The Atlantic-to-Pacific steric variance contrast depended on block length in 2005-2025 upper-ocean estimates

We compared total and temperature-only steric sea level in NOAA basin means for the upper 700 m. We used 84 seasons from 2005 to 2025 and removed a linear trend and fixed seasonal means. Total-to-thermal residual variance ratios were 0.728 in the Atlantic with Arctic, 1.134 in the Pacific and 1.222 in the Indian Ocean. The Atlantic-to-Pacific log ratio was -0.443. Its conditional 95% interval was -0.683 to -0.025 with three-year residual blocks, but -0.676 to +0.007 with four-year blocks. Each individual basin ratio interval included one. Negative component covariance explained the lower Atlantic point ratio. With consecutive-season differences, the Atlantic ratio rose to 1.094 and the log contrast fell in size to -0.097. Thus, a stable basin contrast was not established across the checks. One product, 21 years, source error, possible salinity bias and resampling assumptions limit inference. The result does not identify a physical cause. Code, source hashes, aggregate evidence, bootstrap scores, an audit and a figure are included.

steric sea level · salinity compensation · basin variance · NOAA · block bootstrap · robustness