EARLY DETECTION ACROSS SCALES Remote sensing for early detection and mapping of the European spruce bark beetle outbreaks in the South-Eastern Alps

Climate change increases the vulnerability of European forests by intensifying extreme weather events and favouring outbreaks of pests like the European spruce bark beetle Ips typographus (L.). In the Southeastern Alps, severe I. typographus infestations were triggered by the 2018 Vaia storm and quickly spread over many spruce forests survived to Vaia. Early detection of infested trees is a critical but challenging point, especially at large scales. This thesis evaluates early detection performance using multispectral Remote Sensing (RS) data, ranging from open-source satellite imagery to high-resolution drone imagery. Aims include assessing early detection limitations at large scale, identifying key factors affecting early symptom detection, and testing whether high-resolution data can improve early detection upscaling. At the large scale, results showed that coarse-resolution satellite imagery (from 3 to 10 m resolution) can detect infestation hotspots even in case of small or mixed infestation spots and can improve detection of early symptoms especially when combined with intensive ground data. In this sense, participatory ground survey (i.e. field data collection by non- scientific personnel) proved to be an important complement to satellite data, especially for training and validating satellite-based observations. At smaller scales, analysis of drone imagery showed that epidemic beetle populations triggered faster symptom onset and successfully separated infested from healthy trees, sometimes weeks before symptoms were visible to the human eye. Despite processing limits when upscaling early detection using high-resolution imagery over large areas, NDRE (Normalized Difference Red Edge) and GNDVI (Green Normalized Difference Vegetation Index) were top-performing indices, highlighting the value of red-edge and green bands for early detection tasks. Since early detection success was influenced by bark beetle population density, tree size, local environmental conditions, and image-processing challenges when scaling up, future work should focus on the integration of field reference, high-resolution RS and coarse- resolution RS for improved early detection especially over vast areas. Such multi-scale approach could effectively support forest management to promote forest resilience against disturbances induced by global change.