loading page

Understanding the Dynamic Nature of Catchment Response Time through Machine Learning Analysis
  • +1
  • Mistaya Langridge,
  • Bahram Gharabaghi,
  • Hossein Bonakdari,
  • Rachel Walton
Mistaya Langridge
University of Guelph

Corresponding Author:mlangridge@hotmail.ca

Author Profile
Bahram Gharabaghi
University of Guelph
Author Profile
Hossein Bonakdari
University of Guelph
Author Profile
Rachel Walton
University of Guelph
Author Profile

Abstract

Understanding the hydrologic response to rainfall events is vital for flood forecasting and design for peak flows. The Time to Peak (Tp) is used to characterize the speed of catchment response, as the time from the start of a rainfall event to the time the peak flow is reached in a stream. Advancing our understanding of a catchment’s temporal response to rainfall is key to our overall understanding of hydrologic processes. In this study, more than 1400 storm hydrographs were isolated and utilized to calculate the Tp value for decades of storms spanning Great Britain. Previous works into understanding Tp have been static, with no variability due to storm magnitude or antecedent conditions, providing a single static value for each catchment. Using this data and machine learning techniques, dynamic Tp values were predicted for each storm within the hundreds of catchments, to allow for fuller understanding of the catchment response. Artificial Neural Networks are utilized in this study to create models which account for antecedent conditions of the catchment, and the storm size, to predict the storm-specific, dynamic Tp value.