Research 

The FOReST research group are addressing the need for accurate forest description to support effective planning and forest management, especially in small-scale, fragmented forests or for alternative species. We aim to overcome the limitations of traditional field inventory methods by integrating diverse data types, such as structural, spectral, and environmental information. Our approach combines advanced remote sensing tools, including LiDAR, and multispectral imagery, with traditional field data.

We prioritise high-resolution spatial capture using UAVs for fine-scale analysis, and for larger areas, we integrate airborne LiDAR with passive optical sensors. To analyse this complex data, we use a range of machine and deep learning techniques. We also apply statistical models to integrate environmental factors and predict growth and survival.

Our work addresses challenges in plantations, focusing on species like radiata pine and eucalyptus species. By blending high-resolution remote sensing with advanced modelling, we aim to deliver accurate, spatially explicit estimates of forest resources and provide vital information for sustainable forest management.

The FOReST research group investigate how to build high-resolution, automated geospatial methods that quantify individual tree metrics and reveal how urban tree canopies change  over time. Our work responds to a growing need for precise, efficient measurement of urban forests, where development decisions, extreme weather events, or pests and disease often determine which trees are removed, retained, or able to recover.

We rely on high-resolution remote sensing—LiDAR and aerial imagery, often captured by UAV, and ground-based imagery transformed into 3D models. These datasets are anchored by traditional dendrometry and destructive sampling for validation and enriched with administrative information such as building consents and property-level economic variables. For city-wide monitoring, we assemble multi-temporal airborne datasets that capture canopy change at fine spatial scales.

Our analysis combines deep learning for automated canopy delineation with established techniques. To understand the mechanisms of canopy loss, we model the links between tree metrics, spatial context, and urban pressures using machine learning and traditional regression frameworks. We aim to deliver tools that reveal how individual trees and city canopies respond to rapid urban change.

The FOReST research group advance multi-scale remote sensing and machine learning to classify and monitor forest ecosystems that are dynamic, complex, and often difficult to access. Our work tackles the challenge of mapping fine ecological processes—such as encroachment, degradation, and species turnover—while still covering vast and remote landscapes.

We integrate data across resolutions and sensors, combining LiDAR and multispectral or hyperspectral imagery with medium-resolution satellite time series. Phenological signals extracted from these time series, together with terrain models and rigorous field observations, allow us to distinguish vegetation types with precision. UAV surveys bridge the gap between plot-based measurements and satellite imagery, enabling reliable upscaling to the landscape and regional levels.

Our analyses rely on robust machine learning to reveal the drivers of long-term land-cover change.We apply this remote sensing and analytical framework to diverse and dynamic ecosystems—from temperate grasslands threatened by shrub encroachment to Miombo woodlands, indigenous New Zealand forests, and open landscapes vulnerable to invasive conifers—creating tools that capture ecological change with clarity and scale.