EVALUATION OF THE KALMAN FILTER FOR THE PREDICTION AND ESTIMATION OF DAILY STREAM FLOWS
Abstract
The problem considered in this dissertation is the modeling of hydrologic time series, in particular, stream flow data, by the signal plus noise model. It is assumed that the signal can be described by a linear difference equation. The reasonableness of this approach becomes clear when one acknowledges the fact that hydrologic data are heavily affected by noise. It should be recognized that thus far, very little attention has been given to this aspect of hydrologic modeling. To account explicitly for the observation noise, one possibility is to express the linear stochastic difference equation in state variable form and to introduce an observation model. The discrete Kalman filter algorithm may then be used to obtain estimates of the state variable vector. Typically, in hydrologic systems, transition matrix, system noise statistics and measurement noise statistics are unknown. Thus, these quantities are simultaneously estimated with the state variables. The forecasting performance of the conventional lumped error model is compared with that of the signal plus noise model. In the study, simulated data as well as real data are used.
Degree
Ph.D.
Subject Area
Mechanical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.