Exploring the utility of high-resolution UAV imagery for mapping nectar-rich pollinator floral resources in a temperate montane heathland.

• Main Author: Gernot, Niamh
• Other Authors: Nichol, Caroline
• Format: Master Thesis
• Language: English
• Published: The University of Edinburgh 2023
• Subjects: Unmanned aerial vehicles; Remote sensing; Multispectral; RGB; Mapping; Pollinator; Supervised classification; Arctic
• Online Access: https://hdl.handle.net/1842/41203; https://doi.org/10.7488/era/3939

Montane arctic scrub, such as Salix lapponum, forms an important part of temperate montane heathlands in Scotland, however, due to centuries of muirburn and herbivore grazing, remaining populations have become isolated and fragmented. Suggested strategies for aiding the recovery of Salix lapponum at Glen Gaick, Scotland, and adjacent glens is to ensure provision of nectar-rich floral resources surrounding populations to attract pollinators. Vaccinium myrtillus andVaccinium vitis-idaea are important early-season nectar-rich floral resources for pollinators at the site, flowering in synchrony with Salix lapponum. At the time of this study, no research had been conducted on mapping nectar-rich floral resources using UAV imagery at Glen Gaick, for assessing floral resource distribution and cover, specifically around existing populations of Salix lapponum. The aim of this dissertation was to investigate supervised classification methods for quantifying nectar-rich floral resource cover in a temperate montane heathland at Glen Gaick estate, Scotland. This dissertation collected 25m RGB, 30m RGB, and multispectral imagery over a 1ha area at Glen Gaick, containing a population of Salix lapponum. Pixel- and object-based classifications were conducted and compared across the imagery. Spectral separability analysis of the imagery was statistically assessed. This dissertation successfully mapped floral cover of Vaccinium myrtillus and Vaccinium vitis-idaea across RGB and multispectral imagery using pixel- and object-based classification methods, achieving Overall Accuracies between 78-98%. Spectral separability of training samples across 25m and 30m datasets were statistically different from each other. These results demonstrate the opportunities UAV imagery can provide for mapping nectar-rich floral resources at landscape scale.