This paper shows the usefulness of the scale-space filtering technique
for the analysis of stream-bed roughness. In the first part of the pa
per, the scale-space filtering technique originally developed for sign
al analysis is adapted for the analysis of stream-bed microtopographic
profiles. The modified technique provides a multiscale description of
bed profiles, which allows the objective identification and measureme
nt of individual roughness elements at all scales of observation. Thes
e bed elements then are classified as either grain or form roughness e
lements according to the roughness scales identified from a semivariog
ram analysis of the bed profile. Roughness indices describing each rou
gh,less scale are obtained by calculating the mean height and spacing
of the grain and form roughness elements. Because they measure directl
y the geometry of the bed, these indices provide a better description
of stream-bed roughness than the traditional method based on a single
grain index. in the second part of the paper, the technique is applied
to the analysis of stream-bed roughness and resistance reflow of nine
different gravel-bed stream reaches. The roughness indices calculated
from this approach are used to obtain predicted values of resistance
to pow which are in good agreement with the values observed in the fie
ld. The quality of this prediction is compared to the one obtained fro
m a traditional description of bed roughness based on a single grain-s
ize index. The results indicate that the values of resistance to flow
calculated using the traditional approach are significantly less accur
ate than the values calculated using the new procedure which separates
stream-bed roughness into a grain and a form component.