Spatio-temporal assessment of <i>Toumeyella parvicornis</i> spread in mediterranean coastal pine forests through sentinel-2 data

Mediterranean coastal pine forests offer vital ecological and cultural benefits but face threats from climate change, urbanization, and pests. Among invasive agents, Toumeyella parvicornis is a severe threat to Pinus pinea along Italy's coast, causing canopy loss and death. Despite many studies on outbreak detection, the regional-scale dynamics and risk factors are poorly understood. Thus, an interpretable remote sensing framework using multitemporal Sentinel-2 composites to track T. parvicornis infestation in Campania, Italy, from 2017 to 2025 was developed. Seasonal medoid composites and spectral indices were generated in Google Earth Engine. Random Forests and SHAP interpretation assessed infestation patterns, vulnerability, and risk. The workflow combined spectral, temporal, and spatial predictors, including interannual differences, proximity to previous infestations, and transportation infrastructure. Annual attack-transition, vulnerability, and risk maps were produced and validated with photointerpretation data. The attack-transition model had an AUC of 0.79, whereas the vulnerability model showed only moderate discrimination and should be interpreted as a high-sensitivity screening layer rather than as a binary prediction product. The risk model achieved an AUC of 0.80, providing more robust operational prediction of near-future infestation. This scalable framework aids in monitoring invasive pests and understanding environmental conditions and exposure-related spatial factors associated with T. parvicornis spread.