| Abstract: | This paper presents an evaluation of the feasibility and the reliability of a visual characterization technique for gravel–cobble river bed surface substrate. Based on principal axis regressions, using phi scale (?), comparisons of visual estimation and grid sampling techniques show that useful predictive relations (R2 = 0·78–0·88) exist between visual estimates of the surface d16, d50 and d84 and estimates obtained for the same percentiles with the grid sampling technique. Comparisons of visual estimation and the surface‐bulk sampling technique also indicate a predictive relation (R2 = 0·70) between the d50 of the two methods. Trained operators can visually estimate gravel–cobble bed surface d16 to uncertainties of 41 per cent, d50 to 15 per cent and d84 to 11 per cent (for example, there is a 5·5 mm error on a d84 size of 50 mm). Furthermore, evidence shows that if operators are properly trained, a calibration relation for each percentile can be applied independently of operators. This visual characterization allows effective detailed mapping of spatial patterns in substrate size distribution along extensive reaches of gravel‐bed rivers. The technique can be very useful in creating terrain models for various geomorphological, hydrological and biological applications such as the determination of entrainment thresholds, hydraulic roughness and substrate suitability for benthic insects or salmonid habitat. Copyright © 2001 John Wiley & Sons, Ltd. |
