Towards Generalizable Mapping of Hedges and Linear Woody Features from Earth Observation Data: a national Product for Germany
New workflow uses deep neural nets to identify linear woody features from satellite data...
Hedges and other linear woody features are critical for ecosystem services in agricultural landscapes, providing habitat for pollinators and aiding climate adaptation. But mapping them systematically at scale has been a challenge due to sensor diversity and landscape variability. Now, researchers from Germany have developed a modular workflow that tackles this with two independently optimizable components: a flexible input data interface that consolidates heterogeneous Earth observation data into a binary woody vegetation mask, and a deep neural network trained to separate linear from non-linear shapes within those masks.
The team demonstrated the workflow by deriving three national-scale maps for all of Germany from three input sources using a single trained model without any retraining. Evaluation against refined reference data from four federal state biotope mapping campaigns and comparison with two existing maps showed competitive results across all sites. The modular design and proven applicability at national scale provide a foundation for scalable and generalizable linear woody feature mapping beyond Germany, promising to make ecological monitoring more efficient and accessible.
- Modular workflow with two independently optimizable components: an input data interface and a deep neural network for shape separation.
- Single trained model applied to all of Germany using three different input sources without retraining, achieving national-scale coverage.
- Evaluated against four federal state biotope mapping campaigns, showing competitive results across all sites.
Why It Matters
Enables scalable, generalizable mapping of vital ecological features for climate and biodiversity efforts worldwide.