Robust estimations of areal grain size distribution from geometric surface roughness in a proglacial outwash area

Abstract

Rapid glacier mass loss in high-alpine catchments exposes unconsolidated sediments to erosion by meltwater-fed mountain streams, with impact on the catchment-scale sediment dynamics. Proglacial outwash areas (OWA) are a key spatial element in this sediment flux cascade, with grain size distributions (GSD) as an integral variable in current and future glacio-fluvial processes and associated effects.In this study, surface grain size estimation methods based on linear correlations between geometric surface roughness and GSD data are adapted to test this new mapping approach for areal GSD patterns and their changes in a proglacial OWA. Surface roughness is computed with adaptive kernel sizes from high-resolution topographic point clouds derived from UAV-based photogrammetry. For model calibration, ground truth data is manually obtained from 30 dry-exposed GSD sampling patches of 1 m2 each. The novelty and potential of this surface roughness regression model lies in its capability to robustly map GSD percentiles at 1-m cell resolution in highly dynamic proglacial environments and inherent data scarcity, drastically reducing the portion of labour-intensive and time-consuming manual surveying.Our optimized linear regression function for the key figure D84, which is widely used in practice, returns a coefficient of determination of 76.4 %. The linear relation found between the D84 of field GSD and the 84th distribution percentile of surface roughness metrics has a slope of α = 6.01, which is consistent with the range of recently published studies using LiDAR-based and photogrammetric surveys. The presented surface roughness regression model provides realistic surficial estimations of areal GSD patterns in a proglacial domain of sediment remobilization. In addition, this work analyses and discusses detected changes in the estimated D84 over a three-month period. A systematic shift is identified for sediment compositions with an initial D84 above 60 mm. This is thought to be an effect of the original image quality and variable lighting conditions between photogrammetric surveys and indicates that best practice should include careful survey planning for most uniform flight settings. Nevertheless, visually verified significant changes in D84 along frequently wet low-flow channels can be detected automatically, and thus the findings demonstrate the feasibility of this GSD monitoring approach for tracking fluvial re-working of surface sediment.