Terrain-aware RF propagation analysis for VHF/UHF (amateur, GMRS, rail) using real elevation data. A Python orchestration layer around SPLAT! (Longley-Rice / Irregular Terrain Model) with a per-band foliage/clutter correction layer and PostGIS persistence for spatial querying and QGIS visualization.
For VHF/UHF work in hilly, wooded terrain, coverage is line-of-sight limited and dominated by terrain geometry and antenna height -- not transmitter power. (Example: 20 W -> 45 W is only ~3.5 dB / ~half an S-unit, while a single blocking ridge can cost 20-40+ dB of diffraction loss, and foliage at 460 MHz adds several to many dB on top.) Answering "will site A actually hit site B over these ridges?" requires software that reads real terrain and computes path viability -- not eyeballing a paper map.
- SPLAT! already implements the Longley-Rice/ITM physics correctly. Rewriting it would be a months-long trap with no payoff. We drive it.
- Frequency is a parameter, not a separate model. One ITM engine handles 2 m through 70 cm / GMRS by parameter.
- Foliage/clutter is genuinely frequency-specific and ITM does NOT model
vegetation -- so it is applied as a separate correction layer on top of
ITM output (Weissberger / ITU-R P.833). See
splat_rf/models/foliage.py. - PostGIS is the keystone. It turns ad-hoc runs into a queryable regional RF system: terrain as raster, sites/results as geometry, spatial queries (e.g., "which candidate relay spots have a usable shot at the repeater"), and native QGIS visualization.
SRTM tiles ──► PostGIS (raster terrain)
│
Python layer ──────┤ reads terrain, drives SPLAT!, applies foliage model
│
SPLAT! (ITM) ──► results ──► PostGIS (coverage/path geometry)
│
QGIS (native PostGIS connection)
Early scaffold (Phase 0/1). SPLAT! is installed system-wide; PostGIS is
available locally. See docs/ARCHITECTURE.md for the phased plan.
splat_rf/ Python package (orchestration layer)
models/foliage.py Frequency-specific foliage correction models
config/ Presets. *.example.* tracked; *.local.* PRIVATE/gitignored
sites.example.yml Sites, transmitters, relay candidates (COORDS PRIVATE)
antennas.example.yml Antenna presets (from public datasheets)
bands.example.yml Band / ITM parameter presets + foliage model choice
database.example.yml PostGIS connection template
sql/01_schema.sql PostGIS schema (rf_analysis namespace)
data/srtm/ SRTM .hgt tiles (gitignored -- re-fetchable)
data/sdf/ SPLAT! .sdf terrain (gitignored -- derived)
outputs/ Path profiles, coverage maps, KML (gitignored)
docs/ Architecture / design notes
The reusable tooling is public-friendly. Your actual site coordinates,
operating locations, callsign, and DB credentials are private and live only
in *.local.* files and the gitignored data/ tree. Only *.example.*
templates are committed. This is enforced by .gitignore from day one so it
never has to be retrofitted before going public.
- SPLAT! --
sudo apt install splat(providessplat,splat-hd,srtm2sdf,usgs2sdf). Verified installed: 1.4.2. - PostgreSQL + PostGIS -- verified: PostGIS 3.6.4, DB
gis_dev. - GDAL -- verified:
gdalinfo,gdal_translatepresent.
# 1. Python environment
conda env create -f environment.yml
conda activate splat_rf
# 2. Database schema (into existing gis_dev, namespaced under rf_analysis)
psql -d gis_dev -f sql/01_schema.sql
# 3. Local config (COPY templates, fill in real values -- these stay private)
cp config/database.example.yml config/database.local.yml
cp config/sites.example.yml config/sites.local.yml
cp config/antennas.example.yml config/antennas.local.yml
cp config/bands.example.yml config/bands.local.yml
# edit *.local.yml ...
# 4. Terrain data (Phase 1 fetcher -- see ARCHITECTURE.md)
# Acquire SRTM .hgt tiles for the AOI -> convert to .sdf via srtm2sdf.MIT License — see LICENSE. Copyright (c) 2026 Nate Murry.
Note on SPLAT!: this project invokes SPLAT! (splat / splat-hd) as a
separate command-line binary; it does not incorporate or link SPLAT!'s source
code. SPLAT! is independently licensed under the GNU GPL v2 by its author
(John Magliacane, KD2BD). Running a separately-licensed binary as an external
process does not extend SPLAT!'s copyleft to this MIT-licensed orchestration
layer.