Canadian Water Resources Association (CWRA) 2023: A Machine-Learning Framework for Modeling and Reconstructing Historical Monthly Streamflow Time Series
Date:
Abstract: This study focuses on the challenge of predicting monthly streamflow in data-scarce environments, where recurring data gaps can lead to unfavorable outcomes such as loss of critical information, ineffective model calibration, inaccurate timing of peak flows, and biased statistical analysis in various applications. To overcome this challenge, an ensemble machine-learning regression framework is introduced, which utilizes historical data from multiple monthly streamflow datasets in the same region to predict missing monthly streamflow data. The framework selects the best features from all available gap-free monthly streamflow time-series combinations and identifies the optimal model from a pool of 12 machine-learning models. The study shows that the gradient boosting regressor with bagging regressor produced the highest accuracy in 7 of the 26 instances, and the models using this method exhibited an overall accuracy range of 0.9737 to 0.9968. The framework was employed to accurately extend the missing data for all 26 stations, including two crucial stations located in the economically significant lower Athabasca Basin River in Alberta province, Canada, which had approximately 70% of their monthly streamflow data missing. The accurate extensions allow for further analysis, including grouping stations with similar monthly streamflow behavior using Pearson correlation.
Conference page: https://conference.cwra.org/