Improve microphysics in GCMs

Every Earth system model contains a dedicated aerosol scheme, a smaller model that represents the life cycle of particles: how they form, how they grow as vapor condenses onto them and as they collide and merge with each other, how settling and transport remove them, and how they scatter and absorb radiation. Two examples are the Global Model of Aerosol Processes (GloMAP), which runs inside the UK Earth System Model, and the Modal Aerosol Module (MAM), which runs inside CESM2-WACCM.

Different research groups developed these aerosol schemes independently, each pursuing its own scientific questions and calibrating against its own body of observations such as volcanic eruptions, wildfire plumes, urban air quality, dust storms, and cloud formation. Each scheme therefore performs well in the regime its developers built it for rather than serving as a general-purpose representation of aerosol physics. The schemes consequently diverge, meaning that they produce different sizes of particles and concentrations with a given quantity of SO₂. This divergence is one of the principal sources of uncertainty in projections of SAI (Visioni et al., 2021) (Irvine et al., 2014).

Researchers have tested these schemes for SAI almost exclusively against volcanic eruptions, which are not a like-for-like analogue. Volcanic eruptions deliver very large quantities of SO₂ to the stratosphere in concentrated plumes that also carry ash and other co-emitted gases, and instruments have characterized that composition only approximately. SAI would instead involve smaller, more sustained releases from aircraft at locations, altitudes, and rates chosen deliberately. The eruption record therefore constrains a physical system that differs from SAI in emission rate, plume composition, and spatial scale alike.

No mechanism currently exists for feeding new observations back into microphysical code. If better measurements arrived tomorrow, no modelling group has a systematic procedure for translating them into a revised scheme. Calibration remains a one-off exercise, repeated separately by each group and rarely documented in a form others can reuse.

An aerosol microphysics experiment will measure how SO2 converts to aerosol particles under controlled, well-instrumented stratospheric conditions. Combined with the record from past volcanic eruptions, on-going stratospheric measurements, and from new observational campaigns, these measurements will constrain the microphysical schemes in the regime SAI would actually occupy. Building the pipeline that turns observations into model revisions is the essential first step, and will require Observing System Simulation Experiments (OSSEs), which test how a model's predictions respond to different scenarios and measurement designs.

The initial approach involves three main steps:

1. Refining aerosol microphysical models — researchers will use data from the aerosol microphysics experiment along with observational data from volcanic eruptions to refine aerosol microphysical models, including at least three models such as MAM, the Community Aerosol and Radiation Model for Atmospheres (CARMA), and GloMAP. This refinement will follow methods developed in the activity designed to determine observational requirements for the aerosol microphysics experiment and will focus on improving how these models simulate sulfate aerosol formation, growth, and properties under stratospheric conditions.
2. Cross-model comparison — researchers will compare the improved models against each other to understand what changed and how much those improvements matter for predicting aerosol size and number. This comparison will validate whether the improvements are moving all models in the right direction or if key disagreements persist.
3. Coupling into full climate models — researchers will incorporate these improved models into full climate models to see how the improvements affect predictions of surface climate. Given the complexity of fully coupling new aerosol schemes into climate models, the initial approach will be to update a standalone atmospheric model first, use those results to generate boundary conditions, and then feed that into a fully coupled climate model (which includes oceans and land). This step-by-step approach avoids the need for extensive retuning of the entire climate model while still capturing how improved aerosol predictions translate into different climate impacts.