Integrating unmanned aerial vehicle-derived habitat metrics with movement persistence and species distribution models to characterize seasonal habitat use of invasive turtles in urban wetlands

Invasive freshwater turtles — the Red-eared Slider ( Trachemys scripta elegans ) and River Cooter ( Pseudemys concinna ) — pose serious ecological threats, yet fine-scale seasonal habitat-use patterns remain poorly understood due to GPS signal loss from semi-aquatic behaviour. We evaluated whether Movement Persistence (MP) modelling, combined with high-resolution unmanned aerial vehicle (UAV) imagery (~2 cm resolution), can identify seasonal habitat suitability and management priority areas for invasive turtles in an urban wetland. Nine individuals were tracked at the Ansan Reed Wetland, South Korea, and season-specific MaxEnt species distribution models (SDMs) were built from MP-reconstructed occurrence proxies. Boosted Regression Trees were additionally fitted to validate variable importance across algorithmically independent approaches. SDMs revealed clear seasonal contrasts: habitat suitability coverage was broader in spring (0.35) than in fall (0.20). Road proximity dominated spring predictions (MaxEnt Percent Contribution: 41.7–54.9%), while non-vegetated islands were the primary fall driver (46.4–53.2%). Both models identified the same leading predictor in the seasonal Total models. Trap deployment guided by SDM outputs achieved 100% spatial overlap with the maximum training sensitivity plus specificity-defined high-suitability zone — covering only 19.2% of the study area (enrichment ratio ≈ 5.2) — yielding over 18 captures in 2022. We propose a seasonally adaptive management strategy and demonstrate an integrated approach combining MP-based trajectory reconstruction, UAV habitat mapping, and multi-model SDMs for invasive chelonian management in urban wetlands.