‹ back to GIS Glossary

Land Cover

Definition

Land cover describes the physical material at the Earth’s surface—forests, grasslands, croplands, impervious surfaces, bare soil, snow/ice, wetlands, and water. It differs from land use, which is how humans utilize land (residential, industrial, recreation). Land cover is commonly derived from satellite imagery using spectral indices, seasonal signatures, and machine-learning classifiers. Products vary in resolution (10 m to 300 m+), class definitions, and update cadence. Good mapping incorporates training data across seasons, handles mixed pixels at edges, and validates with ground truth. Change detection requires consistent methods across years and careful attention to phenology; a field might look like bare soil in spring and dense vegetation in summer. Misclassifications typically occur between spectrally similar classes—urban versus bare soil, irrigated crops versus wetlands—so post-processing with elevation, slope, and proximity to water can help.

Application

Planners use land cover to model runoff and heat islands, conservationists to assess habitat fragmentation, and carbon programs to estimate biomass change. Agriculture tracks cropping intensity and fallow cycles. Disaster managers evaluate burn scars and flood extent. Land cover forms a base layer for many other spatial models, from epidemiology to transportation.

FAQ

How do land cover and land use interact in planning?

Land cover sets biophysical context; land use encodes policy. A park (use) may include forest and turf (cover). Many analyses require both to understand performance and constraints.

Why do global products differ from local maps?

Class schemes, sensor mix, and training data vary. Global products emphasize consistency and may miss local classes; local maps tailor classes but can be harder to compare across regions.

How much ground truth is needed?

Enough to represent each class across geographies and seasons—often thousands of samples. Stratified sampling and independent validation prevent optimistic accuracy estimates.

Can crowdsourcing help?

Yes. Verified photo labeling and apps like field notebooks augment training data, but must be quality-controlled to prevent bias toward easy-to-reach areas.