Transient vegetation degradation reinforced rapid climate change (RCC) events during the Holocene

The study addresses why rapid climate change (RCC) events punctuated the otherwise relatively warm and stable Holocene and seeks mechanisms explaining their timing and hemispheric expression. Previous work identified eight Bond events (11.1, 10.3, 9.4, 8.1, 5.9, 4.2, 2.8, 1.4 ka BP) with ~1500-year spacing, and additional events (~7.2 ka and Little Ice Age, ~0.4 ka). RCCs were characterized by NH cooling, weakened tropical monsoons, and low-latitude drying, with major ecological and societal impacts across Eurasia. While many studies link RCCs to North Atlantic processes and AMOC weakening, skepticism persists, as some events (e.g., 4.2 ka) lack freshwater forcing in transient simulations. The authors hypothesize that dynamic terrestrial vegetation changes and their climate feedbacks are a key, previously underappreciated driver of Holocene RCC events, especially in semi-arid transition zones like North China. They test this with Holocene-length transient Earth system model experiments that isolate roles of dynamic vegetation, orbital forcing, and greenhouse gases, and benchmark against high-resolution East Asian proxies and North Atlantic HSG (IRD) records.

Prior research attributes Holocene RCCs to North Atlantic IRD/Bond events and AMOC variability, solar activity, volcanic forcing, hemispheric interactions, and tropical ocean feedbacks, but results are regionally inconsistent and debated. Some simulations suggest key events (e.g., 4.2 ka) are not associated with North Atlantic freshwater input. Proxies from East Asia (stalagmite δ18O, pollen-inferred precipitation) show early to mid-Holocene wetness followed by drying aligned with summer insolation decline, with centennial-scale events at ~8.2 and ~4.2 ka widely recorded, while others are less documented. Vegetation-climate feedbacks (albedo, evapotranspiration, roughness) are known to modulate climate; charcoal records suggest synchronicity between Holocene fires, IRD, and RCCs. Foundational work (Charney) and subsequent studies highlight positive feedbacks between vegetation and precipitation, especially in dryland transition zones. However, comprehensive transient modeling explicitly assessing dynamic vegetation’s role across all Holocene RCCs has been limited.

Modeling framework: Community Earth System Model CESM1.2 with CAM4 (atmosphere), CLM4 (land with DGVM), CICE4 (sea ice), POP2 (ocean), CPL7. Horizontal resolution: atmosphere/land 0.9°×1.25° (26 and 17 vertical layers), ocean/sea ice 0.5°×0.5° (ocean 60 vertical levels). Acceleration: 20-year acceleration (1 model year = 20 calendar years) applied for transient runs after a 600-year spin-up at 14 ka conditions to equilibrate and allow DGVM to establish vegetation from bare ground. Experiments (Holocene-length transients):

• DV (dynamic vegetation): Transient orbital parameters and GHGs; DGVM active; fully coupled atmosphere–land–ocean–sea ice.
• nDV (no dynamic vegetation): Transient orbital and GHG forcing; DGVM off (vegetation fixed at pre-industrial); fully coupled.
• ORB (orbital-only): Transient orbital forcing; GHGs fixed at pre-industrial; DGVM off; fully coupled.
• VF (vegetation forcing): Atmosphere and land active; ocean and sea ice fixed at pre-industrial; DGVM off but terrestrial vegetation PFTs prescribed annually from DV outputs; orbital and GHGs fixed at pre-industrial. Proxy synthesis and RCC definition: Collated four high-resolution East Asian records (Daihai and Gonghai pollen-based annual precipitation; Dongge and Yulong cave stalagmite δ18O) and compared with subpolar North Atlantic hematite-stained grains (HSG) representing Bond events. An RCC event is defined when at least three of the four East Asian proxies indicate a drought event aligned (within dating uncertainties) with Bond events, yielding eight RCCs: 9.2, 8.2, 7.2, 5.2, 4.2, 2.6, 1.2 ka, and the Little Ice Age (LIA). Detection of model RCCs: Applied the KDJ index (J component of Lane’s stochastic oscillator) to annual precipitation time series at each grid cell (global 192×288 grid), with window m=21 years (~420 years in accelerated time) and thresholds J≤20 or ≥80 to identify abrupt drops (droughts) or peaks. K, D, J computed recursively; J is the sensitive indicator. Also used empirical orthogonal function (EOF) and singular value decomposition (SVD) analyses to identify vegetation–precipitation coupled modes in East Asia, and examined radiative (albedo) and non-radiative (latent/sensible heat, Bowen ratio, evaporation) flux changes. Event compositing: For spatial maps, anomalies computed as differences between averages of the 200 years prior to the start and 200 years after the end of each RCC period. Compared DV, nDV, ORB, and VF outputs to isolate roles of dynamic vegetation, orbital forcing, and prescribed vegetation anomalies. Calculated NH and North Atlantic SST and AMO-like indices to evaluate linkage with RCC timings.

• Dynamic vegetation is essential to reproduce Holocene RCCs: The DV experiment captured eight RCC drought events in North China (NC) with sharp precipitation declines at timings consistent with proxies: 9.14, 8.14, 7.26, 5.10, 4.22, 2.62, 1.20, and 0.42 ka BP (KDJ≥20 threshold satisfied). In contrast, nDV captured only a subset (9.2, 5.2, 4.2, LIA with timing offsets), and ORB captured only 8.2 ka correctly (and detected 10.3 ka; a delayed ~8.95 ka event; and a 3.05 ka decline), indicating orbital forcing alone cannot explain all RCCs.
• Vegetation feedback intensifies and times RCCs: The difference DV−nDV isolates vegetation feedback, showing additional precipitation reductions during RCCs. Estimated DV−nDV precipitation decreases: ~40 mm at 9.2, 8.2, and 2.6 ka; ~5.9 mm at 7.2 ka; 26.78 mm (4.2 ka); 50.66 mm (5.2 ka); 67.96 mm (1.2 ka); 43.1 mm (LIA). The VF experiment, driven solely by prescribed vegetation changes with fixed ocean/ice/orbit/GHG, reproduced 6 RCCs (8.2, 7.2, 5.2, 4.2, 1.2, LIA) with precipitation decreases of 20.16, 8.4, 23.45, 35.15, 8.99, and 22.02 mm, respectively; the lowest VF precipitation occurs at 2.3 ka (−48.25 mm), consistent with a vegetation-induced delay.
• Vegetation–precipitation coupling: EOF/SVD analyses over East Asia show strong coupling between total vegetation cover (TotalVeg) and annual precipitation: correlations of 0.62, 0.58, and 0.62 for the first three mode pairs (p<0.01); variance contributions 16.75%, 12.92%, 10.51%, respectively. TotalVeg minima coincide with precipitation minima during RCCs; sensitive PFTs (e.g., temperate broadleaf deciduous shrubs) are reduced, increasing surface albedo.
• Thermodynamic pathways: During RCCs in the wet–dry transition zone, vegetation degradation increases albedo (radiative cooling) and decreases evapotranspiration, shifting turbulent fluxes toward sensible heat (higher Bowen ratio) and reducing evaporation. DV and VF both show surface cooling aligned with droughts in NC for nearly all RCCs (exception: 7.2 ka in VF), indicating vegetation-driven cooling.
• Hemispheric extent: In DV, Northern Hemisphere (NH) land and global precipitation decrease during all RCCs, with pronounced drying from East/Central Asia through North Africa and central–southern North America. Some regions (e.g., North Africa to Iranian Plateau) exhibit >30% precipitation reductions during 9.2, 8.2, 2.6, and 1.2 ka events. NH temperatures also drop during all RCCs in DV, whereas nDV and ORB generally lack synchronous cooling, implicating vegetation dynamics.
• North Atlantic forcing is not sufficient: DV’s NH SST and AMO-like indices do not show consistent cooling for all RCCs; early to mid-Holocene North Atlantic SSTs peak during some RCCs (9.2, 8.2, 7.2, 5.2, 4.2 ka), while late Holocene (2.6, 1.2 ka, LIA) shows cooling. Given model limitations (accelerated coupling and fixed ice sheets), results suggest multi-factor origins of RCCs, with dynamic vegetation a prerequisite for widespread drought/cooling signatures, especially in semi-arid transition zones.
• Detection approach: The KDJ index proved efficient and robust for global RCC detection with low false/miss rates, enabling gridded identification consistent with proxy-defined events.

The simulations demonstrate that including dynamic terrestrial vegetation in an Earth system model is crucial to reproducing the timing, intensity, and hemispheric expression of Holocene RCC events. The DV experiment aligns with multiproxy evidence in East Asia and reveals widespread NH drought and cooling during RCCs, whereas experiments without vegetation dynamics fail to capture most events or their co-occurring temperature drops. This indicates that vegetation degradation and associated land–atmosphere feedbacks (albedo increases, reduced transpiration, altered turbulent flux partitioning) act as a prerequisite and amplifier for RCCs, particularly in semi-arid transition zones like North China. While North Atlantic processes (IRD/AMOC) contributed, especially in the late Holocene, they do not provide a universal explanation; early to mid-Holocene RCCs lack consistent North Atlantic cooling in the model. Therefore, RCC events likely arise from a combination of external drivers (orbital insolation, GHGs, volcanic/solar variability) and internal feedbacks, with dynamic vegetation playing a dominant role in translating relatively smooth forcings into abrupt regional hydroclimate and temperature shifts. The findings underscore the importance of representing vegetation dynamics in paleoclimate simulations and interpreting regional proxy disparities through the lens of vegetation–climate coupling and regional sensitivity.

This study provides a comprehensive transient-model assessment showing that dynamic terrestrial vegetation feedbacks reinforced and effectively orchestrated Holocene RCC events. Using CESM1.2 with active DGVM, the authors captured eight RCCs consistent with East Asian proxies and demonstrated widespread NH drought and cooling during these episodes. Vegetation degradation emerged as the key internal feedback, operating via albedo and evapotranspiration pathways, that enabled abrupt hydroclimate and temperature changes in response to gradual external forcings. The results challenge the notion that North Atlantic cooling alone drove RCCs and call for multi-factor frameworks that include vegetation dynamics. Future work should: (1) incorporate interactive ice sheet dynamics and remove acceleration to better resolve ocean–ice–atmosphere interactions; (2) examine seasonal and regional insolation anomalies and their coupling with vegetation; (3) investigate model component sensitivities (e.g., land–atmosphere coupling strength) and PFT-specific responses; (4) expand event-by-event regional analyses with additional high-resolution proxies; and (5) refine detection and attribution methods for abrupt changes in transient simulations.

• Accelerated coupling: A 20-year acceleration was applied, limiting accurate representation of surface–deep ocean energy exchanges and potentially affecting North Atlantic SST variability and AMOC-related teleconnections.
• Fixed cryosphere: High-latitude ice sheets were held at pre-industrial conditions, excluding iceberg calving/freshwater forcing, likely contributing to the failure to capture early Holocene (11.1 and 10.3 ka) Bond events and some North Atlantic cooling signatures.
• Vegetation prescription in VF: While isolating vegetation effects, VF fixed ocean/sea ice/orbit/GHGs at pre-industrial, potentially misrepresenting coupled feedbacks and timing (e.g., delayed 2.6 ka signal to ~2.3 ka).
• Regional/proxy uncertainties: Proxy sensitivities, dating uncertainties, and spatial heterogeneity can affect event alignment and magnitude assessments.
• Model structural uncertainties: Differences between DV and nDV NH precipitation levels suggest influences from other CESM components beyond vegetation; these were not explored in depth.
• Detection method constraints: The KDJ-based detection relies on chosen windows and thresholds; while efficient, it may miss or shift events under different parameter choices.