Abrupt reduction in shipping emission as an inadvertent geoengineering termination shock produces substantial radiative warming

The Earth's atmosphere has warmed because human activities increased greenhouse gases that create positive radiative forcing. Anthropogenic aerosols, however, increase atmospheric reflectivity directly and indirectly via cloud interactions, imparting a net cooling effect whose magnitude is crucial for understanding climate sensitivity and future warming. Marine cloud brightening (MCB) aims to enhance cloud reflectivity using aerosols; ship tracks observed in satellite data have served as opportunistic small-scale MCB experiments, showing increased cloud droplet number and altered cloud liquid water path (LWP) and cloud fraction (CF). On January 1, 2020, the International Maritime Organization (IMO) implemented IMO2020 regulations reducing maximum sulfur content in international shipping fuel from 3.5% to 0.5%, leading to an abrupt, large reduction in SO2 emissions (~80%). This constitutes an inadvertent termination shock to long-running aerosol-driven cooling from shipping, effectively a reverse MCB that reduces cloud droplet number concentration (Nd) and dims marine clouds. The study aims to quantify the radiative forcing from this event using a combination of satellite observations and chemical transport modeling, and to assess its implications for recent and near-term warming and regional climate impacts.

The paper situates its work within research on aerosol-cloud interactions and solar radiation modification. Prior satellite observations documented ship tracks as evidence of aerosol-induced cloud brightening and microphysical changes (e.g., Twomey and Albrecht effects). Studies have shown that aerosols affect both droplet number and macrophysical properties such as LWP and cloud fraction. Modeling and observational analyses have estimated radiative forcing from shipping emissions and highlighted uncertainties in aerosol indirect effects. Recent analyses indicate reduced ship-track occurrence post-IMO2020 and cloud dimming signatures in specific corridors (e.g., South Atlantic). Multiple modeling groups report global forcing on the order of 0.1 W m−2 from IMO2020-related changes, and independent observational approaches yield comparable lane-scale forcing, providing cross-validation. The literature also discusses MCB as a potential geoengineering strategy and emphasizes challenges due to spatial and temporal heterogeneity and potential impacts on the hydrological cycle.

The study combines a chemical transport model with satellite observations and empirically derived cloud adjustment relationships to estimate radiative forcing from the IMO2020-induced aerosol reduction.

• GEOS-GOCART simulations: NASA GEOS Earth System Model with the GOCART aerosol module was run in replay mode using MERRA-2 reanalysis for dynamics. Simulations spanned 2019-10 to 2020-12 (first 3 months spin-up), at ~50 km resolution with 72 vertical layers. Two experiment sets were used: (1) Business-as-usual (BAU) emissions using CEDS (2019 emissions repeated into 2020) with nine sectors plus QFED biomass burning, volcanic SO2 (OMI-based), biogenic (MEGAN), and online dust/sea-salt; and (2) Covid-adjusted emissions applying daily sectoral scale factors derived from Google/Apple mobility data to 2019 CEDS monthly emissions. Within each set, three sulfur shipping scenarios were run: full shipping emissions, reduced emissions per IMO2020 standards, and no sulfur shipping emissions. The impact of IMO2020 was taken as the difference between the full and reduced shipping scenarios. Aerosol loading differences were translated to changes in aerosol/droplet number.
• Deep-learning estimation of Nd: Because the operational GEOS setup does not prognose CCN or Nd, a diagnostic approach was used involving two neural nets. MAMnet emulates the Modal Aerosol Module (MAM7) to predict internally mixed modal number and composition from bulk aerosol mass (sulfate, sea salt, dust, black carbon, organics) and atmospheric state, enabling CCN estimation. Wnet predicts subgrid vertical velocity variability (σw) from atmospheric state and turbulence metrics (Richardson number, scalar diffusivity), trained on global storm-resolving simulations and constrained by observations. The Abdul-Razzak and Ghan activation scheme uses MAMnet size distributions and Wnet σw to estimate Nd.
• Aerosol indirect forcing calculation: Following Yuan et al., the approach includes the Twomey effect and adjustments to LWP and cloud fraction. Susceptibility of cloud or scene albedo to Nd or Na is used to compute shortwave TOA radiative forcing, combining modeled ΔNd (scaled by local modeled-to-observed Nd ratios to correct biases) with observations of background cloud properties. LWP and CF adjustment sensitivities (d ln LWP / d Na and CF responses) were derived from a large sample of ship-track observations and parameterized as functions of background conditions (e.g., Na, SST, estimated inversion strength, low cloud fraction). Forcing is computed monthly, and multiple functional forms (including an Na-only lower-bound form and various two-variable combinations) are averaged, with spread reported as uncertainty. CERES EBAF-TOA provides TOA fluxes and absorbed solar radiation used for energy balance and hemispheric contrast analyses.
• Energy balance model: A one-layer energy balance model with heat capacity C = 8.2 W yr m−2 K−1 and climate feedback parameter λ = 1.2 W m−2 K−1 is used to translate forcing into transient temperature response for an abrupt forcing. For F = 0.2 W m−2, equilibrium warming is ~0.17 K with an e-folding timescale of ~7 years, implying ~0.24 K/decade additional warming rate. Uncertainty in λ (±0.25 W m−2 K−1) yields equilibrium warming ~0.14–0.21 K.
• Sensitivity/seasonality: Seasonal variations in forcing components are analyzed over the North Atlantic by holding individual variables (ΔNa, background Nd, CF) seasonally invariant in turn to attribute contributions to seasonality after removing insolation seasonal cycle. Observational comparisons of TOA flux trends and hemispheric absorbed shortwave contrast (CERES) are used to contextualize the forcing magnitude.

• Global-ocean mean radiative forcing from IMO2020-induced marine low cloud dimming is +0.2 ± 0.11 W m−2. Component means over the global ocean are approximately: Twomey effect 0.077 W m−2 (~40% of total), LWP adjustment ~0.004 W m−2 (near neutral), and cloud fraction adjustment 0.114 W m−2 (~60% of total).
• Strong regional heterogeneity: The North Atlantic exhibits the largest forcing with peaks ~1.4 W m−2 and basin-wide mean ~0.56 W m−2; notable forcing also in the North Pacific and South Atlantic, consistent with shipping intensity and low-cloud prevalence.
• Modeled aerosol changes: Mean modeled Na reduction is ~0.5 cm−3 globally, with regional reductions up to ~3 cm−3 in the North Atlantic, Caribbean, and South China Sea. The ΔNd due to IMO2020 can reach >50% of the total anthropogenic change (PI to present) in parts of the tropical North Atlantic; ΔAOD peaks ~0.01 along major shipping lanes.
• Temperature response: Using the energy balance model, F ≈ 0.2 W m−2 implies ~0.16–0.17 K equilibrium warming with ~7-year timescale, an added warming rate ~0.24 K/decade. This could more than double the recent historical warming rate since 1880 and boost the 2020s warming above the 1980-onward trend; observed record warmth in 2023 falls within expected trajectories. The 2020s mean temperature anomaly is projected to be ~0.3 K higher than the 2010s.
• Planetary energy balance: The 0.2 W m−2 cloud-dimming forcing is nearly 80% of the measured increase (~0.25 W m−2) in planetary heat uptake since 2020. CERES TOA net radiation trend increased from 0.46 to 0.67 W m−2 per decade post-IMO2020, a difference (~0.21 W m−2) consistent with the estimated forcing.
• Hemispheric contrast: The forcing induces substantial interhemispheric contrast (~0.22 W m−2), with NH ~0.32 and SH ~0.10 W m−2, comparable to observed increases (~0.2 W m−2) in NH–SH absorbed solar radiation contrast since 2020.
• Seasonality: In the North Atlantic, monthly total forcing varies by roughly a factor of two (~0.19–0.38 W m−2). After removing insolation seasonality, variations in ΔNa are the largest contributor (>30% in some months), with background Nd and CF also contributing via impacts on Twomey and macrophysical adjustments.
• Consistency with independent analyses: Lane-scale forcing estimates in the South Atlantic agree well with independent observational approaches, and multiple modeling groups report similar global forcing magnitudes (~0.1 W m−2 order).

The analysis demonstrates that the abrupt reduction in shipping sulfur emissions under IMO2020 constitutes a termination shock to an inadvertent geoengineering experiment, leading to a significant positive radiative forcing through marine low-cloud dimming. The magnitude and spatial distribution of the forcing explain part of the observed acceleration in planetary heat uptake and recent surface warming, particularly during 2023, and suggest anomalously warm conditions in the 2020s. The strong hemispheric asymmetry and regional heterogeneity in forcing imply potential impacts on precipitation patterns and regional climate (e.g., enhanced North Atlantic SST warming), highlighting a key consideration for any deliberate MCB scheme: minimizing interhemispheric and regional contrasts to avoid disrupting monsoons and hydrological regimes. The seasonal analysis clarifies the roles of ΔNa, background Nd, and CF in modulating the forcing, while comparisons with CERES TOA flux trends and hemispheric absorbed shortwave changes provide observational context consistent with the modeled forcing. Overall, findings support the feasibility of MCB to temporarily cool the climate but emphasize challenges due to inherent spatiotemporal variability and the need for careful design to manage unintended consequences.

IMO2020 produced a positive radiative forcing of +0.2 ± 0.11 W m−2 via marine low-cloud dimming, with strong spatial and seasonal heterogeneity and a pronounced interhemispheric contrast. This shock is expected to substantially increase warming rates during the 2020s, contributing to observed record warmth and increased planetary heat uptake. The results provide guidance for solar radiation modification via MCB, indicating potential effectiveness but underscoring the need to minimize hemispheric and regional contrasts and to account for background cloud variability. Future research should employ coupled climate models with robust aerosol–cloud interactions to better quantify regional SST responses and feedbacks, refine observational constraints on ΔNa and ΔNd, and assess the broader hydrological impacts and policy trade-offs between air quality improvements and additional warming.

Key uncertainties include: (1) Magnitude of aerosol/droplet number changes—ΔNa is modeled (annual mean ~0.5 cm−3) and not directly constrained by observations; counterfactual satellite-based constraints may improve with longer post-2020 records. (2) LWP and CF adjustments, while derived from large ship-track samples, have limitations and may depend on variables beyond those parameterized; applying relationships to regions with fewer ship-tracks adds uncertainty. (3) Forcing calculations use surface downwelling SW as a proxy for cloud-top values, likely underestimating forcing; semi-direct effects of absorbing aerosols are not explicitly included. (4) Systematic differences between modeled and observed Nd necessitated scaling of ΔNd; regional discrepancies can be up to ~30%. (5) The analysis does not include feedback processes (e.g., low-cloud feedbacks to ocean warming), which require coupled climate models. (6) The inadvertent nature of IMO2020 implies radiative forcing per unit sulfur reduction (~0.2 W m−2 for ~3.7 Tg S reduction) is not optimized; emitted particle properties and spatial emission patterns affect efficiency. (7) Short observational period since 2020 limits attribution of long-term trend changes in energy balance and hemispheric contrasts.