Who built Kruger National Park's digital map, and when?
On Day 14 I showed how much detail OpenStreetMap holds for Skukuza, the biggest rest camp in the Kruger National Park. For Day 21 I wanted to know where that detail came from. OpenStreetMap keeps every edit, so the map has a history: who added each feature, when, and what they changed afterwards.
So the report, Who Built Kruger's Digital Map?, follows the 14,461 buildings, roads, rivers, natural features and visitor places that OpenStreetMap holds today for Kruger National Park back to the year each was first mapped and the person who mapped it. The answer surprised me. The map wasn't built steadily. It was built in two bursts, by a handful of people, and the data still shows the marks of each project.
Explore the interactive report
The map shows the park’s mapped landscape and visitor places, and its 12 SANParks rest camps.
- Colour the map by What, When or Who: feature group, the year the feature was first mapped, or who first mapped it. The same features, three ways.
- Explore area: pick a rest camp, or "Wider park" for everything outside them. The map zooms in, and the cards, charts and story follow.
- First mapped up to: drag the year to see only the features first mapped by then.
- Click a legend bar or a year: the chart beside the map is the legend. Click a bar to show only those features, or a year in the timeline below it.
- Peak mapping year: the busiest year in whatever you've selected, with its share of features.
Interactive report: Day 21, Openstreetmap
What the map does not show: Kruger's map as it looked in a past year. It shows today's shapes of the features first mapped in the chosen years. Dates are edits, not construction, and contributor counts measure recorded activity, not accuracy. It also leaves out about 2,300 features under other tags: power lines and poles, water works and tanks, airfield runways (including Skukuza Airport), the old Selati railway, place names, and 29 monuments and memorials.
Starting with OpenStreetMap's own history
Current OSM extracts only tell you who last touched a feature. To see who created it you need the full history, so I used the ohsome API from HeiGIT, which serves OSM's history from October 2007. Its data runs to 27 July 2026, so that's this report's cut-off.
I asked for every contribution and each feature's latest shape in Kruger, one request per key: buildings, highways, natural, waterways, tourism, amenity, leisure, shops, land use, barriers and entrances. A single request for all keys timed out at ohsome's 600-second limit; separately, most took under a minute, and highways took about 20 minutes. That came to 22 files and 43 MB. I chose these keys for the landscape and visitor places, and left out infrastructure, aviation, railways, place names and heritage. Those deserve their own look.
ohsome gives each edit's changeset number but not who made it, so I looked up all 7,527 changesets through the OSM API, 100 at a time, for the user, editor, comment and imagery source.
The park outline is OSM relation 1752987, first drawn as a simple way in April 2008 and now on its 76th version.
From edits to "first mapped"
What "first mapped" really means
ohsome marks a feature as created when it first matches the filter, not necessarily when it was first drawn. Adding or fixing a tag counts. The termite mounds taught me this: one changeset of 1,589 "creations" turned out to be a typo fix, from termite-mound to termite_mound. So the report says "first mapped with these tags" throughout, and I checked large single-changeset creations against their comments.
Contributors as pseudonyms
Every contributor is a pseudonym: Contributor 001 has the most recorded contributions across the park and camps, and the number never changes between views. Usernames stay in the raw data. OSM history is public, so this is a display choice, not anonymity, and the post doesn't name anyone.
Clipping to the real boundary
My first park download used a simplified outline, about 200 m tolerance. At river bends, it cut across the real boundary, and slivers of neighbouring reserves and farmland showed up along the edge. I rebuilt the map with every feature clipped to the full-detail boundary, which removed 114 features lying wholly outside. Neighbouring reserves still overlapped Kruger's outline by about 1.6 km² (Sabi Sand by 106 ha), so I left them off the map, and kept Makuleke Contract Park, which is part of Kruger.
Finding the rest camps
SANParks lists 23 camp maps. For the 12 rest camps I looked for each camp's fence in OSM. Ten were easy to find by name. Berg-en-Dal and Crocodile Bridge had fenced outlines with no name, and the polygons carrying their names turned out to be just their campsites, 3.8 ha and 0.4 ha. A feature belongs to a camp only if it lies wholly inside the fence, so roads and rivers passing through stay with the wider park.
Building the map in Icon Map Slicer
One table, three colourings
Every map row carries three colours, one for each view. The map's colour is a DAX measure that reads the What, When or Who button and picks the matching column, so switching views needs no extra layers.
A legend you can click
The chart beside the map is the legend. Click "2017" or "Contributor 003", and the map, cards and timeline all follow.
Boundaries as reference layers
The park boundary and the 12 camp fences are GeoJSON overlays embedded in the map, at full OSM detail. The park outline is 1,688 points and 42 KB. Drawing the camp fences as outlines, not data rows, fixed an odd effect: Lower Sabie's fence is tagged as a resort, so it had filled the whole camp green, while Punda Maria's is tagged as forest, so it had filled yellow.
Choosing colours that don't compete
Each view has its own kind of palette, checked with a colour validator before use:
- What: eight categorical colours, with river areas grouped with waterways in blue. A river's centre line and its area polygon are separate OSM features, and they'd been in different groups.
- When: a single orange ramp, light for the earliest years and dark for the most recent.
- Who: purple, teal, pink and yellow for the four named contributors, and grey for everyone else. A colour stays with its contributor everywhere, never re-ranked by area.
Land cover is drawn at 80% transparency, so roads and rivers show through it.
What the map reveals
1. In 2013, one contributor traced the rest camps. In 10 of the 12 rest camps, 2013 is the busiest year and Contributor 003 is the leading first mapper: 77% of Satara, 82% of Pretoriuskop, 86% of Mopani. Much of it happened on one day, 28 July 2013, when three changesets added most of Skukuza's buildings. Park-wide, mapped buildings jumped from 341 to 1,804 that year, and 1,551 of the 2,276 features first mapped in 2013 (68%) are inside the rest camps. Skukuza is the exception: Contributor 006 first mapped 39% of it, close behind Contributor 003's 45%.
2. In 2017–18, two other contributors mapped the wider park. The mapped park nearly tripled, from 4,543 elements at the end of 2016 to 12,336 at the end of 2018. Two people account for 97% of what was added: Contributor 001 (59%: termite mounds, tracks, land use and streams) and Contributor 002 (37%, mostly rivers and their water areas). 52% of today's features were first mapped in 2017. The two rest camps whose busiest year is 2017, Crocodile Bridge and Orpen, belong to this second push.
3. Two-thirds of the park's natural features are termite mounds. 2,542 were added in late 2017 by Contributor 001, in six changesets traced from Bing imagery along the Phalaborwa–Letaba road, next to "adding streams, koppies, pans". I wondered whether they came from a research project. I found no sign of one: the published Kruger mound surveys used airborne LiDAR in the south of the park, and the OSM points carry no survey tags. Ranked by count, termite mounds look like Kruger's main natural feature, but they're one person's project. That's the clearest example of how an individual's interests shape what a map contains.
4. The map's edges need attention. Two rest camp fences are unnamed, four carry odd tags (Satara and Letaba are tagged as a single chalet), and three bushveld camps have no camp outline at all. OSM's park boundary covers 18,994 km², 0.9% less than the 19,172 km² in the government's protected areas database, most of the difference around Orpen. These are leads for contributors, not proven errors.
The caution: activity isn't accuracy. A burst of features in one year shows who was mapping, not how good the map is, and later years are mostly editing. In 2023, about 600 buildings were deleted and redrawn, so they show as "first mapped" in 2023, even though the building count barely changed.
What I learned from Day 21
- History beats a snapshot. A current extract can't tell you who built a map; the full history can, and it changes the story.
- Define "first" carefully. In OSM history, a tag fix can look like a new feature. Check big jumps against their changeset comments.
- Check the boundary you query with. A simplified outline brought in neighbouring reserves and dropped strips of the park. Clip to the detailed one.
- Keep a colour with its entity. Re-ranking contributors per area would have recoloured the same person between views.
- One table can carry three stories. A colour measure and a clickable legend let the same features answer what, when and who.
Final thoughts
Every OpenStreetMap feature carries its own history, and asking who mapped it, and when, turns a map into a record of people's work. The same approach works anywhere OSM matters to you: pull the history for your area, find the bursts, read their changesets, and see whose interests shaped the map you rely on. It's a quick way to see where the map is well cared for and where it needs a second look.
Sources and attribution
- OpenStreetMap data and history: © OpenStreetMap contributors, ODbL (https://www.openstreetmap.org/copyright). History via the ohsome API v1.10.4, HeiGIT (https://api.ohsome.org/v1; https://ohsome.org/copyrights), data to 27 July 2026, retrieved 19 November 2026. Changesets and the park relation (1752987, version 76) via the OSM API 0.6.
- Park boundary comparison: South Africa Protected Areas Database (SAPAD) Q4 2025, Department of Forestry, Fisheries and the Environment, via the public DFFE screening service; used for comparison only, not redistributed.
- Rest camp list: SANParks Kruger camp maps (https://www.sanparks.org/parks/kruger/travel/maps); camp names only; the maps themselves are not reproduced.
- Termite mound research (for comparison): Levick et al. 2010, Nature Communications; Davies et al. 2014, Ecography; Davies et al. 2020, Environmental Research Letters.
Methodology note: Features are OSM elements in Kruger National Park with at least one of 11 keys (building, highway, natural, waterway, tourism, amenity, leisure, shop, landuse, barrier, entrance), in their latest state at the ohsome cut-off (27 July 2026), clipped to the full-detail park boundary (OSM relation 1752987). Neighbouring reserves and the 12 camp fences are excluded from the counts. "First mapped" is the earliest ohsome creation, the first time an element matched these tags, so a retag or redraw counts as new. Each element has one feature group, by tag precedence, and river areas (natural=water) are grouped with waterways. Contributors are pseudonyms numbered by recorded contributions; "first mapped by" is the user of the changeset that created the element. A feature belongs to a rest camp when it lies wholly inside the camp's fence, with a 20 m allowance. The 2017-18 shares come from the park extract, including elements since deleted. 26 large polygons are simplified by 0.6-1.1 m to fit Power BI's text limit. Counts measure recorded mapping activity, not completeness or accuracy.