2019 CSCE Annual Conference - Laval (Greater Montreal) Conference
Dr. Mohamed Meguid, McGill University
Dr. Fateh Chebana, Eau Terre Environnement, Institut National de la Recherche Scientifique
Dr. Mohammed Elenany, Urban Systems Ltd., Canada
Dr. Bahaa Khalil, Helwan University, Egypt
River flow modeling and forecasting studies have been dominated by statistical and heuristic based techniques. Such Techniques are shown to successfully produce useful inferences in absence of information on the variables that drive the evolution of the flow, of which a useful subset is typically utilized for physically inspired forecasting models. Like any time series generated by complex systems, river flow evolution can be represented by a time-varying parameter (TVP) model. TVP modeling frameworks often assume that the system evolution exhibits superstatistical random walks, and a statistical cohort of other assumptions is followed upon, to infer better forecasts about the system. Also, the TVPs should hone a predefined model ability to capture systems recurrence, which translate to a number of state-based model structures in this case. To the best of our knowledge, there has been no research effort made for assessing the possibility of modeling river flows on extended horizons and more continuously. In this work, we develop a computationally efficient method that extracts useful cyclic-type TVPs that are strongly coupled over extended, but more continuous and less aggregated, time horizons. A multi-level modeling framework is suggested to facilitate the creation and direct use of updated parameter states to produce reliable models on the considered temporal resolutions. We also show that the modeling framework produce stable results of the parameter variations at different time strings. The proposed framework have can incorporate exogenous descriptors at different levels without loss of generality, and can be used in forecasting applications.
Keywords: River flow forecasting; varying parameter framework; system identification; clustering