Paleocean energy atlas
Reconstructed from Gebbie’s 2021 review, mean ocean temperatures, the Common Era inversion, and the actual Zanna / NOAA / Cheng files. The 12,000–20,000 ZJ deglacial number is real. So is the 1,500 ZJ Medieval peak versus the Little Ice Age. So is 427 ppm. Dotted terms open a textbook note — hover or tap.
The comparison
Deglaciation
12,000–20,000ZJ
Last Glacial Maximum → Holocene. Best estimate from noble-gas MOT: 14,084 ZJ.
MCA, ~1000 years ago
1,500ZJ vs LIA
Gebbie inversion: MCA stored 1,000 ZJ more than year 2000, so ~1,500 ZJ above the LIA trough.
Modern warming era
500ZJ
Gebbie 2021 industrial-era total to ~2017. Zanna 1870–2018 full-depth: 464 ZJ. One-third of the climb back to MCA.
Each square is 500 — the entire modern-era uptake in Gebbie’s ledger. Deglaciation is 28 of those squares (range 24–40). MCA is 3. Today is 1.
Deglaciation ~14,000 ZJ
28 × 500 ZJ
MCA vs LIA 1,500 ZJ
3 × 500 ZJ
Modern era 500 ZJ
1 × 500 ZJ
“Deglacial ocean uptake estimates vary from 12,000 to 20,000 ZJ and represent an awesome amount of energy that dwarfs the excess energy stored during the modern warming era by a factor of at least 20.”
Geoffrey Gebbie, Annu. Rev. Mar. Sci. 2021. doi:10.1146/annurev-marine-010419-010844
Where the joules actually went
Same units, same scale. Melting the ice sheets took as much energy as warming the ocean. MCA and the industrial era are a rounding error on that plot — as inventories. Recreated from the Gebbie 2021 numbers the thread quoted, plus Baggenstos 2019 ice-melt energy.
12,000 / 500 = 24×. 20,000 / 500 = 40×. The thread’s “20–30 times” is the low-to-mid envelope of those two ratios. Ice-melt energy is not in the tweet; it is in the papers.
21,000 – 8,000 years ago
relative to the . Age axis is thousands of years before 1950, reversed so the ice age is on the left.
Mean ocean temperature vs present
500-year heat uptake rate (mW/m² of Earth)
Gebbie’s published peak is 600–700 mW/m²; deglacial mean 99 ± 9 mW/m². Modern is 500–700 mW/m² — similar rate, two orders of magnitude less time, so two orders less inventory.
Sea level, same millennia — the other half of the energy bill
Schematic 130 m rise locked to the deglacial chronology. Baggenstos et al. (2019): to melt that ice is ~15,800 ZJ — the same order as the 14,084 ZJ that went into the ocean. Modern excess heat is ~90% oceanic because those ice sheets are already gone.
The last 2,000 years
Spline through the published anchors — not the 1.7 GB 3-D OPT-0015 field. MCA peak 1,500 ZJ vs LIA; LIA trough 0; year 2000 500 ZJ. The 2010–2017 uptick is the industrial continuation, still well below MCA on this inversion.
50-year heat uptake rate
MCA–LIA loss 90 ± 40 mW/m² (1250–1750). Sign change at the end of the LIA is the modern era beginning to win against a still-cooling abyss.
Law Dome CO₂, same centuries
Preindustrial CO₂ is quiet (~277–285 ) across the MCA and LIA. The ocean-heat swing of 1,500 ZJ happened with almost no ice-core CO₂ trend. reconstructions of the same centuries are not that flat — see Library.
1870 – 2025
Zanna 1870–2018 full
464 ZJ
NOAA 0–700 m 1955→2025
-32 → 228
NOAA 0–2000 m pentad
-93 → 295
IAP 2025 vs 1981–2010
317 ZJ
Cheng IAP, 2021–2025 (0–2000 m vs 1981–2010)
Cheng et al. (2026): 0–2000 m has gained heat at 11.4 ± 1.0 ZJ/yr since 2007. That is ~0.7 W/m² — a modern rate in the same neighborhood as the peak, sustained for two decades so far, not seven centuries, and not ten millennia. Inventory is still small against deglaciation. Rate is not small against the . Different baseline than Gebbie’s 500 ZJ.
The distinction the thread blurs
Inventory
Deglaciation stored ~14,000 ZJ because it ran for ~10,000 years at ~0.1 W/m². Melting 130 m of sea-level equivalent ice took another ~15,800 ZJ of (Baggenstos 2019) — the cryosphere was an energy sink of the same order as the ocean. Modern excess heat is ~90% in the ocean because the ice sheets are already gone.
Rate
Modern ocean heat uptake 0.5–0.7 W/m² (Gebbie) to ~0.7–0.9 W/m² in the years. Deglacial mean 0.099 W/m². MCA–LIA loss 0.09 W/m². The industrial ocean is taking heat five to nine times faster than the deglacial average, and comparable to the YD spike, without an ice-sheet collapse to pay for it.
Equivalent planetary energy imbalance, Gebbie 2021 numbers. Modern bar uses the 0.5–0.7 W/m² band midpoint. Sign of MCA–LIA is ocean heat loss.
How the numbers are made
Data in. Kr/N₂, Xe/N₂ and Xe/Kr in air trapped in Antarctic ice (WAIS Divide, EPICA Dome C). Solubility of Xe and Kr in seawater is strongly temperature-dependent; N₂ less so.
How it works. When the global ocean cools, more Xe and Kr dissolve, so their atmospheric mixing ratios fall. A two-box ocean–atmosphere model inverts the measured elemental ratios for a single number: the volume-weighted mean ocean temperature. Three ratios are independent thermometers and are combined into a best-estimate record. Corrections: gravitational settling in the firn, thermal fractionation, and Xe/Kr undersaturation in the modern ocean.
What it produces. One global MOT time series, not a map. Bereiter 2018: +2.57 ± 0.24 °C from LGM to early Holocene. Converted here at 5,480 ZJ per °C → ~14,100 ZJ. Haeberli 2021 extends snapshots across 700,000 years: glacials ~3.3 °C colder.
Limits. No spatial pattern; saturation-state uncertainty is the largest error term. Holocene mean sits ~0.36 °C below ‘present’ in the original scale — more than industrial MOT warming — so the absolute Holocene offset should not be over-read. Age is the ice-core gas age, not a calendar of the deep Pacific.
What the records actually repeat
CO₂: Bereiter et al. 2015 Antarctic composite (NOAA). Antarctic temperature: Jouzel et al. 2007 EPICA Dome C. The 100-kyr beat is the subject of the papers, not a suppressed finding.
July insolation at 65°N — 800,000 years
Schematic Berger-like 21/41/100-kyr envelope, not a full Laskar integration. Today is a modest interglacial, not a deglacial summer. Orbital forcing is not walking the system toward another 14,000 ZJ pulse. The cycle is real; 427 ppm sits off the ice-core envelope either way.
~100,000 years (with 41-kyr and 21-kyr beats)
Glacial–interglacial (100 kyr)
12,000–20,000 ZJ per deglaciation; reverse on glaciation
The dominant ocean-heat cycle in the data. Ice sheets + CO₂ + Southern Ocean ventilation. Not missed — it is the subject of Bereiter, Haeberli, Baggenstos, and Gebbie.
Millennial, irregular; YD ~12,900–11,700 yr BP
Dansgaard–Oeschger / Heinrich / Younger Dryas
Early YD MOT pulse: 600–700 mW/m² for centuries (Gebbie 2021), comparable to or above the modern planetary imbalance, but not sustained.
AMOC disruptions redistribute heat. The YD ocean warming is ‘enigmatic’ in Bereiter — faster than modern MOT warming for ~700 years. That is a rate story, not a 14,000 ZJ inventory.
~11,000–4,000 yr BP peak, then slow cooling into the LIA
Holocene thermal maximum → neoglacial
Rosenthal-type intermediate waters warmer in the mid-Holocene; Common Era inversion then loses ~1,000–1,500 ZJ into the LIA.
A Holocene cooling into the Little Ice Age is in the papers. It is a millennial drift, not a demonstration that 2025 OHC is ‘just the cycle coming back’ — the surface and upper ocean have already left the LIA.
MCA ~950–1250 CE; LIA ~1300–1850 CE
Medieval Climate Anomaly → Little Ice Age
MCA +1,000 ZJ vs year 2000; LIA trough ~−500 ZJ vs year 2000; 1,500 ZJ MCA vs LIA. Loss rate 90 ± 40 mW/m² (1250–1750).
This is the number in the thread. It comes from Gebbie & Huybers 2019, not from a suppressed literature. Deep Pacific cooling into the 20th century is the mechanism: the abyss is still remembering the LIA while the upper ocean records modern warming.
Proposed ~1,500 yr Holocene rhythm (Bond et al. 1997)
Bond / 1,500-year drift-ice cycle
Not quantified as a global ZJ integral. North Atlantic IRD pulses; global coherence is disputed.
Often invoked as the ‘cycle they won’t mention.’ PAGES 2k and later Holocene syntheses do not recover a clean global 1,500-year sinusoid in temperature. Using it to cancel a 0.5–0.9 W/m² planetary imbalance is not supported by the OHC data here.
11 years; ~80–90 years; Maunder-type minima
Solar (11-yr / Gleissberg / grand minima)
TSI Holocene amplitude is ~1 W/m² at the top of the atmosphere, ~0.2 W/m² globally after albedo — far below deglacial or modern OHC rates when averaged.
The LIA coincides with several grand minima, and Gebbie’s inversion does not need a CO₂ explanation for MCA–LIA ocean cooling. That does not make the Maunder Minimum the 21st-century energy source: satellite TSI is not in a grand maximum that can supply 500 ZJ in 150 years.
in the never exceeds ~300 in 800,000 years. 2025 is 427 ppm. That ice-core ceiling is still the best well-mixed global mean, but it is a heavily smoothed one: and bubble close-off low-pass the atmosphere over decades to centuries. reconstructions — limited, noisier, and locally biased — remain valuable for that reason. They record Holocene and Common Era swings tens of ppm above the ice envelope, which is what a filtered archive would hide. The ice core is not the unfiltered it has been sold as; last week’s Stomata vs Ice comparison is the side-by-side. today is not walking the system toward another 14,000 ZJ deglaciation. The honest reading of Gebbie is narrower and more interesting: the deep ocean still carries the , heat content was higher than year 2000, and public charts that start in 1955 hide both facts.
Downloads and archives
This ledger
Spreadsheets and the JSON the charts actually plot. Ready to open; no login.
Stomata vs ice
Last week’s comparison, and the files behind it. Ice is smoothed; stomata are limited and still the check.
Original archives
NOAA, Zenodo, IAP — the unmodified sources this atlas was compiled from.
Key papers
Open-access where it exists; otherwise the DOI page.
Hover dictionary
Textbook
A zettajoule is 10²¹ joules. It is the natural unit for planetary heat inventories: the global ocean’s heat capacity is so large that a 1 °C change in mean ocean temperature is several thousand ZJ.
In this ledger
This ledger uses 5,480 ZJ per °C of mean ocean temperature. Deglaciation is ~14,000 ZJ. Gebbie’s industrial-era ocean, to ~2017, is ~500 ZJ — one square on the tile chart.
Textbook
The thermal energy stored in seawater, usually written as an anomaly relative to a chosen baseline: OHC = ρ Cp ∫ ΔT dV. Because water’s heat capacity is high, the ocean holds more than 90% of Earth’s recent energy imbalance.
In this ledger
Every chart here is an OHC anomaly with the zero named in the caption. Zanna zeros 1870. NOAA zeros its own climatology. Gebbie’s Common Era zeros year 2000 or the Little Ice Age trough. Mixing those zeros is how public charts hide the Medieval peak.
Textbook
The volume-weighted average temperature of the global ocean. Unlike sea-surface temperature, MOT is dominated by the deep ocean and therefore changes slowly.
In this ledger
Bereiter et al. (2018) invert krypton and xenon in the WAIS Divide ice core for MOT. Last Glacial Maximum to early Holocene: +2.57 ± 0.24 °C → 14,084 ZJ at 5,480 ZJ/°C.
Textbook
The collapse of continental ice sheets at the end of a glacial period. The last deglaciation ran roughly 21,000–8,000 years ago and raised global sea level by about 130 metres.
In this ledger
It is an ice-sheet-scale energy event: 12,000–20,000 ZJ into the ocean, plus ~15,800 ZJ of latent heat to melt the ice. That is twenty to forty times Gebbie’s modern-era ocean inventory.
Textbook
The most recent interval of maximum ice-sheet volume, about 26,000–19,000 years ago. Sea level was ~120–130 m lower; the deep ocean was several degrees colder and saltier.
In this ledger
The deglacial OHC curve is plotted relative to LGM. Noble-gas MOT puts the LGM ocean 2.57 °C colder than the early Holocene; Haeberli’s 700-kyr snapshots put peak glacials ~3.3 °C colder.
Textbook
The current interglacial epoch, beginning ~11,700 years ago after the Younger Dryas. Ice sheets had mostly retreated; CO₂ sat near 260–280 ppm in Antarctic ice until the industrial rise.
In this ledger
A Holocene thermal maximum, then a slow cooling into the Little Ice Age, is in the Mg/Ca and Common Era inversion literature. That millennial drift is real. It is not a 14,000 ZJ deglaciation running in reverse.
Textbook
A multi-century interval of relatively warm, regionally uneven climate, commonly placed around 950–1250 CE. It is reconstructed from tree rings, corals, sediments, and boreholes — not from a single global thermometer.
In this ledger
Gebbie & Huybers (2019) invert those surface proxies through a Green’s-function ocean and find the MCA ocean held 1,000 ZJ more than the year-2000 ocean, or 1,500 ZJ above the Little Ice Age trough.
Textbook
A cooler interval of the late Holocene, roughly 1300–1850 CE, marked by glacier advances in both hemispheres. Forcing candidates include volcanic clusters and solar grand minima; CO₂ was nearly flat.
In this ledger
The inversion says the LIA trough sits ~500 ZJ below year 2000. Deep Pacific waters were still cooling from that transition into the 20th century — the abyss is slower than the charts that start in 1955.
Textbook
An abrupt millennial cold reversal in the North Atlantic, ~12,900–11,700 years before 1950, usually attributed to a collapse or diversion of the Atlantic overturning circulation.
In this ledger
Bereiter’s MOT shows an enigmatic early-YD pulse: 600–700 mW/m² for centuries — a modern-sized rate, not a deglacial-sized inventory. Rate and pile of joules are different questions.
Textbook
A global array of autonomous profiling floats that measure temperature and salinity in the upper 2,000 m every 10 days. Near-global coverage began around 2005 and is the backbone of modern ocean-heat estimates.
In this ledger
Cheng et al. (2026): 0–2000 m has gained heat at 11.4 ± 1.0 ZJ/yr since 2007. That is a modern rate in the Younger Dryas neighborhood, sustained for two decades, not seven centuries.
Textbook
Air trapped as bubbles in glacial ice. Unlike most paleoclimate proxies, this is a direct sample of ancient atmosphere. Firn diffusion and gradual bubble close-off mix that air over years to centuries, so the record is a low-pass filter of the true mixing ratio.
In this ledger
The Antarctic composite never exceeds ~300 ppm in 800,000 years. That ceiling is a real measurement of well-mixed air — and a heavily smoothed one. High-accumulation Law Dome resolves about a decade; Dome C and Vostok smear centuries.
Textbook
Land plants reduce the density or index of stomata (leaf pores) as atmospheric CO₂ rises. Fossil or herbarium leaves can therefore be calibrated to CO₂. The proxy is local, species-dependent, and noisy, but it is not mixed through Antarctic firn, so it can retain decadal-to-centennial swings that ice cores average away.
In this ledger
Kouwenberg, van Hoof, Wagner and Finsinger reconstructions run tens of ppm above the ice envelope across the Holocene and Common Era — limited, still valuable. Last week’s Stomata vs Ice comparison is the side-by-side: ice is not the unfiltered proxy it has been sold as.
Textbook
Snow on an ice sheet becomes firn, then ice. Gases diffuse through the firn for decades to millennia until bubbles close. The enclosed air is therefore a weighted average of many years, not a single season. Gas-age width scales with accumulation: Law Dome ~8 years FWHM; WAIS ~19; Dome C centuries.
In this ledger
A 50-year 400 ppm spike in the true atmosphere would be strongly attenuated in deep Antarctic cores. Stomatal series still show Holocene and Common Era swings the ice never records. That is the filtering, not a proof that ice CO₂ is fabricated.
Textbook
A mixing ratio: micromoles of CO₂ per mole of dry air. Pre-industrial ice-core CO₂ is ~277–285 ppm. The 2025 Mauna Loa annual mean is 427.35 ppm.
In this ledger
Ice-core CO₂ stays inside ~180–300 ppm for 800,000 years. Stomatal reconstructions of the same millennia often sit 30–40 ppm higher on average, with much larger amplitude. Both can be true if one archive is low-pass filtered.
Textbook
NOAA’s baseline CO₂ observatory on Hawai‘i, started by Charles David Keeling in 1958. It measures the well-mixed marine boundary layer, not a city plume.
In this ledger
2025 annual mean: 427.35 ppm. The ice-core composite never reaches that in 800 kyr of trapped air. The instrumental record is not a proxy and is not smoothed by firn.
Textbook
An indirect recorder of a past climate variable: tree rings, foraminifera, leaf stomata, ice-core isotopes. Ice-core CO₂ is unusually direct (trapped air) but is still filtered by firn. Stomata, Mg/Ca and noble gases are classical proxies with calibrations and error terms.
In this ledger
The ledger never pretends a single thermometer sat in the abyss for 20,000 years. Methods disagree on the third significant figure and agree on the first: deglacial heat is 10⁴ ZJ, Common Era 10³, industrial 10²–10³.
Textbook
Xenon and krypton dissolve more readily in colder seawater. When the global ocean cools, atmospheric Xe and Kr fall. Their ratios to nitrogen in ice-core air can be inverted for a single number: mean ocean temperature.
In this ledger
Three independent ratios (Kr/N₂, Xe/N₂, Xe/Kr) are combined. Largest error is saturation-state uncertainty. No map — one global mean. That is why 12,000–20,000 ZJ is a range once Mg/Ca and pore fluids are folded in.
Textbook
A linear method that treats the ocean as an advective–diffusive filter. G(x, τ) is the fraction of water at interior point x that left the surface τ years ago. Interior temperature is past sea-surface temperature convolved with G.
In this ledger
Zanna 2019 uses this for 1870–2018 from SST alone (464 ZJ full-depth). Gebbie & Huybers 2019 invert Common Era proxies plus HMS Challenger, so the deep Pacific is allowed to still be cooling from the Little Ice Age.
Textbook
The 1872–1876 oceanographic circumnavigation that took the first systematic deep-ocean temperature casts. Comparing Challenger with modern hydrography is one of the few direct checks on a century of deep change.
In this ledger
Gebbie & Huybers use Challenger versus WOCE/Argo as a constraint. The deep Pacific cooled — consistent with a still-draining Little Ice Age, not with a well-mixed ocean that started warming in 1955.
Textbook
The Atlantic’s conveyor: warm water north, deep cold water south. Disruptions (Heinrich events, Younger Dryas) redistribute heat on millennial timescales without requiring a global 14,000 ZJ pulse.
In this ledger
If AMOC or Southern Ocean ventilation changed, the Green’s-function lags are wrong. That is the leading caveat on the MCA–LIA deep-ocean result — and why Challenger-vs-modern cooling is the corroboration, not a 1,000-year thermometer.
Textbook
Slow changes in Earth’s eccentricity (~100 kyr), obliquity (~41 kyr) and precession (~21 kyr) that alter the seasonal and latitudinal distribution of sunlight. Summer insolation at 65°N is the classic pacemaker of ice sheets.
In this ledger
Orbital forcing today is a modest interglacial, not a walk toward another 14,000 ZJ deglaciation. Insolation did not produce 427 ppm. The 100-kyr CO₂ cycle in ice (180–280 ppm) is real; the present greenhouse sits off that envelope.
Textbook
Incoming solar radiation at the top of the atmosphere, in W/m². Paleoclimate usually quotes a seasonal, latitudinal slice (e.g. June at 65°N), not the global annual mean of ~340 W/m².
In this ledger
The 65°N July curve in this atlas is a Berger-like harmonic envelope, not a full Laskar integration. It is shown so the 21/41/100-kyr beats are visible; it is not overplotted on OHC (wrong units).
Textbook
Net heat uptake of the Earth system, in W/m² of Earth’s surface. Positive means the planet is gaining energy. Most of that energy goes into the ocean.
In this ledger
Deglacial mean 99 ± 9 mW/m². Younger Dryas peak 600–700. MCA–LIA loss 90 ± 40. Modern ocean uptake 0.5–0.7 W/m² in Gebbie, ~0.7 W/m² in the Argo years. Similar modern and YD rates; two orders of magnitude less time, so two orders less inventory.
Textbook
The magnesium-to-calcium ratio in the calcite shells of bottom-dwelling foraminifera increases with temperature. Combined with δ¹⁸O (temperature plus ice volume) and pore-fluid chlorinity, it separates deep-ocean warming from ice-sheet melt.
In this ledger
The warm end of Gebbie’s 12,000–20,000 ZJ deglacial envelope comes from these sediment reconstructions. Spatial coverage is a handful of drill sites, not a global integral — hence the range, not a point.
Textbook
A high-accumulation coastal Antarctic ice dome. Fast snowfall closes bubbles in ~8 years, so Law Dome is the least-smoothed ice-core CO₂ record of the last two millennia and overlaps the Cape Grim instrumental series.
In this ledger
Plotted against the Common Era OHC inversion. Pre-industrial CO₂ is quiet (~277–285 ppm) across the MCA and LIA. The 1,500 ZJ ocean-heat swing happened with almost no ice-core CO₂ trend.
Textbook
West Antarctic Ice Sheet Divide ice core. High enough accumulation for a relatively tight gas-age distribution, deep enough for a 68-kyr climate record, and the site of the noble-gas MOT measurements.
In this ledger
Bereiter 2018’s Kr and Xe ratios come from here. It is also one of the ice CO₂ series in the Stomata vs Ice comparison (Bauska/Marcott/Ahn compilations).
Textbook
A community reconstruction of surface temperature over the last 2,000 years from geographically spread proxies. MCA and LIA appear as real, spatially messy features, smaller at the surface than the 20th-century rise.
In this ledger
Surface PAGES 2k and the ocean inversion can disagree on amplitude because the deep field lags. A warmer MCA ocean with a still-cooling 20th-century abyss is allowed.
Textbook
A proposed Holocene rhythm in North Atlantic ice-rafted debris (Bond et al. 1997), often generalized to a global 1,500-year climate oscillator.
In this ledger
Not quantified as a global ZJ integral. PAGES 2k and later Holocene syntheses do not recover a clean global sinusoid. Using it to cancel a 0.5–0.9 W/m² planetary imbalance is not supported by the OHC data here.
Textbook
Inventory is the time-integral of a flux: joules stored. Rate is the flux itself: watts, or W/m². A small rate held for millennia builds a large pile; a large rate held for decades may not.
In this ledger
Deglaciation stored ~14,000 ZJ because it ran ~10,000 years at ~0.1 W/m². The industrial ocean takes heat five to nine times faster than that average, without an ice-sheet collapse to pay for it. The thread that quotes only the pile blurs this.
Textbook
Energy absorbed when ice melts without changing temperature, ~3.34 × 10⁵ J/kg. Melting a 130 m sea-level-equivalent ice sheet therefore consumes a planetary-scale energy budget of its own.
In this ledger
Baggenstos 2019: ~15,800 ZJ to melt the last deglaciation’s ice — the same order as ocean heat uptake. Modern excess heat is ~90% in the ocean partly because those ice sheets are already gone.
Textbook
A probe dropped from ships that records a temperature profile as it falls. Fall-rate errors biased 20th-century subsurface temperature until they were corrected; different groups still use different corrections.
In this ledger
Pre-Argo OHC disagreements (NOAA vs IAP vs Ishii) are partly this. None of those series is Gebbie’s 500 ZJ industrial total, which includes the deep field the XBTs never reached.
Textbook
A well-mixed greenhouse gas. Ice-core air gives its glacial–interglacial envelope (roughly 180–280 ppm). The instrumental atmosphere is now above 420 ppm.
In this ledger
Ice never exceeds ~300 ppm in 800 kyr. Stomata say the Holocene was not that flat. Orbital forcing is not walking the system toward another deglaciation. 427 ppm is off the ice envelope either way.
Bibliography
Gebbie, G. (2021). Combining Modern and Paleoceanographic Perspectives on Ocean Heat Uptake Annual Review of Marine Science 13:255–281 · 10.1146/annurev-marine-010419-010844
The review quoted in the thread. Deglacial uptake 12,000–20,000 ZJ; MCA +1,000 ZJ vs year 2000; modern-era ~500 ZJ; rates in mW/m² across timescales.
Bereiter, B., Shackleton, S., Baggenstos, D., Kawamura, K., Severinghaus, J. (2018). Mean global ocean temperatures during the last glacial transition Nature 553:39–44 · 10.1038/nature25152
Noble gases (Kr, Xe, N₂) in WAIS Divide ice. LGM–Holocene mean ocean temperature +2.57 ± 0.24 °C. Early Younger Dryas 700-year pulse faster than modern MOT warming.
Supplementary noble-gas tables; NOAA paleoclimate archive.
Gebbie, G., Huybers, P. (2019). The Little Ice Age and 20th-century deep Pacific cooling Science 363:70–74 · 10.1126/science.aar8413
Common Era ocean inversion (OPT-0015). Deep Pacific still cooling from the MCA–LIA transition. Deep heat loss since 1750 offsets ~1/4 of upper-ocean gain. Upper 2000 m 1875–2005: 330 ZJ.
https://www.ncei.noaa.gov/pub/data/paleo/gcmoutput/gebbie2019/
Zanna, L., Khatiwala, S., Gregory, J.M., Ison, J., Heimbach, P. (2019). Global reconstruction of historical ocean heat storage and transport PNAS 116:1126–1131 · 10.1073/pnas.1808838115
Green’s-function reconstruction 1870–2018 from SST. Full-depth uptake 464 ZJ by 2018 (this download). Method used in Gebbie 2021 Fig. 5.
Cheng, L. et al. (2017). Improved estimates of ocean heat content from 1960 to 2015 Science Advances 3:e1601545 · 10.1126/sciadv.1601545
IAP objective analysis of subsurface temperature. One of the modern curves in Gebbie Fig. 5.
Cheng, L. et al. (2024). IAPv4 ocean temperature and ocean heat content gridded dataset Earth System Science Data 16:3517–3542 · 10.5194/essd-16-3517-2024
Updated 0–2000 m trend 1955–2023: 6.4 ± 0.3 ZJ/yr. Deep 2000–6000 m since ~1993: 2.0 ± 0.3 ZJ/yr.
Cheng, L. et al. (2026). Ocean heat content sets another record in 2025 Advances in Atmospheric Sciences
IAP/CAS 0–2000 m relative to 1981–2010: 2025 = 317 ZJ. 2007–2025 rate 11.4 ± 1.0 ZJ/yr. Different baseline than Gebbie’s 500 ZJ industrial total.
Levitus, S., Boyer, T.P., Reagan, J. et al. (NCEI) (2026). Global ocean heat and salt content — yearly and pentadal fields NOAA NCEI World Ocean Database analysis
In-situ mapped OHC. Files in 10²² J, converted ×10 to ZJ. Pentadal 0–2000 m 1957–2023: −93 to +295 ZJ vs NOAA climatology.
https://www.ncei.noaa.gov/data/oceans/woa/DATA_ANALYSIS/3M_HEAT_CONTENT/DATA/basin/
Baggenstos, D. et al. (2019). Earth’s radiative imbalance from the Last Glacial Maximum to the present PNAS · 10.1073/pnas.1905447116
Energy budget LGM to present from ice-core noble gases and ice-sheet melt. Ice-melt latent heat ~15,800 ZJ for ~130 m sea-level rise — comparable to ocean heat uptake itself.
Shackleton, S. et al. (2020). Global ocean heat content in the Last Interglacial Nature Geoscience · 10.1038/s41561-019-0498-0
Noble-gas MOT during MIS 5e. Last Interglacial ocean heat comparable to or above Holocene.
Haeberli, M. et al. (2021). Snapshots of mean ocean temperature over the last 700,000 years using noble gases in the EPICA Dome C ice core Climate of the Past 17:843–867 · 10.5194/cp-17-843-2021
Peak glacial MOT consistently ~3.3 ± 0.4 °C cooler than Holocene. Pre–Mid-Brunhes ‘lukewarm’ interglacials ~1.6 °C cooler. Glacial–interglacial ocean-heat cycle is the repeating structure.
Bereiter, B. et al. (2015). Revision of the EPICA Dome C CO₂ record from 800 to 600 kyr BP Geophysical Research Letters · 10.1002/2014GL061957
Antarctic composite ice-core CO₂, 800 kyr. Plotted here from NOAA study 17975.
https://www.ncei.noaa.gov/pub/data/paleo/icecore/antarctica/antarctica2015co2composite.txt
Jouzel, J. et al. (2007). Orbital and millennial Antarctic climate variability over the past 800,000 years Science 317:793–796 · 10.1126/science.1141038
EPICA Dome C deuterium temperature. Plotted vs last-1000-year mean.
https://www.ncei.noaa.gov/pub/data/paleo/icecore/antarctica/epica_domec/edc3deuttemp2007.txt
MacFarling Meure, C. et al. (2006). The Law Dome CO₂, CH₄ and N₂O ice core records extended to 2000 years BP Geophysical Research Letters 33:L14810 · 10.1029/2006GL026152
High-resolution Common Era CO₂. Preindustrial ~277–285 ppm; industrial rise begins ~1750–1850.
https://www.ncei.noaa.gov/pub/data/paleo/icecore/antarctica/law/law2006.txt
Rubino, M. et al. (2019). Revised records of atmospheric trace gases CO₂, CH₄, N₂O and δ¹³C-CO₂ over the last 2000 years from Law Dome, Antarctica Earth System Science Data 11:473–492 · 10.5194/essd-11-473-2019
The Law Dome discrete + spline files used in the Stomata vs Ice comparison. High-accumulation, ~8-year gas-age width — the least-smoothed ice CO₂ of the last two millennia.
https://www.ncei.noaa.gov/pub/data/paleo/icecore/antarctica/law/law2018co2.txt
Lan, X., Tans, P., Thoning, K. (NOAA GML) (2026). Atmospheric CO₂ at Mauna Loa, annual means NOAA Global Monitoring Laboratory
Direct atmosphere. 2025 annual mean 427.35 ppm.
https://gml.noaa.gov/webdata/ccgg/trends/co2/co2_annmean_mlo.txt
Fox-Kemper, B. et al. (2021). Ocean, cryosphere and sea level change IPCC AR6 WG1 Chapter 9
Fig. 9.9: LGM-to-present OHC rise >17,000 ZJ (MOT >3.1 °C) from Bereiter/Baggenstos/Shackleton. Also the long-term committed warming under CO₂.
Huybers, P., Curry, W. (2006). Links between annual, Milankovitch and continuum temperature variability Nature 441:329–332 · 10.1038/nature04745
Used in Gebbie Fig. 1. The ocean is a low-pass filter: millennial surface variance is strongly attenuated at depth.
Rosenthal, Y., Kalansky, J., Morley, A., Linsley, B. (2017). A paleo-perspective on ocean heat content: Lessons from the Holocene and Common Era Quaternary Science Reviews · 10.1016/j.quascirev.2016.10.017
Mg/Ca Holocene intermediate-water temperatures. Holocene thermal maximum warmer than late Holocene in several basins — a Holocene cooling into the LIA, then modern rebound.
PAGES 2k Consortium (2019). Consistent multidecadal variability in global temperature reconstructions and simulations over the Common Era Nature Geoscience 12:643–649 · 10.1038/s41561-019-0400-0
Surface temperature over the last 2,000 years. MCA and LIA are real, spatially messy, and smaller than the 20th-century rise at the surface — the ocean inversion can still store more heat at MCA if the deep field lags.
Wunsch, C. (2016). Global ocean integrals and means, with a note on the importance of location in an imaginary world Annual Review of Marine Science
On what a global ocean integral actually is, and why LGM abyssal isotopes are hard. Context for why 12–20 thousand ZJ is a range, not a point.
Kouwenberg, L., Wagner, F., Kürschner, W., Visscher, H. (2005). Atmospheric CO₂ fluctuations during the last millennium reconstructed by stomata of fossil Quercus Geology 33:33–36 (and Kouwenberg 2004 thesis, Tsuga heterophylla) · 10.1130/G20941.1
Stomatal CO₂ of the last two millennia. Runs well above the ice-core envelope, with centennial swings ice does not show. Limited (local, species-calibrated) and still the right check on firn-smoothed air.
van Hoof, T.B., Wagner-Cremer, F., Kürschner, W.M., Visscher, H. (2008). A role for atmospheric CO₂ in preindustrial climate forcing PNAS 105:15815–15818 · 10.1073/pnas.0807624105
Quercus robur stomatal index, 1000–1500 CE. Argues ice-core smoothing understates Common Era CO₂ variability — the same point as last week’s Stomata vs Ice comparison.
Wagner, F., Kouwenberg, L.L.R., van Hoof, T.B., Visscher, H. (2004). Reproducibility of Holocene atmospheric CO₂ records based on stomatal frequency Quaternary Science Reviews 23:1947–1954 · 10.1016/j.quascirev.2004.04.003
Betula stomata across the 8.2 ka event. Deglacial/early-Holocene stomatal CO₂ is erratic (roughly 175–400 ppm) against a gradual ice-core rise.
Finsinger, W., Wagner-Cremer, F. (2009). Stomatal-based inference models for reconstruction of atmospheric CO₂ concentration: a method assessment using a calibration and validation approach The Holocene 19:757–771 · 10.1177/0959683609105300
Betula nana stomatal density, 1860–2000 — the overlap with the instrumental record, where ice, stomata and Mauna Loa are forced to meet.
Science Blog (2026). Stomata vs Ice — interactive comparison of CO₂ proxies Published 26 August 2026
Side-by-side of stomatal-index reconstructions against Law Dome, WAIS, Dome C and Mauna Loa. Ice-core CO₂ is a heavily smoothed well-mixed mean; stomata are limited and still valuable because they retain the variability firn mixing removes.