Bark beetle outbreaks are a significant and increasing concern in forest ecosystems worldwide, but they pose unique challenges in mountainous regions. Protective forests in these areas play a critical role in protecting communities from natural hazards such as rockfall and avalanches. When bark beetle infestations affect these forests, their protective function may be compromised. However, timely and effective intervention is often complicated due to limited accessibility, and resource constraints in remote, rugged terrain. Early detection of bark beetle infestation is essential for effective forest management, particularly in mountain regions. Identifying infestations at an early stage gives foresters valuable time to plan and implement targeted interventions, increasing the likelihood of controlling small- to medium-sized infestations before they expand. Early detection helps to limit the spread of infestations, ultimately reducing the workload for forest management teams.
In recent years, remote sensing has become an increasingly valuable tool for forest monitoring, with drone technology enabling high spatial and temporal resolution data collection over large areas for detecting forest health decline and pest infestations. However, most drone-based studies on the detection of trees in an early phase of bark beetle infestation have focused on lowland forests or smaller areas, with limited testing in complex mountainous landscapes.
Mountain regions present unique challenges for drone-based monitoring, including complex terrain, variable lighting conditions, and unpredictable weather. Despite these constraints, drones were chosen because they provide high spatial resolution, flexible and repeatable data acquisition, and the ability to operate below cloud cover —
capabilities that are critical for detecting subtle, early-stage stress responses at the individual tree level. In contrast, freely available satellite imagery generally lacks the spatial resolution required for tree-level detection, while manned airborne campaigns are costly, infrequent, and less adaptable to rapidly changing conditions.
In this study, we focused on detecting bark beetle infestations during the “green phase”, the critical window where trees show no visible symptoms yet, but where intervention can still mitigate the spread to surrounding trees. Using a drone equipped with a multispectral camera, we flew repeatedly over forest sites in the cantons of Grisons and Ticino at key moments throughout the beetle's active season. From the collected imagery, we calculated indicators of tree health, known as vegetation indices, that can capture subtle changes in tree stress not visible to the naked eye. We then trained a machine learning model to identify trees in the early stages of infestation, while also considering their proximity to recently killed trees and clearcuts, both of which are known risk factors for bark beetle attack.
When tested on an independent site, the model correctly identified 64% of early infested trees, a promising result given the difficulty of the task. The main challenge lies in distinguishing bark beetle-induced stress from stress caused by other factors, such as disease, drought, or physical damage. In addition, variable lighting conditions between drone flights can affect model consistency. Nevertheless, the ability to detect early stress signals from drone imagery represents a significant advance in forest monitoring. In practice, the method provides foresters with a valuable overview of where stressed trees are located, helping them prioritise specific regions for inspection and intervention in complex mountainous environments. Looking ahead, combining drone-based monitoring with trained scent detection dogs capable of identifying bark beetles’ pheromones could provide a powerful and complementary approach to early warning and rapid response in mountain forests.