The Pacific Walker Circulation (PWC), a crucial component of global weather and climate, exhibits significant interannual to decadal variability. Understanding its response to external forcings like anthropogenic and volcanic aerosols remains a challenge, with discrepancies between empirical data and climate model simulations. Most climate models predict a PWC weakening in response to global warming, yet observations reveal a strengthening from 1992 to 2011. This discrepancy highlights the need to disentangle the roles of anthropogenic forcing, volcanic activity, and internal variability. This study addresses this gap by utilizing a novel, high-resolution reconstruction of the PWC over the past millennium, allowing for a more robust assessment of forced changes against the backdrop of intrinsic variability. The PWC's inherent variability makes it difficult to discern forced signals from observational data alone, particularly since observational records are too short to fully characterize both. Previous reconstructions, often limited by decadal resolution and the use of proxies sensitive to hydroclimate rather than atmospheric circulation, have hampered efforts to analyze interannual variability and the PWC's response to major forcings like volcanic eruptions. This study aims to overcome these limitations by offering a comprehensive, annually resolved PWC reconstruction spanning 1200-2000 CE.

Existing research on the PWC's response to external forcings reveals significant disagreement. Some studies suggest a strengthening of the PWC due to anthropogenic forcing, while others indicate a weakening or no detectable influence. Most observational datasets show a considerable strengthening of the PWC between 1992 and 2011, but whether this is due to external forcing or internal variability remains uncertain. This uncertainty is exacerbated by the consistent absence of this strengthening in climate model simulations. Annually resolved ENSO reconstructions have provided insights into the ENSO response to volcanic eruptions, but similar assessments for the PWC have been hindered by the difficulty of reconstructing atmospheric variability directly from proxies. Existing inferences of preindustrial PWC variability rely on decadally resolved records and proxies sensitive to various hydroclimate aspects, insufficient for assessing interannual changes. The lack of a long-term, high-resolution PWC reconstruction has hampered efforts to understand the impact of external forcings on this critical component of the climate system.

This study constructs an annually resolved PWC reconstruction from 1200 to 2000 CE using an ensemble approach that incorporates uncertainties arising from data, methods, and chronology. The reconstruction targets anomalies in the trans-Pacific atmospheric sea-level pressure (ASLP) gradient, a commonly used PWC indicator. Higher ASLP values represent a stronger PWC (more 'La Niña-like' conditions), while lower values indicate a weaker PWC (more 'El Niño-like' conditions). The reconstruction leverages 59 palaeoclimate proxy records, including 54 globally distributed records of stable isotopic composition of precipitation and other meteoric waters, and 5 annually resolved non-isotope based records related to PWC or ENSO. The Iso2k database, which integrates water-isotope signals into palaeoclimate reconstructions, was utilized. To maximize the use of available data, the reconstruction is performed in five overlapping temporal subsets (1200–2000, 1400–2000, 1600–2000, 1800–2000, and 1860–2000), using records with >66% coverage in each subset. Five statistical methods – composite plus scale (CPS), principal component regression (PCR), pairwise comparison (PaiCo), and two variants of principal component analysis (PCA) – are employed to account for methodological uncertainties. The PCR reconstructions are performed using all proxy records and a subset significantly correlated with ASLP. The CPS reconstructions utilize the entire dataset and subsets focused on the tropical Pacific and records without seasonal bias. The reconstructions are trained on ASLP calculated from three gridded SLP products: HadSLP, ICOADS, and ERA-20C, using a 1900–2000 calibration interval. Chronological uncertainty is explicitly incorporated by sampling multiple realizations from banded age-depth model ensembles for each record. Up to 15% of records are randomly removed in each iteration to account for potential dependence on specific subsets. Reconstruction skill is assessed using a separate 1951–2000 calibration interval and an independent 1900–1950 validation interval. Superposed Epoch Analysis (SEA) is applied to analyze the PWC response to volcanic eruptions using the 4,800 ASLP reconstruction ensemble members. SEA composites volcanic eruption events to show a composite ASLP response to explosive volcanism. The proportion of ensemble members exhibiting significant positive or negative ASLP anomalies in the years following eruptions is determined. The study also analyzes climate model data (CESM1 LME and PMIP3/4 models) to compare simulated PWC and Niño 3.4 SST responses to volcanic eruptions with the reconstruction results. Correlations between the ASLP reconstruction and global mean surface temperature (GMST) reconstructions are also examined across various time periods.

The annually resolved PWC reconstruction reveals substantial interannual to decadal variability throughout the past millennium. A weak positive ASLP trend is observed from 1200 to 1750, followed by a slight decrease to around 1800, and a period of low inter-method agreement (potentially due to non-stationary climate covariation and volcanic eruptions). The 20th century shows fluctuations around a stable mean, culminating in a positive trend over the past two decades. Spectral analysis reveals the highest power in the interannual (2–9-year) band, consistent with ENSO influence. A shift to higher power at lower frequencies (4–9-year periods) is observed in the industrial era (1850–2000) compared to the preindustrial period (1200–1849), suggesting a subtle shift in variability. The distribution of ASLP values in the industrial era is slightly skewed towards higher (more La Niña-like) values than in the preindustrial period, but the difference in mean ASLP is not statistically significant in 81% of the reconstruction ensemble members. This indicates a lack of significant PWC mean state change in response to anthropogenic forcing, contrasting with climate model predictions of PWC weakening. SEA reveals a significant El Niño-like PWC weakening (negative ASLP anomaly) in the 0–2 years following large volcanic eruptions, with a rapid recovery to pre-eruption conditions. This response is consistent across reconstruction methods and observational products, although it becomes less clear when including older eruptions (likely due to increased chronological uncertainty). The magnitude of the post-eruption ASLP response does not appear to scale with eruption magnitude. Climate model simulations (CESM1 LME and PMIP3/4 models) reproduce the El Niño-like PWC weakening after volcanic eruptions, with the ASLP response showing greater consistency than the associated Niño 3.4 SST response across model members. Comparisons with GMST reconstructions show no reliable anticorrelation between PWC strength and GMST, indicating that if a thermodynamic effect of GMST on PWC exists, it's masked by other forcings or dynamic responses. The unusually strong 1992–2011 PWC strengthening is unlikely to be solely attributed to recovery from the Mount Pinatubo eruption; it may reflect anthropogenic aerosol forcing (concentrated in the Northern Hemisphere) or natural decadal variability.

The findings highlight the substantial intrinsic variability of the PWC, emphasizing the need for long-term context when assessing trends. Both volcanic eruptions and anthropogenic forcing produce detectable PWC changes, although the nature of the responses differs. The significant El Niño-like PWC weakening following volcanic eruptions, consistent across the reconstruction and climate models, indicates a robust atmospheric response to this forcing. The lack of a significant PWC trend since the onset of anthropogenic forcing is noteworthy, suggesting that either competing forcings obscure a thermodynamic weakening effect, other mechanisms are at play, or the change is too subtle to emerge from intrinsic variability. The 1992–2011 PWC strengthening, though anomalous, demonstrates that decadal variability can produce substantial short-term fluctuations. This study's findings, based on a comprehensive high-resolution PWC reconstruction, offer valuable empirical insights that can refine climate model simulations and improve understanding of PWC dynamics.

This study presents a novel, high-resolution reconstruction of the Pacific Walker Circulation (PWC) over the past millennium, revealing substantial intrinsic variability. The analysis demonstrates significant El Niño-like PWC weakening following volcanic eruptions, a pattern supported by climate model simulations. While no significant industrial-era PWC trend is observed, a shift toward lower-frequency variability and the anomalous recent strengthening suggest a subtle, yet detectable, anthropogenic influence. Future research should focus on further data-model comparisons to understand model biases and the complex interplay of forcings and internal variability in driving PWC changes.

The reconstruction's skill decreases prior to approximately 1600 CE due to reduced data coverage and increased chronological uncertainty. This limits the confidence in the findings for the earliest part of the reconstruction. The analysis focuses on the PWC's response to broad-scale forcings; more detailed investigation into regional-scale interactions would enhance understanding. The study acknowledges the influence of pre-existing El Niño conditions on the PWC response to some twentieth-century volcanic eruptions, which might affect the interpretation of the composite responses. Further research is needed to completely resolve the complex relationship between different types of forcing and the PWC's response.