MeteoSwiss Data Factory

This repository contains a Python utility to automatically download MeteoSwiss station data from the Automatic Weather Stations (SMN) Open Data via the official STAC API hosted at data.geo.admin.ch.

The routine is implemented as a class and is designed for integration into larger scientific or modelling workflows.

Only the module meteoswiss_api.py is required. A runnable example is included in the __main__ section.

---

What the script does

meteoswiss_api.py defines a class MeteoSwissDataFactory that:

1. Takes latitude and longitude coordinates.

2. Selects the nearest MeteoSwiss station using the Haversine distance formula.

3. Automatically downloads metadata:

   • Station list
   • Data inventory
   • Parameter catalogue

4. Downloads historical data for a user-defined year range.

5. Supports multiple temporal resolutions:

   • 10-minute data
   • Hourly data
   • Daily data
   • Monthly data
   • Annual data

6. Automatically:

   • Handles decade-based file segmentation (where applicable)
   • Parses timestamps
   • Filters data to the requested year range

7. Provides convenience accessors for:

   • Raw datasets
   • Temperature-only datasets
   • Irradiation-only datasets
   • Station and parameter metadata

---

Data Source

The code uses the official STAC API endpoint:

https://data.geo.admin.ch/api/stac/v1

Collection used:

ch.meteoschweiz.ogd-smn

The following metadata assets are retrieved automatically:

• ogd-smn_meta_stations.csv
• ogd-smn_meta_datainventory.csv
• ogd-smn_meta_parameters.csv

Official documentation:

• https://opendatadocs.meteoswiss.ch/a-data-groundbased/a1-automatic-weather-stations
• https://opendatadocs.meteoswiss.ch/general/download
• https://www.geo.admin.ch/en/rest-interface-stac-api

---

Installation

Dependencies:

pip install pandas numpy requests

Python ≥ 3.9 recommended.

---

Basic Usage

from meteoswiss_api import MeteoSwissDataFactory

latitude = 46.947144
longitude = 7.444751
year_start = 2005
year_end = 2016

weather = MeteoSwissDataFactory(
    latitude=latitude,
    longitude=longitude,
    year_start=year_start,
    year_end=year_end,
)

df_hourly = weather.get_raw_data_hourly()
df_daily = weather.get_raw_data_daily()
df_monthly = weather.get_raw_data_monthly()

---

API Overview

MeteoSwissDataFactory(latitude, longitude, year_start, year_end)

Initialises the class and:

• Loads metadata tables
• Determines the closest station
• Stores requested year range
• Prepares lazy-loading of datasets

---

Raw Data Access

The following methods return copies of the internally cached data:

• get_raw_data_10min()
• get_raw_data_hourly()
• get_raw_data_daily()
• get_raw_data_monthly()
• get_raw_data_yearly()

Internally, datasets are fetched only once per resolution and then cached.

---

Convenience Data Access

Hourly Temperature

df_T = weather.get_hourly_data_temperature()

Returns:

Column	Description
T_mean_degC	Hourly mean air temperature (2 m)
T_min_degC	Hourly minimum temperature
T_max_degC	Hourly maximum temperature

---

Hourly Irradiation

df_I = weather.get_hourly_data_irradiation()

Returns:

Column	Description
I_global_mean_W_per_m2	Global radiation (hourly mean)
I_diff_mean_W_per_m2	Diffuse radiation (hourly mean)

---

Daily Temperature

df_T_daily = weather.get_daily_data_temperature()

---

Daily Irradiation

df_I_daily = weather.get_daily_data_irradiation()

---

Metadata Access

• info_stations_meta_data()
• info_parameters_meta_data()
• info_selected_station()
• info_selected_station_available_data()

These provide full transparency regarding:

• Available stations
• Parameter definitions
• Data coverage of the selected station

---

Internal Logic

Decade Handling

Hourly and 10-minute historical data are stored in decade-based files (e.g. 2000–2009). The class automatically:

• Determines the required decades
• Downloads and concatenates the relevant files
• Filters data to [year_start, year_end]

Daily, monthly, and yearly data are retrieved via their respective single-asset endpoints.

---

Timestamp Handling

• Timestamps are parsed from reference_timestamp
• Stored as pandas datetime
• No timezone is attached (naive datetime)
• Data is filtered to the requested year range (inclusive)

---

Notes

• HTTP errors are surfaced via raise_for_status().
• Returned DataFrames are copies to avoid accidental mutation of internal cache.
• Not all parameters are available at all stations; consult metadata functions for coverage.

---

Example Run

The __main__ section demonstrates:

• Data retrieval at different resolutions
• Extraction of temperature and irradiation subsets
• Inspection of metadata tables

Run:

python meteoswiss_api.py

---