- University: Technische Universität Dresden
- Place of defence: Dresden
- Hosting institution(s): Department of Landscape Ecology, Helmholtz Centre for Environmental Research (UFZ)
- Keywords: Individual based model, Resource selection function, White stork, Agricultural landscape, Habitat selection
Abstract
Understanding how white storks (Ciconia ciconia) interact with dynamic agricultural landscapes requires integrating empirical habitat selection data with mechanistic models that explain individual behavioural variability. The aim of this study was to This study examines how landscape transformation shapes white stork habitat selection, movement decisions, and breeding status using an integrated modelling approach. Using GPS tracking data from 33 storks in Germany, this study quantified resource selection with mixed-effects Resource Selection Functions (RSFs) using the empirical data and embedded these model coefficients into an Individual-Based Model (IBM) to evaluate how landscape structure influences movement, energy balance, and breeding status. RSF showed that vegetation indices (NDVI, NDWI) provided weak explanatory power, largely due to the narrow environmental gradients and rapid vegetation turnover in the study area. Instead, land-use and land-cover (LULC) classes were the primary determinants of habitat use, with strong selection for herbaceous croplands and opportunistic use of impervious surfaces. Storks also preferred landscapes with high edge density and low evenness, indicating a reliance on fragmented agricultural systems that maximise prey accessibility. The IBM reproduced these empirical selection patterns and revealed that LULC classes directly influences energetic performance. Simulated breeding outcomes increased in years with greater availability of preferred cropland habitats, demonstrating clear linkages between landscape composition and reproductive fitness. However, strong individual variability reduced predictive accuracy in both the empirical RSF (mean AUC = 0.58) and IBM-derived RSF models (overall mean AUC = 0.51). Breeder-specific models performed substantially better (AUC up to 0.61), reinforcing that combining behavioural states masks true selection signals. Overall, the RSF-IBM integration demonstrated that white stork habitat use is shaped primarily by LULC classes and landscape structure rather than vegetation indices, and that breeding status is sensitive to temporal changes in landscape composition. These results highlight the ecological risks posed by homogenisation of agricultural systems and underscore the need for conservation strategies that prioritise fragmented open agricultural landscapes, and individual variability across populations.
