FLOOD04478 - Laserfiche WebLink

Home Browse Search

94 300 FL. Ogdtn el al. / }OUffl(J/ 0/ Hydrology 228 (1000) 82-100 . 200 "E . ~ . ~ 6 100 o 18:00:00 ~ - ~ Gages (Thiessen) Gagas(IDS} -f:lelerenceRun , \ \ \ \ \ \\ "~I \\ \\ \\ \\ \'\,,: 21 :00:00 00:00:00 03:00:00 Time, July 28/29, 1997 (MDT) Fig. 8. Simulation resuhs III ~(alion UI using nUn gage input and two rainfall spalial irllerpolalion methods. from the bias-adjusted WSR-88D simulation is 4.5% higher than the reference simulation, further indicat- ing that there are differences between the two rainfall fields although the storm total volume ofrainfall is the same. Calibration of the WSR-88D rainfall pgains! rain gages rather than the polarimetric radar estimates indicated Ih<ll the WSR-88D underestimated the storm tolal rainfall by 50% (Landel et at, 1999). A simulation with a bias adjustment factor of 2.0 resulted in a 16% increase in slOrm total rainfall compared wilh the polarimetric radar estimates. The peak discharge for this simulation was overestimated by 18.5% al station VI. 13, Simulations with rain gage input The spatial description of the rainfall field is poor when Thiessen Polygons are used 10 estimate the spatial distribution of precipitation due to the fact that the entire watershed is essentially covered by only two rain gages, This is equivalent, in a sense, to picking two of the 70 hexagonal radar pixels thaI cover the walershed and assuming that these two pixels represent the entire rainfall field. As expected, the poor gag.e coverage results in differences in the lotal storm rainfit]] and simulated discharge. Simu- lated hydrogr.1phs based on rain gage input are compared with the reference simulation at station VI in Fig. 8. The simulation using rain gage data and Thiessen polygon interpolation overestimates the storm tOlal rainfall volume by 6.5% and the runoff volume by 6.8%. The peak discharge at slation VI was underestimated by 12.5%. This indicates that the existing rain gage network did not adequately capture Ihe space-time structure of the precipitation. Further proof of the inadequacy of the rain gage spatial resolution is the large difference in simulation results when the Inverse Distance Squared (IDS) interpolation method was applied. IDS weighing is often used as an alternative to Kriging when there are insufficient data to compute the rainfall covariance funclion. The storm tOlal rainfall on the watershed is underestimated by 13% using IDS interpolation rela- tive to the reference simulation and 18.5% compared with Thiessen Polygon interpolation. The simulation with IDS weighting underestimates the runoff volume by 17% and the peak discharge by 34.5% at station VI. In the IDS simulation, the hydrologic model predicts railroad embankment overtopping 30 min later than in the reference simulation. 14, Watershed characteristics uncertainty Unlike the Horsetooth watershed, the Spring Creek drainage basin is highly urbanized, as shown on Fig. 6. 300 F.L Ogdtt" <II ",/./ JOllmtll of Hydrology 128 (2000) 82-100 95 200 R~ . ~ ~ is , - //- /1 // ,'I // /- .; ,/,' /? ,'./ .,-/ ,// ,1(/ '1 , 100 Test #1 -- -- Test #2 --- Test 113 --- Tesl#4 -ReterenceRun o 21:00:00 22:30:00 Time,July26129,1997(MDT) 00:00:00 Fig. 9. Simulatiou J:esulls at slation UI using different levels oftand-sulface detait. This portion of the analysis seeks to identify the effect of watershed characteristics uncertainty on the hydro- logic model results. Given thaI there is uncertainty regarding the actual magnitude of land-suriace para- meters, a series of tests are periormed to determine the model sensitivity to watershed characteristics input magnitude and level of detail for this extreme event. The impact of land-use and land cover can be quan- tified by running the model with different levels of detail of land-use information. Four lests with differ- ent land surface descriptions were chosen for compar- ison to the reference simulation, which includes all available land surface details. The literature indicates that land-surface details are less imponant for extreme events (e.g. Saghafian et aI., 1995; Winchel et aI., 1998). These simulations were conducted to evaluate this premise using real data from an extreme event in an urbanized watershed. All four tests used polari- metric radar.rainfall as input. In the first and second tests, impervious areas were omined. Soil hydraulic parameters from the Horse- tooth calibration were used. Soil parameters, overland flow roughness coefficient and retention depth were varied in space as in the reference run using the values listed in Tables I and 2. In the second lest, spatially- uniform retention depth is assumed, calculated as the arithmetic mean of the spatially-varied retention used in test I. Results shown in Fig. 9 indicate that differ- ences between these two simulations are minor. The use of spatially-averaged relention has an effect only early in the stonn. Given the extreme quantity of rain- fall during this storm, rt;;tention is quickly over- whelmed. Comparing tests I and 2 with the reference simulation (Fig. 9) reveals that omission of impervious area reduces the peak dis~harge by 9% relative to the reference simulation. Numerical values for the peak discharge, runoff and infiltration volumes, and runoff efficiencies at station V 1 are given in Table 3. Note that the rainfall volume is identical in each simulation. In the thud and fourth tests, impervious areas are included, but soil hydraulic propenies were taken as the arithmetic mean of the calibrated values and applied unifonnly over the watershed. Spatially varied overland flow roughness was input for both tests 3 and 4. Spatially unifonn overland flow reten- tion was applied in test 3, while retention storage was omitted in test 4. The hyrlrographs shown on Fig. 9 indicate that assuming spatially uniform soil hydraulic characteristics has a significant impact on the predicted hydrograph. The peak discharge in lests 3 and 4 are both approximately 12% lower than the reference simulation. Retention plays a minor role compared to the effect of spatially uniform soil hydraulic characteristics. The omission of retention does not compensate for the errors caused by applying spatially uniform soil properties. The influence of errors in K, on runoff predictions