Landcover Classification

Landcover Classification is Kumi Analytics' land-cover mapping service for nature-based solutions (NbS) projects, national reporting and planning. Land-cover maps made at country scale and at project scale often use different legends, so teams have to reconcile them before they can compare national trends with site conditions; Kumi Analytics also argues that generic global land-cover products use classes too coarse for heterogeneous landscapes.

The service classifies satellite imagery into wall-to-wall national or regional maps at 10 m or 30 m resolution and into higher-resolution project-level screening maps, using one taxonomy at every scale. Optical and SAR imagery are combined where appropriate, with composites and multi-temporal analysis, to keep the classification stable in cloudy regions and across seasons. Outputs include deforestation baselines, forest/non-forest classifications and ecosystem-type maps of the kind required by Verra, Gold Standard and REDD+ national accounting frameworks, plus land-cover change layers and time series for deforestation, urban expansion, agriculture and restoration.

• Standards-aligned legends: UN-FAO LULC-style options, or legends tailored to national systems such as CORINE, ministry schemes or project-specific classes.
• Feasibility screening: fast, coarse classification to shortlist candidate project areas before detailed surveys.
• Validation: accuracy assessment and, where needed, uncertainty quantification so outputs hold up in reporting and audits.
• Cross-sector use: the same maps serve carbon projects, agriculture, urban and spatial planning, biodiversity and habitat assessment, water and catchment analysis, disaster risk and ESG reporting.

Kumi Analytics reports that one of its satellite classifications in North Sumatra reached 94.58% accuracy in distinguishing mangroves, palm plantations and tropical forest.