Can airborne lidar point clouds quantify grain size contributions to ground sediment facies?
Abstract
Substrate facies monitoring at high temporal and spatial resolution is critical for the understanding of river geomorphologic and ecohydraulic processes. Recent technological advancements in remote sensing with airborne and ground-based lidar as well as drone imagery are stimulating a new era of substrate mapping and quantification. However, most emerging studies focus on mapping substrate based on mean grain size of a specific class of substrate or using only assumed descriptive typologies with no grain size quantification. Further, drone image-based studies tend to be limited to small areas and suffer from uneven lighting, whereas airborne lidar datasets have the potential to capture long river corridors with the same active laser lighting. A key to substrate facies quantification with lidar hinges on quantification of the relationship between microtopography characteristics and grain sizes, especially in size mixtures. Therefore, the aim of this study is to develop methods to generate size-mixture substrate maps and quantify exposed (subaerial) surficial riverbed sediment facies using airborne lidar data of a gravel-cobble river corridor. The selected testbed was the 37.5-km regulated lower Yuba River in north central California, USA mapped in sub-meter resolution in 2014 and again in 2017. We established random ground truth substrate sample plots across the study reaches. Grid-by-number sampling were designed to estimate particle grain size using a gravelometer template. Field-based training data were used to identify mixture substrate classes by multivariate cluster analysis. A multiple supervised classification process was conducted using airborne lidar data and machine learning algorithms programmed in R. Calibration and tuning methods were optimized to analyze classifier performance. Six different substrate mixture classes were identified. Accuracy and uncertainty assessment were performed. A map of the spatial distribution of mixture substrate in the river shows an overall accuracy of 85%. We conclude that the combination of spectral data, specifically Green intensity band and topography information, is highly effective for distinguish the mixed classes of substrates. The size-mixture substrate approach provides valuable new insight regarding river sediment facies patterning.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMEP11C2137D
- Keywords:
-
- 9805 Instruments useful in three or more fields;
- GENERAL OR MISCELLANEOUS;
- 5464 Remote sensing;
- PLANETARY SCIENCES: SOLID SURFACE PLANETS;
- 8040 Remote sensing;
- STRUCTURAL GEOLOGY;
- 8485 Remote sensing of volcanoes;
- VOLCANOLOGY