UK Cell Tower Database

1 Line-of-Sight Raycasting

This is the core validation technique. For each candidate mast, we cast 72 rays outward (every 5°) from the antenna tip, sampling the terrain and Ofcom coverage at 200m intervals. The goal: determine whether this mast location is consistent with the coverage patterns Ofcom shows on its live maps.

How raycasting works

Each ray starts at the mast's antenna tip (ground elevation + antenna height) and extends outward to a distance determined by the site's environment:

• Urban: 1 km — dense sites, short range
• Suburban: 2 km — moderate range
• Rural: 3 km — longer range, terrain-dependent

At every 200m step along each ray, we check two things:

Terrain obstruction check

At each 200m sample point, we compute the elevation angle from the mast antenna to that point, accounting for:

• Ground elevation from 30m DEM tiles (AWS Terrarium)
• Earth curvature drop — at 3 km this is ~0.6m, small but cumulative
• Atmospheric refraction (k = 0.13, conservative) which slightly bends radio waves over the horizon
• Receiver height of 1.5m (person holding a phone)

If any intervening terrain creates a steeper elevation angle than the current maximum, it blocks line-of-sight to all points behind it — a shadow zone. The algorithm tracks the maximum blocking angle as it progresses along each ray.

Ofcom coverage check

For sample points with clear LOS, we check Ofcom's official predicted coverage tiles (BNG projection). Coverage is classified by signal strength:

• ■ Good indoor — strong, reliable signal
• ■ Good outdoor / variable indoor
• ■ Good outdoor only
• ■ Variable outdoor — weak signal

We require at least "good outdoor" coverage (Ofcom's purple designation) at the correct operator for a point to count.

2 Voronoi Nearest-Mast Attribution

Finding coverage with line-of-sight isn't enough. A mast in a city might have clear LOS to coverage that's actually served by a different, closer mast of the same operator.

For each coverage point found with clear LOS, we check whether the test mast is the nearest same-operator mast to that point, searching within 3× the ray casting range. If a closer same-operator mast exists, the coverage is attributed to it instead.

A site is validated if at least one sample point meets all three criteria:

• ✓ Ofcom shows coverage at the correct operator
• ✓ Clear line of sight from the mast antenna
• ✓ This mast is the closest of its operator to that point

The algorithm exits early on the first attributed point — if even one ray finds attributable coverage, the site is confirmed.

Worked Example: Helmsdale, Scottish Highlands

Helmsdale is a small village on the A9 coast road in Sutherland. It's a useful test case because the terrain is dramatic — steep hills rising directly from the coast — and coverage is sparse.

In this environment, the terrain-aware validation is critical. A naive proximity check would wrongly validate masts hidden behind ridges. Our raycasting correctly identifies which masts can actually see the coverage areas they claim to serve.

3 OpenCellID Cross-Reference & Enrichment

Proximity deduplication

We first identify OpenCellID sites that correspond to a validated Sitefinder site using tight proximity thresholds:

• Urban: within 200m → duplicate
• Suburban: within 300m → duplicate
• Rural: within 500m → duplicate

Matched sites are removed from the candidate pool. The remaining sites represent genuinely new masts built since 2012.

Enrichment

For each validated Sitefinder site, we collect all nearby OpenCellID cells and attach their metadata. This updates the 2012 records with:

• Current radio technologies — What's actually transmitting today. A site listed as "GSM" in 2012 is likely running LTE and possibly 5G NR now.
• Cell identifiers — LAC and CID for each cell on the site.
• Observation counts — How many independent handset observations back up this location.
• Last seen date — When the site was last confirmed active.

The result: Sitefinder's precise surveyed locations and antenna heights combined with OpenCellID's current technology and activity data.

4 LOS Validation of New OpenCellID Sites

Unique OpenCellID sites — those that don't correspond to any Sitefinder entry — go through the same LOS + Voronoi validation as Step 1. This filters out false or inaccurate crowdsourced entries while confirming genuinely new sites.

These validated sites carry their own cell metadata (radio types, cell IDs, observation counts) from the OpenCellID dataset.

5 WTR Cross-Reference

All validated sites are cross-referenced against the Ofcom Wireless Telegraphy Register. A nearby WTR match (licensed microwave backhaul) provides additional confidence the site is active current infrastructure.