DMI weather observations as GeoParquet
Observations from the Danish Meteorological Institute's weather stations (air temperature, dew point and wind speed), republished as monthly GeoParquet files that you can load straight into GeoPandas, pandas or DuckDB.
How it works
- Every 3 hours the updater downloads observations from DMI's open data metObs API, one parameter and month at a time. The last few days are re-downloaded each run to pick up late observations.
- As soon as all parameters for a month are in, they are merged into that month's GeoParquet file: one row per observation with the station's location as a point geometry.
- The files are served from this server over HTTP(S) and as a read-only S3 bucket. The table below shows the progress of each month.
How to use it
- S3: endpoint
, buckets3, anonymous access, data unders3://s3/raw/, partitioned byyear=…/month=…. See the GeoPandas and DuckDB examples. - In the browser: run Python with maps and plots without installing anything.
- Plain files: browse raw/ and download a month, e.g. with
curl; progress is in _status.json. - Data © DMI, CC BY 4.0.
Processing status
Each block is one parameter for one month; a month's file is built as soon as its blocks finish. loading…
Load the data with GeoPandas
The dataset is also a read-only, anonymous S3 bucket named s3 on this
server, so readers discover the year=…/month=… partitions themselves and only
download the months you filter on. Each row is one observation: station_id,
parameter_id, observed_at, value and the station's
latitude/longitude/geometry.
Or run a version of it in your browser with Pyodide, with maps and plots (it reads the same files over plain HTTP, since S3 clients don't run in browsers).
…or query it with DuckDB