Novel High-Resolution Three-Dimensional Mapping of Vegetation Using Unmanned Aerial Vehicles (UAV) and Structure from Motion Photogrammetry (SfM).

Forests store around half of the carbon found in terrestrial ecosystems and are a significant sink for rising atmospheric CO2, but monitoring the fine-scale structural changes that are indicative of impacts of climate change and disturbances on long-term forest dynamics is expensive and impractical at large scales. Recent advances in terrestrial and airborne remote sensing are producing, for the first time, high resolution three-dimensional measurements of vegetation structure that have the potential to revolutionise our understanding of biodiversity, species distributions, and ecosystem function. However, the high cost and impracticalities of implementing these technologies limits their application. For example, airborne remote sensing of vegetation usually requires Light Detection and Ranging (LiDAR) and hyperspectral sensors mounted on a plane, meaning studies are often only cost effective over large areas (> 100,000 ha), whilst terrestrial remote sensing, requires expensive laser scanners (TLS), and is limited to small, easily accessible plots (typically > 1ha). Structure from Motion Photogrammetry (SfM) data collected on unmanned aerial vehicles (UAVs) is a promising new method to cheaply and easily collect high resolution 3D data over large areas. Although SfM has been applied in the fields of geomorphology and archaeology for over a decade, it's use in forest ecology is in its infancy with many unresolved uncertainties around best practice for data acquisition and processing.
This project will establish a robust methodology for extracting forest structural traits from SfM and determine accuracy and best-practice in different ecosystems. Predictive models will be developed for accurately estimating forest structure, function and carbon storage from mapped individual tree crowns, a key component of forest dynamics. This methodology will be applied to a range of forest types, demonstrating a new method of low-cost, multi-temporal forest monitoring, capable of quantifying individual tree and whole-forest responses to environmental stressors such as drought and disease. Methods that are developed here are widely applicable, delivering a new framework to overcome common data acquisition issues in some of the most critically understudied regions on the planet; for example, tropical forests which are areas with some of the highest biodiversity on Earth.