Page 39

4
Mechanisms and Predictability

We have seen that climatic attributes directly affect human activities and viability, and that these attributes are directly connected to the major processes acting in the Earth's climate system. To exploit beneficial climate variations and mitigate the effects of the harmful variations, we must first understand the operation of the system and then project what could happen in the future.

By now it is well known that some features of our climate can be forecast beyond the two-week limit of atmospheric deterministic predictability. This chapter details the nature of predictability on dec-cen (and shorter) time scales, and endeavors to distinguish between the predictable aspects of natural variability and the predictable consequences of possible anthropogenic changes. It examines how predictability is related to the mechanisms of climate variability and assesses the possibility of predicting climate on dec-cen time scales, especially the attributes discussed in Chapter 2. Last, it speculates on the applications of such predictions should they be attainable.

The Nature of Climate Prediction

Weather Prediction

The state of the atmosphere and ocean is governed by what is generally agreed to be a set of deterministic equations: If the initial state is exactly known, the future evolution of the system is determined for all time. On the other hand, the atmosphere and ocean are rife with instabilities and nonlinearities that imply that the climate system is chaotic: Two sets of initial conditions that appear to be very close to one another (but not identical) will evolve along trajectories that inevitably diverge. Their divergence cannot continue forever, though, given the bounds imposed by the system' s finite energy. As a result, trajectories continuously approach, as well as diverge from, each other, generating the system' s chaotic, albeit deterministic, behavior.

A numerical weather-forecast model also has deterministic solutions, but any small error in the initial state guarantees that the resulting forecast trajectory will diverge from a trajectory that began with the "correct" initial state. The inevitable growth of the initially specified error, and the subsequent mixing of ever-changing trajectories, limits predictability. The pioneering work of Lorenz (1963, 1969), and the research it spurred over the years, established that the doubling rate of errors for large-scale atmospheric flows is on the order of 2-3 days (Lorenz, 1982), so that the global atmosphere is predictable only on scales of two weeks or so, given the currently achievable accuracy of initial-state determination. Similar studies for the detailed predictability of the oceans are in their infancy, but the slower growth and saturation times of oceanic instabilities suggest potentially longer predictability times, on the order of months rather than weeks. Still, all these times are much shorter than the dec-cen time scale of interest here.

Climate Prediction

The obvious question is, if the ultimate limit of detailed prediction for atmosphere and ocean weather is on the order of weeks to months at best, how could we possibly expect to predict climate on time scales of years or decades? The answer follows directly from the definition of climate: Climate is the statistics of the atmosphere (and other components of the climate system). We have known for a long time that atmospheric statistics are determined entirely by the boundary conditions of the atmosphere; every atmospheric modeler uses this paradigm. The boundary conditions for an atmospheric model (in particular SST and land and sea ice) are specified, and the atmospheric circulation and hydrologic cycle are allowed to come into equilibrium with these specifications. Other boundary quantities are then determined by internally coupling the atmospheric model to a land model, in particular land-surface moisture and vegetation, and land snow and ice cover. Whether or not climate is uniquely determined by the boundary conditions is still undetermined, but for the purposes of argument, we will assume here that it is.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 39
Page 39 4 Mechanisms and Predictability We have seen that climatic attributes directly affect human activities and viability, and that these attributes are directly connected to the major processes acting in the Earth's climate system. To exploit beneficial climate variations and mitigate the effects of the harmful variations, we must first understand the operation of the system and then project what could happen in the future. By now it is well known that some features of our climate can be forecast beyond the two-week limit of atmospheric deterministic predictability. This chapter details the nature of predictability on dec-cen (and shorter) time scales, and endeavors to distinguish between the predictable aspects of natural variability and the predictable consequences of possible anthropogenic changes. It examines how predictability is related to the mechanisms of climate variability and assesses the possibility of predicting climate on dec-cen time scales, especially the attributes discussed in Chapter 2. Last, it speculates on the applications of such predictions should they be attainable. The Nature of Climate Prediction Weather Prediction The state of the atmosphere and ocean is governed by what is generally agreed to be a set of deterministic equations: If the initial state is exactly known, the future evolution of the system is determined for all time. On the other hand, the atmosphere and ocean are rife with instabilities and nonlinearities that imply that the climate system is chaotic: Two sets of initial conditions that appear to be very close to one another (but not identical) will evolve along trajectories that inevitably diverge. Their divergence cannot continue forever, though, given the bounds imposed by the system' s finite energy. As a result, trajectories continuously approach, as well as diverge from, each other, generating the system' s chaotic, albeit deterministic, behavior. A numerical weather-forecast model also has deterministic solutions, but any small error in the initial state guarantees that the resulting forecast trajectory will diverge from a trajectory that began with the "correct" initial state. The inevitable growth of the initially specified error, and the subsequent mixing of ever-changing trajectories, limits predictability. The pioneering work of Lorenz (1963, 1969), and the research it spurred over the years, established that the doubling rate of errors for large-scale atmospheric flows is on the order of 2-3 days (Lorenz, 1982), so that the global atmosphere is predictable only on scales of two weeks or so, given the currently achievable accuracy of initial-state determination. Similar studies for the detailed predictability of the oceans are in their infancy, but the slower growth and saturation times of oceanic instabilities suggest potentially longer predictability times, on the order of months rather than weeks. Still, all these times are much shorter than the dec-cen time scale of interest here. Climate Prediction The obvious question is, if the ultimate limit of detailed prediction for atmosphere and ocean weather is on the order of weeks to months at best, how could we possibly expect to predict climate on time scales of years or decades? The answer follows directly from the definition of climate: Climate is the statistics of the atmosphere (and other components of the climate system). We have known for a long time that atmospheric statistics are determined entirely by the boundary conditions of the atmosphere; every atmospheric modeler uses this paradigm. The boundary conditions for an atmospheric model (in particular SST and land and sea ice) are specified, and the atmospheric circulation and hydrologic cycle are allowed to come into equilibrium with these specifications. Other boundary quantities are then determined by internally coupling the atmospheric model to a land model, in particular land-surface moisture and vegetation, and land snow and ice cover. Whether or not climate is uniquely determined by the boundary conditions is still undetermined, but for the purposes of argument, we will assume here that it is.

OCR for page 39
Page 40 If, therefore, we can predict the boundary conditions of the atmosphere at a specific time, particularly SST and sea ice, we will have some information about the statistics of the atmosphere at that time. We will not be able to predict the precise state of the atmosphere, because it can vary in equilibrium with the predicted boundary conditions, but we will know something about its average conditions. We may be able to predict the average monthly or seasonal precipitation over a region even if we cannot say on what specific day the precipitation will fall. Knowing only a mean value at a given time can still be helpful if the associated variability is small. Predictions of tropical boundary conditions at a certain time are likely to be useful because tropical climate variability is low, while predictions of mid-latitude boundary conditions would be less useful because mid-latitude variability, especially during winter, is high. Two more important points need to be made. First is the distinction between initialized and uninitialized prediction. To make a prediction about a specific time in the future, say the summer of 2009, there must be some connection to the actual conditions now. We call this estimation of the actual beginning state of the system "initialization" (while recognizing that this term is sometimes used elsewhere to mean the act of bringing a model system to a state of equilibrium without estimating its current conditions). If we do not make this initial estimation, we will not be able to forecast the time at which the climate will assume a given state, though we may still draw conclusions about its statistics (that is, changes of its mean and the nature of its variability). The difference between initialized and uninitialized prediction becomes important in discussing greenhouse-warming predictions versus ENSO predictions. The second important point is the potential for making empirically or statistically based (analog) predictions. Sufficient information is available from past climate records to allow predictions to be made (with specified uncertainty) whenever specific climate states exist that have in the past been accompanied or followed by particular regional or local climate conditions. "Climate state" is defined here, as in NRC (1975), as the average of the complete set of atmospheric, hydrospheric, and cryospheric variables over a specified period of time in a specified domain of the earth-atmosphere system. For climate prediction on all time scales, whether initialized or not, the tool for predicting the boundary conditions of SST and sea ice is the coupled climate model—a model that consistently links the atmosphere, ocean, and ice together in responding to a specified external forcing. Short-, Medium-, and Long-Range Climate Prediction There is no accepted terminology describing the various time scales for prediction. This report will use "short-range climate prediction" to denote prediction on time scales up to interannual, "medium-range climate prediction" to mean prediction at decadal time scales, and "long-range climate prediction'' (sometimes called "greenhouse prediction'') for prediction on centennial time scales—the scale of a human lifetime. Short-range climate prediction is an established enterprise: Skill has been demonstrated for predicting the SST changes in the tropical Pacific that are characteristic of the ENSO phenomenon on lead times of 6 to 12 months. Atmospheric properties elsewhere may then be inferred from these forecasts. These predictive skills, which vary as a function of several factors (including season, model type, and decade), have been well documented (Battisti and Sarachik, 1995; Glantz, 1996; Latif et al., 1998). ENSO prediction is initialized prediction (in the sense defined above), so a real-time observing system in the tropical Pacific was put in place by the TOGA research program. It has been kept in place even though TOGA has ended, which should permit us to develop our skill further. Long-range climate prediction has so far been limited to predicting forced climate change in response to the anthropogenic addition of radiatively active gases and aerosols to the atmosphere. Because this type of prediction is essentially uninitialized, it cannot predict the actual state of the boundary conditions at some specific future time. It can, however, be used to derive the statistics of the boundary conditions (and therefore the statistics of the atmosphere in equilibrium with the statistics of the boundary conditions) at some future time. Thus, initialized short-range climate prediction can predict the SST in the tropical Pacific for January of 1999, say, while greenhouse predictions can only say that annually averaged SST will be warmer in the year 2050 by some specified amount, or within a certain range. Such greenhouse predictions are still valuable if the forcing changes the mean boundary conditions enough for a difference beyond natural variability to be apparent; again, small shifts of the mean may be noticeable in the tropics where the variability is low, while larger shifts may be masked in mid-latitudes where variability is high. Long-range forecasts permit the assessment of shifts in average precipitation, or length of the growing season, or changes in patterns of runoff; as indicated by Karl et al. (1996), even subtle shifts in the mean state can have considerable implications for the frequency and magnitude of extreme climate events. Medium-range climate prediction, prediction on time scales of a decade or so, is the most problematic type of prediction. Its value as uninitialized prediction is limited: The year-to-year variability of climate, together with the relatively slow approach of the climate system to equilibrium with anthropogenically added radiatively active atmospheric constituents, limit the value of prediction of the statistics of boundary conditions a decade in advance. Even this type of prediction may be useful under certain circumstances, however. When regional changes are fast and crossing the threshold of a new climate state can be predicted,

OCR for page 39
Page 41 preparations can be made for change even if the exact time of occurrence is not known. The possibility of initialized medium-range climate prediction is a real and intriguing one. As the fully coupled system is allowed to evolve freely over the course of a decade from its initial state, water parcels from deeper parts of the ocean reach the surface and imprint themselves on the SST field. The question, of course, is whether or not the evolving imprint of the initial ocean state on the SST can survive both the mixing in the ocean (as parcels wend their way to the surface) and the inevitable noise from high-frequency atmospheric forcing. Prediction and Mechanisms The existence and nature of climatic predictability depends on the nature of the mechanisms responsible for the variability. We can distinguish two basic types of mechanisms for decadal variability: variability forced by processes external to the climate system, and variability generated by processes internal to the climate system. Externally Forced Atmospheric Variability In the external-forcing category are forcings by varying solar output; by addition of aerosols due to major volcanic eruptions, biomass burning, and industrial sources; and by the addition of radiatively active gases to the atmosphere. Their effects are discussed in detail in Chapter 5, "Atmospheric Composition and Radiative Forcing." It is generally true that variability generated by external forcing is unpredictable when the forcing itself is unpredictable; the decadal variations of solar radiative output and the geodynamics of future eruptions of volcanoes are both poorly understood. When volcanic aerosols are added to the stratosphere, there is a period of a year or two in which the aerosols stay in the stratosphere, so their radiative and chemical effects can be predicted over the following year (see, e.g., Fiocco et al., 1996). The addition of radiatively active gases to the atmosphere produces mean warming of the Earth, regional changes of mean temperature and precipitation, and possible changes to natural climate cycles, such as ENSO and the Pacific North American pattern, but prediction of these mean changes relies on our ability to know the past and future emissions of these gases. Those radiatively active gases that also contain chlorine or oxides of nitrogen affect the stratospheric burden of ozone, so to the extent that the concentration of these gases is known and the chemistry of ozone is understood, the concentration of ozone may be predicted. Internally Forced Atmospheric Variability We may classify the internal decadal-variability mechanisms into three distinct categories: those arising from high-frequency forcing of the slow components of the climate system by the more rapidly varying atmosphere; those arising from slow internal variations in the ocean, atmosphere, cryosphere, or biosphere; and those arising from the coupling of components of the climate system that individually would not have such an effect (Sarachik et al., 1996). Specific mechanisms associated with each of the components of the climate system will be presented in Chapter 5. Chapter 3 described a variety of patterns of variability, and some of their co-varying aspects. In this section we discuss the extent to which these patterns, or other broad-scale features of the atmosphere, appear to be coupled to other parts of the climate system, predominantly the ocean. These links are important to predictability; when a fast and a slow component are coupled, and the latter has mechanistic control of the pattern, the longer time scales of the slower component can be capitalized on to make more distant forecasts or predictions of the faster component (in this case the atmosphere) than would otherwise be possible. The Hasselmann (1976) theory of climate variability is a convenient starting point for understanding the relevance of different climatic time scales to climate prediction. Hasselmann's theory asserts that the atmosphere produces, through instabilities of various types, high-frequency variability that presents itself as weather. At these frequencies, the variability may be considered random. When a slower-reacting reservoir, such as the ocean, is forced by such high-frequency variability, the high-frequency variability is damped in the slower component. This basic climate mechanism of Hasselmann seems to account for a great deal of observed variability (see, e.g., Frankignoul and Hasselmann, 1977) and of modeled natural variability in long, coupled climate simulations (Manabe and Stouffer, 1996). If the Hasselmann mechanism were the only operative mechanism—say, the atmosphere acting on the ocean—the temporal extent of the predictability would be the autocorrelation time of the sea surface temperature. In this case, the best forecast of SST would be a forecast of persistence (i.e., no change) that fades to the norm with a time constant consistent with the autocorrelation function. Over many parts of the world ocean, persistence is on the order of months. If, however, the SST generated by the Hasselmann mechanism feeds back to the atmosphere in a coherent way, coupled modes may result, increasing predictability notably. Similar couplings may exist with other parts of the climate system, such as the perennial ice or snow fields, though this possibility has not been explored to any great extent. Extension of these ideas to the stochastically forced coupled atmosphere-ocean system suggests that such coupling will act to significantly enhance the very-low-frequency variance in the atmosphere (see, e.g., Barsugli and Battisti, 1998). This is the simplest theory accounting for the presence of very-low-frequency variability in the climate system, and is usually assumed to be correct in the absence of other information. The Hasselmann mechanism would

OCR for page 39
Page 42 also determine the maximum level of predictability of climate anomalies, if the coupling between the atmosphere and the ocean (or biosphere or cryosphere) did not support patterns of its own. However, there are several climate phenomena on the decadal time scale that are most likely the result of processes other than those inherent in the Hasselmann mechanism. The best documented of these coupled phenomena are discussed below, and the possible mechanisms are summarized. The essential reason for studying the mechanisms of decadal variability (aside from their intrinsic scientific interest) is that determining which mechanisms are operative will determine the extent to which climate can be predicted. Some mechanisms (e.g., external forcing by volcanic eruptions) we will assume to have no predictability, and thus to offer no improvement in predicting climate. Some mechanisms have moderate predictability (e.g., random forcing of SST by the atmosphere), and some have significant predictability (e.g., SST variations in tropical Pacific caused by coupled atmosphere-ocean modes). These are well worth exploring. Coupled Modes Chapter 3 described a large-scale decadal mode of variability in the Pacific that has ENSO-like SST characteristics in the tropical regions and strong out-of-phase covariation of SST in the North Pacific. The mechanisms responsible for this variability are not yet known. Various investigators have postulated various reasons for it, including: • inherent nonlinearities in the physics of ENSO (e.g., Munnich et al., 1991); • interaction between ENSO and the seasonal cycle (e.g., Jin et al., 1994, 1996); • interaction between ENSO and other unstable coupled atmosphere-ocean modes (e.g., Mantua and Battisti, 1994); • stochastic forcing of a linearly stable coupled system (e.g., Penland and Sardeshmukh, 1995); and • low-frequency changes in the shallow equatorial thermohaline circulation that may lead to changes in the amplitude and frequency of the interannual ENSO variability (e.g., Pedlosky, 1987; McCreary and Lu, 1994; Gu and Philander, 1997). Whatever the cause, a significant portion of the low-frequency variability in the global climate system is ENSO-like in structure, as can be seen in Figure 3-7. Decadal-scale changes in the state of the tropical Pacific atmosphere-ocean system might also affect the predictability of the higher-frequency ENSO variability. For both of these reasons, the Pacific is an important focus for understanding dec-cen variability in the climate system. In addition to the ENSO-like coupled phenomenon, there is ample evidence of variability in the mid-latitude North Pacific atmosphere-ocean system on interannual time scales. Specifically, wintertime variability is largely forced locally by the atmosphere; about half of the variance in interannual SST anomalies is forced by ENSO, and communicated to the North Pacific Ocean via the atmospheric teleconnections. The variability in the North Pacific climate on decadal and longer time scales is not yet fully documented. It is known that, over the last half-century, a substantial portion of the variability in the atmosphere-ocean system on these time scales is associated with the global ENSO-like structure displayed in Figure 3-7. Recently, Latif and Barnett (1996) have suggested a mechanism by which coupling between the atmosphere and upper ocean that takes place in the mid-latitude North Pacific basin may give rise to climate variability in the North Pacific and North America on the multidecadal time scale. The spatial structure of the model anomalies in the Latif and Barnett study is remarkably similar to that of the mid-latitude anomalies in the global ENSO-like structure of the observations displayed in Figure 3-7, which is also primarily a low-frequency pattern of variability. However, the observed anomalies clearly involve the tropical atmosphere-ocean system. Hence, if the mechanism suggested by Latif and Barnett does indeed operate in nature, it will be necessary to sort out how much of the variance in the mid-latitudes comes from forcing in the tropical Pacific that is teleconnected to the mid-latitudes, and how much from atmosphere-ocean interactions that are specific to the North Pacific. Delworth et al. (1993) found multidecadal variability in the North Atlantic of the coupled atmosphere-ocean GCM of Manabe and Stouffer (1988). The pattern of the SST anomalies associated with this variability is somewhat similar to that found in the observations by Kushnir (1994) and in simpler coupled models (Chen and Ghil, 1996). However, the mechanism associated with the variability in both types of atmosphere-ocean models seems to be internal to the ocean thermohaline circulation, and not inherently a coupled atmosphere-ocean phenomenon. Sarachik et al. (1996) review the current theories of mechanisms for producing dec-cen climate variability from an ocean and ocean-atmosphere modeling perspective. These theories include: stochastic forcing of the ocean by white-noise, synoptic-scale variability of the atmosphere (e.g., Hasselmann, 1976; Mikolajewicz and Maier-Reimer, 1990), internal ocean variability (e.g., Delworth et al., 1993; Chen and Ghil, 1995), coupled ocean-atmosphere modes (Hirst, 1986; Latif and Barnett, 1994), and ENSO variability (Trenberth and Hurrell, 1994; Wallace et al., 1995). Yin and Sarachik (1994) proposed an oceanic advective and convective mechanism. A completely different type of interannual (in the North Atlantic) to decadal (in the North Pacific) variability has been modeled by Jiang et al. (1995) via the nonlinear dynamics of the double-gyre circulation's strength, heave, and wobble (see also Cessi and Ierley, 1995, and Speich et al., 1995) and by Spall (1996) via the nonlinear

OCR for page 39
Page 43 dynamics of the Gulf Stream and the deep western boundary current. Prospects for Climate Prediction As stated, the nature of the mechanisms causing dec-cen variability will determine whether or not it is possible to make deterministic climate predictions on these time scales. It is clear that the mechanisms for some of the most notable decadal phenomena are unknown at present. If the natural decadal phenomena are forced externally, they cannot be predicted a decade in advance because those forcings are assumed to be unpredictable. If the Hasselmann mechanism is the operative one, then the scale of predictability is limited to the order of the autocorrelation time of the slower component of the system, usually the ocean. If the ocean is the driver of long-term climate-system variability, however, the fact that its circulation is sluggish offers some hope for initialized decadal-scale prediction: Initial experiments with low-resolution coupled models (Griffies and Bryan, 1997) have indicated that errors imposed on the initial (internal) state of the North Atlantic grow slowly enough that SST and sea level may be predictable as much as a decade in advance. Skill in forecasting equatorial Pacific SST, which has been demonstrated out to two years (Chen et al., 1995), may be extended to longer forecast periods, perhaps approaching a decade. The best hope for initialized prediction is offered by the existence of true atmosphere-ocean modes. Depending on the degree of coupling and the internal ocean phenomena that set the time scale for the oscillation, the initialization of those ocean phenomena captures the wherewithal for the coupled evolution in part of its cycle, and guarantees that parts of its evolution will continue into the future. ENSO on shorter time scales is an example of this; the initial state is the thermocline's configuration, which, when specified at the initial time, determines the future evolution of the upper tropical ocean and hence its temperature structure. This kind of prediction also seems to work for the modal structures involved with the Atlantic subtropical dipole (Chang et al., 1998). The possibility of long-range (uninitialized) greenhouse prediction has been demonstrated by many coupled predictions over the years (see IPCC, 1996a, for a complete review). As yet, however, only the grossest measures of warming have been used. Also, it is clear that other climate system components besides temperature must change in response to a significant change in greenhouse gases—for example, the warming may be kept to a minimum by changes in the atmospheric moisture distribution. These changes, such as those in the hydrologic cycle or cloudiness, may actually be more important to society than the temperature effect. Also, the regional usefulness of current predictions is limited both by the uncertainty of the prediction of the mean itself and by the regional variability that masks modest changes in the mean. The general usefulness of uninitialized greenhouse prediction is defined by the magnitude of the forced response relative to the existing variability. These factors, combined with the impossibility of demonstrating the correctness of such a prediction until the extremely long prediction time has passed, make the use of such predictions particularly challenging. The Uses of Climate Prediction This section considers the uses of decadal and longer prediction, both initialized and uninitialized, should the skill of such prediction turn out to be significant. In order to determine whether the climate attributes that affect humankind (discussed in Chapter 2) are likely to be predictable, we must first demonstrate that some related physical quantity (e.g., SST) is predictable. While we do not know enough yet to do this properly, we can indicate what is known and what still needs to be known in order to make predictions of each of the climate attributes. Uses of Medium-Range Climate Prediction (Initialized) The discussion of decadal-scale prediction here deals with initialized predictions only. There are two ways they can apply: directly (for the region from which the prediction is derived) or remotely (when teleconnections exist between that region and another). Precipitation and Freshwater Availability At this time, there are a few predictable phenomena that clearly affect rainfall over vulnerable populations: The Atlantic subtropical dipole that influences rainfall in Brazil and Africa (Hastenrath and Heller, 1977; Lamb, 1978; Hastenrath, 1990) is one, and the extreme ENSO states influencing western U.S. rainfall events (Cayan and Peterson, 1989) are another. Other phenomena affect rainfall and are probably predictable (e.g., the large-scale variability over the northern Pacific), but so far they have not been shown to be predictable on decadal time scales. The variation of rainfall in the Brazilian Nordeste clearly has both interannual and decadal components. It has been related to the location of the ITCZ over the tropical Atlantic, which in turn has been related to SST variations in both the tropical and subtropical Atlantic and Pacific (most recently in Uvo et al., 1997). The possibility therefore exists that prediction of SST will lead to better estimates of the location of the Atlantic ITCZ and of consequences of future rainfall in Brazil and in the Sahel. SST has come to be predicted better on short (seasonal-to-interannual) time scales in the tropical oceans; the population dislocations in the Nordeste caused by years with low precipitation during

OCR for page 39
Page 44 the growing season have gradually been reduced by actions taken on the basis of predictions made a few months to a year in advance (Moura, 1994). Short-term forecasts for Morocco have proven quite successful as well. These are clear examples of how useful prediction of future climate can be to society. The indication by Chang et al. (1998) that the decadal variation of tropical to subtropical Atlantic SST anomalies may be predicted years in advance offers the possibility of longer-range planning for the allocation of water resources to agriculture, drinking water, and hydroelectric power for northeastern Brazil and northwestern Africa. A prediction of dry conditions for the next five years might shift resource planning from power generation to irrigation, while a prediction of wet conditions might do the opposite. To the extent that precipitation in the Sahel and Morocco could be predicted, resources could be directed to agricultural infrastructure on a forecast of good rainfall, or to food-relief infrastructure on a prediction of poor rainfall. Temperature Temperature is the variable most relevant to the intensity and timing of cryospheric melting. To the extent that rivers are fed by glaciers and/or snowpack, their streamflow is partly determined by temperature. A prediction of increased average temperature would imply earlier melt, and therefore an early peak in streamflow. For those regions requiring summer water, the earlier peak may leave the summer unreasonably dry. Believable decadal (half-decadal) forecasts of temperature and precipitation—and, more important, streamflow, particularly in arid and semi-add regions such as the U.S. West—will be critical to water managers in developing new operating rules that are better adapted to the changing conditions. High temperatures over land also affect vegetation and soil moisture, especially during summer, both directly, by increasing evaporation, and indirectly, by affecting precipitation. A decrease in soil moisture has a feedback effect that increases surface temperature, as does reduced evapo-transpiration by vegetation. Believable early warning of a prolonged period of increased temperature and decreased precipitation, especially in the prime agricultural lands of the Northern Hemisphere, would provide opportunities to avert global food insecurity and attendant disruptions: Irrigation infrastructure could be built, markets could be stabilized by futures hedging, drought-resistant seed could be stocked, and so on. Changes in atmospheric temperature over the oceans eventually work their way into the interior of the ocean, where they affect ocean volume and may lead to noticeable changes in sea level. Although multiple decades generally pass before the impact of temperature change is felt, approximately 50 percent of the sea-level rise in this century may be attributed to the global warming over this time period. Storms Advanced numerical models can simulate and predict the short-term evolution of storms with a remarkable degree of success. However, the science of describing and predicting the statistical properties of storms on time scales of a month or longer is still in its infancy. Of major interest to climatologists is the relationship between extreme weather events (which are associated both with tropical and extra-tropical storms and with climatic factors such as regional SST) and the strength and position of the semi-permanent features of the atmospheric circulation. It has been demonstrated in several observational studies that a close relationship exists between the intensity and distribution of storms and climate variability. Changes in tropical cyclone activity have been observed to be linked to variations in regional SST and SLP (Shapiro, 1982; Emanuel, 1987), as well as to the entire state of the tropical atmosphere associated with ENSO (Gray et al., 1992; Nicholls, 1985). Extratropical cyclonic storms are linked to the state of the "teleconnections" in the monthly mean circulation (Lau, 1988; Rogers, 1990). Similarly, the high-frequency storm events exert an influence on low-frequency climatic fluctuations (Gray, 1979; Lau and Nath, 1991). Existing observations are not adequate to elucidate the climatology of severe storms and their link to climate variability. The limited extent of the instrumental record and inconsistencies in reports of storm activity are major obstacles to a satisfactory resolution of these issues. Attempts have been made to extract useful information on severe weather from GCMs, but the results are at best rudimentary (Broccoli and Manabe, 1990; Haarsma et al., 1993). Once the evolution of large-scale climatic factors can be predicted, it is plausible to envision the emergence of an ability to predict the degree of storminess. In fact, schemes to predict up to a year in advance the characteristics of an upcoming tropical-cyclone season, given atmospheric and ocean variables, have already been demonstrated with some success (Nicholls, 1985; Gray et al., 1992). Ecosystems Temperature changes will alter the stability of the oceans. Changes to the stability-sensitive upwelling of nutrient-rich waters, as well as direct thermal effects (among other things), may modify the patterns of biological productivity in the world's oceans (IPCC, 1996b). Only recently has it been shown that the salmon fisheries of the Pacific Northwest undergo decadal variability in phase with an index of the North Pacific SST (Mantua et al., 1997; see Figure 2-18). Currently Alaska is producing a prodigious amount of salmon, while the Columbia River basin is producing very little. Just knowing this variability exists is useful; much time, effort, and money are spent in reviving salmon fisheries, when in fact their decline might be due to natural factors rather than fishing practices. A prediction of this pattern

OCR for page 39
Page 45 several years in advance would allow resources to be applied more efficiently. Uses of Long-Range Climate Prediction (Uninitialized) This section deals with uninitialized prediction on interdecadal time scales, and in particular the response of climate to the anthropogenic addition of radiatively active constituents (gases and aerosols) to the atmosphere. This topic has been treated in considerable detail by the IPCC which was established to assess available scientific information on climate change, assess the environmental and socioeconomic impacts of climate change, and formulate response strategies. Indeed, the three volumes of the second IPCC (1996a,b,c) assessment devote some two thousand pages to doing just that, so only a few salient points connected with the climate attributes of Chapter 2 are noted here. Precipitation and Freshwater Availability Since the earliest days of coupled modeling and its use to determine the climatic effects of increases in radiatively active gases, an unambiguous and robust result has been that the hydrologic cycle will run faster in a warmer world (Manabe and Wetherald, 1975). What has been in some dispute is where and how the additional rainfall would occur, what the impact of the increased evaporation potential on land and vegetation would be, and how the frequency and intensity of extreme precipitation events would be altered. In most models, the increase in surface irradiance leads to increases in summer evaporation and reduced soil moisture (with subsequent vegetation stress). The greater evaporation also implies higher surface temperatures in continental interiors, which will reduce precipitation in those regions. The excess precipitation occurs over oceans and in high latitudes (see, e.g., IPCC, 1996a) where it is not beneficial to human activity. There are also suggestions that extreme precipitation events would increase (see, e.g., Trenberth, in press). The patterns identified earlier (NAO, PNA, the Atlantic dipole, and ENSO-like decadal variability) all influence rainfall, by controlling the location of the storm tracks, by controlling the location of the ITCZ, or, in the case of ENSO, by controlling rainfall directly through SST expression or remotely through teleconnections. Thus it becomes important to know how these patterns of variability change as radiatively active constituents are added to the atmosphere. Unfortunately, current climate models cannot answer this question unambiguously. Since much of the distribution of precipitation and its subsequent return to the ocean is topographically determined, it becomes necessary to more fully resolve model orography and streamflow. This cannot be done with the current generation of models that predict the response to radiative gases; for practical reasons, they cannot have high resolution if they are to be run for a hundred years. There is increasing evidence (see, e.g., Giorgi and Marinucci, 1996) that embedding high-resolution models within global-prediction models gives better distributions of precipitation, and offers the hope that future predictions of climate response to radiatively active gases will be specific enough for more accurate planning. In addition, statistically-based climate-downscaling methodologies are showing promise (Hewitson and Crane, 1996; Zorita et al., 1995). The implications of having accurate medium-range forecasts of precipitation are immense. They are particularly important for water-resource management and planning, though they are relevant to a great many other aspects of society as well. More regionally accurate precipitation forecasts will be important in planning for the world's food security in the face of rising population. Temperature Until now, the basic question has been how the surface temperature of the globe, and of various regions, would be altered by the addition of radiatively active constituents to the atmosphere. As we have seen, this question must be broadened to include possible changes to the naturally occurring climate patterns, since such changes affect both the mean and variations of climate. Until we know whether natural cycles will change under the addition of radiatively active constituents, we will not be able to predict the regional and global temperature responses. It should also be noted that statements about detection of global warming are usually statements about detecting a shift of the mean in the presence of a naturally varying background. When the mean is partly the result of the phase-locking of various natural cycles, and these cycles themselves may be affected by the mean they help to produce, the question of surface temperature alteration must be deepened and reformulated. Storms Projections of changes in storminess due to anticipated changes in greenhouse-gas concentration remain debatable (IPCC, 1990; Hall et al., 1994; Lighthill et al., 1994). In the words of the IPCC (1996a): "Overall, there is no evidence that extreme weather events, or climate variability, has increased, in a global sense, through the 20th century, although data and analyses are poor and not comprehensive. On regional scales there is clear evidence of changes in some extremes and climate variability indicators. Some of these changes have been toward greater variability; some have been toward lower variability." Further discussion of changes in storms, as they relate to changes in flood frequency, is included in the "Hydrologic Cycle" section of Chapter 5.

OCR for page 39
Page 46 Sea Level Long-term prediction is particularly relevant for changes in sea level. The changes in relative sea and land levels measured at the coast by tide gauges can be usefully divided four ways, into the local and the global changes of the land and the sea level. These distinctions are important for understanding the time scales and causes of ongoing changes, and predicting their future evolution. Land levels are significantly influenced by global-scale tectonic effects—the adjustment of the Earth's mantle to the removal of the glacial-era icesheets, for example. Local changes in land level can result from altered sedimentation rates, or from subsidence due to extraction of groundwater or oil. Sea level is also subject to local changes forced by local winds, river runoff, and the passage of oceanic waves of various frequencies. The global sea level is determined primarily by the mass of water in the ocean and its temperature structure. Particular effort has gone into understanding the global sea-level component of the tide-gauge measurement, because it is expected to change with climate. Tide-gauge records longer than 50 years are needed to eliminate spurious trends due to low-frequency variability. Once records have been corrected for post-glacial rebound, a trend over the past 100 years of about 1.8 mm per year emerges (Douglas, 1991). There is no firm evidence of an increase in the rise, nor would it be expected from the change in climate that has occurred over that period. Archeological and geological studies indicate that the variation of sea level over the previous two millennia was no more than a few tens of centimeters. The time of onset of the current rise is not known. While uncertainty about the measured rise remains, because of the lack of global coverage and the possible influence of coastal subsidence, uncertainties about the components of the rise are far larger. Two factors contribute significantly to the change of global sea level with climate: thermal expansion of the ocean, and redistribution of water between land and sea. Surface thermal anomalies penetrate down into the ocean's interior via the wind and thermohaline-driven overturning. These circulations have a range of time scales from decadal to millennial. Existing direct observations of ocean temperature are insufficient to reveal the past global warming of the ocean, although significant local changes have been observed. Models of ocean circulation have therefore been used to calculate the thermal-expansion part of the observed sea-level rise. These models yield estimates ranging from 0.2 to 0.7 mm per year (IPCC, 1996a). This calculation is inherently uncertain, however, since the boundary conditions—wind stress, surface temperature, and salinity—are not well known. The calculation becomes even more tenuous when made for future climate scenarios with additional greenhouse gases. Ninety-nine percent of the world's land ice is contained in the Greenland and Antarctic ice sheets. The response of these ice sheets to climate change is difficult to predict (see, e.g., Oppenheimer, 1998). Since the mass balance of these ice sheets reflects long time scales, they are likely still adjusting to past climate changes. In general, the increased supply of moisture in a warmer climate is expected to dominate the increased melting for the Antarctic ice sheet, while the reverse is expected for the Greenland ice sheet. Current observations are insufficient to detect a mass imbalance in either. Here, a climate prediction might attempt at least to determine the relative change in the mass balance of the ice sheets, when models can determine temperature and precipitation in the high latitudes. To interpret observations of sea-level and ice-volume changes, they must be placed in the context of the past and compared with projections of the future. It is clearly of interest to know when the current rise began and whether there were past rises of comparable magnitude and duration. Paleo-studies and data "archeology" (recovery of unpublished records) can help address these issues. Most projections of sea-level response to anthropogenic forcing have been based on simple models (e.g., one-dimensional up-welling-diffusion ocean models). Sea level is fully embedded in the climate system, however, and a coupled ocean-atmosphere-ice model must be used to maintain consistency in all the elements. Furthermore, the dynamic response of the ocean to climate change gives rise to regional changes in sea level that may be of a magnitude comparable to that of the global mean change. Continued improvement of these sophisticated models will be necessary if useful projections are to be made. Such projections will prove invaluable, though, because sea-level rise can have such a large and devastating impact on the vastly developed and densely occupied coastal regions of the world. Ecosystems Parameterizations of climate-induced ecosystem changes are rapidly improving. To predict ecosystem changes under scenarios of elevated greenhouse gases, earlier models simply mapped the recently observed biomes to the GCM-predicted locations with similar climatic conditions. Some of the latest models include vegetation interactions with nutrients, CO2 fertilization, and fire (VEMAP, 1995; IPCC, 1998). Recent integrated-assessment models of climate change even include climate-vegetation and carbon cycle-vegetation feedbacks, as well as the effects of changing land use (see, e.g., CIESIN, 1995). While the climate scenarios that have been explored with these models are often derived from transient coupled ocean-atmosphere GCMs, the ecosystem models themselves tend to be designed to simulate an equilibrium land-surface biosphere, rather than the transient ecosystem compositions that will precede the equilibrium state. The veracity of potential (i.e., omitting land-use changes) vegetation-distribution predictions made from uninitialized climate forecasts is as yet unknown. Because

OCR for page 39
Page 47 between one-third and one-half of terrestrial biological production is used or dominated by human action (Vitousek et al., 1997), and human-induced land-use changes are difficult to anticipate, the prediction of future vegetation distributions is a difficult undertaking. However, even in the absence of local-scale ecosystem forecasting skill, there is value in assessing the large-scale response of modeled ecosystems to uninitialized climate-model forecasts. Such models can be used to assess the future, large-scale response of terrestrial carbon sinks to altered climate or to anthropogenic inputs, for instance, or the effects of large-scale, vegetation-related albedo or surface roughness changes on climate. As longer time series of the relevant vegetation data become available for testing ecosystem-climate models, and these models are more rigorously validated and improved, the value of their forecasts will increase, particularly to the societies and institutions that depend most directly on ecosystems.