Spring flow improved early June hypoxia forecasts after a level update and gave no clear seasonal gain in Chesapeake Bay in 2015-2025

We tested whether a published spring river-flow relation forecasts later Chesapeake Bay hypoxia. We fitted linear rules to 1985-2012 monitoring-based mainstem volumes below 2 mg/L dissolved oxygen, then tested 2015-2025. For the primary May-October average, the fixed January-May flow rule had RMSE 0.245 cubic miles, against 0.229 for the previous five-observation mean. Its MSE skill was -13.7%, with a conditional 95% three-year block interval of -107.6% to +56.0%. A recent level update reduced bias but gave skill of -0.1% [-65.8%, +47.1%]. The June-September average also had no clear gain. A secondary early June test gave a narrower positive result. Keeping the old slope and updating the recent level cut MSE by 44.4% [36.2%, 60.1%] in ten available years. Removing any one year left 41.9-47.0% gain. The conditional percentile band after adjustment for 22 outcome tests was 32.5-75.2%. Gains were also positive against a recent median; some other benchmark intervals included zero. Exploratory April-ending flow gave 34.9% gain. Few years, missing early June 2020, revised observations, sparse monitoring and conditional uncertainty limit inference. These are retrospective forecasts, not archived real-time tests. The package supplies original code, aggregate evidence, figures and exact external-source provenance. It tests a published relation, not the printed anoxia or nutrient equations.