Hydrological Modelling of River Flow Forecasting in Cold Regions and Its Application over the Athabasca River

• Main Author: Belvederesi, Chiara
• Other Authors: Hassan, Quazi, Achari, Gopal, Rangelova, Elena, Gupta, Anil
• Format: Doctoral or Postdoctoral Thesis
• Language: English
• Published: Schulich School of Engineering 2023
• Subjects: ungauged; cold hydrology review; empirical; Athabasca River Basin; Engineering; Engineering > Environmental; Athabasca River
• Online Access: http://hdl.handle.net/1880/115958; https://doi.org/10.11575/PRISM/40807

Floods are disasters that represent a growing threat to the communities living close to rivers. To maximize community resilience, the main objective was to formulate a transferable framework for river flow forecasting in cold and poorly gauged/ungauged regions. First, the literature was reviewed, summarizing the recent findings in river flow forecasting in these regions. Here, hydrological processes greatly vary seasonally and annually, translating into increased model uncertainty. Regionalization, spatial calibration, and other methods were implemented into process-based and empirical models. Although process-based models provided a wide understanding of a watershed’s hydrology, they were often complex and computationally demanding. Empirical models produced fewer calibration parameters although generated biased results when insufficient descriptors were available. The results from this review highlighted some efforts necessary to improve river flow forecasting, including: coping with limited data; providing user-friendly interfaces; advancing model structure; developing a universal method for transferring parameters; standardizing calibration and validation; integrating process-based and empirical models. In addition, a machine learning-based model was developed using a single-input sequential adaptive neuro-fuzzy inference system (ANFIS) in the Athabasca River Basin (ARB) in Alberta, Canada. After estimating the optimal lead time between four hydrometric stations, data measured near the source were used to compute flows near the mouth, over approximately 1,000 km. This technique was compared to nonsequential and multi-input ANFIS, which used data from all the four hydrometric stations. The results showed that sequential ANFIS could accurately predict flows (r2 = 0.99, Nash–Sutcliffe = 0.98) with a longer lead time (6 days) using a single input. Finally, a novel simplistic method for short-term (6 days) forecasting called Flow Difference Model (FDM) was developed and compared against existing hydrological ...