A new approach for flood forecasting of river flows

Abstract

Flood warning mainly depends on reliable flood forecast models. Literature is rich in flood modelling techniques, but failures of these models, especially on the very short scale such as hourly flows, do often cause devastating impacts on the communities affected by these floods, and on many occasions result in loss of lives. This paper presents a new approach for flood forecasting of river flows based on the projection theorem in Hilbert space. The new modelling process obtains the projection of hourly flow rates on hourly rainfalls over the catchment at previous hours to the projected flow rate. A total of 25 flow events observed for the Leith River in Dunedin, New Zealand, along with their corresponding observed rainfalls at two sites in the catchment have been identified and applied to calibrate and validate the derived model. The proposed modelling technique was capable of simulating the flow process for the Leith River, and is a promising tool for flood forecast when other models fail. The proposed model is easy to apply, doesn’t imply a lot of assumptions or parameters, as other models usually require, and can be used for long term forecast based on forecasted hourly rain one day or more before the event, or real time forecast during the event itself based on rainfall which has been already gauged. However, for real time (short term) forecast, the forecast time can be a few hours based on the catchment area and its topography which can lead to a fast flow to the outlet.