Step changes in OISST v2.1 product noise cause most of the 1982-2025 rise in daily SST-anomaly persistence; 126 moored buoys show no clear rise

Question: Lee et al. (2025) used daily NOAA OISST and reported that sea-surface temperature (SST) anomalies became much more persistent after 1982. Is this an ocean change or a change in product noise? Less white noise makes the autocorrelation decay more slowly and lowers the high-frequency variance, with no change in the ocean. Approach: We computed yearly autocorrelation e-folding times of detrended daily OISST v2.1 anomalies for 1982-2025 (29,129 cells, 60S-60N, ice-free). We located step changes in day-to-day variance and compared them with the OISST error field. We compared OISST with 126 long-record moored buoys (52 NDBC, 74 GTMBA tropical moorings) at the same cell and on the same days. Results: The naive 1/e crossing of the global mean autocorrelation increased from 9.5 to 16.7 days (1982-1991 to 2016-2025, +75%; +2.05 days per decade, 95% interval 1.54 to 2.60). The day-to-day variance of OISST fell in abrupt steps on 1985-01-12, 2007-01-02, 2020-07-07 and 2021-11-01, from about 0.045 to 0.010 K^2. The OISST error variance fell in the same steps. At the buoys, the OISST e-folding time increased by +2.54 days per decade (1.80 to 3.30), but the buoys' own value increased by only +0.61 (-0.12 to 1.34). OISST changed from noisier than the buoys (day-to-day variance ratio 1.57 in 1985-2006) to smoother than the buoys (0.38 in 2022-2025). After control for ENSO and for the noise steps, the global trend is about half of the raw trend (+0.93 to +1.03 days per decade). A noise-insensitive 5-20 day tail fit gives a smaller rise (+15%, +0.75 days per decade, 0.29 to 1.23); the buoys neither confirm nor exclude it. The duration trend of detrended heatwave-like events (+1.17 days per decade) becomes +0.32 +/- 0.45 after the noise control. Conclusion: Step changes in OISST input and noise cause most of the reported doubling of daily SST persistence. A smaller rise of about 1 day per decade remains in OISST and is not confirmed in situ. Limitations: buoy data are OISST inputs; the buoys do not sample the Southern Ocean or subtropical gyres; we did not reproduce the exact method of Lee et al.; the last noise step occurs at the same time as the strong 2023-2024 warm events.

sea surface temperature · OISST · persistence · autocorrelation · thermal inertia · marine heatwaves · measurement noise · data inhomogeneity · TAO/TRITON · NDBC buoys · change points · replication