18.2 Climate modeling and projections

Climate Models and Scenarios

Climate models translate the physics of Earth's atmosphere, oceans, and land surface into mathematical equations that a computer can solve. They're the primary tools scientists use to project how the climate system will respond to rising greenhouse gas concentrations, changes in land use, and other forcings. This section covers how those models work, what scenarios they run, and why their projections carry inherent uncertainty.

General Circulation Models and Radiative Forcing

General Circulation Models (GCMs) are the workhorses of climate science. They simulate the full climate system by numerically solving equations that represent physical processes in the atmosphere, ocean, and land surface.

Here's how they're structured:

1. The Earth is divided into a three-dimensional grid, with cells stacked vertically through the atmosphere and ocean.
2. Each grid cell tracks variables like temperature, pressure, humidity, and wind speed.
3. At each time step, the model calculates the transfer of energy, mass, and momentum between neighboring cells.
4. The results feed into the next time step, and the simulation marches forward.

Grid resolution matters. A typical GCM grid cell spans 100–300 km horizontally, which means small-scale processes like individual thunderstorms can't be resolved directly. Instead, they're approximated through parameterizations, simplified representations of sub-grid processes.

Radiative forcing is the concept that connects human activity to climate change within these models. It measures the change in Earth's energy balance (in watts per square meter, $\text{W/m}^2$) caused by a given factor:

• Positive radiative forcing (e.g., increased $CO_2$, methane) means more energy is retained in the system, leading to warming.
• Negative radiative forcing (e.g., sulfate aerosols reflecting sunlight) means less energy is retained, leading to cooling.

GCMs use radiative forcing values as inputs. By adjusting these values, scientists can simulate how the climate responds to different combinations of forcings over time.

Emission Scenarios and Representative Concentration Pathways

To project future climate, models need assumptions about how human emissions will change. That's where emission scenarios come in. They describe plausible futures based on different assumptions about population growth, economic development, technology, and climate policy.

Representative Concentration Pathways (RCPs) are a set of four greenhouse gas concentration trajectories adopted by the IPCC for its Fifth Assessment Report (2014). Each RCP is named after its approximate radiative forcing value in the year 2100:

These pathways don't predict what will happen. They bracket a range of possibilities so that policymakers can see the consequences of different choices. Climate models take RCPs as input and project the resulting changes in temperature, precipitation, sea level, and other variables.

Worth noting: the IPCC's Sixth Assessment Report (2021) introduced Shared Socioeconomic Pathways (SSPs), which pair emission trajectories with broader socioeconomic narratives. SSPs are gradually replacing RCPs in newer research, but the underlying logic is the same.

Feedback Mechanisms in Climate Models

A feedback mechanism is a process where the climate system's initial response to a forcing either amplifies or dampens itself. Feedbacks are what make climate projections both powerful and uncertain.

Positive feedbacks amplify the initial change:

• Ice-albedo feedback: Warming melts Arctic sea ice, exposing darker ocean water. Darker surfaces absorb more solar radiation, which causes further warming, which melts more ice. This is one of the reasons the Arctic is warming roughly two to four times faster than the global average.
• Water vapor feedback: Warmer air holds more water vapor, and water vapor is itself a greenhouse gas. So initial warming from $CO_2$ leads to more atmospheric water vapor, which traps more heat.

Negative feedbacks dampen the initial change:

• Planck response: As Earth's surface warms, it radiates more longwave energy to space, which partially offsets the warming. This is the most fundamental stabilizing feedback.
• Some cloud feedbacks: Warmer temperatures can increase evaporation and low-level cloud formation. Low clouds reflect sunlight, which has a cooling effect.

Cloud feedback is actually one of the trickiest parts of climate modeling. Different cloud types at different altitudes can produce either positive or negative feedbacks, and models disagree on the net effect. This disagreement is a major reason why the range of projected warming remains wide.

Climate models must represent all of these feedbacks simultaneously, and the strength assigned to each one significantly affects the model's projections.

Model Refinement Techniques

Downscaling and Ensemble Modeling

GCM grid cells of 100–300 km are useful for global patterns, but too coarse for regional planning. If a city needs to know how flood risk will change by 2060, it needs finer resolution. Downscaling bridges that gap using two main approaches:

• Dynamical downscaling nests a Regional Climate Model (RCM) within a GCM. The GCM provides boundary conditions (large-scale atmospheric patterns), and the RCM simulates climate at higher resolution (often 10–50 km) over a limited area. This captures regional features like mountain effects and coastlines but is computationally expensive.
• Statistical downscaling uses observed statistical relationships between large-scale atmospheric variables and local climate variables. It's cheaper and faster than dynamical downscaling, but it assumes those statistical relationships will hold in the future, which isn't guaranteed under novel climate conditions.

Ensemble modeling tackles a different problem: uncertainty. No single model run gives you the "right" answer, because:

• Different models represent physics differently.
• The climate system has internal variability (natural fluctuations that occur even without any forcing change).
• Small differences in initial conditions can lead to divergent outcomes over time.

To address this, scientists run ensembles, collections of multiple simulations:

• Multi-model ensembles (like CMIP5 and CMIP6) combine output from dozens of independently developed models worldwide. Where models agree, confidence is higher.
• Perturbed-physics ensembles take a single model and systematically vary uncertain parameters (like cloud feedback strength) across many runs to map out the range of possible outcomes.

The spread across an ensemble gives you a measure of projection uncertainty, which is just as important as the central estimate.

Model Validation and Climate Sensitivity

Before trusting a model's projections of the future, you need to check how well it reproduces the past and present. Model validation does exactly that by comparing simulations against:

• Instrumental records (thermometer data, ocean buoys, weather stations) from the past ~150 years
• Paleoclimate reconstructions (ice cores, tree rings, ocean sediments) for deeper time periods
• Satellite observations for variables like sea ice extent, cloud cover, and radiation balance

Validation doesn't prove a model is "correct," but it identifies which models perform well for which variables and regions, and it highlights systematic biases that developers can work to fix.

Climate sensitivity is one of the most important numbers in all of climate science. It quantifies how much global mean surface temperature will change in response to a doubling of atmospheric $CO_2$ concentration (from the pre-industrial level of ~280 ppm to ~560 ppm).

Two versions of this metric matter:

• Equilibrium Climate Sensitivity (ECS): The temperature change after the climate system has fully adjusted to doubled $CO_2$ and reached a new equilibrium. This can take centuries.
• Transient Climate Response (TCR): The temperature change at the moment $CO_2$ reaches double its initial concentration during a gradual (1% per year) increase. TCR is always lower than ECS because the ocean hasn't finished absorbing heat yet.

The IPCC's Sixth Assessment Report (2021) narrowed the likely range for ECS to 2.5–4.0°C, with a best estimate of 3°C. (The Fifth Assessment Report had a wider range of 1.5–4.5°C.) This narrowing came from better understanding of cloud feedbacks and paleoclimate constraints.

Why does this number matter so much? Because ECS acts as a multiplier for all emission scenarios. A world with ECS at 2.5°C and a world with ECS at 4.0°C will experience very different outcomes even under the same emissions pathway. That's why pinning down climate sensitivity remains one of the highest priorities in climate research.