The short answer

For trustworthy polygon areas, use an equal-area (equivalent) projection whose area of use covers your region — never Web Mercator and never raw WGS 84 degrees. The three workhorse choices are a country or continental Lambert Azimuthal Equal Area (LAEA) grid, a regional Albers Equal Area Conic, and, for small extents that fit one zone, UTM (which is conformal but distorts area by well under 0.1% inside a single zone). If you would rather not choose a projection at all, compute geodesic area on the ellipsoid directly.

The reason this matters is geometric, not cosmetic: no flat map can preserve shape, distance, and area simultaneously. You must pick which property to protect, and for area measurement you protect area.

Why projections distort area

Projecting the curved Earth onto a plane stretches some directions and compresses others. The local scale factor describes that stretch. In a conformal projection (Mercator, Lambert Conformal Conic, UTM) angles are preserved, so the scale factor is equal in all directions at a point — but it grows with distance from the projection's standard lines, and because area scales with the square of linear scale, area error compounds quickly.

Web Mercator is the extreme case. Its scale factor is 1/cos(latitude), so area is inflated by 1/cos²(latitude):

  • At 45°N, area is overstated by about 2x.
  • At 60°N, by about 4x.
  • At 70°N, by about 8.5x.

That is why a forest polygon in Scandinavia digitised over a Web Mercator basemap and measured in EPSG:3857 can report several times its real hectares. An equal-area projection instead fixes the area scale at 1 everywhere, distorting shape and angles to compensate — which is exactly the trade you want when the output is a number of hectares.

Choosing among equal-area options

Lambert Azimuthal Equal Area (LAEA)

Best for compact or continental extents centred on a point. The pan-European reference is ETRS89-LAEA, EPSG:3035, the standard for EU area statistics and the 1 km reference grid. For a national or sub-continental study with no official equivalent, define a custom LAEA centred on your region's centroid.

Albers Equal Area Conic

Best for regions wider east–west than north–south at mid latitudes — countries, states, large basins. EPSG:5070 (NAD83 / Conus Albers) is the conventional choice for the contiguous United States. The two standard parallels should bracket your area of interest (a common rule is to place them at about one-sixth and five-sixths of the latitude span) to minimise the residual linear distortion, though area remains exact by construction.

UTM

UTM (EPSG:326xx / 327xx) is conformal, not equal-area, but within a single 6°-wide zone the area error stays under ~0.1%, which is negligible for most parcel- and field-scale work. Use it when the polygon fits inside one zone and you also want sensible distances. Avoid it for anything spanning multiple zones — stitching across the zone boundary reintroduces error.

Worked example: area of a polygon three ways

Take a regional habitat polygon stored in EPSG:4326. In QGIS, add a field with the Field Calculator and compare:

  • $area after reprojecting to EPSG:3857 — inflated and wrong.
  • $area after reprojecting to EPSG:3035 (LAEA) — correct planar area in m².
  • area($geometry) with ellipsoidal measurement enabled (Project ▸ Properties ▸ General ▸ Measurements, ellipsoid set to WGS 84) — geodesic area, the reference value.

With GDAL/OGR you can do the equal-area reprojection explicitly:

ogr2ogr -t_srs EPSG:3035 habitat_laea.gpkg habitat_4326.gpkg
ogrinfo -dialect SQLite -sql \
  "SELECT SUM(ST_Area(geom))/10000.0 AS hectares FROM habitat_laea" \
  habitat_laea.gpkg

In PostGIS, the cleanest path skips projection entirely and uses the geography type for a geodesic result:

SELECT SUM(ST_Area(geom::geography)) / 10000.0 AS hectares
FROM habitat;            -- geom in EPSG:4326

For a projected check, transform first:

SELECT SUM(ST_Area(ST_Transform(geom, 3035))) / 10000.0 AS hectares
FROM habitat;

The geography result and the EPSG:3035 result should agree to a fraction of a percent over a region the size of a country; large divergence is a signal that something upstream (CRS declaration, invalid geometry) is wrong.

Common pitfalls and why they happen

  • Measuring in the project/basemap CRS by reflex. Web Mercator is the default of most web maps, so analysts measure in it without thinking. The fix is to set an explicit measurement CRS per task.
  • Using $area in QGIS without ellipsoidal measurement. The Field Calculator's $area returns planar area in the layer's CRS; if that CRS is geographic, the number is in square degrees. Either reproject to an equal-area CRS or use area($geometry) with an ellipsoid configured.
  • Picking an equal-area CRS outside its area of use. A custom Albers tuned for North America applied to Africa is still "equal area" mathematically but the WKT's area-of-use mismatch can trigger poor datum transformations and shape distortion that makes QA harder.
  • Invalid geometries. Self-intersecting polygons can yield negative or doubled areas. Run ST_MakeValid (PostGIS) or the "Fix geometries" algorithm (QGIS) first.

QA and validation

  • Compute area two independent ways (equal-area projection and geodesic) and confirm they agree within tolerance.
  • Verify the chosen CRS's area of use (visible in gdalsrsinfo or epsg.io) actually contains your extent.
  • Spot-check one polygon of known true size — a surveyed parcel, a standard grid cell — against your computed value.
  • Confirm geometries are valid before summing.

Bathyl perspective

We pick the area CRS from the deliverable, not the basemap: continental statistics get EPSG:3035 or a regional LAEA, national inventories get Albers, and parcel work stays in the local UTM zone or national grid. Whichever we use, the reported area carries a note of the CRS and method so the figure is reproducible and auditable.

Related reading

Sources