Jack Tivey Geospatial Portfolio

Vestige is an iOS app for golfers who want a record of where they've actually played. Every course in England on one map, filling in as you work through them, with progress tracked by county and nationally.

The idea is simple. It just rests entirely on the map underneath it. A course is only as findable and as countable as its boundary is accurate, and there was no usable spatial dataset of English golf courses. So I built one.

Coverage starts as a bulk extract from open mapping data, which gets you most of the country and almost none of the accuracy. Boundaries drawn to no consistent rule, inconsistent tagging, courses missing, entries that closed or relocated years ago. The dataset is the correction pass, not the extract.

County by county, every polygon is checked against satellite imagery and redrawn in QGIS and geojson.io where it is wrong, to a single rule. Operational extent: playable ground plus clubhouse, car park and driving range where present. The legal freehold would have meant guessing, since it is not visible from imagery. Each county is then reconciled against independent listings to catch closures, relocations and duplicates.

That is 1,773 separate decisions about where a golf course stops, made one at a time, across 47 counties. I built the React admin tool that runs the review, partly so the work would be auditable and mostly so I would never have to do it in a spreadsheet. It loads a county, puts each imported feature next to its source tags, and takes keep, delete or merge. I built it with Claude Code, which is how a two-person project ends up with proper tooling instead of good intentions.

Whilst Vestige is my brainchild, it is a shared project. Tom writes the code for the iOS app and I expand and maintain its dataset. In-app features and decisions are always split, and through this we have created something we are very proud of.

Visit vestige.golf →

Vestige is a commercial project operated through Pinehollow Studios Limited, co-founded with Tom Slater.

The question was whether the UK could grow enough oilseed rape to fuel its own aviation. The short answer is no. The interesting part is by how much, and where the land that could is.

Seven criteria drawn from the agronomic literature (precipitation, mean air temperature, soil pH, growing degree days, sunshine hours, drought severity and distance to processing) were reclassified onto a common 1 to 10 suitability scale, weighted by pairwise comparison using the Analytic Hierarchy Process, and combined into a single national surface. Precipitation carried the heaviest weight at 30.5%, temperature 25.3% and soil pH 19.7%. Slope was handled separately as a binary constraint, excluding anything above 15%.

Clipped to land already farmed, the theoretically suitable area came down to roughly 6 million hectares, about half what full domestic demand would need. Which is a real answer, even if it is not the one the title was hoping for. Submitted as Assessing the UK's Agricultural Capability for Producing Sustainable Aviation Fuels (SAFs) from Oilseed Rape.

The method generalises. Narrowing a whole country to the land meeting a defined set of conditions is the same problem whatever the conditions are.

A hazard management plan for the Krakatoa region. Elevation, population density, age structure, purchasing power and distance to hospitals combined into a multi-criteria vulnerability surface, then inundation modelled against it for two eruption scenarios: 13m, which happened in 2018, and 41m, which happened in 1883.

The analysis put Bandar Lampung at the top: 75km north-east of the volcano, 1.16 million people, and a hospital distribution that assumes nobody needs to leave in a hurry. The plan costed the interventions against it. What that work actually taught me had nothing to do with the modelling. It was that a hazard map is read by someone under time pressure who did not ask for a legend with nine classes on it.

Turning spatial analysis into something a non-specialist can act on is most of the job.

Kawsay is a research station two hours downriver from Puerto Maldonado. I spent a month there as a volunteer, which in practice meant doing whatever the researchers needed doing: tagging tree species on ecological walks, carrying seeds and flowers back to camp to be identified, running trail cameras, counting whatever else moved. The camp ran on eight hours of electricity and six hours of WiFi a day. By the second week that starts to feel generous.

The station's long project is spider monkeys. They had been hunted out of the surrounding forest completely, to the point of local extinction. Years of rehabilitating confiscated animals and releasing them into the conservation area have put two distinct wild groups back into it. While I was there we introduced another individual, recovered from the black market, to one of those groups.

Porto Velho, in the remote sensing work above, sits 718km east of the station: the same deforestation frontier, one country over. I have now looked at that forest from both ends: as an NDVI anomaly measured against a six-year baseline, and as a tree I was standing next to with a tag in my hand.

By the time anyone analyses ecological baseline data it is a spreadsheet. It is worth knowing what it costs to collect.