Remote Sensing

Land Cover Classification

Assigning each pixel of an image to a category such as forest, cropland, water or built-up area.

Land-cover classification turns reflectance data into meaningful categories, allowing analysts to map, for instance, how much of a project area is forest, wetland or agriculture. It underpins area-based carbon accounting.

Classification may be rule-based, using spectral thresholds, or learned, using training samples to teach an algorithm to recognise classes. Accuracy depends on the quality and representativeness of those samples.

Because classes must be defined consistently, classification schemes are fixed in advance so that the same category means the same thing across time and across projects.

Frequently asked questions

Why do classification schemes matter?

Consistent definitions ensure that 'forest' means the same thing at every date and site, making change detection meaningful.

What improves classification accuracy?

Representative training data, good imagery, and validation against independent reference points.

Related terms