Strategic Framework for Biodiversity Conservation: AI and Open Source Data for Protected Area Prioritisation

Ecological Niche Modelling Reinforcement Learning Data Pre-Processing Strategic Framework for Biodiversity Conservation: AI and Open Source Data for Protected Area Prioritisation Katharina Horn1, 2, Christine Wallis1, Birgit Kleinschmit1, 3, Daniele Silvestro4, Annette Rudolph2 Introduction Biodiversity is increasingly under threat due to a range of human-driven processes, including climate change, deforestation, land use change, and habitat degradation [1]. We developed a framework to prioritize nature conservation zones in forests under current and future climate conditions. Here, we apply the framework to the federal state of North Rhine-Westphalia (NRW) in Germany to showcase the potential of data fusion and AI tools for nature conservation purposes. Data Methodology Preliminary results Geospatial data (2016-2024) Vegetation - Tree species, - Tree cover density, - Canopy height, - Forest type, - Distance to forest edges Land cover - LULC (Ecosystem Atlas Germany), - Distance to roads, urban settlements, water bodies, - Protected areas in Germany Climate - Agroclimatic indicators (e.g., biological effective degree days, wet days, summer days, frost days etc.) - Bioclimatic indicators (CHELSA) Topography - Slope, - Aspect, - Elevation Soil - Organic carbon density, - Soil moisture index, - Clay content, - Percentage of sand in the soil, - etc. Satellite data - Sentinel-2 & Landsat: NDVI, EVI, NDWI, NDRE1, CCI, - Sentinel-1: RVI, VH, VV, Gray-level cooccurrence matrix, Standard deviation, Entropy Sources References [1] Habibullah, Muzafar Shah; Din, Badariah Haji; Tan, Siow-Hooi; Zahid, Hasan (2022): Impact of climate change on biodiversity loss: global evidence. In: Environ Sci Pollut Res 29 (1), S. 1073–1086. DOI: 10.1007/s11356-021-15702-8. [2] Silvestro, D., Goria, S., Sterner, T., & Antonelli, A. (2022). Improving biodiversity protection through artificial intelligence. Nature sustainability, 5(5), 415-424. Resampling Pre-Processing Cropping Filtering GBIF bird observation points Ecological niche modelling (MaxEnt) Creation of background points Ecological niche models for forest indicator bird species •Opportunity costs for nature protection, •Climate, •Forest mask, •Nature conservation areas Resampling Protected areas in German forests Fig. 3 Mean ecological niche of Anthus trivialis and Dryocopus martius from 2016-2024. Fig. 4 Preliminary CAPTAIN prediction results in North Rhine-Westphalia (Germany). Fig. 1 Study area for the first CAPTAIN analysis. Conclusions 1Geoinformation in Environmental Planning Lab, Technische Universität Berlin, 10623 Berlin, Germany 2Artificial Intelligence and Land Use Change, Technische Universität Berlin, 10623 Berlin, Germany 3Thünen Institute, 38116 Braunschweig, Germany 4Department of Biosystems Science and Engineering, ETH, Zurich, Switzerland Fig. 2 CAPTAIN AI model workflow (figure © Silvestro et al. 2022) [2]. Development of a robust framework to prioritise nature conservation zones Modelling of the most valuable forest areas in NRW based on ecological niche models of birds and the AI CAPTAIN model Transferability to other regions possible Support for environmental planning institutions and nature conservation agencies Preliminary results indicate that only a few valuable forest areas are located in Natura 2000 areas Outlook Preprocessed data GBIF bird observation points Katharina Horn [email protected] Get in touch! Mean ecological niche (2016-2024) of the black woodpecker (Dryocopus martius) Mean ecological niche (2016-2024) of the tree pipit (Anthus trivialis) Ecological niches across all years (2016-2024), 0 = low, 1 = high