L'hypercubeFrom scenarios to climate projections
Studying future climate does not mean forecasting the weather on a specific day in 2080. Projections seek to determine how the statistical properties of climate may change over several decades: average temperature, very hot days, precipitation, droughts, winds, and extreme events.
They are essential for assessing the vulnerability of buildings, infrastructure, and urban spaces and for adapting projects to conditions unlike those observed today.
Climate projection and weather forecast
A weather forecast starts from the current state of the atmosphere and predicts its short-term evolution. It seeks to describe a specific chronology and can be refined down to the local scale.
A climate projection instead starts from assumptions about the long-term evolution of the climate system: greenhouse-gas emissions, atmospheric concentrations, land use, and other forcings. Over several decades, the exact sequence of weather situations is neither predictable nor the information being sought.
A projection first describes large-scale balances and trends at the global or continental scale. It can then be regionalised to study their effects on a territory. Even at high spatial resolution, its results must be interpreted statistically: a simulated day in 2080 is not the forecast for that date, but one possible occurrence within the climate being considered.
Climate scenarios
Projections rely on assumptions about how societies, emissions, and climate forcings may evolve. Two families of terms are commonly encountered: RCPs and SSPs.
RCPs — concentration pathways
Representative Concentration Pathways (RCPs), notably used by the CMIP5 generation of simulations, describe greenhouse-gas concentration pathways leading to different radiative forcing levels in 2100. Examples include RCP2.6, RCP4.5, and RCP8.5.
The number represents neither a temperature increase nor a probability. It approximately indicates the radiative forcing reached in 2100, expressed in W/m². An RCP therefore describes the physical constraint applied to the climate, but not the exact socioeconomic development leading to it.
SSPs — socioeconomic pathways
Shared Socioeconomic Pathways (SSPs) describe possible societal developments involving population, urbanisation, economic development, technology, energy use, international cooperation, and the capacity to adapt.
In the CMIP6 generation, these socioeconomic pathways are associated with forcing levels. In SSP2-4.5, for example, SSP2 identifies the socioeconomic pathway and 4.5 indicates radiative forcing of approximately 4.5 W/m² in 2100.
SSP2-4.5 is therefore not simply a new name for RCP4.5: their forcing levels are similar, but their socioeconomic assumptions, emissions, and some other forcings may differ.
These scenarios are not forecasts assigned a precise date and probability. They explore several internally consistent futures and the climate’s response to different pathways.
Source: Betting on negative emissions
From scenarios to regionalised projections
The emissions, concentration, and land-use pathways associated with the scenarios are converted into forcings applied to climate models. A Global Climate Model (GCM) then simulates the response of the Earth system over several decades: atmosphere, oceans, land surfaces, cryosphere, and interactions between these components.
GCMs represent the major climate balances at the planetary scale, but their grid does not describe relief, coastlines, or some regional phenomena in detail. Their results can therefore drive a Regional Climate Model (RCM), which refines the simulation over a limited domain.
This chain resembles the nesting of models used in weather forecasting, but its purpose is different: it produces long-term climate statistics and trends, not the chronology of weather on a specific date.
Source: IPCC
CORDEX — coordinating regional projections
CORDEX , the Coordinated Regional Climate Downscaling Experiment, is a World Climate Research Programme initiative. It is not a single climate model, but a shared international framework for producing, documenting, and comparing regionalised climate projections.
The world is divided into large CORDEX domains, including Europe, Africa, the Mediterranean, and the Arctic. For Europe, EURO-CORDEX notably includes simulations at a resolution of 0.11°, or approximately 12 km, identified by the code EUR-11. Simulations at 0.44°, known as EUR-44, are also available.
An ensemble of GCM–RCM combinations
A CORDEX simulation generally combines:
- an emissions or climate-forcing scenario;
- a GCM, which simulates large-scale climate change;
- an RCM, driven at the boundaries of its domain by the GCM;
- a particular realisation and simulation period.
One RCM may be driven by several GCMs, and one GCM may be downscaled by several RCMs. These combinations form a GCM–RCM matrix, which is sparse: not every regional model has downscaled every global model.
This diversity makes it possible to examine both a central tendency and the spread of model results. A single projection cannot represent the full range of possible futures.
Periods, variables, and time steps
CORDEX datasets include historical simulations, future projections and, depending on the experiment, evaluation runs driven by a reanalysis. They provide numerous variables, including minimum, mean and maximum temperature, precipitation, humidity, wind, and solar radiation.
Not every combination of variable, model, and time step is available. Daily data are common, while sub-daily availability depends on the variable and simulation. Hourly building simulations may therefore require temporal disaggregation or interpolation.
From the regional grid to the project site
A 12 km resolution is fine for a climate projection, but it does not directly describe the microclimate of a street or building. Outputs may also retain biases relative to observations. Depending on the application, bias correction, additional downscaling, or temporal reconstruction may be required.
Files are primarily distributed in NetCDF format through the federated ESGF infrastructure. Selecting data requires specifying the domain, scenario, GCM, RCM, variable, period, and time step. The terms of use must also be checked for each dataset.
CORDEX is therefore mainly intended for specialist engineering consultancies, researchers, and users able to select, process, and compare large climate datasets. DRIAS and the prepared products presented below are generally more accessible for an initial territorial assessment or direct application in France.
Source: Ouzeau et al.
DRIAS — Future climate
The DRIAS — Future climate portal distributes France’s reference regionalised climate projections for adaptation. It is operated by Météo-France with the French climate-science community.
Explore indicators
The Discovery area displays geolocated indicators—temperature, precipitation, hot days, or frost—depending on the available datasets.
Download data
The Data and Products area provides variables and indicators for advanced processing. Reference products notably use a grid of approximately 8 km for mainland France. Recent products related to Explore2 and the TRACC supersede older datasets still cited in some historical documentation.
TRACC — a reference pathway for adaptation
The TRACC (Trajectoire de réchauffement de référence pour l’adaptation au changement climatique, or reference warming trajectory for climate-change adaptation) defines the climate for which France must prepare. Relative to the pre-industrial period, it sets the following levels for mainland France and Corsica:
- +2°C in 2030;
- +2.7°C in 2050;
- +4°C in 2100.
The 2100 level corresponds to approximately +3°C globally. The TRACC is neither a new model nor an additional scenario: it defines reference warming levels that regionalised projections translate into indicators for individual territories.
The trajectory was incorporated into the French Environmental Code by the decree of 23 January 2026. See the French Ministry for Ecological Transition’s pages presenting the TRACC and its regulatory incorporation .
From a global scenario to a project
From TRACC levels to hourly weather series
The TRACC defines reference warming levels, but a thermal simulation requires a complete chronology of weather variables. To meet this need, ADEME , Météo-France, CSTB, and DHUP have produced prospective climate files intended notably for building and local-authority studies.
These files translate the three TRACC levels into hourly series for the eight stations associated with the reference climate zones of metropolitan France:
- France at +2°C , the 2030 reference level;
- France at +2.7°C , the 2050 reference level;
- France at +4°C , the 2100 reference level.
Each dataset provides a full year at an hourly time step with 17 meteorological and astronomical parameters. The files are constructed from high-resolution CNRM-AROME simulations corrected against observations. Variables that are not simulated directly are reconstructed using empirical methods.
Two families of files are provided for each warming level:
- a future Typical Meteorological Year, representing average conditions at the relevant TRACC level;
- a typical year containing a characteristic heatwave episode: France +2°C , France +2.7°C , and France +4°C .
The selected heat event is close to the median heatwave among the 5% most severe events in the TRACC-2023 ensemble. It is then inserted into the typical year while retaining its position in the calendar.
The detailed technical note describes their construction method and limitations.
Constructing future weather files
Climate projections often require transformation before they can be used in thermal or energy simulation software.
Future typical year
A future TMY represents average climate around a given horizon or warming level. It can be used to study heating, cooling, comfort, natural ventilation, and energy consumption. Like a current-climate TMY, it tends to smooth out rare events. The ADEME files presented above are a concrete application of this approach to the TRACC levels.
Specific future years
Climate simulations also produce continuous series in which particularly hot years, dry summers, or severe heatwaves can be identified. Selection must account for models, scenarios, years, and intensity levels. Defining a target risk level or quantile is often more robust than automatically selecting the “worst year”.
What about future extreme events?
Adaptation cannot rely on average climate alone. Heatwaves can be detected in climate projections and classified by duration, intensity, and severity to produce hourly files. This method is detailed on the Heatwaves page.