Climate Models Explained: From CMIP6 to Risk Assessment

Climate models are the foundation of every climate risk projection, from corporate TCFD reports to national adaptation plans. Over 30 global climate models contributed to the CMIP6 dataset that underpins the IPCC AR6 assessment, producing projections of temperature, precipitation, and extreme weather out to 2100 and beyond. Climate risk modeling takes these projections and translates them into financial exposure estimates for specific assets and portfolios.

Yet for most risk practitioners, climate models remain a black box. What goes in, what comes out, and how reliable are the projections? Understanding how climate models work is essential for interpreting the risk assessments built on top of them.

Below, you will learn what climate models are, how the CMIP6 framework organizes them, how model output becomes usable risk data, and what limitations to keep in mind.

What Are Climate Models?

Climate models are mathematical simulations of Earth’s climate system. They represent the interactions between the atmosphere, oceans, land surface, ice sheets, and biosphere using physics-based equations. By running these equations forward in time under different emissions assumptions, climate models project how temperature, precipitation, wind, and other variables will change in the future.

The most comprehensive climate models are called General Circulation Models (GCMs) or Earth System Models (ESMs). These divide Earth’s surface into a three-dimensional grid and calculate how energy, moisture, and momentum move between grid cells at each time step. A typical GCM runs at 100-250 km horizontal resolution, meaning each grid cell covers a region roughly the size of a small country.

Climate models are not weather forecasts. Weather models predict specific conditions days ahead. Climate models project statistical patterns (averages, extremes, trends) decades into the future. A climate model cannot tell you whether it will rain on a specific day in 2045, but it can project whether that region will experience more intense rainfall events on average.

How Many Climate Models Exist?

Over 100 climate models have been developed by research institutions worldwide. The most widely used are organized through the Coupled Model Intercomparison Project (CMIP), currently in its sixth phase (CMIP6). CMIP6 includes output from over 30 modeling centers across more than 50 countries.

Major climate models contributing to CMIP6 include:

No single climate model is considered “best.” Each has strengths in different regions and for different variables. Risk assessments typically use multi-model ensembles (averaging across many models) to account for structural differences between models and reduce the influence of any single model’s biases.

What Is CMIP6?

CMIP6 (Coupled Model Intercomparison Project Phase 6) is the current generation of coordinated climate model experiments. Run under standardized protocols, CMIP6 ensures that different modeling groups use the same emissions scenarios, the same historical forcing data, and the same experimental design, making their outputs directly comparable.

CMIP6 is the data backbone of the IPCC AR6 assessment. It pairs climate models with Shared Socioeconomic Pathways (SSPs) to explore how different development trajectories affect climate outcomes. The most commonly used scenario pairs for risk assessment are SSP2-4.5 (moderate emissions, approximately 2.7 degrees C warming by 2100) and SSP5-8.5 (high emissions, approximately 4.4 degrees C warming by 2100).

CMIP6 provides five key climate variables used in physical risk assessment: daily mean temperature (tas), daily maximum temperature (tasmax), daily minimum temperature (tasmin), precipitation (pr), and surface wind speed (sfcWind). These variables drive the analysis of hazards from heat waves and wildfires to drought and flooding.

From Climate Models to Risk Assessment

Raw climate model output cannot be used directly for location-level risk assessment. The data requires several processing steps before it becomes actionable.

Step 1: Downscaling. GCMs run at 100-250 km resolution, too coarse for site-level analysis. NASA’s NEX-GDDP-CMIP6 dataset downscales CMIP6 output to 25 km resolution (0.25 degrees) using bias-correction and spatial disaggregation. The result is daily climate data for every 25 km grid cell on Earth, from 1950 through 2100.

Step 2: Variable extraction. For a given location, the system extracts the relevant grid cell data across all available models for the selected scenarios. Five variables are pulled: temperature (mean, max, min), precipitation, and wind speed.

Step 3: Multi-model averaging. Rather than relying on a single model, risk assessments average across the ensemble of available models. NEX-GDDP-CMIP6 includes output from over 30 CMIP6 models, and averaging across them produces more robust projections than any individual model.

Step 4: Hazard calculation. The averaged climate data feeds into hazard-specific algorithms. For example, heat wave days are calculated by counting days where maximum temperature exceeds historical thresholds. Drought months are estimated using precipitation ratios as a proxy for the Standardized Precipitation Index. Each of the 12 physical hazards has its own calculation methodology drawing from climate risk data sources.

Step 5: Risk scoring. Hazard values are translated into risk ratings (Low through Extreme) using calibrated thresholds. These ratings feed into a composite risk score that summarizes overall physical risk exposure at each location across baseline, 2030, 2040, and 2050 time horizons.

Types of Climate Models

Climate models range from simple energy balance calculations to full Earth system simulations. Understanding the types helps clarify what different models can and cannot do.

Energy Balance Models (EBMs). The simplest type. EBMs calculate global average temperature based on incoming solar radiation, outgoing infrared radiation, and greenhouse gas concentrations. Useful for understanding basic climate sensitivity but too simple for regional projections.

General Circulation Models (GCMs). The workhorses of climate science. GCMs simulate the three-dimensional circulation of the atmosphere and ocean on a global grid. They resolve large-scale weather patterns, monsoons, jet streams, and ocean currents. CMIP6 models are predominantly GCMs.

Earth System Models (ESMs). GCMs extended with biogeochemical cycles: carbon cycling, vegetation dynamics, atmospheric chemistry, and ice sheet processes. ESMs can simulate feedbacks between climate change and the carbon cycle (for example, how warming affects forest carbon uptake).

Regional Climate Models (RCMs). Higher-resolution models that cover a limited geographic area. RCMs take boundary conditions from GCMs and add regional detail at 10-50 km resolution. Useful for capturing local topography effects on precipitation and temperature.

Limitations of Climate Models

Climate models are powerful tools, but they have known limitations that users should understand when interpreting risk assessments.

Resolution constraints. Even downscaled datasets like NEX-GDDP-CMIP6 (25 km) cannot capture microclimates, urban heat islands, or individual building exposure. Site-level factors like elevation, proximity to water, and local land cover add variability that models smooth over.

Model spread. Different climate models produce different projections for the same scenario. Temperature projections show relatively good agreement across models, but precipitation projections diverge significantly, especially at regional scales. Multi-model averaging reduces but does not eliminate this uncertainty.

Missing processes. Current climate models do not fully represent tipping points (ice sheet collapse, permafrost methane release, Amazon dieback) or compound events (simultaneous heat wave and drought). Some variables like relative humidity show inconsistent bias patterns across models and are excluded from certain assessments.

Scenario dependence. All projections depend on which emissions pathway the world follows. Climate models do not predict which scenario will occur; they show what happens if a particular pathway is followed. Using multiple scenarios (such as SSP2-4.5 and SSP5-8.5) brackets the range of plausible outcomes.

Climate Models and Scenario Analysis

For organizations conducting TCFD-aligned physical climate risk assessment, climate models provide the quantitative foundation. The choice of which models and scenarios to use directly affects the results.

Most corporate risk assessments use the CMIP6 multi-model ensemble under at least two scenarios: a moderate pathway (SSP2-4.5) representing current policy trajectories, and a high-emissions pathway (SSP5-8.5) representing limited climate action. Comparing results across scenarios helps organizations understand the range of exposure they may face.

Time horizons matter as well. Climate models show relatively modest divergence between scenarios out to 2030 (near-term warming is largely locked in by past emissions). By 2050, scenario differences become more pronounced, and by 2100, the gap between moderate and high-emissions pathways is substantial.

Frequently Asked Questions

What are different climate models?

Climate models range from simple Energy Balance Models that calculate global average temperature, to General Circulation Models (GCMs) that simulate atmospheric and ocean circulation on a 3D grid, to Earth System Models (ESMs) that add carbon cycling and vegetation dynamics. Regional Climate Models provide higher-resolution projections for specific areas. Over 30 GCMs and ESMs contribute to the current CMIP6 dataset.

How many climate models exist?

Over 100 climate models have been developed globally. The CMIP6 framework, which underpins the IPCC AR6 assessment, includes output from more than 30 major modeling centers across 50+ countries. These models vary in resolution, complexity, and which physical processes they simulate, which is why risk assessments use multi-model averages rather than relying on any single model.

What is CMIP6 and why does it matter?

CMIP6 (Coupled Model Intercomparison Project Phase 6) is the standardized framework through which global climate modeling centers coordinate their experiments. It ensures all models use the same emissions scenarios (SSPs), historical data, and experimental protocols. CMIP6 data is the foundation of the IPCC AR6 assessment and most corporate climate risk assessments.

How accurate are climate models?

Climate models have demonstrated strong accuracy for global temperature projections. Models from the 1990s correctly predicted the warming observed since then. Regional accuracy varies: temperature projections are generally more reliable than precipitation projections. Model spread (disagreement between models) is a key source of uncertainty, which is why risk assessments use multi-model ensembles.

What climate data do climate models produce?

Climate models produce daily and monthly projections of key variables including mean temperature, maximum temperature, minimum temperature, precipitation, and surface wind speed. These five variables are the primary inputs for physical risk assessment, driving calculations of hazards from heat waves and drought to flooding and wildfire risk.

Conclusion

Climate models are the engine behind every forward-looking climate risk assessment. The CMIP6 framework standardizes how these models run, NASA’s NEX-GDDP dataset makes their output accessible at 25 km resolution, and hazard algorithms translate the raw data into actionable risk ratings. Several climate data APIs now provide access to both raw CMIP6 output and processed risk scores, depending on whether you need research-grade data or ready-to-use assessments. Understanding how climate models work, their types, and their limitations helps you interpret the projections that drive physical risk assessments, scenario analysis, and adaptation planning.