FLOOD04478 - Laserfiche WebLink

Home Browse Search

84 F.L Ogden el a/./ JOUT1\Q1 of Hydrology 228 (2000) 82-/00 available on the www at the following URL: hup://honon .engr. uconn .edu/FortColl i ns/Mai n .hun I. overtopped, which undermined several railroad ties and caused a moving train to derail. Four railroad cars left the tracks. The next railroad car. which thank- fully did not derail, was carrying liquid Chlorine. The unit discharge (Costa, 1987) in Spring Creek upstream from the railroad embankmem was nearly 15 m) S~I km-2. The severity of the f1.00d and the relatively short warning both comribUlcd to the tremendous damages and loss of life. 2. Objectives The overall objective of this study is to document the rainfall and runoff thaI resulted in the catastrophic flash flood on Spring Creek, and use this information to evaluate the effect of rainfall and land~surface uncertainty on hydrologic predictions of extreme evenLS in urban environments. This overall objective necessitates several specific tasks; characterize the watershed conditions that existed before and during the flash flood; calibrate and verify radar-rainfall esti- males from the CSU-CHILL S-band dual-polarization research radar located near Greeley, CO and the S- band single-polarization WSR-88D (NEXRAD) radar located near Cheyenne, WY; apply the rainfall and watershed data to simulate the runoff event using the two-dimensional, physically-based hydrologic model CASC2D (Julien et al.. 1995) to identify the IlIlpact of rainfall and watershed characteristics data uncertainlY on runoff predictions. The role of land surface characteristic uncenainty in runoff prediction is examined by removing each level of land-surface detail and comparing the result- ing model predictions with a reference hydrograph and the USGS indirect discharge measuremenLS. Simulations driven by single- and dual-polarization radar-rainfall estimates are compared to identify the impact of the WSR-88D rainfall product estimation errors on hydrologic model performance. Hydro- logic simulations driven by rain gage data are analyzed to address the representativeness of rain gage measurements of the rainfall field for hydro- logic modding of flash floods in urban environ- ments. The dataset assembled for this study was documen- ted by Ogden et al. (1999). The data are currently 3, The storm The heavy rains that occurred in Fort Collins on 27-28 July 1997 were preceded by atypically moist atmospheric conditions for July in Colorado because of an unusually strong flow of monsoonal moisture. The mesoscale meteorology of the storm is described in detail by Petersen et a!. (1999). Doesken and McKee (1998) give a detailed description of the rainfall of27- 28 July and the antecedent atmospheric conditions. The first rainfall event started at l700 MDT (moun- tain daylight time) on 27 July. The extreme western portions of Fort Collins received heavy rainfall during this one and one-half hour event. Another heavy rain period started around 0800 MDT, 28 July, during which 20-50 rom of rain fell over western sections of Fort Collins. Clouds persisted all afternoon on 28 July, and the series of storms that caused the worst flooding began fonning before 1800 MDT over the foothills of the Rocky Mountains southwest of Fort Collins. The storm originated as the nearly saturated low-level easterly flow was lifted past the level of free convection by the ridges of the Front Range of the Rocky Mountains, Other factors that may have contributed to triggering this storm include: outflow boundaries from nearby storms, latent and sensible heating gradients, and pressure perturbations that caused an increase in upslope flow (Petersen et aI., 1999). From about 1800-2030 MDT, a series of convective cells moved in a northeasterly direction over Fort Collins. After 2030 MDT the storm became quasi-stationary over the city and the heaviest rain feU between 2030 and 2215 MDT. With little or no nel storm motion,' there were significant changes in the storm intensity and platfonn morphology after 2030 MDT (Landel et aI., 1999). Between 2030 and 2130 MDT the storm orientation changed from perpendicular to parallel to Spring Creek when the highest rain rates were observed. Radar observations reveal that the temporal evolution of small.scale storm structure produced an apparent net storm motion in the downstream direction on Spring Creek. It is unlikely that the net storm motion had an appreciable effect on the runoff hydrograph . FL Ogdell (I ul./ Journal of Hydrology 128 (2000) 82-100 " ---- . SprmgCreek Fig. 2. Relalionship of horse tooth reservoirconlribuling area to lne sprmg creek watershed. with topography (COlltour imen.-a16.! mOil Spring Creek. 12.2 m on the HorsetOOlh catchmem) because the apparent storm speed was considerably faster than the hydraulic wave speed (Ogden et aI., 1995) in Spring Creek. The peak observed 24.hour rainfall was approxi- malely 260 mm and the maximum rain rate at Ihe 5 min time-scale exceeded 128 mmlh, at 2120 MDT. Some of the striking features of this storm include: small net storm motion, absence of hail and infrequent lightning, and low radar echo centroid structure. These characteristics are more typical of tropical convective rainstorms than high-plains summer convectioll. More details on the stoml and its motion can be found in Landel et al. (1999). 4. Rain gage data Data from 14 recording rain gages with hourly or finer temporal resolution over the entire storm period were obtained. Seven of these recording rain gages are located in or within 10 km of the Spring Creek watershed. The five recording rain gages closest to Spring Creek are shown on Fig. 2. Details of precipi- tation data collection and compilation can be found in Doesken and McKee (1998). 5. Flood control and drainage features Spring Creek originates in the foothills of the Rocky Mountains, and flows generally from west to east through Fort Collins, as shown in Fig. 1. Flows in Spring Creek originate from snowmelt, small spills from irrigation canals, groundwater seepage, and local storm water. A complex network of detention and retention basins is used to manage flood hydro- graph volume. Stonnwater management in the Spring Creek watershed depends on the railroad embankment that parallels College Avenue (see Fig..l) to detain storm water in a large detention basin. This detention basin was designed to attenuate the IOO-year flood to prevent downstream damage. The creation of the detention basin required plugging of a 4.3 X 3.7 m~ (l4X 12 ft~) box culvert through the railroad embankment. Spring Creek was re-routed through three 2.1 m diameter culverts through the embank. ment. 6. Watershed data The Spring Creek watershed has an area of25 km~. A 30 m OEM resolution was used 10 describe the land surface features of the Spring' Creek watershed, yet keeping the number of CASC2D computational grid cells to a minimum. A large number of studies have dealt wilh the effect of spatial variability in land- surface parameters on runoff. W oolhiser et al. (1996) showed that Hortonian runoff is strongly inftu4 enced by spatial variability in soil hydraulic conduc- tivity at the hillslope scale. Grayson et al. (1995) found a large influence of stonn characteristics on the effects of random and organized hydrologic condi- tions on catchment runoff. Ogden and Julien (1993), and Saghafian et al. (1995) found that the influence of spatial variability in watershed, soil hydraulic and rainfall characteristics diminishes as the watershed approaches equilibrium runoff conditions, as occurs during extreme events. Merz and Plate (1997) showed that the effects of spatial variability are smull for very small and for large events and large for medium-sized for infiltration excess runoff. The effect of rainfall on the runoff sensitivity to spatial variability, particularly for Hononian runoff, is the main reason why many researchers find that the idea of "representative elementary area" proposed by Wood et al. (1988) has limited utility in catchment hydrology. More discussions on the topic can be found in Fan and