Longitudinal segmentation and characterization of river features based on Remote Sensing

Although our understanding of fluvial processes has made significant progress over the last decades, river classification and the derived knowledge about river systems has so far been based on discontinuous sampling along the river course through field works and/or (subjective) interpretation of aerial images. Pioneer continuous longitudinal characterization of river reach or network used aerial photographs and highlighted planimetric patterns (mainly channel style, width and sinuosity) but did not consider elevation which is a very important parameter of channel forms. LiDAR being now more frequently available at large scale, a new step is to introduce such information in the longitudinal river characterization.

We propose a river characterization, applied to 60km of the river Orco, located in Piedmont region, Italy, based on remote sensing data available at regional scale: namely color infrared orthophotos at 40 cm and a LIDAR derived DTM at 5 m acquired simultaneously in 2009-2010. Thirteen geomorphic variables are extracted virtually continuously along the river describing three river components: channel planform features (e.g. number of water channels and sinuosity), floodplain features (e.g. valley bottom shape), channel settings (e.g. width, confinement and slope), in-channel topography (e.g. bed relief index). A multi-dimensional river segmentation is performed applying the Hubert test: river styles are distinguished based on the density of discontinuities in channel planform features. Discontinuities in floodplain and channel settings are related to river style transitions and analyzed quantitatively. Specific river geomorphic signatures based on the distributions of the geomorphic variables are defined for each river style. The proposed segmentation matches remarkably well stretches where distinct channel adjustment processes of river bed aggradation and incision occurred, as shown by 1974 and 2003 cross-section comparisons. The method is completely data driven and should be soon applied at regional scale paving the way towards more quantitative large-scale river hydromorphological characterizations.