Published Lake Erie climate rules had mixed test skill in 2021-2024, and a revised bloom record changed their errors

We tested the printed climate-only bloom rules of Tewari et al. (2022) in Lake Erie years outside their 2002-2020 fit. In 2021-2024, the common rule had RMSE 1.62 against the revised EPA severity record, compared with 1.14 for the training mean and 0.62 for a trailing five-year mean. The mean of all 19 rules had RMSE 1.62. One published NAO alternative had RMSE 0.84. It beat the training mean but not the trailing mean. With the earlier NOAA severity values, the common and NAO rules had RMSE 3.24 and 3.30. Revisions lowered mean test severity by 2.80 units. The common rule still had negative point skill after a constant refit, source-definition changes, or removal of any one test year. Paired year uncertainty was wide. We supply a test across observation versions that separates changes in mean error from forecast skill. Limits include four test years, only two strictly after publication, index-definition uncertainty, and incomplete recovery of the original input archive. We do not test nutrient models or the NOAA operational forecast.