Title: Observed Multi‐Decadal Acceleration of Globally Averaged Abyssal Ocean Warming

URL Source: https://repository.library.noaa.gov/view/noaa/74318/noaa_74318_DS1.pdf

Number of Pages: 10

Markdown Content:
Observed Multi‐Decadal Acceleration of Globally Averaged Abyssal Ocean Warming Gregory C. Johnson 1,2 

> 1

School of Oceanography, University of Washington, Seattle, WA, USA, 2NOAA Pacific Marine Environmental Laboratory, Seattle, WA, USA 

# Abstract Given sparse historical data in the deep and abyssal ocean, previously only multi‐decadal temperature trends have been estimated from observations there on a global scale. Full‐depth CTD sampling started circa 1970, with the first decadal global ship‐based survey occupied in the 1990s, and the first regional pilot array of Deep Argo floats started circa 2016. Here we fit second‐order polynomial functions versus time to all available full‐depth CTD profile temperature data in local spatial bins to estimate changes in the rates of these multi‐decadal temperature trends. We find a statistically significant increase of the heating rate of the abyssal (4,000–6,000 dbar) ocean, from 5.4 ( ±4.9) TW in 1988 to 20.2 ( ±3.9) TW in 2018. In contrast, we find no statistically significant change in the heating rate of the deep ocean, estimated at 29.4 ( ±22.1) TW in 1988 and 25.0 ( ±17.5) TW in 2018. 

# Plain Language Summary Much of the abyssal ocean (deeper than 4,000 m) is filled with water that sinks from the continental shelves around Antarctica after it has been made denser through cooling and concentration of salt as a result of ice formation there. Previous studies have used ocean temperature observations taken from ships at approximately decadal intervals around the globe since the 1980s to document that these abyssal ocean waters have been warming, especially strongly around Antarctica, and thus estimated the ocean deeper than 2,000 m has taken up about 10% of the excess heat energy entering Earth's climate system owing to the human‐caused increase of greenhouse gasses in the atmosphere. Here we analyze ocean temperature profile data from shipboard surveys since 1970 and from regional arrays of autonomous robotic profiling floats called Deep Argo floats since circa 2016 to show that abyssal heat uptake has approximately quadrupled from 1988 to 2018. However, no statistically significant change is detectable in the rate of deep (from 2,000 to 4,000 m) ocean heat uptake. Increasing the number of Deep Argo floats to expand the arrays from regional arrays to a global array would allow improved tracking of changes in deep and abyssal ocean temperatures. 

## 1. Introduction 

Observational analyses suggest that the abyssal (here 4,000 to 6,000 dbar) and deep (here 2,000 to 4,000 dbar) oceans have generally been, when globally averaged, warming over the past several decades (Desbruyeres et al., 2016; Johnson & Purkey, 2024; Kouketsu et al., 2011; Purkey & Johnson, 2010). Multi‐decadal trends estimated in these studies reveal the strongest warming rates in the abyssal ocean around Antarctica, with warming trends in the abyssal ocean ventilated by Antarctic Bottom Water decreasing with distance from that continent. These trends may be driven by freshening from glacial melt on the Antarctic Shelf in the Ross Sea, Adelie Lands, and surrounds (Silvano et al., 2018), decadal wind variability (Li et al., 2023) or open ocean polynya variability (Spence et al., 2025; Zhou et al., 2026) in the Weddell Sea. Cooling trends have generally been found in the regions of the deep and abyssal ocean ventilated by North Atlantic Deep Water, although those have reversed recently in the Western North Atlantic (Chomiak et al., 2025). In global integrals of ocean heat content changes from 2,000 to 6,000 dbar, the warming trends dominate. The amount of heat taken up by the global ocean accounts for ∼90% of the heat energy entering Earth's climate system over the past several decades (von Schuckmann et al., 2023). A previously estimated sum of the global ocean deep and abyssal warming trends from ∼1987 to ∼2015, 34.5 ( ±8.3) TW (Johnson & Purkey, 2024), is 

∼14% of the rate of energy uptake in the climate system from 1971 to 2020 or ∼9% of the rate from 2006 to 2020 (von Schuckmann et al., 2023). Argo data in the upper 2,000 dbar of the global ocean have also revealed an 

## RESEARCH LETTER 

10.1029/2026GL124104 

> Key Points:

• Quadratic fits of temperature versus time from Deep Argo float and historical ship‐based CTD data esti-mate local acceleration rates • Globally integrated abyssal (4,000– 6,000 dbar) warming acceleration is significantly different from zero, but deep (2,000–4,000 dbar) is not • Abyssal warming acceleration integrates globally to an increase in heat uptake rate from 5.4 ( ±4.9) TW               

> ∼1988 to 20.2 ( ±3.9) TW ∼2018
> Correspondence to:
> G. C. Johnson, gcj3@uw.edu
> Citation:
> Johnson, G. C. (2026). Observed multi‐decadal acceleration of globally averaged abyssal ocean warming.
> Geophysical Research Letters ,53 ,e2026GL124104. https://doi.org/10.1029/ 2026GL124104 Received 4 MAY 2026 Accepted 13 JUL 2026
> Author Contributions: Conceptualization: Gregory C. Johnson
> Data curation: Gregory C. Johnson
> Formal analysis: Gregory C. Johnson
> Investigation: Gregory C. Johnson
> Methodology: Gregory C. Johnson
> Project administration: Gregory C. Johnson
> Resources: Gregory C. Johnson
> Software: Gregory C. Johnson
> Supervision: Gregory C. Johnson
> Visualization: Gregory C. Johnson
> Writing – original draft: Gregory C. Johnson
> Writing – review & editing: Gregory C. Johnson © 2026 The Author(s). Geophysical Research Letters published by Wiley Periodicals LLC on behalf of American Geophysical Union. This is an open access article under the terms of the Creative Commons Attribution‐NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

JOHNSON 1 of 10 acceleration of upper ocean warming rates from 2005 through 2020, consistent with an acceleration of heat energy entering Earth's climate system measured at the top of the atmosphere (Loeb et al., 2021). At least one global analysis using repeated shipboard section data investigated the possibility of changes in deep and abyssal ocean temperature trends from decade to decade (Desbruyeres et al., 2016), finding some substantial regional changes post‐2000 versus pre‐2000, but no statistically significant change in the global integrals of the temperature trends, or even the integrals of those trends for each ocean, between the two time periods analyzed. Subsequent regional analyses including more recent data found evidence for acceleration of warming in the abyssal waters of the South Pacific Ocean between the 1990s and the 2010s (Purkey et al., 2019) as well as the abyssal waters of the Bellingshausen and Amundsen Basins even more recently (Johnson et al., 2024). A decade of measurements has accumulated since that previous analysis (Desbruyeres et al., 2016). So here we revisit the attempt to estimate changes in the rates of change of deep and abyssal ocean temperatures, but using more historical shipboard CTD data (not just repeat sections) as well as incorporating substantial amounts of Deep Argo (Zilberman et al., 2023) CTD data. We augment the methodology of Johnson and Purkey (2024), by fitting a quadratic function of time relative to 2005 to local temperature data rather than just a linear function of time, in locations where the data distribution is deemed sufficient to do so. 

## 2. Data and Methods 

We built on the methodology outlined in Johnson and Purkey (2024; hereafter JP24), so here mostly changes or updates to that methodology are mentioned to avoid unnecessary repetition of previously published material. We downloaded Deep Argo CTD data from the Argo Global Data Assembly Centre (Argo, 2026) on 1 January 2026 and high‐resolution ship‐based CTD (conductivity‐temperature‐depth instrument) data from the World Ocean Database (Boyer et al., 2018) for all profiles extending at least to 2,100 m on 1 January 2026. The amount of data available, especially from Deep Argo floats, increased substantially since the download for JP24 19 months prior. We followed JP24 to calculate absolute salinity (SA) and conservative temperature (CT) for all of the CTD profiles using the GSW TEOS‐10 toolbox (McDougall & Barker, 2011) and to linearly interpolate all the profiles to a uniform 10‐dbar pressure grid, discarding values that were interpolated over a pressure span exceeding 100 dbar. Likewise, we sorted the profiles into 167 km × 167 km square bins to calculate trends in CT. Different from JP24, if there were three or more CT values at any pressure level within a bin with the final and initial year separated by at least a decade and at least one data point in the middle 50% of the time range, we fit a quadratic function of time as follows: Di = CTo + ∂CT /∂t · (ti  to ) + 1/2 · ∂2CT /∂t 2 · (ti  to )2, (1)

Solving Equation 1 for constants CTo, the value of CT on 1 January 2005, ∂CT/∂t, the trend in CT with time, and ∂2CT/∂t 2, the rate of change of that trend. Here Di are the data at a given pressure in a given bin, ti are the times those data were collected, and the subscript o signifies 1 January 2005, the global 2,000 to 6,000 dbar volume‐ averaged mean year of data for all quadratic fits after extreme outliers were removed as described below. If the above conditions were not met, but there were two or more CT values with the final and initial year separated by at least a decade on a given pressure level within a given bin, we omitted the rightmost term from Equation 1 and just solved for CTo and ∂CT/∂t. We discarded extreme outliers of either ∂CT/∂t or ∂2CT/∂t 2 using the following procedure: Following Tukey (1977) we computed the first and third quartiles of ∂CT/∂t and ∂ 2CT/∂t 2 (if available) at each pressure level and then subtracted the first from the third for each data set to find the interquartile ranges. We then discarded any values that were less than the first quartile minus six times that range or greater than the third quartile plus six times that range for either of the variables under consideration. Global findings were robust to reducing that multiplier to as little as three, the value suggested by Tukey (1977), with means changing little and statistical uncertainties reduced. However, six was chosen for reasons given below. At pressures between 2000 and 6,000 dbar, this quality control step eliminated ∼3% of the bins for the ∂CT/∂t estimates and ∼4% of the bins for the ∂ 2CT/∂t 2 estimates, with a standard deviation of ±1% for each variable over the 2,000–6,000‐dbar pressure range (Figure 1a). These rather high discard rates for the very expansive range used to screen outliers is indicative of the long tails for the probability distributions of the estimates, likely owing to high geophysical noise levels 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 2 of 10 (e.g., noise from internal waves, internal tides, mesoscale eddies, seasonal cycles, interannual signals, etc.). These discard rates suggest that the calculation pushes the limits of the sparsely sampled data available. An additional benefit of the use of the multiplier of six is that it kept the discards randomly distributed geographically. If the standard multiplier of three were used, the outliers would have been concentrated in certain geographic regions, indicating an undesired removal of regional variations of ∂CT/∂t and ∂ 2CT/∂t 2.

Figure 1. (a) Fraction of bins with outliers removed versus pressure for the temperature trend (∂CT/∂t, blue line) and acceleration (∂ 2CT/∂t 2, red line) estimates. (b) Fraction of the total global ocean volume sampled by the bins in which trend (∂CT/∂t, blue line) and acceleration (∂ 2CT/∂t 2, red line) estimates are calculated and (c) Initial years (left lines), mean years (middle lines), and final years (right lines) of volume‐weighted trends (∂CT/∂t, blue lines) and accelerations (∂ 2CT/∂t 2, red lines). 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 3 of 10 After this quality control step at 2,100 dbar there were 82,656 profiles in 3,564 bins used to estimate trends and 67,166 profiles in 2,227 bins used to estimate accelerations. Those numbers decreased steadily with increasing pressure to 1,039 profiles in 111 bins for trends and 787 profiles in 80 bins for accelerations at 6,000 dbar. The counts for both profiles used and bins sampled increase for this analysis relative to JP24, as expected given the additional data available after 19 months had elapsed. Between 2000 and 6,000 dbar, the percentage of ocean volume represented by the bins with sufficient sampling was ∼29% for ∂CT/∂t estimates and ∼18% for the ∂ 2CT/∂t 2,with a standard deviation of ±1% for each variable over the 2,000–6,000‐dbar pressure range (Figure 1b). Since the area of the ocean decreases with increasing pressure, the fractions of the ocean sampled remained relatively constant with pressure, even though the number of bins sampled decreased with increasing pressure. From 2,000 to 6,000 dbar the volume‐averaged initial, mean, and final years for the data points used in the trend estimates were 1988, 2008, and 2016 (Figure 1c). Those same quantities for the acceleration estimates were 1988, 2005, and 2018. We used the latter times for setting to, discussing, and plotting results, as we were most interested in the new acceleration estimates here. We only evaluate the global statistics from the fits between 1988 and 2018 to be conservative in our analysis, although they are estimated locally using data starting circa 1970 and running through the end of 2025. From 2,000 to 6,000 dbar the standard deviations of all these times for sampled bins ranged from 8 to 9 years, reflecting the considerable spreads in initial, mean, and final times for data used in these estimates in different locations (see JP24, their Fig. 1b). We followed JP24 in converting remaining values into rates of change of ocean heat content (in W), adding an analogous conversion for the acceleration term (in W s  1). Likewise, we used the remaining values of ∂CT/∂t and ∂ 2CT/∂t 2 at each pressure level to interpolate spatially to the bins without direct estimates using the natural neighbor option in the Matlab griddata interpolation algorithm. We estimated uncertainties for ∂CT/∂t and ∂ 2CT/∂t 2 following the methodology of JP24. Likewise, when summing vertically within layers, we assumed that uncertainties were perfectly correlated, and added uncertainties for each pressure level to get the summed uncertainty. However, when using ∂ 2CT/∂t 2 to calculate changes in ∂CT/∂t with time, we assumed that the uncertainties for the estimates of ∂CT/∂t and ∂ 2CT/∂t 2 were independent, likely, given their different spatial patterns discussed below, and propagated those uncertainties using the square root of the sums of the squares of those terms, with the time variable used in computing the acceleration uncertainties relative to 1 January 2005. 

## 3. Results 

A zonally averaged section of the ∂CT/∂t versus latitude and pressure (Figure 2a) shows very large bottom‐ intensified values in the coldest, densest waters around Antarctica, tailing off from the bottom and to the north. The zonally averaged ∂CT/∂t values are quite small in the abyssal Northern Hemisphere. There are also very large values in the deep Greenland‐Iceland‐Norwegian Seas. These results are very similar to the JP24 results (their Figure 3a), albeit slightly noisier. Hence, including an acceleration term in the fit or not does not sub-stantially alter the patterns observed in the trend term. A zonally average section of ∂ 2CT/∂t 2 versus latitude and pressure (Figure 2b) reveals generally positive values in the abyssal (4,000–6,000 dbar) ocean, larger in the Southern Hemisphere than the Northern Hemisphere, but not the same bottom and Antarctic‐intensified pattern seen for ∂CT/∂t. In the deep (2,000–4,000 dbar) ocean zonally averaged ∂ 2CT/∂t 2 values are patchy, and larger in amplitude than in the abyssal ocean. Maps of ∂CT/∂t in the deep and abyssal ocean from this study are very similar to the JP24 maps (their Figure 2), so they are not duplicated, or discussed further here, other than to say that the temperature trends in the deep and abyssal ocean were slightly noisier when including the acceleration term in the fits than when not, even though we used more data than JP24. To examine spatial variations in the ∂ 2CT/∂t 2 term, we converted them to an acceleration term for heat content and summed them vertically over the deep and abyssal layers (Figure 3). These acceleration maps are generally patchier than the trend maps (JP24, their Figure 2), which is not surprising since the former are a second order term in the fit, indicative of an additional time derivative. The acceleration term for heat content in the abyssal layer (Figure 3b) is more positive than negative overall. Largely positive values are found in many basins in the Southern Hemisphere, somewhat downstream of basins in which bottom water is formed: the northern Brazil Basin and the Cape Basin in the South Atlantic Ocean; the Amundsen and Bellinghausen abyssal plains in the South Pacific Ocean; and the Madagascar, Crozet, West Australian, and South Australian Basins in the Indian 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 4 of 10 Ocean. Largely negative values are found in the Weddell Sea, Argentine Basin, southern Brazil Basin, Ross Gyre, western Southwest Pacific Basin, and Australian‐Antarctic Basin. The dominance of positive values is reflected in the zonal averages of ∂ 2CT/∂t 2 (Figure 2b). The acceleration terms for heat content in the deep ocean (Figure 3a) are about twice the magnitude of those in the abyssal ocean, and patchier. The main pattern that stands out in the deep ocean map is positive acceleration in most of the sampled bins in the subpolar North Atlantic Ocean. In other locations, the acceleration terms in the 

Figure 2. Zonally averaged (a) temperature trends (∂CT/∂t, colors, in m°C yr  1) and (b) temperature accelerations (∂ 2CT/∂t 2,colors, in m°C yr  2) plotted versus pressure and latitude with 2005 CT values (black lines, contoured at 0.5°C intervals). 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 5 of 10 deep ocean are variable in sign, and seldom coherent over more than a few sampled bins, likely reflecting the noisiness of the calculation there. Globally averaged values of ∂CT/∂t versus pressure (Figure 4a) reach a minimum of ∼0.3 m°C yr  1 from 2,500 to 3,500 dbar, increasing with decreasing pressure shallower than that range, and exhibiting an abyssal maximum of 

Figure 3. Maps of the depth‐integrated decadal heat content acceleration term (colors, in nW s  1 m 2) for (a) the deep (2,000–4,000 dbar) ocean and (b) the abyssal (4,000–6,000 dbar) ocean with the 2,000 and 4000‐dbar isobars of bottom bathymetry contoured, respectively (black lines). The deep ocean color scale has twice the range of that for the abyssal ocean. The 167 km × 167 km bins with directly measured trends are bordered by light black lines, and the remaining bins have values interpolated by nearest neighbor, following JP24. 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 6 of 10 ∼0.7 m°C yr  1 from 4,200 to 5,200 dbar, again similar to JP24. The 5%–95% uncertainties are about ±0.08 m° C yr  1 at pressures >3,000 dbar, and increase with decreasing pressure <3,000 dbar to about ±0.15 m°C yr  1 by 2,000 dbar. Globally averaged values of ∂ 2CT/∂t 2 versus pressure (Figure 4b) are generally positive in the abyssal ocean, on the order of 0.02 m°C yr  2, but in the deep ocean, while they are a similar magnitude, their sign varies with pressure, tending to be negative in the shallow portions of the range. The 5%–95% uncertainties for ∂ 2CT/∂t 2

Figure 4. Global average values (thick lines) and 5%–95% uncertainty ranges (blue shading) of (a) temperature trend (∂CT/∂t, purple in m°C yr  1) versus pressure (b) temperature acceleration (∂ 2CT/∂t 2, green in m°C yr  2) terms versus pressure, and (c) globally integrated heat uptake rates in deep (2,000–4,000 dbar, orange in TW) and abyssal (4,000–6,000 dbar, blue in TW) layers versus time. 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 7 of 10 are about ±0.01 m°C yr  2 at pressures >5,000 dbar and increase with decreasing pressure to about ±0.03 m° C yr  2 at 2,000 dbar. The globally integrated heat content values for the deep and abyssal layers plotted versus time taking into account the acceleration term (Figure 4c) nominally show opposite time evolution in the deep and abyssal layer, with considerably larger uncertainties in the deep layer than the abyssal layer. The rate of heat being absorbed in the abyssal layer increases from 5.4 ( ±4.9) TW in 1988 to 20.2 ( ±3.9) TW in 2018, with the central 2005 value being 13.8 ( ±1.9) TW. The rate of globally integrated heat uptake nearly quadrupled in three decades in the abyssal layer, with the increase being statistically significant. In contrast, the rate of heat being absorbed in the deep layer nominally decreased slightly from 29.4 ( ±22.1) TW in 1988 to 25.0 ( ±17.5) TW in 2018, with the central 2005 value being 26.9 ( ±7.0) TW. Hence, the globally integrated values of heat uptake in the deep layer have not changed over time within their statistical uncertainties. 

## 4. Discussion 

Analysis of temperature data using all available spatially adjacent historical shipboard deep CTD stations and Deep Argo CTD profiles estimates a globally averaged acceleration of heat uptake in the abyssal ocean, increasing from 5.4 ( ±4.9) TW in 1988 to 20.2 ( ±3.9) TW in 2018 (Figure 4c). This finding contrasts with results from an analysis a decade ago using repeat hydrographic section data only, that found no statistically significant difference in warming rates before and after the year 2000 in the global average for either the deep or abyssal ocean (Desbruyeres et al., 2016). The change in results could be owing to the difference in methods (those used here allow use of more data than those used in the previous study), the additional data collected and made available in the decade elapsed since the previous study, or acceleration of abyssal warming in the past decade or two. Relatively recent regional studies have found evidence for an acceleration of abyssal layer warming in the South Pacific Ocean (Purkey et al., 2019) and in the Bellingshausen and Amundsen basins (Johnson et al., 2024), in support of that last possible reason. Furthermore, a continued acceleration of warming in the abyssal ocean over the coming decades is projected with high‐resolution numerical models of the ice‐ocean system (e.g., Xie et al., 2025), also consistent with that last possible reason. The spatial patterns of this abyssal warming acceleration (Figures 2b and 3b), while noisy, suggest stronger accelerations in basins just downstream of the strongest warming trends (Figure 2a, JP24 their Figure 2b). This pattern is intuitively appealing, with the strongest acceleration of a transient signal occurring at the leading edge of that signal. However, the ocean response to variations in Antarctic Bottom Water formation is complex, with planetary waves (i.e., Kelvin and Rossby waves) relatively quickly (i.e., over decades) carrying the signal far downstream, even to the subpolar North Pacific (Masuda et al., 2010), but with the slower advection of changes also playing substantial roles in some locations, especially deep basins containing and adjacent to bottom water formation regions (Zanowski & Hallberg, 2017). In any event, the strongest accelerations should not necessarily be expected to be co‐located with the strongest trends. A multi‐decadal increase in the warming rate of Antarctic Bottom Water over time implies a substantial change of the abyssal ocean structure with global ramifications (Rintoul et al., 2026). Already observed substantial de-creases in abyssal stratification over the past decades have implications for turbulent mixing and the strength and closure of the global ocean circulation (Tan & Thurnherr, 2023). Deep and abyssal warming already accounts for about 5% of global sea level rise owing to thermal expansion over the past few decades (Kouketsu et al., 2011), so an increase in that contribution over time will have increasing consequences, especially in regions where abyssal and deep ocean warming is strong. When considering the time history of ocean heat uptake below 2,000 dbar, the statistically significant acceleration of abyssal warming we find is partly offset by estimated changes in the global heat uptake rate of the deep ocean and effectively obscured by their larger uncertainty (Figure 4c). The deep ocean warming rate decreases slightly with time when considering the acceleration term, from 29.4 ( ±22.1) TW in 1988 to 25.0 ( ±17.5) TW in 2018. The uncertainties are sufficiently large to render that decrease statistically insignificant. When summed, the warming rates including acceleration terms for the deep and abyssal layers yield 34.7 ( ±22.6) TW in 1988 and 45.2 ( ±17.9 TW) for 2018, which again are not statistically different within the larger uncertainties generated when estimating the acceleration terms. The estimates of the heat uptake rates at the mean year of sampling of the deep and abyssal ocean, 2005, from this study, 13.8 ( ±1.9) TW and 26.9 ( ±7.0) TW in the deep and abyssal layers, respectively, agree within uncertainties with the estimates of long‐term heat uptake rates of 12.9 ( ±1.8) 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 8 of 10 TW and 21.6 ( ±6.5) estimated earlier without including an acceleration term by JP24. The magnitudes of the uncertainties in the estimates presented here are likely related to variability on sub‐decadal time‐scales associated with vertical heave of isopycnals (and isotherms) from internal waves, internal tides, mesoscale eddies, the seasonal cycle, and interannual signals. These noise sources are larger in amplitude in the deep ocean than the abyssal ocean because the vertical thermal stratification is much stronger in the deep ocean than the abyssal ocean. Also, there is a coherent near‐global signal of contraction of bottom waters of Antarctic origin in the abyssal ocean whereas changes in the deep waters are larger, but vary more regionally, even in sign. The fraction of the global deep and abyssal ocean volume sampled sufficiently to estimate multi‐decade trends in temperature is about 29%, and the fraction sampled sufficiently to estimate change in the rates of those trends is about 18% (Figure 1). Given the long tails and spatial patchiness of the acceleration estimates, the small fractions of volumes sampled may reduce confidence in the reported statistical significance of their global integrals. While Deep Argo has greatly increased the number of full‐depth CTD profiles collected in the past decade (Zilberman et al., 2023), the array is still only at about 20% of its design strength. A fully deployed global Deep Argo array would much better sample the deep and abyssal ocean spatially and temporally, greatly reducing uncertainties for studies like these, allowing assessment of year‐to‐year, rather than multi‐decadal, changes in temperature and salinity, and providing more confident estimates in regional patterns of temperature and salinity changes in the deep and abyssal ocean over time. 

## Conflict of Interest 

The author declares no conflicts of interest relevant to this study. 

## Availability Statement 

This study used freely available ship‐based CTD data from the World Ocean Database (Boyer et al., 2018) downloaded from https://www.ncei.noaa.gov/products/world‐ocean‐database on 1 January May 2026, Deep Argo CTD data (Argo, 2026) downloaded from an Argo Global Data Assembly Center on 1 January 2026, and ETOPO1 Bathymetry (Amante & Eakins, 2009; NOAA National Geophysical Data Center, 2009). The freely available Gibbs‐Seawater Oceanographic Toolbox (McDougall & Barker, 2011) software was also used and is available at https://www.teos‐10.org/software.htm. Matlab scripts used to analyze the data and make the figures for this paper are available from Johnson (2026). 

## References 

Amante, C., & Eakins, B. W. (2009). ETOPO1 1 arc‐minute global relief model: Procedures, data sources and analysis . NOAA Technical Memorandum NESDIS NGDC‐24. National Geophysical Data Center, NOAA. Retrieved from https://www.ngdc.noaa.gov/mgg/global/relief/ ETOPO1/docs/ETOPO1.pdf Argo. (2026). Argo float data and metadata from Global Data Assembly Centre (Argo GDAC) [Dataset]. SEANOE . https://doi.org/10.17882/4 2182 Boyer, T. P., Baranova, O. K., Coleman, C., Garcia, H. E., Grodsky, A., Locarnini, et al. (2018). World ocean database 2018. A.V. Mishonov, Technical Ed., NOAA Atlas NESDIS 87 Retrieved from https://www.ncei.noaa.gov/products/world‐ocean‐database Chomiak, L. N., Volkov, D. L., Johns, W. E., Hooper, J. A., & Smith, R. H. (2025). Deep ocean cooling and freshening from Subpolar North Atlantic reaches subtropics at 26.5°N. Communications Earth & Environment , 6(1), 235. https://doi.org/10.1038/s43247‐025‐02170‐y Desbruyeres, D. G., Purkey, S. G., McDonagh, E. L., Johnson, G. C., & King, B. A. (2016). Deep and abyssal ocean warming from 35 years of repeat hydrography. Geophysical Research Letters , 43 (19), 10356–10365. https://doi.org/10.1002/2016gl070413 Johnson, G. C. (2026). Software for “Observed multi‐decadal acceleration of globally averaged abyssal ocean warming”. Zenodo . https://doi.org/1 0.5281/zenodo.19957898 Johnson, G. C., Mahmud, A., Macdonald, A. M., & Twining, B. S. (2024). Antarctic bottom water warming, freshening, and contraction in the eastern Bellingshausen Basin. Geophysical Research Letters , 51 (13), e2024GL109937. https://doi.org/10.1029/2024gl109937 Johnson, G. C., & Purkey, S. G. (2024). Refined estimates of global ocean deep and abyssal decadal warming trends. Geophysical Research Letters , 51 (18), e2024GL111229. https://doi.org/10.1029/2024GL111229 Kouketsu, S., Doi, T., Kawano, T., Masuda, S., Sugiura, N., Sasaki, Y., et al. (2011). Deep ocean heat content changes estimated from observation and reanalysis product and their influence on sea level change. Journal of Geophysical Research , 116 (C3), C03012. https://doi.org/10.1029/20 10jc006464 Li, Q., England, M. H., Hogg, A. M., Rintoul, S. R., & Morrison, A. K. (2023). Abyssal ocean overturning slowdown and warming driven by Antarctic meltwater. Nature , 615 (7954), 841–847. https://doi.org/10.1038/s41586‐023‐05762‐w Loeb, N. G., Johnson, G. C., Thorsen, T. J., Lyman, J. M., Rose, F. G., & Kato, S. (2021). Satellite and ocean data reveal marked increase in Earth's heating rate. Geophysical Research Letters , 48 (13), e2021GL093047. https://doi.org/10.1029/2021gl093047 Masuda, S., Awaji, T., Sugiura, N., Matthews, J. P., Toyoda, T., Kawai, Y., et al. (2010). Simulated rapid warming of abyssal North Pacific waters. 

Science , 329 (5989), 319–322. https://doi.org/10.1126/science.1188703 McDougall, T. J., & Barker, P. M. (2011). Getting started with TEOS‐10 and the Gibbs seawater (GSW) oceanographic toolbox (p. 28). SCOR/ IAPSO WG127. 

Acknowledgments 

We thank all those who contributed to the collection, processing, quality control, and calibration of the ship‐based and Deep Argo CTD data used in this analysis, and the Fund for Science and Technology for their Deep Argo support. These data were collected and made freely available by the International Argo Program and the national programs that contribute to it (https://argo.ucsd.edu, https://www.ocean‐ ops.org). The Argo Program is part of the Global Ocean Observing System. GCJ was supported by NOAA Global Ocean Monitoring and Observation Program and NOAA Research. Comments by John Lyman, Sarah Purkey, and an anonymous reviewer improved this work. PMEL Contribution Number 5869. 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 9 of 10 NOAA National Geophysical Data Center. (2009). ETOPO1 1 arc‐minute global relief model [Dataset]. NOAA National Centers for Environ-mental Information . https://doi.org/10.7289/V5C8276M Purkey, S. G., & Johnson, G. C. (2010). Warming of global abyssal and deep southern ocean waters between the 1990s and 2000s: Contributions to global heat and sea level rise budgets. Journal of Climate , 23 (23), 6336–6351. https://doi.org/10.1175/2010jcli3682.1 Purkey, S. G., Johnson, G. C., Talley, L. D., Sloyan, B. M., Wijffels, S. E., Smethie, W., et al. (2019). Unabated bottom water warming and freshening in the South Pacific Ocean. Journal of Geophysical Research‐Oceans , 124 (3), 1778–1794. https://doi.org/10.1029/2018jc014775 Rintoul, S. R., Stewart, A. L., Johnson, G. C., Zhou, S. J., Foppert, A., Li, Q., et al. (2026). Antarctic bottom water in a changing climate. Nature Reviews Earth & Environment , 7(2), 86–102. https://doi.org/10.1038/s43017‐025‐00750‐2 Silvano, A., Rintoul, S. R., Peña‐Molino, B., Hobbs, W. R., van Wijk, E., Aoki, S., et al. (2018). Freshening by glacial meltwater enhances melting of ice shelves and reduces formation of Antarctic Bottom Water. Science Advances , 4(4), eaap9467. https://doi.org/10.1126/sciadv.aap9467 Spence, P., Menviel, L., Morrison, A. K., Foppert, A., & Silvano, A. (2025). On the relationship between the Weddell Polynya and Antarctic bottom water trends. Journal of Climate , 38 (24), 7249–7268. https://doi.org/10.1175/JCLI‐D‐24‐0179.1 Tan, S., & Thurnherr, A. M. (2023). On the global decrease in the deep and abyssal density stratification along the spreading pathways of Antarctic Bottom Water since the 1990s. Geophysical Research Letters , 50 (11), e2022GL102422. https://doi.org/10.1029/2022GL102422 Tukey, J. W. (1977). Exploratory data analysis . Addison‐Wesley. von Schuckmann, K., Minière, A., Gues, F., Cuesta‐Valero, F. J., Kirchengast, G., Adusumilli, S., et al. (2023). Heat stored in the Earth system 1960‐2020: Where does the energy go? Earth System Science Data , 15 (4), 1675–1709. https://doi.org/10.5194/essd‐15‐1675‐2023 Xie, C., Zhang, Z., Chen, Y., & Wang, C. (2025). Substantial contraction of Dense Shelf Water in the Ross Sea under future climate scenarios. 

Geophysical Research Letters , 52 (6), e2024GL112581. https://doi.org/10.1029/2024GL112581 Zanowski, H., & Hallberg, R. (2017). Weddell Polynya transport mechanisms in the abyssal ocean. Journal of Physical Oceanography , 47 (12), 2907–2925. https://doi.org/10.1175/jpo‐d‐17‐0091.1 Zhou, L., Ayres, H., Gülk, B., Narayanan, A., De lavergne, C., Ödalen, M., et al. (2026). Review article: Weddell Sea Polynya formation, cessation and climatic impacts. The Cryosphere , 20 (1), 285–308. https://doi.org/10.5194/tc‐20‐285‐2026 Zilberman, N. V., Thierry, V., King, B., Alford, M., Andre, X., Balem, K., et al. (2023). Observing the full ocean volume using Deep Argo floats. 

Frontiers in Marine Science , 10 , 1287867. https://doi.org/10.3389/fmars.2023.1287867 

# Geophysical Research Letters 10.1029/2026GL124104 JOHNSON 10 of 10
