Radar RS Analyst
Radar RS Analyst is a native Rust Level II radar viewer focused on fast, high-fidelity CPU rendering. It loads public NEXRAD Level II data, shows a lightweight map-style radar canvas, overlays live NWS/SPC hazards, and keeps the native super-resolution pixel pattern visible for analyst workflows.
This is early analyst software. The priority is a small, fast desktop app that preserves radar detail and keeps interaction fluid while the feature surface grows.
Download
Download the newest build from the GitHub Releases page:
- Windows x64:
radar-rs-analyst-windows-x64.zip
- Windows ARM64:
radar-rs-analyst-windows-arm64.zip
- macOS Apple Silicon:
radar-rs-analyst-macos-apple-silicon.zip
- macOS Intel:
radar-rs-analyst-macos-intel.zip
- Linux x64:
radar-rs-analyst-linux-x64.tar.gz
- Linux ARM64:
radar-rs-analyst-linux-arm64.tar.gz
Windows users can unzip the archive and run radar-rs-analyst.exe.
macOS users can unzip the archive and open Radar RS Analyst.app. Current builds are unsigned, so macOS may require right-clicking the app and choosing Open the first time.
Linux users can extract the archive and run radar-rs-analyst/radar-rs-analyst.
Features
- Native Level II Archive II decode for public NEXRAD data.
- Public realtime Level II chunk loading for faster access to in-progress scans when available.
- Full-resolution CPU viewport rendering, with no quality downgrade at zoom.
- Reflectivity, velocity, dealiased velocity, storm-relative velocity, spectrum width, ZDR, RHO, PHI, and CFP product selection when present in the scan.
- Built-in reflectivity and velocity color tables plus user-imported
.pal/.pal3 color tables.
- Lightweight vector basemap with state/county boundaries and zoom-aware town/county labels.
- Live NWS warnings/advisories, SPC mesoscale discussions, and watch-style hazard polygons with click details.
- Radar-site selection and right-click nearest-site loading without forcing the map to recenter.
- Multi-radar overlays with independent live Level II chunk refresh, visibility, opacity, center, refresh, and promote controls.
- VROT/source-gate readout for velocity interrogation.
- Arrow-key product and tilt stepping for fast analyst workflows.
- Honest timing readout split into lookup, fetch, read, decode, render, worker, texture, and cache stages.
- Cache-aware latest-scan loading so repeated loads are fast while still allowing new scans to download.
Notes
- First-time latest-scan loads depend on public NOAA/AWS network latency and file size. Repeated loads use the local cache when possible.
- Live hazard and realtime Level II availability depends on upstream public endpoints.
- Dealiasing is conservative and lightweight; it is designed to improve obvious wraps without hiding raw velocity access.
- The app is experimental analyst software, not an official NWS tool, and should not be used as the only source for life-safety decisions.
- Public Level II data availability and latency are controlled by upstream data providers.
Build From Source
Install a recent stable Rust toolchain, then run:
cargo build --release -p app_ui --bin radar-rs-analyst
The executable will be written to:
target\release\radar-rs-analyst.exe
Run it with an optional local Level II file path:
target\release\radar-rs-analyst.exe C:\path\to\KTLX20260607_162229_V06
Performance Workflows
For a fastest local build on the machine that will run the app, build with native CPU instructions:
$env:RUSTFLAGS="-C target-cpu=native"
cargo build --release -p app_ui --bin radar-rs-analyst
Do not use target-cpu=native for broadly distributed binaries unless the target CPU family is controlled.
The renderer probe measures cold decode, preview-to-final decode, direct viewport render, sample-cache build/render/reuse, DVEL, SRV, DSRV, and multiple viewport sizes:
cargo run --release -p render2d --example perf_probe -- --runs 12 --decode-runs 8 --viewport 1320x820 C:\path\to\level2-file
Use that probe as the training/check workload for PGO or renderer backend experiments. The CPU renderer remains the deterministic correctness path; a future GPU backend should beat these same probe stages without changing radar pixel semantics.
Repository Layout
crates/app_ui - desktop UI and interaction logic.crates/data_source - public radar site lookup and Level II object download/cache helpers.crates/nexrad_io - Level II decode path.crates/render2d - high-fidelity CPU viewport renderer and perf probes.crates/radar_core - shared radar volume, cut, radial, and moment data structures.tools/generate_basemap_data.py - regenerates the baked vector basemap data.
Basemap Data
The baked basemap uses public US Census 2024 cartographic state/county boundaries and Natural Earth populated places. The source archives are not committed; the generated Rust module is committed so the app builds without runtime map downloads.
License
Licensed under either of:
- Apache License, Version 2.0
- MIT license
at your option.