8.2 General Circulation Models (GCMs) and Earth System Models (ESMs)

GCMs and ESMs: Core Components and Processes

Climate models are the primary tools scientists use to simulate and predict how Earth's climate behaves over time. General Circulation Models (GCMs) solve fundamental physics equations to simulate atmospheric and oceanic circulation, while Earth System Models (ESMs) build on GCMs by adding biogeochemical cycles (carbon, nitrogen, phosphorus) and ecosystem dynamics. Understanding how these models work, and where they fall short, is central to interpreting any climate projection you'll encounter.

Atmospheric and Oceanic Processes

GCMs are built from three main components that exchange energy and matter with each other:

• Atmospheric component: Models radiative transfer (how energy from the sun moves through the atmosphere), cloud formation, precipitation, and atmospheric chemistry. This is where greenhouse gas forcing gets calculated.
• Ocean component: Simulates large-scale circulation patterns (like the Gulf Stream and Antarctic Circumpolar Current), heat transport, salinity distribution, and sea ice dynamics. The ocean stores vastly more heat than the atmosphere, so getting this right matters enormously for long-term projections.
• Land surface component: Represents vegetation cover, soil moisture, surface runoff, and energy exchange between the ground and the atmosphere. Processes like evapotranspiration (water moving from soil through plants into the air) and albedo changes (how reflective the surface is) are handled here.

ESMs take all of this and layer on biogeochemical processes, which you'll see in the sections below.

Parameterizations and Sub-grid Scale Processes

No model can simulate every physical process at its true scale. A thunderstorm might be 10 km across, but a typical GCM grid cell is 100 km wide. Parameterization is the technique of approximating these sub-grid scale processes using simplified equations based on larger-scale variables the model can resolve.

Commonly parameterized processes include:

• Cloud microphysics: Droplet formation, ice crystal growth, and how clouds interact with radiation
• Convection: Thunderstorms, tropical cyclones, and vertical air movement within a grid cell
• Boundary layer turbulence: Surface wind stress and heat fluxes near the ground
• Ocean eddies: Mesoscale circulation features (roughly 10–100 km) that transport heat and nutrients

Parameterizations are a necessary compromise. They make global simulations computationally feasible, but they also introduce uncertainty because the approximations don't perfectly capture the real physics. Much of the difference between competing climate models comes down to how they parameterize these same processes.

Coupled Atmosphere-Ocean Models for Climate Simulations

Atmosphere-Ocean Interactions

In the real climate system, the atmosphere and ocean constantly exchange heat, moisture, and momentum. Coupled models simulate these interactions together rather than treating each system in isolation.

This coupling is what allows models to reproduce important modes of climate variability. El Niño-Southern Oscillation (ENSO), for example, involves a feedback loop between tropical Pacific sea surface temperatures and atmospheric circulation patterns. Without coupling, you can't simulate ENSO or its global teleconnections (the way it influences weather thousands of kilometers away).

Coupled models also represent ocean heat uptake, which is critical for understanding Earth's energy balance. The ocean absorbs and redistributes enormous amounts of heat, and this process helps explain why surface warming doesn't always track emissions in a straight line. The apparent slowdown in surface warming during the early 2000s, for instance, was partly linked to increased heat uptake in the deep ocean.

Ocean Circulation and Climate Impacts

Beyond surface interactions, coupled models simulate how changes in ocean circulation ripple through the climate system:

• The Atlantic Meridional Overturning Circulation (AMOC) carries warm water northward and influences North Atlantic and European climate. Models project that AMOC could weaken under continued warming, with significant regional consequences.
• Sea ice feedbacks connect ocean circulation to polar climate. As sea ice melts, it exposes darker ocean water that absorbs more solar radiation, amplifying warming. This is a key driver of Arctic amplification, the observation that the Arctic is warming roughly two to four times faster than the global average.
• The ocean's role as a carbon sink is also captured in coupled models. Oceans have absorbed approximately 30% of anthropogenic $\text{CO}_2$ emissions, but warming and acidification are reducing this uptake capacity over time.

Biogeochemical Cycles in ESMs

Carbon Cycle Integration

The carbon cycle is the most important biogeochemical addition that distinguishes ESMs from standard GCMs. ESMs simulate $\text{CO}_2$ exchange among the atmosphere, land biosphere, and ocean through processes like photosynthesis, plant and soil respiration, and ocean dissolution.

Why does this matter? Because it enables climate-carbon feedbacks. In a standard GCM, you prescribe atmospheric $\text{CO}_2$ concentrations as an input. In an ESM, the carbon cycle responds to warming, which can then change $\text{CO}_2$ concentrations, which changes warming further. Two major feedback concerns:

• Permafrost thaw: As Arctic soils warm, they release stored organic carbon as $\text{CO}_2$ and methane ($\text{CH}_4$), potentially accelerating warming.
• Weakening ocean carbon sink: Warmer water holds less dissolved $\text{CO}_2$, and ocean acidification disrupts marine organisms that sequester carbon. Both effects could reduce future ocean uptake.

These feedbacks are why ESMs are essential for assessing different emission pathways and estimating the remaining global carbon budget.

Nutrient Cycles and Ecosystem Dynamics

Carbon uptake by ecosystems doesn't happen in a vacuum. Plants need nitrogen and phosphorus to grow, and these nutrients are often in limited supply:

• Nitrogen limitation is the primary constraint on plant growth in boreal forests and many temperate ecosystems. Even if $\text{CO}_2$ levels rise (which can stimulate photosynthesis), plants can't take full advantage without sufficient nitrogen.
• Phosphorus limitation dominates in tropical rainforests, where ancient, weathered soils have been depleted of this nutrient over millennia.

ESMs also simulate how land use changes affect greenhouse gas budgets. Deforestation releases stored carbon, cropland expansion alters surface albedo and moisture fluxes, and fertilizer application increases nitrous oxide ($\text{N}_2\text{O}$) emissions, a potent greenhouse gas. These interactions create feedback loops between vegetation growth, nutrient availability, carbon uptake, and climate that only ESMs can capture.

Spatial and Temporal Resolutions of Climate Models

Spatial Resolution Characteristics

Spatial resolution refers to the size of the grid cells a model uses to divide up the Earth. It directly controls what phenomena the model can resolve versus what must be parameterized.

• Global models typically run at 100–25 km horizontal resolution. At 100 km, features like individual mountain ranges and small island nations are poorly represented.
• High-resolution global models push down to ~10 km or less, which starts to resolve tropical cyclones and narrow ocean currents, but at enormous computational cost.
• Vertical resolution usually includes 30–100 layers in both the atmosphere and ocean, capturing how conditions change with altitude and depth.
• Regional climate models (RCMs) are nested within global models to achieve 1–10 km resolution over a specific area. This allows much better representation of local topography, coastlines, and land-use patterns.
• Variable-resolution grids offer a compromise: finer grid spacing in areas of interest (say, over a mountainous region) and coarser spacing elsewhere, optimizing computational resources.

Temporal Resolution and Process Representation

Most model processes are calculated on timesteps of minutes to hours. Radiation calculations, which are computationally expensive, are often performed less frequently (every few hours) since radiative conditions change more slowly than, say, wind patterns.

The choice of temporal resolution involves a direct trade-off with computational cost:

• Sub-daily resolution is necessary to capture diurnal cycles (day-night temperature swings), precipitation intensity, and the timing of extreme events.
• Higher temporal resolution improves the simulation of fast-evolving phenomena like atmospheric rivers and convective storms.
• Coarser timesteps save computing time but can miss short-lived events that have outsized climate impacts.

Computational Requirements and Limitations of Climate Models

Hardware and Resource Demands

Running a GCM or ESM is not something you do on a laptop. These models require high-performance computing clusters or supercomputers, and the demands scale steeply with ambition.

• Computational cost increases roughly as the cube of resolution improvement (doubling horizontal resolution in all directions and halving the timestep increases cost by about 8–10x).
• Model output storage routinely reaches petabytes (millions of gigabytes), especially for long-term projections or large ensembles where the same scenario is run dozens of times with slightly different initial conditions.
• Modelers face constant trade-offs: you can have higher resolution or longer simulations or more ensemble members or more Earth system components, but rarely all at once.

Model Complexity and Advancements

Parameterization uncertainties remain the largest source of spread between different climate models. Small differences in how clouds or convection are parameterized can lead to meaningfully different projections of future warming.

Several developments are helping address these limitations:

• GPU computing: Graphics Processing Units handle parallel calculations far more efficiently than traditional CPUs, enabling faster simulations at higher resolutions.
• Machine learning: Neural networks are being used to develop faster, more accurate parameterizations by training on high-resolution simulations or observational data.
• Earth System Modeling Frameworks (ESMFs): These modular software architectures make it easier to swap in new components (say, an improved ice sheet model) without rebuilding the entire model from scratch.
• Model intercomparison projects: Standardized experiments like CMIP6 (Coupled Model Intercomparison Project, Phase 6) allow researchers to compare output across dozens of independent models, helping identify robust projections versus model-dependent results.