Extended Abstract (192K)9A.5
A test of two distributed hydrologic models with WSR-88D radar precipitation data input in Arizona
Steven Hunter, U.S. Bureau of Reclamation, Denver, CO; and B. Vieux, F. Ogden, J. Niedzialek, C. Downer, J. Addiego, and J. Daraio
The U.S. Bureau of Reclamation's water management officials along the Lower Colorado River wish to accurately determine inflows to the river from numerous sidewash basins. These inflows are problematic to reservoir operations in the event of a heavy widespread rainfall into those basins. Distributed-parameter hydrologic models with radar quantitative precipitation estimate (QPE) input allow the prediction of streamflow at various points in a watershed. Only until recently, however, have advances in computing power made such a system possible for use as near-real-time decision assistance to water operations management. To quantitatively assess this assertion, the distributed hydrologic models GSSHA (a refinement of CASC2D) and Vflo™ were tested in the case of a heavy rainstorm over west-central Arizona. The heavy rain was produced by Tropical Storm Nora, 25-26 September 1997. The test area is the Santa Maria, an unregulated headwater basin flowing from 2100 m elevations west of Prescott toward the Bill Williams River in the lowland desert of western Arizona. The Bill Williams discharges into the mainstem of the Lower Colorado River near Lake Havasu City. The area of the Santa Maria basin is 3,727 square kilometers.
Radar reflectivity data from the operational Flagstaff, Arizona WSR-88D Doppler radar were input to the two hydrologic models. This radar is located about 150 km from the Santa Maria headwaters. The QPEs were derived from Reclamation's new Precipitation Accumulation Algorithm (PAA), which used a reflectivity-rain rate (Z-R) relationship and a radar/gauge bias for the 25-26 September event. Of principal interest are not only streamflow output accuracy but also the time required to run model simulations, since that is crucial to determine feasibility for near-real-time decision assistance. Both models were tested with identical GIS and QPE data input and similar computer processors, to ensure a fair test. Model simulations of the Santa Maria case reveal that for 500 m grid resolution, run time for Vflo™ was about a half minute and for GSSHA about one and a half minutes. Extrapolation of these run times to multiple sidewash basins indicates that both models run quickly enough for timely decision assistance to water management.
Some model calibration was necessary to reproduce the bimodal, fast-responding hydrograph for 26-27 September at the stream gauge near the Santa Maria basin outlet. In particular, channel widths upstream of this gauge were seen as critical input to the simulations. These channel widths were calculated for GSSHA from aerial photographs available online and were instrumental in the resulting accurate simulation of the gauge hydrograph.
These efforts are part of the Agricultural Water Resources Decision Support (AWARDS) system, which has been supplying web-based radar QPEs to water managers and users in the Lower Colorado region. Reclamation plans to integrate AWARDS into Riverware, an object-oriented generalized river basin modeling environment that integrates the multiple purposes of reservoir systems. Riverware is already in use by Reclamation's water operations managers along the Lower Colorado.
Supplementary URL: http://www.usbr.gov/pmts/rivers/awards/publ/radconf_2003_paper_full.pdf
Session 9A, Hydrology
Sunday, 10 August 2003, 4:00 PM-6:00 PM
Previous paper Next paper
Recorded presentation