Climate Models

Climate models

Climate models, also known as general circulation models or GCMs, use mathematical equations to characterize how energy and matter interact in different parts of the ocean, atmosphere, land. They are closely related to meteorological models used for weather-forecast.

Approximately 100 global circulations models participate in the Coupled Model Intercomparison Project phase 6 (CMIP6). To limit the resources required for the studies, especially in terms of time needed for data access and computing power, ClimateVision uses a subset of models.

Models for atmospheric indicators

To ensure consistency and comparability, the same climate models are used for all atmospheric indicators. In ClimateVision V1.1.0, these indicators include mean temperature, maximum temperature, minimum temperature, mean precipitation, maximum precipitation, maximum wind speed, sea level rise, and maximum significant wave height.

Models were selected based on 3 criteria:

• Data availability
• Equilibrium climate sensitivity
• Model independence

Due to the recent publication of model data, studies on their regional performance are still scarce and usually only consider a subset of the models currently available. As a result, this criterion was not considered in ClimateVision V1.1.0 but will be added in later versions.

Data availability

The models were selected primarily based on data availability, as some projections are unavailable. Climate models were prioritized based on the range of climate parameters they provide. To maximize the number of models available, only three emissions scenarios are considered in the current version:

• SSP1-2.6: low emission scenario, representative of an emission trajectory that keeps global warming below 2°C
• SSP2-4.5: intermediate emissions scenario, close to current emission trajectory
• SSP5-8.5: very high emissions scenario that can be used as a worst case

Other scenarios not included are: SSP1-1.9, SSP3-7.0 and SSP4-6.0. Both SSP1-1.9 and SSP4-6.0 are "tier 2" (or low priority) SSPⁱ. And SSP3-7.0 is a high emissions scenario intermediary between SSP2-4.5 et SSP5-8.5.

Equilibrium climate sensitivity

Equilibrium climate sensitivity (ECS) is the long-term warming that would occur if the concentration of CO₂ in the atmosphere were to double. According to the IPCC 6th Assessment Report the best estimate for equilibrium climate sensitivity is 3°C with a likely range of 2.5 to 4°C and a very likely range of 2 to 5°Cⁱⁱ.

Several GCMs of CMIP6 have ECS that fall outside this range either below (low-likelihood, low warming) or above (low-likelihood, high warming). Because one of those models can significantly alter the results on a small ensemble, an increasingly common practice in the scientific literature is to downweigh models with ECS values outside the assessed range or simply exclude them from ensembles^iii,iv.

In practice, this criterion leads to the exclusion of the Russian model (INM-CM5-0) from our list of candidates^v. The latter is left on the list and is used until new models are added. It will be removed shortly.

Model's independence

Even when they are exploited by different institutions, climate models are not necessarily independent of each other: they can use similar assumptions or common modules. The use of several closely related models in an ensemble could therefore result in an overweighting of the corresponding trajectory.

The studies on the correlations between models are still significantly less complete than for the CMIP5 generation^vi. One publication investigating 33 CMIP6 model suggests the following dependency tree^vii:

Models branching further to the left are more dependent, and models branching further to the right are more independent. An estimate of climate natural variability is indicated using gray shading, models that have a distance similar to this value (for instance CanESM5 and CanESM5-CanOE) are statistically indistinguishable. Models branching after the dotted line are reasonably independent.

Models set

The ideal models' ensemble should meet the following conditions:

1. It includes enough models (at least 10),
2. It provides projections for all desired variables,
3. It provides projections for all scenarios,
4. It does not include models with ECS outside the IPCC's very likely range,
5. It prioritizes models that perform well on the study area, and ideally on the whole planet,
6. If possible with respect to the previous conditions, it does not include closely related models.

The ensemble of models that best fit these objectives is the following:

*not used in V1.1.0 but available