Nine watersheds in the Colorado Front Range with steep bedrock channels were used to test the accuracy of paleohydraulic reconstruction of large flash floods using boulder deposits. The nine basins consist of eight small ungauged basins ranging in size from 1.6 to 29 km² and the Big Thompson River at the mouth of the Big Thompson Canyon, draining 790 km². Between 1923 and 1976, all nine basins had had one catastrophic flash flood, the magnitude of which has been estimated by the conventional slope-area method.

In each basin, coarse boulder deposits of the large flash floods were identified, and three axes of the five largest boulders were measured, along with at least two profiles of the valley cross section. A simple arithmetic average of two theoretical and two empirical relationships was used to estimate average flood velocity using boulder size and shape. Average depth was estimated as the arithmetic average of four values computed from the Manning equation, a regression equation for boulder size and unit stream power, a relative smoothness equation, and a modified Shields' relationship. The appropriate flood width for the estimated average depth was found by iteration, using the valley cross sections.

The paleohydraulic discharges thus computed generally underestimate conventional slope-area discharge estimates on small streams by as much as 75%, although the average amount is only 28% too low, and the reconstructed discharge in one stream was 31% too large. The Big Thompson River flood of 1976 was overestimated by 76%. Reasons for discrepancy in reconstructed peaks could include (1) the possibility that floods may have been able to move boulders larger than those available to be moved; (2) overestimation of the slope-area discharge because high-water marks were set prior to erosion of the channel; (3) underestimation of original roughness coefficients; and (4) macroturbulent effects during fast, deep flows.

The paleohydraulic technique is applied to two other streams in Colorado with sedimentological evidence of large flash floods, but no conventional indirect discharge estimates. A small tributary to the Big Thompson River draining 1.8 km² has a paleohydraulic reconstructed flood peak of about 60 m³/s from a flood in 1976. Using boulders excavated from a foundation site in Holocene alluvium along Boulder Creek in Boulder, Colorado, a paleohydraulic reconstructed flood peak of between 860 and 1,512 m³/s is calculated. This is 1.4 to 2.4 times the magnitude of the estimated 500-yr flood.