Forest Health Index (2022-2026)
Project Overview
Forest health is a qualitative term referring to the general condition of a forest, encompassing resilience to disturbances such as insect infestations, disease, fire, and flooding. Despite its importance for conservation and ecological sustainability, there are no well-accepted methodologies to quantitatively measure forest health. Existing frameworks remain high-level, lacking specific formulas to operationalise assessment of a given forest area. IoT sensors can accurately measure environmental parameters at ground level, but deploying large-scale sensor networks in forests is challenging due to hardware costs, energy provisioning, and communication difficulties. Drones offer scalability by observing forests from above, yet they cannot directly sense conditions at the soil level. This project aims to combine IoT sensor data with drone imaging to develop a scalable forest health index by training machine learning models that use aerial imagery to predict ground-level environmental conditions.
Both IoT devices and drones are deployed in a test forest environment to train a machine learning model. The model learns to use drone imagery to predict IoT sensor outcomes and, subsequently, forest health. Satellite imagery may also complement these data sources. Drones rely on proxy measurements derived from canopy characteristics, which are correlated against ground-truth sensor readings to build predictive capability.
The ultimate objective is to develop an AI system that can produce a forest health index from drone images alone. This would require only a limited number of ground-truth IoT sensors for calibration, making the approach scalable to large forest areas where dense sensor deployment is impractical.
Team
