Comparison of software accuracy to estimate the bed grain size distribution from digital images: A test performed along the Rhine River

The quantification of the bed grain size distribution (GSD) river surfaces is primarily conducted through manual approaches in field. These methods are time consuming and not able to accurately represent spatial diversity rivers. Recently, several software programs procedures have been developed using semi-automatic automatic estimate GSD from digital imagery. purpose this study compare softwares accuracy between reference GSDs estimated geometric (Basegrain a procedure on ImageJ), statistical (digital [DGS] PebbleCounts softwares), machine learning framework (SediNet). This evaluates ten images recorded along Rhine River downstream city Basel. results showed that all considerably underestimated manually measured GSDs. Nevertheless, it possible significantly improve estimation by applying calibration laws. Both DGS Basegrain reliable GSD, while three others accurate for percentiles equal higher than D50. After linear regression correction, mean normalized root square error percentile errors did exceed 13% software, 22% coarser