Skip to content
3D Elevation Map

Methodology

This page explains how every map and number on the site is produced — the data it starts from, how the figures are computed, and, just as importantly, what the maps cannot tell you. One rule runs through all of it: every number, name and claim traces to a specific source. Where authoritative data is missing, the gap is disclosed rather than filled in.

1. The elevation data

A digital elevation model, or DEM, is a grid of numbers — each one the height of the ground at a single point. Laid over a map, a DEM turns terrain into something you can measure.

Every map here is built from Copernicus GLO-30, the European Space Agency’s open global DEM. It is derived from the TanDEM-X mission — a pair of satellites that measured the Earth’s surface with radar interferometry — and distributed by ESA under the Copernicus programme.

GLO-30 samples the surface at one arc-second spacing, about 30 metres at the equator. Each cell is roughly a 30-metre square of ground, so a single country is represented by anywhere from thousands to hundreds of millions of height samples. It is used here because it is global, openly licensed, and produced to one consistent specification worldwide — which is what makes it possible to render and compare every country on exactly the same terms, instead of stitching together national datasets of differing quality.

One detail matters for the accuracy notes further down: GLO-30 is a surface model. It records the top of what the radar sees — including tree canopy, snow and ice, and in dense cities some built structure — not the bare earth beneath.

2. How the statistics are computed

For each territory the elevation grid is clipped to that place’s official outline (see Boundaries). Only cells inside the boundary count. From those in-boundary heights the site computes the mean, the median, the full range from lowest to highest, and the elevation distribution — the share of the land lying above each threshold (above 500 m, 1,000 m, 2,000 m, and so on).

Both the mean and the median are reported, because they answer different questions. The median is the typical elevation — half the land lies above it, half below. The mean is the average, and a few very high peaks pull it upward. Bolivia, for instance, has a mean elevation over 1,200 m but a median around 300 m: most of the country is low, and a high but small-area Andean region lifts the average. Reading the two together shows whether a place is genuinely high, or simply has high ground in one corner.

These are first-party figures, calculated here directly from the elevation grid — not copied from another site. That is the core of what this project does: it measures, rather than repeats.

3. Highest & lowest points

A 30-metre grid averages a sharp summit downward — a single peak is smaller than the cell it sits in. So the highest point is not simply read off the grid. The model is re-read at its native resolution around every candidate high zone, and each candidate is anchored to the researched coordinates of the territory’s known high point rather than to polygon containment. A generalized boundary cannot cleanly split a summit that sits on a border — Mount Everest on the Nepal–China line, or Ojos del Salado on the Chile–Argentina line — so the coordinate anchor, not the polygon, decides.

The refined reading is then reconciled with an authoritative, source-cited survey figure. The two usually agree closely. Where they do not, the site is explicit about it — and that transparency is deliberate.

Where the model reads low

A sharp or snow-and-ice-covered summit often reads a little lower in the model than its surveyed height: the grid smooths the point, and radar behaves differently over ice than over rock. In these cases the site headlines the authoritative surveyed figure and notes that the model reads lower — it does not quietly publish the under-read as if it were the summit.

Where the model reads high

Where the model reads higher than the survey — typically a low, forested summit where the radar sits on the tree canopy rather than the ground — the surveyed bare-earth figure is used, and the difference is understood as canopy, not terrain.

A page is published only after its high point passes this check against the researched figure; if the two cannot be reconciled, the page does not ship. Lowest points are handled the same way, including genuine below-sea-level land — the shore of the Dead Sea, the floor of Death Valley, the Dutch polders — which is reported at its real negative elevation, not clamped to zero.

4. Boundaries

Territories are defined with geoBoundaries gbOpen, an open, academically maintained global boundary dataset. Countries use its ADM0 (national) outlines and US states its ADM1 outlines, so every territory is defined on consistent terms from a single source.

Boundaries are generalized. At the 30-metre grid scale a coastline or a border is an approximation — drawn slightly inland of, or outside, the true line in places. This mostly affects the very edge of a map, such as a fjord coast or a river border, and those edges are handled gracefully rather than presented as pixel-exact.

5. The 3D rendering

The elevation grid becomes a 3D terrain mesh — a surface whose height at every point is the measured elevation. It is coloured by a hypsometric tint: a fixed elevation-to-colour ramp running from greens at low elevations, through tans and browns, to white at the highest ground. Because the ramp is fixed and shared, a given colour means a similar height on every map. A multidirectional hillshade — simulated light from several angles — is baked in so that slopes, ridges and valleys read as relief rather than flat colour.

Vertical exaggeration

Terrain heights are scaled up so the shape is legible. A country can be hundreds of kilometres wide but only a few kilometres tall, and at true scale even mountain ranges look almost flat. Depending on how much relief a place has, heights are exaggerated by roughly 6× for the most mountainous places up to about 25× for the flattest, and a slider lets you change it. The exaggeration is only visual — the elevation numbers on every page are always the real, unexaggerated values.

Water

The sea is drawn flat, at sea level. Major lakes are drawn at their own surface elevation rather than at sea level, so a high lake such as Titicaca (about 3,800 m) or a below-sea one such as the Caspian (about −28 m) sits at its true height in the scene.

6. Accuracy & limitations

Copernicus specifies GLO-30’s vertical accuracy at a few metres for the raw model, but the terrain shown here should be treated as accurate to the order of tens of metres. Three things widen the margin: the model is a surface, not bare earth, so dense vegetation or ice can add to a reading; a 30-metre grid smooths features smaller than a cell; and large territories are downsampled before rendering.

Adaptive resolution

To stay responsive in a browser, very large territories are rendered from a coarser grid — a whole continent, or a country the size of Russia, cannot be drawn at full 30-metre detail without an enormous download. Small and mid-sized countries render at full working resolution; the largest are adaptively generalized. The elevation statistics and the refined high point are always computed at full resolution — only the 3D mesh is coarsened — but a giant map’s surface is visibly smoother than a small one’s.

Voids and artifacts

Radar DEMs contain occasional voids and artifacts — gaps where the signal failed, or noise over water and very steep terrain. Known gaps, such as a block of missing tiles over the Caucasus, are filled from a second open elevation model where they fall inside a territory, and that fill is disclosed.

Do not use these maps for

Navigation, aviation, engineering, construction, surveying, boundary determination, safety-of-life or legal purposes. These are terrain models for understanding the shape of the land — not survey instruments, and not a substitute for authoritative data where a decision depends on it.

7. Disputed & sensitive territories

Some borders are disputed. This site maps terrain, not politics. It follows the gbOpen boundary as published, takes no position on any territorial dispute, and describes only the land within the mapped area.

Where a boundary is contested, the site uses neutral, descriptive language: it reports the terrain inside the mapped extent, names features factually, and notes relevant context — for example, that an area is administered by one state and claimed by another — without endorsing any claim. Where a well-known high or low point falls just outside the mapped extent, that is disclosed rather than quietly included or excluded.

This is a deliberate editorial discipline, applied consistently: describe the ground, cite the source, and leave the sovereignty question to the parties and to the international bodies whose role it is.

8. Sources & attribution

The maps and figures are built from open data:

Above all, the site follows a verified-or-omitted principle. Every figure traces to one of two places: a value computed here from the elevation model, or a researched fact with a cited source. Nothing is filled in from memory or generalized from what is usually true. Where an authoritative figure does not exist, the site says so instead of inventing one. A shorter page is treated as correct; an embellished one is treated as a mistake.

The site is built and maintained by Marko Visic — more on the author and the project on the About page.

Elevation data: Copernicus GLO-30 (ESA). Boundaries: geoBoundaries. Reference layers: Natural Earth, GeoNames. Built by Marko Visic.