FLOOD04478

• County: Larimer
• Community: Fort Collins
• Basin: South Platte
• Title: Hydrologic Analysis of the Fort Collins Flash Flood 1997
• Date: 12/15/1999
• Prepared For: Journal of Hydrology
• Prepared By: Elsevier Science Publishers
• Floodplain - Doc Type: Flood Documentation Report

"

FL Ogden ef Dt.1 JOllmo/ of Hydrolog)' 22E (2()()()) 82-]00
300

'" 200
"E
.
"
.
.
~
c5 100

U""'''~'.Mi
DiochargeMeUUfS""'nl I \
alSlallOnUl \
1\
I \
I \
{ \
{ \
I t"c
{
{
{
I
I
I
{'
{
/

o
18:00:00

--- SIationU1
Station U2
- Slalion U3

21 :00:00 00:00:00 03:00:00
Time, July 28/29,1997 (MOT)

Fig. 5. Reference simulalionrcsuhs.

affected this flood is the large detention basin behind
the 6 ill high railroad embankment (see Fig. I). The
Importance of other lakes and detention basins was
secondary but they were also included in model A
large lake formed upstream from the railroad emba~k-
mem as the three 2.1 ill (7 ft)diameterculverts did not
have the capaen}' to pass the inflow to the detention
~asln. Dunng the morning storm on 28 July, simula-
lIOns show the lake behind lhe main detention basin
grew to an area of about 0.02 km2 and then shrank to a
ffil.n.Jm,um area of 0.009 km2 before starting to expand
agam as the major storm stalled in the early evening
of ~8 July 1997. The lake ultimately expanded 10 an

area of more than 0.6 km2 inundating two kilometers of
channel reach and overtopping the railroad embank-
menl Shortly aftelWards, at about 22:55 MDT the box
culvert started to unplug. The unplugging of the culvert
~as assumed to take 18 min and the area of the opening
IS assu~ed to grow exponentially before expanding to
the full area of the c~lve~ cross-section. The timing of
the culvert unplugglOg IS estimated from records of
emergency telephone calls in the immediate area
downstream from the railroad embankment.
Referen.ce simulation results ploned Fig. 5 show
~hat the discharge at the watershed outlet is greatly
lllfluenced by the detention basin and by the culvert

Fig. 6. Polanm"lrlC SlOmHOlal rad.ar esti I d . f Jl
ma e ram a accumulalions (mmI17:JO-23:30 MDT. 28 July 1997.

I-- 3 km---l

300

FL 08d~n ~r ,,1.1 j""ffl,,1 ,,{Hydrology 228 (2000) 82-)00

93

ft!!!. 200
E
.
o
.
~
o 100

{
I
I
I
I
I
,
/

o
18:00:00

... AdJUSt. WSR-68D
_ _ - Unadj. WSR-88D
_ Rete/ence Run

",
\
\
\
\
, ,
"-

21 :00:00 00:00:00 03:00:00
Time. July 28129, 1997 (MDT)

Fig. 7. Simulalion resuhs at slalion VI using unadjusted and adjuSled WSR.88D precipitation estimates.

unplugging. Notice the very sharp rise in outflow at
station U2 (see Fig. 1) due to culvert unplugging.
Given the uncertainty regarding the timing and rate of
culvert unplugging, the discharge at station U 1 is the
moSI useful in evaluating the effects of input uncertainty
on model performance. Furthermore, most of the runoff
originates upstream from slation U 1.
Fig. 6 shows polarimetric radar-estimated storm total
rainfall accumulations in rom between 17:30 and 23:00
MDT over the Spring Creek watershed. The gradient in
stonn-total rainfall over the eastern half of the watershed
is very steep, up to 40 mm/km. The land-use map is
shown in the background in Fig. 6 to illustrated the
fact that large portions of the watershed are urbanized
where the most intense rainfall occurred.
The simulated peak discharge in the reference run
at station U I was 17.5% higher than the USGS indir-
ect measurement but still within the uncertainty range
of ::!:25%. The model estimated the peak discharge at
the outlet of the catchment to be 172 m3Js (3.5%
higher than the indirect estimate) which was well
within the uncertainlY range of :!:15%. Reference
run simulation results are listed in Table 3.

12. Simulations witb WSR-88D radar-rainfall
estimates

The uncalibrated single-polarization WSR-88D

observations significantly underestimated the rainfall
from the stann. The WSR-88D estimated rainfall
volume on the watershed area was 42% less than the
reference rainfall volume. The runoff volume calcu-
lated using unadjusted WSR-88D rainfall estimates
was 52% less than the reference run value indicating
the rainfall estimation error is amplified as it propa~
gates through the runoff model.
The WSR.88D estimates were adjusted by multi.
plying by a constant bias-adjustment factor (Smith et
at, 1996) equal to 1.72 so that the slOrm total rainfall
volume is equal to the reference rainfall volume. Fig.
7 compares simulated hydro graphs at station U I for
the reference run, unadjusted and adjusted WSR-88D
rainfall inputs. Although the peak discharge at station
VI is 7% higher using the adjusted WSR-88D rainfall,
the total runoff volume is 1.2% lower than the refer-
ence value. This finding indicates that the bias-adjust-
ment faclOr is nol constant over the life of the storm.
Notice that the simulated hydrograph produced using
storm-Iotal bias adjusted WSR-88D rainfaU estimates
changes the temporal distribution of runoff, as evident
by differences between the reference and adjusted
WSR-88D hydrographs on Fig. 7. Differences in
runoff volume and hydrograph timing between simu-
lations have a large effect on the simulated conditions
in the detention basin making it meaningless to
compare the simula'ted hydrographs at the watershed
outlet for these simulations. The volume of infiltration