Avalanche Simulation with Rust and WebGPU

Try it yourself!

Most modern browsers now support WebGPU (check the current status). The application has been tested on Windows, Linux, and Android, though support on mobile devices may still be limited.

You can explore one of the predefined AvaFrame examples using the dropdown menu, or upload your own Austrian GPX file to run a custom simulation.

This project was initiated to streamline the development of avalanche simulations using WebGPU, building on concepts from weBIGeo. It provides an easy way to visualize results directly in the browser.

The core has since been rewritten in Rust, with Python and WebAssembly (WASM) bindings available. The underlying model is a block-based approach without particle interactions, enabling fast initial estimates of runout distance and flow routing.

Get started with Python

pip install avalanchers[viz]

import avalanchers

sim = avalanchers.PySimulation.new()
sim.create_example("frontend/data/avaframe/avaMal.png")
sim.run()

# needs the viz option
avalanchers.plot2d(sim, "max_velocity")
avalanchers.plot3d(sim, "max_velocity")

Known Issues with Chrome

• Chromium on Windows currently ignores the high-performance option in powerPreference WebGPU flag if you have multiple GPUs [Issue]. Options are:
  1. Run slow on integrated GPU
  2. Start Chrome with high performance gpu flag "C:\Program Files\Google\Chrome\Application\chrome.exe" --force_high_performance_gpu
  3. Activate flag in Chrome chrome://flags/#force-high-performance-gpu. I got slow runs about 50% of the time even though it actually runs on the fast GPU. Might work by now.
  4. Activate the dedicated GPU for Chrome in the system settings.

Development Setup

Windows

Install Rust Install Python

# wasm bindings
cargo install wasm-bindgen-cli
cargo install wasm-pack

# coverage
cargo install cargo-tarpaulin

# python bindings
python -m venv .venv
.venv\Scripts\Activate.ps1  
python -m pip install maturin

Linux

Install dependencies

Install Rust (WSL (2026.04.04): curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh) Install Python

sudo apt update
sudo apt install -y python3-venv python3-dev build-essential clang lld pkg-config libssl-dev

# wasm bindings
cargo install wasm-bindgen-cli
cargo install wasm-pack

# python bindings
python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install maturin

Build

# cli tool
cargo build -p cli --release

# python bindings
maturin develop 
maturin build --release

# wasm bindings
wasm-pack build crates/wasm_bindings --target web --out-dir ../../frontend/js/pkg --no-opt
wasm-pack build crates/wasm_bindings --target web --out-dir ../../frontend/js/pkg --release

Run

Python

python avalanchers_example.py

Frontend

1. Install OpenSSL
2. Go to the project directory
3. Start the development server: python .\frontend\dev_server.py This launches the app with caching disabled and HTTPS enabled using self-signed certificates, which are required for WebGPU support in most browsers.
4. Open your browser and go to: https://localhost/ If you are accessing the server from another device on the same network, replace localhost with the host machine’s IP address.

Benchmarks

cargo bench -p simulation

Code Guidelines

Before pushing use:

Install pip install ruff

cargo fmt
cargo clippy -- -D warnings
cargo test -p compute_core -p data_processor -p simulation
ruff check ./python_module
pytest python_module

Data Sources

Test examples are from AvaFrame under EUPL-1.2 license

Vals data: D'Amboise Christopher J. L., Neuhauser Michael, Teich Michaela, & Fischer Jan-Thomas. (2021). Maverick-bfw/Flow_py_inputs_results: First releases (Version v1) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5154787