Subaerial gravel size measurement using topographic data derived from a UAV-SfM approach
Accurate and reliable methods for quantifying grain size are important for river science, management and in various other sedimentological settings. Remote sensing offers methods of quantifying grain size, typically providing; (a) coarse outputs (c. 1 m) at the catchment scale where individual grains are at subpixel level, or; (b) fine resolution outputs (c. 1 mm) at the patch scale. Recently, approaches using unmanned aerial vehicles (UAVs) have started to fill the gap between these scales, providing hyperspatial resolution data (< 10 cm) over reaches a few hundred metres in length, where individual grains are at suprapixel level. This ‘mesoscale’ is critical to habitat assessments. Most existing UAV-based approaches use two-dimensional (2D) textural variables to predict grain size. Validation of results is largely absent however, despite significant differences in platform stability and image quality obtained by manned aircraft versus UAVs. Here, we provide the first quantitative assessment of the accuracy and precision of grain size estimates produced from a 2D image texture approach. Furthermore, we present a new method which predicts subaerial gravel size using three-dimensional (3D) topographic data derived from UAV imagery. Data is collected from a small gravel-bed river in Cumbria, UK. Results indicate that our new topographic method gives more accurate measures of grain size (mean residual error -0.0001 m). Better results for the image texture method may be precluded by our choice of texture measure, the scale of analysis or the effects of image blur resulting from an inadequate camera gimbal. We suggest that at our scale of assessment, grain size is more strongly related to 3D variation in elevation than to the 2D textural patterns expressed within the imagery. With on-going improvements, our novel method has potential as the first grain size quantification approach where a trade-off between coverage and resolution is not necessary or inherent.
This research was funded by the University ofWorcester (PhD studentship for Amy Woodget) and a student researchaward from the Geological Remote Sensing Group, a special interestgroup of The Geological Society of London and the Remote Sensingand Photogrammetry Society.
- Social Sciences
- Geography and Environment