Structure from motion (SfM) techniques may estimate three-dimensional structures of river environments from two-dimensional image sequences. A range of possible applications have been developed to collect digital images, including small Unmanned Aerial Systems (sUAS) which are increasingly used for river research and management (Astegiano et al., 2015). Through several data collection campaigns in different river systems in Italy and Japan, in this paper we assess the capabilities of SfM techniques to describe and evaluate instream physical habitat for aquatic fauna.

Specifically, we identified advantages and challenges of this technology to map aquatic and riparian geomorphic features, i.e. wetted area, morphological unit distributions, bathymetry, water surface gradient, grain size distribution, shelters and cover sources for fish. Two different high-resolution (non-metric) cameras (Nikon J1™ and Go-Pro Hero 3 Black Edition™) were used. Images were collected by sUAS to construct a 5-cm resolution orthomosaic image and a digital surface model (DSM). The study assessed the quality of collected data and to initialize common habitat simulation tools (e.g., MesoHABSIM, Mesohabitat Simulation Model, Parasiewicz et al., 2013, Vezza et al., 2014).

Preprocessing of data and object detection procedures were used to automatically extract physical habitat characteristics. High resolution and efficiency in data collection, low cost procedures and flexibility for possible applications were found as advantages of the presented approach. However, limitations due to deep and turbid water, presence of vegetation and white waters were identified for bathymetry extraction and substrates assessment. Results allow us to define the application domain as a reference for future studies.