Atmospheric CO2 forcing on Mediterranean biomes during the past 500 kyrs

Mediterranean forests are biodiversity-rich ecosystems that operate under water-stressed conditions and provide vital ecosystem services. With ongoing anthropogenic CO2 emissions and projected increases in drought frequency and severity over the next decades, concern is rising regarding the resilience of Mediterranean forest biomes. Climate scenarios suggest structural and compositional changes unprecedented in the Holocene, threatening ecosystem services and biodiversity. To assess resilience to future forcing, the study aims to determine the nature, rates, and mechanisms of past changes in Mediterranean vegetation under the full range of Quaternary climate variability. The Tenaghi Philippon (Greece) peat archive offers a uniquely long, continuous, and high-resolution terrestrial record spanning the past 500 kyrs, sensitive to abrupt climate change and representative of vegetation near lower moisture-tolerance limits. The research question is whether atmospheric CO2 variability causally influenced Mediterranean vegetation, particularly via moisture availability, and whether thresholds in precipitation governed abrupt biome shifts.

Previous work identifies Mediterranean forests as vulnerable to drought and climate change, with anticipated biome reorganization and biodiversity loss under future scenarios. Past studies at Tenaghi Philippon and nearby refugial sites (e.g., Ioannina, Lake Ohrid) documented rapid vegetation responses during glacial and interglacial periods and highlighted the sensitivity of these systems to climatic forcing. Interglacial dynamics, orbital pacing, greenhouse gas variability, North Atlantic temperature, ITCZ shifts, and methane have been implicated in regional hydroclimate variability. However, the lack of paired high-resolution precipitation and vegetation proxies from a single site has limited validation of hypothesized moisture thresholds and causal mechanisms linking CO2 to vegetation via hydroclimate.

Data and site: A composite sequence from three drillcores at Tenaghi Philippon (NW Greece) provides a continuous ~500 kyr record. The study interval (4–91 m composite depth) excludes the disturbed uppermost 4 m. Lithologies include predominantly fen peat with intervals of organic-rich mud and lake marl. Pollen analysis: 2347 samples, 2–6 cm spacing (mean temporal resolution ~209 years), with new counts for MIS 6–7 and MIS 9–13 and augmented counts for MIS 1–5 and 8. Standard palynological preparation and counting protocols were used; percentages calculated on terrestrial pollen sums excluding Cyperaceae/Poaceae for main percentage base. Geochemistry and precipitation proxy: XRF core scanning (average temporal resolution ~32 years) produced elemental counts; PCA on Al, Ca, Fe, K, Mn, P, S, Si, Ti identified components distinguishing carbonate-rich (high groundwater) vs siliciclastic-rich intervals. The log(Ca/Fe) ratio was adopted as a local precipitation proxy: higher values indicate higher groundwater table and precipitation via karst-fed Ca input; lower values indicate reduced karst activity and enhanced Fe-bearing siliciclastic input under drier conditions. Chronology: Age model integrates 67 AMS 14C dates, 10 dated (crypto)tephras (out of 57 detected), and cyclostratigraphic analysis of log(Ca/Fe) showing eccentricity (~80–106 kyr) and obliquity (~33–40 kyr) pacing. The log(Ca/Fe) was tuned to an ice-sheet model, aligning maxima/minima with minima/maxima in global ice volume. Resulting age uncertainties are <5 kyr, supported by tephra tie points. Statistical analyses: Rolling correlations explored time-varying relationships among log(Ca/Fe) and GHGs. Convergent cross-mapping (CCM) using rEDM assessed causality among tree-pollen, log(Ca/Fe), and external climate proxies (Antarctic CO2, CH4; North Atlantic SST proxy N. pachyderma sinistral at ODP 983; Eastern Mediterranean Ti/Al and Ba/Al from ODP 967/968 as dust and ITCZ indicators; orbital precession and obliquity). All records were linearly interpolated to 1-kyr steps. Optimal embedding parameters (E, τ) were chosen via simplex projection; significance tested against 100 phase-randomized surrogates with ROC/AUC evaluation. Time-delayed CCM with ±10 kyr prediction horizons identified optimal lags, distinguishing direct causation from synchrony and estimating response times. Generalized additive models (GAM) characterized the non-linear tree-pollen vs log(Ca/Fe) relationship. Spectral analysis: REDFIT and evolutive harmonic analysis identified orbital-scale periodicities in log(Ca/Fe).

• Two dominant vegetation regimes alternated over the past 500 kyrs at Tenaghi Philippon: steppe (tree-pollen <35%) during glacials and forest (tree-pollen >65%) during interglacials, with rapid reorganizations at terminations and frequent abrupt oscillations during interglacial declines. Regime shifts typically lasted <200 years, and decadal-scale shifts are indicated in higher-resolution intervals.
• Terminal interglacial forests were increasingly drought-tolerant and pine-dominated, with declines lagging peak interglacial warmth (MIS 5e, 7e, 9e, 11c) by ~5–20 kyr, implying resilience to temperature decreases and pointing to moisture availability as the key determinant of forest persistence during glacial inception.
• The log(Ca/Fe) precipitation proxy peaks during interglacial maxima and declines towards glacials, evidencing gradual moisture reduction, while forest contractions are abrupt, indicating threshold behavior.
• CCM demonstrates significant causal association between precipitation (log(Ca/Fe)) and forest cover (tree-pollen). Time-delayed CCM shows precipitation changes lead vegetation responses rapidly: tree-pollen responds in <1 kyr to precipitation changes, whereas any feedback from vegetation to precipitation is much slower (>4.5 kyr), supporting moisture-threshold driven regime shifts.
• Among external drivers, only atmospheric CO2 shows causal effects on both precipitation and vegetation at Tenaghi Philippon, robust to ±5 kyr age-model perturbations. CH4 causally affects precipitation but not vegetation directly; no significant causality was found for orbital precession/obliquity, ITCZ migration proxies (Ba/Al), or dust (Ti/Al). A tentative link with North Atlantic temperature was sensitive to chronology and not retained.
• Non-linear, bimodal relationship between tree-pollen and log(Ca/Fe) indicates alternative stable states: forests prevail above log(Ca/Fe) ~0.8; steppe dominates below ~0.25; regime shifts occur around a threshold at log(Ca/Fe) ≈ −0.5. This threshold corresponds to ~400 mm/year annual precipitation required for subtropical dry forest biomes in the Mediterranean.
• Both biomes can persist within ±0.3 units around the threshold, indicating multiple equilibria influenced by migration, longevity, and community dynamics.
• Peak precipitation and maximum forest cover occurred persistently when CO2 >260 ppmv (interglacial conditions).
• Based on full-interglacial conditions at Tenaghi Philippon (log(Ca/Fe) >1.5; forests maintained above 0.8), a minimum 40–45% decrease in annual precipitation from interglacial maxima could trigger a forest-to-steppe regime shift under natural CO2 ranges. Projected near-future CO2-driven warming of ~2°C could reduce Mediterranean annual precipitation by up to ~30%, raising concern for crossing the moisture threshold, compounded by reduced runoff and summer soil moisture.

The findings provide empirical evidence that atmospheric CO2 influences Mediterranean vegetation primarily by modulating regional moisture availability, with a secondary direct physiological pathway. The high-resolution, paired vegetation and precipitation proxies reveal that abrupt forest-to-steppe transitions are triggered when precipitation crosses a threshold, not by gradual temperature decline. The lagged forest demise relative to peak interglacial warmth underscores temperature resilience, while CCM results establish precipitation as the proximate driver of rapid vegetation responses. The identification of CO2 as the only external driver with robust causal links to both precipitation and vegetation implies that anthropogenic CO2 increases can influence Mediterranean biomes through hydroclimate changes, consistent with model projections of drier conditions. The non-linear, bimodal vegetation-moisture relationship and the quantified threshold (~400 mm/yr; log(Ca/Fe) ≈ −0.5) demonstrate the potential for rapid regime shifts once resilience is eroded below critical moisture levels. Given projected precipitation declines of up to ~30% and increased water demand/evapotranspiration, Mediterranean forests may approach or cross the moisture threshold, risking abrupt biome shifts and associated loss of ecosystem services.

Using a 500-kyr, high-resolution record from Tenaghi Philippon, the study shows that Mediterranean forest dynamics are governed by moisture thresholds, with abrupt regime shifts between forest and steppe occurring when precipitation falls below a critical level. Convergent cross-mapping establishes causality between CO2 and both precipitation and vegetation, indicating both an indirect climatic pathway and a direct physiological influence. Forests at Tenaghi Philippon remained resilient to temperature declines but were vulnerable to moisture deficits. The quantified precipitation threshold (~400 mm/yr; log(Ca/Fe) ≈ −0.5) and the estimated 40–45% decline from interglacial maxima required to trigger collapse provide benchmarks for assessing near-future risk. With projected anthropogenic drying up to ~30% and additional hydrological stress, an abrupt forest-to-steppe shift in parts of the Mediterranean appears plausible, underscoring the need for conservation and water management strategies that enhance forest resilience. Future research should refine regional thresholds across sites, integrate physiological responses under non-analog CO2 levels, and couple paleo-informed thresholds with process-based ecosystem and climate models to improve forecasts of biome stability.

Age-model uncertainties (though constrained to <5 kyr) and chronological uncertainties in external proxy records can affect causality inference; sensitivity tests indicate robustness for CO2 links but ambiguity for North Atlantic temperature forcing. The precipitation proxy relies on log(Ca/Fe) as a local hydroclimate indicator tied to karst hydrology; while well supported by geochemical and mineralogical evidence and PCA, it remains an indirect measure with potential site-specific influences. The analysis centers on a single site close to lower moisture-tolerance limits; broader generalization across the heterogeneous Mediterranean requires multi-site validation. CCM requires interpolation to 1-kyr steps and assumes deterministic dynamics; bidirectional coupling and potential synchrony with unobserved variables cannot be entirely excluded. Future projections involve non-analog CO2 concentrations; plant physiological acclimation may partially offset moisture stress, introducing uncertainty when transferring Quaternary relationships to future conditions.