Nested Threshold Autoregressive (NeTAR) Models for Studying Sources of Nonlinearity in Streamflows

Abstract

Historical development of nonlinear stochastic modeling of streamflows is discussed. Physical considerations and graphical investigations of daily streamflows revealed that air temperature and state of basin storage are the most important sources of nonlinearity in catchments with seasonal snow accumulation. The average temperature for the preceding three days, and the flow one or two days earlier were adequate proxies for the temperature and storage conditions of a catchment. The Nested Threshold Autoregressive (NeTAR) model, which considers these sources of nonlinearity simultaneously, was applied to two years of daily flows of the Oldman River near Brocket in Alberta, Canada. A third year of daily data was used for validating the model. The final NeTAR model provided useful insights into the dynamics of this streamflow system.