Climate models: What they calculate, and why there are so many of them.
When we talk about climate forecasts, climate models are what make them possible. These are computer programs that simulate the interactions between the atmosphere, oceans, land surfaces, and ice. But what exactly do they calculate, why are there dozens of them, and what are the main differences between them?
What a climate model does
Imagine the Earth divided into a three-dimensional grid: cubes with sides about 100 kilometers long, extending dozens of layers high into the atmosphere and deep into the oceans. In each cube, the model calculates physical parameters such as temperature, pressure, humidity, wind, and ocean currents every 30 to 60 minutes, based on known physical equations. A single simulation spanning 100 years requires several million hours of computing time on supercomputers.
Why are there so many climate models?
Many critical processes occur on a scale smaller than a grid cell. Clouds typically measure 1 to 10 kilometers and therefore do not fit within a 100-kilometer grid. Every research center must therefore work with simplified equations—known as parameterizations—that approximate these processes. Different teams make different assumptions, such as how clouds behave when they warm up.
As a result, approximately 49 research groups worldwide run their own models and input the same scenarios. The result is an ensemble of independent simulations. Where the results agree, there is greater confidence in the findings. Where they differ, it indicates that the underlying physics is not yet fully understood.
Data source: IPCC AR6 WG1 (2021), CMIP5/CMIP6 model data | Graphic: glaciers.today
What is physically measured
Some parameters are measured directly and form the basis of all models. According to the IPCC AR6, the total anthropogenic radiative forcing is approximately 2.72 W/m² (±0.96) relative to 1750. The natural variation due to the 11-year solar cycle ranges from 0.1 to 0.25 W/m². The CO₂ concentration has risen from a pre-industrial level of 280 ppm to over 425 ppm today. This is a measurable physical fact. Equally measurable: Total Solar Irradiance (TSI) is approximately 1361 W/m², and its variation over the solar cycle is about 0.1 percent.
These figures are scientifically validated. How the climate system responds to these changes—that is, how strong the feedbacks will be—remains the central unanswered question.
Where the biggest differences lie
Clouds: A two- to three-fold uncertainty. The IPCC estimates climate sensitivity (warming resulting from a doubling of CO₂) to be likely between 2.5 and 4.0 degrees (best estimate: 3 degrees). However, the 27 CMIP6 models show a wider range of 1.8 to 5.6 degrees. The difference stems almost exclusively from how models calculate cloud feedback. British models (Met Office Hadley Centre, ECS ~5.5 °C) show strong feedbacks, while American models such as GFDL (~2.7 °C) or NASA GISS (~2.7 °C) yield significantly more moderate values. Researchers such as Nic Lewis and Judith Curry argue, based on empirical observational data, for even lower values of around 1.5 to 1.7 degrees.
Aerosol-cloud interaction: high uncertainty. Fine particulate matter influences cloud properties and cools the climate. According to IPCC AR6, the effect of aerosol-cloud interaction is −1.0 W/m², with a range of −1.7 to −0.3 W/m². This is physically significant: if the cooling effect observed so far is large, there is more hidden warming in the system. If it is small, the climate is less sensitive to CO₂.
Solar amplification mechanisms: inconsistently modeled. The models were not calibrated to accurately reproduce the observed temperature response to solar variations. Different models exhibit widely varying sensitivities. Furthermore, Solar Cycle 24 (2008–2019) was overestimated by the CMIP5 models by 0.1 to 0.5 W/m². Potential amplification mechanisms, such as UV-ozone coupling or solar influences on cloud formation, are inconsistently represented or not represented at all in the models.
Tipping points: not simulated. Most models do not capture abrupt transitions, such as a potential collapse of the Atlantic circulation. This means that there are currently neither reliable data for specific thresholds nor models capable of realistically simulating these dynamics.
What that means
Climate models are based on sound physics, including radiative balance, thermodynamics, and fluid dynamics. However, the key feedback mechanisms—particularly those involving clouds and aerosols—are not yet fully understood from a physical standpoint. The range of model results does not reflect differing opinions, but rather the actual state of our physical understanding, including its limitations.
Source links for the article:
IPCC AR6 WG1 Chapter 7 (2021) Chapter 7: Earth’s Energy Balance, Climate Feedbacks, and Climate Sensitivity https://www.ipcc.ch/report/ar6/wg1/chapter/chapter-7/
CMIP6 model data Official CMIP6 data archive of the World Climate Research Programme (WCRP): https://esgf-node.llnl.gov/projects/cmip6/
Kopp & Lean (2011) “A new, lower value of total solar irradiance: Evidence and climate significance.” Geophysical Research Letters, 38, L01706. https://doi.org/10.1029/2010GL045777
Lewis & Curry (2018) “The impact of recent forcing and ocean heat uptake data on estimates of climate sensitivity.” Journal of Climate, 31(15), 6051–6071. https://journals.ametsoc.org/view/journals/clim/31/15/jcli-d-17-0667.1.xml
NOAA Mauna Loa CO₂ Measurements (2025) NOAA Global Monitoring Laboratory, Trends in Atmospheric CO₂: https://gml.noaa.gov/ccgg/trends/
Stevens, B. & Bony, S. (2013) “What Are Climate Models Missing?” Science, 340(6136), 1053–1054. Key point: The inadequate representation of clouds and moist convection is the main limitation of today’s climate models. Additional model complexity has multiplied rather than reduced uncertainties. https://www.science.org/doi/10.1126/science.1237554
Zelinka, M. D. et al. (2020) “Causes of Higher Climate Sensitivity in CMIP6 Models.” Geophysical Research Letters, 47(1). Key finding: Cloud feedbacks are the main driver of differences in climate sensitivity between models. The variability has increased in CMIP6 compared to CMIP5. https://doi.org/10.1029/2019GL085782
Morrison, H. et al. (2020) “Confronting the Challenge of Modeling Cloud and Precipitation Microphysics.” Journal of Advances in Modeling Earth Systems, 12(8). Key message: The complexity of microphysics schemes has outpaced the current state of cloud physics and the possibility of observational verification. https://doi.org/10.1029/2019MS001689
Randall, D. A. et al. (2003) “Breaking the Cloud Parameterization Deadlock.” Bulletin of the American Meteorological Society, 84(11), 1547–1564. Key point: The development of cloud parameterization has reached an impasse. Fundamentally new approaches are needed. https://doi.org/10.1175/BAMS-84-11-1547
Note
The study by Lewis and Curry (2018) was published in the Journal of Climate (American Meteorological Society) and underwent a standard peer-review process. However, their method—known as the energy-balance approach—has been criticized by several research groups. The main objection concerns so-called pattern effects: the spatial distribution of past warming differs from what would be expected in equilibrium with a doubling of CO₂. Critics such as Marvel et al. (2018) and Dessler et al. (2018) argue that this difference leads energy balance methods to systematically underestimate climate sensitivity. Lewis and Curry dispute this assessment and point out that the observed pattern effects fall within the range of natural variability. The question has not been conclusively resolved from a physical standpoint.
