Effects of prescribing aerosols in climate models

Stratospheric aerosol injection (SAI) research is bottlenecked because very few Earth system models (ESMs) have the necessary ingredients to simulate it: a stratosphere resolved with enough vertical detail to capture stratospheric processes, and an interactive aerosol scheme that can dynamically compute how injected sulfur dioxide (SO₂) becomes sulfate aerosol and evolves over time in the stratosphere.

That lack of model diversity limits more than the number of SAI simulations available—it also means only a handful of models can currently simulate how SAI affects the rest of the climate system (tropospheric dynamics, ocean, sea ice, land, and vegetation). Expanding the pool of models that can participate in the Geoengineering Model Intercomparison Project (GeoMIP) can proceed along two paths: building new models with both capabilities directly, or extending models that lack one or both by supplying them a prescribed aerosol field computed by another model. Models that take this second route, however, miss the feedback between aerosol and stratospheric dynamics — for example, stratospheric heating from aerosol absorption alters circulation (Visioni et al., 2020) in ways a prescribed field cannot reproduce.

This activity quantifies the size and nature of the biases introduced by prescribing aerosol fields, through two complementary studies. The first asks a physical question: how large are the climate feedbacks that prescribed-field models miss? This analysis has two parts: first, whether the climate's response back onto the aerosol itself (its size, location, and timing) is sufficiently small, and second whether the way SAI-altered stratospheric dynamics affects the troposphere (the lower atmosphere where weather occurs) is also sufficiently small.

The second study asks a different, more technical question: even setting aside any real physical feedback, does simply importing an aerosol field computed by a different model create numerical mismatches with the receiving model's own dynamics? Together, the two studies will establish whether prescribing aerosol fields is a viable path to model diversification, or whether meaningful expansion of the GeoMIP ensemble beyond Expand GeoMIP necessarily requires the much larger task of developing new model components.

Study 1 will conduct and compare two sets of SAI simulations in the CESM model family. The first simulates SAI directly as SO₂ injection in CESM2-WACCM. The second runs the same scenario in CESM2's low-top atmosphere version (CAM) with prescribed aerosol fields equal to the aerosol distribution produced in the first.

Comparing surface climate changes between the two will reveal how much is lost by prescribing rather than simulating aerosols interactively. Existing simulations may also be informative: CESM2-WACCM aerosol fields were already prescribed in CESM2 for CMIP6 historical simulations, providing a potential additional comparison without requiring new runs.

Study 2 will investigate the consequences of prescribing aerosol fields from one model into a different model. One approach is to prescribe fields from CESM2-WACCM into UKESM1 and vice versa, where both models have stratospheres and interactive aerosols. This will reveal whether the inconsistency between the prescribed field and the receiving model's own dynamics propagates through to significant surface climate impacts — a question about numerical integrity rather than physical fidelity.