Introducing the variation of advective time delay (ATD) to Transfer Function models

In this paper, a variable advective time delay (ATD) is introduced to Transfer Function (TF) models. This element enables the TF model to follow the diversity in response of a catchment system much better than using a constant advective time delay value as heretofore. An investigation of different approaches to define the variation of the ATD of the catchment studied is presented here. The resulting variable ATD is dependent on the variations of the flow or rainfall, depending on which is used as input to the applied model in order to produce the flow forecast. The variable ATD is tested on both linear and nonlinear TF models; the latter being TF models with state dependent parameters. Both types have been widely used in real-time flow forecasting applications. The proposed methodology is assessed using rainfall and flow time series from the Pinios River in Greece.