Time-synchronous, multiscale remote sensing and multimodal data fusion for biodiversity monitoring

Since the 1970s, conservation efforts have centered on legislation for protected areas, such as the UNESCO Man and the Biosphere Program and the EU Birds and Habitats Directives. The 2022 Kunming-Montreal Global Biodiversity Framework and the 2024 EU Nature Restoration Law establish measurable goals for restoring biodiversity, underscoring the necessity of extensive and effective monitoring strategies. High-quality monitoring is crucial for tracking these efforts.

We propose a time-synchronous, multiscale, remote sensing approach for biodiversity monitoring. The key goal is to provide decision-ready information for essential biodiversity variables (EBVs) derived from analysis-ready data (ARD) stemming from multispectral remote sensing. EBVs represent a quantitative yet theory-driven approach – in contrast to previously accepted data-driven approaches – that helps describe how the state of biodiversity changes over time and space. They can be seen as the connection between basic primary observations and advanced indicators. Priority lists of biodiversity metrics are available that are observable from space. However, standard procedures and protocols for deriving EBVs from various domain-specific field observations and providing these data as machine-readable ARD in digital format are lacking. To provide decision-ready information to stakeholders, ARD multimodal fusion from different domains and scales is essential. Validation of this information is also critical. Unmanned aerial systems (UAS) equipped with multispectral sensors operating in the same spectral range as those on Copernicus/Sentinel-2 provide a flexible, close-range remote sensing option for acquiring very high-resolution data at high frequencies. This helps to satisfy the need to capture data at specific times depending on the phenological state of vegetation, for example.

In the context of the ongoing EU Horizon Europe project BioMonitor4CAP (project number 101081964), we analyzed and compared selected EBVs derived from close-range and multispectral remote sensing data from three spatial scales: Copernicus/Sentinel-2 (10 m GSD), PlanetScope/SuperDove (3 m GSD), and UAS (5 cm GSD). We then compared these EBVs to in situ data, such as vegetation species distribution, acoustic monitoring, and eDNA soil sampling. In our contribution, we critically reflect on this interdisciplinary, holistic approach and discuss its advantages and challenges for multimodal data integration and analysis.