Article https://doi.org/10.

1038/s41467-024-51879-5

Florida Current transport observations

reveal four decades of steady state
1,2
Received: 27 April 2024 Denis L. Volkov , Ryan H. Smith 2, Rigoberto F. Garcia 1,2,
3
David A. Smeed , Ben I. Moat 3, William E. Johns4 & Molly O. Baringer 2
Accepted: 21 August 2024

The potential weakening of the Atlantic Meridional Overturning Circulation

Check for updates (AMOC) in response to anthropogenic forcing, suggested by climate models, is
at the forefront of scientific debate. A key AMOC component, the Florida
1234567890():,;
1234567890():,;

Current (FC), has been measured using submarine cables between Florida and
the Bahamas at 27°N nearly continuously since 1982. A decrease in the FC
strength could be indicative of the AMOC weakening. Here, we reassess
motion-induced voltages measured on a submarine cable and reevaluate the
overall trend in the inferred FC transport. We find that the cable record
beginning in 2000 requires a correction for the secular change in the geo-
magnetic field. This correction removes a spurious trend in the record,
revealing that the FC has remained remarkably stable. The recomputed AMOC
estimates at ~26.5°N result in a significantly weaker negative trend than that
which is apparent in the AMOC time series obtained with the uncorrected FC
transports.

Originating as waters from the Gulf of Mexico Loop Current that enter A unique, sustained observing system in the Florida Straits, con-
the western Straits of Florida, the Florida Current (FC) is constrained sisting of voltage measurements recorded from a submarine tele-
by the relatively shallow bathymetry ( < 800-m depth) between the communication cable installed between Florida and Grand Bahama
Florida Peninsula, Cuba, and the Bahamas before entering the open Island, paired with regular ship surveys in the Florida Straits at 27°N
North Atlantic and becoming the Gulf Stream. This strong, upper- (Fig. 1), was established in 19824–8. Voltage perturbations induced on
ocean flow, where the surface currents can reach velocities of 2 m s−11, the cable by the varying FC flow through the Earth’s magnetic field
represents the beginning of and the swiftest part of the Gulf Stream (EMF; henceforth, used interchangeably with the geomagnetic field)
System. The FC carries approximately 32 Sv (1 Sv = 106 m3 s−1) of warm are calibrated and processed into volume transport estimates using
and salty tropical waters northward and plays an important role in the transports calculated from ship section data, collected with Pegasus
regional and global climate systems2,3. Bolstered by western boundary floats in 1982–19885,6,9, with free-falling dropsonde floats since 199110,
intensification, the FC provides most of the northward volume and and with Lowered and Shipboard Acoustic Doppler Current Profilers
heat transport in the subtropical North Atlantic and accounts for the (LADCP and SADCP) since 20018. Since 2000, this observing system
bulk of both the upper limb of the Atlantic Meridional Overturning has been maintained by the National Oceanic and Atmospheric
Circulation (AMOC) and the western boundary component of the Administration’s (NOAA) Atlantic Oceanographic and Meteorological
subtropical gyre circulation. Therefore, observing and investigating Laboratory (AOML), in Miami, Florida, as part of the Western Boundary
the variability of the FC is of particular importance for detecting and Time Series (WBTS) program. As of now, the cable measurements have
better understanding changes in the Earth’s climate system, as well as provided nearly 40 years of quasi-continuous, daily estimates of the FC
for interpreting regional variations in weather, sea level, and volume transport. Beginning in 2004, WBTS partnered with the Rapid
ecosystems. Climate Change (RAPID) and Meridional Overturning Circulation and

1
Cooperative Institute for Marine and Atmospheric Studies, University of Miami, Miami, FL, USA. 2NOAA Atlantic Oceanographic and Meteorological
Laboratory, Miami, FL, USA. 3National Oceanography Centre, Southampton, UK. 4Rosenstiel School of Marine, Atmospheric, and Earth Science, University of
Miami, Miami, FL, USA. e-mail: denis.volkov@noaa.gov

Nature Communications | (2024)15:7780 1

Article https://doi.org/10.1038/s41467-024-51879-5

Fig. 1 | The Western Boundary Time Series observing network in the Straits of (white circles), two bottom pressure gauges on both sides of the FC at 27°N (yellow
Florida. The network consists of the submarine telecommunications cable stars), and along-track satellite altimetry measurements (magenta dotted
between West Palm Beach and Grand Bahama Island (cyan curve), ship sections line). CTD Conductivity-Temperature-Depth, LADCP Lowered Acoustic Doppler
across the Florida Current (FC) at 27°N with in situ measurements at nine stations Current Profiler, XBT expendable bathythermograph.

Heatflux Array (MOCHA) programs, linking the Florida Straits WBTS down on a centennial timescale, its upper branch, primarily com-
cable recording system with RAPID/MOCHA tall moorings deployed posed of the FC and the Gulf Stream in the subtropical North
between the Bahamas and Africa at about 26.5°N. The resulting RAPID- Atlantic, may also be weakening31, although the trends downstream
MOCHA-WBTS array (henceforth RAPID array) is the oldest trans-basin and upstream of Cape Hatteras can be different32. The four-decade
AMOC observing array in existence (e.g.11,). long FC transport time series inferred from cable measurements at
The AMOC is the principal driver for oceanic meridional heat 27°N represents the longest observational record of a key AMOC
transport in the subtropical North Atlantic12. Most climate models component. This record has exhibited a small negative trend of
that contributed simulations to the Coupled Model Intercomparison about −0.3 Sv per decade. This trend has been monitored and
Project Phase 6 (CMIP6) suggest a decline of the AMOC by 34–45% of documented in several State of the Climate reports published as
its present-day strength towards the end of the 21st century13. A supplements to the Bulletin of the American Meteorological Society
variety of published proxy records indicate that the AMOC may (e.g.33–38,). Depending on the year of publication, the trend has been
already be slowing down on a centennial time scale14–17. Some proxy- deemed either insignificant or marginally significant, therefore, not
based conclusions, however, have lately been criticized, as being attracting particular attention. Several recent studies, however,
based only on a subset of proxy records18. The ability of temperature- have used the overall trend in the FC transport to conjecture
based indices to represent the AMOC has also been questioned19. In about the possible centennial decline of the Gulf Stream and
addition, it has recently been shown that the observed reduction in the AMOC39–41.
sea surface temperatures in the subpolar North Atlantic (also known The objective of this study is to reassess the record-long change
as the warming hole), often assumed to be an indication of the AMOC in the FC strength over the four decades of observations. Specifically,
slowdown, could largely be driven by the atmosphere alone without we reconsider the processing of voltages measured on the submarine
any changes in ocean circulation20. Direct observations at the RAPID cable that are used to derive the FC volume transport time series.
array have shown that the AMOC was in a state of reduced over- Most importantly, we apply a correction for the secular change in the
turning in 2008 and 2016 as compared to the 2004–2008 period21. EMF to the voltage data to derive corrected FC volume transports
On the other hand, the AMOC transport time series shows a small, since 2000. This correction was accounted for in the earlier
but not statistically significant, increase between 2009 and 201822. (1982–1998) part of the cable record7 but neglected later due to a
Similarly, direct measurements obtained from repeat sections across change in data processing more heavily reliant upon calibration with
the Gulf Stream, between New Jersey and Bermuda (beginning in ship section data. We show that with the increased length of the cable
1992), do not support suggestions that the AMOC has started slowing record, explicit consideration of local geomagnetic field variations
down substantially23–25, and no significant weakening of the AMOC has become necessary, especially when studying processes on dec-
has been detected in reconstructions based on sparse hydrographic adal and longer time scales. The analysis of the corrected time series
section data26–28. presented herein challenges the previously made assertions on the
It is possible that the AMOC time series derived from direct statistically significant decline of the FC at 27°N39,40,42. Finally, we
measurements are still too short, and primarily showcase discuss the implications of the corrected FC time series on the AMOC
interannual-to-decadal variations29,30. If the AMOC is indeed slowing estimates at the RAPID array.

Nature Communications | (2024)15:7780 2

Article https://doi.org/10.1038/s41467-024-51879-5

Table 1 | Linear trends of the Florida Current volume transport

Time period 1982–2023 1982–1998 1993–2023 2000–2023
Cable −0.4 ± 0.2 0.1 ± 0.7 − 0.5 ± 0.3 − 0.7 ± 0.4
CableEMF − 0.1 ± 0.2 0.1 ± 0.7 − 0.1 ± 0.3 0.0 ± 0.3
All cruises 0.1 ± 0.3 0.4 ± 0.9 0.1 ± 0.4 1.1 ± 0.7
Dropsonde − 0.1 ± 0.5 1.1 ± 0.8
LADCP 1.0 ± 1.3
SADCP/ 0.7 ± 1.1
LADCP
Altimetry − 0.1 ± 0.2 − 0.1 ± 0.3
EMF
Trends from cable transports corrected for the Earth Magnetic Field.
The trends (Sv per decade) are estimated using measurements from different platforms for
different time intervals.

Fig. 2 | The Florida Current volume transport. Daily transport estimates from the
cable record (black; prior to corrections applied in this study); estimates from the per decade) is obtained from all ship sections in 1993–2023. Within the
Pegasus (orange diamonds) and Pegasus in dropsonde mode (Pegasus-DM; light uncertainty bounds, this trend agrees with the altimetry-derived trend
blue circles) sections; estimates from the dropsonde sections (blue circles); and of −0.1 ± 0.2 Sv per decade for the 1993–2023 period (Fig. 3b). It is
estimates from the Lowered Acoustic Doppler Current Profiler (LADCP) sections interesting to note that the altimetry-derived transports for the
(red circles). The linear trends for 1982–2023, 1982–1998, and 2000–2023 periods 2000–2023 period show the same, statistically insignificant negative
are shown by the orange, cyan, and magenta lines, respectively. trend of −0.1 ± 0.3 Sv per decade (cyan line in Fig. 3b). On the other
hand, the transports estimated from the dropsonde, LADCP, and the
combination of LADCP and SADCP (LADCP/SADCP) sections in
Results 2000–2023 exhibit positive trends of about 1 Sv per decade, although
Long-term changes in the Florida current strength only the trend obtained from the dropsonde sections (1.1 ± 0.8 Sv per
The daily FC volume transports inferred from cable voltages and decade) is found to be significantly greater than zero (Fig. 3c). Both the
estimated from ship section data at 27°N from 1982 to 2023, prior to LADCP and the LADCP/SADCP sections yield insignificant trends of
corrections applied in this study, are presented in Fig. 2. The cable 1.0 ± 1.3 Sv per decade and 0.7 ± 1.1 Sv per decade, respectively (Fig. 3d,
record is nearly continuous except several gaps that occurred due to e). It should be noted that ship section data is relatively sparse and is
operational issues and instrument failures (“Methods”). The cable biased towards good weather days as no calibration cruises are carried
transport shows a time mean of 31.7 Sv and a standard deviation of out in adverse weather conditions.
3.4 Sv. The ship sections yield a time mean of 32.3 Sv and a standard The obtained trends contrast with the overall trend and especially
deviation of 3.1 Sv. The FC transport variability is dominated by sub- with the 2000–2023 trend determined from the cable data (Fig. 2;
seasonal (periods less than a year) signals, while the interannual-to- Table 1). To investigate the observed discrepancies between the trends
decadal variability is weak, with a standard deviation of less than in transports, we compute the concurrent differences between the
1 Sv3,43,44. The 1982–2023 linear trend of the FC volume transport from cable transports and the transports estimated from individual ship
the cable was −0.4 ± 0.2 Sv per decade (the uncertainty indicates the sections (dropsonde, LADCP, and LADCP/SADCP) and from satellite
95% confidence interval; see Methods). Note that this trend is sig- altimetry, as well as their trends from 2000 to 2023 (Fig. 4). Ideally, if
nificant and greater than the trend of about −0.3 Sv per decade the cable data is well calibrated, we should expect the differences to be
reported in earlier studies33–36,39,40. The trend became significant rela- spread around the zero line with no or very small trends over the
tive to the noise as the time series length increased and due to low period of observations. Instead, all the differences exhibit negative
transports in the most recent 2–3 years of observations now included trends, statistically significant for altimetry and LADCP/SADCP, indi-
in the record. During these recent years, the FC transport was about cating that the cable transports are consistently becoming lower than
1.5 Sv lower than the record average45. the transports inferred from all other observing platforms. This ana-
It is important to note that the negative trend in the time series is lysis indicates that there is likely an issue with the cable calibration,
only evident in the data after 2000, when cable measurements reasonably assuming that the ship section data yields more reliable
resumed following a 17-month long (October 1998—March 2000) transport estimates8. The fact that the cable transports are diverging
cessation (“Methods”). Indeed, the 1982–1998 trend of 0.1 ± 0.7 Sv per from both the ship-based and altimetry-based transports in
decade is very small and insignificant, while the 2000–2023 period 2000–2023 raises a question of whether the statistically significant
yields a statistically significant trend of −0.7 ± 0.4 Sv per decade. This decline in the FC transport estimated from cable measurements in
observation is consistent with39, where it was reported that a sig- 2000–2023 (−0.7 ± 0.4 Sv per decade) is a real oceanic signal.
nificant decline in Gulf Stream transport has only become detectable
during the past decade. We note, however, that since the decline can Long-term change in voltages measured on the submarine cable
be traced back to the resumption of cable measurements in 2000, it is To investigate whether the decline in the FC transport observed in the
possible that the observed decline in the FC transport could be due to a cable data since 2000 is a real oceanic signal or a spurious drift due to
change in the data processing procedure (see “Methods” for details) data processing issues, the one-minute voltages measured on the cable
rather than the real oceanic signal. in 2000–2023 were analyzed (Fig. 5). The voltages depend on the
To validate the trends in the cable data, we also estimate the linear depth-averaged velocity, the vertical component of the EMF, the width
trends in the transports estimated from ship section data and from of the strait at the cable location, the electrical conductivity of the
cross-stream satellite altimetry measurements (“Methods”). The trend ocean, and the conductance of sediments and upper conducting crust
estimates for different observing platforms and time intervals are beneath the ocean (Eq. (1) in “Methods”; 7). The time-mean voltage in
presented in Table 1 and in Fig. 3. The estimates from all ship sections 2000–2023 is about 1.25 V. The voltage record exhibits a decreasing
combined in 1982–2023 show a small and statistically insignificant trend, with an overall 2000–2023 voltage reduction of about 0.14 V
trend of 0.1 ± 0.3 Sv per decade (Fig. 3a). The same trend (0.1 ± 0.4 Sv (black line Fig. 5).

Nature Communications | (2024)15:7780 3

Article https://doi.org/10.1038/s41467-024-51879-5

Fig. 3 | The Florida Current volume transport estimates from ship sections and sections, and (e) the combination of Shipboard Acoustic Doppler Current Profiler
satellite altimetry. Linear trends (red lines) of the Florida Current volume trans- (SADCP) and LADCP sections. The cyan line in (b) shows the linear trend in
ports estimates (black dots) from (a) all ship sections, (b) satellite altimetry, (c) 2000–2023.
dropsonde sections, (d) lowered acoustic doppler current profiler (LADCP)

Since the conductivity weighting (Cw) and the conductance depth number that cannot be neglected, in particular when attributing a
(He) in Eq. (1) (“Methods”) can be considered constant, then the relatively small and marginally significant trend in the time series to a
2000–2023 decrease in voltage is either due to a decreasing FC real climate signal. Recall that the 2000–2023 trend in the cable record
transport (TFC) or the vertical component of the EMF (Fz). The EMF is was −0.7 ± 0.4 Sv per decade (Fig. 2, Table 1), which corresponds to a
created by the planet’s moving, molten iron core. Therefore, its poles decline of about 1.7 Sv during this time.
are not stationary but slowly and continuously move over geological
time scales. Based on the International Geomagnetic Reference Field Corrected estimates of the Florida Current volume transport
(IGRF) model46, the north magnetic pole has moved from the Canadian From the analysis presented above, it is clear that with the increased
Archipelago towards Russia over the last century, and it is now located length of the cable record it has become absolutely necessary to cor-
about 450 km away from the geographic North Pole. While in the early rect the 2000–2023 voltage and/or transport data for the secular
years of the cable record it was reasonable to assume that Fz was change in the EMF. The easiest way to accomplish this would be to
steady4, the 1982–1998 cable record was corrected for the secular apply this correction offline to the estimated transports. However, the
change in the EMF by multiplying the motion-induced voltages by a adjustments of the offsets β (Eq. (2) in “Methods”), performed by
factor inversely proportional to the secular change in Fz7. However, for comparing the cable and ship section data during the times of known
the reasons mentioned in Methods, the 2000–2023 cable record has problems with the recording system, are likely to have partially com-
not been corrected for the secular change in Fz. pensated for the EMF change. Therefore, we modified the equation
Using the IGRF model in the NOAA National Center for Environ- used to derive the transports from voltages to explicitly account for
mental Information’s Magnetic Field Calculator (www.ngdc.noaa.gov/ the secular change in Fz (Eq. (4) in “Methods”). We then recalibrated
geomag/calculators/magcalc.shtml), we find that Fz in the Florida the obtained cable record and readjusted the offsets in problematic
Straits (at 27°N, 79.5°W) decreased almost linearly from 3.96 × 10−5 time intervals, before finally calculating the new transport estimates
Tesla (T) in 1999–3.60 × 10−5 T in 2023, equating to the total reduction (see “Methods” for further details).
of about 3.6 × 10−6 T (~150 nT per year) or about 9% from the initial Displayed in Fig. 6a are the obtained FC volume transports in
value. The decreasing Fz in the Florida Straits leads to decreases in 2000–2023 corrected for the secular change in the EMF. The
voltages and the amplitude of voltage variations measured by the 2000–2023 mean FC transport is 31.8 Sv and the standard deviation is
cable (Eq. (1) in “Methods”). Assuming that the FC volume transport is 3.0 Sv. While there are some differences with the earlier record due to
constant (TFC = 32 Sv), the conductivity weighting is 1.06, and the the recalibration of the offsets, in particular in 2000–2005, the most
conductance depth is 1050 m4, the reduction in Fz would correspond significant impact of applying the EMF correction is the elimination of
to a voltage reduction of about 0.12 V measured on the cable (cyan line the negative trend in the FC transports. The trend in 2000–2023
in Fig. 5). Thus, it appears that the voltage reduction due to the secular changed from −0.7 ± 0.4 Sv per decade in the earlier record to
change in the EMF almost fully accounts for the actual voltage 0.0 ± 0.3 Sv per decade in the new record. When we combined the
decrease measured on the cable in 2000–2023 (0.14 V). Given the 2000–2023 cable record corrected for the EMF with the prior
calibration factor α = 24.42 Sv V-1, the voltage reduction of 0.12 V 1982–1998 record, we obtained the overall 1982–2023 trend of
translates to a decrease of ~2.9 Sv in transport. This is a very large −0.1 ± 0.2 Sv per decade, i.e., statistically insignificant and four times

Nature Communications | (2024)15:7780 4

Article https://doi.org/10.1038/s41467-024-51879-5

Fig. 4 | Differences between the concurrent estimates of the Florida Current Acoustic Doppler Current Profiler (LADCP) sections, and (d) the combination of
volume transport. The differences are computed between the transports esti- Shipboard and Lowered Acoustic Doppler Current Profilers (SADCP/LADCP) sec-
mated from cable measurements (prior to corrections applied in this study) and tions. The linear trends of the differences are shown by the red lines.
from (a) dropsonde sections, (b) satellite altimetry measurements, (c) Lowered

and the 2000–2023 trends in the cable and altimetry transports are
close within the uncertainty limits (Table 1).
To further assess the corrected and recalibrated cable transports,
we again consider the differences between the concurrent cable and
ship section transports (Fig. 6b). The time-mean and the RMS of the
differences between the old (not corrected for the EMF) cable trans-
ports and the ship section transports are −0.3 Sv and 1.9 Sv, respec-
tively. As has already been mentioned, the old cable transports in
2000–2023 were consistently becoming lower relative to the ship
section transports. The differences between the old cable and ship
section transports exhibit a negative trend of about −0.7 Sv per decade
(blue line in Fig. 6b). In contrast, the difference between the new
(corrected for the EMF) cable transports and the ship section trans-
ports has no trend (red line in Fig. 6b), meaning that the calibration of
the corrected cable record has improved. Furthermore, the RMS dif-
Fig. 5 | One-minute voltage recorded on the submarine cable between Florida ference between these estimates has reduced to 1.5 Sv. This analysis
and Grand Bahama Island. The linear change in the cable voltage from 2000 to demonstrates that the newly corrected cable record is more accurate
2023 is ΔVCable = −0.14 V (black line). The almost linear change in voltage due to the than the old record. Most importantly, by accounting for the secular
secular change in the vertical component of the Earth’s magnetic field (EMF) is change in the EMF, we show that the FC volume transport has not
ΔVEMF = − 0.12 V (cyan line).
declined significantly over the past 40 years of observations, but in fact
has remained remarkably stable.

smaller than for the cable record not corrected for the EMF Implications for the Atlantic Meridional overturning circulation
( − 0.4 ± 0.2 Sv per decade; Table 1). It is interesting to observe that the The submarine cable has been a key component of the RAPID AMOC
1993–2023 trend in the cable record corrected for the EMF observing array at ~26.5°N since April 2004. Because applying a
(−0.1 ± 0.3 Sv per decade) is the same with the 1993–2023 trend in the correction for the secular change in the EMF makes a significant
altimetry-derived FC transports ( − 0.1 ± 0.2 Sv per decade; Fig. 4b), change in the overall trend of the FC transport, the next logical

Nature Communications | (2024)15:7780 5

Article https://doi.org/10.1038/s41467-024-51879-5

Fig. 7 | The northward transport of the Atlantic Meridional overturning cir-

culation (AMOC), estimated from measurements collected at the RAPID
observing array a ~26.5°N. The seasonal cycle was removed and the residual time
series were low-pass filtered using a 2-year cutoff window. The blue curve and blue
bold line (red curve and red bold line) show the AMOC time series (linear trend)
computed with the Florida Current transport not corrected (corrected) for the
secular change in the Earth Magnetic Field (EMF).

transport increases to 17.1 Sv. The overall AMOC trend becomes

−0.8 ± 0.7 Sv per decade, i.e., it is reduced by about 40% and becomes
only marginally significant (red curve in Fig. 7). With the corrected FC
Fig. 6 | Florida Current (FC) volume transports corrected for the secular change
transport, the AMOC is slightly reduced during the first years of the
in the Earth’s Magnetic Field (EMF). a The time series of the daily FC volume record, but it is increased starting from about 2009, becoming larger
transport: (blue) not corrected for the secular change in the EMF, (red) corrected by almost 1 Sv compared to the earlier estimate toward the end of the
for the secular change in the EMF. The linear trends of the FC transport not cor- time series. It is interesting to note that, using the revised FC trans-
rected and corrected for the EMF are shown by the blue and red lines, respectively. ports, the second maximum in the AMOC time series observed in 2018
b The differences between the cable and ship section transport for the cable data (~18.6 Sv) becomes much closer to high transports in 2005–2006
(blue squares) not corrected for the EMF and (red circles) corrected for the EMF. (~19.3 Sv). These findings provide a stronger support for the previously
The linear trends of the differences (ΔT) not corrected and corrected for the EMF made conclusion that the observed AMOC time series at ~26.5°N
are shown by the blue and red lines, respectively. showcases an interdecadal variability rather than a long-term decline.

Discussion
question to address is how the newly corrected FC transport time There is growing scientific interest in how large-scale ocean circulation
series affects the AMOC calculation. The AMOC at ~26.5°N is composed is changing or will change in the future in response to anthropogenic
of (i) the northward FC transport inferred from cable measurements, forcing. Particular attention is paid to the potential decline of the
(ii) the predominantly northward near-surface meridional Ekman AMOC, fueled by the results based on state-of-the-art climate models
transport, and (iii) the southward upper-midocean transport (between and proxy data reconstructions. Although solid observational evi-
the Bahamas and Africa) down to the deepest northward velocity at dence is still lacking as direct observations of the AMOC are very short
~1100 m. Here, we recalculate the RAPID AMOC time series covering compared to climate time scales, some studies have used relatively
the period from April 2004 to February 2022 using the corrected FC small and often marginally significant trends in observational data as
transports and compare them to the previously published RAPID support for the notion of AMOC slowdown or even collapse (e.g.14,15,17,).
AMOC time series47. Some conclusions based on observational products are oftentimes
In the latter time series, the upper limb of the AMOC at ~26.5°N made without detailed consideration of how the raw data were pro-
had a mean northward transport of about 16.8 Sv, composed of the cessed. Arguments about the robustness of tendencies in observa-
mean FC transport of 31.3 Sv, the Ekman transport of 3.7 Sv, and the tional data products usually rely only on statistical significance39–41. Yet,
upper-midocean transport of −18.2 Sv. Observations at the RAPID array it is important to realize that an inherent nature of observational
showcased that the AMOC was in its strongest state in the first years of records is that they may be subject to spurious drifts and biases,
observations (2004–2007; blue curve in Fig. 7). Then, the AMOC caused by instrumentation and data processing issues. As a result,
transport decreased abruptly reaching its record low value in routine calibration, re-evaluation, and correction of observational data
2009–2010. The recovery of the AMOC from the 2009–2010 dip was records, sometimes years after the initial data were released, is an
not complete and a state of reduced overturning persisted until 201721. integral and inevitable part of any long-term observational program.
The negative AMOC tendency between 2004 and 2018 has been widely Therefore, small and marginally significant (in statistical sense) ten-
used in support of a likely AMOC slowdown, suggested by climate dencies in data products should be treated with caution in any analysis
models and reconstructions based on ocean proxy data (e.g.13,14,16,17,). of long-term changes.
The overall 2004–2022 trend in the previously published RAPID time In this study, we reassessed the 40-year record of the FC volume
series is −1.3 ± 0.7 Sv per decade (blue curve in Fig. 7), thus also sup- transport inferred from motion-induced voltages measured on sub-
porting the AMOC slowdown. marine telecommunication cables in the Florida Straits since 1982. The
After the EMF correction is applied, the mean FC transport over cable record constitutes the longest observational record of a
the RAPID time interval increases to 31.7 Sv, and the mean AMOC boundary current in existence. Therefore, these cable observations

Nature Communications | (2024)15:7780 6

Article https://doi.org/10.1038/s41467-024-51879-5

together with regular cable calibration cruises that directly measure for detection and mechanistic understanding of climate-related
the FC volume transport and water mass properties provide a unique changes and for validating and improving ocean and climate models.
and invaluable dataset to study climate variability in the North Atlantic
on decadal time scales. Because the FC is the major contributor to the Methods
northward transport of both the upper-limb of the AMOC and the To analyze the long-term change in the Florida Current (FC) strength,
subtropical gyre circulation in the North Atlantic, it is reasonable to we use the ~40-year (1982–2023) records of (i) the daily estimates of
expect that on sufficiently long time scales a FC decline could indicate the FC volume transport inferred from voltages measured on sub-
a slowdown in the AMOC and/or the gyre circulation. It should also be marine telecommunication cables between Florida and Grand Bahama
recalled that not all the western boundary flow is contained in the FC. A Island and (ii) the snapshot transport estimates based on direct in situ
small, but highly variable, part of this flow is carried by the Antilles measurements at the 27°N section. In addition, we use (iii) the trans-
Current48. port estimates inferred from cross-stream sea surface height (SSH)
The long history of cable measurements in the Florida Straits differences measured by satellite altimetry in 1992–2023. All data are
involved contributions from several generations of scientists. This has publicly available from the NOAA WBTS project website (https://www.
resulted in some differences in the processing methods of raw voltage aoml.noaa.gov/phod/wbts/).
data during the 1982–1998 and 2000–2023 periods (“Methods”). In
addition, the use of different recording systems over time and recur- Submarine cable measurements
rent instrument failures made it necessary to correct the cable voltages While a robust relationship between cross-stream transports and vol-
and inferred transports for spurious offsets using direct measure- tages and its oceanographic application are not guaranteed every-
ments of the FC transport carried out by ship surveys. After investi- where in the ocean, submarine cables installed between Florida and
gating the differences in data processing during the two cable epochs Grand Bahama Island (Fig. 1) have successfully been used to estimate
(1982–1998 and 2000–2023), we focused our attention on the impact the full-water-column transports across the cable2–4,7. The sustained,
of secular changes in the geomagnetic field on motion-induced vol- nearly continuous measurements started in March 1982 using the
tages measured by the cable, which the earlier part of the record was cable between Jupiter (Florida) and Settlement Point (Grand Bahama
explicitly corrected for. While we could not re-examine the 1982–1998 Island). Since 1993, a nearby cable between West Palm Beach (Florida)
raw voltage record as part of it had been lost3, the small negative trend and Eight Mile Rock (Grand Bahama Island) has been used.
in the FC transport record is associated only with the 2000–2023 The estimates of the FC volume transport from voltage mea-
portion of the time series. The 1982–1998 transports have no mean- surements on submarine cables are based on principles of electro-
ingful trend. Since 2000, high-frequency geomagnetic signals have magnetic induction. When ions present in seawater move with an
been effectively removed by applying a 3-day low-pass filter to the ocean current through the Earth’s magnetic field (EMF), an electric
cable record. However, during this second epoch, the direct correction field is induced perpendicular to the direction of water motion49,50.
for the secular change in the geomagnetic field had previously not Because seawater is a conductive media, the induced electric field is
been done. In our analysis, we find that the observed decrease in vol- proportional to the vertically averaged horizontal flow.
tages over the 2000–2023 period is almost fully accounted for by the The local motion-induced, cross-stream voltages measured on a
secular change in the local EMF. Therefore, as the length of the cable submarine cable can be approximated as follows4:
record continues to grow, correcting it for secular EMF changes is
crucial. F T
Δϕ ≈ C w
z FC
 ð1Þ
The 2000–2023 cable record was reprocessed by first correcting He
the cable voltages for the secular change in the EMF and then recali- D E
brating the FC transports using ship section data. The corrected FC where C w = σv v
σ  is the conductivity weighting, σ is the electrical con-
transport time series is made freely available through the WBTS pro- ductivity, v is the downstream velocity, <> represents a cross-channel
ject’s web page (https://www.aoml.noaa.gov/phod/wbts/). The most average, and the overbar indicates depth-averaging, Fz is the vertical
RW
important outcome of this study is that the overall trend of the FC component of the EMF, T FC ðt Þ = 0 H ð x Þv ð x, t Þdx is the volume
volume transport since 1982 has been revised. The updated trend transport, W is the flow width, t is time, H e = Hð1  λÞ is the
estimate for the 1982–2023 period is −0.1 ± 0.2 Sv per decade, which is conductance depth, H is the water depth, and λ is the ratio of the
not statistically different from zero and 3–4 times smaller than it was solid earth’s conductance to the ocean’s conductance. If there were no
thought previously before accounting for the EMF change. This result electric currents flowing into the sea floor, then λ would be zero. Based
impacts the recently made assertion of the robust weakening of the on collected profiling velocity and temperature data and assuming
FC39. Rather, it appears that the FC has remained remarkably stable that conductivity is a linear function of temperature, the conductivity
over the 40 years of nearly continuous observations. Furthermore, the weighting Cw and the conductance depth He in Eq. (1) were estimated
change in the FC cable transports necessitates a revision in the AMOC to be 1.06 ± 0.02 and 1050 ± 30 m, respectively, and thus assumed to
transports at ~26.5°N since 2004 derived from the RAPID array. The use be constant in time4. The Eq. (1) states that a cross-stream voltage
of the corrected FC transport time series reduces the negative trend in depends on the depth-averaged velocity, the vertical component of
the AMOC from 2004 to 2022 by about 40% and makes it only mar- the EMF, the stream width, the electrical conductivity of the ocean, and
ginally significant in statistical sense. the conductance of the sediments above a resistive zone.
The results of this study indicate that, if climate models are cor- The motion-induced voltages measured on a submarine cable
rect that the AMOC is slowing or will soon slow down, this slowdown have been converted into volume transport estimates via linear
has not yet been reflected in the FC and the AMOC transports, or the transfer coefficients7:
observational records are still too short to detect it with confidence.
This is in accord with other observation-based studies23–28. Never- T FC ðt Þ = αΔϕðt Þ + β + εðtÞ ð2Þ
theless, the likelihood of a future AMOC slowdown and the importance
of both the FC and the AMOC in the regional and global climate where α = 24.42 Sv V–1 is the voltage calibration factor, β is the offset,
variability emphasizes the value of sustained observations in the and ε is the error. The calibration factor α and the offset β were ori-
Florida Straits and in the subtropical North Atlantic at ~26.5°N. The ginally determined by comparison with direct ship-based volume
existing records are just starting to resolve decadal-scale signals rele- transport estimates obtained at the Jupiter – Settlement Point cable
vant to climate variability. Continued observations are thus necessary site in the 1980s7,9.

Nature Communications | (2024)15:7780 7

Article https://doi.org/10.1038/s41467-024-51879-5

Table 2 | Instrument accuracies

Time period 1982–1990 1991–1998 2000–2005 2006–present
Instrument Pegasus Dropsonde Dropsonde LADCP Dropsonde LADCP
RMSE 1.4 3.2 2.2 (1.5) 2.9 (2.0) 1.7 (1.6) 1.6 (1.3)
Accuracy 1.0 1.3 0.8 1.3 0.8 1.3
Cable Error 1.0 2.9 2.1 (1.3) 2.6 (1.5) 1.5 (1.4) 0.9 (0.0)
Root mean squared differences (RMSE) between the cable- and ship-measured transports and individual accuracies of transports inferred from ship sections (8) and cable measurements. The cable
error (bottom row) is inferred from RMSE and the instrument accuracy using Eq. (3) The values obtained from the comparison between the cable transports corrected for the Earth Magnetic Field and
ship section transports are shown in brackets. The zero cable error in the 2006 to present time interval (in brackets) suggests that the accuracy of the estimates obtained from the Lowered Acoustic
Doppler Current Profiler (LADCP) measurements is probably overestimated.

The present-day cable recording system consists of a computer- demonstrates that the impact of ocean conductivity changes on vol-
ized voltmeter that makes a measurement every minute and an auto- tages, for example during the FC warming since 201212,52,53, is negligible.
mated, real-time processing system that involves a three-day lowpass
filter with a 2nd order Butterworth filter passed both forward and back Cable data quality and accuracy
to remove tidal and high-frequency geomagnetic field variations3. The cable record is nearly continuous except several relatively large
gaps that occurred due to operational issues and instrument failures.
Ship-borne measurements The longest gap of 17-months occurred between October 1998 and
The cable recording system is subject to hardware problems that result June 2000 when there was a temporary lapse in the project funding
in data gaps and often in spurious drifts and offsets in the recorded and the building where voltage recording equipment was housed in
voltages. Therefore, routine validation of cable-derived transports is West Palm Beach was closed down. A nearly 2-month gap occurred in
required to ensure the stability of the recording equipment and the September-October 2004, when Hurricanes Frances and Jeanne
robustness of the cable record. This validation has been carried out damaged the building housing similar equipment at Eight Miles Rock
with direct ship measurements at nine stations along the 27°N section on Grand Bahama Island. More recently, there was a 1-month gap in
(Fig. 1). In 1982–1988, the ship section data were collected by Pegasus July 2019 due to a lightning strike at the same facility in the Bahamas,
profilers5, which were used in dropsonde mode after 1984 to cut the and an almost 2-month gap in June-July 2021 due to an instrument
cost of maintaining acoustic sound sources. In the modern era, the failure. Shorter gaps are also present throughout the cable record.
direct measurements of the FC volume transport have been conducted Overall, gaps constitute about 7% of the entire time series. Beginning in
with free-falling GPS dropsonde floats since 1991 (8-10 cruises per year) November 2023, a mechanical issue with the cable has resulted in an
and with Lowered and Shipboard Acoustic Doppler Current Profilers interruption of the daily FC transport time series. As of July 2024, the
(LADCP and SADCP) since 2001 (up to 6 cruises per year). As of June cable record has not yet been resumed, but we are working to resolve
2021, there were 248 dropsonde sections and 104 LADCP/SADCP the problem as soon as possible.
sections. Unfortunately, in mid-2021 the Commonwealth of the Baha- The accuracy of the cable-inferred transports, εcable , can be esti-
mas implemented new regulations for foreign marine scientific mated by comparing them to the transports calculated from ship
research clearance, which resulted in a cessation of research in Baha- sections, using the published accuracies of the dropsonde and LADCP
mian waters until December 2023. sections8:
The methods involved in calibrating and validating the cable
 1=2
records with dropsonde and LADCP measurements are well docu- ð3Þ
εcable = ε2total  ε2ships
mented (e.g.3,8,). In this study, we also use the transport estimates
based on the velocity sections produced by combining the LADCP and
SADCP measurements. The SADCP data are processed and quality where εtotal is the root mean squared (RMS) difference between the
controlled using Common Ocean Data Access System (CODAS) transports obtained from cable measurements and from ship sections,
developed by the University of Hawaii51. The main advantage of these and εships is the accuracy of transports estimated from ship section
combined sections is that they cover almost the entire strait from coast data. In Eq. (3) we assume that the two transport estimates are inde-
to coast, as opposed to the dropsonde and LADCP measurements that pendent, which is true except for the relatively short period when ship
are collected at nine stations across the FC. Unlike the single LADCP section transports were used to determine the linear relationship
velocity profiles associated with each of the nine sampling stations, the between voltage and transport. The difference between the cable- and
SADCP profiles incorporated into the LADCP/SADCP sections are ship-based transports as well as the RMS differences over four time
averaged to 5-minute intervals along the cruise track. As the vessel’s intervals: 1982–1988, 1991–1998, 2000–2005, and 2006–2023, prior to
speed and position varies during the hydrographic survey, so too does corrections applied in this study, are displayed in Fig. 8. We divide the
the spatial distribution and total number of SADCP profiles generated record in this manner to isolate two periods characterized by the
for each survey. Although it is reasonable to expect that the accuracy largest RMS differences and errors in the cable transports (shown in
of LADCP/SADCP sections is greater than the accuracy of LADCP sec- red in Fig. 8). These periods resulted from (i) the active use of the cable
tions, we will conservatively assume that the accuracy of the LADCP/ for telecommunication services in the 1990s54, and (ii) a problematic
SADCP sections is equal to the accuracy of the LADCP sections. The recording system in 2000–20053.
accuracies of transport estimated from the Pegasus, dropsonde, and The lowest RMS difference of ~1.4 Sv is observed in 1982–1988
LADCP are presented in Table 2. using the Pegasus sections (Table 2). The errors of transports esti-
Using 104 Conductivity-Temperature-Depth (CTD) and LADCP/ mated from the Pegasus sections are of the order of 1 Sv4,5. This means
SADCP sections between 2001-2021, we recalculated the conductivity that the error of the cable transports in the 1980s is also ~1 Sv. Using the
weighting parameter Cw in the Eq. (1) and obtained a time mean of observed RMS differences and published accuracies of the dropsonde
1.055 and a standard deviation of 0.014, i.e., almost identical to the one sections8, the errors of the cable transports are 2.9 Sv in the 1990s,
reported in ref. 4. There is a very small and statistically insignificant 2.1 Sv in 2000–2005, and 1.5 Sv in 2006–2021. A comparison with the
linear decrease of −0.003 over the 20 years of ship surveys and there is LADCP sections since 2000 yields the errors of 2.6 Sv in 2000–2005
no correlation with the strength of the transport. This analysis and 0.9 Sv in 2006–2021 (Table 2). The comparison between the cable

Nature Communications | (2024)15:7780 8

Article https://doi.org/10.1038/s41467-024-51879-5

secular change in the EMF. Nevertheless, a problem with this latter

approach is that relatively small secular changes in the geomagnetic-
induced voltages can only be detected when the cable record is suffi-
ciently long. This is one of the reasons why it took about 23 years to
realize that an explicit correction for the EMF is necessary to avoid
misinterpretation of the trend. Therefore, in this study, we first correct
the entire modern era cable voltage data (since 2000) for the secular
change in the EMF, then recalibrate cable transports using recent ship
section data, and finally readjust β in problematic time intervals.

Correcting voltage data for the secular change in the EMF

The yearly vertical component of the EMF, Fz, used in this study comes
from the latest, thirteenth generation International Geomagnetic
Reference Field (IGRF). The IGRF is a set of spherical harmonic coef-
ficients, derived from observations recorded by satellites, ground
Fig. 8 | The differences between the concurrent transports estimated from the observatories, and magnetic surveys46. These coefficients are used as
submarine cable (prior to corrections applied in this study) and ship section input into a mathematical model, which describes the large-scale, time-
measurements. The colored circles show the following differences: (orange cir- varying EMF. Because of the secular variation in the EMF, the IGRF
cles) cable—Pegasus, (light blue circles) cable—Pegasus in dropsonde mode models are specified every 5 years (epochs), with linear interpolation
(Pegasus-DM), (blue circles) cable—dropsonde, and (red circles) cable—Lowered used for intermediate times.
Acoustic Doppler Current Profiler (LADCP). The root mean squared differences, Errors in the coefficients lead to uncertainties in the resulting
computed for each of the four time intervals (1982–1988, 1991–1998, 2000–2005, model field, which are summarized in terms of the corresponding RMS
2006–2023), are shown at the top of the plot. The red rectangles delineate time vector uncertainty of the field when averaged over the Earth’s
intervals with problematic cable data quality, while the blue rectangles delineate surface55. It is suggested that after 2000, when satellite data is avail-
time intervals with better cable data quality.
able, the RMS vector uncertainty is close to 10 nT. It should be noted
that at any specific location, a 10 nT vector error is shared between the
three orthogonal components (x, y, and z). For example, a 10 nT global
and ship sections demonstrates that the quality of the cable record is RMS vector error can correspond to global RMS values of about 5, 5,
not homogeneous over time, which needs to be taken into account and 7 nT for Fx, Fy, and Fz, respectively56. Since 2000, the local Fz in the
when studying long-term changes of the FC. Florida Straits (at 27°N, 79.5°W) has been declining at a rate of ~150 nT
per year. Hence, the RMS error constitutes about 5% of the typical
Differences in the cable data processing yearly decline and can be neglected.
It should be noted that cable voltages recorded before and after the According to Eq. (1) the secular change in the EMF has caused a
1998–2000 gap were processed somewhat differently. Dr. Jimmy Lar- negative trend in recorded voltages and a decrease in the magnitude of
sen, who led the Florida Current project from 1982 to 1998, applied voltage variations. In the 1982–1998 record, these effects were
corrections for high-frequency voltage variations induced in the ocean removed by multiplying the motion-induced voltages by a factor
by the time varying ionospheric and magnetospheric electric currents inversely proportional to Fz7. Following the same methodology in this
(geomagnetic-induced voltages) and for the secular change in the study, we computed the FC volume transports since 2000 using a
EMF7. These corrections were derived from the shore-based magnetic modified version of Eq. (2)
data at San Juan (Puerto Rico) and Fredericksburg (Virginia) observa-
tories. Because these observatories are located far from the Straits of F 0z
T FC ðt Þ = α ΔϕðtÞ + β + εðtÞ ð4Þ
Florida, when cable voltage measurements resumed in 2000 by the F z ðt Þ
WBTS program, it was decided that smoothing the data with a 3-day
low-pass filter would be sufficient to remove the high-frequency geo- in which F 0z is the reference vertical component of the EMF. F 0z was set
magnetic-induced voltages and the impact of tides. Additionally, as the to Fz on Jan 1, 2019, which is about the temporal mid-point of ship
WBTS project began collecting voltage data, the initial record was surveys used to recalibrate cable voltages into volume transports.
relatively short, and it was reasonably assumed that geomagnetic-
induced secular changes were small. Thus, since the 2000, no cor- Recalibration of cable transports and their accuracy
rection for the secular change in the EMF has been applied. To calculate transports using Eq. (4) we recalibrated linear transfer
However, regular calibration cruises have continued to be used to coefficients, α and β, using modern era voltage measurements and ship
detect systematic biases between the cable and ship section trans- section data. Recall that the original α was estimated for the old cable
ports. These biases occurred during the time intervals usually follow- between Jupiter and Settlement Point, so the recalibration for the
ing the replacements and failures of the recording system that modern cable between West Palm Beach and Eight Miles Rock is well
happened several times, mostly in 2000–2005 (e.g., when it was justified. For the recalibration, we chose a subset of cable voltages and
destroyed by hurricanes in 2004). Once detected, these biases are ship section transports between January 1, 2016 and November 12,
minimized by calculating the mean difference between the concurrent 2023. In total, there were 44 ship sections conducted during this time
ship section and cable transports and adding it to the prior value of β in interval, consisting of 22 dropsonde and 22 LADCP sections. The newly
Eq. (2) This adjustment to β represents a piecewise constant function, calculated linear transfer coefficients are: α = 22.5 Sv V-1 and β = 4.2 Sv.
different from zero only in problematic time intervals. The calibration These coefficients will be used for future cable measurements until we
factor α has remained unchanged through the entire 1982–2023 determine that another calibration and/or offset adjustment is
record. Despite the different processing methods used during the required, based on comparisons between the cable and ship section
different cable periods, they did not result in particularly different transports.
spectra3. The implementation of the EMF correction and the recalibration
It has also been assumed that the adjustments of the offset β in of cable transports has led to a reduction in cable error estimates since
problematic time intervals could effectively remove the impact of the 2000 (Table 2). The RMS difference between the concurrent cable and

Nature Communications | (2024)15:7780 9

Article https://doi.org/10.1038/s41467-024-51879-5

ship section transports in 2000–2023 has been decreased from 1.9 Sv Code availability
to 1.5 Sv. In 2000–2005, the RMS difference between the cable and the The codes written to analyze data for this work are available at https://
dropsonde (LADCP) section transports has been decreased from 2.2 Sv codeocean.com/capsule/1291301/tree/v1.
(2.9 Sv) to 1.5 Sv (2.0 Sv), resulting in a conservative cable error of
1.5 Sv. Since 2006, the cable error has become 1.4 Sv based on the References
comparison with dropsonde sections. The RMS of the difference 1. Archer, M. R., Shay, L. K. & Johns, W. E. The surface velocity struc-
between the cable and LADCP section transports in 2006–2023 has ture of the Florida Current in a jet coordinate frame. J. Geophys.
decreased to 1.3 Sv, so that it has become equal to the published Res.: Oceans 122, 9189–9208 (2017).
accuracy of LADCP section transports. This indicates that the accuracy 2. Baringer, M. O. & Larsen, J. C. Sixteen years of Florida Current
of LADCP section transports is probably overestimated and needs to transport at 27°N. Geophys. Res. Lett. 28, 3179–3182 (2001).
be revisited in a future study. 3. Meinen, C. S., Baringer, M. O. & Garcia, R. F. Florida current trans-
port variability: An analysis of annual and longer-period signals.
Transports inferred from SSH gradients Deep-Sea Res. 57, 835–846 (2010).
While the cable recording system provides cost-effective means for 4. Larsen, J. C. & Sanford, T. B. Florida current volume transports from
determining the FC volume transport in near real-time, there have voltage measurements. Science 227, 302–304 (1985).
been efforts to explore the utility of alternative (backup) observing 5. Molinari, R. L., Wilson, W. D. & Leaman, K. Volume and heat trans-
platforms for inevitable system failures or possible future cable ports of the Florida current: April 1982 through August 1983. Sci-
breaks44,57. As part of the WBTS project, a pair of bottom pressure ence 227, 295–297 (1985).
gauges have been deployed in shallow waters (~12 m depth) on both 6. Leaman, K. D., Molinari, R. L. & Vertes, P. S. Structure and variability
sides of the Straits of Florida at 27°N since July 2008 (yellow stars in of the Florida current at 27N: April 1982–July 1984. J. Phys. Ocea-
Fig. 1). Cross-stream pressure differences calibrated into the FC nogr. 17, 565–583 (1987).
transports can explain up to 55% of the cable-derived FC transport 7. Larsen, J. C. Transport and heat flux of the Florida current at 27N
variance57. However, because the pressure gauges are subject to derived from cross-stream voltages and profiling data: theory and
instrument drifts, data gaps, and their records are relatively short, they observations. Philos. Trans. R. Soc. Lond. A 338, 169–236 (1992).
are still of limited use in determining long-term changes in the FC 8. Garcia, R. F. & Meinen, C. S. Accuracy of Florida current volume
volume transport. On the other hand, satellite altimetry has provided transport measurements at 27°N using multiple observational
gap-free and homogeneous-quality along-track sea surface height techniques. J. Atmos. Ocean. Technol. 31, 1169–1180 (2014).
profiles across the FC roughly every 10 days since November 1992 (the 9. Spain, P. F., Dorson, D. L. & Rossby, H. T. PEGASUS: A simple,
satellite track used for the FC transport estimates is shown in Fig. 1). acoustically tracked velocity profiler. Deep-Sea Res. 28A,
Similar to the bottom pressure gauges, the FC transports inferred from 1553–1567 (1981).
the cross-stream SSH gradients measured by altimetry explain about 10. Richardson, W. S. & Schmitz, W. J. A technique for the direct mea-
60% of the 10-day cable-derived FC transport variance44. The accuracy surement of transport with application to the Straits of Florida. J.
of transports estimated from satellite altimetry data is 2.1 Sv, which is Mar. Res. 23, 172–185 (1965).
based on the comparisons with both the cable and the ship sections. 11. Frajka-Williams E., Ansorge I. J., Baehr J., Bryden H. L., and co-
The main advantage of the remote sensing of the FC with satellite authors. Atlantic meridional overturning circulation: Observed
altimetry over in situ observations (cable, ship sections, and bottom transports and variability, Front. Mar. Sci., https://doi.org/10.3389/
pressure gauges) is that it is not impacted by adverse weather fmars.2019.00260 (2019).
conditions. 12. Johns, W. E. et al. Towards two decades of Atlantic Ocean mass and
heat transports at 26.5N,. Philos. Trans. R. Soc. A. 381,
Statistical analysis 20220188 (2023).
The long-term changes in the FC strength from the cable and other 13. Weijer, W., Cheng, W., Garuba, O. A., Hu, A. & Nadiga, B. T. CMIP6
observing platforms were computed by least squares fit of the linear models predict significant 21st century decline of the Atlantic
trend model. The uncertainty in the trend represents the formal error meridional overturning circulation. Geophys. Res. Lett. 47,
of the fit (deviations of observations from the trend), and the random e2019GL086075 (2020).
measurement error using the 95% confidence interval. Recall that the 14. Rahmstorf, S. et al. Exceptional twentieth-century slowdown in
random measurement errors for the cable and ship section measure- Atlantic Ocean overturning circulation. Nat. Clim. Change 5,
ments are time dependent. In this study, for estimating the uncer- 475–480 (2015).
tainties of trends in the cable data, we used the conservative 15. Boers, N. Observation-based early-warning signals for a collapse of
measurement errors for the cable transports shown in Table 2, i.e., 1 Sv the Atlantic Meridional Overturning Circulation. Nat. Clim. Chang.
in the 1980s, 2.9 Sv in the 1990s, 2.6 Sv in 2000–2005, and 1.5 Sv since 11, 680–688 (2021).
2006. The contribution of random measurement errors to the trend 16. Caesar, L. et al. Observed fingerprint of a weakening Atlantic Ocean
uncertainty was estimated using a Monte Carlo style approach for each overturning circulation. Nature 556, 191–196 (2018).
uncertainty estimate by generating 10,000 random time series, with 17. Caesar, L., McCarthy, G. D., Thornalley, D. J. R., Cahill, N. & Rahm-
variable standard deviations equal to the estimated measurement storf, S. Current Atlantic Meridional Overturning Circulation weak-
errors in different time intervals as specified in Table 2. est in last millennium. Nat. Geosci. 14, 118–120 (2021).
18. Kilbourne K. H. et al. Atlantic circulation change still uncertain. Nat.
Data availability Geosci., https://doi.org/10.1038/s41561-022-00896-4 (2022).
All the data used in this work is publicly available, and the references to 19. Little, C. M., Zhao, M. & Buckley, M. W. Do surface temperature
this data are provided in the main text. The Florida Current volume indices reflect centennial- timescale trends in Atlantic Meridional
transports estimated from the cable and ship section measurements Overturning Circulation strength? Geophys. Res. Lett. 47,
are available from the Western Boundary Time Series project web e2020GL090888 (2020).
page: https://www.aoml.noaa.gov/phod/floridacurrent/data_access. 20. He, C. et al. A North Atlantic warming hole without ocean circula-
php. The Atlantic Meridional Overturning Circulation transport data tion. Geophys. Res. Lett. 49, e2022GL100420 (2022).
are available from the RAPID project web page: https://rapid.ac.uk/ 21. Smeed, D. A. et al. The North Atlantic Ocean is in a state of reduced
data.php. overturning. Geophys. Res. Lett. 45, 1527–1533 (2018).

Nature Communications | (2024)15:7780 10

Article https://doi.org/10.1038/s41467-024-51879-5

22. Moat, B. I. et al. Pending recovery in the strength of the meridional 44. Volkov, D. L. et al. Inferring Florida current volume transport from
overturning circulation at 26° N,. Ocean Sci. 16, 863–874 (2020). satellite altimetry. J. Geophys. Res.: Oceans 125, e2020JC016763
23. Rossby, T., Palter, J. & Donohue, K. What can hydrography between (2020).
the New England Slope, Bermuda and Africa tell us about the 45. Volkov D. L. et al. Meridional overturning circulation and heat
strength of the AMOC over the last 90 years? Geophys. Res. Lett. transport in the Atlantic Ocean, [in State of the Climate in 2022].
49, e2022GL099173 (2022). Bull. Amer. Meteor. Soc., 104, S181-S184, https://doi.org/10.1175/
24. Rossby, T. et al. Oleander is more than a flower: Twenty-five years of BAMS-D-23-0076.2.
oceanography aboard a merchant vessel. Oceanography 32, 46. Alken P. et al. International geomagnetic reference field: The thir-
126–137 (2019). teenth generation. Earth Planets Space 73, https://doi.org/10.1186/
25. Rossby, T., Flagg, C. N., Donohue, K., Sanchez-Franks, A. & Lilli- s40623-020-01288-x (2021).
bridge, J. On the long-term stability of Gulf Stream transport based 47. Moat B. I. et al. Atlantic meridional overturning circulation observed
on 20 years of direct measurements. Geophys. Res. Lett. 41, by the RAPID-MOCHA-WBTS (RAPID-Meridional Overturning Cir-
114–120 (2014). culation and Heatflux Array-Western Boundary Time Series) array at
26. Caínzos, V. et al. Thirty years of GO-SHIP and WOCE data: Atlantic 26N from 2004 to 2022 (v2022.1). NERC EDS British Oceanographic
overturning of mass, heat, and freshwater transport. Geophys. Res. Data Centre NOC. https://doi.org/10.5285/04c79ece-3186-349a-
Lett. 49, e2021GL096527 (2022). e063-6c86abc0158c (2023).
27. Fu, Y., Li, F., Karstensen, J. & Wang, C. A stable Atlantic meridional 48. Meinen, C. S. et al. Structure and variability of the Antilles Current at
overturning circulation in a changing North Atlantic Ocean since 26.5°N. J. Geophys. Res. Oceans 124, 3700–3723 (2019).
the 1990s. Sci. Adv. 6, eabc7836 (2020). 49. Sanford, T. B. Motionally-induced electric and magnetic fields in the
28. Worthington, E. L. et al. A 30-year reconstruction of the Atlantic sea. J. Geophys. Res. 76, 3476–3492 (1971).
meridional overturning circulation shows no decline. Ocean Sci. 17, 50. Stommel, H. The theory of the electric field induced in deep ocean
285–299 (2021). currents. J. Mar. Res. 7, 386–392 (1948).
29. Smeed, D. A. et al. Observed decline of the Atlantic meridional 51. Hummon J. M. et al. CODAS+UHDAS Documentation
overturning circulation 2004–2012. Ocean Sci. 10, 29–38 (2014). (282:c289b3782d3a). Zenodo, https://doi.org/10.5281/zenodo.
30. Roberts, C. D., Jackson, L. & McNeall, D. Is the 2004-2012 reduction 8371260 (2023).
of the Atlantic meridional overturning circulation significant? Geo- 52. Domingues, R., Goni, G., Baringer, M. & Volkov, D. What caused the
phys. Res. Lett. 41, 3204–3210 (2014). accelerated sea level changes along the US East Coast during
31. Chen, C., Wang, G., Xie, S.-P. & Liu, W. Why does global warming 2010–2015? Geophys. Res. Lett. 45, 13–367 (2018).
weaken the Gulf Stream but intensify the Kuroshio? J. Clim. 32, 53. Volkov, D. L., Lee, S.-K., Domingues, R., Zhang, H. & Goes, M.
7437–7451 (2019). Interannual sea level variability along the southeastern seaboard of
32. Dong, S., Baringer, M. O. & Goni, G. J. Slow Down of the Gulf Stream the United States in relation to the gyre‐scale heat divergence in the
during 1993–2016. Sci. Rep. 9, 6672 (2019). (2019). North Atlantic. Geophys. Res. Lett. 46, 7481–7490 (2019).
33. Volkov, D. L. Meridional overturning circulation and heat transport 54. Larsen, J. C. Transport measurements from in-service undersea
in the Atlantic Ocean, [in State of the Climate in 2020]. Bull. Am. telephone cables. IEEE J. Ocean. Eng. 16, 313–318 (1991).
Meteor. Soc. 102, S176–S179 (2021). 55. Lowes, F. J. An estimate of the errors of the IGRF/DGRF fields
34. Volkov, D. L. Atlantic meridional overturning circulation and asso- 1945–2000. Earth Planet Sp. 52, 1207–1211 (2000).
ciated heat transport, [In: State of the Climate 2019]. Bull. Am. Met. 56. Lowes F. J. IGRF Health warning, errors, and limitations. IAGA
Soc. 101, S159–S163 (2020). Working Group VMOD, https://www.ncei.noaa.gov/products/
35. Baringer, M. O. Meridional overturning and oceanic heat transport international-geomagnetic-reference-field/health-warning
circulation observations in the North Atlantic Ocean, [in State of the (2022).
Climate in 2017]. Bull. Am. Meteor. Soc. 99, S91–S94 (2018). 57. Meinen C. S., Garcia R. F., & Smith R. H. Evaluating pressure gauges
36. Baringer, M. O. Meridional overturning and oceanic heat transport as a potential future replacement for electromagnetic cable
circulation observations in the North Atlantic Ocean, [in State of the observations of the Florida Current transport at 27°N. J. Oper.
Climate in 2016]. Bull. Am. Meteor. Soc. 98, S84–S87 (2017). Oceanogr., 1–11, https://doi.org/10.1080/1755876X.2020.
37. Baringer, M. O. Meridional overturning circulation observations in 1780757 (2020).
the North Atlantic Ocean, [In: State of the Climate in 2014]. Bull. Am.
Met. Soc. 96, S78–S80 (2016). Acknowledgements
38. Baringer, M. O. et al. Meridional heat transport in the Atlantic We would like to express great appreciation to all who have been
Ocean, [In: State of the Climate in 2013]. Bull. Am. Met. Soc. 95, involved in maintaining the Florida Current observing system at 27°N
S69–S71 (2014). and the RAPID-MOCHA array, including collecting data at sea. We are
39. Piecuch, C. G. & Beal, L. M. Robust weakening of the Gulf Stream grateful to Dr. Kandaga Pujiana for providing feedback on the initial
during the past four decades observed in the Florida Straits. Geo- version of the manuscript.This work was supported by the NOAA’s
phys. Res. Lett. 50, e2023GL105170 (2023). Global Ocean Monitoring and Observing program grant #100007298
40. Pietrafesa, L. J., Bao, S., Gayes, P. T., Carpenter, D. D. & Kowal, J. C. (D.L.V., R.H.S., R.F.G., M.O.B.), NOAA’s Climate Variability and Predict-
Variability and trends of the Florida current and implications for the ability program grant #NA20OAR4310407 (D.L.V.), Natural Environment
future of the Gulf Stream. J. Coast. Res. 38, 1096–1103 (2022). Research Council grants # NE/Y003551/1 and NE/Y005589/1 (B.I.M.,
41. Park, J. & Sweet, W. W., Accelerated sea level rise and Florida D.A.S.), National Science Foundation grants #OCE-1332978 and #OCE-
Current transport. Ocean Sci. 11, 607–615 (2015). 1926008 (WEJ). DLV, RHS, RFG, and MOB received some support from
42. Piecuch, C. G. Likely weakening of the Florida Current during the the WBTS program and from the NOAA Atlantic Oceanographic and
past century revealed by sea-level observations. Nat. Commun. 11, Meteorological Laboratory for this work. D.L.V. and R.F.G. were also
3973 (2020). supported in part under the auspices of the Cooperative Institute for
43. Pujiana, K. et al. Genesis of the Gulf Stream subseasonal variability Marine and Atmospheric Studies (CIMAS), a cooperative institute of the
in the Florida Straits. J. Geophys. Res.: Oceans 128, e2022JC018555 University of Miami and NOAA, cooperative agreement
(2023). NA20OAR4320472.

Nature Communications | (2024)15:7780 11

Article https://doi.org/10.1038/s41467-024-51879-5

Author contributions
Conceptualization: D.L.V., M.O.B.; Methodology: D.L.V., R.H.S., R.F.G., Publisher’s note Springer Nature remains neutral with regard to jur-
D.A.S., B.I.M., W.E.J.; Investigation: D.L.V., R.H.S., R.F.G., D.A.S., B.I.M.; isdictional claims in published maps and institutional affiliations.
Visualization: D.L.V.; Writing—original draft: D.L.V.; Writing—review &
editing: R.H.S., R.F.G., D.A.S., B.I.M., W.E.J., M.O.B. Open Access This article is licensed under a Creative Commons
Attribution-NonCommercial-NoDerivatives 4.0 International License,
Competing interests which permits any non-commercial use, sharing, distribution and
The authors declare no competing interests. reproduction in any medium or format, as long as you give appropriate
credit to the original author(s) and the source, provide a link to the
Additional information Creative Commons licence, and indicate if you modified the licensed
Supplementary information The online version contains material. You do not have permission under this licence to share adapted
supplementary material available at material derived from this article or parts of it. The images or other third
https://doi.org/10.1038/s41467-024-51879-5. party material in this article are included in the article’s Creative
Commons licence, unless indicated otherwise in a credit line to the
Correspondence and requests for materials should be addressed to material. If material is not included in the article’s Creative Commons
Denis L. Volkov. licence and your intended use is not permitted by statutory regulation or
exceeds the permitted use, you will need to obtain permission directly
Peer review information Nature Communications thanks Lequan Chi from the copyright holder. To view a copy of this licence, visit http://
and the other, anonymous, reviewer for their contribution to the peer creativecommons.org/licenses/by-nc-nd/4.0/.
review of this work. A peer review file is available.
© The Author(s) 2024
Reprints and permissions information is available at
http://www.nature.com/reprints

Nature Communications | (2024)15:7780 12