Land Cover Fusion
Provided by
Global land cover products exist and are free, but they are fixed: fixed classes, fixed epochs, fixed accuracy, and never quite the classes a given customer’s regulation or business asks for. This service produces land cover to order, for the area, the date window and the class scheme requested.
Its accuracy comes from fusion. Optical imagery carries the spectral information that distinguishes crops from grass from trees; radar carries structure and, crucially, is unaffected by cloud, which is how a map gets made over a region that is overcast for half the year. Combining Sentinel-1 and Sentinel-2 across a season also exploits phenology: a winter cereal and a summer crop are identical in one image and unmistakable across twelve.
Where the customer has their own nomenclature (e.g.: a national scheme, a Corine-compatible legend, a bespoke set of classes for a permit) the service trains against their reference polygons rather than forcing them onto someone else’s classes. Every output carries a per-pixel confidence layer, because a land cover map without one invites more trust than it has earned.
