30 Day Map Challenge 2026 - Day 17 (Icon Map Pro) - Light and Dark

How far do you have to drive to escape the light?

South Africa has some of the darkest skies on Earth. Under them you can see things nobody in Europe or North America can see from home: the Magellanic Clouds, and in winter the galactic centre passing straight overhead. But many of the country's population live under the orange domes of Johannesburg and Cape Town.

For Day 17 of the 30 Day Map Challenge, the theme is Light and Dark. I wanted a map that shows both halves of it: the light that hides the stars, and the dark you have to travel to find them. The title of the report sums it up: To See the Light, You Need to Go to the Dark.

There's a commercial angle too. The Department of Science, Technology and Innovation's National Astro-Tourism Strategy positions the Karoo as Africa's premier stargazing corridor. So the report asks the question a tour operator or lodge would ask: which stargazing destinations offer the best balance between dark skies and driving time from Cape Town or Johannesburg?

Explore the interactive report

Pick where you are driving from (Cape Town or Johannesburg) and where you are going (one of five stargazing sites). The map highlights that drive, and the chart below it shows how dark the sky gets every 5 km along the way.

  • Distance slider: drag it to drive the route. A blue marker moves along the road on the map, and the matching marker moves along the chart. The slider reads in kilometres from the start, in 5 km steps.
  • Destination card: the site you picked shows a card with what to do there, its modelled sky, and the sunset and "properly dark" times on new-moon night. The other sites stay as stars; hover over one for its name.
  • Four headline cards: drive time and distance, how far you drive before the sky is properly dark, the modelled sky at the destination, and how much of the drive is under a pristine sky.
  • Sky class buttons: each button shows a class and its SQM range. Click one or more to show only those areas of light pollution on the map, and hover for what you'd actually see under that sky.

Interactive report: Day 17, Light & Dark

One thing to keep in mind as you explore: the sky brightness is modelled from satellite night-time lights, not measured with a meter. It shows where artificial light is likely to brighten the sky. It doesn't show cloud, moonlight, or whether a park lets you stay out after dark.

Starting with night-time lights, and a licence lesson

The obvious source for sky brightness is the Falchi World Atlas of Artificial Night Sky Brightness. It's licensed CC BY-NC 4.0, so it can't be used commercially, and nor can values taken from sites built on it. That ruled it out for this series.

Instead I started from WorldPop's 100 m VIIRS night-time lights for South Africa (2023, CC BY 4.0). It's derived from the Earth Observation Group's VIIRS annual composites, and comes as a single GeoTIFF. I also downloaded 2015 to look for a trend, but it didn't make the final map.

The other inputs were:

  • Drive routes: 10 routes from OSRM on OpenStreetMap roads, from Cape Town and Johannesburg to each of five sites.
  • Town names: 2,067 OpenStreetMap places, exported with Overpass Turbo.
  • A measured reading: the published Sky Quality Meter (SQM) average of 21.6 at !Xaus Lodge, from DarkSky International.

From radiance to skyglow

VIIRS measures light going up from the ground. A stargazer cares about light scattered back down from the sky, called skyglow. A town 80 km away can brighten your horizon even if the ground under your feet is pitch black. So the radiance had to be turned into a modelled sky brightness.

Building a simple skyglow model

I aggregated the 3-arcsecond (about 100 m) raster to 30 arcseconds (about 0.9 km) across mainland South Africa. I read the GeoTIFF tiles with a small Pillow-based decoder, so no GDAL install was needed.

For every cell, I summed the light from every lit cell within 300 km. Each contribution was weighted using a softened form of Walker's law, where glow falls off roughly with distance to the power of −2.5, plus an extinction term so very distant cities fade out.

The total converts to an estimated SQM, measured in magnitudes per square arcsecond. Higher is darker: 22.0 is a pristine natural sky, and every 5 magnitudes is a factor of 100 in brightness.

Checking against a real measurement

Site Modelled SQM Class
!Ae!Hai Kalahari Heritage Park (!Xaus Lodge) 22.00 Pristine
Sutherland (SAAO) 21.99 Pristine
Tankwa Karoo National Park 21.99 Pristine
Cederberg Astronomical Observatory 21.94 Pristine
Kruger far north (Punda Maria) 21.74 Excellent

At !Xaus Lodge the model says 22.0, and the measured average is 21.6. That gap is expected. A real meter also picks up the Milky Way, airglow and zodiacal light, and a model of artificial light leaves all of those out. It's a useful reminder to keep measured and modelled values side by side and clearly labelled. In the report the destination card shows the model as "≈22.0", not "22.00": two decimals would suggest a precision the model doesn't have.

Seven classes, drawn as exclusive rings

I used the following seven SQM class breaks and descriptions:

Class Est. SQM What you see
Pristine 21.8–22.0 Milky Way casts shadows; zodiacal light obvious
Excellent 21.5–21.8 Milky Way richly detailed; faint glow low on the horizon
Dark rural 21.2–21.5 Milky Way clear overhead; light domes on the horizon
Rural 20.9–21.2 Milky Way visible but losing detail
Rural edge 20.4–20.9 Milky Way faint, only high overhead
Suburban 19.4–20.4 Milky Way gone; a few hundred stars
City below 19.4 Only the brightest stars and planets

Pristine is left undrawn, so the dark basemap is the dark sky.

My first version drew cumulative "at least this bright" polygons stacked on top of each other. That looks fine until you filter: select "Excellent" and a purple blob covers Johannesburg. Switching to exclusive rings fixed it. Each class covers only its own SQM range, with holes where brighter classes sit.

Power BI truncates text values at 32,766 characters, so the larger multipolygons are split into WKT chunks of under 29,000 characters. That gives 26 glow rows in total.

A darkness profile for every drive

I sampled the skyglow grid every 5 km along each of the 10 routes, giving 1,890 points. Five kilometres is fine enough: the grid cells are about 0.9 km and town labels sit at least 15 km apart, so sampling every kilometre would mostly repeat itself, and would mean nearly 9,500 marker rows on the map.

I labelled the points with nearby OpenStreetMap places:

  • Cities within 8 km, which catches bypasses like the N1 around Bloemfontein.
  • Towns within 3.5 km.
  • Villages only where they cause a visible brightness spike.

For each drive I then recorded the first point after which the sky stays at 21.5 or darker (Excellent or Pristine) for at least 25 km. The 25 km rule stops a short dark stretch between two towns from counting as "dark".

The OpenStreetMap places export had a few surprises. "Jimmy Jones, South Africa" is tagged as a town, and Giyani and Elim each appear twice in different provinces. OSM is a brilliant resource, but it isn't always complete or tidy, so read your labels before you publish them.

One map, many layers, from a single table

Everything on the map comes from one map_layers table: glow rings, routes, departure cities, sites and the 5 km step markers. Icon Map Pro draws them in category sort order, so a sort_order column controls the layering. Glow bands draw first, with brighter bands on top, then routes, cities, sites, and finally the moving marker.

A handful of measures decide what appears and how:

  • [Show on map] in the visual-level filter keeps only the routes from the selected city, the marker at the selected kilometre, and the glow classes picked on the buttons.
  • [Route width] and [Route colour] make the selected drive bright blue and 5 px wide, and dim the other drives from the same city.
  • [Map image] feeds the image well. It gives the selected destination a star with a gold ring and halo, dims the other stars, and dims the departure pin for the city you didn't pick.
  • [Card title], [Card text] and [Card footer] return blank for every site except the selected one, so only the destination you chose shows its card.

That last one fixed the busiest part of the map. Sutherland, Tankwa and Cederberg sit close together, and with all five cards on screen they piled on top of each other. Now there is one card, and the other sites are just stars.

Designing for the dark

The basemap is Icon Map Pro's Dark Matter style at 78% transparency, with subtle hillshading from Mapterhorn. The hillshade adds the other half of the theme: the shadows of the Great Escarpment and the Drakensberg, and the high plateau where the Karoo's dark skies sit. It's kept faint so the light pollution stays the brightest thing on the map.

The class colours run from deep indigo for Excellent through purple and rust to amber for City. Transparency falls as the sky gets brighter, so the cities glow and the dark country recedes.

The symbols follow one rule each:

  • Blue pin: where you start.
  • Blue line: the road you drive.
  • Blue dot with a white centre: where you are now, on both the map and the chart.
  • Yellow star: where you go to see the stars.

The darkness chart as a second map

Below the map, a native line chart plots estimated SQM against kilometres driven:

  • Class bands: shaded Y reference lines paint the classes behind the line, so you can read the sky class at any point.
  • Two thresholds: dashed lines at 21.5 (dark sky) and 21.8 (pristine, in gold) give the reader a target to watch the line cross.
  • "You are here": a blue X reference line and marker bound to the distance slider, matching the marker on the map.
  • Town and site labels: custom data labels from a [Point label] measure show town names at the dips, plus the destination name at the end.

The dips are the interesting part. Leaving a city isn't a steady climb into the dark: on the drive from Cape Town to !Xaus Lodge the line keeps falling back as you pass Calvinia, Keimoes and, deepest of all, Upington.

A slicer that doubles as the legend

The sky class buttons sit beside the chart. Each one is filled with its map colour via a [Class colour] measure, with text switching to near-black on the two brightest classes. Each button carries its SQM range, and a tooltip with the "what you see" description. The slicer is disconnected from the model and read directly by [Show on map], so filtering the glow never removes the routes, sites or cities.

A layer I took out

I spent a good while on a sunset sweep: a time slider driving Icon Map Pro's built-in Daylight Terminator, so you could watch civil, nautical and astronomical twilight move west across the country on new-moon night. It worked, and it looked good. But it told a second story, about time, on a map whose question is about distance. Its blue bands also sat over the glow rings and made them harder to read. So it came out.

The useful part stayed. Each site card still gives the sunset and the end of astronomical twilight on 9 November 2026, the new moon. Punda Maria is properly dark at 19:35, while Cederberg waits until 20:49: more than an hour's difference across one country.

What the map reveals

Cape Town is well placed. Three pristine-sky options sit within three to five hours:

  • Cederberg Astronomical Observatory, 3h15
  • Sutherland, 4h43
  • Tankwa Karoo, 5h09

On the drive to Tankwa, the sky is properly dark after about 135 km.

The dark arrives long before the destination. On the 1,165 km drive from Cape Town to !Xaus Lodge, the sky is properly dark after 220 km, and 63% of the whole drive is under a pristine sky. That reframes the decision. The question isn't only "which site is darkest?" but "which site gives the best stargazing for the least driving?"

Johannesburg has to work harder. Heading for the Karoo, the sky turns dark after about 130 km and pristine at about 290 km, but those sites are 14–16 hours away. The closest option is far-north Kruger at 6h52, yet that drive only turns dark at 540 km of 559, at the park gate, and never reaches "Pristine". Light from Thohoyandou and the Limpopo towns follows you most of the way.

!Xaus Lodge is the darkest, and the hardest to reach. It sits in Africa's first International Dark Sky Sanctuary, co-owned by the ǂKhomani San and Mier communities. But at over 14 hours from either city, it's a destination, not a weekend.

Why not Durban? I started with it. But modelled darkness is only half the story. The humid, cloudy east coast has far fewer reliably clear nights, so I dropped it as an origin. I also dropped Carnarvon as a site. The town itself models as "Rural edge" and the MeerKAT core is scientists only, though its community-run Carnarvon Astro Guides deserve a map of their own.

The map shows where artificial light is likely to brighten the sky. It doesn't show:

  • Cloud: there's no forecast or clear-night data on the map.
  • The moon: moonlight is left out of the model.
  • Lights beyond the border: Lesotho, eSwatini and neighbouring countries are excluded.
  • Real journey times: drive times are free-flow estimates with no stops, and the gravel roads to Tankwa and !Xaus will be slower.

What I learned from Day 17

  • Check the licence before you fall in love with a dataset. The best-known light pollution atlas is non-commercial. Building a transparent model from open radiance data took longer, but now I can explain every number on the map.
  • Measured and modelled values belong side by side. The 21.6 vs 22.0 gap at !Xaus Lodge isn't an embarrassment. It shows exactly what the model does and doesn't include.
  • Cumulative polygons look great until you filter them. If a layer will be sliced, draw exclusive rings.
  • Not every clever layer earns its place. The twilight sweep was useful and the first thing to go. One question per map.
  • Show one card, not five. Tying the card measures to the slicer cleared the label pile-up without losing any detail.
  • Check your symbol colours against your data colours. An amber city marker on an amber city glow is invisible.

Final thoughts

Light and dark turned out to be a story about distance: how far the city reaches into the sky, and how far you have to travel to escape it. The same pattern works for any value that fades with distance from a city, such as noise, air quality, phone signal or house prices. You need a gridded surface, a set of routes, and a profile chart that shows what changes along the way. Icon Map Pro brings the layers, the draw order and the cards together on one map.

Sources and attribution

Methodology note: Sky brightness is modelled from 2023 VIIRS night-time radiance using a distance-weighted skyglow kernel. The kernel is Walker's law, softened within 2 km, with a 200 km extinction scale and a 300 km radius. It is calibrated so the brightest cell equals SQM 17.0. These figures are estimates of artificial skyglow, not measured sky brightness, and shouldn't be compared directly with meter readings. South Africa only; lights outside the border are excluded. Drive times are modelled free-flow OSRM routes. The sky is sampled every 5 km along each route; "dark sky reached after" is the first point after which the modelled sky stays at 21.5 or darker for at least 25 km. Sunset and end-of-twilight times are for 9 November 2026 (new moon), from NOAA solar position formulae.