Data Sources
Sources, and the fine print.
Here’s every upstream feed CurrentWeather.info relies on, what it contributes, and the known failure modes. The app doesn’t “average everything into one truth” — it selects per field, validates inputs, and tags each value with provenance so you can audit what you’re looking at.
Observations
Station Obs (NWS / METAR)
Real station measurements used per field when fresh and passing validation
(METAR / ASOS / AWOS when available).
Failure modes: stale timestamps, station noise, partial/missing fields. When that happens, only the affected
fields fall back — not the entire “current” view.
Forecast / Model
Open-Meteo
Forecast + analysis for future periods, and per-field fallback when station obs are missing, stale, or invalid.
Failure modes: spatial smoothing, model drift, and missed microclimates — especially over short distances.
That’s why “model” is always labeled, never disguised.
Radar / Satellite
RainViewer
Visual instrumentation: precipitation radar tiles + satellite cloud cover + playback timeline. Great for “what’s moving where”
— not a scalar truth source for temperature/humidity/wind.
Failure modes: tile gaps, latency, regional coverage limits. We treat it as a visual layer, not a model.
Geocoding
Reverse-geocode
Maps coordinates → human place labels and nearby station context. Useful for navigation and “where am I?”
UX — not a weather signal.
Failure modes: boundary weirdness, naming variance, occasional mismatched locality labels. We treat labels as metadata, not truth.
How sources map to provenance
Hybrid is a view-level summary. Per-field provenance labels remain the source of truth.
obs
Station observations (when fresh + valid). Selected per field.
model
Open-Meteo forecast/analysis (future periods) + fallback for missing/stale/invalid obs fields.
derived
Computed values/labels from upstream winners (obs/model, sometimes a hybrid mix).
Last updated: December 2025