Climate modeling works by translating the physical laws governing Earth’s atmosphere, oceans, land surface, and ice into mathematical equations that computers can solve. Scientists divide the planet into a three-dimensional grid, apply these equations to each grid cell, and run the calculations forward in time to simulate how the climate evolves. The sections below unpack each layer of that process, from the raw data that feeds a model to the policy decisions that depend on its output.
What data goes into a climate model?
Climate models are built on observational data describing the physical state of the Earth system. This includes atmospheric temperature, humidity, wind speed, ocean temperatures, sea ice extent, land surface properties, and greenhouse gas concentrations. Without accurate input data, even the most sophisticated model will produce unreliable results.
The sources of this data are remarkably diverse. Satellite instruments measure surface temperatures and cloud cover from orbit. Weather stations and ocean buoys provide ground-level readings. Ice cores and sediment records supply historical baselines stretching back hundreds of thousands of years. Radiosondes attached to weather balloons capture vertical profiles of the atmosphere several times a day at stations around the world.
Beyond physical measurements, models also require boundary conditions. These are factors treated as fixed or slowly changing inputs, such as solar irradiance, volcanic aerosol loading, and land-use patterns. Getting these inputs right is foundational to climate research, because errors in initial conditions can propagate and amplify across a long simulation run.
What are the main types of climate models used in research?
The main types of climate models range from simple energy balance models to fully coupled Earth system models. Each type trades computational complexity for either speed or physical realism, and researchers choose the appropriate model based on their specific research question.
- Energy balance models (EBMs): The simplest category, these calculate how incoming solar energy and outgoing heat radiation balance at the planetary or regional scale. They are fast to run and useful for understanding broad climate sensitivity.
- Atmospheric general circulation models (AGCMs): These simulate the three-dimensional dynamics of the atmosphere, including wind patterns, precipitation, and temperature gradients across the globe.
- Coupled atmosphere-ocean models (AOGCMs): By linking atmospheric and oceanic circulation, these models capture the heat exchange between air and sea that drives much of Earth’s long-term climate variability.
- Earth system models (ESMs): The most comprehensive type, ESMs incorporate biogeochemical cycles such as the carbon cycle, vegetation dynamics, and sea ice physics alongside atmosphere-ocean coupling. These are the workhorses of modern climate change research.
Numerical weather prediction models share many of the same principles but operate on shorter timescales, typically days to weeks, whereas climate models project trends across decades or centuries. The distinction matters because climate simulation is less about predicting exact weather on a given day and more about capturing statistical patterns over time.
How do supercomputers power climate simulations?
Supercomputers power climate simulations by performing trillions of calculations per second to solve the equations governing each grid cell in a model at every time step. A single high-resolution global climate simulation can require millions of processor hours, making high-performance computing infrastructure essential to modern climate research.
A typical Earth system model divides the globe into a horizontal grid with cells ranging from roughly 100 kilometers across in older models down to 10 kilometers or less in cutting-edge high-resolution versions. The atmosphere is also divided vertically into dozens of layers. At each time step, which might represent 20 to 30 minutes of simulated time, the model calculates the state of every variable in every cell and exchanges information between neighboring cells.
The computational demand grows rapidly as resolution increases. Doubling the horizontal resolution of a model increases the number of grid cells fourfold and also requires shorter time steps, multiplying the total computation by a factor of roughly eight or more. This is why the world’s most advanced climate modeling centers operate some of the largest supercomputers on the planet, and why international collaboration between research institutions is so important for sharing both infrastructure and results.
How do scientists validate climate model predictions?
Scientists validate climate model predictions by comparing model output against historical observational records. If a model can accurately reproduce past climate conditions it was not explicitly tuned to match, researchers gain confidence that it captures the underlying physical mechanisms correctly and can be trusted for future projections.
Several validation strategies are commonly used in climate research:
- Hindcasting: Running a model over a historical period and comparing its output to observed temperature records, precipitation patterns, and sea level data.
- Paleoclimate benchmarking: Testing whether a model can reproduce well-documented past climate states, such as the Last Glacial Maximum or the mid-Holocene warm period, using proxy data from ice cores and pollen records.
- Process-level evaluation: Checking whether specific physical processes, such as the El Niño-Southern Oscillation or monsoon dynamics, behave realistically within the model.
- Multi-model ensembles: Comparing results across many different models. When independent models built by different teams converge on similar projections, confidence in those projections increases significantly.
No single validation test is sufficient on its own. Scientists treat model evaluation as an ongoing process, continuously refining parameterizations as new observational data becomes available and as understanding of physical processes deepens.
What are the biggest limitations of current climate models?
The biggest limitations of current climate models include grid resolution constraints, incomplete representation of small-scale processes, and uncertainties in how future human behavior will shape greenhouse gas emissions. These limitations do not invalidate climate predictions, but they define the boundaries of what models can and cannot tell us with confidence.
Cloud feedbacks remain one of the most significant sources of uncertainty in climate prediction. Clouds can both cool the surface by reflecting sunlight and warm it by trapping outgoing heat, and the net effect depends on cloud type, altitude, and regional context. Representing cloud formation accurately requires resolving processes at scales far smaller than current global model grids allow, so modelers use simplified parameterizations that introduce uncertainty into projections of climate sensitivity.
Other notable limitations include the coarse representation of ocean eddies, the difficulty of modeling ice sheet dynamics at the scale relevant to sea level rise, and the challenge of capturing land-atmosphere feedbacks such as permafrost thaw and its release of stored carbon. Socioeconomic scenarios also introduce uncertainty that is fundamentally non-physical: models cannot predict whether global emissions will follow a high, medium, or low trajectory without assumptions about policy, technology, and human choices.
How are climate models used to inform policy decisions?
Climate models inform policy decisions by translating physical science into actionable projections about future temperature rise, sea level change, extreme weather frequency, and regional climate impacts. Policymakers use these projections to design mitigation strategies, set emissions targets, and plan adaptation measures for vulnerable communities and infrastructure.
The most influential pathway from climate modeling to policy runs through the Intergovernmental Panel on Climate Change (IPCC), which synthesizes results from dozens of modeling centers worldwide into assessment reports used by governments and international negotiators. The standardized emission scenarios used in these assessments, known as Shared Socioeconomic Pathways, allow policymakers to compare projected outcomes under different levels of climate action.
At the national and regional level, downscaled climate projections help governments make concrete decisions about flood defenses, agricultural planning, water resource management, and urban heat island mitigation. The connection between global climate simulation and local decision-making depends heavily on the capacity of national research institutions to translate and apply model outputs within their specific geographic and socioeconomic contexts.
How WAITRO supports climate research capacity worldwide
Translating climate modeling into meaningful policy action requires research institutions that have the technical expertise, infrastructure, and international connections to do that work effectively. This is where we at WAITRO play a direct role. Through our global network of research and technology organizations, we help institutions build the capacity needed to engage with climate research at a serious level.
Specifically, we support our members and partners by:
- Connecting research organizations across regions to share methodologies, data, and modeling expertise through cross-border partnerships
- Providing institutional capacity building programs that strengthen the technical and organizational foundations of research institutions in emerging and developing economies
- Facilitating collaboration between national research bodies, universities, and industry partners to bring climate science closer to practical application
- Supporting innovation ecosystems aligned with the UN Sustainable Development Goals, including those directly tied to climate action and sustainable development
Whether your organization is looking to deepen its engagement with climate modeling, build stronger research partnerships, or find pathways to translate scientific findings into policy impact, we are here to help. Explore WAITRO membership to learn how joining our network can amplify your organization’s contribution to global climate research.
