We missed our 6‑hour warning target by 90 minutes on the North Fork last Thursday after radar‑only QPE overshot totals about 20%, so we’re shifting our trigger logic. If you’ve seen better performance using HEFS short‑range traces over the AHPS ensemble median for gate ops and reverse‑911 thresholds when USGS 06730200 rise rate tops 0.5 ft/hr, I want the receipts.
We were about 70 minutes late on a similar North Fork setup when radar-only QPE ran about 20% hot — . > performance using HEFS short‑range traces over the AHPS ensemble median for gate ops and reverse‑911 Once USGS 06730200 tops 0.5 ft/hr we weight HEFS p25–p35 for the next 6 h and damp radar QPE by 0.8 with MRMS gauge-corrected, which pulled our gate and reverse-911 triggers about an hour earlier; caveat: in backwater/snowmelt we keep AHPS median for magnitude and only use HEFS for timing.
Switched to HEFS short‑range and using the P75 trace for triggers; that pulled our 6‑hr target to within 10–20 min on three events, whereas the AHPS median was 45–70 min off when USGS 06730200 was >“0.5 ft/hr.” Caveat: on tight convective runs the P75 can still lag, so we escalate to P90 if the first-hour gauge rise beats HEFS by 0.2 ft. @sevans41 we keep AHPS in the plot, but it’s sanity check only now.
Quick tip: we added a fixed travel-time offset (32 min from 06730200 to our control point) to the HEFS short‑range P80 trigger, and our 6‑hr alert was within about 15 min on the last two North Fork events. , when the radar bias ran +20% the AHPS median slipped late, but a simple +0.2 rainfall bias on the P50 as a fallback kept gate ops from whipsawing. @sevans41’s “use the higher percentile when rise rate >0.5 ft/hr” matches our experience.