Kumamap: Bear Incident Tracker

August 2025: a brown bear killed a hiker on Mount Rausu, Hokkaido.

I looked into bear incident data across Japan. Every prefecture tracks it—30+ government websites, all Japanese-only, different formats.

The Problem

Prefecture boundaries don't match reality. Hikers planning routes through the Japanese Alps cross multiple jurisdictions. Bears don't respect borders.

Each prefecture publishes independently: Google Maps pins, PDFs, Excel files, custom APIs. For foreigners, the data is invisible.

The Solution

Kumamap centralizes 75,000+ incidents from all prefectures with translation and unified mapping.

Data Pipeline

Custom Python scrapers for each prefecture. Automated collection runs twice daily via GitHub Actions:

Distance-Based Matching

Prefecture boundaries don't work for hikers. A bear 2km from a trail might fall in a different jurisdiction.

Kumamap matches incidents to 400+ locations (hiking trails, cities, tourist spots) by distance: 12km radius for trails, 18km for tourist spots, 25km for cities. Search Mount Fuji and see all incidents within hiking distance, regardless of prefecture.

Community Reporting

Anyone can report sightings. Click the map, drop a pin, upload photos, submit. No account required. Reports go live immediately with no moderation. Rate limited to 5 per hour per IP.

Interface

Interactive Leaflet map with color-coded markers: red for last 48 hours, blue gradient for older incidents. Click any marker for photos, descriptions, and coordinates. Mobile users toggle between map and feed. Desktop shows both.

The news page visualizes trends with timelines, heat maps, and species breakdown.

Stack

SvelteKit, Tailwind, Leaflet.js. Firebase Firestore. Python scrapers with OpenAI API for translation and Nominatim for geocoding. GitHub Actions for automation.