Asymmetric basin-wide responses of coastal phytoplankton biomass to the Indian Ocean dipole

Abstract

Frequent Indian Ocean Dipole (IOD) events exert substantial ecological pressure on marine ecosystems of the Indian Ocean. However, how coastal ecosystems respond to IOD events remains unknown. Here, combining field observations, remote sensing, and model-derived product, we reveal a basin-scale response of surface and subsurface phytoplankton along the vast eastern Indian Ocean coastline; further, the response is asymmetric between IOD phases, with a 27.0 ± 17.0% increase during positive IOD but only a 13.1 ± 5.2% decrease during negative IOD. Upwelling Kelvin waves (KWs) generated by equatorial easterly wind anomalies during positive IOD raise the thermocline and mixing intensity, increase nutrient availability, and phytoplankton biomass. In contrast, downwelling KWs generated by equatorial westerly wind anomalies during negative IOD further constrain the oligotrophy-limited phytoplankton baseline, resulting in a substantially weaker biological response. Within the KWs framework, underwater topography enhances regional sensitivity to KWs forcing, thereby amplifying the asymmetric phytoplankton biomass responses. Our findings imply that the projected increase in extreme positive IOD events will likely enhance primary production, with possible implications for regional fisheries.

This study shows that Indian Ocean Dipole events drive asymmetric, basin-scale changes in coastal phytoplankton through Kelvin waves, while underwater topography amplifies regional biological responses.

Introduction

Global warming not merely raises average temperature but also intensifies the frequency and severity of extreme climate events^1,2, driving increasing perturbations in marine ecosystems³. The Indian Ocean Dipole (IOD) is a dominant climate mode in the tropical Indian Ocean^4–8, analogous to the El Niño-Southern Oscillation (ENSO) in the Pacific, and manifests as an east-west dipole of sea surface temperature anomalies⁹. The IOD drives coupled ocean-atmosphere interactions and modulates oceanic processes such as upwelling⁹, thereby potentially influencing ecosystem dynamics throughout the water column¹⁰. Global warming is expected to drive a change in IOD frequency and intensity^11,12. The western Indian Ocean is warming faster than the east¹³, contributing to a mean-state change toward a more positive IOD (pIOD)-like condition⁸. As a result, extreme pIOD events are projected to increase substantially in the 21st century^6–8. Understanding how marine ecosystems respond to IOD variability is therefore essential for predicting future ecosystem change.

In oligotrophic and strongly stratified tropical oceans, phytoplankton growth is primarily controlled by nutrient supply via vertical advection and diffusion¹⁴, thereby exerting bottom-up control on marine ecosystems^15,16. In the Indian Ocean, the upwelling region off Sumatra represents the eastern core of the IOD events⁹, where IOD-induced wind anomalies modulate Ekman-driven upwelling intensity and drive pronounced fluctuations in surface phytoplankton biomass^14,17, forming the basis of the so-called biological dipole concept^18–21. Wind-driven upwelling has been identified as the primary driver of phytoplankton variability in the eastern core under both present-day and future climate conditions²². During pIOD events, phytoplankton biomass increases more strongly owing to enhanced upwelling induced by easterly wind anomalies^23,24. However, most previous studies have focused predominantly on the Sumatra upwelling system, providing limited insight into ecosystem responses along the broader eastern Indian Ocean coastline.

The extensive eastern Indian Ocean coastline supports diverse coral reef ecosystems²⁵, which exhibit vulnerability to extreme climate events^26–28, potentially leading to fluctuations in fisheries sustainability^28–31. In contrast to the Sumatra upwelling region, highly stratified coastal areas—particularly in the northeastern basin—exhibit stronger subsurface than surface biological responses. In such regions, ocean-color remote sensing alone is often insufficient to capture the phytoplankton responses to IOD events³². Kelvin waves (KWs), as coastally trapped signals originating from equatorial wind forcing, provide a key physical link between the equatorial Indian Ocean and the eastern boundary^33,34. They may play an important role in regulating nutrient supply and subsurface phytoplankton growth along the coastline during IOD events³⁵. However, whether coastal phytoplankton responses to IOD events are coherent along the eastern Indian Ocean coastline and what mechanisms govern these responses remain poorly understood.

Here, we integrate multi-source datasets including satellite observations, targeted field measurements, and model-derived biogeochemical estimates to reveal that phytoplankton biomass along the eastern Indian Ocean coast exhibits asymmetric responses to the pIOD and negative IOD (nIOD) phases. Based on in situ observations, we further elucidate that KWs-mediated variability in thermocline depth and nutrient supply constitutes the dominant pathway linking IOD forcing to coastal ecosystem responses, while steep local topography acts as an important regulator that amplifies the resulting asymmetry.

Results

IOD-driven basin-wide asymmetry in phytoplankton biomass

Over the past three decades (1993–2022), the Indian Ocean experienced five pIOD and four nIOD events, providing a unique opportunity to examine phytoplankton responses to contrasting IOD phases at the basin scale (Fig. 1a). IOD events exhibit a distinct seasonality, typically developing during boreal summer (June–August), maturing in autumn (September–November), and decaying in winter (December–February)^8,9. Satellite observations reveal coherent coastal Chl-a variability across the IOD life cycle, with positive phases associated with enhanced Chl-a and negative phases with reduced Chl-a (Fig. 1b–d). This variability first emerges along the Sumatra-Java coast during summer, and subsequently expands spatially during autumn and winter to encompass the broader eastern Indian Ocean coastline (Supplementary Fig. 1). Notably, satellite-derived Chl-a exhibits a marked interphase asymmetry, with positive anomalies during pIOD phases consistently exceeding the magnitude of negative anomalies during negative phases (Fig. 1b–d).

Fig. 1. Satellite-observed chlorophyll-a (Chl-a) responses along the eastern Indian Ocean coast during different Indian Ocean Dipole (IOD) phases.

a Dipole mode index (DMI) time series. IOD events are identified when the three-month moving average of the DMI exceeds ± 0.4 °C for at least three consecutive months from June to November. b Spatial distribution of climatological mean Chl-a concentration for 1997 to 2022, averaged from boreal autumn to winter, in the eastern Indian Ocean. c,d Spatial distributions of mean Chl-a anomalies during positive Indian Ocean Dipole (pIOD) events (c) and negative Indian Ocean Dipole (nIOD) events (d) from boreal autumn to winter. The pIOD anomalies are calculated from four events: 1997/1998, 2006/2007, 2018/2019, and 2019/2020, whereas the nIOD anomalies are calculated from three events: 1998/1999, 2005/2006, and 2016/2017. In (a), the occurrence periods of the pIOD and nIOD events are indicated by pink and purple shaded rectangles, respectively. The black solid line represents the DMI time series, while the dashed lines indicate the ± 0.4 °C thresholds. In (b–d), the solid black box delineates the eastern IOD pole region (90°E–110°E, 10°S–0°), and the gray shading indicates land. Dashed pink lines indicate the propagation pathways of equatorial and coastal Kelvin waves. Satellite observations reveal coherent basin-wide Chl-a responses along the eastern Indian Ocean coast across IOD phases, with substantially stronger positive anomalies during pIOD events than negative anomalies during nIOD events.

However, satellite observations cannot adequately capture subsurface biological signals, which are expected to play an important role in stratified coastal regions. To address this limitation, we employ the rigorously validated NEMO-PISCES model-derived biogeochemical product to quantify phytoplankton responses to IOD variability (full model validation is described in Methods). Beyond broadly reproducing the observed phase-dependent pattern of surface Chl-a anomalies (Supplementary Fig. 2), the model framework enables a basin-scale assessment of both surface and vertically integrated biological responses. Model results reveal a pronounced asymmetry in euphotic-zone depth-integrated phytoplankton biomass between opposite IOD phases throughout the IOD life cycle (Fig. 2). Relative to the neutral phase (25.2 ± 1.6 mg m⁻²), biomass increases by 27.0 ± 17.0% in the positive phase (32.0 ± 4.3 mg m⁻²), while it declines by only 13.1 ± 5.2% in the negative phase (21.9 ± 1.3 mg m⁻²) (Fig. 2a, b). Although the magnitude of phytoplankton biomass varies seasonally, this asymmetric response persists throughout the IOD life cycle from summer to winter (Supplementary Figs. 1, 3). Moreover, a distinct subsurface Chl-a maximum (SCM) layer persists along the eastern Indian Ocean coastline from summer through winter (Supplementary Fig. 3), indicating that subsurface phytoplankton variability constitutes an important component of the basin-scale response to IOD forcing.

Fig. 2. Model-derived differences among Indian Ocean dipole (IOD) phases in mean Chlorophyll-a (Chl-a) integral within the euphotic zone along the eastern Indian Ocean coast from boreal summer to winter.

a Mean Chl-a integral within the upper 100 m water column along the coastal Kelvin wave (KW) pathway from summer to winter during different IOD phases, derived from the NEMO-PISCES model output. Shaded areas indicate the interannual standard deviation of the Chl-a integral for each phase. The positive Indian Ocean Dipole (pIOD) events include 1994/1995, 1997/1998, 2006/2007, 2018/2019, and 2019/2020 (n = 5); negative Indian Ocean Dipole (nIOD) events include 1996/1997, 1998/1999, 2005/2006, and 2016/2017 (n = 4); neutral years include five years free of both IOD events and El Niño–Southern Oscillation (ENSO) events (n = 5). b Mean Chl-a integral averaged along the entire eastern Indian Ocean coastal pathway. Error bars represent the standard deviations across multiple years corresponding to each phase. c Seafloor topography and potential density section along the propagation path. Gray shading indicates seafloor topography, while the colored shading and solid black contours represent potential density. d Propagation path of coastal KWs (dashed pink line). For (a) and (c), the horizontal axis indicates the distance from the equator along the propagation path shown in (d), and the x-axis tick labels correspond to the numbered locations along the pathway in (d). The solid horizontal arrows at the bottom indicate the propagation directions of the KWs on the northern and southern sides of the equator, respectively. Model simulations reveal an asymmetric response of Chl-a within the entire euphotic zone along the eastern coast, with a 27.0 ± 17.0% increase during positive phases but only a 13.1 ± 5.2% decrease during negative phases.

Together, the satellite observations and model-derived estimates reveal a coherent basin-scale phytoplankton response within the euphotic zone to IOD variability along the eastern Indian Ocean coastline, characterized by a pronounced asymmetry between positive and negative phases. We next use targeted in situ hydrographic and biogeochemical observations to investigate the physical mechanisms linking IOD forcing to subsurface phytoplankton variability.

Remote regulation through propagation of Kelvin waves

The spatiotemporal distribution of Chl-a anomalies along the eastern Indian Ocean coastline is closely linked to the propagation of coastal KWs (Supplementary Figs. 4, 5). During pIOD (nIOD) events, anomalous easterly (westerly) winds generate upwelling (downwelling) KWs, respectively, which propagate eastward along the equator^33,34. Upon reaching the Sumatra coast, these waves bifurcate into northward and southward coastal branches (Fig. 2d). The northward branch enters the Andaman Sea (AS) through the Great Channel, continues along its eastern coast, and finally enters the Bay of Bengal (BoB) through the Preparis Channel³⁶ (Fig. 2d).

To investigate how KWs regulate subsurface phytoplankton biomass, we conducted in situ observations in the strongly stratified Preparis Channel region (Fig. 3a), located along the downstream section of the northern KWs branch. This region enables us to examine whether ecological modulation persists over long-distance wave propagation (~3500 km from the equatorial source region), while also assessing the influence of complex underwater topography on these responses. Field observations were conducted in winters under neutral IOD (2017/2018) and extreme pIOD (2019/2020) phases within the Preparis Channel (92.75–93.75°E, 14.5–15.5°N, Fig. 3a). Observations reveal a pronounced enhancement of subsurface biological responses during pIOD conditions, consistent with the basin-scale modeling-derived estimates. Specifically, SCM concentrations increased to 2.01 ± 1.49 μg L⁻¹ and shoaled to ~10 m during pIOD, compared to a concentrations of 0.32 μg L⁻¹ at ~40 m under neutral conditions (Fig. 3b). This shoaling SCM elevated euphotic zone depth-integrated Chl-a from 19.51 mg m⁻² in 2018 to 70.72 mg m⁻² in 2020, corresponding to a 3.6-fold enhancement and highlighting the vital control of subsurface processes on upper-ocean productivity. Physical and nutrient conditions also changed substantially during pIOD. The thermocline shoaled by 36 m (68 ± 8 m in 2020 versus 104 m in 2018), while nitrate inventories increased by ~46% (1333.63 mmol m⁻² in 2020 versus 913.23 mmol m⁻² in 2018) (Fig. 3c, d). In addition, the subsurface turbulent dissipation rate ε (4.8 × 10⁻⁷ W kg⁻¹ in 2020 versus 6.0 × 10⁻⁸ W kg⁻¹ in 2018) and diffusion coefficient K_ρ (1.0 × 10⁻³ m² s⁻¹ in 2020 versus 2.9 × 10⁻⁵ m² s⁻¹ in 2018), estimated based on Thorpe methods³⁷, both increase by approximately one order of magnitude (Fig. 3e, f), enhancing upward nutrient diffusion fluxes and creating a favorable environment for phytoplankton blooms (Supplementary Fig. 6).

Fig. 3. Interannual comparison of hydrodynamic, chemical, and biological conditions in the Preparis Channel.

a Distribution of field observation stations and subsurface currents averaged over 50–150 m measured by shipboard acoustic Doppler current profiler (ADCP). b Vertical profiles of Chlorophyll-a (Chl-a) concentration (μg L⁻¹) during the winters of 2017/2018 and 2019/2020. c–f Mean vertical profiles of potential temperature (°C) (c), nitrate concentration (μmol L^–1) (d), turbulent dissipation rate (W kg^–1) (e), and diffusivity coefficient (m² s^–1) (f) around the channel region during the winters of 2017/2018 and 2019/2020. The turbulent dissipation rate and diffusivity coefficient were estimated from high-resolution temperature and salinity profiles using the Thorpe-scale method. Means and standard deviations for the Preparis Channel region were calculated based on observations within the 1°×1° region (92.75–93.75°E, 14.5–15.5°N). Two stations were included for 2017/2018 and five for 2019/2020. Shaded areas indicate the standard deviation across multiple stations. In (c), filled red and blue circles denote the mean thermocline depth during the two cruises, respectively, and are used to delineate the approximate range, indicated by yellow shading in (c–f). PC Preparis Channel, BoB Bay of Bengal, AS Andaman Sea.

These pronounced interannual differences are consistent with remote forcing by IOD-driven KWs. During the 2019/2020 pIOD event, persistent equatorial easterly wind anomalies from summer to winter (Fig. 4d, blue line) generated sustained upwelling KWs that propagated along the equatorial and coastal waveguides, and ultimately reached the Preparis Channel (Fig. 4), coinciding with the observed shoaling of the thermocline and nitracline, enhanced nutrient inventories, and elevated phytoplankton biomass (Fig. 4). In contrast, deeper thermocline and nitracline depths during the 2017/2018 observations were consistent with downwelling KWs, resulting in reduced nutrient availability and lower phytoplankton biomass. Local wind-induced Ekman dynamics represent an alternative mechanism capable of modulating thermocline depth in this region. To disentangle the relative contributions of remote KW propagation and local Ekman forcing, we quantified the cumulative Ekman effects over the two weeks preceding each cruise. Before the 2018 cruise, the net cumulative Ekman effect lifted the Ekman layer by 3.8 m, whereas the net Ekman effect before the 2020 cruise slightly deepened the Ekman layer by 0.08 m (Fig. 4d, red line). This pattern contrasts with the shallower thermocline observed during the pIOD phase, and the magnitude of the Ekman effect is quite weak (Supplementary Fig. 7).

Fig. 4. Temporal evolution of depth-integrated euphotic Chlorophyll-a (Chl-a) and dynamic conditions along the northern branch of the Kelvin waves (KWs) passing through the Preparis Channel, covering the duration of the field observations.

a–c Hovmöller diagrams of model-derived Chl-a integral anomalies (a), sea surface height anomalies (SSHA) (b), and model-derived thermocline depth (indicated by the depth of 22°C isotherm, Z22) anomalies (c) along the northern branch of KWs. d equatorial zonal wind anomaly averaged over 80.0–90.0°E, 2.5°S–2.5°N and Ekman pumping anomaly averaged over the Preparis Channel region (92.75–93.75°E, 14.5–15.5°N) from July 2017 to July 2020. Anomalies were calculated by subtracting the corresponding 1993–2022 climatology. In (a–c), the horizontal axis represents the distance along the KW pathway from 80°E, 0°. In each panel, the black triangles on the right side indicate the timing of the field observations. During the positive phase, anomalous equatorial easterlies excite upwelling equatorial-coastal KWs, which shoal the thermocline and nitracline and support phytoplankton blooms. During the negative phase, westerly wind anomalies trigger downwelling KWs, leading to a decline in phytoplankton biomass.

We further applied lead-lag correlations to separate the relative contributions of remote forcing by KWs and local wind forcing through the Ekman effect along the entire pathway (Supplementary Fig. 8). Correlation coefficients for local wind forcing were generally weak (mean $∣\text{R}∣$ = 0.24 ± 0.16) with irregular lag times. In contrast, KWs-associated correlations remained robust along both the northern and southern branches (mean $∣\text{R}∣$ = 0.74 ± 0.12), although decaying gradually with increasing propagation distance. Corresponding lag times also increased progressively downstream, reaching approximately two weeks across the entire pathway (Supplementary Fig. 8). These results show that KWs act as the primary driver of physical variability along most of the eastern Indian Ocean coastline. Local wind forcing appears to play a secondary role, mainly relevant along the southwestern Sumatra-Java coastal region and near the northern tip of Sumatra, with negligible influence across most parts of the northern KWs branch. Therefore, these findings suggest that IOD-excited KWs constitute the key physical bridge connecting equatorial climate variability to physical and ecological responses along the eastern Indian Ocean coast.

Basin-scale analyses identify KWs as the primary driver of phytoplankton variability. Consistent with this interpretation, a one-dimensional biogeochemical model reproduces the asymmetric phytoplankton response to KW-induced thermocline perturbations under oligotrophic conditions (Supplementary Fig. 9 and Supplementary Note 1), demonstrating that vertically mediated processes are sufficient to generate the observed asymmetry. A model-derived nutrient-budget analysis further shows that horizontal advection accounts for less than 10% of nutrient supply to the euphotic zone, supporting KW-driven vertical transport as the dominant mechanism (Supplementary Fig. 10). The magnitude of the phytoplankton response, however, varies among regions and seasons (Supplementary Fig. 3). These differences are largely controlled by regional stratification and wind forcing. Off southwestern Sumatra during summer and autumn, relatively weak stratification and strong local Ekman pumping act in concert with the KWs forcing. Thus, phytoplankton anomalies here are significantly amplified. In contrast, in Northern Hemisphere regions where Ekman influences are much weaker, stratification intensity becomes the key factor determining whether remote KWs-induced thermocline fluctuations can exert any detectable ecological influence (see details in Supplementary Note 2).

Amplified biological responses driven by local topography

Beyond the basin-scale response to KWs forcing, complex underwater topography can further amplify ecological variability at local scales. This effect is particularly evident in regions characterized by islands and narrow channels, where Chl-a anomalies are substantially larger than those observed in adjacent open-ocean waters (Supplementary Fig. 11). These observations suggest an important role of topographic processes in enhancing biological responses to KWs forcing.

We utilize cross-topographic section field observations obtained in the Preparis Channel during the 2019/2020 pIOD winter (Fig. 5) to examine the hydrodynamic and nutrient dynamic mechanisms behind the substantial interannual biological oscillations around steep underwater topography. We find that the SCM above the channel is more pronounced and notably shallower compared to the adjacent open ocean (Fig. 5c). The maximum SCM (2.85 μg L⁻¹ at S5) far exceeds that observed in the central BoB (0.57 μg L⁻¹ at S1) and AS (0.82 μg L⁻¹ at S10), with the shallowest SCM (11 m at S6), which is shallower than central BoB (67 m at S1) and central AS (42 m at S10). Consequently, the Chl-a integral over the upper 100 m water columns in the channel region (81.33 mg m⁻² at S5) is higher than in the surrounding open regions (21.05 mg m⁻² at S1 and 30.77 mg m⁻² at S10) (Fig. 5c).

Fig. 5. Sectional distributions of hydrological, chemical, and biological parameters across the channel under topographic influence during the 2019/2020 winter.

a–f Sectional distribution of potential temperature (°C) (a), nitrate concentration (μmol L^–1) (b), Chl-a concentration (μg L⁻¹) (c), turbulent dissipation rate (W kg^–1) (d), diffusivity coefficient (m² s^–1) (e), and nitrate diffusive flux (mmol m^–2 d^–1) (f). The turbulent dissipation rate and diffusivity coefficient were estimated from high-resolution temperature and salinity profiles using the Thorpe-scale method. The nitrate diffusive flux was calculated as the product of the vertical nitrate gradient and the diffusivity coefficient. In (a–c), solid black contours indicate isolines of the corresponding variables. In (b), the pink solid line represents the nitracline (depth of 1 μmol L^–1 isoconcentration for nitrate). White dashed lines and the labels above the (a–b) indicate station locations; gray shadings indicate the underwater topography. In (d–f), the color scales are logarithmic. Cross-channel observations reveal that during the positive phase, intensified phytoplankton blooms over topographic features indicate that Kelvin waves effects are topographically amplified.

Shipboard acoustic Doppler current profiler (ADCP) observations identify a strong subsurface current (50–150 m) flowing from the AS to the BoB at 0.48 m s⁻¹ (Fig. 3a). As this subsurface current passes over the steep topography, it generates intense isotherm jumps (Fig. 5a) and vertical mixing downstream (Fig. 5d, e), indicative of typical Lee waves³⁸. The thermocline is overall uplifted downstream, and the turbulent dissipation rate ε and diapycnal diffusivity K_ρ at the 100 m depth are much higher than those in the central BoB and AS (Fig. 5a, d, e and Supplementary Table 2). Subsequently, hydrodynamic processes regulate the nutrient availability within the euphotic zone. Firstly, shallower nitracline depth (13 m above the channel versus 71 m at BoB and 36 m at AS) (Fig. 5b), which correlates with thermocline depth (t-test, R = 0.79, p < 0.01) (Supplementary Fig. 12), effectively enhances the nitrate inventory in the upper 100 m water columns above the channel (1651 mmol m⁻² above the channel versus 509 mmol m⁻² at BoB and 1206 mmol m⁻² at AS). Secondly, diapycnal nitrate flux is crucial for nutrient supply within the euphotic zone. This intensified vertical mixing (Fig. 5d, e), coupled with the nitracline uplift, results in a strong nitrate diffusion flux into the euphotic zone above the channel (Fig. 5f). In contrast, the deeper nitracline and strong barrier layer in the central BoB and AS restrict the upward nutrient replenishment.

The intensified hydrodynamic conditions and enhanced nutrient supply above the channel’s topography create favorable conditions for robust phytoplankton blooms, and these physical and biogeochemical responses are tightly coupled (Supplementary Fig. 12). Within the KWs framework, underwater topography acts as a local amplifier that increases regional sensitivity to external forcing, thereby enhancing the ecological effects of KWs. Apart from the Preparis Channel, KWs encounter several topographic obstacles along their path. To quantify the sensitivity of different regions to external hydrodynamic forcing, we evaluated topographically induced vertical motions as a proxy for local amplification (Supplementary Fig. 13). A linear regression analysis shows that the integrated Chl-a response increases significantly with the magnitude of topographically induced vertical motions across the entire propagation pathway (t-test, R = 0.67, p < 0.01). This result demonstrates that topographic amplification exerts an important control on regional phytoplankton variability and serves as a key local mechanism enhancing ecological responses to KWs forcing (Fig. 6).

Fig. 6. Schematic diagram of biological regulation by Kelvin waves (KWs) during different Indian Ocean Dipole (IOD) phases.

a,b The conceptual diagram illustrates the hydrodynamic, nutrient, and biological responses associated with KWs during positive (a) and negative (b) IOD phases. During positive (negative) IOD phases, anomalous equatorial easterlies (westerlies) wind generates upwelling (downwelling) equatorial-coastal KWs that propagate along the eastern Indian Ocean coastline. As these upwelling (downwelling) KWs travel, they uplift (depress) the thermocline/nitracline and enhance vertical mixing. In (a), during positive IOD phases, a shallower nitracline and greater diffusive nutrient flux relative to the neutral phase enhance nutrient supply to the euphotic zone. In (b), during negative IOD phases, a deeper nitracline and reduced diffusive nutrient flux under strong stratification intensify nutrient limitation, resulting in a weaker biological response under inherently oligotrophic conditions. Steep topography and wind-driven Ekman pumping act as local modulators of KW-driven phytoplankton variability. In regions with steep topography, KWs induce changes in the hydrodynamic environment by affecting the thermocline depth and altering currents, thereby intensifying the already vigorous physical processes and leading to a pronounced phytoplankton response. In regions influenced by Ekman pumping, such as the coastal waters off Sumatra, KWs and Ekman pumping reinforce the phytoplankton response when they act in the same vertical direction, whereas opposing effects result in a net response determined by their relative strengths.

Discussion

Our findings suggest that the asymmetric response of basin-scale phytoplankton biomass along the eastern Indian Ocean coast to IOD events is driven by a remote regulation pathway mediated by KWs (Fig. 6). By modulating thermocline depth and mixing intensity, KWs alter nutrient availability within the euphotic zone and thereby control phytoplankton biomass. During pIOD events, upwelling KWs shoal the thermocline and enhance nutrient supply, leading to substantial increases in phytoplankton biomass. In contrast, downwelling KWs during nIOD events deepen the thermocline and reduce nutrient availability, suppressing phytoplankton growth. However, because phytoplankton biomass is already low under the strongly oligotrophic background conditions of the eastern Indian Ocean, the magnitude of biomass reduction during nIOD events is inherently constrained. As a result, biomass increases during pIOD events substantially exceed biomass decreases during nIOD events, producing the pronounced asymmetry observed across the basin. Moreover, superimposed on this basin-scale response, complex coastal topography further amplifies ecosystem sensitivity to KWs forcing, thereby enhancing the asymmetric phytoplankton response.

Because KWs dynamics also provide an important pathway linking Pacific climate variability to the eastern Indian Ocean, this mechanism offers a useful framework for evaluating the relative influences of IOD and ENSO. Extreme climate events during the past three decades were classified using the IOD and Niño-3.4 indices (Fig. 7 and Supplementary Table 3). Ecologically, pIOD events resemble El Niño events, whereas nIOD events resemble La Niña events. However, independent pIOD events (e.g., 2019/2020 extreme pIOD events) induce larger phytoplankton biomass increases than independent extreme El Niño events (e.g., 2015/2016 extreme El Niño), while independent nIOD events exert a stronger suppressive effect than independent La Niña events (Fig. 7 and Supplementary Table 3). These results suggest that IOD exerts a stronger influence than ENSO on phytoplankton biomass variability in the eastern Indian Ocean. In addition, ENSO and IOD events often occur in phase; pIOD events coinciding with El Niño events, and nIOD events coinciding with La Niña events, produce larger phytoplankton biomass anomalies than either climate mode acting alone.

Fig. 7. Time series of dipole mode index (DMI), Niño 3.4 index, and mean Chlorophyll-a (Chl-a) integral along the eastern Indian Ocean coastline from 1993 to 2021.

a,b DMI (a), and Niño 3.4 index (b) from 1993 to 2021. Red shading indicates the positive Indian Ocean Dipole (pIOD) events or El Niño events, and gray shading indicates the negative Indian Ocean Dipole (nIOD) events or La Niña events. c Mean Chl-a integral along the eastern Indian Ocean coastline from boreal summer to winter. Error bars indicate the spatial standard deviation of Chl-a integral along the eastern Indian Ocean coastline. The years when pIOD and El Niño events occurred simultaneously include 1994/1995, 1997/1998, 2006/2007, and 2018/2019 (red bars filled with red diagonal stripes); the year when pIOD event occurred independently is 2019/2020 (bar filled with only red diagonal stripes); the years when El Niño events occurred independently are 2002/2003, 2004/2005, 2009/2010, 2014/2015, and 2015/2016 (red bars); the years when nIOD and La Niña events occurred simultaneously include 1998/1999 and 2005/2006 (blue bars filled with blue diagonal stripes); the years when nIOD events occurred independently are 1996/1997 and 2016/2017 (bars filled with only blue diagonal stripes); the years when La Niña events occurred independently are 1995/1996, 1999/2000, 2000/2001, 2007/2008, 2008/2009, 2010/2011, 2011/2012, 2017/2018, and 2020/2021 (blue bars). Dashed red, black, and blue lines indicate the mean Chl-a integral for pIOD, neutral, and nIOD events, respectively. Independent Indian Ocean Dipole (IOD) events (e.g., 2019/2020) can trigger stronger Chl-a oscillations along the eastern Indian Ocean coastline than El Niño–Southern Oscillation (ENSO) (e.g., 2015/2016), while IOD events that overlap with ENSO (e.g., 1997/1998) can lead to more extreme Chl-a fluctuations. Chl-a oscillations exhibit asymmetric responses between positive and negative phases, and this asymmetry is further intensified when pIOD coincides with El Niño or nIOD coincides with La Niña.

As global warming increasingly promotes a pIOD-like mean state and raises the frequency of extreme pIOD events^8,12, marine ecosystems along the eastern Indian Ocean coast may experience enhanced primary productivity. Previous studies have reported significant positive correlations between Chl-a concentration and catch rates of yellowfin tuna and sardines in the eastern Indian Ocean^39–42, suggesting that the asymmetric phytoplankton biomass responses documented here could potentially propagate to higher trophic levels. As phytoplankton constitute the foundation of marine food webs and an important food source supporting fish production, enhanced primary production may create favorable conditions for fisheries. However, whether such bottom-up effects ultimately translate into increased fishery yields will depend on food-web structure, trophic transfer efficiency, and future fishery management practices.

Methods

Field observations

As part of two China-Myanmar joint oceanographic surveys, field observation expeditions were conducted in the Preparis Channel region during a neutral winter (January 18–19, 2018) and a pIOD winter (February 16–24, 2020), aboard R/V Xiangyanghong 3 and Xiangyanghong 6, respectively. In 2020, a comprehensive network of fifteen observation stations was strategically deployed, spanning the central BoB, the Preparis Channel region, and the central AS. There are 5 observation stations strategically positioned within a 1°×1° area (92.75–93.75°E, 14.5–15.5°N) in proximity to the Preparis Channel. In 2018, two stations within the same region were occupied, providing a unique opportunity to examine the interannual responses of subsurface phytoplankton biomass under contrasting IOD conditions (Fig. 3a).

In situ observations are conducted using a Seabird 25 plus Conductivity-Temperature-Depth (CTD) to acquire temperature, salinity, and depth. At the same time, the CTD is equipped with a deep-sea submersible ultraviolet nitrate analyzer (Deep SUNA, Satlantic Inc.) and a fluorescence sensor to obtain high-resolution profiles of NO₃⁻ and Chl-a concentration. The observation frequency for the CTD and fluorescence sensors is 24 Hz, and the frequency for the Deep SUNA is 1 Hz. The nitrate and Chl-a observations collected during the 2018 and 2020 cruises are reported here for the first time. Horizontal velocity profiles were measured using a shipboard 38-kHz ADCP with a vertical resolution of 24 m. The subsurface current velocity is calculated as the average velocity between the depths of 50 and 150 m. During the observations, the deployment speed of the instruments is controlled at approximately 0.5 m s⁻¹.

Calibration of sensor data

The fluorescence data and nitrate data from sensors are both further corrected using laboratory analysis data. Nitrate concentrations in seawater samples are measured using an AA3 nutrient analyzer, while Chl-a concentrations are determined using fluorometric methods^43,44. Calibration curves are then developed based on sensor data from the same depths and laboratory analysis data, as depicted in the calibration curve shown in Supplementary Fig. 14.

Estimation of turbulent intensity and nitrate diffusive flux

Turbulent mixing is the primary energy source driving density inversions in stratified water columns⁴⁵, and forms the basis of the Thorpe scale method. Using high-resolution temperature and salinity profiles, density inversions were identified and subsequently used to estimate turbulent kinetic energy (ε) and diapycnal diffusivity (K_ρ) following established parameterizations⁴⁶. Before analysis, CTD data were subjected to standard preprocessing, including thermal-lag correction, salinity spiking removal, outlier filtering, and vertical averaging to 0.1 m³⁷. The Thorpe displacement (d’) was calculated as the vertical distance between the observed depth of a water parcel and its reordered depth in a statically stable density profile. The Thorpe scale L_T was then defined as Eq. (1):

$$L_T=\sqrt{\overline{{d^{\prime }}^2}}.$$ 1

The turbulent kinetic energy dissipation rate ε was estimated as Eq. (2):

$$\mathrm{\epsilon }=a\cdot L_T^2\cdot N^3,$$ 2

where a = 0.8 (empirical constant) and N is the buoyancy frequency calculated by Eq. (3):

$$N^2=\frac{g}{\rho }\times \frac{\partial \rho }{\partial z},$$ 3

with g as the gravity and ρ as reference density. Finally, diapycnal diffusivity K_ρ was subsequently calculated as Eq. (4)

$$K_{\rho }=\Gamma \cdot \frac{\epsilon }{N^2}$$ 4

where Γ = 0.2 is the mixing efficiency. To evaluate the diapycnal nitrate flux F_diff, Fick’s law is applied using the derived diapycnal diffusivity (K_ρ) and vertical nitrate concentration gradient as Eq. (5):

$$F_{diff}=-K_{\rho }\times \frac{\partial {\mathrm{N}O}_3^{-}}{\partial z}$$ 5

Dipole mode index and Niño 3.4 index

The classification of IOD and ENSO events, as well as their neutral phases, was based on the dipole mode index (DMI) and Niño 3.4 index, respectively, following established criteria from previous studies^9,47. The time series of DMI and Niño 3.4 index are obtained from https://psl.noaa.gov/, which are calculated from the HadISST1.1 sea surface temperature (SST) dataset⁴⁸. Positive (negative) values of the DMI and Niño 3.4 index indicate pIOD (nIOD) and El Niño (La Niña) conditions, respectively. The DMI is defined as the anomalous SST gradient between the western equatorial Indian Ocean (50°E–70°E, 10°S–10°N) and eastern equatorial Indian Ocean (90°E–110°E, 10°S–0°)⁹. The Niño 3.4 index is calculated by the area-averaged SST anomaly from 170°W–120°W, 5°S–5°N⁴⁷. IOD events are identified when the three-month moving average of the DMI exceeds ± 0.4 °C for at least three consecutive months from June to November, following the operational criterion adopted by the Japan Meteorological Agency (JMA). ENSO events are identified when the three-month average of the Niño-3.4 index exceeds ± 0.5 °C for five consecutive months, following the operational definition used by NOAA. Periods during which the DMI remained within ± 0.4 °C, and the Niño 3.4 index remained within ± 0.5 °C were classified as neutral IOD and neutral ENSO conditions, respectively.

Propagation pathways and large-scale environmental datasets

Based on the distribution of the 400 m isobath, we establish the propagation pathways of coastal Kelvin waves along the eastern Indian Ocean (Fig. 1b–d). To characterize the propagation of KWs and their associated physical and ecological responses over longer time scales and broader spatial scales, we investigate the Sea surface height anomaly (SSHA), wind, temperature profile, and Chl-a datasets spanning 1993–2022.

To identify KW signals, we use a daily sea surface height dataset, amalgamated from multiple satellite sources by the Copernicus Marine Environment Monitoring Service (CMEMS), a daily gridded dataset at 0.25° resolution with data availability from 1993 to 2022⁴⁹. SSHA is calculated relative to the 30-year climatology. Along the KWs pathway, SSHA is derived by averaging values within a 0.5°×0.5° box centered on each pathway point.

Wind forcing is characterized using daily ECMWF Reanalysis v5 (ERA5) reanalysis data at a 0.25° spatial resolution⁵⁰. Ekman pumping velocity (${\mathit{\omega }}_{\mathbf{e}}$, positive for upwelling) is calculated as Eq. (6):

$${\mathit{\omega }}_{\mathbf{e}}=\frac{1}{{\mathrm{\rho }}_{\text{w}}}\times \left(\nabla \times \frac{\mathit{\tau }}{f}\right),$$ 6

where τ is the vector of wind stress, ${\rho }_w$ is the seawater density (1024 kg m⁻³), and f is the Coriolis parameter.

Phytoplankton variability is characterized using the CMEMS multi-satellite Chl-a product. Daily and monthly Chl-a fields are used to quantify the spatial distribution and temporal variability of surface phytoplankton biomass along the eastern Indian Ocean coastline.

NEMO-PISCES model products and derived variables

Model-derived temperature and bottom-current fields are obtained from the GLORYS ocean reanalysis 12v1 ocean reanalysis product distributed through CEMES⁵¹, which is based on the NEMO ocean model⁵². The product has a horizontal resolution of 0.083° × 0.083° (~9 km), 50 vertical levels, and a daily output frequency. Model-derived Chl-a fields are obtained from the CEMES NEMO-PISCES physical-biogeochemical coupled product⁵³, which is based on the NEMO ocean model coupled to the PISCES biogeochemical model, and provides daily outputs at a horizontal resolution of 0.25° × 0.25° and 75 vertical levels.

According to the CEMES product documentation, the underlying model system is forced at the surface by ERA5 atmospheric reanalysis data⁵⁰. Tracer advection is represented using a total-variance-diminishing (TVD) scheme^54,55, whereas vertical mixing is parameterized using a two-equation k–ε turbulence closure scheme⁵⁶. The products have been extensively validated against satellite and in situ observations in previous studies⁵⁷.

Thermocline depth is represented by the depth of the 22°C isotherm, which closely approximates the nitracline depth in the eastern Indian Ocean. Phytoplankton variability within the water column is quantified using the depth-integrated Chl-a concentration within the upper 100 m, representing the euphotic zone.

To evaluate the potential amplification effect of local topography, topographically induced vertical motions are estimated as Eq. (7)

$${\mathbf{w}}_{\mathbf{bottom}}=∣{\mathbf{u}}_{\mathbf{bottom}}\cdot \nabla \mathrm{\cdot }\mathrm{▽}H∣,$$ 7

where u_bottom is the near-bottom current velocity and H is bottom depth⁵⁸.

Validation of NEMO-PISCES model output

While the NEMO-PISCES physical-biogeochemical coupled model has been validated against satellite and Argo float observations, we conducted additional validation of model outputs using 1) field observations from the Preparis Channel region (Supplementary Fig. 15), 2) sectional and interannual comparative observations from the eastern equatorial Indian Ocean (Supplementary Fig. 16), and 3) publicly available biogeochemical Argo profile measurements across the eastern Indian Ocean to ensure model reliability in this region (Supplementary Figs. 17–19).

First, compared to observations from the Preparis Channel and eastern equatorial Indian Ocean, the model effectively captures the interannual variability and spatial patterns of depth-integrated Chl-a within the euphotic zone, showing a strong linear correlation (Preparis Channel: t-test, R = 0.72, p < 0.01, RMSE = 4.80 mg m⁻²; eastern equatorial Indian Ocean: t-test, R = 0.84, p < 0.01, RMSE = 2.16 mg m⁻²), although it somewhat underestimates the magnitude of differences between positive and neutral IOD phases (Supplementary Figs. 15, 16). Second, after gridding the available BGC-argo profiles in the eastern Indian Ocean at 2° × 2° resolution and computing monthly averages for comparison with model outputs, significant correlations demonstrate the model’s capability in simulating water column-integrated Chl-a in this region (t-test, R = 0.29, p < 0.01, RMSE = 8.99 mg m⁻², Supplementary Fig. 17). Finally, comparisons between five long-term BGC-Argo time series and corresponding model simulations reveal that the model successfully reproduces subsurface phytoplankton biomass responses, in terms of both the depth and magnitude of the SCM (Supplementary Figs. 18, 19). Significant correlations (t-test, p < 0.01) are obtained for both SCM concentration and the depth of SCM across the majority of time series, demonstrating the model’s capability to capture the temporal variability and vertical structure of subsurface phytoplankton biomass in the eastern Indian Ocean. In conclusion, the rigorously validated physical-biogeochemical coupled model reliably captures phytoplankton dynamics in the eastern Indian Ocean under IOD influences.

Regional topographic data

The topographic data for the eastern Indian Ocean is obtained from global bathymetry data SRTM15 + , with a spatial resolution of 15 arc minutes. The bathymetry is produced using a combination of shipboard soundings and depths predicted using satellite altimetry⁵⁹.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

H.L.L., J.F.C. and J.J.Z. contributed to the conceptualization of the manuscript. X.Y.W. processed instrument data, analyzed nutrient concentrations in water samples, and drafted the manuscript. H.L.L. and J.J.Z. conducted on-site observations and guided manuscript preparation and revision. With the assistance of C.E.M., X.Y.W. processed the model output data of NEMO-PISCES. S.J.K and W.J.C. made substantial contributions to revising the manuscript. J.L.X., R.B.D., D.R.P., A.L., and C.E.M. provided substantial input on figure preparation and discussions during the manuscript writing process. Y.Q. and X.W.Z. provided a portion of the field observation data used for comparative analysis. F.Z., L.W., L.S., Y.L.Z., W.W.T., S.M.L., and C.A.T. made important contributions to field data acquisition and sample processing. All authors contributed to the discussion of the results and provided substantial input to the manuscript.

Peer review

Peer review information

Nature Communications thanks Deyong Sun, Kyung-Min Noh and other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was supported by the Natural Science Foundation of China (grant nos. 42330412 to J.F.C. and 42476047 to J.J.Z.), the Special Fund for Basic Scientific Research of the Second Institute of Oceanography, MNR (grant no. SZ2403 to H.L.L.), and Global Change and Air-Sea Interaction II Program (grant nos. GASI-01-EIND-STwin and GASI-04-WLHY-03).

Data availability

The field observational data and Source Data underlying the figures generated in this study have been deposited in the Zenodo repository⁶⁰ under https://doi.org/10.5281/zenodo.22177166. This study also uses E.U. Copernicus Marine Service Information, including Sea Level Anomalies⁶¹ (SLA; https://doi.org/10.48670/moi-00148), satellite-derived Chl-a⁶² (https://doi.org/10.48670/moi-00281), GLORYS12 ocean reanalysis temperature and current fields (https://doi.org/10.48670/moi-00021)⁶³, and NEMO-PISCES biogeochemical model outputs⁶⁴ (https://doi.org/10.48670/moi-00019). Wind fields were obtained from the ERA5 reanalysis provided by the Copernicus Climate Change Service⁵⁰ (https://cds.climate.copernicus.eu/datasets). Biogeochemical Argo (BGC-Argo) profile data were obtained from the Global Data Assembly Center (https://biogeochemical-argo.org). The DMI and Niño-3.4 index were obtained from the NOAA Physical Sciences Laboratory (https://psl.noaa.gov).

Code availability

The custom computer codes that were used in this study to generate figures and more are available from the Zenodo repository under the accession link⁶⁰: https://doi.org/10.5281/zenodo.22177166.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77809-1.