The radar that happened, beside what HRRR and RRFS forecast for that same hour when
issued 1 to 36 hours earlier. Two real cases from 20–22 September 2026, chosen by the data — the
most convective hour in the record and a moderate one — not by hand. Same reprojection and NWS colour
table as the app; MRMS is the observed radar.
Compare in placeGrouped by how far out the forecast was made, not by day. Tap Actual, HRRR
or RRFS above an image — the frame stays put, so you can flip between them and watch what moves.
FSS30Placement score within about 30 km, 0–1, at ≥20 dBZ (any
echo) and ≥35 dBZ (avoid). 1.0 is perfect coverage placement.
area ×Forecast area at ≥35 dBZ divided by observed. 1.0× is honest;
1.8× paints nearly twice the weather that fell.
dBZ20253035405060
+1hour
1 hour out
Valid Sun 20 Sep 0700Zrun 20 Sep 06Z944 blocks of ≥ 35 dBZ observed
Actual · MRMS Sun 20 Sep 0700Z
HRRR 20 Sep 06Z · +1 h
FSS30 ≥20 0.88≥35 0.84area ≥35 0.91×
RRFS 20 Sep 06Z · +1 hNo RRFS run reached this far out yet at this valid time.
This is a fixed snapshot, captured 22 September 2026 from three days of the pipeline's own
cache. It does not update on its own — we're keeping it as a stable, citable pair of examples
rather than a live feed. Ask us for a current read if
you want one.