From the product review (iOS/PRODUCT_REVIEW_2026-07-01.md, finding #4).
In the Open-Meteo radar path, RadarService.createDirectionalSectors returns the same status for all 8 directions (single-point data), and "nearest precipitation" distance/direction is computed as minutes × windSpeed with direction = opposite of surface wind (RadarService.swift ~414-467) — not a radar observation. So "Light rain, 12 mi west, moving east at 15 mph" is an inference from the local wind vector, presented in the summary card with radar-level confidence. Fronts don't move with surface wind, so it's often wrong.
Options:
- Label it clearly as an estimate from wind, not an observation; and/or collapse the 8 identical sectors into one honest "precip approaching in ~N min" line.
- Real directional nowcasting needs actual gridded radar (NEXRAD/MRMS/RainViewer) sampled at offset points — noted as dropped for licensing previously.
Team notes there are ideas for this on the nowcast branch — this issue tracks it.
From the product review (
iOS/PRODUCT_REVIEW_2026-07-01.md, finding #4).In the Open-Meteo radar path,
RadarService.createDirectionalSectorsreturns the same status for all 8 directions (single-point data), and "nearest precipitation" distance/direction is computed asminutes × windSpeedwith direction = opposite of surface wind (RadarService.swift~414-467) — not a radar observation. So "Light rain, 12 mi west, moving east at 15 mph" is an inference from the local wind vector, presented in the summary card with radar-level confidence. Fronts don't move with surface wind, so it's often wrong.Options:
Team notes there are ideas for this on the nowcast branch — this issue tracks it.