Ocean heat content (OHC) is the most robust single measure of planetary warming. Oceans cover more than 70 percent of Earth’s surface and have an enormous heat capacity. As a result they have absorbed more than 90 percent of the excess energy trapped by rising greenhouse-gas concentrations since the mid-20th century. Surface air temperature fluctuates with weather, El Niño–Southern Oscillation cycles, and volcanic aerosols. OHC integrates heat across the water column and therefore shows a clearer, less noisy signal of the Earth’s energy imbalance.
OHC is expressed in zettajoules (1 ZJ = 10^{21} joules). A single zettajoule is roughly equivalent to the energy released by 15 billion Hiroshima-sized bombs, or several years of total human primary-energy use. Small changes in average ocean temperature therefore represent vast quantities of stored heat.
How scientists measure it
Temperature profiles have been collected for decades by research ships, expendable bathythermographs, and moored buoys. Coverage was sparse before the 2000s, especially in the Southern Ocean and at depth. The Argo array of roughly 4,000 autonomous profiling floats, deployed from the early 2000s onward, transformed the observing system by providing regular, near-global sampling of the upper 2,000 m. Satellite altimetry and gravimetry add complementary information on sea-level change and mass.
A 2026 Nature Communications analysis quantified how much this improvement has reduced uncertainty. The 5–95 percent error range on global upper-2,000 m OHC fell six-fold, from about 111 ZJ in the mid-1950s–1960s to about 19 ZJ in 2014–2023. Most of the gain came from better spatial sampling and mapping methods. Quality-control of outliers in eddy-rich regions is now a leading residual error source. The same study concluded that the upper 2,000 m of the ocean has gained a robust 449 ZJ (range 385–519 ZJ) since 1955.
Different research groups (IAP/CAS in China, NOAA/NCEI, Copernicus Marine, CIGAR-RT reanalyses) produce independent estimates. They agree on the long-term rise and on recent records even when year-to-year increments differ by several zettajoules.
The multi-decadal record
Warming of the upper ocean is not new, but it has accelerated. Linear rates for 1960–2025 are typically near 0.14 W m^{-2} per decade when averaged over the Earth’s surface. For 2005–2025 the rate roughly doubles to about 0.32 W m^{-2} per decade in the IAP/CAS product; other datasets give similar or slightly higher values. A 2005–2023 trend of 10.9 ZJ yr^{-1} (0.68 W m^{-2}) is statistically robust, and a modest further acceleration of 0.34 W m^{-2} per decade over that interval is also detectable.
Heat is not confined to the mixed layer. Roughly 40 percent of the increase since 1960 resides in the top 300 m, another 22 percent between 300 and 700 m, 29 percent between 700 and 2,000 m, and about 9 percent below 2,000 m. The Southern Ocean and Atlantic have taken up a disproportionate share of the heat on a per-area basis; the Pacific stores the largest absolute amount simply because of its size.
Full-depth estimates place the 1960–2025 rise near 481 ± 48 ZJ. NASA’s five-year running anomaly for the upper 2,000 m stood at 372 ± 2 ZJ above the 1955 baseline as of late 2024. These numbers continue to climb.
Consecutive annual records
Annual global upper-2,000 m OHC has set a new record every year from 2017 through 2025—the longest such streak in the instrumental era. The ranking of the five warmest years (relative to a 1981–2010 baseline, IAP/CAS) is:
- 2025: 317 ZJ
- 2024: 294 ZJ
- 2023: 281 ZJ
- 2022: 266 ZJ
- 2021: 247 ZJ
Year-to-year increments were 23 ± 8 ZJ in 2025 and 13 ZJ in 2024 (IAP/CAS). Independent products confirm the same order and similar magnitudes: CIGAR-RT shows 2025 at 346 ZJ (+20 ZJ), NOAA/NCEI (through mid-2025) at 306 ZJ (+13 ZJ). Copernicus Marine reports a larger but more uncertain jump because it incorporates near-real-time data.
The 2025 record occurred even though 2025 was only the third-warmest year for global sea-surface temperature and featured developing La Niña conditions. Cooler surface waters in the eastern tropical Pacific do not stop heat from accumulating at depth; stronger trade winds can actually push warm water downward in the western Pacific. About 16 percent of the ocean surface area reached its highest OHC on record in 2025; one-third of the ocean ranked in its historical top three. Regional records were set or approached in the tropical and South Atlantic, Mediterranean, North Indian Ocean, and Southern Ocean.
Surface temperature records in 2026
Sea-surface temperature (SST) and OHC are related but not identical. SST responds quickly to air–sea heat fluxes and to El Niño. In August 2026 the extra-polar (60°S–60°N) daily average SST reached 21.1 °C, edging past the previous record of 21.09 °C set in March 2024. The timing is unusual: the seasonal peak normally occurs in March–April after Southern Hemisphere summer. August 2026 also exceeded the prior highest August value (20.98 °C in 2023).
A rapidly strengthening El Niño in the tropical Pacific is adding heat on top of the multi-decadal warming trend. Forecasts in mid-2026 indicated the event could become one of the strongest on record. Marine heatwaves already covered large fractions of the global ocean in 2024–2026; the number of marine-heatwave days has more than tripled since the early 1990s.
Why the records matter
Thermal expansion of seawater accounts for 30–40 percent of observed global sea-level rise. Additional heat also melts ice shelves from below, reduces polar albedo, and intensifies the hydrological cycle. Warmer oceans fuel stronger tropical cyclones, expand oxygen-minimum zones, and drive mass coral-bleaching events. Fisheries and coastal communities face both acute heat-wave mortality and longer-term habitat shifts.
Because the deep ocean has a long memory, much of the heat already stored will remain for centuries even if emissions fall. Continued high-quality observations—especially in the deep ocean, under sea ice, and in western-boundary currents—are essential for tracking Earth’s energy imbalance and for testing climate models. The six-fold drop in OHC uncertainty since 1960 shows what an observing system can achieve; the unbroken string of annual records since 2017 shows what that system is now measuring.