|
<br />
<br />"
<br />
<br />F.L Ogden lit Ill./ JQllrnlll a/Hydrology 228 (2000) 82-100
<br />
<br />98
<br />
<br />_60
<br />~
<br />~50
<br />.
<br />.
<br />"
<br />0'"
<br />.
<br />"
<br />;30
<br />~ 20
<br />.
<br />"10
<br />
<br />O----'OWSR-88D
<br />G----€>Polanmelflc
<br />
<br />
<br />o
<br />"
<br />
<br />18 19 20 21 22
<br />lime (h) MOT,28 July 1997
<br />
<br />Fl8_ ] I Temporal varialion of basin-average rainfall fate (mmlb)
<br />from polarimetric and WSR-88D radar-rainfall estimates.
<br />
<br />estimates are quite useful for model predictions when
<br />gage adjusted, It is imponant to nOle that the use of
<br />unadjusted single-polarization radar-rainfall estimates
<br />can lead to significant model underestimation of
<br />flooding when the storm structure differs (Petersen
<br />el a\., 1999) from that assumed in the precipitation
<br />processing algorithm.
<br />Calibration of model soil saturated hydraulic
<br />conductivities using simulations of the rise of Horse.
<br />tooth Reservoir as shown in Fig. 4 is a novel approach.
<br />High antecedent moisture conditions lead to a one-
<br />parameter Green and Ampt soil infiltration model
<br />calibration. The fact that the soils in Spring Creek
<br />originated in the Horsetooth reservoir basin contri-
<br />butes to confidence to the transfer of soil parameter
<br />values.
<br />The Spring Creek reference simulation results
<br />shown in Fig. 5 were developed using polarimetric
<br />radar-rainfall estimates, soil hydraulic characteristics
<br />identified in the model calibration on Horsetooth
<br />Reservoir, and literature values of hydraulic rough-
<br />ness and retemion depth. The reference hydrograph
<br />is the best estimate of the actual runoff during this
<br />catastrophic "ood. Uncertainties in the reference
<br />hydrograph are most closely related to uncertainties
<br />in the polarimetric radar-rainfall estimates, which
<br />according to Fig. 3, are modest.
<br />Thiessen polygon interpolation of rain gage
<br />measurements results in overestimation of the stonn
<br />
<br />"
<br />
<br />total rainfall volume. However, CASC2D underesli-
<br />mated the peak: discharge at station U I using Thiessen
<br />polygon rainfall. This is due to smoothing of the rain
<br />field. The rain gages in and around Spring Creek are
<br />spaced far apart relative to the rainfall gradients as
<br />shown in Fig. 6, which produces a large difference
<br />in the spatially-distributed rainfall fields using Thies-
<br />sen polygon and IDS weighting. Deviations in the
<br />predicted runoff are even larger, illustrating the inade-
<br />quacy of rain gage data for hydrologic modeling under
<br />similar conditions with large (up to 40 mmJkm) rain.
<br />fall gradients.
<br />The impact of land surface detail on simulated
<br />runoff is shown in Figs. 9 and 10. The simulation
<br />results given in Table 3 show that neglecting the
<br />spatial variability of soil hydraulic properties and
<br />omitting impervious area increases infiltration by
<br />over 25%. The omission of impervious areas, e.g.
<br />streets and rooftops, decreases the simulated peak
<br />discharge by 6% with. and 3.6% in the absence of,
<br />retention.
<br />Results shown in Fig. 10 indicate that factor of 2
<br />errors in soil saturated hydraulic conductivity result in
<br />considerably smaller errors in predicted peak
<br />discharge for this extreme event. The basin average
<br />saturated hydraulic conductivity value is 0.42 cmlh,
<br />including impervious areas. This value is small
<br />compared to typical convective rainfall intensities.
<br />Fig. 10 is consistent with previous studies (e.g.
<br />Saghafian et aI., 1995) and shows that uncertainty in
<br />soil infiltration rates has a small effect on runoff
<br />predictions for extreme events with Hortonian runoff
<br />production.
<br />Another interesting aspect of the storm of the
<br />evening of 28 July 1997 is the temporal distribution
<br />of rainfalJ intensity over the duration of the storm.
<br />Typically. a gamma distribution is used to describe
<br />the temporal distribution of rainfall in hypothetical
<br />design storms. However, analysis of the polarimetric
<br />and WSR~88D radar-rainfall estimates reveals that
<br />this stomt occurred in 4 distinct pulses, with each
<br />successive pulse stronger than the previous. Fig. II
<br />shows the basin-average rainfall rate from 17:00 to
<br />23:00 MDT over Spring Creek: from both the polari-
<br />metric and unadjusted WSR-88D estimates. Temporal
<br />trends in both radar estimates are the same, but it
<br />appears as though the single-polarization radar-rain.
<br />fall estimates lack sufficient dynamic range. The
<br />
<br />"
<br />
<br />F.L OgJtlf,1 at. I JQjlrnol of Hydrology 228 (2000) 82~J()()
<br />
<br />99
<br />
<br />polarimetric estimates show peak instantaneous rain-
<br />faU rates over the watershed during the four pulses of
<br />rainfall equal to 10, 32, 38, and 60 mmlh. This final
<br />value is quite amazing, considering the fact that the
<br />majority of the rain fell on the western half of the
<br />watershed. (see Fig. 6).
<br />Hydraulically, this rainfall temporal distribution
<br />represents the worst case scenario because subsequent
<br />waves of runoff can propagate and overtake preceding
<br />waves. The large detention basin in Spring Creek
<br />began filling with water about 18:15 MDT after the
<br />first pulse of rain. Simulation results indicate that the
<br />water surface elevation in the detention basin was
<br />within 2 m of the railroad embankment crest when
<br />the fourth and most intense stann pulse began at
<br />20:35 MDT. The temporal distribution of rainfall
<br />played a major role in Ihe severity of the flood.
<br />The two-dimensional, physically.based hydrologic
<br />model CASC2D produces a variety of spatially-varied
<br />output, including discharges and depths. as well as
<br />volumes and watef surface elevations in detention
<br />basins. This additional output comes at the cost of
<br />requiring large quantities of spatially varied input.
<br />However, the data collection and processing effort is
<br />worthwhile, as physically-based input parameters,
<br />low model sensitivity to land-surface uncertainty
<br />and full-unsteady hydrodynamics lend confidence
<br />when modeling extreme events in urbanized catch-
<br />ments. GIS modules and hydrologic model interfaces
<br />simplify input dataset creation (DeBarry et a!., 1999).
<br />Accurate physically based runoff predictions with
<br />extreme rainfall in urban areas require accurate repre-
<br />sentation of the space-time variability of rainfall as
<br />well as accurate rainfall rate estimates. Errors in both
<br />the rainfall rate and the space-time scaling of rainfall
<br />are amplified in runoff predictions. Uncertainty in the
<br />watershed characteristic of input parameter values
<br />propagates into smaller errors in runoff predictions
<br />for extreme events in an urbanized Hononian catch-
<br />ment. The use of spatially uniform soil properties or
<br />the neglect of impervious areas has a considerably
<br />smaller effect than rainfall estimation and interpola-
<br />tion errors.
<br />
<br />Acknowledgements
<br />
<br />The authors thank the following for providing data
<br />
<br />t
<br />.i
<br />
<br />used in this study: The City of Fort CoBins; Northeast
<br />Colorado Water Conservancy District; US Geological
<br />Survey, Colorado District; CSU-CHILL National
<br />Weather Radar Facility; US National Weather
<br />Service; Colorado State Climatologist Office; Moun-
<br />lain Slates Weather Service, Fort Collins, Colorado;
<br />and Riverside Technology, Inc., Fort Collins, Color-
<br />ado. This study was funded by the following US
<br />National Science Foundation SGER Grants: EAR-
<br />9732400, EAR-9732401, EAR-9732402; and the US
<br />Anny Research Office Grant DAAH04-96-I-0026.
<br />
<br />Refe['ences
<br />
<br />Banan, L.J., 1973. Radar Observation of ~ Atmosphcre. 1bc
<br />Univc1Sity of Chir<lgo Press. Chicago.
<br />Beven.KJ..1989.Changingidellsinbydrology-thecaseofpbysi-
<br />cally-based models. 1. Hydrol. 105, t57-172.
<br />Bloschl. G., Grayson. R.B.. Siva~an. M.. 1995, On the represen-
<br />tlltive elementary area (rea) concept lUld itsutililY fordisaibutcd
<br />r:ainfaJl-runolf modelling. H)'drol. Processes 9. JI3-330.
<br />Cbandrasekar, V.. (iorgucci. E., Scarchilli. G.. 1993. OpLimiz..ation
<br />of mulliparameler rddar estimale~ of rainlall. J. AppJ. Meceorol.
<br />32(7}.1288-129J. -
<br />Chandrasek:ar. V., 1997. Personal Communicalions With F.L
<br />Ogden.
<br />Cosla.1.E., t 987. Hydraulics and basin morphometry of tlie largesl
<br />flash floods in the conterminous UOIled Slales.1. Hydro!. 93.
<br />313-338.
<br />DeBalT)', P.A..d aI., 1999. GIS Modules and Distributed Models of
<br />Ihe Walershed. ASCE Tasl Committee Repon. American
<br />Sociely of Civil Engineer,. ReSll:lLl. VA, ISBN 0.7844-0443.7.
<br />120pp.
<br />Doesken. N.J.. McKee, T.B.. 1998. II.n Analy~b of Rainfall for the
<br />July 1997 Rood in Fort Collins. Colorado, Climatology Repon
<br />98-1. DepaltlJ1enl of AlillOSpileOc SCICOOC. ColoradQ SlalC
<br />Univer5ity. FOrt Colhm. Colorado.
<br />Fan.Y..Bras.R.L.,199S.0ntheconceplofrepre5emauveelemen-
<br />tary area III catchment runoff. Hydro\. Processcs 9,821-832.
<br />Freeze,R.A.. Hartan. R.L., 1969. Blueprint foraphysica1ly.based
<br />digitally-simulated hydrological response model. J. Hydro!. 9.
<br />237-258.
<br />Fullon. R.A., Breidenbach. J.P.. Seo. D.-J., Miller. D.A.. 1998. The
<br />WSR.88D rainfall Ollgonthm, We~ther and Forecasting 13,
<br />410-415
<br />Gorgucci. E.. SCllrchilli, G.. Ch:lIldra.sc:kar. V.. 1994. Rob.ust esti-
<br />malor of rainfall rale using dJfferemial retlectivity.1. Aunos.
<br />OcC;ul. Techno!. II (2).586-592
<br />Gr:lyson, R.B., Blosch!. G.. Moore. I.D.. t995. Distributed para-
<br />meler hydrologic moddlinj; using veClor elevation dOlta'
<br />THALES and TAPES-C. In: Sin~h. V,P. (Ed.). Compulcr
<br />Modds of Watershed Hydrology. Water Res Pub!. Highlands
<br />ROlllrh. CO. pp. 669-696
<br />
|