🌤️ MeteoSwiss Open Data: Local Forecast Demos

This repository provides Jupyter notebook examples for accessing and visualizing local weather forecasts from MeteoSwiss, released through Switzerland's Open Government Data (OGD) initiative.

Access high-resolution forecasts for ~5,600 points across Switzerland — including weather stations, towns, and postal code areas (PLZ).

---

🌤️ The Meteogram Notebook

notebooks/Meteogram.ipynb walks you through the full workflow of accessing MeteoSwiss OGD forecast data, from location discovery to visualization. It is structured as a story with one concept per section:

Section	What you learn
1 · Choose your location	How forecast data is organised by point of interest (POI)
2 · Explore parameters	What MeteoSwiss publishes and how parameters are grouped
3 · Download the data	How to query the STAC API and parse the CSV files
4 · Visualise	How to render a 9-day meteogram
5 · Daily summary	How pictogram codes map to weather descriptions

Key features:

• Integrated POI search — find your location by name or ZIP code directly in the notebook
• 9-day forecast horizon — hourly data combined with daily summaries
• Metadata-driven — units, labels, and panel groupings resolved automatically from OGD metadata
• Accurate day/night shading — sunrise and sunset computed per location using astronomical calculations

What the meteogram shows

The chart is divided into up to six panels, each covering a different aspect of the forecast:

Panel	What is shown
Temperature	Hourly median temperature at 2 m (°C) with a Q10–Q90 uncertainty band; daily minimum and maximum markers; freezing-level altitude (m a.s.l.) on a secondary axis
Precipitation (hourly)	Hourly precipitation amounts (mm) as bars with a Q10–Q90 uncertainty band; precipitation probability (%) as a dashed line on a secondary axis
Precipitation (daily)	Daily total precipitation (mm) as bars with Q10–Q90 whiskers
Wind	10-minute mean wind speed and gusts (km/h) with Q10–Q90 uncertainty bands; wind direction shown as arrows every 3 hours
Sunshine	Hourly sunshine duration (min)
Radiation	Global and diffuse solar radiation (W/m²)
Clouds	Low, mid, and high cloud cover (%) as a stacked area chart

All hourly panels share the same time axis with day/night shading. You can display any subset of panels by setting PANELS = ["Temperature", "Wind"] in the configuration cell.

The plotting code lives in meteogram_plot.py, alongside the notebook. The notebook itself focuses on the data and the API; open the module only if you want to customise the chart.

Daily Weather Summary Table (example output):

Date	Weather	T min (°C)	T max (°C)	Precip. (mm)
Mon 19.05	partly sunny, thick passing clouds	12.3	22.1	0–2
Tue 20.05	very cloudy, light rain	10.8	19.5	3–15
Wed 21.05	high clouds	11.1	21.3	0
Thu 22.05	mostly sunny, some clouds	12.5	23.0	0
Fri 23.05	overcast, some rain showers	11.0	18.7	5–20
Sat 24.05	sunny	10.2	24.1	0
Sun 25.05	mostly sunny, some clouds	11.8	25.3	0
Mon 26.05	partly sunny, thick passing clouds	12.0	22.8	0–3
Tue 27.05	mostly sunny, some clouds	13.0	24.2	0

---

☀️ The Solar Production Notebook

notebooks/SolarProduction.ipynb shows a practical downstream use case for the same OGD data: estimating expected PV (solar panel) output from the 9-day radiation forecast. It reuses the fetch/parse pipeline from the Meteogram notebook, then feeds gre000h0 (global radiation), ods000h0 (diffuse radiation), and tre200h0 (2m temperature) into a pvlib-based model: solar position and panel tilt/azimuth determine plane-of-array irradiance, panel temperature derates the output, and pvlib.pvsystem.pvwatts_dc() converts that into DC power for an illustrative rooftop system.

Note: the system itself is illustrative (a generic capacity/tilt/azimuth, not a real module datasheet), but the underlying physics — solar position, irradiance transposition onto the panel plane, and temperature-dependent PVWatts output — is handled by pvlib rather than a flat radiation-ratio approximation.

The plotting code lives in notebooks/solar_plot.py, following the same notebook/plotting-module split as the Meteogram demo.

---

🚀 Quick Start

Try it instantly — no installation needed

1. Click the badge (ctrl/cmd-click to open it in a new tab) to open the notebook in a ready-to-use cloud environment. Wait for the session to start (1–5 minutes).

2. In the file browser on the left, navigate to opendata-local-weatherforecast-demos/notebooks/

   

3. Open Meteogram.ipynb (or SolarProduction.ipynb for the PV output demo)

   

4. Follow the steps in the notebook you chose.

New to Jupyter notebooks? See the Jupyter documentation for an introduction.

---

🛠️ Technical Details

Data source

Forecast data is fetched directly from the Federal Geodata Infrastructure STAC API.

• Collection: ch.meteoschweiz.ogd-local-forecasting
• Update cycle: Updated hourly
• Horizon: 9 days (D+0 to D+8)

Repository structure

notebooks/
  Meteogram.ipynb       # Main notebook — data access and API walkthrough
  meteogram_plot.py     # Plotting module — all matplotlib code lives here
  SolarProduction.ipynb # Downstream demo — PV output estimate from the radiation forecast
  solar_plot.py         # Plotting module for the Solar Production notebook

Architecture

The notebook is metadata-driven: it reads the OGD parameter CSV at runtime to resolve parameter units, panel groupings, and hourly vs. daily granularity — no hardcoded labels.

The plotting module (meteogram_plot.py) is intentionally separate so the notebook stays focused on explaining the data. It exposes a single entry point:

from meteogram_plot import plot_meteogram
plot_meteogram(df_hourly, df_daily, ...)

---

💻 Local Installation

git clone https://github.com/MeteoSwiss/opendata-localforecast-demos.git
cd opendata-localforecast-demos
python3 -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate
pip install poetry
poetry install
poetry run jupyter lab