FAIRiCUBE UC2 · September 2025
Agriculture dominates the Dutch landscape, with more than half of the land area used for farming. These agricultural lands are vital not only for food production but also for biodiversity, especially farmland birds. However, biodiversity in Europe is in decline, making it essential to understand how farming practices can be adapted to support ecological resilience.
As part of the FAIRiCUBE project, Wageningen Environmental Research addresses these topics by applying data cube technology, machine learning, and causal inference methods. The goal is not just to map biodiversity patterns but to reveal the causal effects of farm level interventions, such as crop rotation and mowing intensity, on species richness.
Methodological Approach
• Spatial Data Cubes: Environmental, agricultural, biodiversity, and remote sensing datasets were integrated into a cube-based infrastructure. This setup allows efficient access to multi-source, multi-temporal data and makes it possible to run consistent workflows across different regions and time periods.
• Machine Learning for Pattern Discovery: ML models, such as MaxEnt, were used to predict species richness of farmland birds based on environmental variables like NDVI and land use. These models identify spatial patterns and highlight areas of higher or lower biodiversity potential.
• Causal Modelling: Beyond correlations, graph-based inference methods were applied to test causal relationships. By combining ML predictions with indicators of farm management (e.g., crop rotation index), the approach estimates the actual effect of agricultural practices on biodiversity outcomes. Importantly, the framework includes refutation tests to confirm robustness of the causal links.
• Scalability Across Space and Time: The integration of cubes and causal inference enables workflows that are transferable across regions. Although the example focuses on Flevoland, the same methods can be applied in other agricultural landscapes to systematically evaluate management impacts.
Why It Matters
This methodology demonstrates how cutting-edge data infrastructures and causal inference can bridge the gap between agricultural data and ecological outcomes. It offers a replicable, policy-relevant framework to assess how farming practices influence biodiversity, supporting initiatives such as the EU Green Deal and helping design management strategies that are both productive and biodiversity friendly.
More info: https://uc2.fairicube.nilu.no/