Influence of El Niño on the variability of global shoreline position

The study addresses how interannual climate variability, particularly ENSO and other large-scale climate modes, influences global shoreline position through key drivers: regional sea level, wave climate, and river discharge. Predicting shoreline evolution at seasonal-to-interannual scales is challenging due to limited observations and complex multi-scale processes. While sea-level rise and river influences dominate at decadal–centennial scales, the interannual influence of waves and rivers is less well quantified globally. Recent advances in satellite shoreline mapping and cloud computing enable global assessments, yet a comprehensive framework linking climate variability to shoreline drivers has been lacking. The authors aim to quantify the relative contributions of sea level, waves, and rivers to interannual shoreline variability worldwide and to attribute these to dominant climate modes (ENSO, IOD, NAO, SAM), accounting for ENSO’s spatial diversity (EP vs CP El Niño) and seasonal modulation, to improve understanding and predictive capability of climate-induced coastal hazards.

Prior global studies emphasized shoreline response to regional sea-level variability, but emerging work shows waves also contribute to interannual coastal water levels and erosion, and that rainfall/river discharge strongly affects sediment supply and coastal sea level. ENSO is known to modulate sea level, waves, and precipitation far beyond the Pacific via teleconnections. Other modes—IOD in the Indian Ocean, SAM in the Southern Hemisphere, and NAO in the North Atlantic—also impact coastal drivers regionally and via swell propagation. Recent ENSO research highlights diverse regimes (Eastern Pacific vs Central Pacific flavors) and nonlinear seasonal interactions (combination modes), motivating a more complete attribution framework beyond canonical Niño3-based approaches.

• Shoreline proxy and sampling: Waterline extracted from Landsat 5/7/8 monthly median NDWI composites (NDWI>0 water, <0 land) across 1993–2019. Global transects every 0.5° (~50 km) at 14,140 coastal points, aggregated regionally (8 consecutive transects, ~400 km) to damp local effects. Seasonal cycle removed; linear trends removed; 8‑month running mean applied to focus on interannual variability.
• Drivers: Regional sea level anomaly (SLA) from SSALTO/DUACS altimetry with dynamical atmospheric correction (MOG2D-G forced by ERA-Interim). Offshore wave energy flux from ERA5 (Hs^2 × Ts). River discharge from ISBA-CTRIP global hydrological model (0.5°), driven by ERA-Interim; annual/seasonal variability used as proxy for fluvial inputs. All datasets interpolated to shoreline transects (nearest neighbor) and alongshore-median smoothed within 100 km.
• Regression framework (Eq. 1): Interannual shoreline anomaly S(x,t) modeled as linear combination of three drivers: S = α(x)·Sea Level + β(x)·Wave Energy + γ(x)·River Flow. Coefficients estimated via multiple linear regression at each transect. Robustness via randomized hindcasts over sub-periods of 10–27 years; performance assessed with correlations and 95% significance thresholds (accounting for degrees of freedom after smoothing).
• Climate-mode attribution for drivers (Eqs. 2–3): Each driver is represented as a function of ENSO’s two independent modes (E_mode for EP, C_mode for CP), plus their nonlinear seasonal combination modes (terms multiplied by cos(2πt/12)). Additional linear terms represent NAO, SAM, and IOD influences. Local multi-linear regressions yield spatially varying coefficients; contributions of each mode to explained variance are computed.
• ENSO-based shoreline models: Using Eq. (1) with driver expressions in Eq. (2), a simplified shoreline model uses only ENSO terms (Eq. 4). An extended version adds NAO, SAM, IOD (Eq. 5). Skill evaluated globally and regionally; gains over canonical Niño3 regressions documented.
• Statistical treatment: After 8‑month smoothing, about 40 independent time steps remain; shoreline/driver memory times <~24 months. Degrees of freedom computed for multi-linear regressions; 95%/99% correlation thresholds derived by Student t test. Variance contributions calculated as the change in total model variance when omitting a contributor.
• Validation and uncertainties: Satellite shorelines compared to select in-situ beach datasets (e.g., Truc Vert, Torrey Pines, Duck, Narrabeen) show correlations 0.38–0.61 at interannual scales despite spatial offset, supporting the regional-scale approach.

• Shoreline–drivers link (Eq. 1): The multi-driver linear model reproduces interannual shoreline anomalies with a global mean correlation of 0.49 (95% significant), significant along 91% of satellite-derived shorelines and 52% of world shorelines between 60°N–60°S.
• Dominant drivers spatial patterns: Waves dominate open-ocean facing coasts; SLA dominates enclosed seas (Gulf of Mexico, Arabian Sea, Bay of Bengal, East Sea) and the western North American coast; rivers dominate near large deltas and river mouths (Amazon, Niger, Zambezi, Indus, Red), especially in monsoon-influenced tropics.
• Climate modes driving the drivers (Eqs. 2–3): Model correlations with observed interannual anomalies are 0.72 (SLA), 0.64 (wave energy), 0.61 (river flow). ENSO collectively explains about 65% of the modeled variance of driver anomalies globally; nonlinear ENSO–annual combination modes add ~25% of ENSO-related variance. Extratropical patterns contribute on average: NAO ~15%, IOD ~15%, SAM ~8%.
• Added skill from multiple modes: Including NAO, SAM, IOD over ENSO-only improves variance explained by +12% (SLA), +16% (waves), +11% (rivers), reaching total explained variances of ~52%, 41%, and 37%, respectively; gains are strongest in high latitudes and the Indian Ocean margins.
• ENSO-based shoreline prediction: ENSO-only shoreline model (Eq. 4) achieves global mean correlation 0.43 (significant along ~83% of satellite-estimated shoreline; ~47% of total shoreline). Adding NAO, SAM, IOD (Eq. 5) increases global mean correlation to 0.62, notably improving skill along European and Southern Hemisphere coasts.
• ENSO regime specifics: CP El Niño events (e.g., 2009/10; several smaller CP events) dominated during the study period and tended to overshadow EP contributions; nonetheless, nonlinear combination modes are non-negligible globally.
• Practical implication: A computationally efficient, statistically based ENSO-centric framework captures and predicts a large fraction of interannual shoreline variability, including beyond the Pacific via inter-basin teleconnections.

The study demonstrates that interannual shoreline variability worldwide can be largely understood as the response to climate-driven changes in sea level, wave energy, and river discharge, with ENSO being the dominant organizing mode. By explicitly representing ENSO’s spatial diversity (EP vs CP) and its seasonal modulation through combination modes, the framework accounts for a wide range of teleconnections and timing effects, improving upon canonical Niño3-based approaches. The attribution analyses show that extratropical modes (NAO, SAM) and the IOD further modulate drivers and shoreline variability regionally, especially in the North Atlantic/European coasts and Southern Hemisphere mid-latitudes, and along Indian Ocean margins. These findings validate a parsimonious, global-scale approach to decouple and quantify climate influences on shoreline position, offering a practical pathway for seasonal-to-interannual coastal hazard outlooks. The results imply that climate-informed shoreline forecasting and risk assessments can leverage ENSO forecasts for many regions, supplemented by regional climate mode information to enhance skill in extratropical and Indian Ocean settings.

• Main contributions: The paper establishes a global, data-driven framework linking interannual shoreline variability to three key drivers (SLA, waves, river discharge) and attributes their variability primarily to ENSO in its diverse regimes and seasonal combination modes. It quantifies the added value of including NAO, SAM, and IOD, and delivers simplified ENSO-based shoreline prediction models with demonstrable global skill.
• Implications: ENSO’s state provides a robust predictive basis for anticipating interannual shoreline changes, including outside the Pacific via teleconnections. Incorporating additional climate modes enhances regional predictability, particularly at higher latitudes and in the Indian Ocean basin.
• Future research: Develop regional physics-based shoreline projections coupling nearshore wave transformation and sediment dynamics; incorporate anthropogenic influences (river regulation, coastal engineering, subsidence); extend analyses to high latitudes affected by sea ice; improve satellite-derived shoreline algorithms and data quality control; integrate vertical land motion and long-term trends to bridge interannual-to-decadal prediction.

• Spatial and process simplifications: Excludes very high-latitude coasts influenced by sea ice seasonality. Uses offshore wave energy (ERA5) without explicit shelf-to-shore wave transformation, potentially limiting local relevance in wide-shelf regions.
• Focus on variability, not magnitude: Emphasizes interannual variability and correlations rather than precise amplitude of shoreline response; local nonlinear processes and short-term dynamics are smoothed by 8‑month running means.
• Data/model constraints: Satellite shoreline extraction via constant NDWI threshold can be sensitive to cloud/turbidity; monthly composites and smoothing mitigate but do not remove errors. ERA5 and other reanalyses are coarse relative to local processes. ISBA-CTRIP river flows omit anthropogenic regulation and irrigation effects.
• Human interventions not resolved: Urban coastal engineering, nourishment, land reclamation, and changing sediment management can dominate local shoreline behavior but are not explicitly included; regional aggregation (~400 km) aims to damp such effects.
• Missing long-term drivers: Vertical land motion (subsidence/uplift) and secular sea-level rise are not included; the study removes trends to isolate interannual scales.
• Predictability limits: NAO, SAM, and IOD variability unrelated to ENSO is largely stochastic and offers limited seasonal predictability, constraining forecast skill improvements from these modes.