Version 1.0 remains the dashboard score. Versions 1.1 and 1.2 run beside it as shadow experiments; different numbers are not evidence of better catch prediction.
What changed
All versions are deterministic and inspectable
Currentv1.0
Recent flow and nearby weather
Flow versus the previous 14-day median · 4 points
Six-hour flow change · 2 points
Previous 12-hour modeled precipitation · 2 points
Nearby modeled cloud cover · 2 points
Shadow candidatev1.1
Seasonal flow and measured temperature
Flow percentile for this time of year · 4 points
Absolute six-hour flow stability · 2 points
Measured water temperature when available · 4 points
Rain and cloud remain context, not score points
Shadow experimentv1.2
Station-month flow and multi-window context
Station-month discharge percentile · 3 points
One- and six-hour flow trends · 2 points
Complete 6-, 24-, and 72-hour modeled precipitation · 3 points
Measured water temperature, when fresh · 2 points
Unusually rapid one-hour rise · up to 1 point removed