Peak flow prediction using fuzzy linear regression: Case study of the Bow River


  • Usman T Khan University of Victoria
  • Caterina Valeo Mechanical Engineering, University of Victoria, BC, Canada


The 2013 floods in Alberta highlighted the need for better flood prediction. Though the mechanisms behind floods and extreme events in urban areas are understood and documented, the uncertainty in data during these events makes it difficult to accurately predict and assess the risk of floods. In this research, a fuzzy number based linear regression model is proposed that incorporates uncertainty and characterizes risk of extreme events in the Bow River at Calgary, Alberta, Canada. The proposed model uses a fuzzy linear regression model to predict peak flow rate using mean daily flow rate. Lagged data from one to seven days is also considered. Results of the research show that using a fuzzy number approach to predict uncertain extreme events outperforms traditional regression methods in the Bow River at Calgary. The developed model can accurately predict daily peak flow, including a flood event in 2005, up to 7 days in advance. In addition to this, fuzzy number model output can be used to further characterize the risk of peak flow magnitude. These results are extremely beneficial for water resource managers who implement flood mitigation and defence strategies.

Author Biography

Usman T Khan, University of Victoria

PhD Candidate Department of Mechanical Engineering






Environmental Modeling, Risk Assessment and Decision Making (EMR)