Rapid 20th century warming reverses 900-year cooling in the Gulf of Maine

The study addresses when recent Gulf of Maine warming began, how unusual it is in a millennial context, and what mechanisms drove it. The Gulf of Maine has experienced exceptional warming in recent decades, including high sea surface temperature trends, marine heatwaves, and subsurface warming, with ecological and fishery impacts. Proposed drivers include atmospheric warming and circulation, Gulf Stream warm core rings, changes in Gulf Stream destabilization point, and internal climate variability, especially links between Gulf Stream position and Atlantic Meridional Overturning Circulation (AMOC) strength. However, instrumental records are short (e.g., Boothbay Harbor SST since 1905), limiting historical perspective. Prior shell-based isotopic work suggested multi-century cooling but lacked continuity. This study reconstructs hydrographic variability over the last 300 years from multiple proxies in Arctica islandica shells, compares with last-millennium CESM-LME simulations, and determines the timing and drivers of the recent warming.

Previous studies document rapid recent warming on the Northeast U.S. shelf and Gulf of Maine, marine heatwaves (2012, 2016), and ecosystem impacts. Multiple works implicate Gulf Stream position and AMOC variability in regional temperature changes, with a more northerly Gulf Stream linked to warmer Gulf of Maine conditions and potentially a weakened AMOC in recent decades. The longest regional SST record (Boothbay Harbor) shows warming since 1905. Wanamaker et al. (2008) used limited Arctica islandica δ18O to suggest 1–2 °C cooling over the last millennium. Broader North Atlantic records (e.g., Scotian Shelf, Gulf of St. Lawrence, Labrador Sea, Cape Hatteras sortable silt) indicate significant late-19th-century circulation and temperature changes, some linked to AMOC strength changes and the end of the Little Ice Age. Global and North Atlantic paleo syntheses and CMIP/CESM studies attribute pre-1850 ocean cooling to volcanic forcing and post-1850 warming to greenhouse gases and aerosols; volcanic forcing also drives multidecadal variability (AMO-like).

Proxy archive and site: Arctica islandica shells were collected near Seguin Island, western Gulf of Maine (43.7°N, 69.8°W) at 38 m depth in 2010–2014. Thirty-four shells were crossdated to build an absolutely dated master chronology extending from 1761–2013, with several shells dated earlier. The site is near the Boothbay Harbor SST station. Shells were prepared by cleaning, embedding in epoxy, sectioning along the growth axis, polishing, and producing acetate peels for chronology; isotope sampling used a reserved block. Crossdating followed dendrochronology techniques to ensure annual age control. Isotope measurements: - Oxygen isotopes (δ18O): Annual shell carbonate increments were milled (100–300 μg) with a micromill; samples were dried at 50 °C, He-flushed, acidified with H3PO4, and analyzed on a ThermoFinnigan Delta Plus XL via GasBench II. Standards NBS18, NBS19, and LSVEC were used; results are VPDB-referenced. Long-term analytical uncertainty ±0.18‰. Annual values from multiple shells were averaged; record spans 1694–2013 with gaps (1703–1746, 1749, 1762–1763, 1767–1768). - Nitrogen isotopes (δ15N): Measured on periostracum (proteinaceous outer shell layer) as a proxy for dissolved nitrate δ15N. Bulk δ15N analyzed on a ThermoFinnigan Delta Plus XI coupled to a Costech EA; standards (IAEA-600, IAEA-N2, IAEA-CH-3, acetanilide) used; values reported versus Air. Long-term analytical uncertainty ±0.36‰. Each sample integrates 1–21 years (mean 5 years, 2σ=10 years). Decadal-resolution record spans 1751–2008; plotted at midpoints. - Radiocarbon (Δ14C): Previously published age-corrected Δ14C measured at NOSAMS via AMS from shell carbonate; Δ14C computed per Stuiver & Polach using absolute shell age. Average 2σ analytical uncertainty ±5.8‰. Decadal record spans 1685–1935; samples integrate 3–30 years (mean 11 years, 2σ=13 years); plotted at midpoints. Instrumental and gridded data comparisons: The δ18O record was compared to Boothbay Harbor SST (since 1905) and gridded SST/temperature products ERSSTv5, HadISST1.1, OISSTv2.1, and EN4 surface and 35 m for the Gulf of Maine area (domain tailored to each grid). Correlations were computed across varying 12-month periods with leads/lags to account for unknown shell growth season, identifying maximum covariance periods, with autocorrelation accounted for via AR(1). Climate model simulations: Used CESM1 Last Millennium Ensemble (CESM-LME) at ~1° ocean (POP2 with sea ice) and 1.9°×2.5° atmosphere (CAM5). Analyses included 13 fully forced ensemble members (orbital, solar, volcanic, GHG, aerosol/ozone, land-use) and single-forcing ensembles (volcanic n=5, GHG n=3, solar n=4, ozone/aerosol n=5). Model output was regridded to 1°×1°. Gulf of Maine region defined as 41.5°–44.5°N, 73.5°–66.5°W; annual means (Jan–Dec) of temperature and salinity at 35 m were computed. For millennium-scale context, the 996–2005 period was selected and 11-year running means applied. Modeled δ18O calculation: From model salinity and temperature, δ18Owater was estimated using a regional salinity–δ18O relationship (δ18Owater = 0.5·Salinity − 17.3), then combined with temperature using a modified Grossman & Ku paleotemperature relation to produce calculated δ18O time series for comparison with shell records. AMOC metric: Annual maximum Atlantic overturning streamfunction between 20°N–90°N (Eulerian mean) was extracted as AMOC strength. Statistics: Pearson correlations (p<0.05) with AR(1) adjustment assessed δ18O–temperature relationships. Segmented regression (one breakpoint) identified trend change points in shell δ18O and model Gulf of Maine temperatures. Monte Carlo sampling (10,000 iterations) of 100-year linear trends from 1001–2000 provided context for the 1901–2000 trend, for both ensemble mean and individual members. Extreme-year analysis identified top/bottom 10% Gulf of Maine temperatures over 1685–2005; North Atlantic anomalies in upper 200 m temperature, salinity, currents, and sea level pressure were composited; significance via bootstrapping with replacement at 95% level for individual ensembles and ensemble mean.

- The shell δ18O record significantly correlates with regional temperatures, validating it as an annual proxy for Gulf of Maine SSTs. Over 1983–2013, maximum Pearson correlations (with AR(1) adjustment) ranged roughly from −0.50 to −0.70 across ERSST, HadISST, OISST, EN4 (surface and 35 m), and −0.52 to −0.61 with Boothbay Harbor SST, typically with the gridded products lagging δ18O by ~1 year and strongest for 12-month periods starting in late summer–fall. - Multi-proxy reconstructions (δ18O, δ15N, Δ14C) indicate centennial-scale variability. Prior to the mid-1800s, δ18O and Δ14C co-vary indicating alternating influences of warmer/younger WSW and colder/older LSW. From the mid to late 1800s, a rapid shift to high δ18O, high δ15N, and low Δ14C suggests a peak in LSW influence and cold/salty conditions, culminating in the 1870s. - Segmented regression shows a δ18O breakpoint at 1875, marking the transition from long-term cooling to subsequent warming. CESM-LME ensemble mean Gulf of Maine temperature shows a breakpoint at 1904, consistent with proxy timing within uncertainties. - Modeled and observed/proxy trends agree on a long-term (last millennium) cooling followed by rapid 20th-century warming. Modeled calculated δ18O matches multi-decadal variability, with differences attributable to internal variability among ensemble members. - The 20th-century (1901–2000) modeled Gulf of Maine warming rate is 0.44 °C per century for the ensemble mean, faster than 99% of all randomly sampled 100-year trends over the last millennium (Monte Carlo). Individual ensemble member trends range from 0.15 to 0.95 °C per century; even the slowest exceeds 78% of historical trends, while the fastest exceeds 99%. - Single-forcing simulations implicate volcanic forcing as the dominant driver of pre-industrial multi-decadal to centennial Gulf of Maine temperature variability (r=0.56, p<1×10⁻⁵, n=1000 with ensemble mean), including post-eruption recoveries. Post-1850, greenhouse gases (partly offset by aerosols/ozone) dominate warming; volcanic influence wanes. - The late-1800s cold period is prominent in fully forced members and in proxy data but not clearly reproduced by any single forcing ensemble mean, indicating a likely role for internal variability. - Extreme-year composites in CESM-LME show that warm Gulf of Maine years are associated with anomalously warm, salty upper-ocean conditions and a northward-shifted Gulf Stream; cold years show the opposite and NAO-like sea level pressure patterns. Proxy trends indicate increasing WSW contributions since the 1870s, consistent with a northward Gulf Stream shift. - Regional context and other proxies (Scotian Shelf SSTs, Gulf of St. Lawrence benthic δ18O, Labrador Sea algal records, Cape Hatteras sortable silt) corroborate dramatic North Atlantic changes in the late 19th century.

The multi-proxy shell records, together with CESM-LME simulations, demonstrate that recent Gulf of Maine warming represents a reversal of a multi-century cooling trend. Volcanic forcing primarily drove pre-1850 cooling and multidecadal variability, while post-1850 warming is largely attributable to increased greenhouse gases, partially offset by aerosols. The δ18O breakpoint near 1875 and modeled breakpoint near 1904 mark the transition from cooling to warming. Hydrographic proxies (δ18O, δ15N, Δ14C) indicate a 19th-century peak in LSW influence followed by increasing WSW contributions through the 20th century, implying changes in western North Atlantic circulation and Gulf Stream position. Model extreme-year analyses support a strong linkage between Gulf of Maine temperatures and Gulf Stream meridional shifts, with NAO-like atmospheric patterns accompanying cold extremes, although the forced (ensemble mean) signal lacks a robust NAO imprint. The 20th-century warming rate is exceptional in the last millennium in the model context and, combined with the especially cold late-1800s baseline in the Gulf of Maine, explains why regional warming exceeds the global ocean average in the last century. While AMOC variability covaries with Gulf of Maine temperatures in simulations, indicating common responses to volcanic forcing, the simulations do not require a recent strong AMOC weakening to produce observed warming; however, AMOC-related Gulf Stream shifts likely modulate regional hydrography and amplify trends. The nitrogen isotope decline earlier in the record is interpreted as reflecting changes in water column mixing and oxygenation rather than increased WSW, illustrating the value of the multiproxy approach.

This study provides a continuous, absolutely dated, 300-year multiproxy reconstruction of Gulf of Maine hydrography from Arctica islandica shells, complemented by last-millennium climate model simulations. It shows that rapid 20th-century warming reversed at least 900 years of cooling largely driven by volcanic forcing and that recent warming, beginning in the late 1800s to early 1900s, is among the fastest 100-year warmings of the last millennium, driven primarily by greenhouse gases. Proxy evidence indicates increased Warm Slope Water influence and a likely northward shift of the Gulf Stream over the 20th century. These findings contextualize modern warming as unprecedented in the past millennium for the region and imply continued vulnerability to future greenhouse gas increases and potential AMOC weakening. Future work should extend water mass source tracers (e.g., δ15N and Δ14C) in source regions, develop bomb-corrected radiocarbon or alternative tracers to extend beyond the 1950s, improve regional ocean model resolution to resolve Gulf of Maine circulation pathways, and expand proxy networks to refine the timing and mechanisms of late-19th-century cold anomalies and subsequent shifts.

- The Δ14C record ends in 1935 due to the mid-20th-century bomb pulse, limiting direct water-mass source inferences for the latter 20th century and beyond. - CESM-LME resolution (~1° ocean) cannot resolve detailed Gulf of Maine circulation, precluding direct comparison of modeled water mass transports with proxy-inferred changes. - Single-forcing ensembles have relatively few members, reducing power to separate forced signals from internal variability during the late-1800s cold period; no single forcing reproduces that cold spell’s magnitude in the ensemble mean. - Gridded observational products and EN4 are sparse/uncertain in the region before ~1960; Boothbay Harbor SST represents a point location and may differ from broader Gulf of Maine conditions. - Potential influence of anthropogenic nitrogen deposition on δ15N complicates interpretation of nitrogen cycle and source changes, particularly in the 20th century. - Differences between shell-based δ18O records (this study vs. earlier non-crossdated records) and model underestimation of variability highlight proxy dating and model variability limitations; uncertainties remain in model volcanic response amplitude.