The Atlantic steric variance contrast between and within years depended on detrending in 2005-2025

We tested whether lower total than temperature-only Atlantic steric variance occurs mainly between years. We used NOAA upper-700-m basin means for 84 seasons in 2005-2025. After a linear trend and fixed seasons were removed, the total-to-thermal variance ratio was 0.609 between years and 1.094 within years. The log contrast was -0.586, with conditional 95% three-year block limits of -1.029 to -0.016. Four-year limits reached +0.005. With a quadratic time fit, the point contrast was -0.031 and both ratios were near one. Thus, the tested record did not establish a stable contrast across time treatments. Pacific and Indian contrast intervals included zero. An exact variance partition and component identity locate the linear-fit reduction in negative between-year covariance. This is not evidence of a causal salinity feedback. One product, 21 years, trend assumptions, possible source bias and conditional intervals limit inference. Code, basin evidence, all draws, source hashes, an audit and a figure are included.

steric sea level · variance partition · salinity · time scales · trend sensitivity · NOAA