Projected rapid response of stratospheric temperature to stringent climate mitigation

The study addresses whether stringent near-term greenhouse gas mitigation produces detectable climate-system changes on short timescales, and specifically whether stratospheric temperatures can provide early, robust evidence of mitigation impacts. Motivated by the 2023 UNEP Emissions Gap Report showing the world off track for the 1.5 °C target and the rapidly diminishing carbon budget, the authors note that detecting mitigation signals in global surface air temperature (GSAT) and other surface indicators is hindered by large internal variability and lagged oceanic responses. In contrast, the globally averaged stratosphere is close to radiative balance, leading to much lower internal variability than in the troposphere. Consequently, externally forced stratospheric temperature trends should be more readily detectable and could provide an earlier indication that the climate trajectory is shifting due to mitigation. The paper evaluates near-term trends in surface and stratospheric temperatures under different future emissions scenarios to determine detectability timelines.

Established theory and observations show a vertical fingerprint of anthropogenic influence featuring tropospheric warming and stratospheric cooling. The IPCC First Assessment Report highlighted stratospheric cooling as an important detection variable, and subsequent work (e.g., Santer et al.) demonstrated that including middle and upper stratospheric data improves detectability of the anthropogenic temperature fingerprint by a factor of five. Over the past ~40 years, observed stratospheric cooling has been primarily attributed to increasing greenhouse gases and ozone depletion, with a smaller role for stratospheric water vapour changes. Looking ahead, CO2-driven stratospheric cooling will be partially offset by height-dependent warming from ozone recovery, but higher signal-to-noise ratios in the stratosphere are still expected, suggesting potential for faster detection of mitigation effects.

The analysis uses large initial-condition ensembles from three CMIP6 climate models—CanESM5, EC-Earth3, and MIROC6—each simulating three ScenarioMIP future emissions pathways: SSP1-1.9 (stringent mitigation aligned with the 1.5 °C goal), SSP2-4.5 (approximate realization of current commitments), and SSP3-7.0 (baseline with weak near-term mitigation). Models span a range of climate sensitivities (CanESM5: ECS 5.64 °C, TCR 2.66 °C; EC-Earth3: ECS 4.10 °C, TCR 2.38 °C; MIROC6: ECS 2.60 °C, TCR 1.52 °C). Scenario simulations are initialized on 1 January 2015 and include evolving anthropogenic forcings (greenhouse gases, aerosols, land use) and natural forcings (solar, volcanic). To address data quality issues, CanESM5 ensemble members exhibiting spurious temperature spikes were filtered by excluding any member exceeding ±3 standard deviations of the inter-ensemble spread in any year during 2023–2042; the remaining ensemble sizes are listed per model and scenario in Table 1 of the paper. Four global indicators were evaluated: GSAT and three synthetic satellite-observable atmospheric layer temperatures—TLS (lower stratosphere), TMS (middle stratosphere), and TUS (upper stratosphere). Synthetic layer temperatures were constructed by applying satellite weighting functions: TLS from Remote Sensing Systems (MSU4; peak ~18 km), and TMS (SSU1; peak ~30 km) and TUS (SSU2; peak ~37 km) from NOAA STAR SSU v3, interpolated to 17 standard pressure levels up to 1 hPa. For CanESM5, sensitivity tests using higher vertical resolution (plev39) indicated negligible impact on near-term TUS trends from including additional levels above 1 hPa. For each ensemble member, least-squares linear trends were computed over 5, 10, 15, and 20-year windows starting in 2023. Statistical separability between scenarios was quantified via the overlap of trend distributions: for each pair of scenarios, 10th and 90th percentile trends in the higher-forcing scenario were mapped to corresponding percentiles in the lower-forcing scenario, and overlap was defined as the percentile-range intersection (bounded at 100%).

- Middle and upper stratospheric temperature trends (TMS and TUS) exhibit high signal-to-noise ratios: ensemble spreads are small relative to ensemble-mean changes, and trends are more readily distinguishable between emissions scenarios than for GSAT or TLS. - In all three models, distributions of TMS and TUS trends begin to separate between scenarios within 5–10 years. Separation is fastest and largest for TUS, consistent with stronger CO2-driven cooling at higher altitudes. - Comparing stringent mitigation (SSP1-1.9) against a weak-mitigation counterfactual (SSP3-7.0), a reduction in upper stratospheric cooling (TUS) is clearly detectable within 5 years; overlap of trend distributions diminishes to near-zero within 5–10 years. - Even when comparing SSP1-1.9 with SSP2-4.5 (approximate current commitments), TMS/TUS trend distributions separate within 5–10 years after accounting for internal variability; overlap can be up to ~20% at 5 years in one model but drops to near-zero by 10 years. - GSAT trends show strong overlap between scenarios for at least the next 10–15 years in all models. Achieving comparable statistical separation in GSAT requires at least ~20 years in CanESM5 and longer in EC-Earth3 and MIROC6 (which have lower TCR/ECS). - TLS shows weak near-term global mean trends with relatively large ensemble spread due to cancellation between upper tropospheric warming and extratropical lower stratospheric conditions, yielding similar global trends across scenarios over the next 5–10 years. - Solar-cycle variability appears in TMS/TUS time series but does not preclude detection given the high signal-to-noise in the ensemble-mean forced response.

The findings demonstrate that stringent near-term greenhouse gas mitigation would produce a weakening of stratospheric cooling that is statistically detectable in middle and upper stratospheric temperature trends within 5–10 years, substantially earlier than detectability in global surface temperature. This addresses the central question by identifying a robust, satellite-observable early indicator of mitigation impacts despite internal variability that obscures near-term GSAT changes. The results underscore the value of leveraging components of the climate system with high signal-to-noise ratios for early detection, and motivate development of a multivariate monitoring framework combining stratospheric temperatures with other indicators. Such early evidence can inform policymakers and the public that mitigation is altering the climate trajectory and support sustained long-term climate action.

This study shows that stratospheric temperature trends, particularly in the middle and upper stratosphere, provide an early and robust indicator of the effects of stringent climate mitigation, with detectable differences between mitigation pathways within 5–10 years—well before comparable separation is attainable in global surface temperature. The work contributes a practical detection strategy using satellite-observable layers and a distribution-overlap metric applied to large-ensemble projections. Future research should: (i) expand the set of high signal-to-noise climate indicators for early detection of mitigation impacts; (ii) apply the methodology across broader model ensembles and additional emissions scenarios; (iii) refine observation-model comparisons, including improved vertical weighting and coverage; and (iv) assess sensitivities to external forcings (e.g., volcanic events) and structural model uncertainties.

- Scenario dependence and policy uncertainty: SSPs are not predictions; real-world policies may diverge from the scenario pathways, affecting detectability timelines. - Model coverage: Results are based on three CMIP6 models; broader model ensembles could sample a wider range of internal variability and forced responses. - Structural differences: Detectability timelines depend on model TCR/ECS; models with lower sensitivities (e.g., MIROC6) show slower separability for GSAT. - Satellite emulation: Constant (latitude-independent) weighting functions and a vertical grid limited to 1 hPa were used; while sensitivity tests suggest minimal impact on near-term TUS trends, more detailed weighting and higher vertical resolution could refine estimates. - Internal variability and external forcings: Although large ensembles reduce variability, episodic natural forcings (e.g., volcanic eruptions) and solar-cycle variability could modulate short-term trends. - Data quality: Some CanESM5 members exhibited spurious temperature spikes and were filtered; residual data issues in archived products could affect analyses. - Indicator scope: TLS shows significant cancellation across regions, limiting its utility as an early indicator; the study focuses on global means rather than regional signals.