Deep ocean cooling and freshening from Subpolar North Atlantic reaches Subtropics at 26.5°N

Abstract

Deep water masses of the North Atlantic Ocean are formed in the subpolar regions through ocean-atmosphere interactions, imprinting unique climate signals like anomalous temperature and salinity on the deep waters that are exported to the Subtropical North Atlantic. Sustained hydrographic, mooring, and seafloor temperature surveys off Abaco Island, Bahamas at 26.5°N have illuminated significant cooling and freshening of the deep ocean (>2000 m) over the past four decades, challenging the paradigm of anticipated deep ocean warming in the Subtropical North Atlantic. Here, we discuss a linkage between the deep Subtropics and Subpolar North Atlantic, finding synchronicity between the observed freshening at 26.5°N and a multi-decadal freshening event in the subpolar basins occurring more than two decades prior. Findings hint at the likely onset of subtropical deep ocean warming and salinification in the near future, which could have notable impacts on deep ocean Atlantic heat content, circulation, and sea level changes.

Introduction

Ocean heat uptake is one of the leading consequences of human-induced warming of the global atmosphere. Amidst a rapidly changing climate, it is critical to observe discrete oceanic changes and accurately predict forthcoming societal impacts. In the Atlantic, large scale ocean circulation is responsible for moving heat, freshwater, nutrients, and carbon from the poles to the tropics. Changes in these transports can yield basin-wide impacts in ocean warming, as well as sea level changes and ecosystem imbalance. Hydrographic evidence has documented wide-spread heat uptake and resultant ocean warming in both the upper (<2000 m) and deep (>2000 m) Atlantic Ocean since the turn of the century^{1,2,3,4,5,6,7,8}. Ocean warming directly impacts coastal sea-level through the process of thermal expansion, and coastal observing systems along the western Atlantic have already seen levels climb past +10 mm per year^9,10,11. As global temperatures continue to rise, heat uptake in the regions of deep water formation that in-part drive large-scale ocean circulation is anticipated to significantly warm the deep ocean; observations have shown that this has already begun⁶.

While property changes (e.g. temperature, salinity) of the surface ocean reflect local heat and freshwater exchanges across the air-sea interface, property changes of the deep ocean are instead reflective of the region of deep water formation and subsequent circulation. These property changes at depth often reflect surface conditions from the past, providing memory to the deep ocean. Through ocean-atmosphere interactions, northward flowing tropical Atlantic surface waters cool, become dense and sink upon arrival in the Subpolar North Atlantic. The Deep Western Boundary Current (DWBC) serves as one of the main equatorward pathways that exports these deep waters to the subtropics at depths below 1000 m. The Abaco 26.5°N hydrographic line (Fig. 1A) is situated at a strategic location within the Subtropical North Atlantic, capturing both the poleward flowing upper-ocean subtropical gyre waters and the deep-ocean equatorward return flows of the DWBC. Over the past 40 years (1985-present), the 26.5°N hydrographic line has been surveyed quasi-annually through collaborative efforts between the National Oceanic and Atmospheric Administration’s (NOAA) Western Boundary Time Series (WBTS) project, the University of Miami Rosenstiel School of Marine, Atmospheric, and Earth Science’s Meridional Overturning Circulation and Heat-flux Array (MOCHA) project, and the National Oceanographic Centre’s Rapid Climate Change (RAPID) program. Sustained monitoring at 26.5°N has been achieved through repeat hydrographic surveys^12,13, the use of Pressure-Inverted Echo Sounders^14,15 (PIES) and continuous mooring installations dating back to 2004¹⁶. Nearly four decades of sustained observations along this latitude make this hydrographic line part of the longest-running trans basin meridional overturning observation program on the planet^16,17,18.

Fig. 1: Nearly four decades of deep ocean freshening and cooling is observed at 26.5°N.

A The Abaco 26.5°N hydrographic line off the coast of Abaco Island, Bahamas. Stations are shown as black dots; bathymetric contours in 1000 m intervals. Deep Western Boundary Current (DWBC) flow is indicated by the gray arrow, with the 140 km DWBC throughflow region highlighted in blue (congruent with Zone 1). B Mean neutral density (γ_n, kg m^-3) structure of the four water masses investigated in this study showing the three partitioned zones of the 26.5°N hydrographic line: Zone 1 captures the equatorward-flowing DWBC, Zone 2 captures the middle region of the transect and the opposing northward return flow of the deep Abaco gyre, and Zone 3 captures the weaker flow regime offshore of the western boundary and into the ocean interior. The four water masses of investigation are shown colored (Intermediate, teal; Deep, light blue; Abyssal, dark blue; AABW (Antarctic Bottom Water), indigo) distinguished by constant neutral density layer boundaries (γ_n, shown). Hydrographic profiles are gridded with 10 km spacing, indicated by the black triangles. C 26.5°N hydrographic trends of salinity (left) and potential temperature (right) of the four defined water masses across each respective zone (Zone 1, circles; Zone 2, crosses; Zone 3, stars) with mean trend per decade shown. Uncertainty represents the 95% confidence interval.

Elsewhere in the North Atlantic, observing systems across the advective path of the DWBC have monitored circulation and variability of the deep water mass properties therein. These deep water masses (e.g. Labrador Sea Water, Iceland Scotland Overflow Water, Northeast Atlantic Deep Water, Denmark Strait Overflow Water), collectively known as North Atlantic Deep Water (NADW), are formed through the transformation of warm and salty subtropical waters into cold and dense deep waters, each specific to its region of formation and local environmental conditions. Studies in recent decades have highlighted the advection of a notable component of NADW, Labrador Sea Water (LSW), along the path of the DWBC by documenting the onset of an anomalously-low temperature and salinity signature characteristic to convectively-formed LSW masses¹⁹. At 26.5°N, four decades of hydrographic data have revealed the arrival of this subpolar convective signal within the LSW layer (approximately between 1000 and 2000 m) in the late 1990s by the onset of a temperature and salinity minimum²⁰, continued persistence over a decade later²¹, and the complete passage and return to warmer and saline conditions in the 2010s²². The onset of cooling and freshening followed by warming and salinification at 26.5°N mimicked deep water formation conditions that occurred in the LSW source region in the Subpolar North Atlantic 10-15 years earlier^21,22, providing a climatic link between these two Atlantic regions. Upstream of 26.5°N, connectivity between the source of LSW and locations elsewhere in the North Atlantic have further affirmed the linkage of the Subpolar North Atlantic to the intermediate layers of the Subtropical North Atlantic via the DWBC and interior spreading pathways^{19,21,22,23,24,25,26,27,28,29,30,31}. The environmental imprints within the exported NADW aid in the understanding of deep ocean circulation pathways and time scales, but more importantly, can serve as predictors for forthcoming deep ocean conditions in the Subtropical Atlantic given what is known about preexisting conditions in the Subpolar formation region.

With a greater understanding of the interplay between the Subpolar and the Subtropical North Atlantic at intermediate depths (1000–2000 m depth range occupied by LSW), we focus here on the deep ocean (>2000 m) at 26.5°N where nearly four decades of ship-based, tall-mooring, and seafloor hydrographic observations begin to unveil a surprising new take on the matter of ocean warming. For the first time at this location, to the best of our knowledge, we find nearly 40 years of significant cooling and freshening of the deep ocean at 26.5°N, challenging the paradigm of anticipated deep ocean warming in the Subtropical North Atlantic.

Results

Deep Cooling and Freshening at 26.5°N

The Abaco 26.5°N hydrographic line off Abaco Island, Bahamas (hereafter 26.5°N) is divided into three longitudinal zones encapsulating the varying dynamics of this subtropical region (Fig. 1B; refer to Methods). Across all zones, we identify four deep-ocean water mass layers (Intermediate constraining LSW, Deep (note, this italicized name is separate from the descriptor ‘deep’ in reference to all water masses exceeding 2000 m) constraining Northeast Atlantic Deep Water (NEADW) and Iceland Scotland Overflow Water (ISOW), Abyssal constraining Denmark Strait Overflow Water (DSOW), and AABW constraining Antarctic Bottom Water (AABW); Fig. 1B) bounded by constant neutral density isopycnals characterized by historical water mass classification^22,32 and seawater property identification using the existing hydrographic dataset.

Hydrographic trends indicate significant freshening and cooling of the deep ocean (>2000 m) in the defined Deep (approximately 2000–3000 m), Abyssal (~3000–4500 m), and AABW (>4500 m) layers across all three zones since the beginning of observations in 1985 (Fig. 1C). The reported mean decadal trends (Fig. 1C) are significant at a 95% confidence level. Relatively steady and synchronous freshening of the Deep (-0.0047 ± 0.0007 practical salinity units (PSU) per decade), Abyssal (-0.0051 ± 0.0009 PSU per decade), and AABW (-0.0051 ± 0.0008 PSU per decade) layers are contrasted, however, by varying cooling rates, with the Abyssal layer showing the largest decline in layer temperature of -0.0433 ± 0.0148 °C per decade (Deep, -0.0352 ± 0.0062 °C per decade; AABW, -0.0291 ± 0.0070 °C per decade). Across all deep ocean masses (>2000 m) along the entire 26.5°N time series record, we observe continual deep ocean freshening and cooling on the order of -0.0250 ± 0.0051 PSU and -0.2523 ± 0.0866 °C, respectively.

To compare with deep trends below 2000 m, we look above 2000 m in the Intermediate layer where an initial cooling and freshening trend reached a minimum in 2008, followed by a warming and salinification trend into present day. The advection of this anomalous signal within the Intermediate layer is studied extensively in prior studies^20,21,22, and is attributed to the LSW mass that convectively formed in the Labrador Sea in the 1990s which advected through 26.5°N via the DWBC and interior Atlantic pathways²⁷. The most recent hydrographic conditions within the Intermediate layer are nearly comparable to those at the start of the surveying program in 1985, suggesting the complete passage of an anomalous climate signal. This complete passage is further elaborated using salinity anomalies across all three spatial zones in Fig. 2, whereas the Deep, Abyssal, and AABW layers instead show gradual freshening apparent across all zones through time. A rebound to more saline conditions, as observed in the Intermediate layer, is not evident with the current hydrographic data in these deeper layers.

Fig. 2: Salinity anomalies of the deep ocean show continual freshening, with no sign of a salinity rebound as observed in the Intermediate ocean.

Salinity anomalies across longitudinal zones 1-3 along the 26.5°N hydrographic line in time for the Intermediate, Deep, Abyssal, and AABW (Antarctic Bottom Water) layers defined in neutral density (γ_n; kg m^-3). Anomalies of each grid point are computed temporally relative to the time mean over a climatological period of 2001-2020. Negative salinity anomalies are denoted in blue shading, indicative of overall freshening. The positions of actual hydrographic observations are overlaid with the black points. Note, not all hydrographic observations are used for the temporal and spatial gridding scheme (refer to Methods).

To enhance the temporal resolution of the 26.5°N hydrographic time series, the existing quasi-annually sampled shipboard hydrographic transect data is supplemented with the hydrographic time series of four fixed tall moorings and bottom-temperature measurements across the 26.5°N transect, providing continuous measurements dating back to initial deployments in 2004 (Fig. 3). Locally at each fixed mooring location (Fig. 3A–D), significant cooling and freshening of the Deep, Abyssal, and AABW layers persists, while the Intermediate layer shows the onset and recovery of a temperature and salinity minimum concurrent with the zonal hydrographic trends. 12-hour continuous and 1-month smoothed sampling of temperature and salinity measurements from the moorings reveal the high-frequency variability (periods from days to a few months) not captured with quasi-annual shipboard hydrographic data. Further assessment reveals differences between some mooring trends and co-located hydrographic trends in density space (refer to trends in Supplementary Note 2, Supplementary Tables 3-6). As hydrographic measurements fall within the ranges of the mooring observations over the same time periods, these differences could suggest influence of sampling resolution which is indeed worthy of a follow up investigation. We find depth-averaged decadal layer trends of the 26.5°N hydrographic data to be congruent with the density-averaged layer trends highlighted across this study (and others⁷), highlighting significant cooling and freshening across the deep ocean locally at 26.5°N.

Fig. 3: Hydrographic, tall-mooring, and continuous bottom temperature records highlight nearly four decades of deep ocean freshening and cooling at 26.5°N.

Four tall-mooring and bottom-temperature locations across the 26.5°N hydrographic line (center) with potential temperature and salinity time series of (A) WB2 and PIES A2 at -76.74°W, (B) WB3 and PIES B at -76.5°W, (C) WB4 and PIES D at -75.7°W, (D) and WB5 and PIES E at -72°W. 12-hour (thin gray line) and 1 month smoothed (black line) continuous tall-mooring hydrographic data is partitioned by water mass layer classifications and supplemented with shipboard collected mean hydrographic data (via a CTD: Conductivity, Temperature, Depth profiling instrument) sampled ±0.2° of the mooring location (circles). Similarly, continuous hourly PIES (Pressure Inverted Echo Sounder) bottom temperature data (thin gray line) and 1 month smoothed (black line) is supplemented with near-bottom shipboard collected mean hydrographic data sampled ±0.2° of the PIES location (circles). Temperature and salinity trends (thick black line) are shown in degrees per decade and practical salinity units per decade, respectively, with uncertainties representative of the 95% confidence interval.

Bottom temperature is measured by temperature sensors located inside Pressure Inverted Echo Sounders (PIES) deployed next to the tall-moorings along the seafloor³³. Significant cooling of the bottom-most shipboard hydrographic measurement is observed at the inshore PIES sites of A2, B, and D. Bottom temperature of the continuous PIES time series shows little variability along the aforementioned inshore locations, however, observations at sites PIES D (Fig. 3C) and offshore at site PIES E (Fig. 3D) may hint at decadal variability. While the continuous PIES bottom temperature may suggest decadal variability at Site E, linear tendencies in the bottom-most shipboard hydrographic measurement at the PIES location do not show a significant decline in temperature (−0.0036 ± 0.0092 °C per decade; Fig. 3D). Further continuous sampling is needed across all four bottom temperature locations to discern any significant pattern or decadal-interdecadal variability.

It is shown herein that our western-boundary defined AABW layer presents an overall cooling and freshening trend, contrary to the estimates of AABW warming along 24.5°/26.5°N west of the Mid Atlantic Ridge by prior studies^34,35. However, we note that our defined AABW layer likely only captures a small presence of AABW in the Subtropical North Atlantic due to the zonal extent of the hydrographic line (west of -69°W), likely only the western extension of the core. We cannot, therefore, rightfully contrast our observed localized cooling and freshening to the overall Atlantic AABW warming trends^{5,7,8,35,36,37}. Site E, being the furthest offshore at -72°W has the largest likelihood of AABW presence compared to what is likely modified AABW (diluted or mixed with near-bottom NADW) or a lack thereof closer to the western boundary. One of these studies³⁵ also found localized AABW cooling at -72°W (Site E)- and even decadal variability farther south at 16°N- but showed that a warming trend at 26.5°N could be observed farther offshore closer to the AABW core, away from the western boundary and NADW influence. As a standalone finding, the observed freshening and cooling of the AABW layer along the confines of the 26.5°N hydrographic line may suggest a large influence of overlying cold and fresh NADW on AABW property modification in the western Subtropical North Atlantic, or alternatively, a limited presence of AABW in the region.

Upstream Connections

Given the compounding evidence highlighting four decades of significant cooling and freshening of the deep ocean along the Atlantic western boundary at 26.5°N, it now questions whether this signal can be traced back to the source region of these deep waters in the Subpolar North Atlantic. The properties of these deep subpolar waters (LSW from the western Subpolar North Atlantic, and NEADW, ISOW, and DSOW from the eastern Subpolar North Atlantic) are heavily influenced by surrounding environmental conditions within their discrete regions of formation^32,38, leaving distinguishable marking characteristics that aid in the identification of these deep masses throughout the global ocean. Further investigation of the salinity anomalies associated with each water mass at 26.5°N is accomplished by looking upstream at several locations along the DWBC, ocean interior, and source regions (Fig. 4), discovering a linkage between the Subpolar and Subtropical North Atlantic.

Fig. 4: Upstream salinity anomalies show a multi-decadal deep ocean Subpolar freshening event, likely linking downstream observations at Subtropical 26.5°N to what occurred more than twenty years prior in the Subpolar North Atlantic.

Salinity anomalies over depth and time across Subpolar source region locations (Labrador Sea, Irminger Sea, Iceland Basin) and downstream Subtropical locations (Cape Cod, Bermuda, and Abaco). Intermediate, Deep, Abyssal, and AABW (Antarctic Bottom Water) neutral density layer boundaries are shown in black contours, where applicable. Labrador Sea, Irminger Sea, and Iceland Basin anomalies are shown using EN4 hydrographic data from the UK Met Office relative to a 1990–2020 climatological mean. Cape Cod anomalies use shipboard hydrographic profiles from the Line W hydrographic section within the Deep Western Boundary Current (DWBC), Bermuda anomalies use shipboard hydrographic profiles from the Bermuda Atlantic Time Series program, and Abaco anomalies use shipboard hydrographic profiles from the 26.5°N hydrographic line (from the NOAA Western Boundary Time Series Program) within the DWBC (see methods in Ref. ²²). Mean salinities of the deep ocean at each location greater than 2000 m (denser than γ_n = 28.01 kg m^-3_, constraining all deep ocean water masses) are shown in the figure insert, with declining salinity trends indicated by the colored trend line. Decadal trends are shown with uncertainties given by a 95% confidence interval. Decadal wintertime North Atlantic Oscillation (NAO) inverted indices are shown relative to >2000 m subpolar salinity trends.

In the source region of these deep water masses - the Labrador Sea, Irminger Sea, and Iceland Basin—it is evident that an unprecedented multi-decadal salinity anomaly (i.e. freshening event) reached a minimum in the mid-1990s to mid-2000s in the Labrador and Irminger Seas, and mid-2000s in the Iceland Basin in the layers deeper than 2000 m (denser than γ_n = 28.01 kg m^-3; Fig. 4). This subpolar period, from the 1990s to mid-2000s, is characterized by increased freshwater supply from the Nordic overflows, and enhanced surface heat loss due to atmospheric forcing induced by an extreme positive state of the North Atlantic Oscillation^39,40,41 (NAO). Across all upstream subpolar locations, substantial anomalous freshening is followed by salinification after 2000, more notably so in the Labrador and Irminger Seas (Fig. 4). Numerous studies have observed and cataloged this multi-decadal subpolar freshening reversal in both the surface and deep ocean farther upstream in the GIN (Greenland, Iceland, and Nordic) seas, highlighting an end to the termed ‘Great-Salinity Anomaly of the 1960s-1990s’^{39,42,43,44,45,46,47}. This deep fresh signal originating from these subpolar locations becomes evident in the deep anomalies farther downstream in the path of the DWBC as this anomaly is imprinted via deepwater formation, exported out of the subpolar region, and advected equatorward in the Atlantic.

Within the DWBC throughflow offshore of Cape Cod at the Line W hydrographic section^29,31, the onset of this deep freshening anomaly (salinity minimum, Fig. 4) appears by the mid-2000s and farther downstream at 26.5°N by the late-2000s. Within the Atlantic interior using the Bermuda Atlantic Time Series (BATS), the same deep anomaly appears by the late-2000s and early-2010s. As Ref. ^39,42 also highlighted with four decades of observed freshening in the Subpolar North Atlantic, mean salinities across locations at depths greater than 2000 m (denser than γ_n = 28.01 kg m^-3 constraining all deep ocean water masses; Fig. 4, inset) show consistent and significant freshening of the source region basins dating back to the early 1970s with maximum freshening occurring in the early to mid-2000s, followed by a rebound in salinification. A recent reversal and decline in deep ocean temperature beginning in the mid-2010s across both the western and eastern Subpolar North Atlantic^40,48 could explain the secondary upstream freshening event simultaneously occurring from the mid-2010s until 2021 following the salinification period (Fig. 4, upper inset). Outside of the subpolar source region, persistent freshening is observed at all locations greater than 2000 m. As Ref. 22 investigated with the passage of the LSW anomalies within the Intermediate layer (γ_n = 27.87–28.01 kg m^-3, approximately 1200–2000 m) from the source region to 26.5°N, a rebound to saline conditions was shown after the full propagation of the anomaly. Whether or not a salinity rebound in the deeper layers greater than 2000 m can be expected in the Subtropical North Atlantic remains the question of interest. Based upon the mean salinity trends of the deep ocean (Fig. 4, inset), the steepest decline in salinity in the Labrador and Irminger Seas from the early 1990s reaches Cape Cod between 2003 and 2004, Bermuda between 2010 and 2013, and Abaco (26.5°N) between 2006 and 2009. The discrepancy in arrival time of the signal to the subtropics may suggest that deeper NADW (>2000 m; denser than γ_n = 28.01 kg m^-3) advects along a faster route via the DWBC to 26.5°N and a slower interior route to Bermuda, if we speculate similarly based upon conclusions of known Intermediate layer (LSW) advective pathways^21,22,27. An increase in salinity is shown in the mean trend at 26.5°N near the end of the record (2019) which may possibly be the leading edge of the expected subpolar salinification signal from 2000. This uptick is not yet clearly evident in the discrete 26.5°N hydrographic record (Figs. 1–3), however, given the upstream evidence of a multi-decadal salinity anomaly in the Subpolar North Atlantic, we anticipate a salinity and temperature increase of the subtropical deep ocean within the coming decade.

Discussion

Nearly four decades of deep ocean observations along the Abaco 26.5°N hydrographic line have shown evidence of cooling and freshening in the water mass layers below 2000 m. These findings present an anomaly in deep ocean warming, which has already been observed elsewhere across the Atlantic since the turn of the century^{1,2,3,4,5,6,7,8}. We suggest a climatic link between the deep Subpolar North Atlantic and the deep Subtropical North Atlantic through salinity anomalies, where the continuous deep cooling and freshening at 26.5°N is attributed to a multi-decadal freshening event (1970s–2000s) in the subpolar source regions that reached a salinity minimum in the late-1990s to early-2000s, imprinting this anomaly to the deep ocean (>2000 m) through deep-layer convection and water mass formation. The exact forcing mechanisms responsible for the 30+ years of persistent deep subpolar freshening are still uncertain, although many studies attribute this salinity anomaly to NAO surface heat flux-forcing in an extreme positive state^40,49 (see Fig. 4 inset) and anomalously large ice (and thus, freshwater) exports from the Beaufort Gyre in the Arctic Ocean into the Subpolar North Atlantic^50,51,52. Continued investigation is necessary to positively identify the driving mechanism(s) for sustained freshening in the Subpolar North Atlantic given the limited observational record present to detect multi-decadal variability. More recently, surface freshening across the eastern Subpolar North Atlantic beginning in the mid-2010s was proposed to be driven by strong atmospheric forcing, pronounced surface heat loss across the basins, and increased Arctic-sourced freshwater inflows^53,54, which has now subsequently imprinted a new salinity anomaly upon the deep waters formed in this region^55,56,57. Nonetheless, the deep-ocean salinity imprint from the multi-decadal subpolar anomaly discussed herein, and likely future anomalies, will help predict downstream ocean conditions given continued knowledge of NADW export, equatorward circulation patterns, and advective time scales. Based on the upstream observations in the Subpolar North Atlantic, a significant increase in deep ocean salinity and temperature in the western Subtropical North Atlantic is anticipated in the near future following the passage of this subpolar salinity anomaly.

The arrival of this postulated forthcoming deep ocean warming and salinification signal in the near future could have ramifications to ocean heat and freshwater content and therefore impact large scale sea level and ocean circulation in the Subtropical North Atlantic. The tendencies observed herein of the deep ocean at 26.5°N are likely attributed to interdecadal variability observed upstream in the Subpolar North Atlantic as aforementioned, however the observational record is still too short to fully resolve and separate it from any anthropogenic forcing. Future work based on longer observational records and/or sufficiently long ocean and climate model simulations will help to test the hypothesized mechanisms discussed herein. Continued hydrographic monitoring of 26.5°N and upstream locations across the North Atlantic is necessary to discern this variability and to predict the societal and ecosystem impacts in the coming decade.

Methods

Hydrographic Data

Hydrographic data along the 26.5°N hydrographic line are compiled from the NOAA Western Boundary Time Series program (https://www.aoml.noaa.gov/phod/wbts/data.php; https://www.aoml.noaa.gov/ftp/phod/WBTS/Global_Class/GC_calibrated_data_archive/). Full-depth, continuous shipboard hydrographic measurements of pressure, salinity, and temperature via a CTD (Conductivity, Temperature, Depth profiling instrument) are captured quasi-annually from 1985-2023 along the transect due east of Abaco Island, Bahamas spanning -77°W to -69°W (Fig. 5). Hydrographic profiles are pre-processed, calibrated, and quality-controlled via the methods described in Ref. ¹² A secondary round of processing and quality control is performed to limit the impact of short-term (<1 year) variability associated with eddies, planetary waves, and/or Gulf Stream or Subtropical Gyre intrusion²². Hydrographic profiles for each occupation are gridded with 10 km zonal and 1db vertical spacing. Gridded time series of temperature and salinity for the 26.5°N tall mooring array (WB2, -76.74°W; WB3, -76.5°W; WB4, -75.7°W; WB5, -72°W) spanning years 2004-2022 are sourced from the RAPID program (https://rapid.ac.uk/rapidmoc/rapid_data/datadl.php). 12-hour, 20 m gridded profiles are produced via the methods in Ref. ⁵⁸ Time series data from mooring WB5 is extended past 5000 m using available MicroCAT time series records along the WB5 mooring sourced from the British Oceanographic Data Centre (www.bodc.ac.uk), vertically and zonally extended with spline interpolation⁵⁸.

Fig. 5: The 26.5°N hydrographic line since the start of sampling in 1985.

Original shipboard hydrographic stations (via a CTD; Conductivity, Temperature, Depth profiling instrument) are shown as dots, the 10 km-gridded shipboard hydrographic profiles are shown as an X. Zone 1 (blue shading) captures the start of the transect at -77°W to -75.5°W, covering the extent of the Deep Western Boundary Current (DWBC) throughflow spanning years 1985–2023. Zone 2 (orange shading) captures the middle of the transect from -75.5°W to -72°W spanning years 1990–2023. Zone 3 (green shading) captures the offshore component of the section from -72°W to -69°W, spanning years 2001–2020. Mooring coverage (and subsequent Pressure Inverted Echo Sounder (PIES) bottom-temperature locations) are shown in blue (orange).

Bottom temperature time series are generated from the WBTS bottom-mounted Pressure Inverted Echo Sounders (PIES; sites A2, B, D, and E; https://www.aoml.noaa.gov/ftp/phod/WBTS/IES/calibrated_bottom_temperature/) located adjacent to the WB2, WB3, WB4, and WB5 moorings, respectively, from 2004 to 2023. Bottom-temperature measurements are precise to 0.0002 °C and are computed and calibrated similarly to the methods used in Refs. ^33,35,59 for assessment in the South Atlantic using near-bottom shipboard hydrographic measurements within 5 km of the site to correct the continuous hourly sampled time series.

Salinity anomalies relative to the time-series mean of Cape Cod, Bermuda, and Abaco downstream locations are sourced from the hydrographic datasets of the Line W (https://scienceweb.whoi.edu/linew/index.php), Bermuda Atlantic Time Series (http://bats.bios.edu/data/), and 26.5°N Western Boundary Time Series surveying programs. Hydrographic data from Line W and 26.5°N are geographically constrained (shown in Fig. 4), representing stations within the throughflow of the DWBC²². Salinity anomalies of the Subpolar North Atlantic source region use 1° EN4.2.2 monthly gridded salinity profiles^60,61 from 1970 to 2021 (https://www.metoffice.gov.uk/hadobs/en4/) for the central locations of the Labrador Sea (58°N, -50°W), Irminger Sea (60°N, -35°W), and Iceland Basin (58°N, -25°W). Anomalies are computed relative to a 30-year climatological mean of monthly profiles between years 1990 and 2020 based on observation weights. Subpolar EN4 salinity records are validated for accuracy against World Ocean Database high-resolution hydrographic profiles of the deep ocean between years 1970 and 2023 (WOD23⁶²) obtained from the NOAA National Centers for Environmental Information (https://www.ncei.noaa.gov/products/world-ocean-database; refer to Supplementary Note 3: Subpolar EN4 Validation).

Wintertime (December-February) North Atlantic Oscillation indices⁶³ (NAO) are used to compare atmospheric drivers to local salinity changes in the subpolar North Atlantic. Monthly NAO indices were obtained from the NOAA Climate Prediction Center (https://www.cpc.ncep.noaa.gov/products/precip/CWlink/pna/nao.shtml). Monthly NAO indices are decadally averaged (e.g. 1990-2000 represents the 1990s, 2000-2010 represents the 2000s).

26.5°N Zonal Partitioning and Water Mass Definition

The 26.5°N hydrographic line is zonally subdivided into three sections to account for spatial and temporal sampling irregularity (Fig. 5). The NOAA AOML Western Boundary Time Series program (WBTS) began in 2001, taking over from the historical sampling of the NOAA AOML Straits of Florida project beginning in 1985. Early cruises were focused on sampling the core of the DWBC, whereas occupations in recent decades have covered the total extent of the DWBC and extended as far east as -69°W. To achieve a multi-decadal, climatological assessment of the entire hydrographic line, a zonal breakdown and spatial gridding is necessary.

For mean water mass layer assessments across each zone, gridded shipboard hydrographic data are first zonally constrained by the longitudinal range of each defined zone and temporally by the maximum zonal sampling coverage (Zone 1: -77°W to -75.5°W, 1985–2023; Zone 2: -75.5°W to -72°W, 1990-2023; Zone 3: -72°W to -69°W, 2001-2020). Zone 1 captures the equatorward-flowing DWBC^64,65,66, spanning 140 km from the continental shelf (20–140 km offshore). Zone 2 captures the middle region of the transect and the opposing northward deep return flow of the Abaco gyre⁶⁶ (140-490 km offshore), while Zone 3 captures the weaker flow regime offshore of the western boundary and into the ocean interior (490-750 km offshore). Mean zonal (Fig. 6A) and meridional (Fig. 6B) velocity structure show 10 km gridded lowered acoustic doppler current profiler (LADCP) profiles of 10 m-bins collected on WBTS cruises 2002–2018 (https://www.aoml.noaa.gov/ftp/phod/WBTS/Global_Class/) used to divide the line into three zones.

Fig. 6: Mean zonal and meridional velocity of the 26.5°N hydrographic section.

A Zonal (u) and (B) meridional (v) velocity map of the 26.5°N hydrographic line across the three defined zones in a distance coordinate. Distance is shown in kilometers as a distance away from Abaco Island, Bahamas at -77°W. Mean velocities are computed from 10 m-binned, 10 km zonally gridded LADCP (lowered acoustic doppler current profiler) data collected on NOAA Western Boundary Time Series cruises spanning 2002–2018. 10 km gridded shipboard hydrographic stations are shown as black triangles. Neutral density isopycnals (kg m^-3; black contours) represent the defined water mass layers across the transect.

Water masses across the 26.5°N section are defined using deep water (>2000 m) mass criteria²², Atlantic historical water mass classifications³², and through an in-depth analysis of 26.5°N shipboard hydrographic profiles throughout the water column (refer to Supplementary Note 1). Water mass layers from the surface ocean to the seafloor are defined based on uniform temperature and salinity changes in neutral density⁶⁷ (γ_n, kg m^-3) space consistent across all three zones along the hydrographic transect. The defined deep water mass layers of Intermediate, Deep, Abyssal, and AABW are used in this study, refer to Supplementary Note 1 (Supplementary Fig. 1, 2; Supplementary Table 1) for further upper-ocean water mass definition criteria.

The Intermediate layer, defined with constant neutral density isopycnal bounds of γ_n = 27.87–28.01 kg m^-3, resides approximately between 1200–2000 m, and constrains the water masses of Labrador Sea Water (LSW) and Mediterranean Overflow Water (MOW). Prior studies have studied temperature and salinity changes to this layer^20,21,22,27, linking the temperature and salinity minima in the late 2000s (~2008, Fig. 1C) to direct changes in water mass formation upstream in the Subpolar North Atlantic 10-15 years prior (i.e. convective formation in the Labrador Sea) based on the propagation of this climate signal to the subtropics at 26.5°N. Although this study focuses on the deep ocean >2000 m, we include this Intermediate layer time series as comparison to the deep ocean observations below 2000 m and to aid in the understanding of propagating subpolar climate signals through this location.

Below the Intermediate layer resides the Deep layer, defined with constant neutral density isopycnal bounds of γ_n = 28.01–28.10 kg m^-3 approximately between 2000–3000 m. The water masses of Northeast Atlantic Deep Water (NEADW) and Iceland-Scotland Overflow Water (ISOW) are constrained within this density layer. The Abyssal layer, defined with the constant neutral density isopycnal bounds of γ_n = 28.10–28.18 kg m^-3, lies approximately between 3000–4500 m and constrains Denmark Strait Overflow Water (DSOW). Between the Abyssal layer and the seafloor lies the AABW layer, constraining Antarctic Bottom Water (AABW, or modified nearshore water masses thereof) at densities greater than γ_n = 28.18 kg m^-3 and >4500 m. AABW has been previously defined near 26.5°N at isotherms lower than 1.8 °C^34,35, which coincides with the density definition herein of γ_n > 28.18 kg m^-3.

The gridded shipboard hydrographic profiles across all zones and the continuous 12-hr mooring time series from each mooring location are then isopycnally constrained by the Intermediate, Deep, Abyssal, and AABW layers defined by neutral density (Intermediate: γ_n = 27.87–28.01 kg m^-3; Deep: γ_n = 28.01–28.10 kg m^-3; Abyssal: γ_n = 28.10–28.18 kg m^-3; AABW: γ_n > 28.18 kg m^-3). Constrained shipboard hydrographic profiles are averaged vertically across the respective density layer and longitudinally across each zone to generate mean zone layer averages for all 26.5°N occupations (N = 52). The continuous 12-hour sampled mooring time series from the four mooring locations are averaged vertically across the respective density layer. The continuous 12-hour sampled mooring time series of each water mass layer and the PIES hourly sampled continuous bottom temperature time series are smoothed with a 1-month low-pass filter and compared with water mass layer [near-bottom] mean shipboard hydrographic measurements within ±0.2° of the mooring [PIES] location. Mean hydrographic data are reported in °C per decade (potential temperature) or practical salinity units (PSU, unitless) per decade (salinity) with uncertainties representing the 95% confidence interval. Calibrated hydrographic data fall within the World Ocean Circulation Experiment standard accuracy limits¹², ±0.002 °C and ±0.002 PSU, where the statistically significant cooling and freshening observed at 26.5°N are above the shipboard hydrographic sensor (SBE3plus temperature sensor, SBE4 conductivity sensor) detection accuracies. The observed cooling and freshening of the tall mooring temperature and salinity records are statistically significant (at the 95% confidence interval, refer to Supplementary Note 2) and above the tall mooring temperature and salinity (SBE37 MicroCAT) sensor detection accuracies of ±0.002 °C and ±0.002 PSU. Refer to Supplementary Note 2 for the reported trends of the mooring and PIES time series.