Turning Satellite Imagery into Actionable Deforestation Evidence

​

As global supply chains face increasing scrutiny over deforestation and forest degradation, businesses need reliable, transparent and geospatially explicit information to assess the environmental risks associated with their sourcing activities. The European Union Deforestation Regulation (EUDR) has reinforced the importance of understanding where commodities originate, how land has changed over time, and whether production areas are associated with deforestation after the regulation’s applicable cutoff date.

​

RADIOM supports organisations in addressing these challenges through satellite-based Earth observation, geospatial analysis and environmental monitoring. By transforming multispectral satellite imagery into spatially explicit information on land-cover change, RADIOM helps businesses and supply-chain stakeholders build a stronger evidence base for EUDR due diligence.

​

This case study demonstrates how Sentinel-2 satellite imagery and supervised land-cover classification can be used to assess tree-cover loss relative to a baseline period. RADIOM applies a machine-learning alghoritm to distinguish relevant land-cover classes and quantify changes between observation periods.

 

The resulting analysis provides a practical example of how remote sensing can contribute to deforestation risk screening, environmental assessment and geospatial evidence preparation.

​

​

​​​The Challenge: Establishing a Baseline for Land-Cover Change

A central challenge in deforestation monitoring is determining whether tree cover has changed within a specific geographical area and identifying where those changes have occurred. Field inspections can provide valuable information, but they may be costly, time-consuming or difficult to conduct across large or remote areas. Satellite imagery offers a complementary approach, enabling repeated observations of the Earth's surface over time.

​

For organisations subject to EUDR due diligence requirements, the ability to examine land-cover conditions at relevant dates is particularly important. Evidence must be linked to the actual production plot or relevant sourcing area, and the timing of any detected land-cover change must be evaluated against the applicable regulatory requirements.

The objective of this case study was to calculate the difference in tree loss relative to a defined baseline using Sentinel-2 imagery. The analysis focused on identifying spatial differences between classified land-cover observations and translating those differences into measurable indicators that can support further environmental investigation.

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

 

The Methodology: Sentinel-2 and Supervised Image Classification

​

Sentinel-2, part of the European Union's Copernicus Earth observation programme, provides multispectral satellite imagery suitable for monitoring vegetation, land cover and environmental change. Its spectral bands capture different characteristics of the Earth's surface, allowing vegetation and other land-cover types to be distinguished using appropriate image-processing and classification techniques.

​

For this case study, Sentinel-2 imagery was processed and classified using B02, B03, B04 and B08. The imagery was prepared for classification, and a supervised land-cover classification workflow was implemented using the Maximum Likelihood algorithm. This method assigns image pixels to predefined classes based on their spectral characteristics and the statistical distributions derived from representative training samples.

​

The classification workflow was designed to distinguish tree-covered areas from other relevant land-cover categories. By applying a consistent classification approach to the baseline and comparison imagery, the resulting maps could be compared to identify areas where pixels previously classified as tree cover were subsequently classified as non-tree cover.

​

The analysis illustrates how multispectral satellite data can be converted into a structured land-cover dataset suitable for spatial comparison. The quality of the resulting information depends on factors such as the selection of representative training samples, image quality, cloud contamination, seasonal variation, classification accuracy and the suitability of the chosen land-cover definitions.

​​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

​

 

Detecting and Quantifying Tree-Cover Loss

​

Following classification, the baseline land-cover map was compared with the later observation to identify potential transitions from tree cover to non-tree cover. This change-detection process provides a spatial representation of areas that may have experienced tree-cover loss during the period under investigation.

By calculating the area of the identified changes, the analysis can produce quantitative indicators of tree-cover loss within the study area. These indicators can include the total area of detected change, its spatial distribution, the proportion of the study area affected and the location of potential change hotspots.

The comparison with a defined baseline is particularly useful because it establishes a reference against which subsequent observations can be evaluated. Rather than relying solely on a single image or a generalised regional indicator, the approach allows changes to be examined at a more detailed spatial scale.

​

Importantly, a transition from tree cover to non-tree cover in a classified satellite image is an indication of potential land-cover change, not automatic proof of deforestation. Temporary vegetation changes, harvesting cycles, seasonal differences, fire damage, image artefacts and classification errors can produce similar signals. Additional observations and contextual evidence are therefore important when interpreting detected changes.

​​

​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

From Satellite Analysis to EUDR Evidence

​

The EUDR establishes requirements relating to deforestation-free production, legality and due diligence for relevant commodities and products within its scope. Satellite-based change detection can contribute to these processes by helping organisations examine land-cover conditions in and around the geographical areas associated with commodity production.

​

The analysis demonstrated in this case study can support several elements of an EUDR evidence workflow.

​

- Geospatial screening of production areas: land-cover classifications and change maps can be overlaid with supplied production-plot boundaries or other relevant geographical information. This helps identify areas where detected changes may warrant further investigation.

​

- Baseline and historical comparison: suitable satellite observations can be used to assess land-cover conditions at relevant dates, including the applicable EUDR cutoff date of 31 December 2020. The selection of imagery must be appropriate to the specific plot, date and evidentiary question being assessed.

Identification of potential deforestation hotspots. Spatially explicit change maps help distinguish areas that appear stable from those where potential tree-cover loss has been detected. This enables more targeted review and prioritisation of higher-risk locations.

​

- Supporting documentation: classified maps, change-detection outputs, imagery dates, processing records, methodology descriptions and validation results can be assembled into a documented geospatial evidence package. Such documentation helps make the analytical process more transparent and easier to review.

​

- Ongoing monitoring: repeat satellite observations can be used to monitor changes over time and flag locations that may require additional assessment, supporting a more systematic approach to environmental risk management.

The value of this approach lies in combining spatial analysis with traceability information and documented due diligence procedures. Satellite imagery provides an independent source of environmental observations, while plot-level geolocation, commodity information, legality documentation and other relevant records help establish the context required for an EUDR assessment.

​

​

RADIOM's Approach: From Earth Observation to Decision Support

​

RADIOM uses Earth observation and geospatial analysis to help organisations understand environmental change and make better-informed decisions. The approach combines satellite data, spatial processing and transparent analytical workflows to turn complex geospatial information into practical outputs for environmental monitoring and supply-chain risk assessment.

​

Depending on project requirements, the services can include satellite-image acquisition and preparation, land-cover classification, baseline comparisons, change detection, calculation of affected areas, geospatial visualisation and the preparation of supporting technical documentation.

​

For EUDR-related applications, these outputs can help clients investigate potential deforestation risks associated with relevant production plots, prioritise areas for additional verification and organise geospatial information for due diligence. The precise workflow is tailored to the commodity, geographical scope, availability of plot boundaries, relevant observation dates and evidence requirements.

​

Where appropriate, satellite-derived findings can be complemented by higher-resolution imagery, additional historical observations, reference datasets, field information and independent validation. These complementary sources are particularly important where tree-cover changes are ambiguous or where a regulatory conclusion depends on the exact location, timing or nature of the land-cover transition.

​

​

Why Satellite-Based Monitoring Matters

​

Satellite-based monitoring offers several practical advantages for organisations working with geographically dispersed production areas. It provides a consistent means of observing large areas, supports repeatable comparisons over time and generates spatially explicit outputs that can be integrated with geographic information systems and supply-chain datasets.

​

Sentinel-2 is particularly useful for systematic land-cover monitoring because it provides freely available multispectral imagery and repeated observations. Although its spatial resolution and observation conditions impose limitations, it can serve as an effective first step in identifying areas of potential change and directing more detailed investigation where necessary.

​

A documented classification and change-detection workflow also improves transparency. Recording the imagery used, classification approach, training data, processing parameters, change-detection rules and accuracy assessments makes it easier to understand how results were generated and what conclusions can reasonably be drawn from them.

​

For businesses seeking to strengthen environmental risk screening, this combination of repeatable observation, spatial analysis and documented outputs can form a valuable part of a broader monitoring and due diligence system.

​​

​

From Detection to Verification

​

An essential distinction in any satellite-based deforestation assessment is the difference between detecting tree-cover loss and determining whether deforestation has occurred within the meaning of the EUDR.

​

Tree-cover loss is a remotely observed change in vegetation cover. Deforestation, as defined under the EUDR, involves the conversion of forest to agricultural use. Consequently, a classified transition from tree cover to non-tree cover does not by itself establish that the change constitutes deforestation under the regulation.

A robust assessment therefore needs to consider the relevant forest definition, land use before and after the observed change, the geographical boundaries of the production plot, the applicable cutoff date and the limitations of the available satellite observations. Classification accuracy and the reliability of the temporal comparison must also be considered.

​

RADIOM positions satellite-derived outputs as supporting evidence for screening, investigation and documented due diligence, rather than as a standalone legal determination or a guarantee of EUDR compliance. This distinction helps ensure that analytical findings are interpreted appropriately and integrated into a defensible evidence workflow.

​​

Making Geospatial Evidence More Accessible

​

This Sentinel-2 case study demonstrates how satellite imagery, supervised land-cover classification and baseline comparison can be combined to identify and quantify potential tree-cover loss. Using QGIS and the Semi-Automatic Classification Plugin, the workflow illustrates a practical route from multispectral imagery to mapped land-cover change and quantitative spatial indicators.

​

For organisations preparing for or implementing EUDR-related due diligence, such methods can support the examination of production areas, the identification of potential environmental risks and the preparation of transparent geospatial documentation. When combined with plot-level traceability, appropriate historical evidence, accuracy assessment and complementary verification, satellite-derived information can contribute meaningfully to a more systematic approach to deforestation risk assessment.

​

RADIOM helps bridge the gap between satellite observations and practical environmental intelligence. Through geospatial analysis and Earth observation, it supports organisations in identifying potential land-cover change, understanding where further investigation may be needed and developing documented evidence to inform responsible sourcing and environmental due diligence.