HR: 0830h
AN: H51D-03    [Abstracts]
TI: An Integrated Multiscale Approach to River Flood Prediction Using a Land-Surface Hydrology Model
AU: * Mackey, B P
EM: mackey@met.fsu.edu
AF: Department of Meteorology, FSU, Love Building Florida State University, Tallahassee, FL 32306-4520 United States
AU: Barros, A P
EM: ana.barros@duke.edu
AF: Department of Civil and Environmental Engineering Duke University, Box 90287 Hudson Hall Duke University, Durham, NC 27708-0287 United States
AU: Krishnamurti, T N
EM: tnk@io.met.fsu.edu
AF: Department of Meteorology, FSU, Love Building Florida State University, Tallahassee, FL 32306-4520 United States
AB: This work outlines and demonstrates a comprehensive hydrometeorological flood forecasting system that is interdisciplinary in nature, multiscale in its approach, and state-of-the-art in its use of forecasting techniques. This integrated approach links both meteorological and hydrological tools in order to realize a more accurate prediction of major flood events three to six days in advance. One focus is on a new non-linear method to improve global multi-model superensemble precipitation forecasts with an emphasis on successful prediction of intense rain areas. On average, the skill from such a technique is higher and the bias lower than any of the individual member models, and the overall character of the precipitation distribution is maintained through the 5-day forecast period. In addition, global and nested regional spectral models are integrated in hindcast mode. Output from such models as well as from the superensemble is used as forcing input to a physically-based, spatially-distributed hydrology model in order to predict streamflow response during selected flood events. In this terrestrial hydrology prediction system, a dynamical water routing scheme and other physical parameterizations are used in conjunction with a 3-D hydrologic model that keeps track of surface water and energy budgets, including surface-subsurface interactions, groundwater and hydraulic river routing. Experiments are run for the Limpopo River basin in southeastern Africa, where massive flooding occurred in both years 2000 and 2001. An important intermediate step in this process is the downscaling of the precipitation forecasts from the course resolution atmospheric models to the much finer resolution (1 to 10 km) required by the hydrology model. For this purpose, we examined the space-time scaling behavior of simulated precipitation fields from the NWP models at different resolutions and devised a simple physically-based multifractal downscaling algorithm that relies on the scaling consistency of multiple resolution forecasts. The physical constraints are imposed at multiple scales based on indices of storm dynamics and, where available, the climatology of storm structure inferred from of high-resolution merged multisensor precipitation data (raingauge, radar, and satellite data). The approach is demonstrated for real-time forecasting over the Oklahoma Mesonet and adjacent land areas during the summer and fall of 2003.
DE: 1821 Floods
DE: 1854 Precipitation (3354)
DE: 1860 Runoff and streamflow
DE: 3210 Modeling
DE: 3250 Fractals and multifractals
SC: Hydrology [H]
MN: 2005 Joint Assembly