The shape of the Earth
Every measurement has to assume some shape for the planet. These are the three assumptions in common use, in increasing order of realism.
- Sphere
- A perfect ball. Wrong, but wrong by less than half a percent, which is why most everyday distance calculations — including the great-circle figures on this site — use one and are perfectly adequate.
- Ellipsoid
- A sphere squashed at the poles. The Earth bulges at the equator by about 21 km, and an ellipsoid captures that with two numbers. It is smooth and mathematical, which makes it the surface satellites compute against.
- Geoid
- The lumpy surface that gravity actually produces — where the sea would settle if it could flow freely everywhere. It deviates from the ellipsoid by roughly +85 m to −106 m depending on where you stand. This is what 'sea level' means.
- WGS84
- The World Geodetic System of 1984: the specific ellipsoid and coordinate frame that GPS, web maps and almost everything else now use. When a latitude and longitude is quoted without qualification, it is usually WGS84.
- Datum
- The chosen reference surface plus its anchoring to the real world. Two coordinates from different datums can name points hundreds of metres apart despite reading nearly identically — a classic source of silent error when mixing old survey data with modern GPS.
Position
How a place gets turned into numbers, and how much those numbers are really telling you.
- Latitude and longitude
- Angles, not distances: latitude north or south of the equator, longitude east or west of Greenwich. A degree of latitude is about 111 km everywhere; a degree of longitude shrinks to nothing at the poles. The full table is here.
- Decimal degrees
- Writing 46.0569° rather than 46° 3′ 25″. Easier to compute with, and the format nearly every mapping tool expects. Six decimal places resolves about 11 cm — well past what consumer GPS can justify.
- Arc-minute and arc-second
- A sixtieth and a 3,600th of a degree. One arc-minute of latitude is the original nautical mile. Gridded datasets are often specified in arc-seconds — a 30 arc-second grid has cells about 928 m tall.
- Geocoding
- Turning a place name into coordinates. Necessarily lossy, because a city is an area and a coordinate is a point, so something has to choose a representative spot — which matters for short distances.
Distance and paths
There is more than one sensible answer to 'how far is it', and the words below name the differences.
- Great circle
- The shortest path between two points across a sphere’s surface — the arc a taut string would trace over a globe. Every distance on this site is measured this way. It is why long flights appear to arc north on a flat map; the size of that effect is tabulated here.
- Rhumb line
- A course of constant compass bearing, which is exactly what a straight line on a Mercator map represents. Easier to steer than a great circle and reliably longer — by 4% across the North Atlantic, and by 36% from Anchorage to Oslo.
- Geodesic
- The shortest path on whatever surface you are actually using. On a sphere it is the great circle; on an ellipsoid it is a slightly different curve requiring more work to compute. 'Geodesic area' means area measured on the curved surface rather than on a flat projection.
- Crow-flies distance
- Everyday name for the straight-line distance, ignoring roads, terrain and water. Always a lower bound on any real journey.
- Detour factor
- The ratio of actual route distance to crow-flies distance, also called circuity. Typically 1.2–1.3 in a regular city grid, 1.4–1.6 in cul-de-sac suburbs, and unbounded where there is water in the way.
- Isochrone
- The region reachable within a given travel time. Almost never a circle — it stretches along fast roads and collapses where the network is poor, which is the crucial difference between it and a radius.
Projections
Flattening a curved surface always destroys something. Which thing gets destroyed is the choice a projection makes.
- Projection
- Any scheme for putting a curved surface onto a flat one. No projection preserves shape, area, distance and direction simultaneously — this is a proved impossibility, not an engineering limitation.
- Web Mercator
- The projection used by essentially every online slippy map, including this one. It preserves local shape and compass bearings, at the cost of inflating area with latitude.
- Scale distortion
- In Web Mercator, lengths are exaggerated by 1/cos(latitude) and areas by its square. At 60° that means distances double and areas quadruple — the reason Greenland looks continental on a world map.
- Conformal
- A projection that preserves angles and local shapes, which is what makes Mercator good for navigation: a constant bearing really is a straight line. Conformality is precisely why area cannot also be preserved.
Terrain and elevation
What 'how high is it' turns out to mean on closer inspection.
- DEM
- Digital Elevation Model: a grid of height values covering the ground. Elevation lookups here read from the Copernicus DEM, which has roughly 90 m spacing.
- DSM and DTM
- A Digital Surface Model includes whatever sits on the ground — trees, rooftops. A Digital Terrain Model is the bare earth beneath. Confusing the two explains most cases where a forest or a city appears mysteriously tall.
- Orthometric height
- Height above the geoid, i.e. above sea level. The convention on topographic maps, road signs and mountain summits.
- Ellipsoidal height
- Height above the ellipsoid, which is what a GPS receiver computes natively. Converting to orthometric height requires a geoid model; skipping that step leaves a discrepancy of tens of metres.
- Elevation gain
- The sum of all the uphill sections of a route, ignoring the downhills. Highly sensitive to how finely the route is sampled, which is why a GPS watch reports more climbing than a coarse profile.
Gridded data
Several of these tools read from global grids. The vocabulary of grids governs what their answers can mean.
- Resolution
- The ground size of one cell. It sets the finest detail obtainable: asking a 1 km grid about a city block gets an answer about the kilometre containing it, not about the block.
- Dasymetric mapping
- Redistributing a statistic within its reporting unit using some other signal — for population, typically satellite-detected buildings. It produces more realistic detail at the cost of depending on the model that generated it.
- WorldPop
- A global population dataset modelled at about 100 m, from the University of Southampton. Dasymetric: census totals are redistributed within each district using buildings, roads and land cover. Used by the population tool.
- GPW
- Gridded Population of the World, from NASA SEDAC. Deliberately not dasymetric: it spread census counts uniformly across each administrative unit, keeping the output traceable to real census returns at the cost of ~1 km resolution. Its live query service went offline when SEDAC was decommissioned in 2025.
- Orthorectification
- Geometrically correcting aerial or satellite imagery so that every pixel sits at its true map position despite terrain and camera angle. Done to a tolerance, which is the floor on how precisely you can trace anything from imagery.
Related: distance and area units and how long is a degree of latitude.