A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 1985

In recent year, Kalman filter techniques have been extensively developed and widely used in many hydrologic forecasting applications. In this study, three useful algorithms of Kalman filter were considered and their forecasting performances were compared. From several runs, it was found that a more complicated and generalized algorithm did not seem to perform better than a simple one. Therefore, a very simple algorithm was adopted throughout the work. The applications considered here involve forecasting of flood flows and forecasting of daily discharge. In the first case, trial-and-error procedures were used to arrive at suitable models which are very simple in its structure. In the second case, an extension of the Linear perturbation (or Hybrid) Model was proposed, where the departure (from the daily mean, or harmonic mean, ) of daily river discharge was linearly regressed upon its past values and past values of the departure of daily rainfall. For all the stations considered, This proposed model produced very good forecasts which were much better than those provided by the original Linear Perturbation Model. It was also found that Kalman filter techniques were not really needed in forecasting daily discharge at these stations.