FLOOD07616

• County: Denver
• Stream Name: Harvard Gulch
• Basin: South Platte
• Title: Application and Evaluation of the Colorado Urban Hydrograph Procedure
• Date: 5/7/2001
• Prepared For: CWCB
• Prepared By: University of Colorado Denver
• Floodplain - Doc Type: Educational/Technical/Reference Information

r

TABLE 5. Difference of Actual vs Predicted for Regressions

PARAMETER DESCRIPTION 5 YEAR 100 YEAR I
Q % Difference Q % Difference ~
Gumbel Regression 763 1544 ,
,
30% Imperv.lSCS CoD 573 -24.9 1987 28.7
40% Imperv.lSCS CoD 872 14.3 2688 74.1
27% Imperv.lSCS CoD 502 -34.2 1805 16.9
30% Imperv.lSCS C-D/fj-1.0 655 -14.1 2081 34.8
30% Imperv.lSCS C-Dlfj=fo-0.5 683 -10.5 2114 36.9
30% Imperv.lSCS B 479 -37.2 1874 21.4
40% Imperv.lSCS B 764 0.1 2568 66.3
Semi-Log Regression 744 1751
30% Imperv.lSCS CoD 573 -23.0 1987 13.5
40% Imperv.lSCS CoD 872 17.2 2688 53.5
Log-Gumbel Regression 765 3397
30% Imperv.lSCS CoD 573 -25.1 1987 -41.5
40% Imperv.lSCS C-D 872 14.0 2688 -20.9
Log-Normal Regression 769 1883
30% Imperv.lSCS B 479 -37.7 1874 -0.5
40% Imperv.lSCS B 764 -0.7 2568 36.4
30% Imperv.lSCS CoD 573 -25.5 1987 5.5
40% Imperv.lSCS CoD 872 13.4 2688 42.8

(greater than observed) bias. In general, the log-normal regression showed
much less high return period bias than the Gumbel method. As stated
previously, the UDFCD precipitation values (and consequently the peak flow
values) show marked exponential growth character and therefore fit much better
with a logarithmic scale. The log-Gumbel regression comparisons for class CoD
soils also correlated well, however neither the AMS data nor the CUHP data fit a
log-Gumbel regression very well (~= 0.9219). In any event, the parameters that
produce peak runoff that best fit the observed data include SCS class B or CoD
soils and an effective imperviousness of 40 percent. Complete details of the
parameters found to best 'calibrate' CUHP are described in table 6.

Application and Evaluation of CUHP

Page 14 of 52