Integrating remote sensing and data-driven machine learning for monitoring Ailanthus altissima in heterogeneous Mediterranean landscapes: a case study from Sicily

Abstract

Invasive alien plants threaten Mediterranean ecosystems, where early detection is essential for effective management and biodiversity conservation. Ailanthus altissima is a fast growing, highly competitive tree invader whose ecological plasticity and allelopathic properties intensify impacts on native vegetation, highlighting the need for reliable monitoring. This study develops and evaluates a remote-sensing workflow based on multitemporal PlanetScope imagery and Support Vector Machine (SVM) classification algorithm to map Ailanthus in ecologically sensitive areas. A training dataset was built in the heterogeneous urban park of “ La Favorita ”, characterized by extensive Ailanthus nuclei and diverse land cover types. The trained model was then applied to the “ Vallone Piano della Corte ” Nature Reserve in central Sicily, a sensitive area currently experiencing Ailanthus invasion. Six SVM kernel configurations were tested and validated using field surveys and UAV-based observations. Among them, Coarse Gaussian SVM provided the most balanced performance, effectively detecting the minority invasive class while reliably discriminating dominant land cover categories. Detection accuracy depended on life stage, canopy density, and spatial configuration of Ailanthus , with juvenile or sparse individuals remaining difficult to detect due to spectral mixing and radiometric smoothing due to cubic convolution resampling of PlanetScope images. The use of multitemporal imagery proved essential for leveraging phenological differences between Ailanthus and co-occurring vegetation. The final distribution map supports monitoring, control planning, and ecological restoration, and can be integrated into ecohydrological or vegetation-dynamics models to assess invasion under changing environmental conditions. Overall, the study provides an operational and transferable workflow for invasive species monitoring in Mediterranean landscapes.