Overhead.

Where South Africa built the most

Growing means built-up land here, not people. We measured how much of the land around 120 South African towns and cities turned into buildings, roads and other built-up surfaces between 2017 and 2023, using free satellite data that anyone can check. This page is the whole answer — including the places where we could not tell.

Diepsloot changed the most. Roughly 840 hectares of land around Diepsloot became built-up between 2017 and 2023. That is about 26% more built-up land than Diepsloot had in 2017. A hectare is about the size of a rugby field.

The places that changed most

Ranked by how much land became built-up, biggest first. Town pages open in batches between October and December 2026 — a name without a link doesn't have its own page yet, but the figure is the measured one.

  1. Diepsloot840 ha
  2. Somaroboro830 ha
  3. Lebowakgomo670 ha
  4. Soshanguve660 ha
  5. Ekangala570 ha
  6. Mkhuhlu560 ha
  7. Mthatha560 ha
  8. Folweni540 ha
  9. Siyabuswa540 ha
  10. Ulundi490 ha
  11. Dwarsloop440 ha
  12. Rustenburg410 ha
  13. Bhisho400 ha
  14. Lukwatini360 ha
  15. Bloemfontein350 ha
  16. Thohoyandou340 ha
  17. Giyani320 ha
  18. Tzaneen300 ha
  19. Mtubatuba290 ha
  20. Mahikeng270 ha
  21. Harrismith260 ha
  22. KwaDukuza260 ha
  23. Ermelo250 ha
  24. Butterworth240 ha
  25. Vryburg240 ha
  26. Lydenburg220 ha
  27. Middelburg, Mpumalanga220 ha
  28. Makhanda210 ha
  29. Paarl200 ha
  30. Polokwane200 ha
  31. Mokopane190 ha
  32. Wembesi190 ha
  33. Aliwal North180 ha
  34. Balfour, Mpumalanga180 ha
  35. Benoni180 ha
  36. eMalahleni180 ha
  37. Whittlesea180 ha
  38. Bethal170 ha
  39. Dundee170 ha
  40. Standerton170 ha
  41. Wolmaransstad170 ha
  42. Plettenberg Bay160 ha
  43. Lichtenburg150 ha
  44. Parys150 ha
  45. Stutterheim140 ha
  46. Stellenbosch130 ha
  47. George120 ha
  48. Knysna120 ha
  49. Upington120 ha
  50. Kokstad110 ha
  51. Pretoria110 ha
  52. Queenstown110 ha
  53. Schweizer-Reneke110 ha
  54. Bloemhof100 ha
  55. Richmond100 ha
  56. Wesselsbron100 ha

Land cover: Impact Observatory, Esri and Microsoft, CC-BY-4.0. Derived from Copernicus Sentinel data. · Full data: /api/overhead/seo/change.json

The places we don't list

We measured 120 places and publish a figure for 56 of them.

For 23, there was little measurable change — the built-up land around them is much the same as it was at the start. That is an answer, not a gap.

For 41, we say nothing. Either too little changed to be worth your attention, or the yearly readings moved up and down in a way we could not honestly read as a trend. We would rather show you nothing than guess.

We also leave out any place where less than 3% of the land was already built up at the start, because the measurement is unreliable that low down. This year that rule excluded no place — the least built-up place we cover was already well above it.

How we measured this

Where the data comes from. A free, open satellite land-cover map published each year by Impact Observatory, Esri and Microsoft, built from Copernicus Sentinel-2 images. Every place on this page was measured the same way, from the same source, in the same run.

The area. A fixed square, 10 km by 10 km, centred on each place. The same size everywhere, never adjusted to flatter a town. That is why the figures can be compared with each other.

The detail. The map is drawn in squares of 10 m by 10 m of ground. Each square is labelled as built-up, water, trees, crops and so on. We count only how many squares became built-up.

The years. 2017 to 2023. That is the full run of this dataset. These figures are not "today" — 2023 is the most recent year measured, and we update once a year when a new one is published.

Water doesn't count. Coastal and dam-side places would look artificially unchanged if we divided by a box that is half ocean, so we only ever measure against the land in the square.

How accurate is it? The people who publish the map put its overall accuracy above 75%. That is good enough to see a town spreading over six years, and not good enough to tell you about a single street. We round every figure down to the nearest 10 hectares so we never overstate one, and we only publish a place at all if at least 100 hectares changed and its yearly readings moved in one consistent direction.

What we deliberately do not tell you. Nothing about trees, crops, grazing land or any other vegetation. Those categories flicker by up to 40% from one year to the next in this data — not because the land changed, but because the satellite read it differently. Any claim we made about them would be wrong often enough to be worthless, so we make none.

This is a satellite estimate, not a survey. It tells you roughly how much land around a place became built-up. It cannot tell you what was there before, who built it, what kind of buildings they are, or anything at all about a particular address.

Land cover: Impact Observatory, Esri and Microsoft, CC-BY-4.0. Derived from Copernicus Sentinel data. · Full data: /api/overhead/seo/change.json

What this does not tell you

This is not a population ranking. A place can gain thousands of people without gaining much built-up land, and land can be built on without anyone moving in. If you want to know how many people live somewhere, Stats SA is the source, not us.

Where to go next

Open the live Overhead app

Overhead is a free guide built on live open weather data from Open-Meteo (CC BY 4.0). It is advisory only — for official warnings and emergencies, follow the South African Weather Service.