Streamflow Regime Classification and Annual Discharge Prediction for Ungauged Catchments in British Columbia, Canada Using the Extreme Gradient Boosting Algorithm
by Ziqiang (Kevin) Hu
Abstract
The prediction of the streamflow regime at ungauged locations in a stream network is a fundamental problem in hydrology. The objective of this paper is to assess the Extreme Gradient Boosting (XGB) machine learning algorithm’s ability to classify streamflow regimes in BC, Canada. The created model is robust, with an 88.9% accuracy, and requires only readily available climate and topographic data. However, the ability of the XGB algorithm to predict streamflow magnitude is modest, and it tends to over-estimate for locations with low flow conditions.
Keywords: hydrology, GIS, streamflow regime, ungauged, machine learning, GWR