Stream flow Predictions using Support Vector Machines
Location
Space Dynamics Laboratory
Event Website
http://water.usu.edu/
Start Date
3-26-2004 2:30 PM
End Date
3-26-2004 2:45 PM
Description
Effective lead-time stream flow forecast is one of the key aspects of successful water resources management in arid regions. In this research we present new data-driven models known as Support Vector Machines (SVMs) that were used to forecast seasonal stream flow volumes. The SVM methodology is based on Statistical Learning Theory (SLT), which was developed to solve linear and non-linear inverse problems. SLT theory starts from stabilizing the solution of inverse problems by introducing a penalty function to account for the complexity of a solution. As a result, the generalized risk function has two components, one related to the goodness-of-fit, and the other to the mapping function complexity. The methodology makes use of uniquely solvable quadratic optimization problem that minimizes the bound on generalized risk, rather than just the mean square error of differences between measured and “predicted” values. Empirical results from these models showed a promising performance in solving site-specific, “real-time” water resources management problems. Stream flow was forecasted using local-climatological data and requiring far less input than physical models. In addition, predictions were improved by incorporating atmospheric circulation indicators. Specifically, use of the North-Pacific Sea Surface Temperature Anomalies (SSTA) improved flow volume predictions.
Stream flow Predictions using Support Vector Machines
Space Dynamics Laboratory
Effective lead-time stream flow forecast is one of the key aspects of successful water resources management in arid regions. In this research we present new data-driven models known as Support Vector Machines (SVMs) that were used to forecast seasonal stream flow volumes. The SVM methodology is based on Statistical Learning Theory (SLT), which was developed to solve linear and non-linear inverse problems. SLT theory starts from stabilizing the solution of inverse problems by introducing a penalty function to account for the complexity of a solution. As a result, the generalized risk function has two components, one related to the goodness-of-fit, and the other to the mapping function complexity. The methodology makes use of uniquely solvable quadratic optimization problem that minimizes the bound on generalized risk, rather than just the mean square error of differences between measured and “predicted” values. Empirical results from these models showed a promising performance in solving site-specific, “real-time” water resources management problems. Stream flow was forecasted using local-climatological data and requiring far less input than physical models. In addition, predictions were improved by incorporating atmospheric circulation indicators. Specifically, use of the North-Pacific Sea Surface Temperature Anomalies (SSTA) improved flow volume predictions.
https://digitalcommons.usu.edu/runoff/2004/AllAbstracts/14