The physical characteristics of the land surface, including the vegetation cover, have a strong effect on the absorption of solar energy and on the fluxes of heat, water vapour and momentum between the surface and atmosphere. These fluxes at any given location strongly influence the local surface climate and have effects on the atmosphere which, in some cases, extend globally. Of particular importance are changes in the extent of highly reflective ice and snow cover, as climate warms, the area of ice and snow will decrease, leading to greater absorption of solar energy and further warming. However, concurrent changes in cloud cover induced by the changes in ice and snow extent complicate the picture considerably. Correct simulation of landsurface changes and their net effect requires models with high spatial and temporal resolution on account of potential interactions with clouds and because of the spatial heterogeneity of the surface (see SAR WGI: Sections 1.4.3 and 4.4). On a time-scale of decades to centuries, changes in the vegetative cover and soil properties will also alter the exchanges of heat, moisture and momentum between the surface and atmosphere, as well as the sources and sinks of a number of greenhouse gases.

The oceans play a number of important roles in the climate system and in climatic change. First, they are a major storehouse of carbon, and have played an important role in absorbing a portion of the anthropogenic CO2 emitted up to the present. This role will continue to some extent in the future. Second, ocean currents transport substantial amounts of heat, thereby exerting a strong influence on regional climates. Changes in oceanic heat transport could significantly affect regional climatic changes, possibly causing some regions to cool temporarily and others to warm by considerably more than the global mean as the global climate warms. Third, the absorption and downward mixing of heat by the oceans considerably slows down the rate of surface warming. This reduces those impacts which depend on the rate of climatic change, but also implies that, until some time after greenhouse gas concentrations have been stabilized, there will be an irreversible commitment to more climatic change than has already occurred. Ocean currents and the rate of absorption of heat by the oceans depend on wind patterns and the exchange of heat and freshwater (through precipitation and evaporation) between the ocean and the atmosphere. At high latitudes, the presence of sea ice has a very strong effect on these exchanges, so the satisfactory simulation of sea ice is of considerable importance (see SAR WGI: Sections 1.4.2, 4.3, and 6.2; and SAR WGI: Chapter 10).

control over the formation, nature and lifespan of clouds. thereby providing a direct coupling to both solar and infrared radiation budgets. Atmospheric heat transport and changes therein will also influence the response of sea ice and land snow cover to global mean temperature changes, thereby providing another link to the Earth's overall radiative balance. Changes in atmospheric winds, or in evaporation and precipitation due in part to changes in atmospheric winds, could also lead to significant and possibly abrupt changes in the oceans' circulation (see SAR WGI: Sections 4.2, 4.3, and 6.2).

The temperature of the Earth tends to adjust itself such that there is a balance between the absorption of energy from the Sun and the emission of infrared radiation from the surfaceatmosphere system. If, for example, there were to be an excess of absorbed solar energy over emitted infrared radiation (as occurs with the addition of greenhouse gases to the atmosphere), temperatures would increase but, in so doing, the emission of infrared radiation to space would increase. This would reduce the initial imbalance, and eventually a new balance would be achieved, but at a new, warmer temperature (see SAR WGI: Sections 1.2 and 1.3.1).

Anthropogenic greenhouse gases and aerosols affect the climate system by altering the balance between absorbed solar radiation and emitted infrared radiation, as discussed in the SAR WGI (Section 2.4). The imbalance is quantified as the "radiative forcing", which is defined as the change in net downward radi ation (combined solar and infrared) at the tropopause when, for example, greenhouse gas or aerosol amounts are altered, after allowing for the adjustment of stratospheric temperatures only. The surface climate responds to the initial change in net radiation at the tropopause rather than at the surface itself or at the top of the atmosphere because the surface and troposphere are tightly coupled through heat exchanges, and respond as a unit to the combined heating perturbation. The adjustment of the stratosphere is included in the radiative forcing because the stratosphere responds quickly and independently from the surface-troposphere system. Non-anthropogenic radiative forc ings relevant at the decade to century time-scales include variations in solar luminosity and volcanic eruptions, the latter producing reflective sulphate aerosols which are effective for several years if injected into the stratosphere.

The radiative forcing for a CO2 doubling is 4.0-4.5 W m2 before adjustment of stratospheric temperatures (Cess, et al., 1993); allowing for stratospheric adjustment reduces the forcing by about 0.5 W m2 to 3.5-4.0 W m2. If temperature were the only climatic variable to change in response to this

12

2.3.2 Fast and Slow Feedbacks

A feedback is a process whereby an initial change in some variable ("A") leads to a change in another variable ("B") which then produces further changes in the initial variable. A positive feedback is such that the change in B leads to further changes in A in the same direction as the original change, thereby tending to amplify the initial change. Anegative feedback, on the other hand, acts to diminish the initial change. Among the feedbacks which have to be considered in the calculation of global mean climatic change are the following: (a) Water vapour amount: in a warmer climate the atmospheric concentration of water vapour will increase. Since water vapour is a greenhouse gas, this represents a positive feedback; (b) Clouds: changes in clouds are difficult to calculate reliably, as noted in Section 2.2.1. Clouds have a strong radiative effect, and are, therefore, likely to produce a noticeable feedback. This feedback depends on changes in the amount, altitude and characteristics of the clouds, as well as on the reflectivity of the underlying surface, so even the sign of the feedback is uncertain; (c) Areal extent of ice and snow: a reduction in the area of sea ice and seasonal snow cover on land as climate warms will reduce the surface reflectivity, thereby tending to produce greater warming (a positive feedback). As noted in Section 2.2.2, however, concurrent changes in cloud cover complicate the picture considerably; (d) Vegetation: changes in the distribution of different biomes or in the nature of vegetation within a given biome can also lead to changes in the surface reflectivity, thereby exerting a feedback effect on climatic change; (e) The carbon cycle: the effect of climate on the terrestrial biosphere and the oceans is likely to alter the sources and sinks of CO2 and CH4, leading to changes in their atmospheric concentrations and hence causing a radiative feedback (see SAR WGI: Sections 1.4, 2.1, 4.2, and 4.4; and Chapters 9 and 10).

Of these feedbacks, those involving water vapour and clouds respond essentially instantaneously to climatic change, while those involving sea ice and snow respond within a few years. We therefore refer to these as "fast" feedbacks. Some vegetation and carbon cycle processes are relevant on a time-scale of decades, whereas others not listed above, such as a reduction in the area of continental ice sheets, dissolution of carbonate sediments in the ocean and enhanced chemical weathering on land (the latter two of which tend to reduce the atmospheric CO2 concentration), require hundreds to thousands of years to unfold. These are referred to as "slow" feedbacks.

Climate Sensitivity: Definition

The term "climate sensitivity" refers to the steady-state increase in the global annual mean surface air temperature associated with a given global mean radiative forcing. It is standard practice to include only the fast feedback processes, including changes in water vapour, in the calculation of climate sensitivity, but to exclude possible induced changes in the concentrations of other greenhouse gases (as well as other slow feedback processes).

As noted above (in the introduction to Section 2.3), the temperature of the Earth tries to adjust itself such that there is a balance between absorbed solar radiation and emitted infrared radiation. If there is an energy surplus, temperatures will tend to increase, thereby increasing the emission of infrared radiation to space. The more strongly that infrared emission to space increases with temperature (that is, the stronger the radiative damping), the smaller the temperature increase required to re-establish zero net energy balance and the smaller the climate sensitivity. Changes in the albedo (reflectivity) of the atmosphere-surface system also contribute (positively or negatively) to the radiative damping. The fast feedback processes, thus, affect climate sensitivity by affecting the ease with which excess heat can be radiated to space — that is, by altering the radiative damping.

It is common practice to use CO2 doubling as a benchmark for comparing climate model sensitivities. As reported in the SAR WGI (Technical Summary, Section D.2), the climate sensitivity for a CO2 doubling is expected to fall between 1.5 and 4.5°C. To the extent that the global mean temperature response depends only on the global mean forcing, any combination of greenhouse gas, solar luminosity and aerosol forcings which give the same net forcing as for a doubling of CO2, will produce the same global mean temperature response in steady state. To the extent that the climate sensitivity is constant, the steadystate temperature response will vary in proportion to the net forcing. However, as discussed below, both of these conditions are only rough approximations.

Given the many non-linearities associated with the fast feedback processes, which determine the climate sensitivity as defined above, one might expect that the climate sensitivity will depend both on the magnitude of the forcing and on the vertical, latitudinal and seasonal distribution of the forcing. However, experiments with a variety of models indicate that, for forcings up to the magnitude that could be experienced during the next century, the climate sensitivity is approximately constant (that is, the global mean surface temperature response is roughly proportional to the global mean forcing). Also, for a number of different forcings, the climate sensitivity is largely independent of the specific combination of factors producing a given global mean forcing. In particular, the global mean temperature response to a mixture of greenhouse gas increases is within about 10 per cent

of the response to a CO2 increase alone having the same global mean forcing as for the mixture of gases (IPCC94: Sections 4.1.1 and 4.8; and SAR WGI: Section 6.2.1.1).

On the other hand, the rough proportionality between global mean forcing and global mean temperature response established for well-mixed gases and solar luminosity variations can break down for cases involving very large and spatially or seasonally heterogeneous forcings (such as those due to variations in the Earth's orbit, which occur over periods of tens of thousands of years), or in which particularly strong interactions between the forcing and clouds occur. This appears to be the case for changes in tropospheric O, and in tropospheric aerosols, both of which produce much stronger spatial variations in the radiative forcing than for changes in well-mixed gases, and which have a decidedly different vertical pattern of forcing (IPCC94: Sections 4.1.1 and 4.8).

In spite of the possibility that the global mean climate sensitivity to aerosol and tropospheric O, changes is different from that for changes in other greenhouse gases, the SCMs used in the SAR WGI (Section 6.3) are such that the same sensitivity is assumed for all of these forcings. However, the climatic response to a given aerosol increase depends on both the climate sensitivity to aerosol increases and on the aerosol forcing, the latter being highly uncertain (ranging from -0.2 Wm2 to -2.3 W m2; see SAR WGI: Section 2.4.2). Thus, the uncertainty in climatic change due to possible differences in the climate response to increases in aerosols and in well-mixed greenhouse gases is, at present, overwhelmed by the uncertainty in the aerosol forcing itself.

13

Irrespective of the extent to which the global mean temperature response depends only on the net global mean forcing, different combinations of forcings involving O3, aerosols and wellmixed greenhouse gases will produce substantially different climatic changes in any given region. This is especially true for increases in tropospheric aerosols, where regional cooling can occur in the midst of global mean warming, and to a lesser extent for stratospheric and tropospheric O, changes (SAR WGI: Chapter 6). Thus, the climatic change in a given region associated with a given global mean forcing depends on the specific forcings involved when combining aerosol and ozone forcings with those of well-mixed greenhouse gases, even if the global mean temperature response is roughly the same. Furthermore, when large net negative forcings occur at the regional scale due to the effects of aerosols, the cooling effects will not be restricted to the immediate regions where aerosols occur, due to the effects of heat transport by winds and ocean

currents.

There will also be strong regional variations in the climatic response to greenhouse gas increases even in the case of wellmixed gases, such as CO2 and CH4, whose forcing is relatively uniform from one region to the next. This is due to spatial variations in the nature and strength of various feedback processes (such as those involving snow cover, sea ice and clouds) and in atmospheric winds and ocean currents, which can be expected to change in response to overall changes in the global climate (see SAR WGI: Chapter 6).

In order to project the impact of human perturbations on the climate system, it is necessary to calculate the effects of all the key processes operating in the climate system. These processes can be represented in mathematical terms, but the complexity of the system means that the calculations can only be performed in practice using a computer. The mathematical formulation is therefore implemented in a computer program, which we refer to as a "model". If the model includes enough of the components of the climate system to be useful for simulating the climate, it is commonly called a "climate model".

A climate model which explicitly included all our current understanding of the climate system would be too complex to run on any existing computer. For practical purposes, some compromises have to be made. The basic question is: in how much detail should the components and processes of the climate system be represented? If the representation is simplified, fewer calculations are needed and the model can be run faster or on a

less powerful computer.

The most detailed model of a particular process is one which is based on fundamental physical principles which we believe to be invariant. Such a model would be applicable to any climate. In order to represent the process in a way which can be used in a climate model, additional, simplifying assumptions have to be introduced. In some cases, empirically-derived relationships are included. When this is necessary, the range of the validity of the model will inevitably become more limited. As far as possible, climate models make use of basic physical principles or of simplifications which introduce minimal uncertainty. This is necessary because the conditions of a changed climate may be quite different from current conditions, so relationships derived empirically or statistically for the current climate will not necessarily hold (SAR WGI: Section 1.6).

In the most complex climate models, physical quantities which vary continuously in three dimensions are represented by their values at a finite number of points arranged in a threedimensional grid. This is clearly necessary because we can do only a finite number of calculations. The spacing between the points of the grid is the "spatial resolution". The finer the resolution, the larger the number of points, and the more calculations there are to be done. Hence, the resolution is limited by the computing resources available. The typical resolution that can be used in a climate model is hundreds of kilometres in the horizontal. Many important elements of the climate system (e.g., clouds, land surface variations) have scales much smaller than this. Detailed models at high resolution are available for such processes by themselves, but these are computationally too expensive to be included in a climate model. Instead, the climate model has to represent the effect of these sub-grid scale processes on the climate system at its coarse grid scale. A formulation of the effect of a small-scale process on the large-scale is called a "parametrization" (SAR WGI: Section 1.6.1). All climate models use parametrization to some extent.

One-dimensional radiative-convective atmospheric models. These models are globally (horizontally) averaged but contain many layers within the atmosphere. They treat processes related to the transfer of solar and infrared radiation within the atmosphere in considerable detail, and are particularly useful for computing the radiative forcing associated with changes in the atmosphere's composition. The change in atmospheric water vapour amount as climate changes must be prescribed (based on observations), but the impact on radiation associated with a given change in water vapour can be accurately computed. Radiative-convective models thus provide one means for determining one of the key feedbacks which are important to climate sensitivity through a combination of observations and wellestablished physical processes.

One-dimensional upwelling-diffusion ocean models. The atmosphere is treated as a single well-mixed box that exchanges heat with the underlying ocean and land surface. The absorption of solar radiation by the atmosphere and surface depends on the specified surface reflectivity and atmospheric transmissivity and reflectivity. The emission of infrared radiation to space is a linearly increasing function of atmospheric temperature in this model, with the constant of proportionality serving as the infrared radiative damping. The ocean is treated as a one-dimensional column which represents a horizontal average over the real ocean, excluding the limited regions where deep water forms and sinks to the ocean bottom, which are treated separately. Figure 2 illustrates this model. The sinking in polar regions is represented by the pipe to the side of the column. This sinking and the compensating upwelling within the column represent the global scale thermohaline circulation. This model is used primarily to study the role of the oceans in the surface temperature response to changes in radiative forcing.