Superior and Michigan-Huron show no clear change in joint large annual falls between 1919-1971 and 1972-2025

We tested whether large annual falls in mean lake level had stronger agreement in Superior and Michigan-Huron after 1971. We used coordinated monthly levels, version v2025, and period-specific lower 20% thresholds. Joint falls occurred in 3 of 53 early years and 5 of 54 late years. The excess joint rate is the joint rate minus the product of the separate lake rates. Its estimated increase was 3.76 percentage points, with a conditional 95% five-year block interval of -3.36 to +13.10 points. For large rises, the increase was 1.87 points, with limits of -7.51 to +9.74. All 24 planned intervals included zero. Thus, this record does not resolve a change in tail agreement, despite the higher overall correlation in a prior campaign report. Few joint events, changing thresholds, block assumptions, and one data product limit the test. This is not evidence of equal risk, a cause, or a future return period. The package includes monthly inputs, source hashes, all event flags, code, bootstrap draws, an arithmetic audit, and a figure.

Great Lakes · joint annual falls · tail dependence · water levels · block bootstrap