The ocean absorbs most of the excess heat from anthropogenic climate change, causing global ocean warming and sea-level rise with a series of consequences for human society and marine ecosystems. While there have been ongoing efforts to address large uncertainties in future projections, to date the projected ocean warming has not been constrained by the historical observations. Here, we show that the observed ocean warming over the well-sampled Argo period (2005–2019) can constrain projections of future ocean warming and that the upper-tail projections from latest climate models with high climate sensitivities are unrealistically large. By 2081–2100, under the high-emission scenario, the upper 2,000 m of the ocean is likely (>66% probability) to warm by 1,546–2,170 ZJ relative to 2005–2019, corresponding to 17–26 cm sea-level rise from thermal expansion. Further narrowing uncertainties requires maintenance of the ocean observing system to extend the observational record.

海洋吸收了人为气候变化产生的大部分多余热量，导致全球海洋变暖和海平面上升，给人类社会和海洋生态系统带来一系列后果。虽然一直在努力解决未来预测中的巨大不确定性，但迄今为止，预测的海洋变暖并未受到历史观测的限制。在这里，我们表明，在采样良好的 Argo 时期（2005-2019 年）观察到的海洋变暖可以限制对未来海洋变暖的预测，并且具有高气候敏感性的最新气候模型的上尾预测大得不切实际。到 2081-2100 年，在高排放情景下，海洋上层 2,000 m 可能（>66% 的概率）相对于 2005-2019 年变暖 1,546-2,170 ZJ，对应于热膨胀导致海平面上升 17-26 厘米。进一步缩小不确定性需要维护海洋观测系统以扩展观测记录。