Which South African towns did mining build, and which still depend on a commodity that is running out?
For Day 6 of the 30 Day Map Challenge, the theme is Vintage. After Day 5's sightlines to Table Mountain, I wanted to use an original map I own: SÜD-AFRICA, an engraved and hand-coloured atlas sheet published by Carl Flemming in Glogau around 1850.
It shows the Cape Colony, Natal and the new Boer republics, but no Kimberley, no Johannesburg and not a single railway. Almost everything that made the interior's towns came after it was printed. So the map is the perfect blank canvas to watch the mining towns appear.
The question behind it is a commercial one. Many South African towns exist because of one commodity. Gold mines are closing, and coal faces the energy transition. Which towns' economies rest on a single commodity, and how exposed are they now? That matters to banks, property investors, retailers and municipal lenders.
Explore the interactive report
The map opens on the 1850 sheet with 21 towns that already existed when it was printed, from Cape Town (1652) to Lydenburg (1849), as well as 70 mining towns. You can use the Play Axis to watch 70 mining towns appear decade by decade, from the Namaqualand copper towns of the 1850s to the coal towns of the 1970s and 1980s.
- Colour is the commodity, size is the number of people living nearby today, and shape is how the town began: ✖ founded for mining, ⭘ an older town that grew with mining, ■ built for a coal-to-fuel or power plant. ◉ marks a town of c.1850.
- Hover over a town to see when and why it began, how many people live within 10 km today, and its exposure.
- Pick a commodity with the buttons under the map, or an exposure class from the dropdown. The Show dropdown switches between all towns, mining towns only and the towns of c.1850.
- The column next to the map shows the sheet's two insets: a town plan of Cape Town and a larger-scale map of the south-western Cape.
Interactive report: Day 6, Born of the Mine
The exposure measure is illustrative. It combines the commodity's outlook with whether an active mine is still nearby. It isn't a forecast or a credit rating, and it has no data on jobs or municipal revenue.
Starting with a map from my own wall
This map came to me from my grandparents, and it lives framed on my wall. Using it for the challenge adds a very personal touch: instead of downloading a scan from an archive, I'm mapping on something that has been in my family.
The printed map is long out of copyright, so the only work was getting a good image of it. I photographed it straight on, in even light, with a mobile phone at its full 50 MP (6,120 × 8,160 pixels).
At about 175 years old, the paper shows its age. There are tide-line water stains in the corners, foxing across the interior and a general yellowing. The two insets also sit on top of the main map, over Namaqualand and over the sea off the Eastern Cape.
Cleaning a 175-year-old sheet
I wanted to remove the damage without redrawing anything. Three steps did most of the work.
- Flattening the paper. I removed the ink with a maximum filter, blurred what was left to get the paper's own light and stains, and divided the photo by it. The tide lines and the uneven light even out, while the engraving and the colouring stay as they were. The result is re-tinted to a warm cream.
- Cutting out the insets. The Cape Town plan and the south-western Cape inset were cut out, corrected for perspective and saved as separate images. On the main map, their windows were filled with plain mottled paper, and the 2° grid was continued through them. No coastline or place names were invented.
- Removing the boundary washes. The sheet's green, yellow, orange and pink boundary colours fought with the commodity colours. Each washed pixel was compared with the local wash colour, so pure wash became paper while the engraved coast, rivers and names underneath stayed as ink. This colouring was probably applied by hand when the sheet was issued, so it's original, and the insets keep theirs.
Georeferencing a map measured from Paris
The map's longitudes are counted east of Paris, not Greenwich. Cape Town sits near the 16° line, not 18.4°.
- Graticule. I traced the 2° grid automatically: 9 meridians and 6 parallels, each fitted as a curve. That gave 43 crossings outside the insets.
- Warp. A third-order polynomial maps Paris longitude and latitude to the photo's pixels. The fit has a mean error of 1.4 pixels, about 0.5 km on the ground.
- Paris to Greenwich. I added 2° 20′ 14″ to every longitude.
- Check. Cape Town, Stellenbosch, Port Elizabeth, Grahamstown, Durban and Graaff-Reinet all fall within a few kilometres of where the map draws them. The rest of the offset is the 1850 survey itself.
The result was reprojected to Web Mercator and trimmed to the largest rectangle that lies wholly inside the neatline. It was saved as a Cloud Optimised GeoTIFF: 2,850 × 2,259 pixels and 15 MB.
Finding the towns that mining built
I wanted towns that exist because of mining, not every town that has a mine nearby.
- Candidates. I started from Wikipedia's Mining communities in South Africa category and added about 35 known mining towns missing from it. Coordinates come from Wikidata, which is CC0. Founding years were checked against each town's history; only facts were used, not text.
- Founding years. Wikidata's "inception" is often wrong for this purpose. For example, it gives Dundee 1835 against 1882, and Musina 1968 against 1904. So every year was set by one rule and recorded with its source.
- How each town began. Each town was classified in one of three ways:
- 57 were founded for mining.
- 15 are older towns that grew with mining, such as Klerksdorp and Rustenburg; for these, the year is when mining began.
- 3 were built for a plant: Sasolburg and Secunda for Sasol's coal-to-fuel works, and Kriel for an Eskom power station.
People and mines today
Population comes from the JRC's GHS-POP 2025 grid, at about 1 km resolution. On the Witwatersrand, 12 of the towns are within 10 km of another, so I counted each person only once, for their nearest mining town. In total, about 8.3 million people live within 10 km of these 70 towns.
Mines come from the USGS Mineral Industries and Related Infrastructure of Africa compilation. It lists 290 facilities in South Africa, with their status as of 2018. 50 of the 70 towns still have an active mine of their founding commodity within 30 km.
An illustrative exposure measure
Each town gets one of five classes. The class depends on the outlook for its commodity and on whether its mine is still there.
| Class | Rule | Towns | People nearby |
|---|---|---|---|
| High | Active gold or coal mine within 30 km | 24 | 2.5 m |
| Medium | Active diamond, platinum or chrome mine | 12 | 0.95 m |
| Low | Active copper, iron ore, manganese or other mine with steady or growing demand | 6 | 0.15 m |
| Mine gone | No active mine of its commodity within 30 km | 20 | 0.28 m |
| Diversified metro | Johannesburg and seven other East and West Rand towns, where mining is now a small share of the economy | 8 | 4.4 m |
The old map as a raster overlay in Icon Map Pro
Icon Map Pro's raster overlay reads a Cloud Optimised GeoTIFF straight from a URL. I uploaded the COG to Azure Blob Storage, with CORS allowed, and pointed the overlay at it. The browser only fetches the tiles it needs, so a 15 MB sheet opens quickly and stays sharp when you zoom in.
Locking the view to the sheet. Beyond the edge of the sheet there is only the blank basemap, which spoils the effect. Icon Map Pro can limit how far the reader pans and zooms, so I locked the view to the sheet's bounds, with a little extra room at the top for Musina's label.
The map took the shape of the sheet. I then resized the map visual to 990 × 873, the sheet's own proportions at that view, so it fills the visual edge to edge. That freed a column beside it for the two insets as image visuals, so the whole sheet is on the page.
Building the towns up decade by decade
The Play Axis custom visual steps through a small table of decades, from the 1850s to the 1980s. That table isn't related to the towns. Instead, a measure reads the selected decade, and the map keeps any town founded by the end of it. So the towns build up decade by decade rather than flashing past. The cards and the commodity chart follow the same decade.
Symbols that match their labels
My first markers were SVG images built by a DAX measure. Each hover card had the town name as its title, and I wanted the card to show the town's commodity too. The answer was Unicode symbols. Icon Map Pro's image layer can draw a character instead of a picture, coloured and sized by measures. The same character can then start the card's title, such as "✖ Barberton", where it takes the title's colour. So the symbol on the map and the symbol on the card always match.
There are two things to watch:
- Size. A symbol is sized like text, so a ● fills only about half of its font size. The symbols needed roughly double the pixel size of the old circles.
- Weight. Thin characters disappear on old paper. ○ was too faint for the growth towns, so I swapped it for the heavy ring ⭘. I also changed the founded towns from ● to ✖, so they couldn't be mistaken for the 1850 towns' ◉.
Keeping the old map on screen
If a filter left the map with no rows, for example coal towns in the 1850s, Icon Map Pro hid the raster overlay as well. The vintage map vanished just when it mattered most.
The fix had two parts:
- Measures do the filtering. The commodity and exposure slicers no longer filter the towns table directly. They use their own tables, and the measures check the selection, so the 1850 towns are never filtered out.
- A hidden placeholder. One extra row sits in the sea south of the Cape. It is drawn as a blank 1 px symbol, has no card and isn't counted anywhere, so the map always has at least one row.
Design choices
Colour. Each commodity has one colour, used everywhere:
- the symbols on the map;
- the hover card titles;
- the buttons under the map;
- the bars in the chart.
The buttons double as the legend. Gold is a deep ochre rather than yellow, so it reads on cream paper, and coal is near-black.
Shape before decoration. On a busy old map, shape does more work than outline or shading. Four distinct characters are enough to tell founded, grown, industrial and c.1850 towns apart, even at small sizes.
The towns of 1850. The 21 older towns are drawn in sepia, close to the map's own ink. They make it clear how empty the interior was before mining, and they are reference points rather than data, so they ignore the commodity and exposure filters.
What the map reveals
- Copper came first, and it has gone. The first towns are Springbok and Concordia in the 1850s, then Nababeep and Okiep, all in Namaqualand. None of the four has an active copper mine nearby today.
- The 1880s changed the map. Fourteen towns appeared in a single decade. Eleven were gold towns, on the Witwatersrand, at Klerksdorp and in the Barberton and Graskop fields. Johannesburg (1886) was founded on land that the 1850 sheet leaves almost blank.
- Gold built more towns than any other commodity, and most of them now face decline. 26 gold towns have about 6 million people living nearby:
- 8 have grown into diversified metros;
- 14 are rated high exposure, including Welkom, Carletonville and Klerksdorp;
- 4 have lost their mine altogether.
- Coal is the newest exposure. Ten of the 11 coal towns, from Witbank to Secunda and Lephalale, still have an active coal mine within 30 km. They face the energy transition directly.
- Twenty towns have already outlived their mine. About 280,000 people live near towns such as Pilgrim's Rest, Barkly West, Alexander Bay and Musina, whose founding commodity is no longer mined nearby.
These are mapped founding stories and modelled populations, not an economic assessment. The mine status is as of 2018, and 30 km is only a proxy for dependence. Jagersfontein's tailings dam failure in 2022, for example, isn't reflected.
What I learned from Day 6
1. Photograph old maps out of the frame. Glass adds glare and perspective, and no amount of processing fully removes them. A flat sheet, even light and a modern phone camera gave a usable 50 MP image.
2. Check the prime meridian. Many 19th-century European maps count longitude from Paris or Ferro, not Greenwich. Using the map's own grid lines as control points makes the conversion a single, exact shift.
3. Clean, don't redraw. Dividing by the paper's own background removes stains and uneven light without touching the engraving. Anything the map didn't show stays empty.
4. Use symbols when the label must match the marker. A Unicode character can be coloured like text, so the same symbol on the map and in the label always agrees.
5. Keep a row on the map when an overlay matters. Filtering through measures, plus one hidden placeholder row, stops the overlay disappearing when a filter matches nothing.
Final thoughts
Vintage maps are often treated as decoration, but they make an honest baseline: a picture of what was there before. Laying today's data over a sheet from 1850 shows something a modern basemap can't show, the empty interior that mining filled. The same method works for any old map you own or can license: photograph it flat, use its grid to georeference it, serve it as a COG, and let a Play Axis show what changed since it was printed.
Sources and attribution
- Base map and insets: SÜD-AFRICA, Druck u. Verlag von C. Flemming, Glogau, c.1850. Original sheet from the collection of Antony Catella, used with permission. Photographed, cleaned and georeferenced for this report. The printed map is in the public domain by age.
- Town coordinates: Wikidata, CC0 1.0, queried November 2026.
- Founding years: set by the author from town histories, including Wikipedia articles. Only facts were used.
- Population: Schiavina, M., Freire, S., Carioli, A., MacManus, K. (2023), GHS-POP R2023A, GHS population grid multitemporal (1975–2030), European Commission, Joint Research Centre, 2025 epoch at 30 arcseconds. doi:10.2905/2FF68A52-5B5B-4A22-8F40-C41DA8332CFE. Licensed under CC BY 4.0.
- Mines: Padilla, A.J., and others (2021), Compilation of Geospatial Data (GIS) for the Mineral Industries and Related Infrastructure of Africa, U.S. Geological Survey data release. doi:10.5066/P97EQWXP. Public domain. Facility status as of 2018.
- Background reading for the exposure classes: Cole, M.J. and Broadhurst, J.L. (2020), Mapping and classification of mining host communities: A case study of South Africa, The Extractive Industries and Society 7(3), 954–964.
Methodology note: The 70 towns are those founded for mining, older towns that mining remade (dated from when mining began) and towns built for coal-to-fuel or power plants. Five towns without a reliable year are excluded. Population is the modelled GHS-POP 2025 count within 10 km of each town point, with each person assigned to the nearest mining town. A town has an active mine if the USGS compilation lists an active mine of its founding commodity within 30 km, as of 2018. Exposure classes are an illustrative combination of that test and a broad commodity outlook. They are not forecasts and do not reflect jobs, municipal finances or events after 2018. The historic base map is georeferenced to a mean accuracy of about 0.5 km; remaining offsets reflect the 1850 survey.