The purpose of this chapter is to assess and quantify projections of possible future climate change from climate models. A background of concepts used to assess climate change experiments is presented in Section 9.2, followed by Section 9.3 which includes results from ensembles of several categories of future climate change experiments, factors that contribute to the response of those models, changes in variability and changes in extremes. Section 9.4 is a synthesis of our assessment of model projections of climate change.
In a departure from the organisation of the SAR, the assessment of regional information derived in some way from global models (including results from embedded regional high resolution models, downscaling, etc.) now appears in Chapter 10.
Studies of projections of future climate change use a hierarchy of coupled ocean/atmosphere/sea-ice/land-surface models to provide indicators of global response as well as possible regional patterns of climate change. One type of configuration in this climate model hierarchy is an Atmospheric General Circulation Model (AGCM), with equations describing the time evolution of temperature, winds, precipitation, water vapour and pressure, coupled to a simple non-dynamic "slab" upper ocean, a layer of water usually around 50 m thick that calculates only temperature (sometimes referred to as a "mixed-layer model"). Such air-sea coupling allows those models to include a seasonal cycle of solar radiation. The sea surface temperatures (SSTs) respond to increases in carbon dioxide (CO2), but there is no ocean dynamical response to the changing climate. Since the full depth of the ocean is not included, computing requirements are relatively modest so these models can be run to equilibrium with a doubling of atmospheric CO2. This model design was prevalent through the 1980s, and results from such equilibrium simulations were an early basis of societal concern about the consequences of increasing CO2.
However, such equilibrium (steady-state) experiments provide no information on time-dependent climate change and no information on rates of climate change. In the late 1980s, more comprehensive fully coupled global ocean/atmosphere/sea-ice/land-surface climate models (also referred to as Atmosphere-Ocean Global Climate Models, Atmosphere-Ocean General Circulation Models or simply AOGCMs) began to be run with slowly increasing CO2, and preliminary results from two such models appeared in the 1990 IPCC Assessment (IPCC, 1990).
In the 1992 IPCC update prior to the Earth Summit in Rio de Janeiro (IPCC, 1992), there were results from four AOGCMs run with CO2 increasing at 1%/yr to doubling around year 70 of the simulations (these were standardised sensitivity experiments, and consequently no actual dates were attached). Inclusion of the full ocean meant that warming at high latitudes was not as uniform as from the non-dynamic mixed-layer models. In regions of deep ocean mixing in the North Atlantic and Southern Oceans, warming was less than at other high latitude locations. Three of those four models used some form of flux adjustment whereby the fluxes of heat, fresh water and momentum were either singly or in some combination adjusted at the air-sea interface to account for incompatibilities in the component models. However, the assessment of those models suggested that the main results concerning the patterns and magnitudes of the climate changes in the model without flux adjustment were essentially the same as in the flux-adjusted models.
The most recent IPCC Second Assessment Report (IPCC, 1996) (hereafter SAR) included a much more extensive collection of global coupled climate model results from models run with what became a standard 1%/yr CO2-increase experiment. These models corroborated the results in the earlier assessment regarding the time evolution of warming and the reduced warming in regions of deep ocean mixing. There were additional studies of changes in variability in the models in addition to changes in the mean, and there were more results concerning possible changes in climate extremes. Information on possible future changes of regional climate was included as well.
The SAR also included results from the first two global coupled models run with a combination of increasing CO2 and sulphate aerosols for the 20th and 21st centuries. Thus, for the first time, models were run with a more realistic forcing history for the 20th century and allowed the direct comparison of the model's response to the observations. The combination of the warming effects on a global scale from increasing CO2 and the regional cooling from the direct effect of sulphate aerosols produced a better agreement with observations of the time evolution of the globally averaged warming and the patterns of 20th century climate change. Subsequent experiments have attempted to quantify and include additional forcings for 20th century climate (Chapter 8), with projected outcomes for those forcings in scenario integrations into the 21st century discussed below.
In the SAR, the two global coupled model runs with the combination of CO2 and direct effect of sulphate aerosols both gave a warming at mid-21st century relative to 1990 of around 1.5°C. To investigate more fully the range of forcing scenarios and uncertainty in climate sensitivity (defined as equilibrium globally averaged surface air temperature increase due to a doubling of CO2, see discussion in Section 9.2 below) a simpler climate model was used. Combining low emissions with low sensitivity and high emissions with high sensitivity gave an extreme range of 1 to 4.5°C for the warming in the simple model at the year 2100 (assuming aerosol concentrations constant at 1990-levels). These projections were generally lower than corresponding projections in IPCC (1990) because of the inclusion of aerosols in the pre-1990 radiative forcing history. When the possible effects of future changes of anthropogenic aerosol as prescribed in the IS92 scenarios were incorporated this led to lower projections of temperature change of between 1°C and 3.5°C with the simple model.
Spatial patterns of climate change simulated by the global coupled models in the SAR corroborated the IPCC (1990) results. With increasing greenhouse gases the land was projected to warm generally more than the oceans, with a maximum annual mean warming in high latitudes associated with reduced snow cover and increased runoff in winter, with greatest warming at high northern latitudes. Including the effects of aerosols led to a somewhat reduced warming in middle latitudes of the Northern Hemisphere and the maximum warming in northern high latitudes was less extensive since most sulphate aerosols are produced in the Northern Hemisphere. All models produced an increase in global mean precipitation but at that time there was little agreement among models on changes in storminess in a warmer world and conclusions regarding extreme storm events were even more uncertain.
Other reports in this collection