A 1988 step reversed two fitted runoff–precipitation timing trends in L2SWBM V3.0 for 1950–2022

We tested the January–June runoff minus over-lake precipitation centroid in Superior and Michigan-Huron for 1950–2022. We used monthly medians from L2SWBM V3.0. The primary linear gap slopes were +0.916 and +0.281 days per decade. Conditional 97.5% five-year block intervals were 0.029 to 1.825 and -0.622 to 1.178. Adding a fixed step at 1988 changed the slopes to -0.964 and -1.191, with intervals that included zero. The step year is the boundary between the source model's fitting periods. Superior results also depended on block length and season definition. Thus, the primary positive Superior slope does not establish a steady physical change. The step check does not prove a source fault. Over-lake precipitation differs from land precipitation, monthly centroids differ from daily runoff metrics, and source posterior uncertainty is omitted. We provide code, source paths and hashes, aggregate results, a figure, and an independent arithmetic audit. Source and annual tables remain external and can be retrieved or regenerated.

Great Lakes · runoff timing · precipitation · centroid · reconstruction · model sensitivity · L2SWBM