glasgow-fpv-land-search

Glasgow FPV Drone Racing — Land Search

Finding a new home for our FPV drone racing club by mining Scotland’s open land data.

We’re a registered charity and a recognised sport, with full BMFA membership and public liability insurance. We currently fly on a football-pitch-sized plot at Craufurdland Castle near Fenwick (East Ayrshire), and we’re looking for a site in and around Glasgow that the club can grow into.

This repo takes the Scottish Vacant and Derelict Land Survey (SVDLS) — a national record of vacant and derelict land — and filters it down to realistic, council-owned candidate sites, then ranks them for how well they’d suit a drone racing course.

Why council-owned land?

As a registered charity and recognised sport, we can make a credible case to a local authority for recreational use of land they’re not developing. So we filter to council-owned sites and favour land the council has no active plans to build on — that means both more secure tenure for us and an easier ask of them.

What we’re looking for

A good FPV racing site is:

• About the size of a football pitch (~0.7 ha) or a bit larger — we filter to 0.4–3.0 ha.
• Open, flat, vacant ground — not derelict rubble or land with buildings on it.
• Clear of airport flight-restriction zones (FRZ) — we flag proximity to Glasgow and Cumbernauld airports.
• Away from dense housing — for line-of-sight, and to keep separation from uninvolved people (a BMFA/CAA consideration) and minimise noise objections.
• Low contamination risk — former recreation grounds, pitches and amenity land beat ex-industrial sites.

The results

• REPORT.md — the shortlist of standout sites, with map links, reasoning, and what to avoid. Start here.
• outputs/all_scored_sites.csv — every council-owned in-region site (289 of them), scored, with lat/long, map links and airport distances.
• outputs/ranked_table.md — the full ranked table (FRZ sites excluded).
• outputs/satellite/ — annotated satellite views of each shortlisted site (crosshair on the centroid, 100 m scale bar).

Tracking outreach

We use GitHub Issues to track which sites we’ve approached and where each conversation stands — one issue per site. Each issue carries the site’s satellite view, its member-travel report (how far each of us drives to it vs the castle now, and the average), and the Community Asset Transfer contact for the relevant council — the statutory route for a charity to take on council land. See docs/outreach-tracking.md for how it works.

Member travel

Add your postcode to data/members.md (postcode only — no names). We geocode it (postcodes.io) and get real driving distance/time (OSRM), both free and keyless, comparing every site against the current base at Craufurdland Castle. After adding members, refresh the issues with ./scripts/backfill_issues.sh.

Housing proximity (CAA/BMFA separation)

Sites with a home within 50 m of the (approximate) edge are omitted from the shortlist entirely — a hard filter, listed in outputs/excluded_by_housing.csv. The screen also appears on each issue: real building footprints from OpenStreetMap, the nearest home to the site, and an annotated map showing the site’s approximate extent and a separation line. Two honest limits:

• The survey gives only a centroid + area, not a surveyed boundary, so the site extent is approximated as a compact square of that area — a visual aid, not a legal boundary. Verify against a site plan or on the ground.
• The required separation is a BMFA/CAA question. The default is 50 m; the standard Open-category A3 distance from residential areas is 150 m, and a BMFA Article-16 authorisation may differ. Change it with --gap, e.g. python3 analysis/site_proximity.py --code <code> --gap 150 --print.

Reproducing the analysis

Pure Python 3, no dependencies to install.

python3 analysis/find_sites.py

This reads data/Vacant_and_Derelict_Land_-_Scotland.csv and regenerates everything in outputs/.

File	What it does
analysis/find_sites.py	Filters, scores and ranks the sites. All the tunable knobs (region, size band, scoring weights) are at the top.
analysis/osgb.py	Converts the survey’s OS grid references to WGS84 lat/long for map links.
analysis/fetch_satellite.py	Downloads & annotates a satellite view per shortlisted site. Needs requests + Pillow (pip install -r requirements.txt).
analysis/members.py	Member-travel report for a site (geocoding + driving times). Needs requests.
analysis/site_proximity.py	Housing-proximity screen + annotated map (OpenStreetMap buildings). Needs requests + Pillow.
analysis/housing.py	Sweeps every candidate for the nearest home, caching to housing_screen.json (feeds the hard filter). Needs requests.
analysis/basemap.py	Shared satellite-tile fetching and lat/lon↔pixel maths.
analysis/issue_body.py	Assembles a site issue’s full Markdown body (single source of truth).

How the score works

Each site gets a score from five factors — site type, the council’s development intentions, the previous use, size, and member proximity (up to +4 for being close to members’ postcodes). The exact weights live at the top of analysis/find_sites.py; edit them and re-run to re-rank. Higher is better.

Proximity is read from outputs/member_locations.json. After editing data/members.md, regenerate it and re-score:

python3 analysis/members.py --dump    # geocode members -> member_locations.json
python3 analysis/find_sites.py        # re-score with the new members
./scripts/backfill_issues.sh          # update the open issues' scores + travel

The housing hard-filter reads outputs/housing_screen.json, built once by a sweep over all candidates (cached, so re-runs are cheap):

python3 analysis/housing.py --build   # screen every candidate for nearby homes
python3 analysis/find_sites.py        # re-rank, omitting homes-within-50m sites

To use a different separation (e.g. the 150 m A3 figure), change HOUSING_GAP_M at the top of find_sites.py and re-run it — no re-sweep needed (distances are cached out to ~175 m).

Caveats

• The survey is a snapshot. Some sites may since have been built on or reallocated — always verify current status with the council.
• Always check airspace in a drone app (e.g. Drone Assist) before flying anywhere — our FRZ flag only covers the two nearest airports.
• A site visit is essential. The data doesn’t capture surface condition, access, fencing or contamination.

Data source & licence

Source data: Scottish Vacant and Derelict Land Survey, published under the Open Government Licence v3.0. Our analysis code in this repo is released under the MIT Licence (see LICENSE).

This site is open source. Improve this page.