Wavelet-based verification of a relative paleointensity record from the North Pacific
KeywordsPaleomagnetism Relative paleointensity North Pacific Wavelet analysis Orbital cycles
vibrating sample magnetometer
anhysteretic remanent magnetization
ARM demagnetized at AF of 10 mT
remanent coercive force
mass-normalized magnetic susceptibility
mass-normalized ARM susceptibility
characteristic remanent magnetization
first-order reversal curve
isothermal remanent magnetization
IRM saturated by DC field of 1 T
IRM demagnetized at AF of 10 mT
maximum angular deviation
saturation remanent magnetization
natural remanent magnetization
principal component analysis
RPI proxy normalized by ARM
RPI proxy normalized by IRM
(local) wavelet power spectra
squared wavelet coherence
Temporal variations in geomagnetic paleointensity are fundamental for understanding the geodynamo process and have therefore been closely investigated via geological materials such as rocks and sediments (e.g., Tric et al. 1992; Laj et al. 2000; Channell et al. 2014). In particular, relative paleointensity (RPI) values recorded in sedimentary sequences provide insights into the long-term variations in the geomagnetic field and show a global coherency over millennial timescales despite environmental differences and possible physical uncertainties of the recording process (e.g., Tauxe et al. 2006; Roberts et al. 2013). With increasing numbers of well-constrained long RPI records, time-varying paleointensity records have been integrated into global models, which now extended through the entire Quaternary (e.g., SINT-800: Guyodo and Valet 1999; SINT-2000: Valet et al. 2005; EPAPIS-3Ma: Yamazaki and Oda 2005; PISO-1500: Channell et al. 2009; PADM2M: Ziegler et al. 2011). Such globally stacked RPI data have a practical advantage in constraining sedimentary chronology, particularly for marine or lake environments that lack carbon or oxygen isotope dating (Roberts et al. 2013). However, high-quality RPI records from various geographic locations are still necessary to improve and evaluate the global models. In recent decades, few RPI records from the North Pacific (e.g., Kent and Opdyke 1977; Roberts et al. 1997; Yamazaki 1999) have contributed to the global stacks for the past 2 Myr, including SINT-2000 and PADM2M. These North Pacific RPIs, covering relatively short periods of geological time (< 0.9 Myr), are also insufficient for understanding the longer-term behavior of the geomagnetic field. Although a longer RPI record for the time interval of ~ 0.35–1.60 Ma has been documented (Yamazaki and Kanamatsu 2007), this record is not included in any of the global stacks.
Previous studies have discovered ~ 40-kyr or ~ 100-kyr cycles from the North Pacific RPI records, using spectral or wavelet analyses, suggesting orbital forcing of geomagnetic field variations (Kent and Opdyke 1977; Yamazaki 1999; Yokoyama and Yamazaki 2000; Yamazaki and Kanamatsu 2007; Yokoyama et al. 2007). Searches for orbital periods in paleointensity have also been performed for high-resolution and globally stacked RPI signals such as SINT-800, SINT-2000, and PISO-1500 (e.g., Heslop 2007; Channell et al. 2009), as well as deep-sea floor magnetization records and beryllium production records (e.g., Thouveny et al. 2008; Saracco et al. 2009). Some studies have suggested relationships between orbital signals and paleomagnetic intensity and directional changes (e.g., Fuller 2006; Thouveny et al. 2008), but no statistically significant orbital periods have been detected from global RPI stack records (e.g., Heslop 2007; Channell et al. 2009; Ziegler et al. 2011).
Xuan and Channell (2008) found that intermittent orbital cycles in some local RPI records are due to lithological or climatic contamination. Because the origin of such orbital cycles can be connected to lithological variations of magnetic minerals, the coherence between RPI and magnetic parameters including the normalizer used in RPI estimation has also been tested (e.g., Yamazaki and Kanamatsu 2007; Xuan and Channell 2008). In particular, in the case of marine sediments containing a mixture of biogenic and detrital magnetites, the different physical properties of these magnetites could result in coherence between RPI and the normalizer owing to incomplete normalization (Roberts et al. 2013), thereby complicating the record of paleointensity (Ouyang et al. 2014). However, such possible contamination of RPI from the two magnetite assemblages has not been fully assessed. The wavelet method is a useful way of detecting periodicities in non-stationary time series, including geomagnetic records (e.g., Guyodo et al. 2000; Heslop 2007; Xuan and Channell 2008; Channell et al. 2009), compared with power spectral analysis without time resolution. In addition, cross-wavelet analysis and wavelet coherence provide a powerful test for a physical link between two time series (Torrence and Compo 1998; Torrence and Webster 1999; Grinsted et al. 2004) and have been applied in investigating the presence of a phase relationship between RPI records and sedimentary lithologies (e.g., Guyodo et al. 2000; Xuan and Channell 2008). For example, Guyodo et al. (2000) were the first to analyze the cross-wavelet powers and phases of RPI versus rock magnetic records and proved a secondary lithological overprint on the RPI record.
In this study, we present a high-quality RPI record for the last ~ 1.1 Myr from the Hess Rise in the North Pacific, verified using wavelet analysis. We test for the possible existence of orbital cycles on the RPI proxies and evaluate wavelet coherence with magnetic parameters, including the normalizer for RPI estimation. Then, the best RPI estimate is determined and further tested for a possible physical relationship with lithological changes in magnetic properties. Finally, the obtained RPI record is compared with global stacks and other records from near the study area.
Sampling and methods
Magnetic susceptibility was measured for all of the prepared samples using a Bartington MS2 magnetometer, and mass-normalized susceptibility (χ) was calculated by wet mass. Measurements of the natural remanent magnetization (NRM) and subsequent stepwise alternating field (AF) demagnetization were carried out at AFs of 5, 10, 15, 20, 25, 30, 40, 50, 70, and 90 mT, using a cryogenic magnetometer system (2G Enterprises model 760R) interfaced with an in-line AF demagnetizer located in a magnetically shielded room. AF demagnetization data were analyzed using PuffinPlot (Lurcock and Wilson 2012). Demagnetization data for each sample were displayed on a vector diagram, and the characteristic remanent magnetization (ChRM) was isolated using principal component analysis (PCA; Kirschvink 1980) with a line-fitting method involving anchoring at the origin. The quality of the ChRM direction was evaluated using the maximum angular deviation (MAD), calculated from the difference between the anchored line and demagnetization data.
After demagnetization experiments, all samples were subjected to the acquisition of laboratory-induced magnetization of anhysteretic remanent magnetization (ARM) and isothermal remanent magnetization (IRM). ARM was acquired by applying a direct current (DC) field of 0.05 mT superimposed on a peak AF of 100 mT. A DC field of 1 T was applied to saturate the remanence (IRM1T) using an ASC Scientific IM-10 impulse magnetizer. Subsequently, both ARM and IRM were stepwise demagnetized in a peak AF of up to 50 mT with a 10 mT increment. The laboratory-induced magnetizations and their AF demagnetizations were used as normalizers to calculate RPI by ratios (NRM/normalizer) and by the so-called pseudo-Thellier method (Tauxe et al. 1995), respectively.
ARM and IRM1T were used as concentration-dependent magnetic parameters after mass normalization; ARM was normalized to a DC field of 0.05 mT (= 39.79 A/m) to calculate the mass-normalized ARM susceptibility (χARM). IRM1T was reacquired and then applied at 0.3 T in the opposite direction to obtain IRM− 300mT. S-ratios were calculated as S300 = 0.5 × (1 − IRM− 300mT/IRM1T). Values of S-ratio close to 1 are regarded as indicating a relative abundance of low-coercivity magnetic minerals such as magnetite (Bloemendal et al. 1992). To determine the grain size of magnetic minerals and their assemblages, hysteresis and first-order-reversal curve (FORC) analyses were performed on selected freeze-dried samples using a Princeton Measurement Corporation vibrating sample magnetometer (MicroMag 3900 VSM). Hysteresis parameters were obtained from 33 hysteresis curves, including the bulk coercive force (Bc), the remanent coercive force (Bcr), the saturation magnetization (Ms), and the saturation remanence (Mrs). Bcr values were obtained from backfield measurements. For 10 selected samples, 156 FORCs were measured with a field increment of 1.5 mT up to a maximum applied field of 1 T. FORC data were processed with the FORCinel software package (Harrison and Feinberg 2008) using the VARIFORC built-in function for variable smoothing described by Egli (2013).
Relative abundances of Ba and Ti were measured using an ITRAX X-ray fluorescence (XRF) core scanner. The XRF measurements were made at a resolution of 0.5 cm for the archive half of the core, and five-point moving averages for the obtained data were used to calculate the Ba/Ti ratio.
To detect the presence of orbital periods in time series for magnetic variations, (local) wavelet power spectra (WPS) were analyzed using the Morlet wavelet following Torrence and Compo (1998) and Grinsted et al. (2004). WPS can identify significant orbital powers in time–frequency space. In addition, to identify the physical relationship between two time series, cross-wavelet spectra (XWS) and squared wavelet coherence (SWC) were analyzed. For the two WPSs, XWS exposes areas with common power and the phase relationship, and the SWC highlights phase-locked behavior within the area where XWS reveals high common power (Grinsted et al. 2004). Before performing these analyses, all input data were resampled by linear interpolation to 1-kyr resolution, linearly detrended, and normalized to zero mean and unit variance.
Rock magnetic properties
Values of S300 are in the range of 0.93–0.96, with a mean of 0.94. Relatively low Bcr values of 31–47 mT with a mean of 40 mT and narrow widths of hysteresis loops (Bc of < 21 mT) are observed in the selected samples. Such high S300 values near 1.0 with low Bcr values indicate a predominance of low-coercivity ferrimagnetic minerals (Thompson and Oldfield 1986) throughout the core.
Concentration-related magnetic parameters show cyclic fluctuations in depth from a few decimeters to a meter with value ranges of less than a factor of 3: χ = ~ 0.1–0.2 × 10−6m3/kg, χARM = ~ 2.0–4.5 × 10−6m3/kg, and IRM1T = ~ 2–4 × 10−3Am2/kg (Fig. 5b–d). Of the parameters, χARM in a sediment core is sensitively acquired by SD magnetite with a weak magnetostatic interaction, such as biogenic magnetite, whereas values of χ and IRM1T are related to the total quantity of all magnetic minerals, with a negligible grain-size effect (Thompson and Oldfield 1986). Therefore, the ratios of χARM to IRM1T and χARM to χ, which are conventional grain-size proxies, are commonly considered as representing the relative contributions of finer biogenic magnetite and coarser detrital magnetic minerals in marine sediments (e.g., Yamazaki 2009; Yamazaki and Ikehara 2012). χARM/IRM1T and χARM/χ show similar variation, with amplitudes smaller than twofold. Above a depth of ~ 400 cm, their amplitudes show similar peaks and troughs to χARM (Fig. 5c, e), indicating that grain-size parameters (χARM/χ and χARM/IRM1T) in this study are strongly linked to quantitative variation in the concentration of finer SD magnetite rather than of coarser magnetic minerals. As identified in FORC analysis, the most distinctive feature in the diagrams is that the biogenic SD magnetite signal is enhanced as the central peak (Fig. 4b) in the sample showing higher χARM (338.5 cm depth) in comparison with the sample showing lower χARM sample (227.5 cm depth). Therefore, it is reasonable to infer that temporal variations in χARM/χ and χARM/IRM1T reflect a relative fluctuation of biogenic SD magnetite, at least above a depth of ~ 400 cm. From the perspective of RPI estimation, the variation in bulk magnetic properties of less than a factor of 2–3 throughout the core represents stable concentrations and grain sizes of NRM-carrying grains, namely SD–PSD magnetites, which fulfills the conventional criteria of Tauxe (1993).
Relative paleointensity proxies
RPI proxies were also calculated using the slope method of Tauxe et al. (1995). The RPI slopes were determined from NRM lost versus ARM lost (or IRM lost) at 10–40 mT demagnetization steps where stable ChRM directions remain (NRM10–40mT/ARM10–40mT and NRM10–40mT/IRM10–40mT; Fig. 6), and are presented as RPIARM and RPIIRM, respectively, for convenience. The linear correlation coefficient (r) was calculated to monitor the quality of the slope fit (Fig. 6), following Channell et al. (2002). The r values are generally close to 1 (> 0.95), although some specific intervals show a drop in r value around the polarity transitions corresponding to intervals of high MAD values (Fig. 3). For the whole core, the mean r values are higher than 0.99 (0.9932 and 0.9928 for RPIARM and RPIIRM, respectively), indicating that both ARM and IRM have similar coercivity spectra with NRM and can thus be used as appropriate normalizers in RPI estimation.
Our magnetic grain-size proxies (χARM/IRM1T and χARM/χ) show distinctive cyclic fluctuations throughout the core (Fig. 7b) that are similar to the fluctuations in ARM/IRM of a nearby core, KR0310-PC1 (Fig. 7a; Yamazaki and Kanamatsu 2007). For sediments older than ~ 700 ka, relatively sharp glacial lows or interglacial highs are correlated despite the smoothed variations of our grain-size proxies (Fig. 7a, b). For the < ~700 ka record, the uncertainty of age correlations was overcome by correlating Ba/Ti (Ba/Al) records to reference cores SO202-39-3 (Korff et al. 2016) and ODP 145-882 (Jaccard et al. 2010) with a mean age window of ~ 50 kyr for correlations (Fig. 7c–e). In the North Pacific, Ba/Ti or Ba/Al records corresponding to productivity changes are closely connected to paleoclimatic variations (Jaccard 2005; Jaccard et al. 2010) and are useful for establishing an age model (e.g., Korff et al. 2016; Shin et al. 2018). Within the Ba/Ti age framework, magnetic grain-size proxies between two adjacent cores also have temporally similar fluctuations (Fig. 7a, b). Consequently, a total of 25 tie points was assigned, and the age of the bottom of the core was estimated to ~ 1.07 Ma. With linear interpolation between tie points, the sedimentation rate was calculated as varying from ~ 0.2 to ~ 1.4 cm/kyr with a mean of ~ 0.7 cm/kyr (Fig. 7h). Age uncertainties of the reference records and our correlations are expected to be within ~ 20 kyr.
In the age model, RPI proxies show overall temporal variations consistent with a global stack record such as SINT-2000 (Fig. 7f, g). However, the delayed age of the M–B boundary (~ 794 ka) is probably caused by surficial mixing and pDRM lock-in delay accompanied by a lower sedimentation rate (~ 0.4 cm/kyr during ~ 900–750 ka; Fig. 7h). Considering the recently determined Be-proxy age constraints on the M–B boundary (~ 772 ka at mid-reversal: Simon et al. 2018 and Valet et al. 2019), the age delay of ~ 22 ka may indicate lock-in depths of ~ 8 cm below the mixing layer for the M–B boundary in the studied core.
Orbital cycles and possible climatic contamination of RPI estimates
Orbital frequencies in RPI records involve some sources of uncertainty, such as sediment ages and resolutions. In this study, low data resolution (i.e., averaging ~ 3 kyr over a sample) and related age uncertainties may obscure the potential signals of short-term orbital forcing such as obliquity (~ 41 kyr) and precession (~ 23 kyr) in the RPI record. In particular, the highly smoothed RPI signals and pDRM lock-in delay for our > ~ 700 ka record can produce noisy signals at both orbital and non-orbital frequencies. However, our RPI proxies for < ~ 700 ka share generally similar peaks/troughs with SINT-2000 at a scale of ~ 10–30 kyr (Fig. 7f, g), suggesting a good age control and detectability of longer orbital parameters such as eccentricity (~ 100 kyr). For example, an orbital cycle with a ~ 100-kyr period has been consistently reported in low-resolution North Pacific RPI records from power spectral analysis (e.g., Yamazaki 1999; Yamazaki and Kanamatsu 2007), although its temporal continuity and possible climatic/lithological contamination were not analyzed. Therefore, we applied wavelet analysis to detect the potential orbital frequencies in our RPI proxies and to test for the presence of coherence with lithological factors.
Magnetic parameters used for the normalizer (ARM and IRM after demagnetization at 10 mT steps) were tested for their possible link with the orbital contamination of RPI proxies. For the time interval 900–700 ka, XWS maps between RPI and the corresponding normalizers (Fig. 8d, e) show significant common power at the 100-kyr orbital period, but their phase arrows (phase relationships) are variable. In the SWC maps of Fig. 8h, i, a significant physical relationship between RPIs and normalizers is not found for the time interval 900–700 ka. Therefore, orbital contamination of the 100-kyr period for RPI during the time interval 900–700 ka (Fig. 8a, b) is not directly induced by the normalizers.
Interestingly, RPIARM has spectral coherence at the 100-kyr period with the ARM record during 400–200 ka, showing high common power and a consistent anti-phase relationship (negative correlation) in XWS and SWC maps (left-pointing arrows in Fig. 8d, h). Such significant wavelet coherence with the anti-phase relationship would indicate inefficient normalization and overcompensation when ARM is used as a normalizer. Previously, spectral coherence between RPI and its normalizer ARM at the 100-kyr period has been suggested to be an effect of the magnetostatic interaction between detrital SD magnetites hampering ARM acquisition (Yamazaki and Kanamatsu 2007; Yamazaki 2008), but our FORC diagrams (Fig. 4b) confirm an intrinsic signature of PSD detrital magnetite with the intraparticle interaction (Roberts et al. 2013). In a mixture of non-interacting biogenic magnetite and detrital magnetite, relative changes in the two different origins could affect ARM normalization in RPI estimation in terms of different coercivity distribution and remanence acquisition efficiency (Ouyang et al. 2014). In the present study, ARM shows a prominent cyclicity of the 100-kyr period at < 800–700 ka (Additional file 1: Figure S1a), reflecting glacial–interglacial cyclicity of fine magnetite concentration after the mid-Pleistocene transition. In contrast, ARM/IRM (calculated using ARM and IRM demagnetized at 10 mT) shows the most prominent 100-kyr cycles at around 400–200 ka and also has a consistent anti-phase relationship with RPIARM (Additional file 1: Figure S1b–d). Such consistency suggests that enhanced cyclic fluctuations of biogenic magnetite over detrital components result in overcompensation of RPI using ARM normalization.
As IRM does not show coherence with RPIIRM (Fig. 8e, i), IRM was considered as the more suitable normalizer and for which wavelet coherence of RPIIRM with grain-size variation was additionally tested (Fig. 8f, j). The SWC maps show little coherence and variable phase relationships at orbital periods, indicating that the 41-kyr and 100-kyr orbital contamination of RPIIRM is not directly induced by the grain-size proxies. We speculatively regard the intermittent presence of orbital periods in RPIIRM as climatic contamination prompted by non-magnetic factors. In addition, smoothed RPI signals due to low resolution and the lock-in effect could be responsible for the significant orbital contamination during 900–700 ka. It is also noted that the coexistence of 41-kyr and 100-kyr orbital periods at ~ 700 ka in the WPS of RPI proxies and NRM (Fig. 8a–c) might reflect climatic contamination linked to the mid-Pleistocene climate transition when dominant orbital power shifted from 41 kyr to 100 kyr.
Global consistency of RPI estimates
Overall, the RPI record estimated from NPGP1401-2A is highly consistent with the millennial-scale pattern of variation seen in the global records (Fig. 9), indicating a well-defined paleointensity record for the last ~ 1.1 Myr. Although our RPI record is not free of orbital contamination during certain time intervals (Fig. 8b), its high level of consistency with the global stacks suggests that the dipolar origin of the geomagnetic field was not receded in the RPI signals. Our well-determined RPI estimates also represent an improvement in the North Pacific records, in particular during the Brunhes Chron (Fig. 9 and Additional file 3: Table S2).
Paleomagnetic and rock magnetic analyses for deep-sea sediments in the North Pacific were performed to obtain a reliable RPI record. Down-core variation in inclination and declination revealed two polarity reversals, correlating to the Matuyama–Brunhes and upper Jaramillo transitions. Magnetic signals were carried dominantly by a mixture of biogenic SD magnetite and detrital PSD magnetite, with bulk concentration and grain size varying by less than a factor of 3. The RPI proxies calculated by ARM and IRM normalization, using a slope method, appear to be well defined, as judged by NRM and normalizers having almost the same coercivity spectra (r values of > 0.99). The age of the bottom sediments of the studied core was determined as ~ 1.07 Ma based on the correlation of our RPI proxies to a global stack. Further wavelet analysis revealed the intermittent presence of ~ 100-kyr and ~ 41-kyr orbital periods in the RPI proxies. In particular, RPIARM shows a significant coherence and anti-phase relationship with the normalizer during 400–200 ka, indicating an over-normalization of ARM. The presence of the 100-kyr orbital period in ARM/IRM is also prominent showing a coherency with RPIARM in the same time interval, suggesting that a differing ability in remanence acquisition between biogenic and detrital magnetite is probably reflected as contamination. In contrast, a coherence test for RPIIRM shows no physical link between orbital periods and magnetic parameters related to the concentration and grain size of magnetic minerals, indicating negligible lithological contamination. Although RPIIRM is not fully free from an amount of orbital contamination, likely caused by climate changes, its high level of consistency with global stacks indicates a dipolar signal.
The authors thank the editor Dr. Chuang Xuan for handling this paper. Valuable comments made by the editor and three anonymous reviewers greatly improved the paper. The paleomagnetic and rock magnetic measurements were made in KIGAM, Korea, by instrumental supports from Dr. Yoon Soo Lee. The XRF measurements were performed by Dr. Dhongil Lim at the Library of Marine Samples, KIOST.
JYS carried out the sample preparation, data collection and analysis, and wrote the manuscript. YY contributed to the discussion. WK designed this work, sampled a study core, and supervised the paper preparation. All authors read and approved the final manuscript.
This work was financially supported by the Ministry of Oceans and Fisheries (PM61151) and the Korea Institute of Ocean Science and Technology (PE99724 and PE99711).
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
- Gradstein FM, Ogg JG, Schmitz MD, Ogg GM (2012) The geologic time scale. Elsevier, AmsterdamGoogle Scholar
- Kirschvink JL (1980) The least-squares line and plane and the analysis of palaeomagnetic data. Geophys J Int 62:699–718. https://doi.org/10.1111/j.1365-246X.1980.tb02601.x CrossRefGoogle Scholar
- Korff L, von Dobeneck T, Frederichs T, Kasten S, Kuhn G, Gersonde R, Diekmann B (2016) Cyclic magnetite dissolution in Pleistocene sediments of the abyssal northwest Pacific Ocean: evidence for glacial oxygen depletion and carbon trapping. Paleoceanography 31:600–624. https://doi.org/10.1002/2015PA002882 CrossRefGoogle Scholar
- Mark DF, Renne PR, Dymock RC, Smith VC, Simon JI, Morgan LE, Staff RA, Ellis BS, Pearce NJG (2017) High-precision 40 Ar/39 Ar dating of pleistocene tuffs and temporal anchoring of the Matuyama–Brunhes boundary. Quat Geochronol 39:1–23. https://doi.org/10.1016/j.quageo.2017.01.002 CrossRefGoogle Scholar
- Ouyang T, Heslop D, Roberts AP, Tian C, Zhu Z, Qiu Y, Peng X (2014) Variable remanence acquisition efficiency in sediments containing biogenic and detrital magnetites: implications for relative paleointensity signal recording. Geochem Geophys Geosyst 15:2780–2796. https://doi.org/10.1002/2014GC005301 CrossRefGoogle Scholar
- Saracco G, Thouveny N, Bourlès DL, Carcaillet JT (2009) Extraction of non-continuous orbital frequencies from noisy insolation data and from palaeoproxy records of geomagnetic intensity using the phase of continuous wavelet transforms. Geophys J Int 176:767–781. https://doi.org/10.1111/j.1365-246X.2008.04057.x CrossRefGoogle Scholar
- Simon Q, Thouveny N, Bourlès DL, Valet J-P, Bassinot F, Ménabréaz L, Guillou V, Choy S, Beaufort L (2016) Authigenic 10Be/9Be ratio signatures of the cosmogenic nuclide production linked to geomagnetic dipole moment variation since the Brunhes/Matuyama boundary. J Geophys Res Solid Earth 121:7716–7741. https://doi.org/10.1002/2016JB013335 CrossRefGoogle Scholar
- Simon Q, Bourlès DL, Thouveny N, Horng C-S, Valet J-P, Bassinot F, Choy S (2018) Cosmogenic signature of geomagnetic reversals and excursions from the Réunion event to the Matuyama–Brunhes transition (0.7–2.14 Ma interval). Earth Planet Sci Lett 482:510–524. https://doi.org/10.1016/j.epsl.2017.11.021 CrossRefGoogle Scholar
- Suganuma Y, Yokoyama Y, Yamazaki T, Kawamura K, Horng CS, Matsuzaki H (2010) 10Be evidence for delayed acquisition of remanent magnetization in marine sediments: implication for a new age for the Matuyama–Brunhes boundary. Earth Planet Sci Lett 296:443–450. https://doi.org/10.1016/j.epsl.2010.05.031 CrossRefGoogle Scholar
- Torrence C, Compo GP (1998) A practical guide to wavelet analysis. Bull Am Meteorol Soc 79:61–78. https://doi.org/10.1175/1520-0477(1998)079%3c0061:APGTWA%3e2.0.CO;2 CrossRefGoogle Scholar
- Torrence C, Webster PJ (1999) Interdecadal changes in the ENSO–monsoon system. J Clim 12:2679–2690. https://doi.org/10.1175/1520-0442(1999)012%3c2679:ICITEM%3e2.0.CO;2 CrossRefGoogle Scholar
- Valet J-P, Bassinot F, Simon Q, Savranskaia T, Thouveny N, Bourlés DL, Villedieu A (2019) Constraining the age of the last geomagnetic reversal from geochemical and magnetic analyses of Atlantic, Indian, and Pacific Ocean sediments. Earth Planet Sci Lett 506:323–331. https://doi.org/10.1016/j.epsl.2018.11.012 CrossRefGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.