A Climate Modelling Primer

CHAPTER 1

Climate

Back in nineteen twenty seven

I had a little farm and I called it heaven

Prices up and the rain come down

I hauled my crops all into town

Got the money… bought clothes and groceries…

Fed the kids… and raised a big family

But the rain quit and the wind got high

Black old dust storm filled the sky

I traded my farm for a Ford machine

Poured it full of this gas-i-line

And started… rocking and a-rolling

Deserts and mountains…to California

(Talking Dust Bowl Blues, Woody Guthrie)

1.1 THE COMPONENTS OF CLIMATE

The term ‘climate’ has a very wide variety of meanings. To a geologist or geomorphologist, the ‘climate’ is an external agent which forces many phenomena of interest. For an archaeologist, the ‘climate’ of an earlier time might have been a crucial influence upon the people being studied, or might have been of little socio-economic significance, yet still so strong an environmental feature that it has left a ‘signature’ that can be interpreted. An agriculturalist probably sees the ‘climate’ as the background ‘norm’ upon which year-to-year and day-to-day weather is imposed, while the average person may speak of moving to a location with a ‘better climate’. To many of us, ‘climate’ often first suggests temperature, although rainfall and humidity may also come to mind. When we think of climatic change it used to be in the time frame of glacial periods. Recently, however, most of us have become aware of the shorter-term impact upon the climate of increasing atmospheric carbon dioxide and other trace greenhouse gases.

The climate is both a forcing agent and a feature liable to be disturbed. It can fluctuate on relatively short time-scales, producing for example the droughts that devastated West Africa in the 1970s and 1980s and, over much longer times, giving rise to glacial epochs. The climate is perceived in terms of the features of the entire climate system which most readily or most usefully characterize the phenomenon of interest. All of these characteristics of the climate are depicted in Figure 1.1. The three axes themselves are fundamental but the intervals are arbitrary and many more could be included.

A single satisfactory definition of climate is probably unobtainable because the climate system encompasses so many variables and so many time- and space-scales. One definition might be ‘all of the statistics describing the atmosphere and ocean determined over an agreed time interval (seasons, decades or longer), computed for the globe or possibly for a selected region’. This definition is broad, but it does serve to emphasize that higher-order statistics, such as variance (variability), can often be more useful in characterizing a climatic state than just the mean (average). The definition also permits further description of a climatic change as the difference between two climatic states, and a climatic anomaly as the difference between a climatic state and the mean state. The variations of the system arise from interactions between different parts of the climate system and from external forcings. Although the greatest variations are due to changes in the phase of water (i.e. frozen, liquid or vapour), the constituents of the atmosphere and ocean and the characteristics of the continental surface can also change, giving rise to a need for consideration of atmospheric chemistry, ocean biogeochemistry and land-surface exchanges.

Figure 1.1 The climate cube. Climate can be viewed as existing in at least three domains: time, space and human perception. The divisions of these domains depicted here are arbitrary – a great many more could be suggested. Historically, individual disciplines have been concerned with single ‘cells’. The extent of the climate system and the importance of interactions between domains are now well recognized

Introduction and outline of the book

In this book, we have set out to introduce and describe the way in which the climate is modelled. The climate models we will discuss are those developed using physically-based formulations of the processes that make up the climate system. We are concerned with explaining the approaches and methods employed by climate modellers and shall not focus directly on meteorology, socio-economic impacts of climatic changes or palaeoclimatic reconstruction, although all of these disciplines and many others will be drawn upon in our descriptions.

In this chapter, we identify the components of the climate system and the nature of their interactions, as well as describing briefly some of the motivations of climate modellers. Chapter 2 contains a history of climate modelling and provides an introduction to all the types of models to be discussed in subsequent chapters. The other chapters are concerned with different model types, their development and applications. Throughout, we have taken climate models to be predictive descriptions of regional- to global-scale phenomena; hence empirically based ‘models’ such as crop prediction equations and water resource management codes have not been included. The reason for this limitation is not that such models are uninteresting, but rather that they have grown from well-identified fields and thus background literature can be readily obtained elsewhere. Climate modelling in the sense in which we use the term, on the other hand, has developed from a wide variety of sources in a somewhat haphazard manner and consequently there is little accessible background to which the uninitiated can refer.

In one sense, the book develops the background material required for understanding of the most complex type of climate model, the fully coupled climate system model, by illustrating principles in other, simpler, model types. Thus, it is necessary to introduce the concept of energy balance, especially planetary radiation balance, before one-dimensional energy balance models (Chapter 3) can be understood. In Chapter 4, models that intentionally consider only a few of the important processes of the climate system are examined. These simpler models are used to gain a deeper understanding of the nature of feedbacks and forcings within the climate system as well as providing a foundation for impact assessment. These models, which have enjoyed a significant renaissance in the last ten years, are now widely known as Earth System Models of Intermediate Complexity (EMICs).

By Chapter 5, the reader should be well prepared to understand the way in which radiative forcing, ocean and atmosphere dynamics, biological processes and chemical changes are included in coupled three-dimensional models of the climate system. In Chapter 6, we explore some of the technical issues faced by climate modellers and look at how models are tested and their results evaluated. We also address how these results can be integrated with impact assessments in the development of social and economic policies.

Twentieth-century Classics (in Appendix A) is not an exhaustive list of references (which can be found on the accompanying Primer CD), but rather an introduction to the seminal works of the climate modelling literature. We have chosen these classics, with the help of a few friends. Appendix B contains a glossary of terms that may be new to readers unfamiliar with climatology/meteorology. As we have used this glossary for definitions rather than interrupt the main thread of the text, reference to it is recommended. The Primer CD (described in Appendix C) contains source code for a range of model types contributed by their developers. These will allow readers to make their own climate simulations ranging from global glaciations to increased CO2 experiments. A set of simulations from a global climate model also permits analysis of the results of a land-use change experiment. Also on the CD, movies illustrate some of the techniques used to analyse and display the results from a range of climate models.

Throughout the book, an effort will be made to underline the importance of simpler models in understanding the complex interactions between various components of the climate system. Complex models are only one, particularly sophisticated, method of studying climate. They are not necessarily the best tools; simple models are often used in conjunction with, or sometimes even to the exclusion of, more complex and apparently more complete models. The literature contains many fascinating examples of very simple models being used to demonstrate failures and illustrate processes in much more complex systems.

Last, but by no means least, any introduction to climate modelling must stress the crucial role played by computers. Without the recent growth in computational power and the reduction in computing costs, most of the developments in climate modelling that have taken place over the last four decades could not have happened. We have intentionally emphasized computing tools over mathematical skills in the description of the simplest type of climate model, the energy balance model (EBM), in Chapter 3. In that chapter, the steps required to construct a simple EBM are described, and the Primer CD includes example EBMs and source code.

It is estimated that a fully coupled ocean–atmosphere general circulation model (OAGCM) takes about 25–30 person-years to code, and the code requires continual updating as new ideas are implemented and as advances in computer science are accommodated. Most modellers who currently perform experiments with the most complex of models modify only particular components of the models. The size of and detail in these models means that only through a sharing of effort can progress be made. As the models have become increasingly complex, increased application of the principles of software engineering has become an essential part of the process and has made it easier to upgrade and exchange parts of the models. Host computers and model physics develop in parallel.

The climate system

The climate system was defined, in a document produced by the Global Atmospheric Research Programme (GARP) of the World Meteorological Organization in 1975, as being composed of the atmosphere, hydrosphere, cryosphere, land surface and biosphere. In 1992, the United Nations’ Framework Convention on Climate Change (FCCC) defined the climate system as ‘the totality of the atmosphere, hydrosphere, biosphere and geosphere and their interactions’. These definitions are similar, but the emphasis on interactions, both in the definition and in the literature, has grown in the thirty years since 1975. Figure 1.2 shows a schematic representation of the climate system components which climate modellers must consider. It complements Figure 1.1 by emphasizing components and processes rather than the space- and time-scales. The order of the components of the climate enumerated in 1975 is also a rough indicator of the historical order in which these elements were considered and, to some extent, the (increasing) magnitude of their time-scales.

Figure 1.2 A schematic illustration of the components and interactions in the climate system (modified from Houghton et al., 1996)

The first modelled component was the atmosphere, which, because of its low density and ease of movement, is the most ‘nervous’ of the climatic subsystems. These early models developed directly from weather prediction models. Precipitation was included early but many aspects of clouds (such as cloud liquid water and the effects of different cloud droplet sizes) are still difficult to incorporate successfully, and linking the major part of the hydrosphere, the oceans, into climate models had to wait for adequate computer resources. This was partly because the critical space- and time-scales of the ocean and atmosphere subsystems differ, but also because the coupling between the subsystems is strongly latitude-dependent. In the tropics, the systems are closely coupled, especially through temperature (Figure 1.3). In mid-latitudes the coupling is weak, predominantly via momentum transfer, whereas in high latitudes, there is a tighter coupling, primarily through salinity, which is closely involved in the formation of sea ice and oceanic deep water. Biochemical processes controlling the exchange of carbon dioxide between atmosphere and ocean also vary as a function of geographical location.

The cryosphere (frozen water) was first incorporated into climate models in the description of simple EBMs, in which the high albedo of the ice and snow dominated the radiative exchanges. The insulating effect of the cryosphere is at least as important as its albedo effect: sea ice decouples the ocean from the overlying atmosphere, and snow has a similar, but smaller, effect on land, causing considerable changes in separated subsystems.

Figure 1.3 A representation of the major coupling mechanisms between the atmosphere and ocean subsystems. The relative importance of these coupling mechanisms varies with latitude. The feedback between atmospheric temperature and oceanic salinity is interesting because it is strong only in the sense of the atmosphere forcing the ocean

Scientists concerned with land-surface processes had described the climate as both an agent and a feature of change for over a century before climate modellers began serious consideration of their theories. The importance of the biosphere has been underlined by the climate impacts resulting from atmospheric carbon dioxide levels dependent upon oceanic and terrestrial biota. Modern studies incorporate the state of the ecology on the continental surface and the growth of marine biota.

The stratospheric ‘ozone hole’, first identified over Antarctica in 1985, was the catalyst for incorporating atmospheric chemistry into climate models. Inclusion of these rapidly changing subsystems is still in its early stages, but it is already clear that Earth system models need to incorporate atmospheric and marine chemistry and transient changes in the world’s biota. The human component of the climate system, manifested particularly in trace gas and aerosol emissions and land use change, is perhaps its most difficult and challenging aspect. Human activities have only recently begun to be parameterized in climate and ‘integrated assessment’ models.

In this rather clumsy fashion and from mixed parentage, the discipline of climate modelling has evolved. Climate modellers have discovered that the system that they had summarized so neatly in 1975 is exceedingly complex, containing links and feedbacks which are highly non-linear and hence difficult to identify and reproduce.

1.2 CLIMATE CHANGE ASSESSMENT

Today, the atmosphere of planet Earth is undergoing changes unprecedented in human history and, although changes as large as those we are witnessing now have occurred in the geological past, relatively few have happened with the speed that characterizes today’s climate changes. Concentrations of greenhouse gases are increasing, stratospheric ozone has been depleted and the changing chemical composition of the atmosphere may be reducing its ability to cleanse itself through oxidation. These global changes threaten the balance of climatic conditions under which life evolved and is sustained. Temperatures are rising, ultraviolet radiation is increasing at the surface, and air pollutant levels are increasing. Many of these changes can be traced to industrialization, deforestation and other activities of a human population that is itself increasing at a very rapid rate.

Over the last fifteen or so years, with increased awareness of the potential impacts of changes in atmospheric concentrations of trace gases and aerosols, there has been an evolving demand from policymakers for the results of climate models. In 1988, the United Nations Environment Programme (UNEP) and the World Meteorological Organization (WMO) established the Intergovernmental Panel on Climate Change (IPCC). The IPCC was directed to produce assessments of available scientific information on climate change, written in such a way as to address the needs of policymakers and non-specialists. The First Scientific Assessment was published in 1990 in three volumes encompassing science, impacts and response. There was a scientific update in 1992 and two further volumes were produced as input to the First Conference of the Parties to the FCCC in March 1995. The Second Scientific Assessment followed in 1996; the Third Assessment was published in 2001 and the Fourth Assessment Report is due to be concluded and published in 2007. Around 700 researchers contributed to the Third Assessment and another 700 reviewed it.

An important result of the IPCC’s assessment of climate forecasts has been to focus interest on climatic reconstruction. The longest available record of proxybased Northern Hemisphere temperatures spans the period from 200 to 2000 AD (Figure 1.4). The proxies employed in making these detailed reconstructions include tree rings, corals, ice cores and written records of events such as floods, droughts, cold spells and even the blossoming of trees. Reconstructions have been made for longer times for the Northern Hemisphere because more data exist. These long records, even recognizing their measures of uncertainty, underline the fact that twenty-first century temperatures are warmer than any experienced over at least the last 1800 years.

The IPCC process aims to determine the current level of confidence in our understanding of the forcings and mechanisms of climate change, to find out how trust worthy the assessments are, and to ask whether we can yet unequivocally identify human-induced climate change. Through an exhaustive review process, the IPCC aims to provide assessments which discuss climate change on a global scale and represent international consensus of current understanding. Throughout the process, the goal is to include only information which has been subjected to rigorous review, although this is balanced by a desire to include the latest information in order that the best possible assessment can be made. These two competing desires mean that the development of the IPCC documents is an extremely timeconsuming process, but ensure that the final result is a powerfully strong statement of the state of current knowledge of the climate system. The IPCC assessment covers three areas, which are handled by three working groups. For the Third Assessment Report, published in 2001, Working Group I dealt with the scientific basis of climate modelling, climate observations and climate predictions, Working Group II dealt with issues relating to the impacts of, adaptations to, and vulnerability to climate change while Working Group III reported on mitigation, i.e. actions to reduce climate change.

Figure 1.4 Comparison of Northern Hemisphere temperature reconstructions with model simulations of Northern Hemisphere mean temperature changes over the past millennium based on radiative forcing histories. Also shown are two independent reconstructions of warm season extra-tropical continental Northern Hemisphere temperatures and an extension back through the past 2000 years based on eight long reconstructions. All reconstructions have been scaled to the period 1856–1980 and are shown with respect to the 1961–1990 based period. This is a slightly modified version of the figure that appeared in EOS Vol. 84, courtesy of Michael Mann. Reproduced by permission of the American Geophysical Union from Mann et al. (2003), EOS 84, 256–257

1.2.1 The scientific perspective

It is generally accepted that physically-based computer modelling offers the most effective means of answering questions requiring predictions of the future climate and of potential impacts of climatic changes. Although there have been great advances made in such modelling over the past 40–50 years, even the most sophisticated models are still far removed in complexity from the full climate system. Further advances are possible, but they need to be associated with increased understanding of the nature of interactions within the real climate system and translated to those within models. Perturbations caused by everything from industrial aerosols to volcanoes, from solar luminosity to climatically induced variation in surface character must be considered. Modelling in such a widely ranging subject is a formidable task and it requires co-operation between many disciplines if reliable conclusions are to be drawn.

Available computing power has increased greatly over the past 40–50 years (Figure 1.5a). Meteorological and climate research establishments have some of the fastest and most powerful computers available. This continuing increase has meant that climate models have expanded in terms of complexity, resolution and in potential simulation time. As computing capabilities have evolved, the components of the Earth system that can be included and coupled have increased, and will continue to increase in number (Figure 1.5b). Multi-decadal simulations, with full diurnal and seasonal cycles and fully coupled ocean and sea ice, are now expected in climate experiments, and transient changes in, for example, the atmospheric greenhouse gases and aerosol loading now replace the previous equilibrium simulations. As our knowledge increases, more aspects of the climate system will be incorporated into climate models, the resolution and length of integrations will further increase and additional components will be incorporated.

Figure 1.5 (a) Peak performance of the most powerful computers between 1953 and 2003. The power is given in millions of instructions per second (MIPS) up to 1975 and in millions of floating point operations per second (MFLOPS) since then. Note that the vertical scale is logarithmic and supercomputer performance shows no signs of levelling off. System performance has been hypothesized to continue on this trend to 2020. (b) Schematic of the interdependency of computer power and model capability. As well as increased resolution, modellers have progressively coupled more models to create today’s unified Earth system models. Foreseeable advances in computer technology will allow simulations with even more sophisticated Earth system models to be constructed

The general trend that, as computer power increases, so do the complexity, resolution and length of climate model simulations is moderated by different contributing specialist groups. For example, biospheric modellers have tended to favour increasing the number of components in their submodels, while the ocean modellers have driven the resolution of their submodels higher. Spatial and temporal resolution compatibility is critical to effective and integrative coupling. Indeed, the drive towards fully coupled ocean–atmosphere biogeochemical models has seen computational demand reach new heights. New model and software engineering designs, offering better numerical representation of the climate system, promise to challenge the fastest computers for years to come. However, it would be a mistake to think that the only measure of success of a climate model is the resolution or the speed of computation achieved. The purpose of the climate models is to gain insight into the climate system and its interactions. While improved resolution and faster computers are very helpful, there are many other modelling avenues to be explored which can aid our understanding of climate.

Figure 1.6 shows the performance of a group of coupled ocean–atmosphere models that participated in CMIP, the Coupled Model Intercomparison Project. Superimposed is the envelope of atmosphere-only performance for models between 1974 and 1984. There has been considerable improvement in model simulation of observed characteristics of the climate system over the last 20 years. Certainly, some of this improvement has come with faster computers, as they have helped to increase the possible size and complexity of models, but simple models have also played a role. Simple models may be sufficient to answer particular, well-specified problems and provide insight that might otherwise be hidden by the complexity of a larger model.

Whether its predictions are correct, for the right reasons, is the ultimate test of any model. Weather forecast models can be tested over a period of a few hours to a few days, but models of climate are required to predict decades to centuries in advance or to simulate periods of the Earth’s history for which validation data are scant. Importantly, climate model ‘predictions’ offer only a general case of the response since the model climate loses its association with the initial conditions within a few weeks. Hence, testing of single simulations is virtually meaningless and ensembles of results are needed to characterize the climate. Despite the limitations placed by chaos theory on our ability to predict the exact state of the atmosphere beyond about 10–15 days into the future, there is good reason to believe that our ability to predict the nature of the ensemble state (the climate) is not impaired.

A useful analogy might be with a gambler, who sees the chaotic processes of the roulette wheel as unpredictable. The casino owner, however, knows the boundary conditions set by the structure and layout of the wheel and the rules of the game, which mean that the casino exists in a winning ‘climate’. A more meteorological illustration would be that we are generally comfortable with the notion of making predictions based on known constraints or a statistical envelope when deciding where to take our annual holiday. We know that certain times of year and certain locations will be acceptable to us (either delightfully sunny or enjoyably snowy), but we cannot guarantee the exact nature of each day of the holiday. The weather depends on the exact state of the atmosphere within a week or so of the beginning of the holiday, rather than the overarching constraints of, for example, seasonal conditions and ocean surface temperature, which are largely similar from one year to the next.

Figure 1.6 Model performance from the Coupled Model Intercomparison Project (CMIP) for selected ocean and atmosphere variables. Solid line indicates observations in the case of (b), (c) and (d) and model mean for (a). An envelope of performance for earlier atmospheric models (hatched) is shown in (c) and (d) from Gates (1985), illustrating the change in model performance over the intervening years. Reprinted from Global and Planetary Change, 37, Covey et al., pp. 103–133. Copyright 2003, with permission from Elsevier

The climate models discussed in this book cover a wide range of space- and timescales. These different types of climate models attract interest from many different disciplines. Long-period modelling may attract glaciologists, geologists or geophysicists. For example, even simple models can predict the effect on mean temperatures of volcanic eruptions such as Mount Pinatubo quite successfully on seasonal or longer time-scales, so we can have confidence that climate predictions are not obfuscated by the same chaotic processes that trouble weather forecasts. Atmospheric chemists, dealing with complex reactions that typically have very short time-scales, are successfully incorporating these processes into three-dimensional climate models. Implications of solar-system-scale phenomena attract planetary physicists and astronomers, while social and economic scientists are interested in the human component of the climate system. In this book, we will attempt to show how these contributions fit together and jointly enhance the science of climate modelling.

1.2.2 The human perspective

Any changes in climate, whatever the cause, may impact human activities. Cropyield models have been used to quantify how food production depends on the weather. It might therefore be postulated, for example, that a change in climate could lead to consistently low or high yields in a particular area, which, in turn, may lead to a human response in terms of a change in agricultural practice. Such simple postulates can be misleading, since they conceal several problems that are inherent in relating climate change to human impact. These concern the nature of climatic changes themselves, the strength of the relationship between climate changes and human response and the availability of (past) climatic and sociological data for evaluation.

It is possible to think of climatic changes as being represented by changes in the long-term mean values of a particular climatic variable. Superimposed on this changing mean value will be decadal fluctuations and year-to-year variations. Such short-period variations may, of course, be influenced by the change in the mean. On the human time-scale, changes in the mean value are likely to be so slow as to be almost imperceptible. For example, the changes over the last few decades can only be detected by careful analysis of instrument records. Much more noticeable will be variability, expressed, for example, as a ‘run of bad winters’. Any human response will depend on such a perception, whether consciously or subconsciously. A ‘large’ climate change may not lead to any response, whereas a much smaller change in a particular feature, expressed as a perceived change in variability, may have a profound impact on human activity. Detection, for example, of climate change in response to increasing atmospheric trace greenhouse gases is very difficult in the early stages if only one response is monitored. For this reason, ‘fingerprint’ methods have been proposed which monitor a set of small changes in a number of variables and require prespecified thresholds in all of them to be passed before a signal can be established.

Any attempt to establish the impact of past climatic changes must use historical information. Pre-instrumental historical records are qualitative and selective and emphasize information about unusual conditions which were perceived as having an impact. Consequently, they can tell us less about normal conditions than about abnormal ones. A great deal, therefore, needs to be inferred about the historical climate and its variability before any suggestions regarding its impacts can be made. Even if a change occurs which potentially has a significant impact on human activity, a societal response will not necessarily follow. Any response to a climate change is governed by a host of non-climatic factors which need to be considered. Clear, and particularly direct, links between climate change and human activity are often difficult to establish. This problem of ‘attribution’ to human-produced greenhouse gas increases following detection of global warming is currently