|
<br />
<br />!I
<br />
<br />Journal
<br />of
<br />Hydrology
<br />
<br />ELSEVIER
<br />
<br />Journal of Hydrology 228 (2000) 82-100
<br />#-/-l-
<br />
<br />www.clscvicr.comllocaleJjh).drol
<br />
<br />Hydrologic analysis of the Fort Collins, Colorado,
<br />flash flood of 1997
<br />
<br />F.L. Ogden"', H.O. Sharif', S.U.S. Senarath', J.A. Smithb, M.L. Baeckb.
<br />J.R. Richardson'
<br />
<br />'Dtpanment of Civil and ElIv;'OIIlMnlal Engifltlering. U-37 Universiry of COMecticul, Storrs. cr 06269. USA
<br />"DefH/nfMnI of Civil lJnd E,wironmenta/ Engineering, Princeton University. Prinufon, NJ 08544, USA.
<br />'DeJX'nment of Civil and Envirorrmemal Enginuring. Univusiry of Missouri. Kansas City. MO 64/10, USA.
<br />
<br />Received 8 MilR:h 1999; received in revised form 26 October 1999; acccpted 15 December 1999
<br />
<br />Abstracl
<br />
<br />On 28 July 1997, an unusually moisl air mass, driven westward towards the foothills of the Rocky Mountains near Fort
<br />Collins, Colorado. produced torrential rainfall. The nearly saturated atmospheric column in conjunclion wilh lighl upper-level
<br />winds resulted in a "warm rain" proo:;:ess convective Slonn with little net mOlion. On the evening of 28 July over 200 mm of rain
<br />fell in western Fort Collins. This eX(feme rainstonn was observed by lhree S-band weather radars, including two National
<br />Wealher Service WSR-88D radars and the dual-polariz.ation CSU-CHILL radar. Fourteen recording rain gages in and near the
<br />affected area recorded the event. The US Geological Survey, Colorado District, periormed indirect peak discharge measure-
<br />ments.In our analysis, the lwo+dimensional, physically-based hydrologic model CASC2D is applied to examine the influence
<br />of rainfall and land surface data uncertainty on runoff predictions in the 25 km~ Spring Creek watershed. Soil salurated
<br />hydrauhc conductivity values are calibrated in simulations of the rise in nearby Horsetooth Reservoir. Results of simulations
<br />driven by polarimetric and single-polariz.ation radar-rainfall estimates and recording rain gage data show thai for this extreme
<br />event in an urbanized watershed, rainfall estimation errors give rise to the most significant errors in runoff predictions.
<br />Hydrologic simulations with various levels of land-suriace detail reveal that uncertainty in watershed characteristics has a
<br />considerably smaller effect on runoff predictions than uncenainty in the space/time distribution of rainfall. The soil saturated
<br />hydraulic conductivity, fraction of impervious area, and the retention depth are the most sensitive land-suriace parameters.
<br />C2000 Elsevier Science B.V. All rights reserved.
<br />Kqwords; Urban hydrology; Aash nood: Radar-rainfall; Hydrologic modeling: Distribuled hydrology: Extreme event; CASC2D
<br />
<br />1. Introduction
<br />
<br />in: waler resources assessment, irrigation design,
<br />prediction of erosion and sedimentation, aquatic ecol-
<br />ogy, evaluation of climate change impacts and subsur-
<br />face water conlaminalion (Refsgaard and Abbott,
<br />1996). Hydrologic modeling is useful for predicting
<br />lhe response of ungaged catchmenls to hydrologic
<br />inputs, flood forecasting, and estimation of missing
<br />flood records from recorded rainfall. Physically
<br />based hydrologic models are particularly useful in
<br />
<br />Since lhe early work of Freeze and Harlan (1969),
<br />physically based hydrologic models have been
<br />applied 10 various water resources problems. Some
<br />of lhe main applications of hydrologic models are
<br />
<br />. COITCspondingaulhur. Fax: +].860-486-2298.
<br />E-mlJillJddrtss:ogden@engr.uconn.edu(F.L.Ogden).
<br />
<br />0022-16941()(tf$ _ see from maner e 2000 Elsevier Science B.V. All rights reserved.
<br />PU: 50022-]694(00)00]46-3
<br />
<br />FL Ogd~" ~llJl. / }ounuJI of H.~drology 228 (2000) 82-JOO
<br />
<br />83
<br />
<br />
<br />~DelenllonBasin
<br />. Perennial Lake
<br />fiJl USGSlndue<:1
<br />L.::..J OlschargeSne
<br />
<br />Fi~. I. Spring Creek watershed hydro~raphy. major stream crossings, and locations of USGS indirecl discharge measurements.
<br />
<br />modeling floods in ungaged catchments where cali-
<br />bration data do not exist for the calibration of concep-
<br />tual hydrologic models, or for simulating extreme
<br />storms beyond the limits of existing cali bra lion data.
<br />Although conceplual models are heavily used in water
<br />resources assessment, distributed physically based
<br />models have the advantage that lheir input parameters
<br />are more closely related 10 Ihe physical attributes of
<br />the catchment. An example of a contrary opinion is
<br />given by Beven (1989). Woolhiser (1996) states Ihat
<br />physically based models can be used for predictive
<br />purposes in small walersheds when overland flow is
<br />generated by the infiltration-excess mechanism.
<br />The use of a physically based hydrologic model is
<br />necessary in this urban watershed because of the high
<br />spalial variability of land surface parameters, and the
<br />absence of calibralion data. The existence of signifi-
<br />cant hydraulic sUllctures adds to Ihe complexity of the
<br />watershed, which further requires the use of a distrib-
<br />uted hydrologic model. In this study, the intent is to
<br />examine the effect of spatially varied rainfall and
<br />watershed characteristics on the runoff modeling
<br />processes, which prevents Ihe use of lumped models.
<br />On the other hand, the availability of temporally and
<br />spatially distributed rainfall, distributed land-use!
<br />land-cover, soil properties and hydraulic data for
<br />Ihis watershed provides an excellent opportunity to
<br />apply a distributed hydrologic model.
<br />The continental US has experiencd a large number
<br />of extreme weather-related events including hurri-
<br />canes. tornadoes, and small-scale and regional-scale
<br />flooding in the 1990s (Keirn, 1998), an outstanding
<br />example is the Great Flood of 1993 (NWS, 1994) on
<br />
<br />i:
<br />
<br />the upper Mississippi and lower Missouri rivers. In the
<br />United States, floods, and in particular flash floods, are
<br />responsible for more weather-related deaths than any
<br />other type of weather event, resulting in an average of
<br />over 100 deaths per year (Shelton and May, 1996).
<br />Failures of hydraulic structures contribute signifi-
<br />cantly to flood damage.'The foothills of the Appala-
<br />chian and Rocky Mountains have been lhe stage for
<br />numerous extreme stonns in'the I 990s (Lande! et aI.,
<br />1999). Some small watersheds have experienced flash
<br />floods with return periods in excess of 400 years, such
<br />as the Rapidan River in Virginia during June 1995.
<br />On 28 July 1997, a flash flood occurred in Fort
<br />Collins, Colorado. The flooding affected large
<br />portions of Fort Collins, but was focussed in the
<br />25 km ~ Spring Creek watershed, which is shown in
<br />Fig. I. The peak discharge at the outlet of Spring
<br />Creek exceeded twice the 500-year flood (Smith,
<br />1997). This flood was triggered by an extreme preci-
<br />pitation event that is currently the largest ever
<br />recorded, in terms of short-duration rainfall, over a
<br />developed urban area in Colorado (Doesken and
<br />McKee, ]998). The flood caused over a 100 million
<br />dollars damage 10 Colorado State University campus
<br />alone, 200 homes were lost and 1500 homes were
<br />damaged (Grigg et aI., 1999).
<br />In the late evening of28 July 1997, an intentionally
<br />plugged box culvert through a 6 m high railroad
<br />embankment across Spring Creek became unplugged.
<br />The subsequent rapid release of floodwater flowed
<br />through a mobile home park immediately down-
<br />stream, resulting in the unfonunate drowning deaths
<br />of five people. The railroad embankment was also
<br />
|