Increased tropical South Pacific western boundary current transport over the past century

The wind-driven meridional overturning circulation between the tropical and subtropical oceans is important for regulating decadal-scale temperature fluctuations in the Pacific Ocean and globally. An acceleration of the overturning circulation can act to reduce global surface temperature as ocean stores more heat. The equatorward low-latitude western boundary current represents a key component of the meridional circulation cell in the Pacific and a major source of water mass for the Equatorial Undercurrent, yet long-term observations of its transport are scarce. Here we demonstrate that the 15N/14N ratio recorded by Porites spp. corals in the western tropical South Pacific is sensitive to the exchanges of water masses driven by the western boundary transport. Using a 94-year coral record from the Solomon Sea, we report that the 15N/14N ratio declined as the global surface temperature rose. The record suggests that the South Pacific western boundary current has strengthened in the past century, and it may have contributed to the reported strengthening of the Equatorial Undercurrent. In addition, the 15N/14N record shows strong decadal variability, indicative of weaker equatorial Pacific upwelling and stronger western boundary transport when the eastern equatorial Pacific is in the warm stage of the Pacific Decadal Oscillation. The low-latitude western boundary current in the South Pacific Ocean strengthened as climate warmed over the past 100 years, according to a coral nitrogen isotope record from the Solomon Sea.

The wind-driven meridional overturning circulation between the tropical and subtropical oceans is important for regulating decadal-scale temperature fluctuations in the Pacific Ocean and globally. An acceleration of the overturning circulation can act to reduce global surface temperature as ocean stores more heat. The equatorward low-latitude western boundary current represents a key component of the meridional circulation cell in the Pacific and a major source of water mass for the Equatorial Undercurrent, yet long-term observations of its transport are scarce. Here we demonstrate that the 15 N/ 14 N ratio recorded by Porites spp. corals in the western tropical South Pacific is sensitive to the exchanges of water masses driven by the western boundary transport. Using a 94-year coral record from the Solomon Sea, we report that the 15 N/ 14 N ratio declined as the global surface temperature rose. The record suggests that the South Pacific western boundary current has strengthened in the past century, and it may have contributed to the reported strengthening of the Equatorial Undercurrent. In addition, the 15 N/ 14 N record shows strong decadal variability, indicative of weaker equatorial Pacific upwelling and stronger western boundary transport when the eastern equatorial Pacific is in the warm stage of the Pacific Decadal Oscillation. Instrumental and palaeoclimate records over the past century and millennia show clear decadal climate variability in the tropical Pacific 1 , which modulates the rate of global temperature change in response to the rising level of greenhouse gases in the atmosphere. For example, decadal surface cooling in the eastern equatorial Pacific may be responsible for the slowdown of global warming between the 1940s and 1970s and post-1998, weakening the anthropogenic warming trend during the past century 2,3 . Decadal variability in the tropical Pacific also modulates the behaviour and characteristics of the El Niño/Southern Oscillation (ENSO) 4 .
Climate models suggest that the decadal variability in the tropical Pacific arises from oceanic processes involving the upper-ocean Article https://doi.org/10.1038/s41561-023-01212-4 shallow flow ('tropical' mode) and the LLWBC transport that is probably forced remotely by subtropical winds ('subtropical' mode) 16,17 . During an El Niño, the reduced equatorial easterlies are associated with negative wind stress curl anomalies in the South Pacific. The decreased equatorial wind stress reduces the strength of the STCs, while the negative wind stress curl anomalies enhance the equatorward transport in the Solomon Sea, bringing more waters of subtropical origin towards the Equator 18 . On decadal and longer timescales, modelling results suggest that the strength of the Solomon Sea LLWBC transport is strongly correlated with the poleward transport in the interior, indicating a tendency for the LLWBC to compensate for the interior transport changes 6,19,20 . However, existing observational time series are not long enough to determine decadal variations in LLWBC transport.

Nitrogen isotopes as a proxy for LLWBC transport
The isotopic composition of marine nitrogen (N) and its temporal variations can help constrain the LLWBC transport in the Solomon Sea. In the western tropical Pacific (WTP), the 15 N/ 14 N ratio (or δ 15 N) in both the upper thermocline nitrate (<400 m) and the near-surface suspended particulate organic nitrogen (PON) show clear tropical-to-subtropical changes on both sides of the Equator (Fig. 1b and Extended Data Fig. 1). The high δ 15 N values close to the Equator in the WTP arise from processes associated with upwelling and productivity along the Equator. Easterly winds shoal the thermocline and drive upwelling of cool, nutrient-rich thermocline waters along the Equator and in the eastern margin of the basin, yet biological production in the upwelling regions is iron-limited 21 . As a result, the eastern and central equatorial Pacific overturning circulation known as subtropical-tropical cells (STCs) [5][6][7] . The STCs are primarily driven by surface wind forcing 8 . Subduction of subtropical water in the eastern Pacific, which flows westward and equatorward in the upper pycnocline layers through both western boundary and interior pathways, feeds into the Equatorial Undercurrent and upwells into surface equatorial water before returning to the subtropics in the surface Ekman layer 8 . Observations suggest a slowdown of the interior transport since the end of the 1970s along with rapid warming of the eastern equatorial Pacific and globally 5 , supporting the importance of the tropical-subtropical link in controlling tropical Pacific decadal variability. However, this remains very challenging to demonstrate observationally, owing to the insufficient instrumental record of oceanic processes.
The Solomon Sea in the western tropical South Pacific provides a western boundary connection between the subtropics and the Equator (Fig. 1a). The equatorward low-latitude western boundary current (LLWBC) transport through the Solomon Sea represents an important fraction of pycnocline transport in the meridional circulation cell 9 and a major source for the Equatorial Undercurrent (EUC) 10 that upwells along its eastward route and directly influences equatorial Pacific sea surface temperature 11,12 . The LLWBC transport in the surface layer of the Solomon Sea also contributes to the variability of the warm water volume (WWV), which has been suggested to impact ENSO dynamics 13 . As a result, changes in either the amount or properties of the water coming through the Solomon Sea have the potential to create global-scale feedbacks 5,14,15 .
Glider-based observations of the Solomon circulation indicate that the equatorward flow is composed of both a tropical-wind-driven (EEP and CEP) are characterized by incomplete consumption of the major nutrients (nitrate and phosphate), with substantial nutrient concentrations remaining in their surface waters (Fig. 1a) to be transported westward and poleward. When nitrate is consumed by phytoplankton, the lighter 14 N isotope is preferentially incorporated, causing the remaining nitrate pool (and thus also the PON subsequently produced from it) to become progressively enriched in the heavier 15 N isotope 22 . As a result, the δ 15 N of the surface nitrate and the suspended PON is observed to increase from east to west 23,24 . The high δ 15 N signal is also incorporated into sinking particles, the remineralization of which can elevate thermocline nitrate δ 15 N on a regional basis 23,25 (Fig. 1b, Extended Data Fig. 1 and Supporting Information). As a result, although surface nitrate is exhausted, the high δ 15 N propagates further into the WTP including the Solomon Sea by cycles of production and regeneration of PON (and possibly also dissolved organic N) that is elevated in δ 15 N (refs. 24,25).
On the other hand, the subtropical waters flowing into the Solomon Sea have distinctively low δ 15 N in the upper thermocline, due to the N input from regional N 2 fixation 26 ( Fig. 1b and Extended Data Fig. 1). As a result, the δ 15 N changes in the upper Solomon Sea water column can be strongly influenced by altering the relative proportion of equatorial vs subtropical waters on different timescales. During El Niño events or under an El Niño-like climate state, the intensification of the equatorward flow in the Solomon Sea would tend to lower the δ 15 N in the upper ocean.
These subtropical waters are carried to the western boundary by the South Equatorial Current (SEC), which bifurcates in the Coral Sea at or before the coast of Australia, turning north into the Solomon Sea towards the Equator, and south into the East Australian Current (EAC) (Fig. 1a). Given the meridional δ 15 N gradient, the western boundary currents would work to redistribute δ 15 N anomalies northward towards the Equator and southward to the subtropical gyre. When the SEC intensifies, such as during and after an El Niño 27 , it increases the equatorward transport to the Solomon Sea 28 and lowers the δ 15 N in the upper ocean.
In the meantime, it should also increase the southward transport along the Australian coast 28 and increases the δ 15 N there. As a result, the δ 15 N changes in the Solomon Sea should also have an opposing phasing with δ 15 N changes along the Australian coast on various timescales.
Lacking seasonal-to-interannual-resolved nitrogen isotope measurements or sample collections in the Solomon Sea, we turn to a subseasonally resolved coral record from the Solomon Sea to test our hypothesis (Extended Data Fig. 2 and Methods). We also compare the Solomon Sea record with a published coral record in the Australian coast (19.15° S, 146.87° E) 29 . Scleractinian corals acquire their N from the environment, primarily by feeding on zooplankton and PON in surface waters under low-nutrient conditions in the WTP 30 . To facilitate the growth of coral skeleton, corals produce a small amount of organic material (for example, polysaccharide and proteins) in the extracellular calcifying medium, which is subsequently preserved in the mineral matrix 31 . The coral skeleton protects the organic matter against the diagenetic loss and exogenous N contamination that introduces uncertainty into non-fossil-bound archives of organic matter 32 . Thus, coral skeletal δ 15 N (CS-δ 15 N) can record upper-ocean δ 15 N changes in the WTP.
On seasonal timescales, the Solomon Sea CS-δ 15 N changes in accordance with variations in the LLWBC transport estimated from glider data 16 and from satellite and proxy data 17 since 2004 (Fig. 1c). The CS-δ 15 N is lower in austral winters when transport of cold and saline subtropical waters into the Solomon Sea is enhanced 33 . On interannual timescales, the Solomon Sea CS-δ 15 N has a negative correlation with the temperature anomalies in the CEP and EEP, with lower CS-δ 15 N when the surface of CEP and EEP are abnormally warm (r = −0.50, P = 0.07; the interannual variability in CS-δ 15 N is defined by ensemble empirical mode decomposition or ensemble empirical mode decomposition (EEMD) analyses 34 and the Niño 3.4 index sea surface temperature (SST) anomaly is smoothed by 3-month running average; Methods). On decadal timescales, the Solomon Sea CS-δ 15 N has a strong negative correlation with changes in the CS-δ 15 N from the Australian coast with a negative phase (r = −0.77, P = 0.01, detrended and 7-year running average for both CS-δ 15 N records) (Figs. 2 and 3; Methods). This evidence supports our hypothesis that the δ 15 N dynamics in the WTP are regulated by basin-wide processes associated with the western boundary currents.
Over the past century, our CS-δ 15 N record in the Solomon Sea is characterized by strong interannual and decadal changes superimposed on a long-term declining trend (Fig. 2b). In the following, we will discuss the decadal variability and long-term trend in the CS-δ 15 N record before we return to the implications for tropical climate, that is, the ENSO cycles.

Decadal changes in the coral skeletal nitrogen isotope ratio
The variability in the Solomon Sea CS-δ 15 N on decadal timescales is strongly correlated with basin-wide changes in the SST ( Fig. 3d and Extended Data Fig. 3). In particular, the CS-δ 15 N changes appear to be strongly correlated with the SST changes associated the Pacific Decadal Oscillation (PDO), which reflect the leading mode of climate variability in the Pacific on decadal timescales ( Fig. 3d and Extended Data Fig. 3). The PDO manifests as a low-frequency El Niño-like pattern of climate variability with a warm tropical Pacific and weakened trade winds On decadal timescales, the Solomon Sea CS-δ 15 N is lower and the Australian coast CS-δ 15 N is higher during positive PDO phases, consistent with basin-wide processes regulating the western boundary currents and redistributing the meridional δ 15 N anomalies in the WTP (Figs. 2 and 3). The trade winds weaken during positive PDO phases, depressing Ekman divergence along the Equator and slowing down the pycnocline convergence in the interior 35,36 . In the meantime, the LLWBC intensifies due to enhanced Sverdrup transport driven by a negative wind stress curl anomaly between the weakened tropical trade winds and subtropical winds 19 (Fig. 3). CS-δ 15 N changes in the Solomon Sea appear to lag behind changes in the tropical winds, especially since the 'climate regime shift' in 1976 (Extended Data Fig. 6), but are generally in phase with changes in the subtropical wind stress curl in the South Pacific (Extended Data Fig. 6). The overall correlation is strong between CS-δ 15 N and wind stress curl in the South Pacific (Extended Data Fig. 6e) (r = 0.52, P = 0.10, with CS-δ 15 N lagging by 9 months; Methods). These data thus suggest that the decadal variability in the equatorward LLWBC transport through the Solomon Sea, as inferred from the CS-δ 15 N record, is remotely controlled by off-equatorial processes.
Assuming a simple linear wind-driven circulation, our data lends qualitative support for the modelling studies that demonstrate subtropical regulation of the STCs and the equatorial Pacific mean climate state 7 . The step-like decrease in CS-δ 15 N in the early 1980s follows the observed thermocline shoaling in the WTP 37 (Extended Data Fig. 7b) and slowdown of the interior pycnocline transport 5 . These data together corroborate the view that variability in the STCs may contribute to decadal changes in the equatorial SST (Fig. 4).

Long-term decline in the Solomon Sea coral skeletal nitrogen isotope ratio
Underlying the interannual and interdecadal changes, CS-δ 15 N has a long-term declining trend (−0.018 ‰ yr −1 , p < 0.001) that parallels the global warming trend (Fig. 2). The Australian coast CS-δ 15 N indicates a relatively stable δ 15 N in the subtropical South Pacific since the early twentieth century (Fig. 2c) 29 , suggesting that the long-term decline in the Solomon CS-δ 15 N cannot be explained by an overall decline in the subtropical δ 15 N over the last century. Declines in the tropical δ 15 N end member may result from reduced upwelling in the EEP and CEP, representing an 'El Niño-like' mean state change over the twenty-first century. However, the pattern of the observed ocean surface temperature trends since the 1950s is characterized by notable warming in the tropical western Pacific and Indian oceans and a slight cooling along the equatorial eastern Pacific, suggesting a 'La Niña-like' mean state change. This suggests that a long-term decline in the tropical δ 15 N is unlikely to explain the Solomon CS-δ 15 N change.
The long-term decline in the Solomon CS-δ 15 N is best explained by an intensification of the LLWBC transport through the Solomon Sea. This is supported by radiocarbon-based evidence for increase of the subtropical waters in the Solomon Sea 38 (Fig. 2d). When the PDO returned to negative phase and the trade winds restrengthened in the late 1990s, the CS-δ 15 N did not return to the high values of the previous negative PDO phase ; this evidence suggests that processes other than the equatorial trade winds may be responsible for the long-term intensification of the LLWBC transport. Both models and historical data confirm a southward expansion of the southern edge of the Southern Hemisphere Hadley cell since 1979 (ref. 39). The associated intensification of the southeasterly trade winds and off-equatorial wind stress curl change in the South Pacific could enhance the LLWBC transport 40 . This mechanism is supported by the observed southward migration in the SEC bifurcation latitude 41 (Extended Data Fig. 7a) and an overall better correlation between CS-δ 15 N and the wind stress curl in the South Pacific (Extended Data Fig. 6).
Because the waters from the Solomon Sea are the main source of the EUC 10,42,43 , strengthening of the LLWBC transport inferred from our record would help to explain the observed long-term increase in the EUC transport 40,44 (Fig. 2e). This may then imply a growing importance of the LLWBC transport in regulating the water mass characteristics, heat/salt budget and climate dynamics in the equatorial Pacific and globally. In addition, recent works suggest that the LLWBC transport through the Solomon Sea is a major source of iron to the EUC 45 , which eventually upwells and fuels productivity in the EEP. The strengthening in the LLWBC may then work to relieve iron limitation in the EEP in the future.

Implications for the tropical climate variability
We now return to discuss how LLWBC transport could contribute to the tropical climate variability dominated by the ENSO cycles on interannual timescales. While we anticipate enhanced LLWBC transport during and after an El Niño event, the data suggest a weakening relationship between ENSO activity and CS-δ 15 N after the 1980s, caused by weakening correlation with La Niña conditions (Extended Data Fig. 8). We attempt to understand this nonlinear response to El Niño vs La Niña events with the 'recharge oscillator' model 13 .
The model states that the depth of the mean thermocline, and hence the WWV above it, plays an important dynamical role in the ENSO cycle by controlling the temperature of the waters upwelled in the eastern equatorial Pacific, with a deeper mean thermocline resulting in the upwelling of warmer waters. As anthropogenic warming stratifies the upper ocean and shoals the mean thermocline along the Equator (Extended Data Fig. 7), upwelling of thermocline waters in the CEP and EEP during La Niña conditions will more effectively strengthen the eastwest temperature gradient, while El Niño conditions would require more WWV to be transported from the western Pacific warm pool 46 . On the one hand, periods of low LLWBC will become less important in raising the frequency of La Niña events. On the other hand, enhanced LLWBC transport would supply additional WWV, compensating for global warming's tendency to shoal the tropical thermocline, allowing the continued development of strong basin-wide El Niño events 47 . These two dynamics may explain the weakening relationship between ENSO and CS-δ 15 N since the 1980s and imply growing importance of the LLWBC in modulating the tropical ENSO dynamics in the future.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41561-023-01212-4.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Methods
Coral core and age model The coral core (12NAK-K) was drilled from a live Porites at 1 m water depth in August 2012 in the western Solomon Islands between Vella Lavella Island and Ranongga Island (7° 52′ 52.83″ S, 156° 30′ 14.61″ E). The site is remote and not close to major freshwater input. The core was subsequently sliced and stored at the University of Texas at Austin. One slice was sent to National Taiwan University for subsequent analyses. The coral core was passed through the Philips Ingenuity computed tomography (CT) scanner at Taiwan Instrument Research Institute. The coral skeletal density was determined from the CT scans 59 and is used to mark annual growth layers and derive the initial age model with annual resolution (Extended Data Fig. 2a). The average annual extension rate of this Porites coral is 1.3 cm yr −1 . We did not observe any distinct anomalously high-density bands that could be indicative of past bleaching events. Because the density bands are not clear before 1918 ad, we did not sample the materials before 1918 ad. We also used the CT scan to identify the axis of primary growth for subsampling (for example, red squares in Extended Data Fig. 2a). Subsamples were cut with a customized automated sawing machine along the primary growth axis. Each sample is 1 mm parallel to the growth axis, 11 mm perpendicular to the growth axis and 2 mm deep. Given the average extension rate, our sampling resolution is approximately 0.9 month. Each sample is collected into a centrifuge tube and gently ground with a pestle. Approximately 0.5-1 mg material from each sample is collected for carbonate δ 18 O and δ 13 C analyses, while the remaining material is stored for CS-δ 15 N analyses. δ 18 O is a measure of the ratio of oxygen stable isotopes 18 O and 16 O with respect to the the reference standard, Vienna Pee Dee Belemnite (VPDB). δ 13 C is a measure of the ratio of carbon stable isotopes 13 C and 12 C with respect to VPDB. Both are expressed in δ notations. We first derive the CT-based annually resolved age model (CT age), then fine tune the age model by correlating annual cycles of δ 18 O and δ 13 C with satellite-observed SST (8° S, 156 °E, NOAA Extended Reconstructed Sea Surface Temperature (SST) V5, monthly resolution) (Extended Data Fig. 2b). The δ 18 O is generally less negative during austral winters when the surface waters in the Solomon Sea are cold and saline 33 , and δ 13 C generally has higher values during austral summers, probably due to the increase in the δ 13 C of the dissolved inorganic carbon (DIC) in the coral reef as a result of greater productivities in the warm seasons 60 . The relatively large sampling size could result in some time averaging across months, so we focus on discussing interannual, decadal and long-term changes in the CS-δ 15 N. After oxidation, the sample is centrifuged, the clear supernatant is transferred to another precombusted 4 ml borosilicate glass vial, and the pH of the supernatant is adjusted to between 4 to 7 with HCl and NaOH. To determine the N content of the samples, we measure nitrate concentration in the oxidation solutions after autoclaving. Nitrate concentration is analysed by reduction to nitric oxide using vanadium(III) followed by chemiluminescence detection 62 . The N blank is also quantified in this way. Consistent with previous findings, Porites corals have an average N content of 2.3 µmol N per gram of clean aragonite, yielding nitrate concentrations in the oxidation solutions of ~15 µM, whereas the blank concentration ranges between 0.15 and 0.4 µM (less than 3%, typically less than 1%, of the total N per sample).

Sample preparation and analyses for
The δ 15 N of the samples is determined using the denitrifier method in conjunction with gas chromatography and isotope ratio mass spectrometry 63 . The denitrifier method involves the transformation of dissolved nitrate and nitrite into nitrous oxide gas (N 2 O) via a naturally occurring denitrifying bacterial strain that lacks an active form of the enzyme N 2 O reductase. The denitrifier Pseudomonas chlororaphis was used for this work. Normally, 5 nmol sample amounts are added to 1.5 ml of bacterial concentrate after degassing of the bacteria. Along with the samples, the organic standards and replicate analyses of nitrate reference material IAEA-NO3 (δ 15 N = 4.7‰ vs air), USGS-34 (δ 15 N = −1.8‰ vs air) and a bacterial blank are also measured. We use the IAEA-NO3 and USGS-34 standards to monitor the bacterial conversion and the stability of the mass spectrometer, and we use the oxidation standards to correct for the oxidation blank. The denitrifier method typically has a standard deviation (1sd) of less than 0.1‰. An in-house coral standard provides a metric for repeatability both within an analysis batch and across batches, which indicates an analytical precision (1sd) of our protocol of 0.26‰.
Because of the slight changes in the extension direction of this coral, we divided the whole core into 13 parts for subsampling to subsample along the maximal growth axis. We collected ~10 subsamples overlapping in time from the adjacent parts for separate analyses. We use δ 13 C, δ 18 O and CT image to correlate the overlapping samples and to combine these parts back into a complete record. In the same time, these samples can be considered as duplicate measurements on carbonate produced during the same time. The average 1sd (n = 87) of CS-δ 15 N is 0.17‰ (from 0 to 0.64‰). This is probably an overestimation for our analytical error, given the uncertainties in the age when correlating the samples. Correlation analyses. Each dataset is first smoothed with different methods described in the main text and detrended before calculating correlation between any two data series. To estimate the correlation significance level, the effective degree of freedom is calculated from the autocorrelation function 64 .

Ethics statement
The conducted study considers diversity, equity and inclusion.

Data availability
Data are archived at NOAA National Centers for Environmental Information: https://www.ncei.noaa.gov/access/paleo-search/study/37698. The strength of the equatorial trade winds in c is computed from a large-scale tropical SLP gradient (ΔSLP) between the central/east Pacific (160°W-80°W) and the west Pacific/Indian Ocean (80°E-160°E) (box c2 and c1). The index is computed with SLP anomalies from monthly climatology and averaged over grid cells within 5° latitude of the equator, then averaged with 7 years window. The tropical and subtropical wind stress curl anomalies in panel e and g are computed from monthly climatology and averaged over the area in box e (5°S-15°S, 165°E-160°W) and in box g (11.5°S-20°S, 150°E-75°W) respectively, then averaged with 7 years window. During positive PDO phase between 1980 and 2000, the trade winds weaken, and the wind stress curl in the tropical and subtropical bands both show a negative anomaly. The decadal changes in the CS-δ 15 N are significantly correlated with changes in the wind at a period of 11 years, but the CS-δ 15 N showed a delayed response from changes in the equatorial trade winds and tropical wind stress curl. Specifically, the CS-δ 15 N changes are largely in phase with changes in the wind stress curl, but were out of phase after the 1980s. But the decadal changes in the CS-δ 15 N are significantly correlated with and are largely in phase with changes in the wind stress curl in the subtropical South Pacific. In d, f, and h, the thick lines indicate periods when CS-δ 15 N is significantly correlated with the wind changes with 99% confidence level.  77 . The detrended changes in CS-δ 15 N are also significantly correlated with Z 20 (r = 0.67, p = 0.01; 3-year running average). c, Decrease in the CS-δ 15 N, thus increase in the western boundary current transport, appears to coincide with decrease in the ocean heat content in the upper 700 m in the tropical/subtropical Pacific 78 . This supports the importance of the STCs in contributing to the heat recharge-discharge dynamics on interannual/decadal time scales 13 . All data are smoothed by 3-yr running mean.