Developer builds offline flood digital twin of Sheffield's River Don using WebGPU and on-device AI
A developer has created Don Watch, a browser-based digital twin of Sheffield's River Don that simulates flood conditions entirely on the user's device, requiring no backend server. The tool reconstructs the Don valley using open LIDAR elevation data and replays the November 2019 Sheffield flood hour by hour using Environment Agency gauge records, while also supporting a live data mode. Rendering is handled via WebGPU and three.js, with a level-driven flood surface that spreads water realistically from river vertices across submerged terrain. An on-device AI model, loaded through WebLLM in the 1–2 billion parameter range, powers five small agents that ingest gauge data, project flood risk, and draft plain-language briefings — but all numerical hydrology calculations remain deterministic code, keeping the AI in a strictly interpretive role. The application is built as a Next.js static export and installable as a progressive web app, functioning fully offline once cached.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)
Log in to join the discussion and vote.
Log in