Climate warming does not force sea-level rise (SLR) at the same rate everywhere. Rather, there are spatial variations of SLR superimposed on a global average rise. These variations are forced by dynamic processes1, 2, 3, 4, arising from circulation and variations in temperature and/or salinity, and by static equilibrium processes5, arising from mass redistributions changing gravity and the Earth’s rotation and shape. These sea-level variations form unique spatial patterns, yet there are very few observations verifying predicted patterns or fingerprints6. Here, we present evidence of recently accelerated SLR in a unique 1,000-km-long hotspot on the highly populated North American Atlantic coast north of Cape Hatteras and show that it is consistent with a modelled fingerprint of dynamic SLR. Between 1950–1979 and 1980–2009, SLR rate increases in this northeast hotspot were ~ 3–4 times higher than the global average. Modelled dynamic plus steric SLR by 2100 at New York City ranges with Intergovernmental Panel on Climate Change scenario from 36 to 51 cm (ref. 3); lower emission scenarios project 24–36 cm (ref. 7). Extrapolations from data herein range from 20 to 29 cm. SLR superimposed on storm surge, wave run-up and set-up will increase the vulnerability of coastal cities to flooding, and beaches and wetlands to deterioration.
We test the hypothesis that a statistically significant observed northeast hotspot (NEH) of accelerated SLR exists by determining its position and dimensions and comparing them with model projections1, 2, 3, 4. We explore correlations between rate changes of observed NEH SLR and of climate indices potentially relevant to NEH formation.
In the late twentieth century, sea levels were relatively low along the North American east coast, particularly north of Cape Hatteras8, 9. Sea-surface gradients sloped down towards the coast away from the Gulf Stream and its continuation to the northeast, the North Atlantic Current10. The sharp pressure gradients balance the Coriolis force to sustain these narrow and strong geostrophic currents, leading to low coastal sea levels.
These low levels could rise with warming and/or freshening of surface water in the subpolar north Atlantic, where less dense water inhibits deep convection associated with the Atlantic Meridional Overturning Current (AMOC). The AMOC weakens and pressure gradients along the North American east coast decrease, raising sea levels. The models considered here simulate this dynamic SLR using Intergovernmental Panel for Climate Change (IPCC) Special Report on Emissions Scenarios warming scenarios2, 3, 4 and/or assumed freshening scenarios1, 4. Gyre system weakening by changes in the North Atlantic Oscillation11, 12 (NAO) could also reduce sea-level gradients and raise sea levels.
To establish the observed NEH, we analyse tide-gauge records along the North American Atlantic coast for increasing rates of SLR (see Methods and Supplementary Information). With least-squares linear regression, rates of SLR were found for the first and second halves of time-series windows and differenced (for example, Supplementary Fig. S7, equation (2)). We also fitted quadratics to each time-series window, computed accelerations, and showed our results were not sensitive to method. As we are concerned with detecting departures from long-term trends, rate differences, or accelerations, can be compared between gauges without first removing signals that are approximately linear over the time series. Processes contributing solely to the longer-term trend (for example, glacial isostatic adjustment) do not affect our analyses13.
Sea-level rate differences (SLRDs) for gauges along the North American east coast show a distinct spatial pattern using time-series windows of 60, 50 and 40 yr (Fig. 1a–c and Supplementary Figs S1 and S2). For 60 yr (1950–2009), the largest SLRDs occur from Cape Hatteras to Boston (mean SLRD=1.97±0.64 mm yr−1; 2σ; confidence intervals account for serial correlation, equations (3)–(5)). South of Cape Hatteras, SLRDs are not statistically different from zero (mean SLRDs=0.11±0.92 mm yr−1), whereas north of Boston, SLRDs are either negative or not different from zero (mean=−0.94±0.88 mm yr−1). The 40-yr window (1970–2009) exhibits the largest mean NEH SLRD (3.80±1.06 mm yr−1), and positive differences continue north of Massachusetts and into Canada. For all three durations, SLRDs south of Cape Hatteras are not significantly different from zero. Similar patterns are found for quadratic accelerations (Supplementary Fig. S2).
Mean NEH SLRD is a factor of ~ 3–4 larger than global SLRD. For the 60-yr window, the global SLRD during 1950–2009 is 0.59±0.26 mm yr−1 (using reconstructed time series14), compared with NEH SLRD of 1.97±0.64 mm yr−1. For the 40-yr window, global SLRD during 1970–2009 was 0.98±0.33 mm yr−1, compared with NEH SLRD of 3.80±1.06 mm yr−1. These strong NEH SLRDs may be associated with AMOC weakening; for observed NEH, model1, 3 results suggest ~ 4.4–19 Sv of weakening by 2100 dependent on scenario and regression window length.
The NEH is unique across coasts of North America between the latitudes of Key West, Florida and St John’s, Newfoundland (Fig. 2 and Supplementary Fig. S3). On the Gulf of Mexico and Pacific coasts, most SLRDs using 60-yr windows are not statistically different from zero or are negative (Fig. 2). Results are similar for 50- and 40-yr windows (Supplementary Fig. S3). The lack of positive acceleration through much of North America is consistent with previous results15 showing that the recent (about 1990) SLR acceleration occurred mostly in the tropics and the Southern Ocean.
The authors of ref. 16 reported ‘little regional dependence’ of SLR acceleration in the US counter to our detection of a NEH. They found mean negative acceleration for 57 US gauges, including 17 in our observed NEH. Fitting a single quadratic equation for the entire time series available at each station, they calculated average accelerations from gauges having record lengths from 60 to 156 yr and compared them. The spatially averaged SLRDs (and accelerations) in NEH are, however, dependent on time-series length (Fig. 3and Supplementary Fig. S4). Statistically significant positive SLRDs were detectable in 40-yr (1970–2009) to 72-yr (1938–2009) windows. SLRDs for windows longer than 72 yr were not significantly different from zero. Seventy-six per cent of the NEH data from ref. 16 were longer than 72 yr. By using variable record lengths, their results are biased towards SLRDs not statistically different from zero, masking the observed NEH.