. The estimated maximum rate of change associated with all time scales of climatic fluctuations shown in figure A2 is plotted as a continuous function of wavelength in figure A30. The family of curves also shown in this figure indicates the relationship between maximum rate of change and wavelength in Markovian red noise for various degrees of redness characterized by the value of the serial correlation coefficient at a time lag of one year Gilman et al. 1963. By comparison with these curves it is suggested that the observed shorter period climatic fluctuations that is fluctuations of the order of 100 to 200 years are not clearly distinguishable from random fluctuations whereas the longer period fluctuations especially those with periods of 20,000 years or more may be appreciably larger in amplitude than would be expected in random noise. The contributions of the longer period fluctuations to present day climatic change are seen nonetheless to be relatively small. Should the longer period fluctuations be non-sinusoidal or episodic in form rates of change perhaps 10 times larger than the magnitudes shown in figure A30 could be possible. Even such rates however would contribute little over and above the normal inter-annual variability of present day global climate and the cumulative change of climate associated with the longer period fluctuations would remain relatively small until several centuries had elapsed. Despite its simplistic view of climatic change this exercise is an instructive one in that it demonstrates how difficult it would be for the longer period sinusoidal fluctuations to contribute substantially to the changes of climate taking place in the 20th century. If the longer period fluctuations are those that primarily determine the cause of the glacial interglacial succession of global climate it would seem that the transition to the next glacial period even if it has already commenced will require many centuries to accumulate to a drastic shift from present climatic conditions. In assessing such projections however we must keep in mind that our ability to anticipate the locally important synoptic pattern of climatic variations is limited. The work of Mitchell 1963 for example has shown that while the northern hemisphere average air temperatures rose only about 0.2 degrees celsius during the period 1900-1940 there were many areas that deviated markedly from this hemispheric average trend. Parts of the eastern United States for example exhibited a 1 degree celsius rise in average temperature 5 times the hemispheric average Parts of Scandinavia and Mexico showed temperature increases of 2.0 degrees celsius 10 times the hemispheric average while in Spitzbergen the warming was 5 degrees celsius 25 times the hemispheric average The corresponding data on other climatic elements are spares but may be expected to exhibit comparable or even greater spatial variants. Likelihood of a major deterioration of global climate in the years ahead. As noted above the longer period climatic fluctuations seem to be associated with larger amplitudes of change than those consistent with Margovian red noise behaviour. The same cannot be said however of the shorter period fluctuations. For the moment let us suppose that all the fluctuations described in this report are actually random fluctuations in the sense that transitions between successive maxima and minima may occur at random Poisson distributed intervals of time rather than at more or less regular intervals. The probability that one or more transitions of a fluctuation will occur in an arbitrarily specified length of time may then be calculated from the negative binomial distribution. Following this approach we can assess the risk of encountering a change of climate in the years ahead as rapid as the maximum rate of change otherwise associated with sinusoidal climatic fluctuations on each of the characteristic details noted above. Such a measure of risk for time intervals between one year and 1,000 years into the future can be inferred by interpolation between the curves of transition probability in figure A31. The proper interpretation of this figure will be apparent from the following examples. 1. The curve labelled 100,000 in the figure indicates the probability of a major transition of climate in either direction that is normally associated with climatic fluctuations on the time scale of 100,000 years a change of global average temperature of up to perhaps 8 degrees celsius in a total time interval of 50,000 years or less. The curve indicates that if successive transitions of this kind recur at random time intervals as assumed here the onset or termination of such a transition will occur in the next 100 years with a probability of about 0.002 and in the next 1,000 years with a probability of about 0.02. 2. The dashed curve labelled 100 in the figure indicates the probability of one transition of climate in either direction that is normally associated with climatic fluctuations on the time scale of 100 years of up to perhaps 0.5 degrees celsius in a total time interval of about 50 years or less. Such a transition is indicated to have a probability of about 0.02 of occurring in the next year a probability of about 0.16 of occurring in the next 10 years and a probability of about 0.35 of occurring in the next 50 years. The solid line labelled 100 in the figure indicates the probability of one or more transitions of the same kind which rises from about 0.2 in the next 10 years to about 0.8 in the next 100 years. If it can be assumed that the typical duration of such a transition when it occurs is not less than 4 or 5 decades and that only one such transition can occur at the same time then the dashed curve would be the appropriate guide for estimating such probabilities in the next few decades otherwise the solid curve would be a more appropriate guide. When figures A30 and A31 are considered together it is suggested that whether climatic fluctuations are or are not quasi-periodic those that are most relevant to the cause of global climate in the years and decades immediately ahead historical fluctuations and not the longer period glacial fluctuations even if the phase of the longer period changes is such as to contribute to a cooling of present-day climate the contribution of such fluctuations to the rate of change of present-day climate would seem to be swamped by the much larger contributions of the shorter period if more ephemeral historical fluctuations we must remember however that this analysis assumes a simple model of climatic change in which climatic fluctuations of various periods are independent and therefore additive the paleoclimatic record presented here does not preclude the possibility that relatively sudden climatic changes could arise through interactions between fluctuations of different periods one may still ask the question when will the present interglacial end? few paleoclimatologists would dispute that the prominent warm periods or interglacials that have followed each of the terminations of the major glaciations have had durations of 10,000 plus or minus 2,000 years in each case a period of considerably colder climate has followed immediately after the interglacial interval since about 10,000 years has elapsed since the onset of the present period of prominent warmth the question naturally arises as to whether we are indeed on the brink of a period of colder climate Cochlear and Matthews, 1972 have already called attention to such a possibility there seems little doubt that the present period of unusual warmth will eventually give way to a time of colder climate but there is no consensus with regard to either the magnitude or rapidity of the transition the onset of this climatic decline could be several thousand years in the future although there is a finite probability that a serious worldwide cooling could befall the earth within the next 100 years what is the nature of the climatic changes accompanying the end of a period of interglacial warmth from studies of sediments and soils Cochlear finds that major changes in vegetation occurred at the end of the previous interglacial figure A14 the deciduous forests that covered areas during the major glaciation were replaced by sparse shrubs and dust blew freely about the climate was considerably more continental than at present and the agricultural productivity would have been marginal at best the stratification of fossil pollen deposits in eastern Macedonia figure A13 also clearly shows a marked change in vegetative cover between interglacial warmth and the following cold periods the oak pine forest that existed in the area gave way to a step of shrub and grass was a dominant plant cover other evidence from deep sea cores reveals a substantial change in the surface water temperature in the North Atlantic between interglacial and glacial periods figure A13 and the marine sediment data show that the magnitude of the characteristically abrupt glacial cooling was approximately half the total glacial to interglacial change itself the question remains unresolved if the end of the interglacial is episodic in character we are moving toward a rather sudden climatic change of unknown timing although as each 100 years passes we have perhaps a 5% greater chance of encountering its onset if on the other hand these changes are more sinusoidal in character then the climate should decline gradually over a period of thousands of years these are the limits that we can presently place on the nature of this transition from the evidence contained in the paleoclimatic record these climatic projections however could be replaced with quite different future climatic scenarios due to man's inadvertent interference with the otherwise natural variation Mitchell 1973a this aspect of climatic change has recently received increased attention as evidenced by the SMIC report Wilson 1971 a leading anthropogenic effect is the enrichment of the atmospheric CO2 content by the combustion of fossil fuels which has been rising about 4% per year since 1910 there is evidence that the oceans uptake of much of this CO2 is diminishing Keeling et al 1974 which raises the possibility of even greater future atmospheric concentrations man's activities are also contaminating the atmosphere with aerosols and releasing waste heat into the atmosphere either or both of which may have important climatic consequences Mitchell 1973a such effects may combine to offset a future natural cooling trend or to enhance a natural warming this situation serves to illustrate the uncertainty introduced into the problem of future climatic changes by the interference of man and is occurring before adequate knowledge of the natural variations themselves has been obtained again the clear need is for greatly increased research on both the nature and causes of climatic variation end of section 10 section 11 of understanding climatic change this is a LibriVox recording all LibriVox recordings are in the public domain for more information and auto volunteer please visit LibriVox.org recording by Avayee in January 2019 understanding climatic change a program for action by the US committee for the global atmospheric research program section 11 appendix B survey of the climate simulation capability of global circulation models introduction much of the present effort within GARP as well as other research programs in the atmospheric and oceanic sciences is aimed toward the development of a quantitative understanding of the behavior of the atmosphere with the immediate objective of improving the accuracy of weather forecasts other research efforts and plans and the research program proposed in this report in particular are directed to the longer range objective of understanding the physical basis of climate and climatic change essential to both of these objectives are the dynamical models of the global atmospheric and oceanic circulation these general circulation models or GCMs have been developed over a number of years in parallel with the growth of computing capability and the increase of atmospheric data coverage the several atmospheric and oceanic GCMs have now reached the point where reasonably accurate simulations of the global distribution of many important climatic elements are possible and where their coupling into a single dynamical system is now feasible this therefore seems to be a useful time to survey briefly these models climate simulation capabilities here we have not attempted to present a detailed discussion of the various GCMs as such descriptions are readily available both in the literature and in documents describing special models model reviews have recently been prepared by Robinson 1971 Wilson 1973 Smagorinski 1974 and Schneider and Dickinson 1974 and general discussions of the use of such models for weather prediction and for studies of the general circulation are available see for example the review by Smagorinski 1970 and also Holtner 1971 and Lorentz 1967 a survey of the physical and mathematical structure of both regional and global atmospheric models is also in preparation for GARP 1974 what has not been assembled in 1984 is the comparative climatic performance of the various models and this appendix is an initial effort to fill this need for both the atmospheric and oceanic global GCMs in general any formulation that relates variables of the climatic system to the external or boundary conditions may be considered a climatic model we can thus identify basically empirical and statistical climatic models as well as those that rest on the system's dynamical equations within the dynamical climate models a wide variety of the type and degree of parameterization may be seen at one extreme are the vertically and zonally averaged atmospheric models that address the mean heat balance at the earth's surface such as those of Boudicot 1969 and Sellers 1973 in such models the transport of heat is parameterized in terms of mean zonal variables which are in turn related to the surface temperature at the other extreme are the high-resolution global general circulation models or GCMs in these models the details of the transient cyclone scale motions are resolved along with the global distribution of the elements of the heat hydrologic balances even these models however parameterize certain physical processes in that they employ empirical or statistical representations of some of the subgrid scale processes in the surface boundary layer and in the free atmosphere and open ocean such as the effects of diffusion and conviction dynamical climate models also display a wide variety of parameterization with respect to time this ranges from equilibrium or steady state models such as that of Salzman and Wernicke 1971 to the GCMs that explicitly calculate the time dependence of the circulation in steps of a few minutes with respect to their treatment of both space and time therefore a wide range of models exist and each is suited to the investigation of particular aspects of the climatic problem the GCMs of both the atmosphere and ocean provide the most detailed representation of the physical processes involved but require large amounts of computation these models have therefore been used up to the present time to study only the climatic variations on time scales of the order of years for the atmosphere to centuries for the oceans the more highly parameterized models on the other hand provide less detail but are capable of treating the longer period climatic variations with much less computation once they are adequately calibrated with respect to observations an important use of the GCMs will be to generate detailed climatic statistics from which parameterizations appropriate to the various statistical dynamical models may be prepared in the remainder of this appendix we give our attention to the principle atmospheric and oceanic general circulation models for the purpose of indicating their present capability to simulate climate before presenting these results however it is useful to review briefly the historical development of numerical modeling in general development and uses of numerical modeling the basis for the mathematical modeling of the behavior of the atmosphere was first unambiguously stated by V. Bjaknis in 1904 it is only in the last 20 years or so however that the means for carrying out such modeling on a practical basis have become available these include adequate observations for model calibration and verification a knowledge of the important physical processes and their parameterization and the computers and numerical methods necessary to perform the calculations the observational base for numerical modeling of the atmosphere has grown steadily since the 1940s and early 1950s when the global radiosonde network began to take shape the IGY provided further expansion but the observational coverage still needs augmentation especially over the oceanic regions the real breakthrough toward the global measurements necessary for numerical modeling has come from the remote sensing capabilities of meteorological satellites with the aid of suitable surface ground truth observations these are capable of providing the first truly worldwide observations of the air and ocean surface temperature moisture and cloudiness and elements of the heat and hydrologic balance by using the numerical models diagnostically there is then the prospect of deducing the accompanying global distributions of other variables such as the wind velocity such a scheme is the observational basis for the proposed first garb global experiment FGGE in 1978 the physical and theoretical basis for numerical modeling has grown significantly with the understanding of the theory of baroclinic instability the parameterization of moist convection and advances in our knowledge of the behavior of the stratosphere and the planetary boundary layer our growing understanding of these processes has increased the prospects for improved weather forecasts these hopes are bounded however by the realization that the atmosphere possesses limited predictability that is a time range beyond which the local variations of weather appear as random fluctuations as far as their explicit prediction by numerical models is concerned present indications are that this limit lies at about two weeks time the key physical processes that control the longer period variations of the atmosphere those that are properly associated with climate are largely unknown although we are beginning to recognize the importance of a number of feedback relationships such as the air-sea coupling and cloudiness temperature feedback numerical models that incorporate such effects are our best tool to develop a quantitative understanding of their role in climate and climatic variation the computational base for numerical modeling has grown during the last 20 years in parallel with the development of successive integrations of high-speed computers as shown in figure B1 this overview makes clear the interrelated development of numerical models, theory and computer speed numerical weather prediction may be considered to have begun with the first successful numerical integration of the vorticity equation Charney et al. 1950 with the demonstration of the ability of baroclinic models to forecast cyclonic development Charney and Phillips 1953 or with the commencement of operational numerical weather prediction in 1955 numerical general circulation studies may be considered to have begun with the simulation of the atmospheric energy cycle in an idealized model with sources and sinks of energy and momentum Phillips 1956 with the first successful hemispheric circulation experiments Smagorinsky 1963 or with the first extended global integration Mintz 1965 numerical climate models for the atmosphere may be considered to have begun with the global simulation of the seasonal and inter-annual variation of the primary climatic elements Mintz et al. 1972 although the modeling of climate by other methods has a much longer history the numerical modeling of climatic variation on the other hand which addresses the coupled ocean atmosphere climatic system has only just begun Brian et al. 1974 Manabi et al. 1974a 1974b the development during the past decade of numerical methods whose stability and accuracy can be suitably controlled has made it possible to carry out such calculations for extended periods of time even with today's fastest computers however the solution of the more detailed global numerical models proceeds only at a rate between 1 and 2 orders of magnitude faster than nature itself and our ability to perform the large number of numerical integrations required for the systematic exploration of climate and climatic change requires the continued development and dedication of new computer resources a similar pattern of development has occurred in the numerical modeling of the oceans except that the rate of progress has been slower due principally to a lack of suitable oceanic observations the database for the oceans is fragmentary in comparison with that for the atmosphere and there is no oceanic counterpart of the radio zoned or weather station network the bathy thermograph has been widely used to measure the thermal structure of the oceans surface layer for the past few decades but even this has not been done on a synoptic basis the bulk of the data for oceanic temperature, salinity and currents has been obtained in the course of occasional oceanographic expeditions or special observational programs even so the number of direct velocity measurements is quite small and our knowledge of the oceanic circulation is largely based on geostrophic estimates from conventional hydrographic observations our knowledge of the dynamics of the ocean circulation is also less complete than is that for the atmosphere while the character of the vorticity balance of the ocean was established by Swerdrup 1947 and Stommel 1948 the role of the thermal haline circulation was demonstrated with a numerical model only a few years ago, Brian and Cox 1968 and the effects of bottom topography have been established even more recently, see for example Holland and Hirschman 1972 numerical models are proving of great value in the study of time dependent behavior of the oceanic general circulation and in the analysis of oceanic mesoscale motions such as those now being revealed by the mode observations the structure of these eddies and the role that they play in the oceanic heat balance is one of the principle unsolved problems in physical oceanography other important questions concern the nature of vertical mixing in the ocean especially in the surface layer and the mechanics of the formation of deep and bottom water each of these can perhaps be most fruitfully studied with appropriate regional numerical models in order to lay the foundation for their parameterization in three dimensional models of the world ocean but perhaps the most important problem of all from the viewpoint of climate is the interaction between the ocean and the atmosphere the numerical modeling of this coupled system offers our best hope of achieving a quantitative understanding of the dynamics of climatic variation numerical models thus lie at the heart of the modern study of climate and climatic change they complement and may even be regarded as a part of the observing system they serve as tools for climatic analysis and diagnosis and they offer the most rational way of assessing the course of future climatic events whether or not climatic forecasting in the time dependent sense ever becomes feasible the use of numerical models to simulate the average or equilibrium climates of the past and the likely climatic consequences of various natural or anthropogenic effects in the future will justify their development atmospheric general circulation models formulation all general circulation models are based on the fundamental dynamical equations that govern the large scale behavior of the atmosphere this system consists of the equation of motion expressing the conservation of momentum the thermodynamic energy equation expressing the conservation of heat energy the equations of mass and water vapor continuity and the equation of state when geometric height z is the vertical coordinate these equations can be written in vector form as follows one partial derivative of vector v with respect to t plus vector v times del vector v plus w times partial derivative of vector v with respect to z plus two times vector omega cross vector v plus one over rho del p equals vector f two partial derivative of p with respect to z plus rho times g equals zero three partial derivative of theta with respect to t vector v times del theta plus w times partial derivative of theta with respect to z equals q four partial derivative of rho with respect to t plus del dot rho times vector v plus partial derivative with respect to z of rho w equals zero five partial derivative of q with respect to t plus vector v dot del q plus w times partial derivative of q with respect to z equals s six p equals rho r capital t here vector v is the horizontal velocity w is the vertical velocity vector omega is the rotation vector of the earth rho is the density p is the pressure g is the gravitational acceleration theta is potential temperature which is related to the ordinary temperature t by the relation theta is equal capital t of p zero divided by p to the power of kappa where p zero equals 1000 millibar and kappa equals 0.286 is the ratio of the specific heats q is the water vapor mixing ratio r is the gas constant for moist air and del is the horizontal gradient operator the terms vector f, q and s on the right hand sides of equations one, three and five represent the sources and sinks of momentum heat and water vapor due to a variety of physical processes in the atmosphere and must be either prescribed or parameterized in terms of the primary dependent variables in order to close the system one to six the net frictional force vector f consists of the frictional drag at the earth's surface and the internal friction in the free atmosphere as well as the changes of large scale momentum due to smaller scale processes the net diabetic heating rate q consists of the latent heat released during condensation the heating due to the exchange of both long wave and short wave radiation and the sensible heating of the atmosphere by turbulent heat fluxes from the underlying surface the net moisture addition rate s consists of the difference between the evaporation rate from both the surface and from cloud and precipitation and the condensation rate an important contribution to each of these source terms is the vertical flux of momentum, heat and moisture which accompanies cumulus scale convection in the atmosphere we may note that such convective scale processes are not governed by the system one to six and must be represented in terms of the larger scale variables this parameterization is particularly critical for the net heating because most of the latent heating in the atmosphere is accomplished by convective motions which are also responsible for much of the cloudiness see figure 3.2 the various atmospheric gcms are each formulated in slightly different ways and employ different treatments of the source terms there is at present insufficient evidence to decide which particular formulation is the most satisfactory and there is even more uncertainty regarding the most correct parameterization of the unresolved physical processes contained within vector f, q and s a summary of some of the features of the better known atmospheric general circulation models is given in table b1 each of the models shown here uses generally similar procedures to determine the ground surface temperature from an assumed heat balance over land and ice the surface hydrology with runoff permitted after saturation of the surface soil and the occurrence of convection from vertical stability criteria depending on the moist static energy each of the models also incorporates the observed large scale distributions of terrain height, surface albedo and sea surface temperature solution methods all the atmospheric gcms considered here employ finite different methods of second order accuracy with the dependent variables generally determined on a spatially staggered grid resolution of several hundred kilometers sea table b1 time differencing is also generally of second order accuracy with time steps between 5 and 10 minutes used to maintain linear computational stability long term non-linear computational stability is inherent in some of the models space differencing schemes while others employ eddy diffusion processes to achieve this end various degrees of smoothing are also employed in the model's solution in addition to that inherent in the finite difference approximations themselves depending on the computer used the number of model levels and the frequency with which the radiative heating calculations are performed global atmospheric gcms generally run between 10 and 100 times faster than real time selected climatic simulations in order to display the level of accuracy characteristic of present day atmospheric gcms in the simulation of climate we have here assembled the results of model integrations drawn from recently published and in some cases as yet unpublished sources to facilitate comparison these are presented in a common format along with the corresponding distributions sea level pressure although the various gcms differ greatly in their resolution of the vertical structure of the atmosphere each simulates the distribution of a number of climatic variables at the earth's surface of these perhaps the distribution of sea level pressure is the most familiar it is shown here as simulated by four different models for the month of January in figure b2 the average sea level pressure simulated by the 11 level gfdl atmospheric model is shown for the months of December, January and February Manabe at al 1974 b figures b3, b4 and b5 show the corresponding average January sea level pressure simulated by the 6 level n car model Kasahara and Washington 1971 by the 2 level rand model Gates 1972 and by the 9 level Gis model Somerville et al 1974 in each case the observed average January sea level pressure distribution is also shown while the models results differ in a number of details these results generally show of accuracy as might be anticipated the largest errors and the greatest differences among the models occur in the middle and higher latitudes of the northern hemisphere where cyclonic activity is the most frequent it should be recalled however that sea level pressure alone is by no means a complete indicator of climate tropospheric temperature and pressure in figure b6 the average January 800 millibar temperature simulated by the 2 level rand model Gates 1972 is shown along with the observed distribution although systematic errors may be noted over the continents the simulated large scale temperature distribution clearly reflects the positions of the major thermal perturbations in the lower troposphere the average January 500 millibar white simulated by the 9 level gis model summerville et al. 1974 is shown in figure b7 along with the observed distribution these results also clearly show that the mean position and intensity of the long waves in the westerlies are portrayed reasonably well in the simulation cloudiness and precipitation among the more difficult climatic elements to simulate accurately in a GCM are the cloudiness and precipitation this is doubtless due to the fact that a substantial portion of the total cloudiness and precipitation observed occurs in connection with connective scale motions especially in the lower latitudes as noted earlier these processes must be parameterized in the GCMs and their accurate calibration is relatively difficult in figure b8 the average January total cloudiness simulated by the 6 level NCAR model Kazahara and Washington 1971 is shown along with a composite observed distribution for January and for December January and February with the exception of the equatorial region and the low latitudes of the northern hemisphere the large scale areas of maximum and minimum cloudiness are reasonably well simulated in figure b9 the annual average precipitation simulated by the 11 level GFDL model Manabe et al. 1974 b is shown along with the corresponding observed distribution in addition to the large scale precipitation pattern in middle latitudes this simulation also portrays the number of the smaller scale features including the zone of heavy precipitation near the equator although this comparison is for a somewhat longer time period than the others shown here the difficulty of correctly parameterizing the precipitation process makes the skill of this simulation impressive oceanic and coupled atmosphere ocean general circulation models estimates based on observed data show that the heat transported by ocean currents plays a major role in the global heat balance von der Haar and Ortt 1973 a model that is to be useful for the study of climatic variation must therefore include the ocean as well as the atmosphere as suggested by the simulations just reviewed the specification of a fixed ocean surface temperature in atmospheric GCMs is a strong boundary condition and may mask weakness in the model's simulation of the heat balance the problem of climatic variation therefore furnishes a major motivation for the accelerated development of numerical models of the oceanic general circulation relative to numerical models of the atmosphere numerical modeling of the ocean is still in a primitive state as previously noted this is primarily due to the lack of sufficient data to perform a careful verification of the models and to parameterize properly the effects of the smaller scale motions the only large body of data presently available for verifying ocean circulation models is the collection of measurements of density structure while these data were sufficient to calibrate the earlier analytic theories of the ocean thermocline numerical models require a much more extensive database for adequate verification it is now recognized that many of the earlier studies such as those by Brian and Cox 1967 and Haney 1974 for idealized basins as well as the higher resolution simulations of Cox 1970 for the Indian ocean and of Friedrich 1970 for the North Atlantic represents transient rather than equilibrium solutions for the boundary conditions imposed the extended integration of even more detailed numerical models will be necessary in the future in order to design and calibrate adequately other simpler models such models will require less calculation and thereby allow more freedom to carry out the large number of numerical experiments required general reviews of numerical modeling of the ocean circulation are given in the proceedings of a recent symposium ocean affairs board 1974 and by Gilbert 1974 formulation the principal dynamical components of an oceanic general circulation model are similar to those of its atmospheric counterpart namely the equations of motion, conservation equations for potential temperature and salinity the continuity equation and an equation of state in addition an oceanic model should contain equations for the growth and movement of pack ice in some problems of oceanic circulation it is not necessary to treat the temperature and salinity separately and these variables can be combined into a single density variable in climatic studies however we are interested in the heat transported by ocean currents explicitly and in many regions of the world ocean particularly the polar seas the density and temperature are not proportional in these regions at least it is therefore necessary to predict salinity as a separate independent variable a changing salinity structure in the ocean may provide the basis of climatic change mechanisms that have not yet received sufficient attention in an ocean model the equation of motion 1 may be simplified by treating the density rho as a constant rho 0 businesque approximation while the hydrostatic equation 2 remains unchanged the thermodynamic energy equation 3 and the water vapor continuity equation 5 are represented in the ocean by conservation equations for potential temperature theta and salinity s of the form 7 partial derivative with respect to t of theta s plus vector v times del theta s plus w times partial derivative with respect to z of theta s equals q sigma where q and sigma denote source functions the continuity equation 4 may be simplified by considering the ocean to be incompressible in which case we may write 8 partial derivative of w with respect to z plus del times vector v equals 0 the oceanic equation of state may be written symbolically as 9 rho equals rho of theta s p where the actual expression is a polynomial of high order whose coefficients have been determined by laboratory experiments to close the system expressions must be chosen for vector f in the simplified form of equation 1 and for q and sigma in terms of the dependent variables as in the case of atmospheric models disclosure is an important problem in the formulation of oceanic models and includes the parameterization of the mesoscale oceanic eddies solution methods the predictive equations for momentum temperature and salinity given in the previous section are generally approximated by centered differences of second order accuracy with care taken to conserve both linear and quadratic quantities the numerical methods that have been used successfully for large scale models of the atmosphere are usually further modified by the exclusion of external gravity waves from the system this permits the use of a time step 50 to 100 times larger than is possible for the atmosphere this is accomplished by requiring the total vertically integrated flow to be divergence free in which case it is possible to specify the total transport by a stream function the numerical time integration of an oceanic GCM formulated in this manner proceeds by a combination marching and jury process involving the explicit prediction of theta, s and vector v solution for the total transport stream function Thakanon, 1974 has recently introduced the implicit treatment of rosby waves which allows a considerably longer time step with little loss in accuracy for problems in which the emphasis is on low frequency oceanic phenomena selected climatic simulations to illustrate the characteristic climatic performance of global oceanic GCMs we here present comparative solutions from the recent models of Thakanon et al, 1974 Cox, 1974 and Alexander, 1974 a number of characteristics of these models are given in Table B2 these models are currently undergoing further development and similar oceanic models are under construction at Ncar Gis it is a general characteristic of all such oceanic models that the circulation is dominated by the large values of viscosity and further efforts are required to extend the solutions into the less viscous and more non-linear range surface current the annual surface current simulated by the 9 level GFTL model Cox, 1974 is shown in figure B10 along with the observed currents for February and March the February surface currents simulated by the 5 level UCLA model Thakanon et al, 1974 and the March 1 surface currents simulated by the 2 level RAND model Alexander, 1974 are similarly shown in figures B11 and B12 in each case the overall pattern of the large scale circulation is simulated successfully although in general the strength of the equatorial and major western boundary currents is underpredicted we may note however that the UCLA model's solution represents a 30 year integration the GFTL solution is for 2.5 years and the RAND solution is for 1.5 years closer examination reveals that the simulated surface currents diverge from the equator somewhat more than do-dose observed due to the model's effective averaging over the depth of the surface Ekman layer C surface temperature the February C surface temperature simulated by the 5 level UCLA model Thakanon et al, 1974 is shown in figure B13 along with the observed distribution to some extent the agreement of the simulation with observation is due to the use of observed components in the surface heat balance condition the prediction of low surface temperatures at the equator however is a feature entirely due to the model's internal dynamics coupled ocean atmosphere models as has been previously noted a dynamical model adequate for the study of climatic variation should include the coupling of the ocean and atmosphere the first attempt at such coupling was made by Manabe and Brian, 1969 for an idealized ocean basin and later extended by Weatherold and Manabe 1972 in such a joint model the net fluxes of heat moisture and momentum at the air-sea interface are determined by the atmospheric model while the ocean model in turn provides the sea surface temperature as a lower boundary condition for the atmosphere these studies at GFDL have recently been extended to the entire world ocean and the results of a coupled numerical integration are now available Manabe et al, 1974 A, Brian et al, 1974 in this study the 9 level GFDL atmospheric model was integrated for 0.85 of a year simulated time while a 12 layer ocean model was integrated for 256 years time the annual sea surface temperatures simulated in this joint model are shown in figure B14 along with the observed distribution the general level of accuracy may be considered satisfactory, especially in view of the absence of any specification of observed quantities at the air-sea interface much further development and testing of such coupled models is required so that their potential for the study of global climatic variations may be realized End of section 11 End of understanding climatic change a program for action by the US Committee for the Global Atmospheric Research Program