A reconstructed PDO history from an ice core isotope record on the central Tibetan Plateau

The study addresses how to robustly interpret low-latitude ice core δ18O records from the Tibetan Plateau (TP) given longstanding challenges in precise annual-scale dating due to weak seasonality. δ18O in TP precipitation and ice cores is influenced primarily by convective intensity and large-scale circulation linked to the Asian Monsoon (AM) and ENSO, rather than local temperature. Prior work has shown strong inverse correlations between central TP δ18O and the Southern Oscillation Index (SOI), implying interannual variability tied to ENSO is preserved in ice. However, most inland TP ice cores lack clear seasonal cycles, complicating chronology and limiting high-resolution climate reconstructions. Building on the observed δ18O–ENSO coherence, the authors hypothesize this relationship is persistent over longer periods and can be exploited to construct a high-resolution time series. They aim to: (1) develop a new dating approach for the QT No. 1 ice core using spectral characterization and ENSO-mode matching via VMD, (2) reconstruct a multi-century δ18O time series, and (3) investigate Pacific decadal variability (PDO/IPO) signals and recent δ18O increases in relation to ENSO behavior.

Multiple lines of evidence link precipitation δ18O across the AM region to ENSO-driven convective variability, including cave speleothem isotopes, Himalayan and TP ice cores, and modeling/simulation studies showing δ18O’s sensitivity to monsoon circulation changes. Prior TP ice core chronology methods have used bomb layers (1963) and various radiometric techniques (14C, 39Ar, 81Kr, 85Kr, δ18O of trapped O2), but resolution is often insufficient for annual dating, leading to debates over Tibetan ice core chronologies. ENSO signals have been retrieved from diverse high-resolution archives (tree rings, corals, stalagmites, giant clams), and decadal-to-multidecadal Pacific variability (IPO/PDO) has been identified in Pacific ice core composites. This study builds on that foundation by applying spectral methods and VMD to isolate ENSO, biennial (TBO-related), and annual modes in the QT core, using the ENSO mode to anchor dating and subsequently examine PDO periodicities.

Study site and sampling: The QT No. 1 glacier (33.3°N, 88.69°E; 5890 m asl) in central TP is influenced by winter westerlies and the summer Indian monsoon. Two bedrock-reaching ice cores (109 m and 110 m) were drilled in May 2014. Core 1’s top 50 m δ18O (samples cut at 3 cm) were analyzed (accuracy ±0.15‰). Outer 1 cm was assayed for total β activity and 137Cs to identify the 1963 bomb layer (8.8–9.2 m). Climate indices: Annual Niño 3.4 SST anomalies (HadISST1; 1870–2011 CE) represent ENSO; for 1650–1869 CE, an integrated reconstructed ENSO index was used. Spectral analysis: The upper 50 m δ18O depth series were divided into five 10 m intervals (D1–D5). The multitaper method (MTM; NW=2, K=3) with robust red-noise testing was applied to each interval to identify significant frequencies. Three consistent frequency bands were interpreted as: ENSO mode (~1.1–1.6 cycles/m), biennial mode (~2.3–3.4 cycles/m), and annual mode (~5.1–6.6 cycles/m), based on peak ratios and known accumulation/sample spacing. Variational mode decomposition (VMD): VMD was used to decompose each interval’s δ18O depth series into intrinsic mode functions (IMFs). With k=6, the ENSO (IMF2), biennial (IMF3), and annual (IMF4) modes were cleanly separated across D1–D5, with center frequencies matching MTM results and no frequency aliasing. Dating approach: The dating hinges on aligning the ENSO IMF (IMF2) from the depth series to peaks/minima in the ENSO index time series (nonlinear alignment to account for varying accumulation rates). Biennial (IMF3; half-cycle ≈ one year) and annual (IMF4; full cycle ≈ one year) modes serve as auxiliary constraints to determine the number of samples per year within each annual layer. The method was validated in the upper ~9.2 m via the radioactive bomb layer (1963) and yields a consistent D1 chronology (2011–1960 CE). The approach was then extended to D2–D5, producing a continuous 335-year time series (1677–2011 CE) for the top 50 m. PDO reconstruction and analyses: From the dated δ18O time series, band-pass components corresponding to PDO bidecadal (25–35 years) and multidecadal (50–70 years) periodicities were linearly superimposed to reconstruct a PDO index for 1677–2011 CE. Comparisons were made to observed PDO/IPO indices and published reconstructions (LMR Online, LMR v2.1, D’Arrigo 2001). Wavelet analysis was used to assess temporal changes in PDO periodicities. Mechanism for recent δ18O increase: The δ18O series was compared with multiple Niño 3.4 SST proxy reconstructions/assimilations (PHYDA, LMR Online, Emile-Geay 2013, Mann 2009). Thirty-year running counts of El Niño and La Niña events (Australian Bureau of Meteorology list) from 1900–2011 CE were used to evaluate frequency changes and their relation to δ18O trends.

- Dating and time series: Using VMD-derived ENSO mode alignment and auxiliary biennial/annual modes, the upper 50 m of the QT core was dated to 1677–2011 CE (335 years). The D1 section (0–10 m) yielded 2011–1960 CE, consistent with the 1963 bomb layer detected at 8.8–9.2 m (this study dated it to 9.05–9.11 m). Average samples per annual layer in D1 were 6.38 with an annual accumulation rate of ~19.14 cm ice depth. Correlations between the dating δ18O series and the ENSO index across intervals were 0.43–0.72 (all p<0.01 in D1; others reported as positive). - Mode identification: MTM spectra consistently revealed significant peaks interpreted as ENSO (~1.1–1.6 cycles/m), biennial (~2.3–3.4 cycles/m), and annual (~5.1–6.6 cycles/m) modes across D1–D5. VMD with k=6 cleanly separated these modes; the ENSO IMF had the highest correlation with the original δ18O depth series. - PDO reconstruction and validation: The PDO reconstruction (1677–2011 CE) from the δ18O time series shows good coherence with observed indices (PDO, IPO) and reconstructions (LMR Online, LMR v2.1, D’Arrigo 2001) in decadal variability. For 1900–2011 CE, the reconstruction correlates with the 9-year running mean observed PDO index at r=0.57 (n=112, p<0.01). Spatial correlations (1900–2011 CE) between annual δ18O and SST show a significant PDO/IPO-like pattern across the North Pacific/All-Pacific. - Frequency shift in PDO: Wavelet analysis indicates that the 50–70-year periodicity dominated from ~1700–1900 CE, while the 25–35-year bidecadal periodicity became more active after ~1900 CE, consistent with expectations of shortened PDO periodicity under global warming. - Recent δ18O increase mechanism: The QT δ18O series exhibits an increasing trend over the past century. It aligns with increasing trends in multiple Niño 3.4 SST reconstructions/assimilations. Running correlations (1950–2000 CE) with PHYDA and LMR Online are 0.65 and 0.60 (p<0.01), respectively. From 1900–2011 CE, there were 26 El Niño and 18 La Niña years; 30-year running counts show increasing El Niño frequency and decreasing La Niña frequency, with a growing EN–LN difference. Given ENSO’s control on convection and precipitation δ18O in the AM/Indo-Pacific region, the increased frequency of El Niño events is identified as a primary factor driving the recent rise in QT δ18O.

By leveraging the robust ENSO–δ18O linkage in the Asian monsoon domain, the authors established a high-resolution chronology for a TP ice core despite weak seasonal layers—overcoming a key obstacle in interpreting inland TP ice cores. The dated δ18O series encodes Pacific decadal variability, enabling a reconstruction of PDO periodicities that is consistent with instrumental indices and other reconstructions. The identified shift from dominant multidecadal (50–70 years) variability pre-1900 to more active bidecadal (25–35 years) variability after 1900 supports hypotheses that global warming strengthens ocean stratification and accelerates Rossby waves, shortening the PDO timescale. The analysis also connects the 20th-century rise in δ18O to a documented increase in El Niño event frequency relative to La Niña, consistent with ENSO’s influence on regional convection and precipitation isotopes. These findings substantiate the use of central TP ice core δ18O as a sensitive recorder of Pacific climate variability and offer an independent line of evidence for PDO frequency shifts in the modern warming era.

The study introduces a spectral and VMD-based dating framework that exploits the preserved ENSO signal in δ18O to derive a 335-year chronology (1677–2011 CE) for the QT No. 1 ice core. Using the dated series, the authors reconstruct PDO variability and identify a century-scale shift in PDO periodicities—from 50–70-year dominance (pre-1900) to enhanced 25–35-year activity (post-1900). They also attribute the recent increase in ice core δ18O primarily to rising El Niño event frequency. This work advances TP ice core interpretation, demonstrates the feasibility of reconstructing Pacific decadal variability from inland TP isotopic records, and provides evidence for PDO timescale changes under global warming. Future research could extend dating to deeper sections with refined sampling, integrate additional proxies (chemistry, dust, gases) and multi-core comparisons, and more comprehensively assess other atmospheric drivers (e.g., westerlies influence) in this transition zone.

- Chronology assumptions: The dating relies on the persistence of the ENSO–δ18O relationship over centuries and on aligning IMF2 (ENSO mode) to reconstructed/observed ENSO indices; deviations in teleconnections over time could introduce uncertainties. - Weak seasonality: Below the radioactive layer (~9.2 m), seasonal signals in δ18O and chemistry are vague, complicating independent annual layer identification and increasing uncertainty. - Depth-age nonlinearity: Variable accumulation and layer thinning with depth can shift frequencies; the study limited analysis to the upper 50 m to mitigate aliasing, leaving the lower core undated in this framework. - Radiometric constraints: While 39Ar and other radiometric methods confirm broad age ranges (to ~1300 years), their temporal resolution is insufficient for near-annual validation of mid-depth chronology. - Regional circulation complexity: The central TP is a transition zone between westerlies and the Indian monsoon; potential influences beyond Pacific circulation (e.g., westerly variability) were not explored here, which may affect isotope signals. - Method sensitivity: VMD parameter selection (e.g., k, penalty α) and spectral bandwidth choices can affect mode separation; although k=6 yielded consistent modes, some sensitivity remains.