Causes and multiyear predictability of the rapid acceleration of U.S. Southeast Sea level rise after 2010

Sea level rise (SLR) is a major consequence of climate warming, increasing coastal flooding risks. Coastal sea level fluctuates from hours to centuries; lower-frequency variability provides a background that modulates high-frequency events such as storms and tides. Superposition of storm-driven SLR on decadal and longer-term rises can produce extreme sea level events with severe socioeconomic impacts. In the past decade, the U.S. Southeast Coast emerged as a hotspot of rapid SLR in the North Atlantic. Observations show a 2010–2022 linear SLR trend of ~10.8 mm/year along the USSEC, 3–4 times larger than 1920–2009 (~2.6 mm/year), contrasting with a weak deceleration north of Cape Hatteras. Competing hypotheses link this acceleration to changes in Gulf Stream (GS)/AMOC strength and position, Florida Current/GS warming, gyre-scale heat convergence related to the NAO, wind-driven Rossby waves, and atmospheric pressure (inverse barometer) effects. Given the flood risk implications, the study aims to identify mechanisms behind the rapid post-2010 acceleration and assess multiyear predictability of USSEC sea level using observations, reanalysis, and initialized decadal predictions with the GFDL SPEAR system.

Prior work identified the US East Coast, particularly the Southeast and Gulf coasts, as regions of accelerated SLR and increased flooding hazard. Studies proposed several mechanisms: (1) Gulf Stream and AMOC changes modulating coastal sea level via geostrophic balance; (2) warming of Florida Current/GS waters and large-scale heat convergence in the North Atlantic subtropical gyre linked to the NAO; (3) wind-driven Rossby waves in the tropical North Atlantic contributing to decadal accelerations; (4) inverse barometer effects from atmospheric pressure changes influencing coastal sea level. Earlier analyses documented an extreme decline in AMOC/GS transport around 2009–2010 and links between AMOC variability at 26.5°N and New England coastal sea level. Literature also established that global warming increases sea level via thermal expansion and land ice melt, and that internal variability can create regional hotspots. However, the relative roles of buoyancy-driven AMOC versus wind-driven variability in the recent acceleration remained debated.

Data: Satellite altimetry sea surface height (SSH) from CMEMS (quarter-degree, 1993–present) and Tide Gauge (TG) records from PSMSL (since early 20th century). TG records are corrected for the inverted barometer (IB) effect using JRA-55 sea level pressure to align with satellite datasets. Linear detrending is applied (gridded and TG) prior to variability analyses to reduce forced trends while acknowledging imperfect separation of forced vs internal signals. Models: The GFDL SPEAR_LO coupled model (MOM6 ocean/ice at ~1° with 1/3° meridional refinement in tropics; AM4/LM4 atmosphere/land at ~100 km) is used in three configurations: (1) a preindustrial control simulation; (2) a SPEAR reanalysis in which atmospheric temperature/winds are nudged to JRA-55 and SSTs (60°S–60°N) are restored to ERSSTv5, yielding an ocean that experiences observation-constrained conditions and buoyancy-driven multidecadal AMOC variability coherent with NAO; (3) initialized decadal hindcasts/forecasts (20-member ensembles) started each 1 January from 1961–2022 from different reanalysis members and integrated 10 years with time-evolving forcings; lead-time-dependent climatology is removed to address drift. Analyses: Empirical Orthogonal Function (EOF) analyses of detrended North Atlantic sea level isolate dominant modes. Low-pass (≥15-year) and high-pass (<15-year) filtering separate multidecadal (AMOC-related) from interannual-to-decadal (wind/NAO-related) variability. Regressions relate sea level principal components to sea level pressure (SLP), Atlantic meridional overturning streamfunction, and meridional heat transport (MHT) convergence/divergence. Steric, thermosteric, and halosteric components are computed by vertical integration of density anomalies; thermosteric variability is linked to advection and surface heat flux. A composite of AMOC transition phases in the control run characterizes associated sea level patterns, GS path shifts, and lagged correlations of AMOC and MHT along latitudes. Predictability is assessed by correlating ensemble-mean hindcast anomalies with observations (TG or satellite), with significance via Monte Carlo resampling. The Average Predictability Time (APT) method diagnoses predictable components (integrated predictability over 1–5 lead years) after removing global warming and AMOC-related components; the fourth predictable mode corresponds to the NA sea level tripole. Wind-stress sensitivity experiments prescribe NAO-like wind stress differences between reanalysis/observations and control to isolate wind-driven sea level responses over ~8-year integrations. Key datasets/metrics: USSEC TG composites; NA sea level tripole EOFs; correlation of monthly sea level tendency with MHT convergence (r=0.62, p<0.01); AMOC at 26.5°N in reanalysis vs RAPID (r=0.59, p<0.05); spectral peaks near ~35 years in control; APT-derived tripole component correlation with EOF3 (r=0.67, p<0.01).

- Observed acceleration: Along the USSEC, 2010–2022 linear SLR trend is ~10.8 mm/year, 3–4× the 1920–2009 trend (~2.6 mm/year), while regions north of Cape Hatteras show weak deceleration. - NA sea level tripole: Leading EOF of detrended NA sea level exhibits a tripole with opposite-signed anomalies in subtropical vs subpolar/tropical regions, with maximum variability near the Gulf Stream path. An extreme low in 2010 coincided with negative NAO and weak AMOC at 26.5°N; from 2010 to 2022 the tripole index rose sharply, contributing to accelerated USSEC SLR. - AMOC linkage: In control and reanalysis, the tripole corresponds to AMOC transition states; streamfunction regressions show patterns matching AMOC adjustment phases. Control run spectra show a ~35-year peak, consistent with internal buoyancy-driven AMOC variability. - Wind-driven contribution: Positive NAO-like wind anomalies (stronger in reanalysis/observations than in control) drive broad subtropical high sea level via heat convergence and gyre spin-up, as confirmed by wind-stress sensitivity runs. Thus, both buoyancy-driven AMOC changes and NAO-driven circulation adjustments jointly caused the post-2010 acceleration. - Processes: Accelerated SLR arises primarily from steric (thermosteric) increases due to meridional heat transport convergence within 25°–45°N. Monthly sea level tendency averaged in subtropical band correlates with MHT convergence/divergence (r=0.62, p<0.01). Halosteric contributions are minor and partly compensating. - Gulf Stream path: During AMOC transition from negative to positive phases, AMOC anomalies propagate southward, GS path shifts northward, and associated northward heat transport anomalies converge heat near the GS path, elevating sea levels there and along the USSEC; opposite in the subpolar gyre. - Predictability: Initialized decadal hindcasts skillfully reproduce post-2010 USSEC sea level rise. Detrended USSEC sea level is predictable up to ~5 years during the satellite era, largely due to correct AMOC initialization; AMOC itself is predictable to ~8 years when verified against reanalysis. Over 1958–2022, USSEC prediction skill is up to ~3 years overall, degrading to ~2 years for high-pass (<15 yr) variability. - Wind-driven tripole predictability: After removing forced and AMOC signals, the most predictable component corresponds to the NA tripole; its time series correlates with reanalysis EOF3 (r=0.67, p<0.01) and is predictable up to 2 years ahead, consistent with Rossby wave adjustment timescales (about two years propagation from 40°W to 60°W at 32°–34°N). - Outlook: Forecasts indicate detrended sea level variability along the USSEC is likely to decrease over the next five years due to weakening wind-driven tripole and slight decreases associated with the AMOC transition state.

The study addresses why USSEC SLR accelerated rapidly after 2010 and whether it is predictable on multiyear timescales. Analyses integrating observations, reanalysis, control simulations, sensitivity experiments, and initialized predictions show the acceleration results from compounded effects of long-term warming, buoyancy-driven multidecadal AMOC variability, and wind-driven NAO-related circulation leading to an NA sea level tripole. AMOC transition states modulate Gulf Stream position and regional heat transport convergence, producing thermosteric sea level increases near the GS path that project onto the USSEC. Concurrent NAO-positive wind forcing enhances subtropical heat convergence and elevates sea level broadly, including the Gulf of Mexico and USSEC. These mechanisms explain the distinct post-2010 acceleration south of Cape Hatteras and weak deceleration to the north. The predictability arises from the slowly evolving AMOC (enabling ~5-year USSEC sea level prediction) and the ocean’s adjustment to NAO forcing (enabling ~2-year predictability via Rossby wave dynamics). These findings underscore the potential to anticipate coastal flooding risks on multiyear horizons, although precisely separating externally forced signals from internal variability remains challenging.

This work links the rapid post-2010 USSEC SLR acceleration to the combined impacts of long-term global warming, buoyancy-driven AMOC variability, and wind-driven NAO-related circulation that produces an NA sea level tripole. It demonstrates multiyear predictability: AMOC-related sea level changes are predictable up to ~5 years, and wind-driven tripole-related changes up to ~2 years. The study’s integrated observational–modeling approach increases confidence in these mechanisms and their forecast relevance. The forecasting system suggests the recent rapid acceleration will likely slow over the next five years as the wind-driven tripole weakens and AMOC-related contributions modestly decrease. Future work should employ higher-resolution models including tides and land ice, account for vertical land motion and other non-climate factors, and further refine methods to disentangle forced and internal components, validating these relationships across multiple modeling frameworks and expanding observational constraints.

- Model resolution and components: SPEAR_LO has relatively low ocean resolution and lacks explicit tide and land ice components; vertical land motion and other non-climate factors are not simulated, limiting direct comparison with coastal impacts. - Forced vs internal separation: Linear detrending and related methods cannot perfectly separate external radiative forcing effects from internal variability; short-term forcings (aerosols, volcanoes, solar) may contaminate modes. - Regional biases: Differences between control and reanalysis tripole patterns suggest that wind stress anomalies play a larger role in observations than in the control, indicating model-dependent dynamics and sensitivities. - Predictability variability: Pre-1995 higher-frequency sea level variability reduces hindcast skill to ~2–3 years over 1958–2022 compared to ~5 years during the satellite era; skill assessments depend on reanalysis verification and may be sensitive to initialization and drift removal. - Observational constraints: TG coverage is pointwise and requires IB corrections; satellite era is short; reanalysis AMOC correlation with RAPID (r~0.59) implies residual uncertainties in AMOC representation.