Advances in Transfer Function Based Flood Forecasting

Abstract

Simple transfer function (TF) models were first used for operational real-time flood forecasting in the 1970s after a number of researchers had shown that they could efficiently characterise dominant rainfall-runoff dynamics. A major advantage of the approach at that time was the small requirement for computing power in comparison with more complex conceptual models. Although computing power is no longer a particularly advantageous feature, recent advances in transfer function modelling and forecasting techniques provide the ability to produce state-of-the-art flood forecasts whilst retaining the advantage of simplicity. This paper challenges some popularly held views on transfer function forecasting techniques, such as “a major disadvantage of TF modelling compared to conceptual models is that it is a black-box approach with no physical process explanation”, and presents a number of recent advances in non-linear transfer function identification, parameter estimation and real-time implementation, concluding with an illustrative example.