Try it live: huggingface.co/spaces/snesbitt/myPSD ·

myPSD is a teaching and exploration tool that shows how the shape of a precipitation particle size distribution (PSD) translates into the observables a dual-polarization weather radar actually sees. You pick a PSD and a precipitation type, and the app computes the full size spectrum together with eleven polarimetric metrics (Zh, Zdr, Kdp, ρhv, LDR, δ, attenuation, Nt, LWC/IWC, and more) using T-matrix scattering in real time.

It accompanies ATMS 410 — Radar Meteorology at the University of Illinois Urbana-Champaign and the textbook Radar Meteorology: A First Course (Rauber & Nesbitt, 2018).

This is a public web app hosted on Hugging Face Spaces as a Docker container — no install, no account required. Source lives at swnesbitt/myPSD on GitHub; every push to main auto-deploys to the Space.

For a given PSD, particle type, and radar wavelength, myPSD:

1. Builds (or reuses from cache) a T-matrix scatter table via rustmatrix, a Rust port of pytmatrix, using habit-appropriate axis ratios and refractive indices.
2. Integrates the normalized-gamma PSD (N(D) = Nw·f(μ)·(D/Dm)^μ·exp[−(4+μ)D/Dm]; Testud et al. 2001) against that scatter table.
3. Returns both the N(D) curve (log concentration vs. diameter) and a metrics table of the polarimetric observables a radar at that wavelength would measure.

Supported precipitation types:

• Rain — oblate drops, Beard & Chuang (1987) axis ratios, water at 10 °C
• Hail — near-spherical solid ice, D up to 75 mm
• Sector-plate snow — low-density plates
• 6-point bullet rosettes — sparse ice volume fraction
• Bullet-rosette aggregates — Brandes et al. (2007) ρ(D)

Snow habits use the Honeyager (2013) spheroid-proxy approach: a Maxwell-Garnett ice-in-air effective medium is built per-diameter, so the T-matrix stays tractable while capturing the density variation that drives Zdr and ρhv for realistic snow.

Controls (left column):

• Precipitation type — pick a habit. The particle shape, density model, and D_max are set automatically.
• Radar wavelength — S (10 cm), C (5 cm), or X (3 cm) band.
• Dm (mm) — mass-weighted mean diameter. Shifts the PSD peak.
• log₁₀ Nw (mm⁻¹ m⁻³) — Testud normalization intercept. Scales total concentration without changing spectral shape.
• μ (shape) — gamma shape parameter. Positive μ narrows the distribution; negative μ fattens the small-D tail.
• Canting σ (deg) — standard deviation of the orientation distribution. Larger σ reduces Zdr and Kdp and raises LDR.

Presets (below the sliders) — one-click jumps to canonical archetypes from the literature: drizzle, light stratiform rain, continental/maritime convection, light plate snow, heavy aggregates, and a 70 dBZ hail core. Each button shows its source citation. Nudging any slider deselects the preset.

PSD plot — N(D) on a log scale, shaded by total concentration. The x/y range fields below the plot let you zoom. Hover the Definitions row above the plot for the math behind the axes.

Metrics table — polarimetric observables at the chosen band, with a hover tip on each row explaining what the metric tells you physically.

Assumptions — the per-species panel at the bottom lists the scattering assumptions that are currently active (refractive index, axis-ratio model, D_max, density law), with literature citations. This is deliberately always visible: the numbers above are only as good as the assumptions under them.

You only need this if you want to modify the code. The public app runs on Hugging Face and doesn't require anything local.

cd backend
uv venv
uv pip install -e .
uv run uvicorn app.main:app --reload --port 8000

cd frontend
npm ci
npm run dev

The Vite dev server proxies /api to localhost:8000. Production builds output to backend/app/static/, which FastAPI serves from the same origin.

docker build -t mypsd .
docker run --rm -p 7860:7860 -e HOME=/tmp mypsd
# open http://localhost:7860

• rustmatrix — Rust-backed T-matrix scattering (drop-in pytmatrix replacement)
• FastAPI — thin JSON API over the scatterer
• React + Vite + Mantine + Plotly.js — interactive single-page frontend
• Docker on Hugging Face Spaces — deployment
• GitHub Actions — auto-sync main → the HF Space

Branded with the CLIMAS group icon.

• myPSD concept & original Bokeh app: Steve Nesbitt, University of Illinois Urbana-Champaign.
• pytmatrix (the original T-matrix Python wrapper): Jussi Leinonen, MeteoSwiss.
• rustmatrix: Rust port of the pytmatrix numerical core.

Full literature references for every physics choice are listed in the References panel at the bottom of the live app.