VOLCASH

Self-learning real-time weather forecast AI · volcanic ash risk API

24/7 Self-Learning Auditable vs KMA Knows When Unsure

A self-learning Korean weather forecast AI built on public meteorological data. "We know when we're wrong" — it abstains when unsure and publishes an auditable skill scoreboard against the national weather service. It also serves a volcanic ash risk (0-500) API for any coordinate. Korean/English/Japanese, free tier.

View Dashboard API Docs Back to List

Self-Learning Forecast Loop

A 24/7 autonomous loop issues forecasts, verifies them against real observations, and continuously corrects its own bias by region and variable. Verified learning persists across sessions.

Metacognition — Knows When It's Wrong

When confidence is low, it defers to the national weather service instead of serving its own forecast. It chooses an honest 'unsure' over false alarms.

Auditable Skill Scoreboard vs KMA

Every forecast is tracked issue→verify→diverge in a public ledger, showing hit-rates vs the national service (event skill, etc.) transparently by variable and region.

Volcanic Ash Risk API (0-500)

Combines satellite, dispersion models, and ground observations to return a 0-500 ash risk score with 5-tier grading (Safe, Caution, Danger, Critical, Emergency) for any coordinate.