Geostatistical estimations of bathymetric LiDAR errors on rivers
Article first published online: 3 JUN 2010
Copyright © 2010 John Wiley & Sons, Ltd.
Earth Surface Processes and Landforms
Volume 35, Issue 10, pages 1199–1210, August 2010
How to Cite
Bailly, J.-S., Le Coarer, Y., Languille, P., Stigermark, C.-J. and Allouis, T. (2010), Geostatistical estimations of bathymetric LiDAR errors on rivers. Earth Surf. Process. Landforms, 35: 1199–1210. doi: 10.1002/esp.1991
- Issue published online: 20 JUL 2010
- Article first published online: 3 JUN 2010
- Manuscript Accepted: 8 DEC 2009
- Manuscript Revised: 17 NOV 2009
- Manuscript Received: 14 SEP 2009
- laser scanning;
- block kriging;
- channel-fitted coordinates;
The geometry of river channels is a key descriptive element for hydromorphology, hydraulics and hydroecology. Gravel bed rivers usually have a mean water depth of ∼0·5 m. For such shallow waters, the accuracy of bathymetric LiDAR data has to be precisely assessed. Alongside this accuracy investigation, methodological questions arise: How to assess the data quality of elevation LiDAR when comparing reference topographic points on river beds to laser beam footprints of several square metres at different locations? What are the consequences of uncertainties and scaling in accuracy estimations? In this study, we designed a methodology to assess the quality of LiDAR topographical data within rivers using a specific geostatistical method that conducts upscaling as well as interpolation of reference data that takes into account uncertainties. This method uses an anisotropic block kriging from DGPS points on LiDAR footprint areas within a channel-fitted coordinate system. This assessment focused on a 1·5 km long reach of the Gardon gravel bed river, in the south of France. DGPS points pseudo-regularly located along the river were acquired at the same time as the LiDAR survey with the HawkEyeII system. LiDAR accuracy results for river bottom elevation show a negative bias for high depth. Added to that bias, a random error with 0·32 m standard deviation was found by considering upscaling and uncertainties in reference data, and a 0·20 m standard deviation was found if they were not considered. Consequently, if LiDAR bias can be corrected, measuring a water depth less than 32 cm, i.e. for 28% of the river area, is unrealistic.
However, this experiment shows that LiDAR provides an accurate representation of the riverbed forms. It also provided a useful, continuous, topographic surface from the underwater river bed up to riparian areas. Copyright © 2010 John Wiley and Sons, Ltd.