EvoLand

The Role of IIASA/NODES in EvoLand 

NODES plays a pivotal role in EvoLand’s success. As leaders in the task of Novel In-Situ and Training Data, NODES is responsible for compiling existing in-situ datasets and filling gaps in land use data as well as improving automated techniques for training data generation—critical for developing next-generation machine learning models. 

Existing in-situ datasets

By addressing some gaps in the existing datasets, NODES strengthens the foundation upon which the rest of EvoLand is built. For example, using data of drivers of biomass change, our work enhances our understanding of forest health and change dynamics—key for sectors like agriculture and forestry that are at the forefront of environmental policy discussions.  

Furthermore, we will ensure that the data products emerging from EvoLand can be seamlessly integrated into policy frameworks, by creating a STAC catalogue with all in-situ datasets that can be made available to the public. This will empower decision-makers with timely and high-quality information, allowing for fact-based governance and a more sustainable future.

Automated techniques for training data generation

The next wave of Earth Observation depends heavily on artificial intelligence, especially machine learning (ML) and deep learning (DL) algorithms. To power these algorithms, NODES is developing new tools for the automated generation of training data. By leveraging homogenous objects from Sentinel-1 and Sentinel-2 satellites, we are pioneering labeling techniques that assign categories to these objects based on existing thematic maps. 

This automatic labeling is essential for improving the speed and scalability of land monitoring. Traditional methods often relied solely on manual labeling, which can be time-consuming and error-prone. With our innovative approaches, complementing manual labeling with Artificial Intelligence, EvoLand is moving towards a future where training data can be generated on demand with greater accuracy and efficiency.

Generate a novel in-situ data set on land cover change

A key output from our work is the creation of pixel-based, high-resolution (10m) land cover change data. This dataset, spanning the years 2020-2022, will offer a detailed look at how landscapes have evolved across prototype sites. The novelty lies not only in the resolution but also in the methodology, which combines expert visual interpretation of very high-resolution (VHR) imagery with the latest geospatial tools. These datasets are crucial for calibrating and validating change detection algorithms that form the core of EvoLand’s CLMS prototypes. The resulting insights will not only help develop better land cover and biomass maps but also inform policymakers, stakeholders, and the Copernicus Entrusted Entities about future directions for Earth Observation services.

In conclusion, EvoLand is more than a technical project—it has clear implications for policy and decision-making. IIASA’s role in the EvoLand project is indispensable in realizing the vision of an evolved Copernicus Land Monitoring Service. By pushing the boundaries of data generation, quality control, and automated training data generation, we are helping to create the infrastructure that will guide Europe’s land management and environmental policy for years to come. 

Involved colleagues from NODES include: Steffen Fritz, Myroslava Lesiv, Santosh Karanam, Monica Shishodia, Anto Subash, Maria Shchepashchenko, Orysia Yashchun, Zoriana Romanchuk, Katya Perez Guzman and others.