Radar Siting Toolkit
Open-source QGIS plugin for weather-radar coverage and siting analysis.
Given a territory, a set of constraints, and a definition of what area matters, this toolkit answers: where should one or more weather radars be sited, and what will each configuration see?
It is not a beam-blockage calculator. Partial beam blockage is one component of the calculation, not the product. The product is reproducible geometric evidence to support an investment recommendation to a government or a donor — every run writes a manifest proving which inputs, parameters, and library versions produced it.
This toolkit covers roughly 15% of a full siting study. See Scope & Status for what is deliberately excluded.
Modes
| Mode | Input | Output | Use when |
|---|---|---|---|
| Evaluate | Point layer of candidate sites | Per-site coverage/blockage summary + manifest | You already have a short list |
| Discover | Eligible-terrain mask | Merit raster over the whole search space | You need to narrow a territory to a short list |
| Network | Set of radars + importance layer | Joint coverage, redundancy, residual gaps | You're combining candidates with existing radars |
| Robustness matrix | Candidate sites + DEM sources + parameter ranges | Ranking stability across k / tower height / DEM | You need to defend a recommendation against assumption changes |
Core method
For each azimuth, the DEM is sampled along the ray at native resolution and
the maximum horizon angle seen up to each range is accumulated, corrected for
earth curvature via an effective earth-radius factor k. From this, minimum
clear elevation and minimum visible height are derived directly, in a single
O(N) pass per ray — no discrete elevation-angle loop.
For short-list candidates, partial beam blockage uses the circular-segment geometry of Bech et al. (2003), always published as the range-cumulative maximum (CBB), never as raw per-bin PBB — publishing per-bin PBB produces wrong maps across the whole sector behind an obstacle.
See Methodology for the full derivation and Core API Reference for the implementation.
Implemented in plain numpy — no wradlib dependency at runtime (wradlib
pulls in scipy, matplotlib, xarray, xradar, and recommends a conda install,
which is not installable from the QGIS Plugin Manager). wradlib is used
only as an offline validation reference, outside the plugin.
Projection is azimuthal equidistant (AEQD) centered on each radar, not UTM — see Methodology → Geometry.
Elevation data
Recommended source: Copernicus GLO-30 (30 m, TanDEM-X derived). Because GLO-30's interferometric phase center falls inside the canopy — neither a faithful DSM nor a DTM — runs should be made with both GLO-30 (surface) and FABDEM (bare earth) and the envelope between them reported as vegetation uncertainty. See Elevation Data & Configuration.
License
GPL v2 or later (LICENSE) — the de facto requirement for QGIS plugins.