Climate Models and Projections in Climate Change Science

Climate models are sophisticated numerical representations of the Earth’s atmosphere, oceans, land surface and cryosphere, which together comprise the Earth system. They simulate interactions among physical, chemical and biological processes to project changes in temperature, precipitation, sea level and extreme events under different greenhouse-gas scenarios. Multi-model ensembles, such as those coordinated under successive Coupled Model Intercomparison Projects, underpin most projections by sampling structural and parametric uncertainties. Key sources of uncertainty arise from internal climate variability, discrepancies in the representation of feedbacks (for example, cloud and carbon-cycle feedbacks) and divergent future emissions trajectories. To narrow these uncertainties, researchers have developed methods including performance-based model weighting, statistical emulators, downscaling and bias adjustment. Recent advances harness machine learning to transfer knowledge from observations to model outputs, while statistical frameworks such as Bayesian model averaging integrate multiple lines of evidence to produce probability-based projections. These developments enhance confidence in regional projections, informing adaptation planning, risk assessment and mitigation strategies worldwide.

Research from Nature Portfolio

A Bayesian framework has been applied to multi-model ensembles to address biases arising from models with very high climate sensitivity. By assigning probability weights based on multiple lines of evidence, this approach yields refined projections of global mean temperature increases under both low and high emissions pathways. The weighted ensemble produces lower central estimates and narrower uncertainty ranges compared with simple multi-model means, while retaining the capacity to consider less probable outcomes. This methodology represents a step forward in combining model performance and expert judgement to generate probabilistic climate scenarios that better reflect the breadth of available scientific evidence.

Climate Models and Projections in Climate Change Science publication trend

The graph below shows the total number of articles in climate models and projections in climate change science across all publications each year (not limited to Nature Index journals).

Technical terms

Earth system model: A comprehensive numerical simulation of atmosphere, ocean, land and cryosphere processes and their interactions.

Multi-model ensemble: A collection of climate model simulations used together to characterise uncertainty in projections.

Climate sensitivity: The equilibrium change in global mean surface temperature following a doubling of atmospheric CO₂ concentration.

Bayesian Model Averaging: A statistical technique that combines multiple model projections into a single probabilistic estimate by assigning weights based on evidence.

Shared Socioeconomic Pathways (SSPs): Standardised scenarios describing future greenhouse-gas emissions, land use and socio-economic developments for climate projections.

References

1. Transferring climate change physical knowledge. Proceedings of the National Academy of Sciences of the United States of America (2025).
2. Bayesian weighting of climate models based on climate sensitivity. Communications Earth & Environment (2023).
3. Effects of multi-observations uncertainty and models similarity on climate change projections. npj Climate and Atmospheric Science (2023).
4. User-tailored sub-selection of climate model ensemble members for impact studies. The Science of The Total Environment (2024).