Radiative forcing geoengineering under high CO₂ levels leads to higher risk of Arctic wildfires and permafrost thaw than a targeted mitigation scenario

The Paris Agreement set targets to limit global mean temperature increase to well below 2 °C, ideally 1.5 °C, but current national policies leave a substantial mitigation gap. Geoengineering is discussed as a temporary means to offset anthropogenic warming while emissions are reduced. Two main classes exist: Carbon Dioxide Removal (CDR), which addresses atmospheric CO₂ directly but is slow to implement at scale, and Radiative Forcing Geoengineering (RFG), which can be deployed relatively quickly by altering Earth’s radiation budget. Three RFG methods considered are Cirrus Cloud Thinning (CCT), Marine Cloud Brightening (MCB), and Stratospheric Aerosol Injection (SAI). While RFG can cool the climate, it does not reduce atmospheric CO₂, leaving CO₂-driven plant physiological forcing (e.g., stomatal closure and CO₂ fertilization) unabated. The Arctic, already undergoing amplified climate change and tightly coupled to global systems via teleconnections, may respond differently to RFG than the global mean, particularly for extremes. This study investigates how RFG applied to reduce RCP8.5 radiative forcing to RCP4.5 levels by 2100 affects Arctic mean and extreme temperatures, the surface energy balance, and implications for wildfires and permafrost.

Prior modeling studies suggest RFG can limit global warming (e.g., keeping global mean warming below 2 °C) but can also produce undesirable side effects, including altered precipitation and monsoons, crop yield impacts, ocean productivity changes, enhanced ocean acidification, and dryland expansion. Work focusing on the Arctic has mainly examined SAI impacts on sea ice and permafrost, finding partial remediation of sea ice with sustained SO₂ injection and slowed permafrost warming under SAI, though not to RCP4.5 levels. Equatorially focused SAI and MCB tend to cool the Arctic less than the global mean due to their latitudinal application, leading to undercooling at high latitudes. CO₂ physiological forcing can remotely warm the Arctic via enhanced heat transport, potentially influencing extremes and processes like wildfires and permafrost thaw. The combined effects of RFG with unabated CO₂-driven physiological forcing in the Arctic remain insufficiently explored, motivating this study.

The study uses the fully coupled, emissions-driven Norwegian Earth System Model NorESM1-ME (CMIP5, GeoMIP participant), at ~1.9°×2.5° resolution with 26 vertical levels. The land component is CLM4 with prescribed, transient vegetation (24 plant functional types), and stomatal conductance modeled via Ball-Berry. Simulations include RCP8.5 and RCP4.5 (single runs, 2006–2100), and three RFG scenarios (CCT, MCB, SAI) each with three ensemble members (2020–2100), designed to reduce top-of-atmosphere radiative forcing from RCP8.5 to ~RCP4.5 by 2100 (~−4 W m⁻²). Implementations: (i) CCT increases cirrus ice crystal fall speed (for T < −38 °C) up to 10× by 2100 to thin cirrus and increase outgoing longwave radiation; (ii) MCB increases sea-salt aerosol emissions (accumulation mode, dry modal radius ~0.13 μm) uniformly over oceans between 45°S–45°N to enhance cloud albedo, with fully prognostic CDNC and sea-salt; (iii) SAI injects SO₂ at ~20 km near the equator, evolving via interactive aerosol microphysics (HAM), with increasing injection rates to offset 4 W m⁻² by 2100. Analyses focus on land north of 50°N and 65°N. Temperature metrics: annual mean Tmean, JJA monthly maximum temperature Txx, and DJF monthly minimum temperature Tnn, computed from daily 2-m temperatures as monthly means then annual statistics; anomalies referenced to RCP8.5 2006–2026. Energy budget components include transpiration (canopy), latent and sensible heat fluxes, clear-sky and all-sky shortwave/longwave fluxes, albedo (from clear-sky SW up/down), precipitation, soil moisture and evaporation. Linear regressions of grid-cell variables versus global mean temperature (Tglob) assess sensitivity; differences in slopes between RFG and RCP4.5 indicate altered responses, with unpaired t-tests evaluating significance. Fire-relevant analysis links observed 2001–2020 MODIS burned area (MCD64) to CRU TS4.01 JJA Txx and Tmean anomalies, and compares 2080–2100 simulated Txx–Tmean frequency shifts under CCT and MCB relative to RCP4.5. Permafrost analysis uses ESA Permafrost_cci (2003–2019) to relate permafrost fraction to DJF Tnn and Tmean, then evaluates shifts in simulated DJF Tnn–Tmean frequency under SAI vs RCP4.5 and estimates permanently frozen area (Tmean < 0 °C for ≥24 months) in 2080–2100.

• All three RFG methods reduce global mean temperatures close to RCP4.5 by 2100, but Arctic temperatures remain higher than under RCP4.5. End-century global near-surface air temperatures are slightly above RCP4.5: CCT +0.28 °C, MCB +0.25 °C, SAI +0.40 °C. MCB and SAI achieved −4 W m⁻² radiative forcing; CCT achieved −3.8 W m⁻². • Mean land temperature north of 50°N is higher than RCP4.5 by: CCT +1.01 °C, MCB +0.94 °C, SAI +0.88 °C. In the Arctic (>65°N), mean land temperature is higher by: CCT +0.91 °C, MCB +0.74 °C, SAI +1.05 °C. • Relative to the 2006–2026 baseline (RCP8.5), the Arctic is 4.4–4.7 °C warmer by 2080–2100 under all RFG methods—0.7–1.1 °C more than under RCP4.5. • Extremes: North of 50°N, JJA maximum temperature (Txx) anomalies are higher under CCT (+1.62 °C) and MCB (+1.48 °C) vs RCP4.5 by 2100. DJF minimum temperature (Tnn) increases more under SAI: +1.49 °C north of 50°N and +1.81 °C north of 65°N vs RCP4.5 by 2100. • Sensitivity to global warming: Under CCT and MCB, Tmean and especially Txx increase more per °C of Tglob than in RCP4.5 across mid–high northern latitudes. SAI exhibits the strongest increase in Tnn per °C of Tglob in the Arctic, indicating winter minimums are less controlled. • Energy budget: All RFG scenarios show reduced transpiration per unit Tglob increase relative to RCP4.5 in high latitudes (indicative of enhanced stomatal closure and higher Bowen ratio). In 2100, sensible heat increases under CCT and MCB (supporting higher boundary-layer temperatures and heatwaves), whereas SAI shows reduced sensible heat vs RCP4.5 despite an increased sensible/latent ratio per warming, due to radiative effects. • Shortwave radiation and albedo: SAI produces a significant decrease in incoming clear-sky shortwave at the surface north of 50°N, cooling the surface and maintaining summer albedo close to RCP4.5; CCT and MCB reduce albedo (linked to reduced summer snow cover), amplifying warming and Txx. • Wildfire conditions: Observations show larger burned area when Txx anomalies exceed Tmean anomalies. Simulations for 2080–2100 show CCT and MCB shift temperature frequency toward higher Txx relative to Tmean (mean Txx–Tmean anomaly difference ≈0.69–0.70 vs ≈0.06 in RCP4.5), indicating conditions more conducive to Arctic fire activity. Under MCB, many Arctic regions (e.g., central/eastern Siberia) also experience decreased precipitation and soil moisture, further elevating fire risk. • Permafrost: Under SAI, DJF Tnn–Tmean frequency shifts toward higher Tnn at given Tmean, especially around Tmean ≈ −10 °C where permafrost fraction is sensitive, making conditions less favorable for permafrost. Permanently frozen area (Tmean < 0 °C for ≥24 months) declines by 7.78% under SAI vs RCP4.5 (from 2,120,064 km² to 1,955,148 km²) by 2080–2100. • Overall, meeting a 2 °C global mean target via these RFG designs leads to worse Arctic extreme temperature outcomes than achieving the same target through emissions mitigation (RCP4.5).

The study set out to determine whether RFG that restores global mean temperatures toward RCP4.5 under a high-CO₂ pathway can also adequately protect the Arctic from extreme warming and associated terrestrial risks. Results demonstrate that despite similar global mean temperatures, Arctic mean and extreme temperatures are less controlled under all three RFG methods than under RCP4.5, with method-dependent differences: CCT and MCB amplify summer maximum temperatures, while SAI inadequately cools winter minima at high latitudes. The analysis links these effects to CO₂-driven plant physiological forcing (reduced transpiration, higher Bowen ratio) that persists under RFG because atmospheric CO₂ remains high, alongside radiative and albedo changes specific to each method. These altered thermal regimes translate into increased likelihood of wildfire-favorable conditions under CCT and MCB and more permafrost-unfavorable winter conditions under SAI, implying heightened carbon-cycle risks from fires and thaw. The findings underscore that managing global mean temperature with RFG does not equivalently manage high-latitude extremes and cryosphere stability, highlighting the importance of emissions mitigation and careful assessment of regional impacts and feedbacks in any consideration of RFG.

Applying CCT, MCB, and equatorial SAI to reduce radiative forcing from RCP8.5 toward RCP4.5 by 2100 cools the global mean but leaves the Arctic warmer than under RCP4.5, with more severe extremes. CCT and MCB particularly elevate summer maximum temperatures, increasing fire-conducive conditions, while SAI raises winter minima, reducing permafrost-favorable conditions and shrinking permanently frozen area by ~7.8% relative to RCP4.5. Differences arise from the interplay of unabated CO₂ physiological forcing and method-specific radiative effects (e.g., shortwave reductions and albedo). These results indicate that using RFG to meet a 2 °C target can produce less favorable Arctic outcomes than achieving the target via emissions mitigation. Future work should assess multi-model robustness, alternative SAI injection strategies (including high-latitude targeting), interactions with AMOC variability, explicit quantification of physiological forcing contributions, higher temporal resolution for fire weather indices, and scenarios combining mitigation with limited-duration or regionally optimized RFG.

• Single-model dependency: Results rely on NorESM1-ME; responses, especially to CCT, can be sensitive to parameter choices (e.g., cirrus ice fall speed) and model physics. • AMOC bias: NorESM1 simulates a relatively strong AMOC; actual AMOC states could alter Arctic warming patterns and the relative efficacy of RFG. A weaker AMOC state might change projected Arctic temperatures across scenarios. • Physiological forcing not isolated: The experimental design was not tailored to isolate plant physiological forcing; changes in transpiration and energy partitioning may also reflect precipitation and temperature shifts. • Spatial and temporal limitations: Most geoengineering experiment outputs are monthly; daily data would better support fire weather index and extreme-event analyses. • Injection strategy specificity: SAI was equatorial; alternative strategies (multiple tropical or high-latitude injections) could mitigate high-latitude undercooling but introduce other regional impacts. • Arctic projection uncertainties: High-latitude climate feedbacks and potential nonlinear interactions with combined CO₂ and aerosol forcings introduce additional uncertainty. • Cost and deployment duration uncertainties: Practical RFG costs (especially for CCT/MCB) and the need for sustained deployment without concurrent CO₂ reduction pose socioeconomic and risk challenges.