From Biodiversity Research to Report-ready Metrics: Biodiversity Indicators as Services based on Geo Engine

The demand for transparent corporate biodiversity reporting, particularly driven by standards like the EU’s ESRS E4, requires moving from qualitative risk assessments to quantifiable, location-specific, and dynamic metrics. The main challenges currently include the lack of standardization, the inherent complexity of biodiversity, and the difficulty of acquiring and processing data with sufficient spatio-temporal resolution across diverse operational footprints.

We propose to address this challenge from both ends. First, by providing tooling for the operationalization of indicators based on scientific research in backend services. Second, by an easy-to-use interface for companies that enables the acquisition of report-ready metrics for provided company properties. The operationalization requires connectors to access diverse remote sensing data sources, a processing pipeline that scales with the data volume, and a framework to incorporate machine learning (ML) models in the computation of biodiversity indicators. While the backend has to provide a rich set of operations, the frontend needs to be simple and focus on the access and exploration of indicators with respect to comprehensibility, applicability (e.g., validated for Western European vegetation), and reusability.

The Geo Engine is our geo-processing toolkit that offers native support for time-series, data connectors to crucial data hubs like the Copernicus Dataspace Ecosystem, a declarative query engine that provides standard operators and reusable processing graphs, and an ML operator that allows the incorporation of trained models into processing. As the Geo Engine is Open Source and the processing is declarative, the computation of indicators is fully auditable.

We introduce the Biodiversity Indicator Service BioIS that serves as a user-facing application, which allows discovering, auditing, and accessing a growing number of biodiversity indicators that are computed and served by Geo Engine. It allows the ad-hoc computation of metrics for uploaded company properties for a selected time period. For BioIS, we focus on indicators derived from remote sensing data, in particular satellite imagery from Sentinel-2. While the resolution is limited in comparison to commercial data, it suffices for large-scale monitoring of key biodiversity-relevant changes. We demonstrate our approach on two selected use cases: land usage and dominant tree species in Germany.