European Weather Alerts (Regions)

Goal

Build a fully automated, explainable pipeline that tracks regional heat, cold, wind, rain, and snow hazards, and surfaces them as interactive stories on my portfolio. Each region and day receives a discrete hazard level from 0 (no alert) to 3 (strong event) per hazard, making the maps and curves easy to read and reason about.

Approach

1. Select a handful of high‑population cities per admin‑1 region (Geonames + simple filters), then treat them as representative points.
2. Use the Open‑Meteo archive API to build and maintain a ~180‑day history of daily max/min temperature, max wind, accumulated rain, and snowfall per point, and the OpenWeather 5‑day / 3‑hour forecast API to refresh today and tomorrow in CI.
3. Aggregate the point data to region level and apply transparent threshold rules to assign hazard levels (0–3) for each hazard and day (for example, mild alerts at level 1, strong events at level 2, and very rare/extreme situations at level 3).
4. Group consecutive days above a minimum level into multi‑day events, storing their duration, maximum level, and an indicative metric such as peak temperature or max 24‑hour rainfall.
5. Export region‑day records and alert events as JSON; a GitHub Action runs the pipeline nightly and keeps the website data fresh.

Guiding questions

• Which regions see the most frequent or persistent hazards over the last 180 days?
• How different are hazard patterns between northern Germany and southern Spain or Italy?
• Can a simple, rule‑based system still give useful spatial and temporal intuition?

Data

• Open‑Meteo for historical regional daily aggregates (max/min temperature, wind, precipitation, and snowfall).
• OpenWeather 5‑day / 3‑hour forecast API for short‑term forecasts (today + tomorrow) used in the daily refresh.
• Geonames city database (via geonamescache) to pick the top‑population cities per region.
• The pipeline code and CI configuration live in the weather‑alert‑pipeline repository.
• Because the project relies on free tiers, some location requests occasionally hit rate limits or timeouts. The pipeline logs and skips those points, so a small number of region–day records may be missing, but the regional patterns remain stable enough for exploration.

Hazard thresholds

To keep the pipeline explainable, all hazard levels come from simple, hand‑rolled rules applied per region and day. Level 1 marks a noticeable event, level 2 a strong episode, and level 3 an unusually intense situation:

• Heat: level 1 ≥ 30°C, level 2 ≥ 35°C, level 3 ≥ 40°C (daily max temperature).
• Cold: level 1 ≤ 0°C, level 2 ≤ −5°C, level 3 ≤ −10°C (daily min temperature).
• Wind: level 1 ≥ 50 km/h, level 2 ≥ 70 km/h, level 3 ≥ 90 km/h (daily max wind speed).
• Rain: level 1 ≥ 20 mm, level 2 ≥ 40 mm, level 3 ≥ 60 mm (daily accumulated rain).
• Snow: level 1 ≥ 10 mm, level 2 ≥ 20 mm, level 3 ≥ 40 mm (daily snowfall, water equivalent).

Headline numbers (current 180‑day window)

• Alerts captured in the current export window: 1691 total (278 heat, 191 cold, 39 wind, 1132 rain, 51 snow).
• Heat and rain drive most alerts in this 180‑day window; cold, wind, and snow rarely cross the configured thresholds, which is consistent with the June–December period covered by the current export.
• Charts often highlight three benchmark regions—Île‑de‑France, Comunidad de Madrid, and Bayern—to keep comparisons readable while still spanning different climates.