Abstract
Typhoon-generated high waves are a recurrent coastal hazard along Guangdong, southern China, however, the extent to which the El Niño–Southern Oscillation (ENSO) can inform annual-scale coastal exposure remains uncertain. This study integrates hourly ERA5 wave and wind data (1979–2025) from six coastal-shelf grid points, tropical cyclone (TC) best track records from the China Meteorological Administration, and NOAA Oceanic Niño Index (ONI) data. A site-specific 95th-percentile event catalogue comprising 1,249 site-events and a regional TC-track sample of 542 TCs entering within 700 km of the site network are analysed at annual and event scales. Analyses reveal that ENSO conditions significantly influence annual patterns of typhoon-induced wave exposure. At the annual scale, the TC-season ONI-window composite is negatively associated with site-event frequency (r = −0.29, p = 0.048) and cumulative wave exceedance (r = −0.41, p = 0.004). Mean annual cumulative exceedance is 937.6 m·h in La Niña years and 499.3 m·h in El Niño years, although substantial interannual variability remains. Track diagnostics identify a robust westward shift in mean TC genesis longitude during La Niña, whereas the phase difference in mean nearest distance is small and uncertain. At the event-scale, regression analyses of 1,242 site-events (clustered into 318 parent-TC; R² = 0.292) identified nearest TC distance and TC wind velocity as the strongest predictors of peak significant wave height (Hs). In contrast, the standardized ONI coefficient is weak and not conventionally significant (−0.063, p = 0.066), highlighting the limited direct influence of ENSO on individual event severity. These findings indicate the scale-dependent impact of ENSO: ENSO variability is more clearly associated with annual TC-wave occurrence and cumulative coastal exposure than with the peak Hs of individual events. ENSO information can offer valuable climatic context to support seasonal coastal hazard preparedness. However, it should not be regarded as a deterministic forecasting tool or as a replacement for event-specific observations, real-time forecasts, or detailed engineering design analyses.
1 Introduction
The Guangdong coast, encompassing the Pearl River Estuary, eastern Guangdong embayments, western Guangdong, and the Leizhou Peninsula (Figure 1), is a heavily utilized coastal region featuring major ports, densely populated cities, offshore infrastructure, and extensive aquaculture areas. As part of the economically advanced Pearl River Delta, this region is highly vulnerable to tropical cyclone (TC)-induced coastal hazards, including hazardous waves, storm surges, rainfall, and strong winds. The complex shoreline and high concentration of coastal assets exacerbate these vulnerabilities (Li et al., 2024). For long-term risk assessment and engineering design, the climatological mean wave height alone is insufficient. Instead, it is critical to quantify the interannual variability, duration, and magnitude of typhoon-induced high-wave events to better understand and manage the risks posed by such coastal hazards.
Figure 1
ENSO is the dominant mode of interannual tropical-Pacific climate variability and is well known to modulate western North Pacific TC activity. Earlier work showed that ENSO-related sea-surface-temperature anomalies shift preferred TC genesis regions, alter steering environments and recurvature probabilities, and change the likelihood of storms entering the South China Sea (Bjerknes, 1969; Trenberth, 1997; Wang and Chan, 2002; Chia and Ropelewski, 2002; Camargo et al., 2007). While the present study does not seek to re-establish those basin-scale teleconnections. Instead, it seeks to explore how these changes manifest specifically in the context of a site-specific coastal wave-exposure catalogue for Guangdong.
For the Guangdong coast, the relevant link is not simply basin-wide TC frequency. Local wave response also depends on whether a storm approaches the northern South China Sea, how long it remains regionally influential, its wind intensity, its trajectory relative to the coast, and local exposure of the selected coastal-shelf grid point. An intense but distant TC may produce a smaller local response than a weaker storm with a more favourable approach geometry. This motivates a framework that separates the climate background, the regional TC population and the site-event wave response.
For this region, the connection between ENSO and wave hazards is not solely determined by basin-wide TC frequency. Local wave responses depend on whether a storm reaches the northern South China Sea, its duration of regional influence, wind intensity, trajectory relative to the coast, and the localized exposure of specific coastal-shelf grid points. For instance, an intense TC originating far offshore may generate smaller local wave responses than a relatively weaker storm that follows a more direct or favourable trajectory toward the coast. These complexities necessitate a comprehensive analytical framework that separates the roles of climatic background, regional storm characteristics, and local site-specific wave responses.
To represent ENSO variability, we employ the Oceanic Niño Index (ONI), derived from three-month running averages of Niño-3.4 sea surface temperature anomalies (NOAA Climate Prediction Centre, n.d.; Barnston et al., 2019). Given the seasonal nature of TC activity, three standard ONI windows—June–August (JJA), August–October (ASO), and October–December (OND)—are analysed along with a composite index spanning the entire TC season. A preceding-winter ONI metric is also included to enable lagged analyses and assess the persistence of ENSO influences. The overlapping-window methodology and its sensitivity checks have been fully detailed in Section 2.3 to avoid misinterpretation of the composite as a simple seasonal mean.
The study leverages ERA5 reanalysis data, which provides hourly atmospheric and ocean-wave fields and has been widely utilized in wave-climate research (Hersbach et al., 2020). While ERA5 demonstrates high skill in capturing general wave climatologies for the South China Sea, certain limitations, such as the smoothing of intense-TC wind fields and nearshore gradients, are well-documented (Zhai et al., 2023). Thus, ERA5 is employed primarily for interannual and comparative analyses rather than site-specific design-wave determinations. Combined with the homogeneous CMA best-track dataset (Ying et al., 2014; Lu et al., 2021) for TC track positions and intensities, this study ensures consistency across the multi-decadal wave record.
The northern South China Sea experiences both monsoonal wave regimes and episodic typhoon-generated waves, with the extreme events typically dominated by typhoons. While global studies have documented interannual variability in wind and wave conditions (Reguero et al., 2015; Young and Ribal, 2019), recent studies in the Indo-Pacific highlight that extreme wave responses can co-vary with large-scale climate modes and TC activity (Kumar et al., 2024). In Guangdong, however, such signals are superimposed on stark regional differences in coastal orientation and exposure between the east and west coasts, requiring localized analysis.
Recent studies have also improved representation of future and event-scale wave conditions. For instance, CMIP6-derived wave datasets allow for comprehensive assessments of future wind-wave and swell change (Jiang et al., 2023; Meucci et al., 2024), and projected changes in TC-generated extreme waves depend on both storm intensity and propagation pathways (Grossmann-Matheson et al., 2024). In the South China Sea, typhoon-wave hindcasts are sensitive to wind-field forcing, wave–current interaction and track–coast geometry (Jia et al., 2023; Wen et al., 2024; Bagnasco et al., 2025). These developments demonstrate the growing need for a regional event catalogue that connects climate background, storm pathway and local coastal exposure.
In this context, the primary contribution of this study is therefore the construction and analysis of a Guangdong TC–wave event catalogue with an explicit two-scale sample hierarchy. The regional TC-track sample focuses on storms entering within 700 km of the six-site network, whereas the site-event sample records individual high-wave responses at those grid points. This distinction avoids treating several coastal responses to one storm as independent TCs and allows climate-scale exposure metrics to be separated from event-scale peak-wave controls.
Three questions are addressed: (1) How do annual site-event occurrence and cumulative wave exceedance vary with ONI? (2) What differences exist in regional TC-track attributes under El Niño, Neutral, and La Niña conditions? (3) After accounting for the effects of nearest TC distance, TC wind intensity and persistent site effects are accounted for, does ONI retain an independent association with event-maximum Hs? The analysis is intentionally associative rather than causal approach, as large-scale circulation fields required for dynamical attribution are not composited here.
Finally, it is important to distinguish among wave metrics used in this study, as they capture complementary dimensions of wave exposure. Mean and high-percentile Hs describe regional operating environments, while event-maximum Hs reflects peak severity. Cumulative exceedance, on the other hand, integrates the duration and persistence of waves exceeding local high thresholds, offering insight into sustained exposure. For instance, a year characterized by a moderate annual maximum Hs can still experience significant cumulative exposure if multiple TCs produce prolonged wave events. This distinction underscores the interpretation throughout this study.
2 Materials and methods
2.1 Tropical cyclone and wave data
TC track and intensity data were obtained from the CMA best-track dataset for the western North Pacific and the South China Sea (Ying et al., 2014; Lu et al., 2021). Each record provides the storm’s location and intensity at regular synoptic intervals. For the track-exposure analysis, a regional TC was defined as a CMA storm with a centre that entered within 700 km of at least one of the six selected ERA5 grid points during 1979–2025. Using this criterion, 542 regional TCs were identified. A subset of these TCs produced high-wave events at one or more sites, forming the parent-TC cluster sample used for event-scale analyses.
Wave and wind data were derived from the ERA5 reanalysis dataset, which provides hourly single-level atmospheric and oceanic fields. The ERA5 wave grid has a spatial resolution of approximately 0.5° × 0.5°, while atmospheric data are available at a finer resolution of 0.25°. The requested point coordinates were GD1 (117.0°E, 23.0°N), GD2 (116.5°E, 22.5°N), GD3 (114.5°E, 22.0°N), GD4 (113.5°E, 21.5°N), GD5 (112.0°E, 21.0°N) and GD6 (111.0°E, 20.5°N). The six points were selected to sample an east-to-west coastal-shelf exposure gradient while remaining on simple ERA5 grid coordinates and outside enclosed harbour basins. Table 1 and Figure 1 provide the exact locations and regional settings.
Table 1
| Site | Lon. | Lat. | Nearshore setting | Mean Hs | P95 Hs | P99 Hs | Max Hs | Mean U10 | Mean Tm |
|---|---|---|---|---|---|---|---|---|---|
| GD1 | 117.00 | 23.00 | outer eastern Guangdong | 1.30 | 2.62 | 3.23 | 8.64 | 7.18 | 5.57 |
| GD2 | 116.50 | 22.50 | eastern Guangdong nearshore | 1.46 | 2.86 | 3.57 | 10.77 | 7.45 | 5.89 |
| GD3 | 114.50 | 22.00 | eastern Pearl River Estuary margin | 1.30 | 2.37 | 3.08 | 10.52 | 6.53 | 6.07 |
| GD4 | 113.50 | 21.50 | western Pearl River Estuary margin | 1.33 | 2.43 | 3.19 | 10.47 | 6.51 | 6.09 |
| GD5 | 112.00 | 21.00 | western Guangdong nearshore | 1.25 | 2.32 | 3.00 | 8.74 | 6.63 | 5.84 |
| GD6 | 111.00 | 20.50 | east of the Leizhou Peninsula | 1.14 | 2.23 | 2.86 | 7.35 | 6.41 | 5.75 |
Exact coordinates and basic wind-wave statistics of the six ERA5 grid points.
Hs is significant wave height, U10 is 10-m wind speed and Tm is mean wave period.
Key variables retained in the archived point files are significant wave height (swh; Hs), mean wave period (mwp; Tm), mean wave direction (mwd), and 10-m zonal and meridional wind components (U10, V10), from which U10 was calculated. The data set spans the period from January 1, 1979, to December 31, 2025 before event detection. While wave period is an important parameter for understanding engineering loads, it was not included as a primary response variable in the present ENSO regression because the event catalogue was defined by high-Hs exceedance. Nevertheless, wave period, direction, spectral characteristics, and local bathymetry should be considered for engineering applications.
The six chosen points represent coastal-shelf conditions rather than representations of specific harbours, bays or structures. GD1–GD2 reflect eastern Guangdong’s coastal exposure, GD3–GD4 correspond to the two sides of the Pearl River Estuary, GD5 characterizes western Guangdong, and GD6 represents the coastal shelf sector east of the Leizhou Peninsula. The configuration ensures variability in factors such as offshore fetch, coastal orientation, and sheltering effects, while encompassing the same regional TC climate.
Geographic context for the selected locations is provided using the ETOPO1 global relief model (Amante and Eakins, 2009), depicted in Figure 1. This bathymetric data is visualized solely for contextual purposes, as the shaded-relief image used is not a numerical depth grid. Bathymetry was not employed as a predictor variable in this study. It is important to note that ERA5 values were not adjusted for site-specific conditions, such as harbour environments or infrastructure foundations. The reported wave and wind metrics reflect reanalysis-based regional dynamics and do not represent engineering design conditions or site-specific predictions. For detailed design assessments, higher-resolution wave models, local observations, and site-specific standards (e.g., Ministry of Transport of the People's Republic of China, 2013) remain essential.
2.2 Identification of Guangdong-impacting typhoons and high-wave events
Three linked sample levels are distinguished. First, the regional TC-track sample contains 542 CMA TCs whose centres were within 700 km of any grid point. This sample is used to investigate TC genesis locations, closest-approach positions, bearing and residence-time diagnostics. Second, the high-wave catalogue contains 1,249 site-events; one storm can generate events at several sites. Third, event-producing parent TCs define the dependence clusters for the regression. Of the 1,249 site-events, seven have zero reported minimum centre-to-site distance and are excluded only from the logarithmic-distance regression, leaving N=1242 events nested within 318 parent-TC clusters (62 El Niño, 175 Neutral and 81 La Niña parent TCs). These levels are kept separate throughout the analysis.
High-wave events were identified following a structured methodology: (1) The site-specific 95th percentile of hourly Hs over 1979–2025 was calculated. (2) Continuous exceedance segments with Hs > Hs,95 were identified. (3) Adjacent exceedance segments separated by a gap of 12 hours or less were merged, allowing a multi-peak storm episode to remain one event. (4) Merged events shorter than 6 hours were removed. (5) Event duration and maximum Hs were recorded, and cumulative wave exceedance was calculated as Equation 1.
Thus C has units of m·h. The term “cumulative exceedance” is used hereafter instead of “intensity” to avoid implying wave energy. (6) CMA track records were searched from 24 h before event onset to 24 h after event end. (7) If several TCs were present, the parent storm with the smallest centre-to-site distance was assigned. The workflow is summarized in Supplementary Figure 1, and the full event catalogue is supplied.
The event catalogue was generated before annual ENSO grouping. Cumulative exceedance quantifies the time-integrated amount of Hs above the local threshold and is intended as a relative exposure metric. It is not wave energy and is not a substitute for return-period or design-wave analysis.
A site-event was classified as TC-related when the assigned CMA storm centre came within 700 km of the site during the event-centred matching window. This 700-km radius is an operational association criterion, not a physical wind-radius boundary. It was selected to retain broad northern-South-China-Sea wave forcing while excluding most remote cyclones. The 500-km and 900-km alternatives are therefore treated explicitly as sensitivity tests rather than as competing physical definitions.
The threshold-event method is distinct from formal extreme-value analysis. Its purpose is to compare occurrence, duration, peak Hs and cumulative exceedance under different climate backgrounds, not to estimate 50- or 100-year design conditions. Classical extreme-value and wave-design methods remain necessary for structural design (Coles, 2001; Goda, 2010; Longuet-Higgins, 1953; Holthuijsen, 2007).
Sensitivity tests are reported quantitatively in Supplementary Table 1. With the baseline 95th-percentile threshold, 12-h merging window and 700-km radius, the catalogue contains 1,249 site-events, with r=−0.291 for ONI versus annual event count and r=−0.408 for ONI versus annual cumulative exceedance. Changing the radius to 500 and 900 km produces 1,018 and 1,438 events, respectively; the cumulative-exceedance correlations remain negative (r=−0.336 and −0.460). P90 and P98 thresholds produce 1,542 and 910 events, while 6-, 12- and 18-h merging windows produce 1,286, 1,249 and 1,242 events. Across these reasonable alternatives, the ONI–cumulative-exceedance correlation remains negative; absolute event counts and m·h totals, as expected, depend on the event definition.
2.3 ENSO indices and annual metrics
ENSO is represented by standard three-month ONI windows (NOAA Climate Prediction Centre, n.d.; Barnston et al., 2019). The primary grouping variable used in the submitted analysis is the “TC-season ONI-window composite”, defined as the arithmetic mean of JJA, ASO and OND. Because these are overlapping three-month windows, this composite is not an equal-month June–December average: August and October receive weight 2/9, while June, July, September, November and December each receive weight 1/9. We retain the original composite so the analysis remains directly comparable with the submitted results, but we now state this weighting explicitly. More importantly, JJA, ASO and OND are also analysed separately (Figure 2), so the interpretation does not depend on the composite alone. Years are grouped as El Niño when the composite is ≥ +0.5 °C, La Niña when ≤ −0.5 °C, and Neutral otherwise; the preceding-winter ONI is analysed separately as a lagged background index.
Figure 2
A subset of strong ENSO years is retained for descriptive sensitivity testing, including five strong El Niño years and four strong La Niña years between 1979 and 2025. It is worth noting that individual extreme seasons may disproportionately influence group means in such subsets. Eastern-Pacific and Central-Pacific ENSO types are not separated because subdivision of the regional TC and annual samples would further reduce precision. ENSO flavour is discussed as a plausible source of heterogeneity rather than tested as an additional categorical factor.
Annual response metrics analysed in this study included the number of TC-related site-events, cumulative wave exceedance (m·h yr−1), annual maximum Hs, mean event-maximum Hs, mean event-to-TC nearest distance and mean event duration. Regional-TC diagnostics include storm count, genesis longitude, closest-approach position, lifetime maximum wind, mean bearing and residence time within 700 km of the six-site network.
Pearson correlations are used as exploratory annual diagnostics. All correlations are reported with N=47 and 95% confidence intervals. To account for potential temporal dependence, lag-1 autocorrelation is calculated for the ONI composite and each annual response series. Effective sample size (Neff) was adjusted using the formula Neff = N(1 − r1,x r1,y)/(1 + r1,x r1,y), where r1,x and r1,y denote lag-1 autocorrelation values. The key ONI correlations remain similar after this adjustment (Supplementary Table 2). Because multiple ONI windows and response metrics are inspected, individual p-values are interpreted cautiously rather than as a stand-alone confirmatory multiple-testing framework.
2.4 Track diagnostics and regression analysis
Regional track diagnostics are evaluated on the 542-TC sample. For each TC we retain genesis longitude and latitude, the track position at closest approach to the six-site network, minimum great-circle distance, lifetime maximum wind, mean bearing from genesis toward the closest-approach segment, and total residence time within 700 km of any site. The 700-km residence metric is an operational measure of regional exposure duration, not a storm-size estimate. Continuous metrics are preferred to subjective track classes, and phase means are accompanied by uncertainty where appropriate.
In this study, “track geometry” refers to the observed spatial attributes of the regional TC trajectory, including genesis position, closest-approach position, bearing and distance. “Pathway exposure” refers to the frequency and persistence of regional storms affecting the study locations based on site-events and residence time. These metrics are descriptive and do not, by themselves, define the dynamical mechanisms underlying ENSO effects.
The event-scale regression uses standardized event-maximum Hs as the response. For a given event i, the regression takes the following form:
where diis nearest TC distance in km, Vi is the maximum assigned-TC wind within the event matching window, ONIy is the TC-season ONI-window composite for that year, and GDs is the site reference. All terms are standardized as z(x) = [x−mean(x)]/SD(x), with distance values undergoing natural-log transformation before standardization. Standard errors are clustered by parent TC because several site-events can arise from the same storm. The fitted sample contains 1,242 events in 318 parent-TC clusters. Table 2 reports both standardized coefficients for relative comparison and unstandardized coefficients for physical interpretation.
Table 2
| Term | Std. coef [95% CI] | Unstd. coef [95% CI] | Cluster p | N | TC clusters | R² |
|---|---|---|---|---|---|---|
| ln(nearest distance) | -0.331 [-0.394, -0.268] | -0.496 [-0.591, -0.401] | <0.001 | 1242 | 318 | 0.292 |
| Matched TC wind | 0.366 [0.280, 0.452] | 0.039 [0.030, 0.048] | <0.001 | 1242 | 318 | 0.292 |
| TC-season ONI | -0.063 [-0.128, 0.003] | -0.113 [-0.231, 0.005] | 0.066 | 1242 | 318 | 0.292 |
| GD2 | 0.300 [0.210, 0.391] | 0.380 [0.266, 0.495] | <0.001 | 1242 | 318 | 0.292 |
| GD3 | -0.040 [-0.206, 0.126] | -0.051 [-0.261, 0.159] | 0.634 | 1242 | 318 | 0.292 |
| GD4 | 0.073 [-0.106, 0.251] | 0.092 [-0.134, 0.319] | 0.422 | 1242 | 318 | 0.292 |
| GD5 | -0.034 [-0.206, 0.137] | -0.044 [-0.261, 0.174] | 0.693 | 1242 | 318 | 0.292 |
| GD6 | -0.207 [-0.371, -0.044] | -0.263 [-0.470, -0.055] | 0.013 | 1242 | 318 | 0.292 |
Cluster-robust event-scale regression for peak Hs.
Standardized and unstandardized coefficients are reported with 95% confidence intervals. N = 1,242 site-events, 318 parent-TC clusters, R² = 0.292; GD1 is the site reference..
The regression is deliberately parsimonious and should not be interpreted as a complete wave-generation model. Translation speed, instantaneous heading, storm size, relative storm quadrant, exposure duration, wind-field asymmetry, wave direction and local bathymetric transformation can also affect peak Hs. Some of these variables are partially represented by distance, wind and site effects, but they are not explicitly modelled. The modest R² therefore reflects both process complexity and omitted covariates. Kruskal–Wallis tests are supplemented with epsilon-squared effect sizes and Holm-adjusted pairwise rank comparisons (Table 3; Supplementary Table 3).
Table 3
| Variable | Kruskal–Wallis H | p-value | Epsilon² |
|---|---|---|---|
| Peak Hs | 6.49 | 0.0391 | 0.004 |
| Duration | 25.34 | <0.001 | 0.019 |
| Nearest distance | 11.67 | 0.0029 | 0.008 |
| Matched TC wind | 30.04 | <0.001 | 0.023 |
Kruskal–Wallis ENSO-phase tests for event-scale variables with epsilon-squared effect sizes.
Holm-adjusted pairwise rank comparisons are given in Supplementary Table S3.
2.5 Scope and limitations
This study focuses on relative, reanalysis-based differences in TC-wave exposure rather than engineering design waves or causal ENSO effects. ERA5 uncertainty is especially relevant during intense TCs and near complex coasts, and errors may depend on storm intensity, translation speed, size and proximity. If those characteristics are distributed differently among ENSO phases, phase-dependent reanalysis error is possible. The statistical associations are also limited by 47 annual observations and incomplete representation of storm and coastal physics. The results therefore require independent buoy/satellite validation and higher-resolution coupled modelling before predictive or design use.
3 Results
3.1 Mean-state wave climate and the high-wave tail
The mean-state wave climate at the six nearshore sites is relatively moderate. The multi-year mean Hs ranges from 1.14 to 1.46 m, with the largest mean value at GD2 and the lowest at GD6. In contrast, the maximum event-scale Hs (typhoon-related) reaches 7.35–10.77 m (Table 1; Figure 1). This contrast confirms that the Guangdong nearshore wave climate is strongly skewed: the mean state reflects recurrent monsoon forcing and coastal sheltering, whereas the upper tail is governed by a small number of energetic TC events.
The spatial gradient is not monotonic along the coast. The eastern Guangdong and Pearl River Estuary sites (GD1–GD4) exhibit relatively high upper-percentile Hs, while the site east of the Leizhou Peninsula (GD6) is more sheltered in the mean. GD2, located along the eastern Guangdong nearshore, records the highest 99th-percentile Hs among the six sites. GD3 and GD4 share similar mean Hs but differ in their high-percentile structure, consistent with differential exposure to offshore waves approaching the Pearl River Estuary margins. This site dependence justifies retaining the site-event sample rather than averaging all sites before event classification.
These basic statistics provide the baseline for the ENSO analysis. Any ENSO effect on high waves is unlikely to be visible in the mean Hs alone, because the mean is dominated by seasonal monsoon waves. The more relevant quantities are event frequency, event duration, cumulative intensity, and the degree to which TC tracks pass sufficiently close to generate high waves simultaneously at multiple nearshore sites.
3.2 Interannual SST background and annual typhoon-wave intensity
Annual cumulative wave exceedance varies strongly during 1979–2025 (Figure 3). Recalculation from the supplied event catalogue gives r=−0.291 between the TC-season ONI-window composite and annual site-event frequency (N=47, p=0.048; 95% CI: −0.533 to −0.004) and r=−0.408 with annual cumulative exceedance (p=0.004; 95% CI: −0.622 to −0.137). Adjustments for lag-1 autocorrelation yielded effective-sample-size of Neff=46.2 (p=0.050) for event frequency and Neff=43.3 (p=0.006) for cumulative exceedance (Supplementary Table 2). While these adjustments slightly affect significance levels, the overall qualitative interpretation remains that these relationships, though statistically significant, are modest in magnitude.
Figure 3
Mean annual cumulative exceedance is 499.3 ± 289.0 m·h (mean ± SD; n=11 years) for El Niño, 937.6 ± 373.8 m·h (n=12) for La Niña and 810.4 ± 355.1 m·h (n=24) for Neutral conditions. The La Niña mean is 87.8% higher than the El Niño mean, but the contrast is uncertain: a year-level bootstrap gives a mean-difference 95% CI of approximately 178–702 m·h (percentage-difference CI approximately 29–191%). These results indicate that Neutral years cannot be assumed to represent a low-exposure baseline, as they can include individual seasons with strong cumulative exceedances.
The seasonal-window analysis (Figure 2) highlights that the most pronounced negative associations with ONI occur for annual event frequency and cumulative wave exceedance. In contrast, correlations with annual maximum Hs and mean event distance are weaker. This supports a scale-dependent interpretation: ONI is more strongly associated with metrics reflecting annual opportunities for TC-induced wave exposure, rather than with the characteristics of the largest individual wave events. To ensure that conclusions are not biased by the overlapping ONI-window composite, results from individual JJA, ASO, and OND windows are presented explicitly.
It is important to note that these annual-scale relationships should not be interpreted as evidence that negative ONI years systematically lead to storms occurring closer to the analysed sites. Reanalysis of the data reveals that ONI correlates only weakly with mean event distance. Additionally, the differences in nearest-storm distances across ENSO phases are small when compared to the high storm-to-storm variability. The key and more consistent signals are observed in event occurrence, cumulative exceedance, and the westward shift of TC genesis during La Niña conditions.
3.3 ENSO-phase differences in TC track exposure
Figure 4 presents raw 6-h TC track-point counts using a unified colour scale across all ENSO phases, rather than normalizing counts separately for each phase. This approach maintains the phase-specific sample sizes and allows direct visual comparison of absolute track-point occurrence within the regional TC sample of 542 storms. The map serves as a descriptive diagnostic of exposure patterns rather than as a circulation-anomaly composite.
Figure 4
Mean genesis longitude is 131.34°E in El Niño years, 131.36°E in Neutral years and 126.87°E in La Niña years (Table 4), indicating a clear westward displacement of the regional-TC genesis background during La Niña. In contrast, mean nearest distances from the regional TC tracks to the analysed sites are 267.0km (El Niño), 254.8km (Neutral), and 255.6km (La Niña). Bootstrap 95% CIs are broad and strongly overlap; the La Niña − El Niño difference is only −11.4 km (95% bootstrap CI approximately −64.9 to +41.5 km; Cliff’s δ=−0.032). Therefore, the earlier interpretation of La Niña producing systematically closer TC approaches is not supported. Figure 5 makes this uncertainty explicit.
Table 4
| ENSO phase | Regional TCs | Site-events | Events per regional TC | Genesis lon. | Closest lon. | Nearest dist. mean [95% CI] | Residence 700 km | Lifetime max wind | Site-event mean peak Hs |
|---|---|---|---|---|---|---|---|---|---|
| El Niño | 119 | 241 | 2.03 | 131.34 | 115.45 | 267.0 [229.3, 305.8] | 83.4 | 36.7 | 3.58 |
| Neutral | 294 | 685 | 2.33 | 131.36 | 114.78 | 254.8 [231.7, 278.0] | 87.7 | 35.6 | 3.87 |
| La Niña | 129 | 323 | 2.50 | 126.87 | 114.67 | 255.6 [219.1, 292.4] | 83.4 | 34.1 | 3.83 |
ENSO-phase sample hierarchy and regional TC/site-event metrics.
Regional TCs are CMA storms entering within 700 km of the six-site network. Nearest-distance brackets give bootstrap 95% confidence intervals; site-event mean max Hs is the mean peak Hs across matched site-events.
Figure 5
The lower mean bearing in La Niña years describes a different pathway orientation, but it should not be conflated with the minimum-distance metric or with longer phase-mean regional residence time. Mean residence within 700 km is 83.4 h in both El Niño and La Niña and 87.7 h in Neutral conditions. The larger La Niña cumulative exceedance is therefore interpreted as the combined outcome of event frequency, site-event footprint, event duration and storm-specific pathway and wind characteristics, rather than as a simple consequence of closer mean approaches.
Taken together, these results support an associative interpretation of pathway exposure: ENSO phase is linked to differences in the spatial and temporal characteristics of regional TCs, most notably through differences in genesis longitude. However, individual storm proximity remains highly variable, emphasizing the need for caution in interpreting the ENSO-phase differences beyond descriptive diagnostic trends.
3.4 Site-event distributions and phase-dependent wave responses
Figure 6 summarizes the annual distributions of the main site-event and wave-response metrics. Neutral years contain the largest mean annual number of site-events, while La Niña years have the largest mean cumulative exceedance. Annual maximum Hs is largest on average in Neutral years, illustrating why cumulative exposure and peak severity should be reported separately.
Figure 6
No single metric captures the phase dependence. Event counts, cumulative exceedance, annual maximum Hs, mean event-maximum Hs, distance and duration shift differently among phases. The results therefore describe several dimensions of exposure rather than a single scalar “ENSO wave hazard.”
The sample hierarchy is also important. The regional TC sample contains 542 storms (119 El Niño, 294 Neutral and 129 La Niña), whereas the site-event catalogue contains 1,249 responses (241 El Niño, 685 Neutral and 323 La Niña, respectively). The regression contains 1,242 events after excluding zero distance from the logarithmic transformation and is clustered by 318 event-producing parent TCs. These corrected counts replace inconsistent site-event totals in Table 4.
Spatial heterogeneity in site responses remains a notable feature of the dataset, as eastern Guangdong, the Pearl River Estuary margins, western Guangdong, and the Leizhou Peninsula exhibit distinct wave-response profiles due to differences in coastal orientation and sheltering. Consequently, ENSO-phase differences in regional TC track populations do not necessarily result in uniform wave responses at all sites. This spatial variability was explicitly included in the regression analysis through site-fixed effects.
Kruskal–Wallis tests indicate phase differences in peak Hs (H=6.49, p=0.039), duration (H=25.34, p < 0.001), nearest distance (H=11.67, p=0.0029) and matched TC wind (H=30.04, p < 0.001), but epsilon-squared effect sizes are small (0.004–0.023; Table 3). Holm-adjusted pairwise rank comparisons (Supplementary Table 3) further highlight that statistical significance does not necessarily indicate substantive phase separation. These results suggest that caution should be exercised when interpreting nominal significance values in the context of ENSO-phase-driven exposure variability.
3.5 Direct controls on peak wave height
Annual associations describe the climate background but do not identify the immediate controls on peak Hs. To address this, we applied the event-scale cluster-robust model described in Equation 2 to estimate conditional associations with nearest TC distance, matched TC wind and ONI while accounting for persistent site differences.
In the regression analysis (Table 2; Figure 7), standardized coefficients indicate strong relationships for nearest TC distance (β=−0.331) and matched TC wind (β=+0.336), both statistically significant (p< 0.001). In physical units, the fitted coefficients are −0.496 m per one-unit increase in ln(distance in km) and +0.0390 m per 1 m s−1 increase in TC wind. In contrast, the ONI coefficient is smaller and lacks conventional statistical significant: standardized β=−0.063, p=0.066, with a 95% CI spanning zero; the unstandardized coefficient is −0.113 m °C−1 (95% CI=0.231 to +0.005). The model explains 29.2% of event-scale peak-Hs variance. These results support distance and wind as direct conditional predictors in this model, but they do not independently demonstrate an ENSO mechanism for individual peak waves.
Figure 7
Figure 7 provides the main event-scale result, while the corresponding bivariate annual ONI plots are shown in Supplementary Figure 4. The conditional regression confirms that nearest distance and TC wind explain more of the event-scale peak-Hs variation than ONI after site effects are included. However, the remaining 71% of unexplained variance highlights the limitations of deterministic modelling, given the omission of key variables such as storm motion, storm size, quadrant effects, and localized wave transformations, as identified in Section 2.4.
3.6 Sensitivity and robustness of the main results
Finally, we evaluate whether the principal ENSO-related exposure patterns were sensitive to the choice of ENSO classification window and event-analysis settings.
Table 5 compares the original TC-season ONI-window grouping with preceding-winter ONI and a small strong-ENSO subset. The seasonal composite gives a larger La Niña–El Niño contrast in mean cumulative exceedance than the preceding-winter grouping, but the seasonal-window correlations in Figure 2 are retained as the primary check against any weighting artefact in the composite.
Table 5
| Classification | Phase | Years | Events yr−1 | Cum. exceed. (m·h yr−1) | Annual max Hs (m) | Mean event dist. (km) |
|---|---|---|---|---|---|---|
| TC-season ONI | El Niño | 11 | 21.91 | 499.34 | 5.75 | 282.00 |
| TC-season ONI | La Niña | 12 | 26.92 | 937.56 | 6.79 | 324.38 |
| TC-season ONI | Neutral | 24 | 28.54 | 810.40 | 7.38 | 296.79 |
| Pre-winter ONI | El Niño | 16 | 25.56 | 797.13 | 6.84 | 328.75 |
| Pre-winter ONI | La Niña | 18 | 28.56 | 862.74 | 7.18 | 285.41 |
| Pre-winter ONI | Neutral | 13 | 25.08 | 608.44 | 6.41 | 286.15 |
| Strong ENSO years | strong El Niño | 5 | 17.20 | 502.73 | 6.04 | 276.50 |
| Strong ENSO years | strong La Niña | 4 | 23.50 | 963.40 | 7.03 | 308.35 |
ENSO-classification sensitivity based on preceding-winter and the TC-season ONI-window composite.
Cumulative exceedance is in m·h yr−1. Strong-ENSO rows (five El Niño and four La Niña years) are descriptive only.
The strong-ENSO comparison is strictly exploratory, covering only five strong El Niño years and four strong La Niña years. Strong La Niña years have a higher descriptive mean cumulative exceedance in this sample, but the estimate is imprecise and can be dominated by individual active TC seasons. We therefore do not use this nine-year subset to validate, strengthen or modify the main conclusion. ENSO spatial diversity is also potentially important; Eastern-Pacific and Central-Pacific El Niño states can produce different TC responses, as shown in other basins and modelling studies (Mueller et al., 2024), but the Guangdong sample is too small for a robust phase-by-intensity-by-flavour subdivision.
Event-scale phase distributions, retained as Supplementary Figure 2, reinforce the importance of distinguishing statistical significance from practical relevance. Significant rank differences exhibit small epsilon-squared effect sizes (Table 3), suggesting that while statistical evidence exists, the magnitude of ENSO-related differences remains modest.
Autocorrelation-aware annual correlation estimates and confidence intervals are summarized in Supplementary Figure 3 and Supplementary Table 2. These supporting diagnostics replace the previous redundant annual coefficient figure and keep the main Results focused on the principal exposure and regression findings.
4 Discussion
The results highlight a clear scale separation in the influence of ENSO on wave dynamics in Guangdong, rather than a singular ENSO control. At the annual scale, the ONI composite is modestly associated with event occurrence and cumulative exceedance. At the event scale, peak Hs is more strongly related to the immediate storm distance and wind covariates included in the regression. This distinction is important because a seasonal climate index can alter the population of storms that reaches a region without uniquely determining the severity of a given event.
A physically plausible interpretation is that ENSO changes the environmental background for TC genesis and steering, and the coastal wave response then depends on the resulting storm population and local exposure. This is consistent with established western North Pacific studies of ENSO-related genesis, intensity, and track variability (Wang and Chan, 2002; Chia and Ropelewski, 2002; Camargo and Sobel, 2005; Chan, 2005; Camargo et al., 2007). However, the sign and magnitude of ENSO–TC relationships are known to vary by region. For instance, recent Indo-Pacific studies have associated extreme waves and TCs with both ENSO and the Indian Ocean Dipole (Kumar et al., 2024), whereas Atlantic modelling studies shows that ENSO flavour can alter the magnitude of TC suppression or enhancement (Mueller et al., 2024). The Guangdong-centric findings presented here are not meant to suggest a new basin-scale ENSO mechanism. Instead, they provide a region-specific perspective on how ENSO-conditioned TC populations translate into localized high-wave exposure.
Within Guangdong, the robust phase signal is the westward La Niña shift in mean regional-TC genesis longitude, not a large change in closest-approach distance. In contrast, the reanalysed nearest-distance metrics show only minor differences, with the mean distance between TCs and sites during La Niña and El Niño years being small and accompanied by confidence intervals that span zero. Similarly, the regional residence time of TCs within 700 km does not show significant variation across ENSO phases. Therefore, the larger cumulative wave exceedance observed in La Niña years should be interpreted as the composite result of increased event frequency, longer event durations, larger affected-site footprints, distinct pathway orientations, and storm-specific wind field characteristics. Untangling the contributions of these factors would require circulation composites and further dynamical analyses that are beyond the scope of this study, including attribution to factors such as monsoon-trough positioning, subtropical-high variability, vertical wind shear, intraseasonal oscillations, and steering flow.
While ERA5 provides a consistent long-term dataset for this analysis, it is important to note that it may not produce phase-independent error. Previous evaluations over the South China Sea have demonstrated that ERA5 performs well for climatological assessments but tends to smooth extreme winds and waves, especially in the vicinity of complex coastlines (Zhai et al., 2023). Typhoon-wave simulations, however, are known to be sensitive to the representation of wind fields and resolution in regional models (Jia et al., 2023; Wen et al., 2024). Since storm characteristics such as intensity, translation speed, size, track geometry, and proximity to the coast can vary between ENSO phases, potential biases in ERA5 estimates may also differ across phases. Consequently, the relative magnitude and even the sign of phase contrasts in wave exposure could be influenced by these biases. Therefore, the findings presented here should be considered as reanalysis-based estimates until validated against observational data from buoys, satellite altimetry, and high-resolution coupled models (e.g., Bagnasco et al., 2025).
The statistical uncertainty also supports a cautious interpretation of the results. While the annual ONI associations are statistically significant, they remain modest in magnitude. For example, the frequency association is only slightly below the conventional 0.05 significance threshold after adjusting for autocorrelation. Similarly, the event-scale regression explains 29.2% (R²=0.292) of the variance in peak Hs. The ONI coefficient in that regression is small and non-significant at the conventional 0.05 level (p=0.066). Additional unresolved variables, such as storm translation speed, size, heading, duration, quadrant asymmetry, wave direction, and localized wave transformations, likely contribute to the unexplained variance. The results therefore constitute evidence of association and exposure modulation, not a seasonal prediction model or causal attribution.
For practical applications, annual cumulative wave exceedance can provide valuable contextual information when considered alongside peak wave statistics. Persistent or repeated high-wave conditions could plausibly influence decisions related to weather windows, infrastructure inspection planning, fatigue management, scour risk, and maintenance schedules. However, these engineering consequences are site-specific and were not directly evaluated in this study. The relationship between ONI and wave exposure could help inform preliminary seasonal preparedness or provide a basis for enhanced monitoring of TC pathways and local metocean conditions during critical periods. Nevertheless, these associations are insufficient to justify deterministic operational decisions and should not replace robust site-specific analyses. Formal infrastructure design and operational planning require empirical observations, high-resolution hindcasts, and rigorous modelling of structural or seabed responses aligned with established engineering standards (e.g., Goda, 2010; Ministry of Transport of the People's Republic of China, 2013).
Overall, the evidence supports a cautious two-scale interpretation. ENSO background is associated with differences in annual TC-wave occurrence and cumulative exposure, particularly through changes in the regional storm population and genesis distribution, whereas event-maximum Hs is more directly related to storm distance, wind and local site factors. The present analysis does not establish a direct causal pathway from Niño-3.4 SST anomalies to individual coastal wave extremes.
5 Conclusions
This study used 1979–2025 ERA5 wave and wind data, CMA best tracks and ONI indices to examine climate-conditioned TC-wave exposure along the Guangdong coast. The main conclusions are as follows.
The six coastal-shelf grid points have a moderate mean wave climate (mean Hs 1.14–1.46 m) but a pronounced upper tail, with record Hs of 7.35–10.77 m during the analysis period. Mean climate and TC-related high-wave exposure therefore need to be interpreted separately.
The TC-season ONI-window composite is modestly and negatively associated with annual site-event frequency (r=−0.29, p=0.048) and cumulative wave exceedance (r=−0.41, p=0.004). La Niña years have higher mean cumulative exceedance than El Niño years in this sample, but the year-to-year dispersion is large and Neutral years can also contain substantial exposure. These relationships are associative rather than predictive.
Regional TC track diagnostics reveal a westward shift in mean TC genesis longitude during La Niña years, although no significant reduction in mean nearest TC distance is observed. The La Niña–El Niño distance contrast is small relative to its uncertainty. Consequently, the elevated La Niña cumulative exceedance is better described as a combined frequency, duration, site-footprint and pathway effect than as a simple “closer-storm” effect. Neutral years have the largest mean annual maximum Hs, illustrating that peak severity and cumulative exposure are distinct dimensions.
At the event scale, log nearest distance and matched TC wind are the strongest direct predictors included in the regression. The ONI coefficient is small, its 95% confidence interval crosses zero and p=0.066. With R²=0.292, substantial event-scale variability remains unexplained, so ENSO should be treated as climate-context information rather than a direct driver or deterministic predictor of individual peak waves.
For coastal hazard management, the practical value of the framework is to complement—not replace—TC track forecasts, local wave observations and engineering metocean analysis. Seasonal ENSO information may help identify periods when cumulative TC-wave exposure deserves greater attention in preparedness, inspection scheduling and risk screening, but site-specific early warning, design and operational decisions require real-time storm information, validated local models and observations.
Statements
Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.
Author contributions
LL: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Visualization, Writing – original draft. MZ: Writing – review & editing, Methodology. MFZ: Writing – review & editing, Data curation, Investigation. HL: Methodology, Writing – review & editing. YZ: Conceptualization, Project administration, Supervision, Writing – review & editing. JL: Resources, Writing – review & editing.
Funding
The author(s) declared that financial support was received for this work and/or its publication. This research was funded by the National Key R&D Program of China (No. 2023YFB4203301).
Acknowledgments
The authors thank the data providers of ERA5, CMA tropical cyclone best tracks and NOAA ONI. The authors also thank colleagues who provided comments on the regional interpretation of Guangdong nearshore wave exposure.
Conflict of interest
Authors LL, MZ, MFZ were employed by the companies SPIC Green Energy Science & Technology Development Co., Ltd. and State Nuclear Electric Power Planning Design & Research Institute Co., Ltd. Authors YZ and JL were employed by the company Qingdao Marine Geological Engineering Survey Institute Co., Ltd. Author YZ was employed by Key Laboratory of Marine Geophysical Prospecting Technology of Qingdao.
The remaining author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Generative AI statement
The author(s) declared that generative AI was used in the creation of this manuscript. Generative AI tools were used solely for language editing and translation during the preparation of this manuscript. All scientific content, including the conceptual framework, data analysis, interpretation of results, and conclusions, was developed independently by the authors. The authors take full responsibility for the accuracy and integrity of the final manuscript.
Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.
Publisher’s note
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Supplementary material
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fmars.2026.1922361/full#supplementary-material
References
1
AmanteC.EakinsB. W. (2009). “ ETOPO1 1 Arc-Minute Global Relief Model: Procedures, data sources and analysis,” in Noaa Technical Memorandum Nesdis Ngdc-24 ( National Geophysical Data Center, NOAA, Boulder, CO, USA). doi: 10.7289/V5C8276M
2
BagnascoT.StocchinoA.VousdoukasM. I.WangJ. (2025). A two-way coupled wave-current high-resolution hindcast for the South China Sea. Appl. Ocean Res.161, 104672. doi: 10.1016/j.apor.2025.104672
3
BarnstonA. G.TippettM. K.RanganathanM.L'HeureuxM. L. (2019). Deterministic skill of ENSO predictions from the North American Multimodel Ensemble. Clim. Dyn.53, 7215–7234. doi: 10.1007/s00382-017-3603-3
4
BjerknesJ. (1969). Atmospheric teleconnections from the equatorial Pacific. Mon. Weather Rev.97, 163–172. doi: 10.1142/9789811294808_0019
5
CamargoS. J.EmanuelK. A.SobelA. H. (2007). Use of a genesis potential index to diagnose ENSO effects on tropical cyclone genesis. J. Climate20, 4819–4834. doi: 10.1175/jcli4282.1
6
CamargoS. J.SobelA. H. (2005). Western North Pacific tropical cyclone intensity and ENSO. J. Climate18, 2996–3006. doi: 10.1175/jcli3457.1
7
ChanJ. C. L. (2005). Interannual and interdecadal variations of tropical cyclone activity over the western North Pacific. Meteorol. Atmos. Phys.89, 143–152. doi: 10.1007/s00703-005-0126-y
8
ChiaH. H.RopelewskiC. F. (2002). The interannual variability in the genesis location of tropical cyclones in the northwest Pacific. J. Climate15, 2934–2944. doi: 10.1175/1520-0442(2002)015<2934:tivitg>2.0.co;2
9
ColesS. (2001). An Introduction to Statistical Modeling of Extreme Values (London, UK: Springer).
10
GodaY. (2010). Random Seas and Design of Maritime Structures. 3rd Ed (Singapore: World Scientific).
11
Grossmann-MathesonG.YoungI. R.MeucciA.AlvesJ.-H. (2024). Global changes in extreme tropical cyclone wave heights under projected future climate conditions. Sci. Rep.14, 31797. doi: 10.1038/s41598-024-82892-9
12
HersbachH.BellB.BerrisfordP.HiraharaS.HorányiA.Muñoz-SabaterJ.et al. (2020). The ERA5 global reanalysis. Q. J. R. Meteorol. Soc146, 1999–2049. doi: 10.1002/qj.3803
13
HolthuijsenL. H. (2007). Waves in Oceanic and Coastal Waters (Cambridge, UK: Cambridge University Press).
14
JiaL.WuS.HanB.CaiS.WuR. (2023). Wave hindcast under tropical cyclone conditions in the South China Sea: Sensitivity to wind fields. Acta Oceanol. Sin.42, 36–53. doi: 10.1007/s13131-023-2227-1
15
JiangX.XieB.BaoY.SongZ. (2023). Global 3-hourly wind-wave and swell data for wave climate and wave energy resource research from 1950 to 2100. Sci. Data10, 225. doi: 10.1038/s41597-023-02151-w
16
KumarP.YadavA.SardanaD.PrasadR.Rajni (2024). Extreme wave height response to climate modes and its association with tropical cyclones over the Indo-Pacific Ocean. Ocean Eng.296, 116789. doi: 10.1016/j.oceaneng.2024.116789
17
LiZ.FungJ. C. H.WongM. F.LinS.CaiF.LaiW.et al. (2024). Future changes in intense tropical cyclone hazards in the Pearl River Delta region: An air-wave-ocean coupled model study. Nat. Hazards120, 7139–7154. doi: 10.1007/s11069-024-06510-7
18
Longuet-HigginsM. S. (1953). On the statistical distribution of the heights of sea waves. J. Mar. Res.11, 245–266. doi: 10.1002/qj.49708135020
19
LuX.YuH.YingM.ZhaoB.ZhangS.LinL.et al. (2021). Western North Pacific tropical cyclone database created by the China Meteorological Administration. Adv. Atmos. Sci.38, 690–699. doi: 10.1007/s00376-020-0211-7
20
MeucciA.YoungI. R.TrenhamC.HemerM. (2024). An 8-model ensemble of CMIP6-derived ocean surface wave climate. Sci. Data11, 100. doi: 10.1038/s41597-024-02932-x
21
Ministry of Transport of the People's Republic of China (2013). Code of Hydrology for Sea Harbor: JTS 145-2-2013 (Beijing, China: China Communications Press).
22
MuellerT. J.Patricola-DiRosarioC. M.Bercos-HickeyE. (2024). The influence of ENSO diversity on future Atlantic tropical cyclone activity. J. Climate37, 3959–3975. doi: 10.1175/JCLI-D-23-0286.1
23
NOAA Climate Prediction CenterCold & Warm Episodes by Season: Oceanic Niño Index (ONI). Available online at: https://www.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php (Accessed May 29, 2026).
24
RegueroB. G.LosadaI. J.MéndezF. J. (2015). A global wave power resource and its seasonal, interannual and long-term variability. Appl. Energy148, 366–380. doi: 10.1016/j.apenergy.2015.03.114
25
TrenberthK. E. (1997). The definition of El Niño. Bull. Am. Meteorol. Soc78, 2771–2777. doi: 10.1175/1520‑0477(1997)078<2771
26
WangB.ChanJ. C. L. (2002). How strong ENSO events affect tropical storm activity over the western North Pacific. J. Climate15, 1643–1658. doi: 10.1175/1520-0442(2002)015<1643:hseeat>2.0.co;2
27
WenY.XuX.WasedaT.LinP. (2024). Energy flux variations and safety assessment of offshore wind and wave resources during typhoons in the northern South China Sea. Ocean Eng.302, 117683. doi: 10.1016/j.oceaneng.2024.117683
28
YingM.ZhangW.YuH.LuX.FengJ.FanY.et al. (2014). An overview of the China Meteorological Administration tropical cyclone database. J. Atmos. Ocean. Technol.31, 287–301. doi: 10.1175/jtech-d-12-00119.1
29
YoungI. R.RibalA. (2019). Multiplatform evaluation of global trends in wind speed and wave height. Science364, 548–552. doi: 10.1126/science.aav9527
30
ZhaiR.HuangC.YangW.TangL.ZhangW. (2023). Applicability evaluation of ERA5 wind and wave reanalysis data in the South China Sea. J. Oceanol. Limnol.41, 495–517. doi: 10.1007/s00343-022-2047-8
Summary
Keywords
coastal exposure, ENSO, ERA5, Guangdong coast, significant wave height, storm track, tropical cyclones
Citation
Li L, Zhang M, Zhang M, Liu H, Zhang Y and Li J (2026) ENSO-related variability in typhoon-induced high-wave exposure along the Guangdong coast. Front. Mar. Sci. 13:1922361. doi: 10.3389/fmars.2026.1922361
Received
29 June 2026
Revised
22 August 2026
Accepted
24 August 2026
Published
11 September 2026
Volume
13 - 2026
Edited by
Jinghua Wang, Hong Kong Polytechnic University, Hong Kong SAR, China
Reviewed by
Juan Jose Munoz-Perez, University of Cádiz, Spain
Angela Fontán Bouzas, Universidade de Vigo, Spain
Xinyu Huang, Dalian University of Technology, China
Bing Tai, Jiangsu University of Science and Technology, China
Updates
Copyright
© 2026 Li, Zhang, Zhang, Liu, Zhang and Li.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Yihan Zhang, waterwatersea@163.com
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.