METAR & TAF Aviation Weather + NWS Observations

Airport weather reports for any ICAO station, US state or map box: METAR and SPECI as raw text plus decoded fields, optional TAF forecasts, and optional NWS station observations, up to 30 days back.

This Actor reads the official aviationweather.gov Data API (NOAA / National Weather Service, Aviation Weather Center) and, if you ask for it, api.weather.gov station observations for US stations. Every report keeps the original text in rawText and adds decoded columns: temperature and dewpoint in °C and °F (tenths of a degree when the report has a T-group), humidity, wind and gusts, visibility, ceiling, flight category (VFR / MVFR / IFR / LIFR), pressure, present weather, precipitation, and the 6-hour and 24-hour maximum / minimum temperature groups used by weather-market traders.

Unofficial tool. Not affiliated with or endorsed by NOAA, the National Weather Service, the FAA, Kalshi or Polymarket. It uses only the documented public APIs, at most 50 requests per minute (the Data API allows 100), with an identifying User-Agent.

Quick start — verified input

Run on 2026-10-04 on the Apify platform: one request, under 10 seconds. An empty input {} runs exactly this.

{
  "stations": ["KJFK", "KORD", "EGLL", "RJTT"],
  "hoursBack": 2
}

Today's history with max/min groups for Central Park and Chicago Midway, plus TAFs:

{
  "stations": ["KNYC", "KMDW"],
  "hoursBack": 24,
  "includeTaf": true
}

Current conditions for every airport in Texas:

{
  "stations": [],
  "usStates": ["TX"],
  "hoursBack": 2,
  "latestOnly": true
}

Common input mistakes

Wrong	Right	Why
"stations": ["LHR"]	"stations": ["EGLL"]	ICAO codes have 4 characters. A 3-letter code gets a K prefix (US only): ORD → KORD.
"stations": ["New York"]	"stations": ["KJFK", "KLGA", "KNYC"]	Place names are rejected; use ICAO codes or a bounding box.
"boundingBox": "-74.3,40.5,-73.6,41.0"	"boundingBox": "40.5,-74.3,41.0,-73.6"	Latitude first: minLat, minLon, maxLat, maxLon.
"hoursBack": 2000	"hoursBack": 720	The API keeps 30 days; larger values are clamped.
NWS for EGLL	METAR for EGLL	NWS observations exist only for US stations; others are skipped with a note.

An invalid value stops the run before any request, with a message that names the field and lists valid values.

Empty results?

A station that sent no report in the window returns nothing (the API answers "no content"). The run still succeeds, the log names the station, and RUN_SUMMARY.stationsWithoutData lists it. Typical causes: a wrong code, a closed or part-time station, or hoursBack: 1 for a station that reports every few hours.

What you get

One record per report, with recordType:

• metar — stationId, stationName, latitude, longitude, elevationM, reportType (METAR or SPECI), observedAt, receivedAt, rawText, temperatureC/F, dewpointC/F, relativeHumidityPct, windDirectionDeg, windVariable, windSpeedKt, windGustKt, visibilityMiles, visibilityText ("10+"), altimeterHpa, altimeterInHg, seaLevelPressureHpa, weather, cloudLayers, ceilingFt, verticalVisibilityFt, flightCategory, maxTemp6hC/F, minTemp6hC/F, maxTemp24hC/F, minTemp24hC/F, precip1hIn, precip3hIn, precip6hIn, precip24hIn, snowDepthIn, pressureTendencyHpa.
• taf (optional) — reportType (TAF / TAF AMD), issuedAt, validFrom, validTo, rawText, and forecastPeriods (from, to, change group FM / TEMPO / BECMG / PROB, wind, visibility, weather, clouds).
• nws_observation (optional, US) — the NWS quality-controlled values for the same columns (wind converted from km/h, visibility from metres, clouds from metres), plus textDescription ("Cloudy").

Every record carries source, sourceUrl, license and retrievedAt. All columns exist on every record (null when they do not apply), so CSV exports keep stable columns. The Temperatures view shows the max/min groups side by side.

Reliability

• No silent truncation. The Data API returns at most 400 reports per request. Requests are sized below that (stations × hours × 3 ≤ 350), and any answer that still comes back with 400 rows is split (fewer stations, a shorter time slice, or a box cut into quarters) and read again; duplicates are removed.
• Limits respected. Requests are sent one at a time, at least 1.2 s apart. A 429 is retried only after the wait the server asks for.
• Partial success. If TAF or NWS reading fails, or the request budget runs out, the METAR records already read are kept and the run ends successfully with complete: false. If nothing can be read at all, the run fails visibly.

Honest limits

• History is limited to what the API keeps (30 days for METAR, about 7 days for NWS observations). For older data use an archive.
• Decoding is done by the Aviation Weather Center; this Actor converts units and adds ceiling and humidity, but does not re-decode remarks beyond the max/min, precipitation and pressure-tendency groups the API provides.
• TAF is the current forecast only, not past TAFs.
• Weather markets settle on official climate reports (for example the NWS Daily Climate Report), which can differ slightly from METAR max/min groups. Check each market's rules for its station.
• NWS observations are read for up to 50 US stations per run, one request each.

Pricing

Pay per record stored in the dataset. A run stops by itself when it reaches the maximum charge you set.