What LULC Analysis Does

The LULC module analyses satellite imagery to classify how land is being used around a given location. It produces annual satellite composites, land cover classification maps, statistical breakdowns, and year-over-year change detection for a six-year period (2020-2025).

Each analysis examines a 5 km x 5 km Area of Interest (AOI) centred on your coordinates. This size captures the broader landscape context around a location — large enough to show meaningful land use patterns while small enough to remain relevant to the specific site.

The report delivers four types of output: RGB satellite imagery (true-colour photos from space), land cover classification maps (colour-coded by land use type), percentage statistics for each land cover class by year, and a change summary highlighting the most significant shifts over the analysis period.

Satellite Imagery

Satellite imagery comes from Copernicus Sentinel-2, a constellation of two satellites operated by the European Space Agency (ESA). Sentinel-2 captures images of every location on Earth approximately every 5 days at 10-metre resolution — meaning each pixel represents a 10m x 10m area on the ground.

The platform uses “surface reflectance” imagery, corrected for atmospheric effects (haze, aerosols, water vapour) so colour values represent actual ground conditions rather than atmospheric conditions on the day of capture.

For each year, a cloud-free composite is created by collecting all available images, removing cloudy pixels using the satellite’s built-in cloud detection flags (QA60 quality band), and computing the median value of each pixel. The median approach is robust to occasional cloud contamination.

Images may look different from Google Maps because they use different spectral band combinations, rendering parameters, and compositing methods.

Land Cover Classification

Classification uses Google Dynamic World V1, a near-real-time dataset produced by a deep learning model trained on expert-labelled Sentinel-2 imagery. Dynamic World classifies every 10-metre pixel into one of nine categories.

Each pixel is assigned the modal (most frequent) class across all observations in a year, smoothing seasonal variations. Dynamic World achieves approximately 75% global accuracy. Common errors include mixed pixels at boundaries and seasonal confusion between Crops and Grass.

Change Detection

Year-over-year changes are calculated as the percentage-point difference in each land cover class between consecutive years. Changes smaller than 1-2 percentage points should be interpreted with caution — they may reflect classification noise rather than actual land use change. Changes larger than 5 percentage points generally indicate genuine transitions.

EUDR Deforestation Screening

The EU Deforestation Regulation (EUDR) requires companies to demonstrate that commodities placed on the EU market are not linked to deforestation occurring after December 31, 2020. Continuuiti provides a screening-level assessment using two complementary datasets:

Hansen Global Forest Change 2024 v1.12 (University of Maryland, 30-metre resolution) provides a year-2000 baseline of tree canopy cover and annual detection of tree cover loss from 2001 through 2024. The exact dataset version is recorded for compliance documentation.

JRC Global Forest Cover 2020 (EU Joint Research Centre, 10-metre resolution) provides a binary forest/non-forest classification as of the EUDR cut-off date (December 31, 2020).

A pixel is considered “forest” if tree canopy cover exceeds 30%, following the FAO international standard. Post-2020 forest loss is detected by identifying Hansen loss-year values of 21 or higher.

The report presents four EUDR risk indicator cards: Forest Status (loss detected or not), Protected Area (inside, near, or none), Primary Forest (old-growth assessment via JRC Forest Subtypes V0), and Ecoregion Risk (NNH conservation priority).

This screening is a due diligence support tool, not a legal compliance determination. Full EUDR compliance requires field verification, supply chain documentation, and geolocation of production plots.

Protected Areas

Proximity is assessed using the World Database on Protected Areas (WDPA), maintained by UNEP-WCMC, containing over 270,000 designated sites. The analysis searches a 20-kilometre buffer and reports: Inside PA, Near PA, or None Nearby.

Protected areas are categorised by IUCN management categories (Ia through VI). The interactive map shows boundaries as polygon overlays (up to 10 of the largest nearby areas).

Ecological Context

Ecological context uses the RESOLVE Ecoregions 2017 dataset (846 ecoregions, 14 biomes, 8 biogeographic realms). The Nature Needs Half (NNH) conservation priority scale assigns each ecoregion to one of four categories:

Limitations

The 10-metre resolution means features smaller than ~10 metres cannot be distinguished. Cloud cover is a persistent challenge in tropical regions. EUDR deforestation screening is not a legal compliance determination. WDPA boundary data may lag actual designations by up to one month. Hansen forest loss detection has known false positives in some landscapes. JRC Forest Subtypes V0 is a preliminary dataset version.

Appropriate use: Supply chain due diligence screening, EUDR pre-assessment, land use change monitoring, landscape-scale environmental risk identification, comparative site analysis.

Not suitable for: Parcel-level legal boundary determination, official EUDR compliance certification, field verification replacement, individual tree inventory, or regulatory submissions without additional professional review.