Measured on the same geometry, Ghana and the United Kingdom differ in true area by 0.02% — and the map on your phone draws the UK 2.87× bigger. This lesson is about the data behind every map: what geospatial information is, six ways it already drives decisions in Ghana, and how to design a map that changes one — starting with why you should never fully trust the picture.
Open Google Earth in a new tab and search for your own town. Find your home, your school, or the place you work. Then come back and answer one question: what can you see, from up there, that someone could act on?
Most people list the same things without any training: which roads are tarred, where the empty land is, which roofs cluster together, where the water goes, where the market crowds are. Every one of those observations is the seed of a real application — road maintenance, land planning, settlement services, flood control, market siting.
One phrase, and it's the spine of everything here: collect → map → analyse → act. Geospatial work that stops before the last step is decoration. Keep that test in mind through every section that follows.
A piece of geospatial information is a location on Earth plus something known about that location. 5.6037° N, 0.1870° W is just a point; add "Kwame's pharmacy, open until 9pm" and it becomes data someone can act on. Everything in this field — satellite images, census maps, delivery routes — is stacks of these where+what pairs, and the stacks are called layers: imagery under roads under buildings under flood zones, each one switchable and, crucially, analysable against the others.
Layers come in two shapes, and choosing the right one is half of any project:
| Shape | What it is | Examples |
|---|---|---|
| Raster | A grid of cells, each holding a value — a spreadsheet draped over the ground | satellite images, elevation, rainfall, population density |
| Vector | Geometry with attributes — points, lines and polygons you can name and count | a clinic (point), a trotro route (line), a district (polygon) |
Rule of thumb: measurements of continuous surfaces are raster; things you can count and name are vector. A rainfall map is raster. A trotro route is a vector line. Market locations are vector points.
And the data comes from five kinds of source, each with a trade-off you'll meet again in Section 4:
| Source | Strength | Limit |
|---|---|---|
| GPS receivers & phones | precise points and tracks, nearly free | one place at a time |
| Satellites | repeat coverage of everywhere — much of it free | fixed resolution; clouds |
| Drones | very high detail, flown on demand | small areas; cost; permits |
| Ground surveys | authoritative, rich in attributes | slow and expensive |
| Mobile-phone data | movement patterns at population scale | privacy — aggregate only, properly governed |
Before trusting maps with decisions, test one. Here is a claim you can check: on the map you use every day, the United Kingdom looks much bigger than Ghana. Scroll any standard web map and it's plainly true on screen. So we measured both countries twice, from the same Natural Earth geometry: once on the sphere (true area), and once through EPSG:3857 — Web Mercator, the projection every standard web-map tile is served in.
Squares are drawn to scale with each other in both modes (side ∝ √area). Tap the toggle and watch the UK grow to 2.87× Ghana's drawn area while the ground truth hasn't moved.
The numbers: on the sphere, Ghana comes out at 239,566 km² and the UK at 239,611 km² — a difference of 0.02%. Projected to Web Mercator at the same scale, Ghana occupies 244,455 km² of map while the UK occupies 701,056 km². Same true size; one drawn 2.87× larger. And it's not just the UK — the same measurement across twelve countries:
| Country | True area (km²) | Drawn area (km²) | Drawn at |
|---|
Both columns computed from the same Natural Earth 1:50m geometry, by website/scripts/build-geospatial-data.py. "Drawn at" is the ratio — 1.00 means the map shows the country at its true relative size.
Mercator's projection has one job: keep angles and local shapes true. Do that on a flat map and something else must give — and what gives is area, more and more as you leave the equator. A patch of ground at latitude φ is drawn at sec²φ times its equatorial size. Drag the slider and watch the same square kilometre swell:
The dashed square is the ground truth; the orange square is its share of the map. At Accra the two nearly coincide — 1.01×. At London the map hands the same ground 2.56× the area it gives Accra.
So why does every tiled web map still use it? Three reasons, all good ones. Conformality means north is always up and corners meet at true angles — the property navigation has relied on since the 1500s, and the reason your blue GPS dot turns correctly as you do. The projection maps the world onto a square, which is what makes the whole tile pyramid — zoom level, x, y — work at all. And at any city zoom, the distortion within your screen is a fraction of a percent: for finding a pharmacy in Osu, the map is effectively perfect. The lie only bites when you compare areas across latitudes — which is exactly what every classroom wall map invites you to do.
Now the point of it all. Each card is one domain where geospatial data already earns its keep in Ghana, told the same way: the problem, the layers, the analysis. Before you tap, guess the decision — the specific action, by a specific actor, that the map enables. If you can't name one, the map is decoration.
Cocoa exporters must now prove beans didn't come from deforested land, and extension officers can't visit every farm.
Overlay farm boundaries on imagery and reserves; estimate area — in an equal-area projection, as Section 3 taught — and monitor canopy health per farm.
The same Accra neighbourhoods flood repeatedly along the Odaw channel, and response is reactive every time.
Combine elevation, rainfall and drainage into flood-risk zones; after an event, route relief around cut roads.
Some communities are hours from any clinic — but which ones, exactly, and how many people live there?
Compute travel-time catchments around every facility; flag the communities that fall outside every catchment, with their population counts.
Settlements grow faster than the services that are supposed to reach them, and plans are drawn from years-old base maps.
Change detection between the two dates shows exactly where and how fast the city grew since the plan was drawn.
Illegal mining turns rivers brown faster than ground patrols can find the sites.
Difference the two dates: new bare-earth scars and turbid river reaches locate active galamsey without a single patrol.
Most Ghanaian addresses were historically descriptions — "adjacent the blue kiosk" — which no ambulance or bank can act on.
The digital address system divides the country into 5-metre squares — about 9.5 billion of them — so every location has a code; routing engines do the rest.
Pick one problem from your own town — flooding, clinic access, market siting, waste collection, or something only you know about — and design the map that would move it. Your answers save in your browser, on this device only; nothing is uploaded.
Four things people walk away believing after their first brush with this field. Each card states the tempting version — tap it for what the evidence actually supports.
Everything below costs nothing but time, and each step is a real capability, not a tutorial treadmill. Do the first one today; it takes two minutes on the phone in your hand.
Google Earth on any device: find your home, then trace where the water around it must flow. You are doing terrain analysis by eye.
Install QGIS (free, any laptop), load an OpenStreetMap basemap, and add one layer of your own — the worksheet's, ideally.
The Copernicus Browser serves Sentinel-2 scenes of anywhere. Load two dates of your area and difference them by eye — that's change detection.
KoboToolbox turns a phone into a survey instrument. Twenty GPS-tagged records around your neighbourhood is a real dataset — the full collect → map → analyse → act loop, executed by you.
Ten questions across the lesson. The explanations are worth reading even when you get one right.
Both areas for all twelve countries were computed by website/scripts/build-geospatial-data.py from the same Natural Earth 1:50m geometry (committed with this site): true area on the authalic sphere, drawn area through Web Mercator's own formula at equator scale. Using one dataset for both sides is the control — generalisation cuts both ways, so the ratios survive it. Verified —.
What this cannot claim. Natural Earth 50m geometry generalises coastlines and drops small islands, which is why our sphere areas sit a percent or two from the official figures — the honest comparison of official areas has the UK about 2.1% larger. The country rows for France and Norway are deliberately absent: Natural Earth bundles their far-flung territories (French Guiana, Svalbard) into one unit, and a single inflation number for such a unit would be a lie. And "the web map" here means tiled Web Mercator maps — Google's desktop globe view and equal-area statistical maps are the documented exceptions, discussed in Section 3.
If you filled the worksheet, you already hold a one-paragraph project: a question, its layers, their sources, and the person who acts on the answer. The next explainer teaches nothing you need for step one — open Google Earth and start looking.
More explainers →