Local-history adjustment left no clear rise in Superior and Michigan-Huron monthly change correlation between 1920-1971 and 1972-2025

We tested whether a rise in agreement between monthly lake-level changes remains after local history is removed. We used coordinated v2025 levels for Superior and Michigan-Huron. Separate fits for each lake and period used calendar-month means and the previous three monthly changes. Residual correlation was 0.415 in 1920-1971 and 0.434 in 1972-2025. The change was +0.018, with a conditional 95% five-year block interval of -0.087 to +0.122. A comparison with only month means removed gave a change of +0.124. One-lag and twelve-lag checks also gave intervals that included zero. Thus the tested local-history adjustment left no clear evidence for a rise in correlation. The result does not prove constant dependence or identify a physical cause. One connected lake pair, revised data, fixed model forms and conditional block intervals limit inference. The package gives source hashes, monthly inputs, code, residuals, draws, a figure and a separate solver check.

Great Lakes · monthly water levels · local persistence · residual correlation · block bootstrap · null result