FLOOD04478 - Laserfiche WebLink

Home Browse Search

FL Ogdell el al./ JowrMl of Hydrology 228 (2000) 82-]00 87 120 0 100 E S 80 , .. " 'E 0 '. a: 80 ~ 0 - , 0 0 J: " 40 ~ . a: 20 + + + + 0 100 120 0 20 40 80 80 Gage Hourly Rainfall (mm) 86 FL Ogdell el a/./ JOWrMl of Hydrology 228 (2000) 82-100 Bras (1995); BJoschl et al (1995), and Seyfried and Wilcox (1995). Horsetooth Reservoir, shown in Fig. 2. provides irrigation storage and serves as the water supply for the City of Fort Collins. II is a permanent water storage facility that receives most of its supply from trans-basins transfers from the upper Colorado river via the Colorado-Big Thompson (CSn diversion project. Horsetooth lake is formed by four dams on a ridge in the foothills of the Rocky Mountains. The sUIface area of the lake on 28 July 1997 was 7.4lcm2, with a contributing area of 37.4 km~. This mountai- nous watershed contains six different soil types, five of which are common to the Spring Creek watershed. As shown in Fig. 2. the Horsetooth watershed is adja- cent to the Spring Cre~k and was affected by the same stonns as Spring Creek on 27-28 July 1997, Interest- ingly, when the Horsetooth Reservoir was constructed in the early 19505, the Spring Creek watershed was hydraulically cut in half. Soil texture maps for both Spring Creek and Horse- tooth watersheds were obtained and digitized. Flood plain maps at I :100 scale. developed by others as pan of the Spring Creek master drainage plan, were used to develop channel cross-sections and 10 obtain details of the major detention basins. Land-use data reveal that 23% of the Spring Creek watershed area is covered by impervious surfaces. The location of the only stage-recording station on the main channel of the Spring Creek was considered the watershed outlet. UnfoI1unately, this slage recorder was damaged by floodwaters, and no data were collected after 22:30 MDT on 28 July 1997. 7, Radar-rainfaU The spatial and temporal resolution of rainfall dala has a great impacl on the accuracy of physics-based distributed hydrologic models as verified by many researchers. Woolhiser (1996) adeptly stated that expecting a physically based hydrologic model to correctly predict storm runoff using erroneous rainfall data is akin to expecting a beam flexure equation to predict the correct beam displacement under an erro- neous loading. The magnitude of runoff prediction errors resulting from rainfall spatial and temporal resolution errors depends on various factors due to the complexity of the flood producing system, i.e. the interaction of the watershed topography and geomorphology, the runoff production mechanism, the ratio of the rainfall rate to the saturated hydraulic conductivity of the soil (Saghafian et al., 1995) and the spatial_and temporal variability of the rainfall field. This wide range of factors might be the reason for varied conclusions among researchers who sfudjed the impacts of spatial and temporal variability on runoff. Although some researchers have downplayed lhe importance of averaging in certain situations, many researchers have shown the opposite in other studies. Ogden and Julien (1994) identified twO diSlinct causes of error due to precipitation spatial aggregation, "storm smearing" and "watershed smear- ing". Stann smearing occurs because spatial aggrega- tion tends to reduce areal-average rainfall rates. Watershed smearing causes uncertainty in the location of precipitation over the catchment due to aggrega- tion. Winchel et al. (1998) showed that spatial and temporal averaging of rainfall affects modeling results more for Hortonian runoff production than for satura- tion excess runoff production. Radars and rain gages have very different sampling characteristics in both space and time. As an example, it is possible that instantaneous radar measurements taken every 30 min can produce more erroneous results than lhe use of hourly integrated averages of rainfall rates (Sharif and Ogden, 1998). Only four recording rain gages cover lhe Spring Creek watershed area when using Thiessen poly- gons to determine the area of influence of each rain gage, Two of lhese four gages cover 90% of the watershed. Furthermore, extremely high rainfall intensity gradients were observed on 28 July 1997. Given the likely inadequacy of the rain gage network, investigation of the accuracy of radar-rainfall estimates was necessary for this study (Pessoa et al.. 1993). Both available single- and dual- polarization rainfall rate estimates were verified using rain gage measurements. Fi;. 3. Radar-raiofall estimates vs. rain galle observations for uncalibJaled polarimetric and WSR-88D radu.rainfall algorithms. radar located near Cheyenne, WY, which is 85 km north-northeast of Fort Collins. The WSR-88D is a single-polarization radar which observes the reflec- tivity factor, Z, which is theoretically given by: Z= V-I iD~ 1'=1 where V is the volume of the sample element (cubic meters), 11 is the number of raindrops in the sample element, and Dj is the diameter (in millimeters) of the jth drop. The reflectivity factor is converted to rainfall rate R through an empirical R-Z power function of the form: R=aZb The R-Z parameters used are a = 0.017 and b = 0.714 for R in mmlh and Z in mmt-/m3. These para- meters correspond to the relation Z = 300R1.4 (Batlan, 1973) which was in use at the Cheyenne. Wyoming, WSR-88D site on 28 July 1997. WSR-88D openlfors have the option of using a tropical rainfall l-R rela- tion (Fulton et a1. 1998), however, this relation was not in use during !he Fort Collins stonn. The operational WSR-88D rainfall products (Smith et al., 1996) underestimated the total rainfall for this storm by a factor of 2.0 compared to rain gage accumulations, Hourly WSR-88D radar-rainfall esti- mates compared with hourly measurements at the 7.1. WSR-88D raillfa" estimates The precipitation field was estimated from the three-dimensional volume scan reflectivity fields observed by the S-band US National Weather Service WSR-88D (Weather Surveillance Radar, 1988-Doppler) (I) seven rain gages nearest to Spring Creek are shown on Fig. 3. WSR-g8D underestimation of rainfall can be attributed to the distinctive doud microphysics of extreme storms, e.g. anomalous drop size distribution, the vertical structure of the storm (Smith et aL. 1996), abnonnal propagation of the radar beam (ground clutter) and radar calibration error. Petersen et al. (1999) present a detailed discussion of radar-rainfall estimation errOrs for the Fort Collins storm. 7.2. Polarimetric radar-rainfall estimates (2) The polarimetric weather radar observations used in this study were recorded by the CSU-CHILL research radar facility located near Greeley, Colorado, which is approximately 35 Ian east-southeast of Fort CoHins. The CSU-CHILL S-band radar is dual, line- arly polarized with one-degree beam width. The radar measures the reflectivity in twO orthogonal (horizontal and vertical) polarizations, ZIl and lv. respectively. When the two reflectivities are measured in an approximately simultaneous fashion, the differential reflectivity (in decibels) can be derived by lOR = 10 10glO(ZH/Zv), Differential reflectivity ZOR provides an indicator of raindrop oblateness and thus raindrop size (Seliga and Bringi, 1976). For an assumed rain- drop size distribution (RSD). lOR provides an estimate of one of RSD parameters, and when combined with ZH results in a more accurate estimate of rainfall rate