Anthropogenic warming has exacerbated droughts in southern Europe since the 1850s

Recent decades have seen unusually severe and widespread summer droughts across Europe, with major socioeconomic, agricultural, and ecosystem impacts. Southern Europe is particularly vulnerable to drought at annual to decadal scales, while central Europe serves as a transition zone but has also experienced drying in the last century. Climate projections identify Europe as a hotspot for future high-intensity droughts and heatwaves, especially in southern and central regions. However, due to limited instrumental records, the timing of the onset of widespread drying and the role of anthropogenic greenhouse gas–induced warming relative to natural variability remain uncertain. To address these questions, the study aims to reconstruct warm-season hydroclimate over the last 300 years using a drought-sensitive tree-ring oxygen isotope (δ18O) record from Bosnia and Herzegovina, calibrate it against April–August SPEI, and assess the influence of anthropogenic forcing on land–atmosphere coupling using CMIP6 model experiments and reanalysis data.

Multiple proxy records have documented European hydroclimatic variability over past centuries. The Old World Drought Atlas (OWDA) highlights multidecadal drought and wetness but mixes seasonal signals, limiting summer-specific interpretations. Tree-ring isotope networks (e.g., ISONET) and regional reconstructions have identified trends toward hotter, drier summers in parts of Western and Eastern Europe and unusual recent hydroclimate conditions. Previous work links European drought variability to large-scale modes such as the North Atlantic Oscillation (NAO), with more positive phases since the mid-19th century linked to anticyclonic conditions over central/southern Europe and cyclonic conditions in the north, fostering a north–south dry–wet dipole. Land–atmosphere feedbacks, particularly soil moisture–temperature coupling, have been implicated in exacerbating drought and heat extremes. Projections suggest increased drought frequency/severity under continued warming, emphasizing the need for long-term, season-specific reconstructions to place recent trends in context.

- Study area and sampling: Tree-ring material was collected from European black pine (Pinus nigra) at two rainforest sites (Lom/Klekovača and Janj/Stolovaš) in the Dinaric Alps, Bosnia and Herzegovina (44.15°–44.46°N, 16.47°–17.28°E; 1115–1250 m a.s.l.). In October 2019, one increment core at breast height was taken from each of 24 trees. - Isotope preparation and measurement: After air-drying and sanding, ring widths were measured and cross-dated (COFECHA). Five well-behaved cores (wider rings, minimal disturbances) were selected for stable isotope analysis. Whole-wood α-cellulose was extracted using a modified Jayme–Wise method. Annual α-cellulose from each ring (0.09–0.25 mg) was analyzed for δ18O with a TC/EA coupled to an IRMS (Thermo Delta V Advantage). Total of 1,582 samples measured; analytical uncertainty ±0.15‰ (n=185). - Chronology construction and quality: Individual δ18O series were normalized over their common period and combined to form a regional chronology. Signal strength was assessed via mean inter-series correlation (Rbar) and expressed population signal (EPS). The most reliable period with EPS>0.85 is 1740–2019, with 1711–1739 retained to evaluate long-term trends despite single-core coverage. - Climate data and sensitivity analyses: Gridded climate datasets used include CRU TS4.05 temperature/precipitation and PDSI, GLDAS soil moisture (0–10 cm), NCEP/NCAR reanalysis moisture fluxes, and gridded SPEI (0.5°). Spatial correlation analyses established the regional representativeness over southern Europe (10°–25°E, 35°–45°N). Pearson correlations were computed between δ18O and monthly climate variables and multi-month SPEI (April–August), including first-differences. - Reconstruction: April–August SPEI was reconstructed from the δ18O chronology using linear regression with split-period calibration/verification (1901–1959; 1960–2019). Model performance metrics included r, R2, RE, CE, sign test, and Durbin–Watson statistics. Extreme dry/wet years were identified using ±1σ thresholds. Regime shifts were detected using a 10-year cutoff at 95% confidence. - Comparative reconstructions and circulation indices: Comparisons made with the ISONET gridded July–August SPEI reconstruction and a reconstructed winter NAO index to explore spatial patterns and potential drivers (e.g., NAO, North Atlantic Jet). - Model experiments and coupling diagnostics: Nine CMIP6 models provided monthly soil moisture, sensible heat flux, and 2 m temperature for experiments with natural-only (NAT), greenhouse gas (GHG), and historical (ALL) forcings (1850–2019/2020). Land–atmosphere coupling strength was quantified using the two-legged ISM-T index: ISM-T = ρ(SM, SH) × ρ(SH, T) × σ(T). ANOVA tested differences in coupling between forcings. - Reanalysis-based coupling: ERA5-Land (0.1°) data (1959–2018) provided near-surface temperature, net radiation, actual latent heat flux, and potential evaporation to compute a diagnostic coupling metric π = e′ × T′ quantifying soil moisture–temperature coupling. LOESS regression estimated trends; standardized Z scores facilitated comparisons. Composite analyses contrasted coupling during southern-only drought years versus concurrent southern and central European droughts.

- The five-tree δ18O series are coherent (Rbar = 0.66); the composite chronology shows strong common signal (EPS = 0.89). Mean δ18O values range 28.8‰–29.4‰. - Tree-ring δ18O exhibits broad spatial sensitivity to April–August hydroclimate across southern Europe and correlates strongly with April–August SPEI during 1901–2019 (r = −0.70, p < 0.001; early period r = −0.72; late period r = −0.71). First-difference correlations remain high with temperature (r = 0.45), precipitation (r = −0.57), and SPEI (r = −0.59), all p < 0.001. - The April–August SPEI reconstruction (1711–2019) explains 49.5% of instrumental variance, identifies two wet (1711–1778; 1813–1849) and two dry (1779–1812; 1850–2019) intervals, and reveals a persistent drying trend starting in the 1850s that continues to present. - Recent decades show unprecedented drought severity in southern Europe over the past 300 years, with no wet years since 1976 by the ±1σ criterion. - A dipole-like pattern in European hydroclimate is evident: southern Europe drying correlates negatively with northern Europe wetting, consistent with NAO influences. The ISONET gridded reconstruction shows positive correlation with southern and negative with northern Europe; NAO correlates negatively with southern European SPEI and positively with northern European SPEI. - Anthropogenic forcing strengthens land–atmosphere coupling: In CMIP6, coupling strength (ISM-T) under GHG and ALL forcings is significantly stronger than under NAT, with coupling differences between GHG and NAT of ~58% (southern Europe) and 59.1% (central Europe). Correlations between modeled sensible heat flux and temperature are higher under GHG/ALL than NAT. Modeled soil moisture shows decreasing trends under GHG consistent with reconstructed drying. - ERA5-Land diagnostics indicate increasing summer soil moisture–temperature coupling strength across most of Europe since 1959, with higher coupling in southern than central Europe and marked enhancement since the 2000s. Coupling is stronger during years when southern and central Europe experience concurrent droughts compared to southern-only drought years. - Relationships with large-scale dynamics: Southern European drought correlates with reconstructed NAJ variability, but central Europe shows no stable NAJ relationship, suggesting complex circulation influences in the transition zone.

The reconstruction establishes that the widespread drying across southern Europe began around the 1850s, aligning with the onset of substantial anthropogenic greenhouse gas emissions. The strong negative correlation between tree-ring δ18O and warm-season SPEI validates δ18O as a robust proxy for spring–summer water balance. The dipole pattern—drying in southern/central Europe and relative wetting in northern Europe—is consistent with more frequent positive NAO phases since the mid-19th century, which promote anticyclonic, warm–dry conditions over southern/central Europe and cyclonic, wetter conditions in the north. Beyond circulation, model and reanalysis evidence indicate that anthropogenic warming has intensified land–atmosphere coupling via soil moisture–temperature feedbacks, enhancing evaporation, reducing soil moisture, increasing sensible heat flux, and further elevating near-surface temperatures. Stronger coupling under GHG/ALL forcings and its recent acceleration match the observed/reconstructed drought intensification, especially when droughts occur simultaneously in southern and central Europe. While NAJ variability relates to southern European drought, the lack of a stable relationship for central Europe underscores the complex interplay of circulation and land-surface processes in the transition zone. Collectively, the findings show that anthropogenic warming has exacerbated European droughts since the 1850s by strengthening land–atmosphere coupling and modulating circulation-linked dipole patterns.

This study provides a 309-year reconstruction of warm-season drought in southern Europe from tree-ring δ18O that reveals a persistent drying trend beginning in the 1850s and unprecedented drought severity in recent decades. The analysis links this trend to anthropogenic warming through enhanced land–atmosphere coupling, supported by CMIP6 experiments (stronger coupling under GHG/ALL versus NAT) and ERA5-Land diagnostics (increasing coupling since 1959, intensified since the 2000s). A north–south dry–wet dipole consistent with NAO dynamics further contextualizes regional differences. These results imply continued risk of widespread, concurrent European droughts under ongoing warming. Future work should expand season-specific proxy networks across Europe, integrate multiple tree-ring metrics (e.g., latewood cellulose, additional species), and combine reconstructions with high-resolution observations to refine SPEI magnitude estimates and assess ecosystem impacts. Incorporating process-based modeling and data assimilation could improve attribution and prediction of drought under various forcing scenarios.

- Early portion of the chronology (1711–1739) is based on a single core; the most reliable interval with EPS > 0.85 is 1740–2019, which may limit confidence in trends before 1740. - The isotope chronology is derived from five trees, which, despite strong common signal (EPS 0.89), may underrepresent site heterogeneity. - Use of whole-wood α-cellulose from Pinus nigra emphasizes spring–summer signals and differs from latewood-only approaches in other species (e.g., oak), potentially contributing to discrepancies with other reconstructions. - Coherence with the ISONET gridded July–August SPEI reconstruction is weak before 1900, indicating spatial/seasonal and methodological differences in reconstructions. - The study did not perform explicit future projections; inferences about future drying rely on historical reconstructions, reanalysis, and CMIP6 historical-forcing experiments.