UAV data upscaling for soil erosion monitoring in high-latitude rangelands, northeastern Iceland
Date
2024
Authors
Yerex, Sebastian
Journal Title
Journal ISSN
Volume Title
Publisher
University of Victoria
Abstract
Soil erosion, while a typical geomorphic process, can be amplified and accelerated by land use and climate change. In Iceland, changes in vegetation cover since settlement in the 9th century have led to increased soil erosion. Current field-based methods for erosion mapping and monitoring are difficult and costly to employ frequently over large regions. The systematic and synoptic nature of satellite remote sensing is well-suited for wide-scale environmental monitoring. However, fine-scaled erosion may be obscured in coarse and moderate-resolution data (30-10 m). Here the synergistic use of RGB uncrewed aerial vehicle (UAV) and multispectral Sentinel-2 data is examined to bridge the gap between ground-based and spaceborne monitoring in northeastern Iceland. High resolution (< 5cm) land cover maps from UAV imagery are produced with a random forest (RF) classifier, for six sites. Field validation shows high overall accuracy (> 90%). These data are upscaled to build a RF regression model estimating bare soil cover, yielding good results (R2 = 0.814). Using governmental land monitoring data, erosion severity classes are defined and a map for a portion of northeastern Iceland is produced. This study highlights the potential of multiscale remote sensing for estimating sub-pixel landscape information relevant to environmental monitoring.
Description
Keywords
remote sensing, soil erosion, UAV (uncrewed aerial vehicle), environmental monitoring, sub-pixel information, vegetation cover