Introduction

Explosive tropical volcanic eruptions are among the most powerful natural perturbations to the Earth’s climate system, capable of altering global temperature and subsequently driving changes in atmospheric circulation and precipitation patterns from interannual to centennial time scales1,2,3. By injecting vast quantities of sulfate aerosols into the stratosphere, these eruptions increase planetary albedo and reduce solar radiation reaching the surface, thereby weakening the hydrological cycle and monsoons4,5,6. Historical evidence and climate model simulations have linked large eruptions—such as Tambora (1815) and Pinatubo (1991)—to widespread droughts over the Indian subcontinent and northern East Asia, although the magnitude and spatial extent of the response vary across events7,8,9.

High-resolution paleoclimate proxy records provide valuable insights into these regional hydroclimatic responses to volcanic eruptions. Tree-ring-based drought atlases, such as the Monsoon Asia Drought Atlas (MADA), reveal statistically significant monsoon rainfall deficits over South and northern East Asia during the 1–2 years following large volcanic eruptions10,11. Moreover, tree-ring and historical documentary records consistently reveal widespread post-eruption droughts across the central Himalayas12, the Tibetan Plateau13,14, and northern East Asia7,15,16. Despite these robust regional signals, the spatial coherence and dynamical mechanisms underlying pan-Asian monsoon drought responses to volcanic eruptions remain poorly understood.

While the in-phase variability of summer monsoon rainfall between South Asia and northern East Asia is known to be modulated by internal modes such as El Niño/Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD)17,18,19,20,21,22,23—attributing the robust post-eruption synchronization to these canonical internal pathways proves problematic. The robustness of the volcanic–ENSO connection remains debated, appearing tenuous and highly sensitive to initial climate states24. Furthermore, eruptions have been linked to a positive phase of the IOD25,26, a condition typically associated with enhanced—rather than suppressed—South Asian precipitation. Consequently, these canonical dynamical pathways are insufficient to fully explain the observed synchronization of severe droughts across these distant regions.

One key circulation mode that may bridge this explanatory gap is the circumglobal teleconnection (CGT)—a quasi‑stationary Rossby wave train propagating along the upper‑tropospheric westerly jet27,28,29. The CGT has long been recognized as a principal driver of boreal summer hydroclimatic variability across Eurasia, modulating drought and flood risks in East Asia and mid‑latitude regions30,31,32,33,34. CGT-related circulation anomalies can reorganize upper-tropospheric wave activity and modulate heat and moisture transport over East Asia, resulting in spatially varied rainfall responses32,33,35. Although CGT variability has been largely attributed to internal dynamics, whether it can be systematically triggered by external volcanic forcing remains untested.

To address this knowledge gap, we integrate annually resolved tree-ring-based CGT reconstructions with last-millennium climate simulations and linear baroclinic model (LBM) experiments to evaluate whether tropical volcanic eruptions systematically excite CGT-like upper-tropospheric wave responses. Specifically, we test whether volcanically suppressed South Asian convection acts as a Rossby wave source that triggers the negative CGT phase and reinforces drought over northern East Asia. This study reveals a volcanic–CGT–drought linkage that has persisted over the past millennium.

Results

CGT response to large tropical volcanic eruptions

The tree-ring-based CGT reconstruction since 1657 demonstrates high fidelity in capturing upper-tropospheric circulation variability, with robust calibration and validation skill across nested intervals (Supplementary Fig. 1). Analysis of this reconstruction (Fig. 1a) shows that the first boreal summer following large tropical volcanic eruptions consistently corresponds to negative CGT anomalies. In its negative phase, the CGT exhibits anomalous northerly winds at 200 hPa over East Asia. Superposed epoch analysis (SEA; Fig. 1c) further corroborates this response and demonstrates that the negative CGT signal is statistically significant in the first year after the eruptions, reaching a mean anomaly of −0.74.

Fig. 1: Impact of tropical volcanic eruptions on the circumglobal teleconnection (CGT).
Fig. 1: Impact of tropical volcanic eruptions on the circumglobal teleconnection (CGT).
Full size image

a Tree-ring-based reconstruction of the CGT index (black), overlaid with the ERA5-based CGT index (1948–2020; red) for comparison. Shading indicates the ±1 root-mean-square error (RMSE) uncertainty range of the reconstruction. Blue dots mark the first boreal summer following large tropical eruptions. b CGT index from the Community Earth System Model Last Millennium Ensemble (CESM-LME; black), with ensemble spread (±1 standard error, SE) in gray shading. c Superposed epoch analysis (SEA) of CGT anomalies relative to a 5-year pre-eruption baseline, composited across n = 9 volcanic eruptions. The red dashed line denotes the eruption year (Year 0); the blue line shows the multi-event mean; gray dashed lines indicate the 95% confidence interval. d Same as (c), but using the CESM-LME ensemble mean, composited across n = 91 eruption events drawn from 13 ensemble members. e Spatial pattern of volcanically induced CGT anomalies relative to the 5-year pre-eruption baseline, based on CESM simulations. Shading shows composite June–July–August 200-hPa meridional wind anomalies in the first post-eruption summer, and contours denote the climatological summer CGT pattern, defined as the leading empirical orthogonal function (EOF) mode of the ERA5 200-hPa meridional wind (solid and dashed lines indicate positive and negative values, respectively). Green triangles mark tree-ring sites used in the reconstruction. Stippling indicates regions where ≥10 out of 13 ensemble members agree on the sign of the anomaly.

To test the robustness of this signal, we compare it against Community Earth System Model–Last Millennium Ensemble (CESM-LME)36 ensemble simulations over the same historical period. The model-derived CGT index similarly shows a distinct post-eruption decline in the first boreal summer (mean anomaly of −0.59) (Fig. 1b). While the model simulations indicate a slightly more prolonged recovery extending into the second year (Fig. 1d), the primary response matches the proxy record in both sign and magnitude. This strong agreement between proxy and model evidence during the common historical interval supports the existence of a robust volcanic impact on CGT variability.

To evaluate the recurrence of this response over longer timescales, we extend the analysis to all large tropical eruptions since 850 CE (n = 21). CESM simulations over this full millennium period show a consistent post-eruption CGT decline (Supplementary Fig. 2a, b), with SEA indicating a comparable post-eruption decline, reinforcing the statistical robustness of this pattern. Sensitivity tests confirm that the post-eruption CGT decline is robust to the inclusion or exclusion of clustered events (Supplementary Fig. 3).

In both periods, the spatial pattern of upper-tropospheric meridional wind anomalies (Fig. 1e; Supplementary Fig. 2c) reveals a negative CGT-like structure, characterized by anomalous ridging and northerly wind over East Asia. This pattern closely resembles the canonical CGT pattern (Supplementary Fig. 4), indicating that volcanic forcing projects onto the dominant mode of variability trapped within the summer jet stream waveguide28. Consistent with this horizontal structure, the volcanic response exhibits a coherent, equivalent-barotropic circulation across the mid-latitudes (30–60°N; Supplementary Fig. 5). This deep vertical structure indicates that the upper-level wave train extends downward in a vertically aligned manner, effectively steering the low-level flow.

Together, these results underscore a robust and recurrent volcanic influence on the CGT, revealing that explosive tropical eruptions can systematically project onto a leading Eurasian circulation mode. This finding motivates further investigation into the associated hydroclimatic impacts and underlying dynamical mechanisms.

Volcanic-induced droughts over monsoon Asia

Large tropical volcanic eruptions induce significant hydroclimatic perturbations over monsoon Asia. Based on Paleo Hydrodynamics Data Assimilation (PHYDA)37 proxy-constrained reconstructions during 1657–2000 CE, we find widespread negative precipitation and drought anomalies during the first boreal summer following eruptions. CESM simulations reveal coherent post-eruption reductions in precipitation, particularly across South Asia (10°–30°N, 70°–90°E) and northern East Asia (35°–50°N, 100°–130°E), with high inter-ensemble agreement (Fig. 2a). This drying pattern is echoed in the gridded Palmer Drought Severity Index (PDSI) anomalies from PHYDA (Fig. 2b). SEA confirms that statistically significant drought response peak in the first boreal summer following eruptions and weaken in the second year across both South Asia and northern East Asia (Fig. 2c, d).

Fig. 2: Volcanically induced hydroclimatic anomalies over monsoon Asia.
Fig. 2: Volcanically induced hydroclimatic anomalies over monsoon Asia.
Full size image

a Composite precipitation standardized anomalies during the first boreal summer following large tropical eruptions, based on the Community Earth System Model–Last Millennium Ensemble (CESM-LME). Stippling denotes regions where ≥10 out of 13 ensemble members agree on the sign of the anomaly. b Same as (a), but for gridded tree-ring-based Palmer Drought Severity Index (PDSI) standardized anomalies from the Paleo Hydrodynamics Data Assimilation (PHYDA) dataset. Stippling marks locations where anomalies exceed the 95% Monte Carlo confidence interval. Black rectangles indicate the analysis domains for South Asia and northern East Asia. c Regional Superposed epoch analysis (SEA) of standardized precipitation and drought anomalies over South Asia, comparing CESM-LME (red bars with ±1σ error bars) and PHYDA (blue bars). d Same as (c), but for northern East Asia. All anomalies are standardized relative to the 5-year pre-eruption baseline.

Extending the analysis to the last millennium (850–2000 CE), CESM simulations and PHYDA reconstructions consistently show negative precipitation and moisture anomalies during the first boreal summer after eruptions, centered over broadly overlapping regions (Supplementary Fig. 6a, b). SEA further supports a robust hydroclimate response peaking in the first summer and persisting into the second year (Supplementary Fig. 6c, d). Moreover, independent proxy-only reconstructions from GEDA (Great Eurasian Drought Atlas, scPDSI)38 and RAP (Reconstructed Asian summer precipitation)39 reveal comparable post-eruption drying patterns over monsoon Asia during the first boreal summer (Supplementary Fig. 7a, b), despite some regional discrepancies (e.g., wetting over Southeast Asia in GEDA).

These lines of evidence—spanning global climate model ensemble simulations, proxy-model syntheses, and fully independent proxy reconstructions—jointly indicate that explosive tropical eruptions consistently trigger rapid-onset and regionally coherent summer droughts across monsoon Asia. The convergence across multiple independent data sources underscores the robustness of this volcanic hydroclimate signal. Building on this foundation, we next examine how volcanic forcing projects onto the CGT to form the dynamical link to pan-Asian monsoon drought, and how this pathway differs from canonical ENSO and IOD teleconnections.

Volcanic–CGT–drought linkage and its robustness across tropical ocean states

To examine the role of the CGT as a dynamical pathway linking volcanic forcing to pan-Asian monsoon droughts, we assessed the sensitivity of the hydroclimatic response to the CGT phase. Consistent with the dominant post-eruption tendency toward a negative CGT, stratified composites reveal a pronounced phase asymmetry. When eruptions followed by a negative CGT (61 cases), a coherent, high-amplitude wave train is maintained, producing persistent drying over both South Asia and northern East Asia (Fig. 3a–d). By contrast, the much fewer post-eruption years followed by a positive CGT (30 cases) exhibit a disrupted circulation and strongly attenuated rainfall anomalies (Fig. 3e, f). The negative-CGT configuration is characterized by a well-defined South Asian anticyclone, anomalous northerlies over northern East Asia, and pronounced subsidence centered on these two regions (Supplementary Fig. 8a, b), demonstrating that volcanic forcing preferentially projects onto the negative CGT, which serves as the primary atmospheric conduit shaping the pan-Asian monsoon drought pattern.

Fig. 3: Robustness of post-volcanic circulation and hydroclimatic anomalies across different internal climate states.
Fig. 3: Robustness of post-volcanic circulation and hydroclimatic anomalies across different internal climate states.
Full size image

a, b Composite anomalies in the first summer after all selected tropical volcanic eruptions. c, d Same as (a, b), but composited over eruptions followed by a negative circumglobal teleconnection (CGT) phase. e, f Same as (a, b), but composited over eruptions followed by a positive CGT phase. g, h Same as (a, b), but restricted to cases with neutral El Niño–Southern Oscillation (ENSO) and Indian Ocean Dipole (IOD) background states. Left panels (a, c, e, g) display 200-hPa meridional wind standardized anomalies (V200; shading). Overlaid black contours denote the climatological summer CGT pattern, defined as the leading empirical orthogonal function (EOF) mode of the ERA5 200-hPa meridional wind (solid and dashed lines indicate positive and negative values, respectively). Right panels (b, d, f, h) display precipitation anomalies (shading, σ) and 850-hPa wind anomalies (vectors). All anomalies are derived from Community Earth System Model–Last Millennium Ensemble (CESM-LME) simulations (1657–2000) relative to a 5-year pre-eruption baseline. Stippling (for shading) and black vectors (for winds) indicate regions where at least 10 out of 13 ensemble members agree on the sign of the anomaly. Numbers in parentheses indicate the total number of events included in each composite.

Moisture-budget diagnostics further confirm this CGT-mediated dynamical linkage. Vertical dynamic advection accounts for the majority of the precipitation deficits—approximately −0.75 mm day⁻¹ over South Asia and −0.15 to −0.19 mm day⁻¹ over northern East Asia—and its spatial footprint closely mirrors the observed drought pattern (Supplementary Figs. 9–10). Stratification by CGT phase reveals a pronounced asymmetry consistent with the circulation-based analysis: during post-eruption years with a positive CGT phase, Dyn(V) weakens markedly and its spatial structure no longer aligns with the drought anomalies (South Asia: −0.38 mm day⁻¹; northern East Asia: near zero), whereas during negative-CGT years it intensifies and remains spatially coherent (−0.98 mm day⁻¹ and −0.18 mm day⁻¹, respectively). Together, these results indicate that the dominant role of vertical dynamic advection in post-eruption drying is conditional on the negative CGT phase, providing a direct physical mechanism through which the CGT-modulated circulation translates into synchronized pan-Asian monsoon droughts.

Having established this physical linkage, we next assess its sensitivity to tropical ocean variability. Composite analysis restricted to strictly ENSO-neutral and IOD-neutral years (45 cases) reveals a well-defined negative CGT pattern (Fig. 3g) and a drought structure (Fig. 3h) that closely resembles—and is even more distinct than—the full-sample response (Fig. 3a, b, 91 cases). Moisture-budget diagnostics confirm that the vertical dynamic advection term retains its dominant magnitude and spatial coherence within these neutral composites (Supplementary Figs. 9, 10), indicating that the drying mechanism is intrinsic to the volcanic response rather than an artifact of internal variability. Furthermore, comparison with non-volcanic ENSO and IOD events shows that internal tropical ocean variability cannot account for the volcanic response, as their teleconnections exhibit wave-train patterns fundamentally different from the post-eruption negative CGT pattern (Supplementary Fig. 11). When eruptions are followed by ENSO or IOD events, the resulting anomalies reflect a linear superposition of the negative CGT and the internal mode (Supplementary Fig. 12), indicating that the volcanic signal is additive rather than derivative of tropical ocean variability.

Statistical diagnostics reinforce these physical findings. SEA applied to ENSO-removed and IOD-removed residuals reproduces the same post-eruption negative CGT anomalies and the same coherent drought patterns (Supplementary Fig. 13). Complementary multiple linear regression (MLR) shows that volcanic forcing is a statistically significant predictor of the CGT index, whereas interaction terms (e.g., ENSO × volcano) remain negligible (Supplementary Fig. 14). Taken together, these physical, statistical, and dynamical lines of evidence indicate that the volcanic–CGT–drought pathway is robust across ENSO/IOD conditions and that the CGT provides the primary atmospheric conduit through which volcanic forcing synchronizes summer drying across monsoon Asia.

Mechanism linking volcanic to negative CGT

To establish the tropical forcing that initiates the CGT response, we first examined how volcanic eruptions modify monsoon convection. CESM-LME Outgoing Longwave Radiation (OLR) shows a clear and statistically robust positive anomaly over the core Indian monsoon region during the first post-eruption summer (Supplementary Fig. 15), indicating weakened deep convection and reduced diabatic heating. This suppression of monsoon convection identifies the tropical heating anomaly required to generate the Rossby wave source of the ensuing CGT pattern.

Having established the weakening of South Asian convective heating, we next assess how this tropical perturbation excites the extratropical circulation. Building on this convective forcing, we probed the dynamical pathways by analyzing upper- and mid-tropospheric wind fields from CESM ensemble simulations and idealized experiments conducted with the LBM40,41 (Fig. 4). Composites of CESM simulations following large tropical eruptions reveal a canonical negative CGT-like meridional wind anomaly pattern at 200 hPa, with a wave train extending eastward from Central Asia (~60°E) toward the North Pacific and a pronounced cyclonic anomaly over eastern Eurasia (Fig. 4a). This pattern is broadly reproduced by LBM simulations forced with idealized diabatic heating centered over South Asia (Fig. 4b; see Supplementary Fig. 16 for the idealized heating pattern), supporting the role of tropical heating in exciting extratropical teleconnections.

Fig. 4: Upper-, mid-, and lower-tropospheric circulation anomalies during the first boreal summer following large tropical volcanic eruptions.
Fig. 4: Upper-, mid-, and lower-tropospheric circulation anomalies during the first boreal summer following large tropical volcanic eruptions.
Full size image

Composite 200-hPa meridional wind anomalies from Community Earth System Model–Last Millennium Ensemble (CESM-LME) (a) and Linear Baroclinic Model (LBM) (b) simulations in response to volcanic eruptions. c, d Same as (a, b), but for 500 hPa. e, f Same as (a, b), but for 850 hPa. Shading in all panels represents standardized anomalies; arrows denote horizontal wind vectors. For CESM-LME panels, stippling (for shading) and black vectors (for winds) indicate regions where at least 10 out of 13 ensemble members agree on the sign of the anomaly. For LBM panels, black arrows denote stronger winds (>0.5), while gray arrows denote weaker winds (≤0.5).

Notably, both CESM and LBM exhibit a circulation adjustment to the imposed heating that resembles a Gill-type response, characterized by a mid-to-lower tropospheric anticyclonic anomaly (Fig. 4c–f) and an upper-tropospheric cyclonic circulation west of the heating center (Fig. 4a, b). At mid-tropospheric levels (Fig. 4c, d) and lower levels (Fig. 4e, f), coherent southward meridional wind anomalies emerge over northern East Asia in both CESM and LBM (Fig. 4c–f), reinforcing the vertical coherence of the response. Supplementary Fig. 17 further illustrates the time evolution of the LBM response, showing that the upper- and mid-tropospheric circulation adjusts rapidly to the imposed heating and stabilizes into a coherent negative CGT-like Rossby wave train within 25 days. These results underscore a dynamically consistent mechanism linking post-volcanic tropical heating anomalies to mid-latitude summer circulation via both local and remote atmospheric adjustments.

Discussion

This study demonstrates that large tropical volcanic eruptions consistently induce concurrent summer droughts over South Asia and northern East Asia. In the boreal summer following large eruptions, suppressed convection over the South Asian subcontinent initiates a Rossby wave response that projects onto the negative phase of the CGT, promoting northerly winds and precipitation deficits over northern East Asia. This teleconnection is evident in tree-ring reconstructions, last-millennium climate ensemble simulations, and idealized modeling experiments. While earlier studies emphasized monsoon weakening following eruptions42,43,44,45, our results clarify how volcanic forcing propagates into midlatitudes through an upper-tropospheric wave train. Although CGT variability is commonly attributed to internal dynamics31,32,33, our findings show that tropical eruptions can externally trigger CGT anomalies via suppressed South Asian diabatic heating.

Crucially, our analysis indicates that the volcanic–CGT–drought linkage reflects a primary atmospheric response to volcanic radiative forcing, rather than being contingent on internal tropical ocean modes. Although ENSO and the IOD influence the Asian monsoon through diverse teleconnection pathways—including the conditional indirect excitation of the CGT28,46—their circulation responses are often dominated by the Pacific–North America pattern and can exhibit a complex mixture of structures47,48. This contrasts with the distinct and spatially coherent post-eruption negative CGT wave train. Notably, the negative CGT anomaly emerges robustly even when composites are restricted to ENSO-neutral and IOD-neutral years following large tropical eruptions, and it persists after statistically removing tropical SST-related signals, indicating limited sensitivity to tropical ocean conditions within our diagnostics. When eruptions coincide with ENSO or IOD events, the resulting anomalies are consistent with an approximate linear superposition of the negative CGT response and the internal mode, suggesting that the volcanic signal is largely additive. Similarly, while the IPO modulates the monsoon background state on decadal timescales49, it is insufficient to account for the abrupt, pulse-like establishment of the coherent CGT wave train observed immediately following eruptions. Together, these results support the CGT as the primary dynamical conduit through which volcanic forcing synchronizes summer drought across monsoon Asia.

The identification of this CGT pathway provides a complementary dynamical perspective on the spatially heterogeneous East Asian monsoon response. Earlier studies have highlighted enhanced rainfall over southern China after eruptions—often linked to El Niño–related circulation anomalies23—yet our results suggest that the negative CGT phase offers an additional pathway contributing to the meridional precipitation dipole. This dipole remains evident even in post-eruption years without El Niño influence and weakens when a positive CGT phase prevails. In this circulation configuration, northerly wind anomalies reduce rainfall over northern East Asia while enhancing southward moisture transport, helping to explain the contrasting hydroclimate outcomes across the region.

Although all datasets consistently show a robust negative CGT response in the first post-eruption year, there are modest differences in the apparent duration of this signal. In the CESM simulations, the weak extension into the second post-eruption year is plausibly explained by the thermal inertia of the coupled ocean–atmosphere system, which can sustain circulation anomalies beyond the peak radiative forcing36,50. By contrast, the tree-ring reconstruction exhibits muted negative anomalies in the second and third years following the eruption that are not statistically significant and are most likely indicative of biological memory or physiological preconditioning rather than a sustained atmospheric response. Intense hydroclimatic stress in the first post-eruption year can reduce non-structural carbohydrate reserves and temporarily suppress radial growth even after large-scale climate conditions begin to recover51,52. In addition, because the reconstruction is calibrated over 1958–1995 due to proxy availability, caution is advised when using this reconstruction to assess modern forced changes. Nonetheless, target‑permutation tests indicate a statistically detectable response to volcanic impulses (Supplementary Fig. 18), lending confidence to its use for diagnosing the event‑scale impacts of historical eruptions.

In summary, our findings establish the CGT as a critical dynamical bridge linking tropical volcanism to pan-Asian monsoon hydroclimate extremes. By elucidating this atmospheric teleconnection pathway, our study refines understanding of externally forced drought risks and underscores the need to improve climate model representation of upper-tropospheric circulation. This is particularly important as future volcanic eruptions will interact with an evolving background state—including shifting jet streams and warming oceans53,54,55,56—likely yielding complex, nonlinear climate responses.

Methods

Meteorological reanalysis, climate ensemble simulation, and paleoclimate proxy datasets

We employed a combination of meteorological reanalysis, global climate model ensemble simulations, high-resolution paleoclimate proxies (i.e., tree-ring), hydroclimate reconstructions, and volcanic eruption records to analyze hydroclimatic responses to tropical volcanic eruptions. All gridded datasets were standardized to boreal summer (June–August) means and regridded to a common 0.5° horizontal spatial resolution using bilinear interpolation for comparison and composite analysis.

Meteorological reanalysis

Upper-tropospheric wind fields, including 200-hPa meridional winds were obtained from the ERA5 reanalysis (0.25° resolution)57, with monthly coverage beginning in 1940, and self-calibrated Palmer Drought Severity Index (scPDSI) from the Climatic Research Unit (CRU TS v4.06, 0.5° resolution)58. For this study, we analyzed the period 1958–2020 from both datasets, a period generally considered more reliable owing to the expansion of observational networks and advances in data assimilation techniques.

Global climate model ensemble simulations

We used the CESM-LME, which includes 13 fully coupled ensemble members covering the period 850–2005 CE. Precipitation, zonal and meridional winds (200 hPa, 500 hPa and 850 hPa), Outgoing Longwave Radiation (OLR; variable FLUT in CESM-LME), and surface temperature were extracted for analysis. The original atmospheric resolution of CESM-LME is approximately 2.5° × 1.9°.

Hydroclimate reconstructions

We used multiple independent gridded datasets of paleoclimate variability over Asia. The Monsoon Asia Drought Atlas (MADA v2)10 provides tree-ring-based scPDSI reconstructions spanning 0–2012 CE. The PHYDA37 product integrates proxy and model data to reconstruct temperature, precipitation, and PDSI over 1–2000 CE. In addition, we used the Great Eurasian Drought Atlas (GEDA)38 and the Reconstructed Asian summer Precipitation (RAP)39 dataset to provide complementary perspectives on drought and rainfall variability.

Volcanic eruption events

To identify large tropical eruptions over the past millennium, we used two independent volcanic forcing datasets depending on the analysis framework. For the proxy-based analyses, we used the eVolv2k (v3) reconstruction developed by Toohey and Sigl59, which provides annually resolved estimates of stratospheric sulfur injection (SSI) inferred from ice-core sulfate records. Events were selected based on the following criteria: (1) latitude between 23.5° S/N, ensuring that eruptions originate from the tropical belt where stratospheric aerosol perturbations exert the strongest global climate impacts. (2) Volcanic stratospheric sulfur injection (VSSI) ≥ 4.5 Tg S, which captures moderate-to-large eruptions capable of producing detectable hydroclimatic responses.

For the model-based analyses using CESM-LME, we employed the official volcanic forcing dataset Ice-core Volcanic Index 2 (IVI2)60, which provides prescribed aerosol forcing used to drive the CESM-LME ensemble. Eruptions were selected using criteria analogous to the proxy-based analysis but tailored to the IVI2 dataset structure: (1) global sulfur injection >10 Tg, ensuring that only climatically significant eruptions were included. (2) Interhemispheric injection ratio (NH/SH) between 0.2 and 5.0, selecting eruptions with relatively balanced hemispheric aerosol loading. In both datasets, the eruption year (Year 0) is defined as the year of peak stratospheric sulfate injection3,6. The complete list of selected eruptions and their forcing magnitudes is provided in Supplementary Table 1 and Supplementary Table 2.

Definition of CGT

The CGT pattern is defined as the first empirical orthogonal function (EOF) mode of the summer mean 200-hPa meridional wind over the domain (20–60°N, 0–150°E)27,29, and the CGT index is defined as the normalized first principal component accordingly. The summer mean is defined as the average of June–August (JJA) and is used throughout this study except where otherwise stated explicitly. To maintain consistency and allow for comparisons among datasets with different temporal coverages, the CGT pattern was extracted by applying EOF analysis to the overlapping periods of different datasets (i.e., 1979–2010). The corresponding CGT index for the full period of each dataset was then obtained by calculating the spatial inner product between the fixed CGT pattern and the yearly 200-hPa meridional wind anomalies. We also tried to obtain the CGT pattern and index by applying EOF analysis to slightly different domains or different time periods, and the results are insensitive to these slight changes.

Reconstruction of CGT index

To reconstruct the long-term variability of the CGT index, we compiled tree-ring measurements from the International Tree-Ring Data Bank (ITRDB), initially comprising 2490 measurements across the Eurasian continent. To remove non-climatic age-related growth trends while preserving low-frequency climate signals, we applied a signal-free standardization implemented in the dplR package61,62 in R. Individual series were iteratively detrended using an age-dependent smoothing spline (initial stiffness = 50 years), and site chronologies were developed using biweight robust means. Only chronologies covering at least the period 1850–1995 CE and maintaining an Expressed Population Signal (EPS) ≥ 0.85 were retained for subsequent analysis, resulting in a quality-controlled pool of 924 chronologies.

Following standardization, we rigorously screened the chronologies based on their statistical relationships with the ERA5-based CGT index to ensure that only proxies containing physically meaningful climate signals were utilized. Three criteria were applied: (1) A statistically significant correlation with the CGT index (p < 0.05); (2) A significant positive correlation with local climate variables (scPDSI or temperature anomalies, averaged over the nine nearest CRU grid points) to exclude series dominated by non-climatic stressors; and (3) A consistency check requiring a positive product of the CGT–climate and climate–proxies correlation coefficients, ensuring the stability of the physical linkage direction.

We opted to reconstruct the full-frequency spectrum directly, rather than decomposing the signal (e.g., via EEMD). This ensures that the reconstruction relies on a physically consistent climate-growth relationship across time scales, without the risk of statistical overfitting introduced by separate frequency calibration. To maximize model robustness, we performed a sensitivity analysis on the correlation threshold used to screen proxy–CGT relationships during predictor selection. We found that a correlation threshold of |r| > 0.39 yielded the optimal reconstruction performance (Supplementary Fig. 19). This resulted in a final network of nine chronologies (Supplementary Table 3). The retained chronologies are geographically widespread and align with major centers of action of the CGT, supporting a large-scale teleconnection signal (Fig. 1e). A pseudo-proxy experiment indicates that the final network size is significantly larger than expected from red-noise pseudo-proxies (p < 0.001; Supplementary Fig. 20). Multicollinearity among these final predictors was assessed and found to be minimal (Supplementary Fig. 21).

We employed a nested Principal Component Regression (PCR) approach to maximize the utilization of temporally varying availability of chronologies while minimizing the risk of overfitting. At each nesting level, the subset of chronologies were z-score standardized over their common period and subjected to Principal Component Analysis. Principal Components (PCs) with eigenvalues >1.0 and an absolute correlation with the CGT index >0.20 were retained as predictors in MLR models. To ensure statistical stability, the number of predictors in each MLR model was capped at four, adhering to the “one-in-ten” rule relative to the length of the calibration period. Nests were constructed by systematically removing the shortest chronologies; the resulting time series were then merged into a single continuous reconstruction by scaling the mean and variance of each nest to match the most replicated (most recent) nest.

Model performance was evaluated using standard split-period validation statistics (Supplementary Fig. 1), including the coefficient of determination (R²), reduction of error (RE), and coefficient of efficiency (CE). In addition, Leave-One-Out cross-validation was applied to assess temporal stability (Supplementary Table 4). To evaluate the robustness of the predictor network and its sensitivity to individual proxies, we conducted a Leave-One-Proxy-Out validation, which demonstrates that the reconstruction is not dominated by any single proxy series (Supplementary Fig. 22). We further assessed potential target-conditioning using a target permutation test with 1000 phase-randomized CGT targets, and the observed validation skill lies well outside the permutation-based distribution (p < 0.001, Supplementary Fig. 18). The final reconstruction uncertainty was estimated as ±1σ derived from the residual bootstrap errors from the nested PCR fits.

Composite and significance

We apply superposed epoch analysis (SEA) to evaluate the response of the CGT and associated regional hydroclimatic anomalies to large tropical volcanic eruptions51,63. Volcanic year 0 is defined as the year of peak stratospheric aerosol loading. All variables are expressed as anomalies relative to the 5-year pre-eruption mean and composited across events. To prevent any single eruption from dominating the composite, each event window is variance-normalized within its window11,64. To mitigate compound forcing and biological memory effects6,65, clustered eruptions (<10 years) were filtered to retain only the second event and compute anomalies relative to the baseline preceding the first eruption6. Statistical significance was assessed using Monte Carlo bootstrapping.

For the OLR diagnostics, global-mean OLR anomalies were removed prior to compositing to isolate regional convective signals from the uniform volcanic radiative cooling.

Moisture budget analysis

To diagnose the physical drivers of volcanic-induced precipitation changes, we apply the standard vertically integrated moisture-budget framework66,67. The anomalous column-integrated moisture equation is

$${P}^{{\prime} }=-\left\langle \bar{\omega }{\partial }_{p}{q}^{{\prime} }\right\rangle -\left\langle {\omega }^{{\prime} }{\partial }_{p}\bar{q}\right\rangle -\left\langle \bar{{{\bf{u}}}}\cdot \nabla {q}^{{\prime} }\right\rangle -\left\langle {{{\bf{u}}}}^{{\prime} }\cdot \nabla \bar{q}\right\rangle+{E}^{{\prime} }+{{\rm{Res}}}$$
(1)

Where \(q\) is specific humidity, \({{\bf{u}}}\) and \(\omega\) denote horizontal winds and vertical pressure velocity, respectively, and \({E}^{{\prime} }\) and \({P}^{{\prime} }\) represent evaporation and precipitation anomalies. Angle brackets \(\left\langle {{\rm{\cdot }}}\right\rangle\) indicate mass integration through the troposphere. The four advection terms correspond to: (i) the vertical dynamic \(-\langle {\omega }^{{\prime} }{\partial }_{p}\bar{q}\rangle\) and thermodynamic components \(-\langle \bar{\omega }{\partial }_{p}{q}^{{\prime} }\rangle\), and (ii) horizontal dynamic \(-\left\langle {{{\bf{u}}}}^{{\prime} }\cdot \nabla \bar{q}\right\rangle\) and thermodynamic \(-\langle \bar{{{\bf{u}}}}\cdot \nabla {q}^{{\prime} }\rangle\) components. The residual term (\({\mbox{Res}}\)) captures transient eddies68,69 and unresolved covariance processes.

Isolating the volcanic–CGT–drought pathway from ENSO and IOD

To separate the atmospheric response to volcanic forcing from internal tropical ocean variability, we applied a multi-step analytical framework combining index definition, stratified compositing, linear signal removal, and MLR. This design allows a controlled assessment of volcanic impacts under different background climate states.

ENSO and IOD indices

The influence of internal variability was characterized using standard indices extracted from CESM-LME simulations and independent proxy reconstructions. Based on the season when its impact on the subsequent Asian summer monsoon is strongest70,71, the ENSO index was defined as the boreal winter (DJF) sea surface temperature (SST) anomalies averaged over the Niño-3.4 region (5°N–5°S, 120°–170°W). The IOD index was calculated as the difference in boreal summer (JJA) SST anomalies between the western (10°S–10°N, 50°–70°E) and eastern (10°S–0°, 90°–110°E) equatorial Indian Ocean. To ensure that inferred signals were not artifacts of model-specific internal variability, we additionally employed independent ENSO72 and IOD70 reconstructions for validation.

Stratification and signal removal

We first isolated the volcanic signal physically by constructing stratified composites and Superposed Epoch Analyses (SEA) restricted to “neutral” background states. Neutral years were rigorously defined as those where the absolute index values were less than 1 standard deviation (\( < 1{{\upsigma }}\))—specifically, \(|{{\rm{Ni}}} {\tilde{{\rm{n}}} {{\rm{o}}}}_{3.4}| < 1\sigma\) in the preceding DJF and \(\left|{\mbox{IOD}}\right| < 1\sigma\) in the concurrent JJA. By analyzing “Dual-neutral” eruptions (where both conditions are met), we effectively removed the influence of ENSO-driven Pacific–East Asian teleconnections and IOD-driven Indian Ocean–Asian monsoon couplings.

To complement this physical stratification, we employed a residual-based statistical removal method. The ENSO- and IOD-independent signals (\({X}^{{\prime} }\)) were calculated as:

$${X}^{{\prime} }=X-{\beta }_{{{\rm{ENSO}}}}\cdot {{\rm{ENSO}}}-{\beta }_{{{\rm{IOD}}}}\cdot {{\rm{IOD}}}$$
(2)

where \(X\) represents the target variable (CGT index or gridded atmospheric fields) and \(\beta\) denotes the regression coefficients obtained via MLR. We subsequently applied SEA to these residuals (\({X}^{{\prime} }\)) to evaluate whether the volcanic circulation response persists after the statistical elimination of linear signals associated with ENSO and IOD.

Regression modeling and superposition assessment

We further quantified the independent contribution of volcanic forcing using a MLR model:

$$\begin{array}{c}Y={\beta }_{0}+{\beta }_{1}{\mbox{Volcano}}+{\beta }_{2}{\mbox{ENSO}}+{\beta }_{3}{\mbox{IOD}}+{\beta }_{4}\left({\mbox{Volcano}}\times {\mbox{ENSO}}\right)\\+{\beta }_{5}\left({\mbox{Volcano}}\times {\mbox{IOD}}\right)+\varepsilon \,\end{array}$$
(3)

where \(Y\) represents the CGT index or precipitation targets. The statistical significance of these interaction terms was evaluated to determine whether the combined forcing operates additively (linear superposition) or via modulation of the internal modes. Furthermore, to ensure the robustness of the attribution, we repeated these diagnostic procedures substituting the CESM-based indices with independent proxy-based ENSO and IOD reconstructions.

Linear baroclinic model

To investigate the dynamical mechanisms suggested by observational diagnostics, we employed the LBM40,41. The LBM solves the linearized primitive equations on the sphere, including the momentum, thermodynamic, continuity, and hydrostatic balance equations, linearized around a prescribed climatological basic state. Diabatic heating is introduced as an external forcing in the thermodynamic equation, while Rayleigh friction and Newtonian cooling are applied to damp transients and guide the model toward a quasi-steady response.

The model was configured with a T21 spectral resolution (~5.6° or ~600 km) and 20 vertical sigma levels, using the June–August climatology for 1981–2010 as the reference basic state. Each experiment was integrated for 25 days to capture the quasi-steady circulation adjustment to imposed diabatic heating. This experimental design allows for an isolated assessment of the linear atmospheric response to prescribed tropical heating, while filtering out internal variability and nonlinear feedback. The choice of a dry model with simplified damping highlights the essential dynamics of heating-induced teleconnections, enabling a clearer diagnosis of the mechanisms driving the Eurasian summer circulation anomalies.

Statistical information

All statistical tests and uncertainty estimates follow standard practices in paleoclimatology and climate dynamics. Reconstruction of the tree-ring–based CGT index was evaluated using split-period calibration–verification statistics and leave-one-out cross-validation, including the coefficient of determination (R²), reduction of error (RE), and coefficient of efficiency (CE). Reconstruction uncertainties were quantified as ±1σ based on residual bootstrap resampling.

Hydroclimatic responses to volcanic eruptions were evaluated using SEA with statistical significance determined using a Monte Carlo bootstrap procedure with 1000 random event sets. For composite analyses of CESM-LME simulations, stippling indicates regions where at least 10 out of 13 ensemble members agree on the anomaly sign. For gridded reconstruction, statistical significance of anomalies was evaluated using two-tailed Student’s t-tests; exact t values and degrees of freedom are reported in the figure captions where relevant. Sample sizes (n) generally reflect data availability and are explicitly reported: n = 26 and n = 21 independent volcanic events (850–2000 CE) for the reconstruction (eVolv2k) and simulation (IVI2) analyses, respectively; and n = 13 independent ensemble members for the CESM-LME. Error bars in figures represent ±1σ inter-ensemble spread unless otherwise noted.