Elevation — how this layer is made

What you are looking at

The height of the land above sea level, everywhere on earth, coloured by height and shaded to show the shape of the ground. Green through tan to brown, and pale grey at the summits — the colouring an atlas has used for a century.

There are no points to click here. Unlike most maps on HueMaps this is not a set of places; it is a value at every pixel, the same way population density is.

Green does not mean green

The colour is height and nothing else. The Sahara is green where it is low, and it is sand. The Amazon and the Congo basin are green because they are near sea level, not because they are forested. Tibet is brown because it is high, and much of it is grassland.

This is the oldest misreading of a hypsometric map and it is worth naming plainly rather than hoping nobody makes it. If it bothers you, the Slate scheme in Settings says the same thing without borrowing the colours of vegetation.

Four ways to colour it

The colour scheme is yours to choose, in Settings. The class breaks never change — the same height is always the same step, whichever you pick — so switching schemes changes how the map reads, never what it says. Each link below opens the map in that scheme.

Atlas is the default: green lowlands through tan to brown, pale grey at the summits. It is the scheme almost every reader has already learnt, and it is at its best on landscape at the scale it was invented for — the western Alps rising out of the Po plain.

Slate does the same job without any colour at all: near-black at sea level, white at the summits. Use it for the same big topography — the same Alps — when you want the map's other information to carry. With the colour channel empty, the rivers stay blue, the contours stay brown and the place names stay black. It is also the scheme to pick if the green-to-brown ramp is hard for you to read: it has no hue to confuse, where Atlas, Bright and Bands all lean on green against brown.

Bright is Atlas with the saturation pushed up, and it earns its place on modest hills that the softer ramp swallows — the ridges west of New York. The two parallel arcs behind Newark and Paterson are the Watchung Mountains, the thin wall along the west bank of the Hudson is the Palisades, and the rougher ground towards Sparta and West Milford is the New Jersey Highlands. None of them reaches 400 m, and all of them are unmistakable here.

Bands is the odd one out and the most useful in flat country. Instead of spreading one sequence over the whole range of land on earth, it repeats the sequence, once between each pair of levels, so any ground with more than a band's worth of relief shows the full run of colour. That makes it the scheme for small features — the valleys feeding the Thames above London, where the dry valleys of the Chilterns branch back into the chalk above Henley, Hambleden and Marlow in a way no ordinary ramp can show.

The cost of Bands is real and is why it is not the default: its colour is not a height. The same green appears at 10 m and at 3,010 m, because it means the bottom of whichever band you are in. Its legend says so — it counts per cent through a band rather than metres.

Measured, not modelled

Every height here was measured by radar from orbit. The Copernicus DEM GLO-30 is derived from the German Aerospace Centre's TanDEM-X mission, which flew two satellites in close formation between 2010 and 2015 and read the shape of the ground from the interference between their two radar returns. Nothing is estimated from a statistic, and nothing is interpolated from a sparser survey.

The measurement is roughly one arc-second — about 31 metres on the ground — and Copernicus quotes better than 4 metres of vertical error over most of the world.

It measures the top of things, not the ground

This is a surface model. The height recorded is the top of whatever the radar hit: over forest that is the canopy, over a city it is the rooflines. In closed-canopy rainforest the difference can be thirty metres or more, and in Manhattan rather more than that.

So a wooded valley and a bare one of the same true depth do not read the same here, and the difference is trees. There is an open dataset that corrects exactly this — FABDEM, which machine-learns the forests and buildings away — and it is licensed for non-commercial use only, which rules it out for a site that intends to offer a paid API. That is a real cost of the licence and not a detail: it is the one weakness of this layer that a better-licensed source would fix.

The sea is not mapped here

Copernicus is a model of the land. There are no tiles over the ocean at all, and the sea is treated as exactly zero.

That is why the colour fades out at the coastline rather than starting hard: a scheme whose lowest colour was opaque would paint every ocean on earth the colour of a lowland. The cost is that genuinely low land fades with it — the Netherlands, the Ganges and Mekong deltas, the Fens — and land at exactly sea level is invisible. Ocean depth is a separate question with a separate source, and it is not on this map yet.

The shading is drawn in your browser

The hillshade — the light and shade that makes ridges and valleys legible — is not a set of pre-made pictures. The tiles carry a height per pixel; your browser reads the slope between neighbouring pixels and lights it from the north-west, live, as you pan.

That means the light is fixed to north rather than to your screen. If it swung round as you rotated the map, the same valley would start reading as a ridge — a genuine illusion, not a matter of taste.

So are the contour lines

The contours are real geometry, not a picture: your browser reads the same terrain tiles and traces the line where the ground crosses each round height. Nothing extra is downloaded for them, and they follow the ground exactly, because they are computed from it rather than drawn over it.

Every fifth line is heavier and carries its height — the index contour, so you can count between them without tracing every line. The interval changes with how far in you are, from 100 metres at a regional view down to a single metre where the national surveys support it, and each new set of lines fades in as you zoom rather than appearing all at once.

They cost your device a little work, since the tracing happens on your machine rather than on a server. On a phone, zooming quickly across mountains is where you would notice it.

The detail is not the same everywhere

The tiles are drawn from a global build that is refined with national LiDAR surveys where a country has published one — Switzerland, France, the United Kingdom, the United States and about forty others have metre-scale data in the underlying archive. Everywhere else is the 30-metre radar baseline.

We use that extra detail where it exists, so zooming into the Alps gets you a genuinely sharper picture than zooming into the Hindu Kush. That is worth stating plainly, because it means a close-up comparison between two countries can be a comparison of their survey programmes rather than of their landscapes. Which government publishes an open LiDAR survey is not a fact about the terrain.

The reason it is worth accepting is that the comparison it distorts is one nobody makes at that zoom. Judging one country's ruggedness against another's happens at a continental view, and every zoom down to the point where the baseline runs out is identical worldwide. By the time you are inside one valley, you are looking at that valley.

Where the baseline is all there is, zooming in softens the image rather than blocking it up. That is a deliberate exception to how this site draws every other layer: a terrain surface is genuinely continuous and smoothly measured, where a statistic reported per district is not, and blocky terrain produces stair-steps in the shading that look like landforms and are not.

What it does not know

The acquisitions are 2010–2015. Terrain barely moves, but this is a surface model, so a district built since then reads as the ground that was there before it.
Radar struggles over steep terrain, still water and dry sand. Accuracy is not uniform, and we have not validated any of it ourselves — the measurement is Copernicus'.
Heights are above the EGM2008 geoid, the orthometric height an atlas quotes, so they will not match the ellipsoid height a GPS receiver shows.
A small number of tiles are withheld from the free Copernicus product for specific countries; the 90-metre product covers the same ground where they are missing.

Source and licence

Elevation from the Copernicus DEM GLO-30, delivered as terrain tiles built and served by Mapterhorn.

produced using Copernicus WorldDEM-30 © DLR e.V. 2010-2014 and © Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved. The organisations in charge of the Copernicus programme by law or by delegation do not incur any liability for any use of the Copernicus WorldDEM-30.

Back to the map