Emulators deliver robust end‑of‑century projections — aerosols remain the weak spot

Key finding

A GMST‑driven spatial climate emulator reproduces end‑of‑21st‑century projections reliably in most regions, even under aggressive overshoot pathways. Noticeable deviations concentrate in areas with rapid changes in aerosol emissions. Including aerosol optical depth (AOD) as an additional driver substantially reduces regional biases.

What is Overshoot? In overshoot scenarios, global mean surface temperature (GMST) temporarily exceeds an agreed limit and later falls below it again through strong emission reductions. Such pathways have become relevant because current mitigation efforts often fall short of avoiding a temporary exceedance. Single global metrics like GMST capture only part of the picture, since local and partly irreversible effects may not scale simply with a global average.

How do climate emulators work? Emulators are computationally light alternatives to full Earth system models. They link changes in GMST to spatial patterns of temperature and precipitation, enabling rapid scenario studies. Many approaches use GMST as the sole driver. This simplifies models but can introduce errors where local responses are non‑linear with GMST or strongly influenced by additional forcings.

Study approach

The study tested how well a GMST‑driven spatial emulator reproduces a full Earth system model (ten‑member ensemble) under an aggressive overshoot scenario. Evaluated variables at century’s end were near‑surface mean temperature, daily maximum temperature, and precipitation. Performance was assessed by comparing emulator biases to the magnitude of internal climate variability and ensemble spread.

Results at a glance

– For roughly 99 percent of land areas, differences between the emulator and the Earth system model lay within the range of natural internal variability and ensemble spread. – Statistically significant, larger deviations appeared mainly in regions experiencing rapid reductions in aerosol emissions. – These biases are not explained by GMST alone: changes in aerosols affect local radiative balance, cloud properties and thus temperature and precipitation patterns.

Aerosols as an additional key

An extended emulator that used both GMST and aerosol optical depth (AOD) showed substantially better agreement with the Earth system model. Regional biases shrank especially in areas with strong changes in aerosol loading. This supports explicitly accounting for aerosol‑driven radiative effects in emulation frameworks.

Role of internal variability

Internal variability denotes natural climate fluctuations that occur even without changed external forcings. By century’s end it is large enough in many regions to mask small to medium emulation errors. However, this buffering is not always sufficient for regionally targeted decisions or for statements about extremes — particularly where aerosols decline rapidly or local processes dominate.

Implications for use and research

– GMST‑driven spatial emulators are suitable for broad, rapid estimates of end‑of‑century projections — including under overshoot pathways. – For regions with fast aerosol changes, or for analyses of extreme temperatures and precipitation distributions, emulator results should be supplemented with ensemble runs from full Earth system models. – Practically, integrating aerosol‑related predictors such as AOD and testing other influences (e.g., land‑use change, regional ocean warming, black carbon) is recommended.

Outlook

Future work should test the AOD extension across models and scenarios to assess robustness and transferability. Hybrid approaches that combine fast emulation with targeted, high‑resolution model runs could improve the balance between efficiency and physical detail — benefiting decisions on regional climate risks and the design of overshoot strategies.