Neural Network Solutions to Flood Estimation at Ungauged Sites

Abstract

AbstractArtificial neural networks are universal approximators that have seen widespread application in a number of fields in science, engineering, medicine and finance. Since their re-popularisation in the mid-1980s, they have been applied successfully to a number of problems in hydrology. However, despite their widespread use, these tools have only been applied in a limited number of studies to the problem of estimating flood magnitudes in ungauged catchments. Using data from the Centre for Ecology and Hydrology’s Flood Estimation Handbook, this chapter aims to show how neural network models can be developed to predict 20-year flood events and the index flood (the median of the annual maximum series) for ungauged catchments across the UK. In addition, the chapter provides a discussion of how different choices in the use of the available data can significantly affect the accuracy of the models that are developed.KeywordsArtificial neural networksflood estimationungauged catchmentsFlood Estimation Handbook