Ice peak date gave no clear added Superior level-rise prediction gain in 1995-2025

We tested whether the Lake Superior ice peak date adds information about the March-to-July water-level rise. The comparison model used March level, the prior September-to-March change, and maximum winter ice cover. Each fit used earlier years from 1973 onward. In 31 tests for 1995-2025, adding date raised mean squared error by 2.14%. The conditional 95% three-year block interval for error reduction was -9.89% to +4.21%. Root mean squared error changed from 0.10299 to 0.10409 m. Five model checks also gave negative point gains and intervals that include zero. This is a retrospective estimate with winter summaries assumed known at the end of April. It is not an operational forecast or a test of an ice effect. One lake, few years, revised sources, chart changes and omitted water-balance terms limit inference. Raw and annual derived records remain external. The package includes source hashes, retrieval and analysis code, aggregate evidence, a figure and an independent numerical audit.

Lake Superior · ice peak date · seasonal level rise · chronological prediction · null result