EvoLand at LPS 2025: Pioneering the future of land monitoring with innovation and impact

EvoLand was proudly featured at the DG DEFIS stand, organised in collaboration with HaDEA (European Health and Digital Executive Agency). As part of a curated selection of EU-funded Earth Observation projects, EvoLand was showcased on a dedicated screen—bringing visibility to its mission of developing next-generation services for the Copernicus Land Monitoring Service.

An e-poster presentation highlighted the scope of EvoLand’s 11 prototype services and their alignment with EU priorities in agriculture, forestry, water, urban dynamics, and general land cover. Being part of this exhibit helped position EvoLand within the broader landscape of impactful European EO initiatives, reaching a wider audience of policymakers, researchers, and programme stakeholders.

At the poster session on “Cover Crop Type Mapping”, Mathilde De Vroey (VITO) presented EvoLand’s prototype designed to support sustainable farming under the EU’s Common Agricultural Policy. This work focuses on extending CLMS capabilities to detect and classify cover crop types using remote sensing.
The prototype leverages satellite data and machine learning to identify cover crops with greater accuracy, addressing a growing need for transparent, scalable monitoring of agricultural practices. It offers policy-relevant insights and paves the way for more targeted environmental interventions at the farm level.

In a live demo session, Daniel Thiex (Sinergise) showed how users can quickly visualise project results using the Copernicus Browser—part of the Copernicus Data Space Ecosystem. Featuring EvoLand as a use case, the presentation guided the audience through uploading and preparing EO data using tools like Evalscripts, legends, and the “Bring Your Own COG” API.

The session made it clear that publishing EO results can be simple, interactive, and accessible—lowering the barrier for scientists and service providers to share their insights with wider audiences.

One of the most talked-about moments of the week came during the session “Super-resolution in Earth Observation”, where Julien Michel (CESBIO/CNES) took the stage to share EvoLand’s latest work on deep learning for Sentinel-2 imagery enhancement. The room quickly filled—over 200 participants joined to hear how EvoLand is advancing super-resolution techniques for EO.

Julien introduced a comprehensive new framework for evaluating the performance of super-resolution models. This included innovative approaches to control geometric distortions, ensuring that spatial structures remain intact, and the introduction of spatial frequency metrics to better reflect the fidelity of upscaled outputs.
The session culminated with the announcement of a publicly available model developed within EvoLand that can super-resolve 10 Sentinel-2 bands to 5-metre resolution—offering radiometric accuracy and immediate usability for researchers and practitioners alike.

Read more in the session summary.

Another high-impact moment came as EvoLand took over the EO Arena for the live demo “New Methods for the Next Generation of Copernicus Land Monitoring”. Led by Phil Harwood (Evenflow), the session invited attendees on an interactive tour of the EvoLand results portal—the central gateway to explore the project’s 11 candidate prototypes.

Throughout the demo, Phil explained how each prototype addresses specific monitoring needs across five key land themes: agriculture, forests, urban areas, inland waters, and general land cover. These services are designed to go beyond the status quo, introducing improved data richness, customisation options, and enhanced policy relevance for future CLMS evolution.

The lively audience engagement and insightful questions highlighted the strong appetite for next-generation EO services, and underscored EvoLand’s potential to support more targeted and timely environmental monitoring across Europe.

Astrid Vannoppen (VITO) showcased ongoing upgrades to the Copernicus Gross Primary Production (GPP) algorithm within EvoLand. The work focuses on producing vegetation-class-specific GPP estimates at 10-metre resolution—significantly enhancing the granularity of ecosystem productivity data.

By integrating improved Sentinel-2 derived FAPAR, high-resolution land cover inputs, and distinctions between C3 and C4 crops, the prototype supports more accurate and detailed assessments of agricultural and natural vegetation productivity.

In a poster co-authored by several key partners, Wai-Tim Ng (VITO) summarised EvoLand’s overarching contribution to the future of the Copernicus Land Monitoring Service. The poster offered an overview of the project’s 11 prototype services and their alignment with emerging policy and environmental needs.

Designed across domains such as agriculture, forests, urban, water, and general land cover, these services reflect EvoLand’s mission: to deliver more customisable, scalable, and timely EO tools for Europe’s evolving land monitoring landscape.

During the session “Urban Land Monitoring with EO”, Eve Poitevin (CLS) presented EvoLand’s prototype for annually updating urban land cover and use changes. The method builds on Sentinel-2 Level 3A mosaics and integrates PCA-based change mapping, Random Forest classification, and seasonal filtering to better isolate urban dynamics.

Designed for automation and scalability, the approach supports more consistent tracking of urban growth across Europe—laying the groundwork for future CLMS updates tailored to fast-evolving urban environments.

Nassim Ait Ali Braham (DLR) presented a powerful approach to EO analysis using self-supervised learning. The poster introduced a vision transformer trained on EnMAP, Sentinel-2, and Landsat-8 data, capable of extracting deep features across multiple sensors.

By combining multispectral and hyperspectral inputs, the model supports robust transfer learning across tasks like land cover and crop type classification—highlighting the role of foundation models in next-generation EO applications.

On Friday morning, EvoLand’s forest monitoring work took centre stage during the session “Forest Monitoring with Earth Observation Data”, presented by Linda Moser (GAF AG). The talk highlighted how the project is using Sentinel-2 time series and unsupervised algorithms like EWMA to detect both subtle and abrupt canopy changes, combined with advanced classification for reliable outputs at 10-metre resolution.

The presentation received strong interest from the community, including ESA and researchers working on similar challenges. Attendees expressed eagerness to see the prototypes in action. Several were interested in future collaboration and benchmarking against other forest disturbance products—clear signs of EvoLand’s growing impact in this vital area.

With over a dozen project members present and strong community interest throughout the week, EvoLand’s participation at LPS 2025 marks a milestone in its journey to co-design future-ready CLMS services.