mapbench

Methodology

Every number MapBench shows you has a documented recipe. This page lists them — and their honest limits.

CalculationMethod & accuracyTool
Distance (haversine)Great-circle distance on the WGS84 mean sphere (R = 6371.0088 km). Error vs. the ellipsoidal geodesic ≤ ~0.3%.Open →
BearingInitial great-circle bearing via atan2(sin Δλ·cos φ₂, cos φ₁·sin φ₂ − sin φ₁·cos φ₂·cos Δλ), clockwise from true north. Back-bearing computed independently, not ±180°.Open →
MidpointSpherical midpoint of the great-circle arc — correct across hemispheres and the date line, unlike naive coordinate averaging.Open →
Polygon areaSpherical-excess integration over the boundary (equivalent to the standard geodesic area approximation). Accurate to well under 1% for real-world plots; handles high latitudes.Open →
Radius circles128-vertex boundaries, each vertex placed by spherical destination-point projection, so circles stay true at any latitude.Open →
Routing & travel timeFOSSGIS Valhalla on OpenStreetMap roads (auto/pedestrian/bicycle costing), OSRM fallback for driving. Durations are free-flow estimates — never presented as live traffic.Open →
IsochronesValhalla's routing-engine isochrone polygons: genuinely network-derived reachable areas, not buffers.Open →
Stop optimisationValhalla's optimize endpoint (TSP heuristic) with the first stop fixed as origin.Open →
CoordinatesFull WGS84 UTM series expansions including the Norway/Svalbard zone exceptions; standard MGRS band/grid letters; Open Location Code (Plus Code) encode/decode.Open →
Sun & moonNOAA solar position algorithm (±1 min for modern decades) with the standard 90.833° sunrise zenith for refraction; lunar phase from synodic-month age.Open →
ElevationOpen-Meteo elevation API on the Copernicus GLO-90 DEM (~90 m). Profiles use 60 great-circle samples; line-of-sight adds Earth bulge x(D−x)/(2R′) with refraction k = 0.13.Open →
Population in radiusSum of curated major-city populations (≈2020 vintage, municipal figures). Published as a transparent estimate with the contributing city list, never as a census count.Open →

Known limitations

  • The haversine sphere differs from the WGS84 ellipsoid by up to ~0.3%. Survey-grade work needs ellipsoidal geodesics and a defined CRS.
  • Free routing servers return free-flow travel times; congestion is not modelled.
  • The 90 m elevation DEM cannot see buildings or tree cover — line-of-sight results are terrain-only.
  • Curated population datasets lag behind real growth; vintages are always labelled.

If you find a case where a result disagrees with a trusted reference by more than the stated tolerance, that's a bug we want to hear about — reproduction steps (coordinates in, values out) make it fixable.