FLOOD09640

• County: Weld
• Basin: South Platte
• Title: Supplementing Flood Frequency Analysis in Eastern Colorado using Rainfall Modeling
• Date: 8/2/1999
• Prepared For: CWCB
• Floodplain - Doc Type: Flood Documentation Report

Bulleim j Ii:> gUidelines. In order to coniinue usmg this method, a procedure is reqUired to adjUst
for the excessive censoring.

Annual flood peaks can generally be considered independent, random quantities (IACWD, 1982;
Salas, 1993). The annual peak flows in censored years may be substituted by a random value
between zero and the minimum recordable flow. The Bulletin 17B procedure can then be
applied to this record to further adjust for high and low outliers and zero-flow years. A sufficient
number of these random events must be generated in order to create an ensemble of equally
likely frequency curves. This analysis used 100 sets of random records for each site. The result
is an upper and lower bound of the frequency curve defined by the maximum and minimum
estimates of the flood frequency relation. A best estimate of the frequency relationship is
derived from the mean ofthe ensemble statistics. Confidence limits are then computed from this
estimate of the frequency curve.

Bulletin 17B was created to provide consistency in flood frequency analysis and methods to deal
with outliers and regional skewness. Since its publication, other methods have become available
and are becoming more popular. One such method is the Expected Moments Algorithm (EMA).
This method is capable of handling censored data and is reported to make more effective use 'Of
historical information over the Bulletin 17B method (NRC, 1999). EMA results are presented
with the Bulletin 17B frequency curves in Section 4.2.4 of this report. A detailed description of
the EMA method is not presented here. The interested reader is referred to referenced
documents (Cohn, Lane, and Baier, 1997; England, 1999) for detailed information.

3.4 Model calibration and validation

Various rainfall-runoff models exist that are capable of facilitating this analysis. The two models
chosen for this study are the Hydrologic Engineering Center Hydrologic Modeling System
(HEC-HMS v1.l) and the Soil Conservation Service Curve Number model (SCS CN). These two
models were chosen because of their popularity and simplicity. One goal of this study is to apply
a reliable, popular model system such that these procedures may be directly applicable to
professionals working on similar topics. These two models are event-based models. Currently,
there is no soil moisture or snow accounting component in HEC-HMS to allow continuous
modeling. Such a component would allow for better simulation of antecedent moisture
conditions and should be considered in future efforts.

The HEC-HMS model is a precipitation-runoff model designed to eventually replace the popular
Uood hydrograph model HEC-I (HEC, 1998). The version utilized here is an early version and
has limited capability. The precipitation and runoff components are fully functional and are all
that are required for the scope of this study. The Green and Ampt infiltration model was used in
HEC-HMS in conjunction with the Snyder unit hydrograph transformation. The Green and
Ampt model parameters are widely published and are largely measurable.

The SCS curve number method is a very simple and widely used method in estimating flood
. hydrogra:phs for small rural basins (Pilgrim and Codery, 1993). The program documentation
suggests this method should be used only on basins less than five square miles in area. All sites
in this study fall below this threshold. An area-weighted curve number is used in all

9

/
V