FLOOD06904 - Laserfiche WebLink

Home Browse Search

50 LAWN LAKE DAM AND CASCADE LAKE DAM FAILURES, COLORADO subcritical flow was assumed. Computational cross sec- tions were spaced at intervals that ranged from 0.02 mi to 0.6 mi, depending on the segment of channel. Im- mediately downstream from each dam, computational cross sections were closely spaced to produce better results where the flood hydrograph was changing rapidly. The time step of the computations was computed automatically; therefore, it was not directly controlled. Generally, the time step varied by the position of the flood wave. That is, during gradual changes in stage, the time step was longest (a maximum of 15 min), and dur- ing the maximum development of the breach and near the peak flow when stages are changing rapidly the time step was shortest (3 seconds). Base flow used to start the model ranged from 250 ft'/s for reaches downstream from Lawn Lake dam to 400 ft'ls for reaches down- stream from Cascade Lake dam. These base flows were higher than actual conditions, but were required by the model for computational purposes. Estimated tributary inflows were 150 ft'/s for the Fall River and 385 ft'ls for the Big Thompson River. These inflows reflected esti- mated base flows of these rivers and other minor tribu- tary base flows. The initial water-surface elevation at each cross section (surveyed and computational) was com- puted by the normal depth method by the model. The model's computational cross-section smoothing option was not used. The hydrologic routing method was used for each dam to compute the dam-breach hydrographs. Storage- capacity data and elevation of water surface at the time of breach are shown in table 3 for each dam. Since Lawn Lake dam probably failed from piping, the initial water surface was set at 0.01 ft below the water surface at time of failure. The initial water-surface elevation for Lawn Lake dam was 10,996.41 ft. It was assumed that the trapezoidal breach function in the model (an overtopp- ing failure) would adequately model a piping failure, since the piping breach developed so rapidly. An inflow discharge of 350 ft'/s was used to raise the water sur- face sufficiently to cause the Lawn Lake dam to fail by overtopping (although the actual inflow probably was about 25 ft'ls). An average side slope of the breach (Lawn Lake dam=0.84; Cascade Lake dam=2.77), eleva- tion of breach crest (Lawn Lake dam=10,975 ft; Cascade Lake dam=8,460 ft) and bottom breach width at end of failure (Lawn Lake dam=55 ft; Cascade Lake dam=80 ft) were obtained from the cross-section data in the Sup- plemental Cross-Section Data section at the end of the report; these characteristics are shown in those sup- plemental data for each dam. 1b operate the model in the Roaring River (discussed later), an outlet discharge of 250 ft'ls was used. Prob- ably less than 5 ft'/s flowed out the overflow. An outlet discharge of 400 ft'ls was used for Cascade Lake dam. Artificially high base flow in a reach of river is often re- quired to operate the hydraulic routing component due to computational difficulties when flow depths are shallow (Cunge and others, 1980, p. 176). These high base flows are small in relation to the observed peaks and do not affect the flood-wave hydraulics. Elevations of the crests of the dams used in the model were 10,996.42 ft (actual water level at time of failure) for Lawn Lake dam and 8,471.55 ft for Cascade Lake dam. The duration of breach development, based on available data and judg- ment, was 10 min for Lawn Lake dam and 1 min for Cascade Lake dam. Although Lawn Lake dam may have allowed a small rate of water to escape for several hours prior to 0530 MITT, it was felt that the erosion of the em- bankment to full breach width took about 10 min. Cas- cade Lake dam toppled instantaneously due to 0vertop- ping, but it was estimated it took about 1 min for the water to move the remains of the dam and allow the water to escape freely. The duration of model simulation was 5 hours, which covered the passage of most of the flood wave. Over one-half the length of the Roaring River channel was extensively scoured, and the flood wave contained large amounts of debris and boulders. Rapid channel changes occurred during the flood passage, and the cross sections probably do not reflect these conditions. Similar- ly, high-water marks probably were set before the majori- ty of scour; therefore, flood depths computed from these data were not accurate. The postflood surveyed cross sec- tions shown in Supplemental Cross-Secti.on Data were us- ed in the hydraulic routing. Debris and boulders had an unknown effect on the flow characteristics (particularly in the Roaring River), but certainly did not meet the model assumption of clear waterflow. These problems were recognized, but accurate solutions, if possible, are very complex. Therefore, the objective of modeling flow in the Roaring River was to develop a breach hydrograph and route it downstream to give reasonable results based on the observed traveltime and documented peak dis- charges farther downstream. Small bridges and culverts in the flood path were assumed to have negligible effects on flow conditions. MODEL CALIBRATION Model calibration was made to obtain the best fit, based on observed peak discharges (table 2), flow depths (table 4), and traveltime (table 6). Parameters available for calibration in the model were Manning's roughness coefficient (n-values), expansion and contraction coeffi- cients. and designation of cross-section active and inac- tive flow areas. As described in the section "Peak Stage and Discharge," the selection of n-values was extremely