Climate varies due to both natural and human causes. Natural influences on climate include variations in solar output and the effects of volcanic eruptions. The most important human influence on climate is that due to increasing concentrations of the greenhouse gases carbon dioxide, methane and nitrous oxide. By adding to the greenhouse effect these additional concentrations contribute to warming of the surface and lower atmosphere, and in the case of carbon dioxide, cause a strong cooling of the stratosphere. Another human factor is that of aerosols. These are small airborne particles associated with combustion that reflect sunlight and can make clouds brighter and more reflective as well as longer-lasting. Aerosols generally act to cool the climate, although some types of aerosol can absorb radiation thereby warming some regions. A further human factor is that of changes in land use, for example through deforestation.

A climate model is a computer programme designed to simulate Earth's climate in order to understand and predict its behaviour. Climate models are largely based on a set of mathematical equations that describe the fundamental physical laws which govern the fluid dynamics and thermodynamics of the atmosphere and ocean and their interactions with other parts of Earth's climate system (e.g. land surface and ice sheets). Processes for which underlying equations are not known or that occur at scales smaller that the grid resolution (e.g. clouds, vegetation) are represented by spatially averaged relationships. The mathematical equations describe how variables such as temperature, pressure and wind change over time and are based on the Navier-Stokes equations. These equations are solved using large supercomputers as no analytical solutions are known. Observations are also used to develop the models, particularly in the testing phase. These observations come from instruments such as ocean buoys, weather balloons, satellites and instrumented commercial aircraft. When the forcing factors (e.g. intensity of the sun, concentrations of greenhouse gases, dust from volcanic eruptions) are prescribed for the model, they can be used to simulate the past and present climates, and possible future climates, given scenarios of future anthropogenic emissions. The models represent Earth's climate by dividing the surface, ocean and atmosphere into a grid; imagine each part of the Earth has its own box, much like pixels in a digital photograph (in some cases, wave-like representations are used instead of grid boxes). In global models, the spacing (or size) of these boxes is typically in the range of 25–300km. In regional climate models, the spacing is typically smaller, 2–50km. For each box, the change in a variable (such as wind, temperature or rainfall) over a specified amount of time is calculated. The time step (the amount of time between each calculation) depends on the size of the grid boxes and is usually a few minutes to about half an hour in order to solve the equations with sufficient accuracy. Models can be made up of millions of grid boxes and are run over millions of time steps; this can result in the simulations taking months to produce. A number of simulations (an ensemble) are run for each scenario to estimate the mean climate and the uncertainty due to internal fluctuations in the climate system. The climate system is very complex, and the great benefit of climate models is that they allow us to test theories in a controlled environment. Climate models can help us predict how the climate might vary in the future. They can also improve our understanding of variables such as temperature, precipitation, ocean currents and sea ice cover. Models are also used to answer questions such as when the next El Niño might occur, what might happen if greenhouse gas concentrations double and what factors controlled Earth's climate in the past? Simulating possible future climate change at regional and national scales for different emission scenarios is important for helping evaluate policy options. Climate models are based on fundamental scientific principles and are our best tool for making predictions about future climate. We are confident that models provide useful information because they are based on well-established physical laws and spontaneously reproduce many features of the observed climate, including jet streams, storms, important ocean effects such as El Nino and many aspects of how climate has changed in the past. Nevertheless, despite extensive evaluation and development, models are not perfect. They necessarily involve approximations, particularly related to small-scale processes that cannot be simulated directly because of limited computer power and insufficient observations. How clouds are represented in climate models is a particular challenge because this involves a wide range of processes and different space and time scales. Predictions of different models are therefore often qualitatively similar on large scale changes but differ in magnitude and regional detail. A large part of climate science involves the evaluation, improvement and testing of climate models. Although uncertainties exist, models are unanimous that the climate has warmed, and will warm further, in response to increased greenhouse gases entering the atmosphere. This paper was developed in collaboration with the Royal Meteorological Society's Climate Science Communications Group. The Society thanks Kevin Trenberth, Mat Collins and Keith Williams for reviewing the paper.

Using optimal detection techniques with climate model simulations, most of the observed increase of near surface temperatures over the second half of the twentieth century is attributed to anthropogenic influences. However, the partitioning of the anthropogenic influence to individual factors, such as greenhouse gases and aerosols, is much less robust. Differences in how forcing factors are applied, in their radiative influence and in models' climate sensitivities, substantially influence the response patterns. We find standard optimal detection methodologies cannot fully reconcile this response diversity. By selecting a set of experiments to enable the diagnosing of greenhouse gases and the combined influence of other anthropogenic and natural factors, we find robust detections of well mixed greenhouse gases across a large ensemble of models. Of the observed warming over the 20th century of 0.65K/century we find, using a multi model mean not incorporating pattern uncertainty, a well mixed greenhouse gas warming of 0.87 to 1.22K/century. This is partially offset by cooling from other anthropogenic and natural influences of -0.54 to -0.22K/century. Although better constrained than recent studies, the attributable trends across climate models are still wide, with implications for observational constrained estimates of transient climate response. Some of the uncertainties could be reduced in future by having more model data to better quantify the simulated estimates of the signals and natural variability, by designing model experiments more effectively and better quantification of the climate model radiative influences. Most importantly, how model pattern uncertainties are incorporated into the optimal detection methodology should be improved.

Advances in the science and observation of climate change are providing a clearer understanding of the inherent variability of Earth's climate system and its likely response to human and natural influences. The implications of climate change for the environment and society will depend not only on the response of the Earth system to changes in radiative forcings, but also on how humankind responds through changes in technology, economies, lifestyle and policy. Extensive uncertainties exist in future forcings of and responses to climate change, necessitating the use of scenarios of the future to explore the potential consequences of different response options. To date, such scenarios have not adequately examined crucial possibilities, such as climate change mitigation and adaptation, and have relied on research processes that slowed the exchange of information among physical, biological and social scientists. Here we describe a new process for creating plausible scenarios to investigate some of the most challenging and important questions about climate change confronting the global community.

Trustworthy probabilistic projections of regional climate are essential for society to plan for future climate change, and yet, by the nonlinear nature of climate, finite computational models of climate are inherently deficient in their ability to simulate regional climatic variability with complete accuracy. How can we determine whether specific regional climate projections may be untrustworthy in the light of such generic deficiencies? A calibration method is proposed whose basis lies in the emerging notion of seamless prediction. Specifically, calibrations of ensemble-based climate change probabilities are derived from analyses of the statistical reliability of ensemble-based forecast probabilities on seasonal time scales. The method is demonstrated by calibrating probabilistic projections from the multimodel ensembles used in the Fourth Assessment Report (AR4) of the Intergovernmental Panel on Climate Change (IPCC), based on reliability analyses from the seasonal forecast Development of a European Multimodel Ensemble System for Seasonal-to-Interannual Prediction (DEMETER) dataset. The focus in this paper is on climate change projections of regional precipitation, though the method is more general.

This report summarizes the findings and recommendations from the Expert Meeting on New Scenarios held in Noordwijkerhout, The Netherlands, 19-21 September 2007. It is the culmination of the combined efforts of the New Scenarios Steering Committee, an author team composed primarily of members of the research community, and numerous other meeting participants and external reviewers who provided extensive comments during the expert review process

A coordinated set of global coupled climate model [atmosphere–ocean general circulation model (AOGCM)] experiments for twentieth- and twenty-first-century climate, as well as several climate change commitment and other experiments, was run by 16 modeling groups from 11 countries with 23 models for assessment in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4). Since the assessment was completed, output from another model has been added to the dataset, so the participation is now 17 groups from 12 countries with 24 models. This effort, as well as the subsequent analysis phase, was organized by the World Climate Research Programme (WCRP) Climate Variability and Predictability (CLIVAR) Working Group on Coupled Models (WGCM) Climate Simulation Panel, and constitutes the third phase of the Coupled Model Intercomparison Project (CMIP3). The dataset is called the WCRP CMIP3 multimodel dataset, and represents the largest and most comprehensive international global coupled climate model experiment and multimodel analysis effort ever attempted. As of March 2007, the Program for Climate Model Diagnostics and Intercomparison (PCMDI) has collected, archived, and served roughly 32 TB of model data. With oversight from the panel, the multimodel data were made openly available from PCMDI for analysis and academic applications. Over 171 TB of data had been downloaded among the more than 1000 registered users to date. Over 200 journal articles, based in part on the dataset, have been published AMERICAN METEOROLOGICAL SOCIETY so far. Though initially aimed at the IPCC AR4, this unique and valuable resource will continue to be maintained for at least the next several years. Never before has such an extensive set of climate model simulations been made available to the international climate science community for study. The ready access to the multimodel dataset opens up these types of model analyses to researchers, including students, who previously could not obtain state-of-the-art climate model output, and thus represents a new era in climate change research. As a direct consequence, these ongoing studies are increasing the body of knowledge regarding our understanding of how the climate system currently works, and how it may change in the future.

This paper investigates the impact of aerosol forcing uncertainty on the robustness of estimates of the twentieth-century warming attributable to anthropogenic greenhouse gas emissions. Attribution analyses on three coupled climate models with very different sensitivities and aerosol forcing are carried out. The Third Hadley Centre Coupled Ocean - Atmosphere GCM (HadCM3), Parallel Climate Model (PCM), and GFDL R30 models all provide good simulations of twentieth-century global mean temperature changes when they include both anthropogenic and natural forcings. Such good agreement could result from a fortuitous cancellation of errors, for example, by balancing too much ( or too little) greenhouse warming by too much ( or too little) aerosol cooling.Despite a very large uncertainty for estimates of the possible range of sulfate aerosol forcing obtained from measurement campaigns, results show that the spatial and temporal nature of observed twentieth-century temperature change constrains the component of past warming attributable to anthropogenic greenhouse gases to be significantly greater ( at the 5% level) than the observed warming over the twentieth century. The cooling effects of aerosols are detected in all three models.Both spatial and temporal aspects of observed temperature change are responsible for constraining the relative roles of greenhouse warming and sulfate cooling over the twentieth century. This is because there are distinctive temporal structures in differential warming rates between the hemispheres, between land and ocean, and between mid- and low latitudes. As a result, consistent estimates of warming attributable to greenhouse gas emissions are obtained from all three models, and predictions are relatively robust to the use of more or less sensitive models. The transient climate response following a 1% yr(-1) increase in CO2 is estimated to lie between 2.2 and 4 K century(-1) (5-95 percentiles).

Sea surface temperature (SST) observations in the North Atlantic indicate the existence of strong multidecadal variability with a unique spatial structure. It is shown by means of a new global climate model, which does not employ flux adjustments, that the multidecadal SST variability is closely related to variations in the North Atlantic thermohaline circulation (THC). The close correspondence between the North Atlantic SST and THC variabilities allows, in conjunction with the dynamical inertia of the THC, for the prediction of the slowly varying component of the North Atlantic climate system. It is shown additionally that past variations of the North Atlantic THC can be reconstructed from a simple North Atlantic SST index and that future, anthropogenically forced changes in the THC can be easily monitored by observing SSTs. The latter is confirmed by another state-ofthe-art global climate model. Finally, the strong multidecadal variability may mask an anthropogenic signal in the North Atlantic for some decades.

Forecasts of climate change are inevitably uncertain. It is therefore essential to quantify the risk of significant departures from the predicted response to a given emission scenario. Previous analyses of this risk have been based either on expert opinion 1 , perturbation analysis of simplified climate models 2 , 3 , 4 , 5 or the comparison of predictions from general circulation models 6 . Recent observed changes that appear to be attributable to human influence 7 , 8 , 9 , 10 , 11 , 12 provide a powerful constraint on the uncertainties in multi-decadal forecasts. Here we assess the range of warming rates over the coming 50 years that are consistent with the observed near-surface temperature record as well as with the overall patterns of response predicted by several general circulation models. We expect global mean temperatures in the decade 2036–46 to be 1–2.5 K warmer than in pre-industrial times under a ‘business as usual’ emission scenario. This range is relatively robust to errors in the models' climate sensitivity, rate of oceanic heat uptake or global response to sulphate aerosols as long as these errors are persistent over time. Substantial changes in the current balance of greenhouse warming and sulphate aerosol cooling would, however, increase the uncertainty. Unlike 50-year warming rates, the final equilibrium warming after the atmospheric composition stabilizes remains very uncertain, despite the evidence provided by the emerging signal.