Austria Official Weather — GeoSphere Stations & Forecasts

Get Austrian weather observations, hourly/daily history and numerical point forecasts from the official GeoSphere Austria Dataset API. Export structured records to CSV, Excel, JSON or your operational pipeline.

Use it for construction scheduling, agricultural weather analysis, energy modelling and insurance research. This Actor normalizes public data and runs persistent monitoring; it does not replace official warnings or certify insurance claims.

Quick start

{"dataset":"current","stationIds":["11035"],"maxResults":5}

The prefilled input retrieves the latest observation for Wien Hohe Warte. One result means one location and timestamp, with all requested parameters together. Missing values remain null; rows where every requested value is missing are omitted.

Datasets and parameter codes

Input dataset	Official resource	Default parameters	Coverage
current	TAWES v1 10min	TL, RR, FF, SCHNEE	Latest station observations
historical10min	TAWES v1 10min	TL, RR, FF, SCHNEE	Recent archive, approximately 3 months
hourly	Klima v2 1h	tl, rr, ff, sh	Station-dependent historical hourly records
daily	Klima v2 1d	tl_mittel, rr, vv_mittel, sh	Station-dependent archive; catalog starts in 1775
forecast	NWP v2 1h 1km	2t, tp, 10u, 10v	Hourly grid forecast, approximately 60 hours

Parameters are validated against live dataset metadata. You can supply other official codes through parameters. Dataset availability, station history and future forecast windows vary. Forecasts are a prerelease API and may change.

Station IDs differ by dataset: Wien Hohe Warte is 11035 in TAWES and 105 in Klima. Forecast station IDs use TAWES metadata and are converted to coordinates. Never reuse a TAWES ID blindly in Klima.

{"dataset":"hourly","stationIds":["105"],"start":"2026-09-01T00:00","end":"2026-09-01T02:00","maxResults":100}

{"dataset":"daily","municipality":"Wien","start":"2026-09-01","end":"2026-09-02","maxResults":100}

{"dataset":"forecast","latitude":48.208,"longitude":16.373,"maxResults":24}

Choose one location selector: up to 20 stationIds, latitude + longitude, or municipality. For observations, coordinates select the nearest active station using its official metadata; the reported coordinates are that station's coordinates. For forecasts they select model grid points. Coordinate input is restricted to the Austrian bounding rectangle; this is not an exact national boundary check.

municipality matches a substring of official active station names (e.g. Wien), not an administrative geocoder. Up to 20 matching stations are returned. A municipality without a matching station requires coordinates. If no selector is supplied, current/10min/forecast defaults to TAWES 11035; hourly/daily defaults to Klima 1 (Aflenz).

Historical inputs require both start and end; boundaries are inclusive. Dates and timezone-free timestamps mean UTC. Current data rejects dates. Forecast dates are optional and must fall in the available run's horizon. Historical queries are split into bounded windows; maxResults counts delivered rows (1–10,000), not raw datapoints.

Output

Each record includes stationId (null for a grid point), name, latitude, longitude, timestamp, values (official codes), unit (unit per code), dataset, forecastReferenceTime, sourceUrl, scrapedAt, source and license.

Convenience fields include temperature, precipitation, windSpeed and snowDepth when available. Forecast windSpeed is derived as hypot(10u,10v), in m/s. NWP snowfall is not snow depth; no snow-depth value is invented. Daily vv_mittel means average wind, while optional ffx is a gust and stays in raw values.

Units and source sentinel values are retained. In Klima daily data, rr=-1 indicates no precipitation, rr=0 indicates less than 0.1 mm, and sh=-1 indicates no snow. Do not treat every negative measurement as an error. Accumulation periods and measurement definitions differ between datasets: compare raw parameter definitions before combining them. Forecast precipitation tp is supplied as kg/m², not silently converted.

Persistent monitoring

{"dataset":"current","stationIds":["11035"],"onlyNew":true,"monitorName":"vienna-weather","maxResults":100}

Use a named key-value store and schedule this input. monitorNome is a compatibility alias for monitorName and takes precedence when supplied. The first run emits records; later runs emit new timestamps or changed parameter values. Forecast reference time is part of identity. Selector/parameter sets have independent monitor scopes. Delivered records are checkpointed every 25 rows and at completion; a sudden process termination may replay the last batch. The store keeps at most 50,000 fingerprints per scope; older records can reappear after eviction. Runs sharing a monitor must not overlap (no cross-run locking). A small maxResults continues progress on later historical runs without marking undispatched records as seen.

Reliability and limits

Requests are sequential and spaced at least 15 seconds apart (under 5 requests/second and 240/hour for one Actor run). GeoSphere quotas are shared by source IP: avoid overlapping runs and account for other users of a proxy IP. Transient 403/429/5xx errors receive bounded backoff retries and fresh proxy sessions when running on Apify. Retry-After is respected. No login/captcha bypass is performed.

Each historical response is planned below 50,000 raw values, with a 12 MB download ceiling. Execution stops only 30 seconds before the platform timeoutAt deadline, preserving output. Without timeoutAt (local execution), there is no internal runtime cap. Large date ranges may require several runs. Invalid input, unavailable sources and internal interruption complete cleanly with zero or partial results; inspect key-value store OUTPUT (error, hint, pushed, scanned, unchanged, partial). Clean completion alone is not proof that records were found.

Pricing

Event	Price (USD)
Actor start, at default 1 GB	$0.05
One location/timestamp result	$0.002

Prices will be configured by the publisher in Apify Console. No premium event is used. The item charge tolerates an undefined event during local/private testing. Increasing memory can increase the platform start-event charge.

Source and license

Data © GeoSphere Austria, CC BY 4.0. The Actor transforms original GeoJSON into location/timestamp rows, adds English aliases and derives forecast wind speed. Keep this attribution and license when redistributing data.

Official references: Dataset catalog, license policy, request limits, official API documentation repository. Access verified anonymously on 2026-09-28. The API host's robots.txt returned 404; the portal's restrictions refer to the portal, not this documented API.

Development

Node 22+, npm ci, npm test, apify run. Tests cover invalid inputs, real source parsing, metadata coordinates, bounded windows, retries/proxy rotation and persistent monitoring. Default Actor memory is 1 GB. Cloud collaudo and publication are separate publisher steps.