Although paleoflood reconstructions are frequently used to provide dis
charge estimates of extreme floods, a concern remains regarding uncert
ainty in Manning's n-values. An analysis of this uncertainty relative
to a commonly used step-backwater model was performed for channel reac
hes in five canyon rivers. Values of n were selected for each study re
ach using four methods. Each set of n-values was then varied by +/-10%
and +/-25%. The percentage change in discharge associated with varyin
g n-values is inversely proportional to channel gradient and to width/
depth ratio, and proportional to roughness. For channels with a gradie
nt less than or equal to approximately 0.01, varying n by +/-25% produ
ces a maximum change of 20% in discharge. The greatest discrepancy amo
ng n-values estimated using the four methods occurred for channels wit
h relatively low gradient and low roughness, but associated discharge
varied fairly consistently between 13 and 27%. These results indicate
that uncertainties in discharge estimation resulting from the roughnes
s coefficient in step-backwater modeling of paleofloods are comparable
to or lower than uncertainties associated with other methods of indir
ectly estimating flood discharges.