Radar Coverage Analysis Tool — MSSR Siting Study

Terrain-aware radar coverage analysis for siting a Monopulse Secondary Surveillance Radar (MSSR) near Nice Côte d'Azur Airport. The tool computes line-of-sight visibility over real elevation data at eight flight levels, weights it by the airspace aircraft can actually fly, and exports the result to Google Earth.

📄 Read the full technical report (PDF) — physical modelling, algorithms, constraint pipeline, and the two proposed sites.

Developed during the Data & Modeling Week at CentraleSupélec (January 2026), in collaboration with Thales, in response to a public tender from the DRAC (French Civil Aviation Authority).

---

What This Project Does

Given terrain elevation data, a candidate radar site and a list of flight levels, the tool answers three questions:

1. Where can the radar see? Ray tracing over the terrain, corrected for Earth curvature and atmospheric refraction.
2. Where can aircraft actually be? A flyability model combining an approach glide slope near the airport with a terrain clearance margin en route.
3. Where does coverage actually matter? The intersection of the two — useful visibility — which is what distinguishes a good site from one that merely covers empty sky.

---

Method

Step	Approach
Coordinates	Local ENU plane centred on the radar; error ≈ (d/R)² — under 0.03% at 100 km
Earth curvature	Targets and terrain dropped by d²/(2R_eff)
Refraction	4/3 effective Earth radius (R_eff ≈ 8495 km, ITU-R P.453), so rays stay straight
Ray tracing	Parametric ray radar → target, bilinear terrain interpolation along the path
Sampling	Distance-proportional (50 samples/km) to match the ~90 m terrain resolution
Performance	Numba JIT with prange parallel loops — full 8-level run drops from tens of minutes to a couple of minutes over 2.3 M grid points

The terrain interpolator is a custom Numba-compatible bilinear routine, written because SciPy's RegularGridInterpolator cannot be called from inside a JIT-compiled parallel loop.

---

Repository Structure

radar-coverage-analysis-tool/
├── radar_visibility/       # Line-of-sight engine
│   ├── config.py           # Radar site, flight levels, Earth model, sampling
│   ├── terrain.py          # Terrain loading, ENU projection, JIT interpolator
│   ├── visibility.py       # Ray-tracing visibility computation (Numba, parallel)
│   ├── export.py           # PNG overlays, KML/KMZ, NPZ output
│   └── main.py             # Entry point
│
├── flyability_map/         # Flyable-airspace model
│   ├── config.py           # Airport position, glide slope, cruise margin
│   ├── flyability.py       # Approach and cruise minimum-altitude surfaces
│   └── main.py             # Entry point
│
├── useful_visibility/      # Visibility ∩ flyability
│   ├── combine.py          # Mask combination and coverage statistics
│   ├── export.py           # Maps and Google Earth export
│   └── main.py             # Entry point
│
├── webapp/                 # Interactive Streamlit application
│   ├── app.py              # UI: site selection, parameters, 3D views
│   ├── compute.py          # Standalone computation backend
│   └── config_manager.py   # Run configuration
│
├── data/
│   └── Nice_Terrain_Data.npz   # Elevation grid, 1296 × 1785 points (~90 m)
│
└── docs/
    └── Radar_Coverage_Analysis_Report.pdf

---

Setup

Requires Python 3.9+.

python -m venv .venv
source .venv/bin/activate    # Windows: .venv\Scripts\activate
pip install -r requirements.txt

---

Run

Interactive application (recommended)

streamlit run webapp/app.py

Load the bundled terrain, place a radar site, tune the parameters and inspect coverage interactively.

Batch pipeline

Run the modules in order — each one consumes the previous module's output:

python radar_visibility/main.py
python flyability_map/main.py
python useful_visibility/main.py

---

Configuration

Everything tunable lives in the config.py of each module.

radar_visibility/config.py

• RADAR_LAT, RADAR_LON, RADAR_HEIGHT — candidate site and mast height
• FLIGHT_LEVELS — target altitudes in metres (FL5 … FL400)
• USE_EARTH_CURVATURE — toggle the 4/3 R correction
• SAMPLES_PER_KM, MIN_SAMPLES, MAX_SAMPLES — ray sampling density

flyability_map/config.py

• AIRPORT_LAT, AIRPORT_LON — runway reference
• APPROACH_SLOPE, APPROACH_MAX_ALTITUDE — glide-slope corridor
• CRUISE_MARGIN — terrain clearance above FL100

---

Outputs

Each module writes to its own output/ folder:

• vis_overlay_FL<alt>.png — per-flight-level coverage overlays
• radar_visibility.kmz — layered Google Earth file with the radar marker
• visibility_data.npz — raw masks, reused by the downstream modules
• useful_visibility.png / .kmz — combined coverage maps

---

Results

Two compliant sites were proposed, representing complementary trade-offs:

	Site A — most accessible	Site B — maximum coverage
Coordinates	43.6500°N, 7.1020°E	43.7469°N, 7.1081°E
Elevation	209 m	895 m
Road / power access	16 m / 100 m	495 m / 495 m
Visibility at FL300+	~100%	100%
Score	9.4 / 10	9.5 / 10

Site A minimises civil-works risk and favours terminal approach surveillance; Site B maximises en-route coverage from a higher, less masked position. Full statistics per flight level are in the report.

---

Notes

• The bundled terrain dataset covers 43.12°–44.20°N, 6.46°–7.95°E.
• A custom .npz can be used instead, provided it contains ter, lat and lon arrays.
• First run is slower: Numba compiles the kernels before the first computation.

---

Team

Data & Modeling Week project, CentraleSupélec — Paris-Saclay University.

Augustin Frenk · Charles Ferragu · Zakary Boublil · Julien Frantz · Amaury Cagnol