HR: 0800h
AN: H11C-0646    [Abstracts]
TI: Using oceanic-atmospheric oscillations for long lead-time streamflow forecasting in the Upper Colorado River Basin
AU: * Kalra, A
EM: akalra@egr.unlv.edu
AF: University of Nevada, Las Vegas, Dept. of Civil and Env. Engg. 4505 Maryland Parkway, Las Vegas, NV 89154,
AU: Ahmad, S
EM: Sajjad.Ahmad@unlv.edu
AF: University of Nevada, Las Vegas, Dept. of Civil and Env. Engg. 4505 Maryland Parkway, Las Vegas, NV 89154,
AB: In the recent past, oceanic-atmospheric oscillations have been used successfully for long lead-time streamflow forecasting. Herein, we present a data-driven model, Support Vector Machine (SVM) for the long lead-time streamflow forecast incorporating oceanic-atmospheric oscillations. The SVM is based on Statistical Learning Theory that uses a hypothesis space of linear functions based on Kernel approach and can be used to predict a quantity forward in time based on training that uses past data. The principal strength of SVM lies in minimizing the empirical classification error and maximizing the geometric margin by solving inverse problems. The SVMs are considered superior to the Artificial Neural Networks (ANNs) due to the tendency of formulating a quadratic optimization problem which ensures a global optimum that is found missing in the traditional ANN approach. The SVM model was applied to four unimpaired gages in the Upper Colorado River Basin (UCRB). The streamflow data for the selected gages was used from 1906¡§C2004. Annual oceanic-atmospheric indexes comprising of Pacific Decadal Oscillation (PDO), North Atlantic Oscillation (NAO), Atlantic Multidecadal Oscillation (AMO), and El Nino-Southern Oscillations (ENSO) for a period of 1906¡§C2001 were used to generate streamflow volumes for three years ahead. The SVM model was trained with 86 years of data (1906¡§C1991) and tested for 10 years of data (1992-2001). The testing criteria used for the model effectiveness was based on correlation coefficient r, root means square error (RMSE) and nash sutcliffe efficiency coefficient e. Predictions during the testing phase showed a good agreement with measured streamflow volumes for the selected gages in UCRB. Rigorous sensitivity analysis was performed to evaluate the effect of individual oscillation. The results indicated a strong signal for NAO and ENSO indexes as compared to PDO and AMO indexes for the long lead-time streamflow forecast. The oceanic-atmospheric oscillations are helpful in providing long range streamflow predictions which can be potentially used for planning and management of water resources for the UCRB.
DE: 1616 Climate variability (1635, 3305, 3309, 4215, 4513)
DE: 1719 Hydrology
DE: 1860 Streamflow
DE: 3305 Climate change and variability (1616, 1635, 3309, 4215, 4513)
DE: 4215 Climate and interannual variability (1616, 1635, 3305, 3309, 4513)
SC: Hydrology [H]
MN: 2007 Fall Meeting