Monitoring invasive and expansive species in the Krkonoše Mts using UAV multitemporal data and botanical research

Abstract. The Krkonoše Mountains are a unique ecological region facing increasing pressure from alien invasive and native expansive plant species, which threaten biodiversity and ecosystem stability. This study investigates the potential of UAV-based remote sensing for detecting and monitoring selected invasive species with focus on Lupinus polyphyllus. The primary objectives were (1) to acquire UAV multispectral data for several plots at multiple time points during the growing season in order to identify the best dates for the species detection, (2) to collect reference botanical data, (3) to test the suitability and reliability of mapping invasive/expansive species from UAV imagery using deep learning methods, and (4) to evaluate the effectiveness of various management interventions. High-resolution UAV imagery was processed using the SegUNet deep learning model, achieving classification accuracies up to 95.7%. The results indicated that species detection was most effective during flowering but also viable in spring due to distinct leaf morphology. One of the key findings of the analysis is that centimeter-range spatial resolution enables the detection and monitoring of Lupinus and other species during their growth and flowering stages, to a significant degree without requiring botanical input data. Our study confirms the applicability of UAV remote sensing for invasive species detection, offering a cost-effective and scalable solution for landscape-level monitoring in the future. Long-term monitoring will be essential for refining detection strategies, improving classification models, and testing the reliability, especially for detection after management interventions.