Understanding our Earth’s climate has been one of humankind’s greatest challenges. Scientists want to know things like, “Why do climates change?” and “How does the climate respond to increased levels of greenhouse gases?” Society also has important questions like, “How will the climate change in the future?” and “How can we reduce the severity of climate change?”. Climate models are a tool that can help us answer these questions.
Climate scientists have created mathematical models of the Earth that include components such as the atmosphere, land, ocean, and more. These “virtual laboratories” are used to test questions about the climate and provide society with predictions of future conditions.
In many ways, climate models are like the models used to predict tomorrow’s weather. However, whereas weather models make forecasts for specific regions over time spans of days and weeks, climate models are meant to project global changes over years, decades, and even centuries.
This learning module will take you through the fundamentals of climate models, and help you answer important questions like: How are climate models made? How do they work? Why can we trust them? What are their limitations? And finally, who runs them?
Climate models ‘slice’ the globe into three-dimensional grid cells, the number of which is limited by computing power.
Early climate models ran on the largest supercomputers of the day, but still only included a few grid cells to represent simple ocean, atmosphere, and land components. Scientists used these early models to learn about fundamental climate and weather processes.
Grid cells: A series of vertical and horizontal “boxes” that represent sections of the Earth. Due to the complexity of the climate system, it would be impossible to model each of it’s processes for every cubic meter of the globe. Instead, climate models calculate the average climate over a larger area (a grid cell) throughout time. Q&A: How do climate models work? | Carbon Brief
Supercomputers: Extremely powerful computers that perform computations at a volume and speed much greater than personal computers.
As time passed, computing power improved rapidly, along with understanding of how different components of the climate system interact.
These advances have allowed for increased climate model resolution, and finer scale representation of the Earth’s climatic processes.
Spatial Resolution: The size of climate model grid cells. High resolution indicates more, smaller grid cells of smaller size, that collectively show a greater level of detail.
Temporal Resolution: the size of the time steps used in models, which dictates how often the various properties being simulated in the climate model are updated.
Today, global climate models (commonly known as Earth System Models) include thousands of systems and processes. Not only do these models represent the atmosphere, oceans, and land, but also other key climate components, such as sea ice, plants, glaciers, rivers, ocean waves, chemical cycles, cities, and more. They also contain parameterizations of important – but small scale – climate processes, like thunderstorms.
In order to run climate models, scientists input the key boundary condition processes that influence the climate over time. These processes include:
Once models have this information, they can be run forward to produce datasets of ‘simulated’ climate. This data is then studied – using techniques very similar to those used to study real climate and weather data – to understand how the climate system will react and change in the future.
Parameterizations: Climate models represent processes that occur on scales smaller than the resolution of individual climate model grid points using parameterizations. Parameterizations capture the effect of these processes by using our physical understanding of these processes to develop relationships between large-scale climate conditions and smaller-scale processes.
Boundary Conditions: Boundary conditions determine how the climate system behaves and changes. Boundary conditions must be specified by climate modellers and are sometimes referred to as climate “forcings”.
Before using climate models to plan for the future, it is important to ensure they can reasonably reconstruct the climate. To do this, scientists put their models through some tough tests!
Models are run many times to recreate past climate. Scientists then test how closely the climate model is able to match past observations. If the models do a good job of validating these observations, then scientists are more confident that the model will also do a good job of projecting future changes.
Validation: Ensuring that climate model simulations of past weather and climate compare reasonably well against direct historical atmosphere, ocean, land and ice observations.
To understand future climate, the model is provided with scenarios of future greenhouse gas emissions, land use changes, and other factors. Once equipped with this information, the models are run and return an estimate of how the climate system will respond to each scenario.
Climate models are excellent planning tools and one of our only practical means of exploring future climate in a consistent way. However, they are global in scale and may not capture local-scale variability and change very well. To overcome this, additional methods, known as downscaling and bias adjustment, have emerged to translate global climate model information to regional scales. For example, the Canadian Regional Climate Model (CanRCM4) translates global model data into more detailed information over North America.
Developing global climate models takes specialized groups many years, using large supercomputers. For this reason, global climate model teams are typically supported by national-scale research and development programs. These teams work closely across borders with each other to improve model designs, share results, and coordinate global climate modelling projects such as the Coupled Model Intercomparison Project (CMIP).
