12.1 Types of climate models and their development
5 min read•july 22, 2024
Climate models are essential tools for understanding and predicting Earth's complex climate system. From simple Energy Balance Models to sophisticated Earth System Models, these digital representations simulate various climate processes and interactions.
As climate science advances, models evolve to incorporate new knowledge, higher resolutions, and additional components. This progression allows for more accurate simulations of climate dynamics, from global trends to regional impacts, enhancing our ability to forecast future climate scenarios.
Types of Climate Models
Types of climate models
Top images from around the web for Types of climate models
Why trust climate models? It’s a matter of simple science View original
Focus on the balance between incoming solar radiation and outgoing terrestrial radiation
Treat the Earth as a single point or a small number of latitude bands
Do not explicitly represent atmospheric or ocean dynamics (winds, currents)
Useful for understanding basic climate processes and feedbacks (greenhouse effect, albedo)
More complex than EBMs but simpler than General Circulation Models
Include some representation of atmospheric and ocean dynamics
Often have reduced spatial resolution or simplified physics compared to GCMs
Used for long-term climate simulations (thousands to millions of years) and understanding climate feedbacks (, ice sheets)
Examples: UVic ESCM, CLIMBER
General Circulation Models (GCMs)
Most comprehensive and complex type of climate model
Divide the Earth's atmosphere and oceans into a 3D grid
Solve mathematical equations representing physical processes within each grid cell (fluid dynamics, thermodynamics, radiative transfer)
Coupled Atmosphere-Ocean GCMs (AOGCMs) simulate interactions between the atmosphere and oceans
Earth System Models (ESMs) build upon AOGCMs by including additional components such as the carbon cycle, vegetation (dynamic vegetation models), and ice sheets
Examples: CESM, HadGEM, IPSL-CM
Development of climate models
Incorporation of new scientific understanding
As knowledge of climate processes advances, models are updated to reflect this understanding
Examples include improvements in representing cloud processes (microphysics, convection), aerosols (direct and indirect effects), and the carbon cycle (land and ocean carbon uptake)
Increased spatial resolution
Over time, computational power has increased, allowing for finer grid resolutions in models
Higher resolution enables better representation of small-scale processes (topography, land-sea contrast) and extreme events (tropical cyclones, heavy precipitation)
Typical resolutions have improved from ~500 km in the 1970s to ~25 km today
Addition of new model components
Climate models have evolved to include more components of the Earth system
Examples include the incorporation of dynamic vegetation (land use change, carbon uptake), ice sheets (sea level rise), and atmospheric chemistry (ozone, methane)
Enables more comprehensive understanding of climate feedbacks and impacts
Improved parameterizations
Parameterizations are simplified representations of complex processes that occur at scales smaller than the model grid (subgrid-scale processes)
As understanding of these processes improves, parameterizations are updated to better capture their effects
Examples include parameterizations for cloud microphysics, boundary layer turbulence, and radiative transfer
Initiatives such as the compare results from different climate models
These projects help identify strengths and weaknesses of individual models and guide future model development
CMIP provides a framework for coordinated experiments and analysis, enabling more robust assessments of climate change
Parameterization in climate modeling
is necessary because some processes occur at scales smaller than the model grid
Examples include cloud formation (microphysics), turbulence (boundary layer), and small-scale convection (cumulus convection)
These processes cannot be explicitly resolved by the model's grid
Parameterizations use simplified mathematical equations to represent the effects of these processes on the larger-scale climate
For example, cloud parameterizations estimate the fraction of a grid cell covered by clouds and their radiative properties based on variables like humidity and temperature
The choice of parameterization schemes can significantly impact model results
Different parameterizations may have different sensitivities to changes in climate forcing (CO2, aerosols)
Parameterizations can introduce uncertainties and biases into model simulations
Different models may use different parameterizations for the same process, contributing to model uncertainty
For example, models may use different schemes for deep convection (Arakawa-Schubert, Kain-Fritsch) or boundary layer turbulence (K-profile, Mellor-Yamada)
Improving parameterizations is an ongoing area of research in climate modeling
Advances in understanding of physical processes and availability of high-resolution observations (field campaigns, satellites) help refine parameterizations
New approaches such as machine learning and stochastic parameterizations are being explored to improve subgrid-scale representations
Validation of climate models
involves comparing model output to observations to assess model performance
This can be done using historical climate data, such as temperature and precipitation records (instrumental records, proxy data)
Validation can also use data from satellite observations (radiation budgets, sea surface temperatures) and field campaigns (atmospheric profiles, ocean measurements)
Evaluation helps identify strengths and weaknesses of models
Areas where models perform well (large-scale patterns, long-term trends) increase confidence in their projections
Areas where models perform poorly (regional biases, extreme events) highlight the need for improvement
Example: Models generally simulate global temperature trends well but may struggle with regional precipitation patterns
Evaluation can guide model development by prioritizing areas for improvement
Identifying systematic biases or weaknesses can inform the refinement of model components or parameterizations
Example: Evaluating model performance in simulating El Niño events can guide improvements in ocean and atmosphere dynamics
Model evaluation is an ongoing process as new observations become available and models are updated
Newer generations of models (CMIP6 vs CMIP5) can be evaluated against a larger set of observations and under different forcing scenarios (historical, future projections)
Techniques for model evaluation include:
Statistical measures of model-observation agreement, such as correlation (R), root-mean-square error (RMSE), and bias
Visual comparison of spatial patterns (maps) and time series (trends, variability)
Detection and attribution studies that assess the ability of models to simulate observed climate change in response to different forcings (greenhouse gases, aerosols, natural variability)
Example: Comparing modeled and observed patterns of temperature change to detect the human influence on climate