Some of the earliest climate models, developed in the 1960s, consisted of just a few equations that represented Earth’s radiative-transfer processes and the atmosphere’s convection. Modern models, however, incorporate much more than just radiative transfer, and they solve all the equations that represent the various processes at dozens of vertical levels in the atmosphere. Despite the simplicity, the early models produced estimates of the climate’s sensitivity to greenhouse gases that are roughly in agreement with results from the sophisticated atmospheric climate models used today. (For more on early climate modeling, see Physics Today, December 2021, page 14.)
Now Nicholas Davis and colleagues have taken one of those models—the Community Earth System Model Version 2—and evaluated whether it still produces a quality climate simulation with a simpler representation of atmospheric chemistry. The state-of-the-art configuration simulates one model day using the equivalent processing power of thousands of personal computers, but for the same simulation, the simpler configurations use 35–86% less processing power and still capture key features of Earth’s climate.
The atmospheric configuration of the model typically runs with chemical interactions between the various layers of the atmosphere, from the troposphere (the lowest-lying layer) to the stratosphere, mesosphere, and lower thermosphere, near outer space. In the simple configurations, Davis and colleagues purposefully neglected organic species and other constituents that may be important for air quality, for example, but aren’t necessary for accurately modeling other physics, including stratospheric ozone.
The simpler modeling scheme did still include greenhouse gas chemistry and went through a bevy of tests: a historical simulation of the industrial era, a preindustrial 1000-year run, and a 150-year simulation in which the atmosphere’s carbon dioxide was quadrupled.
The model with simplified chemistry yielded the same result as the model with full chemistry for many climate variables, including sensitivity to changes in carbon dioxide, variability of the polar vortex, and the area and volume of Arctic ice. Perhaps unsurprisingly, the simplified chemistry wasn’t as accurate for some variables, such as aerosol interactions in the upper troposphere and the lower stratosphere. Although coarsening the model resolution from 1° to 2° of latitude and longitude yielded a more significant change compared with the choice of chemistry implementation.
For many research topics—including assessing the effect of future wildfire emissions or studying air quality—resolving the many chemical reactions and constituents in the troposphere is critical. But the simplified atmospheric chemistry may be sufficient for researchers analyzing how geoengineering may offset greenhouse gas warming or studying stratospheric variability.
It even opens up new modeling possibilities. Because the simulations take a fraction of the time, researchers could run longer experiments or a larger number of simulations. With bigger data sets, they could better study historical climate variability or use the freed computer resources to more carefully explore climate phenomena that operate over long time periods. (N. A. Davis et al., J. Adv. Model. Earth Sys. 15, e2022MS003579, 2023.)