Prediction of Daily Streamflow Data Using Ensemble Learning Models

Abstract

Estimating river streamflow is a key task for both flood protection and optimal water resource management. The high degree of uncertainty regarding watershed characteristics, hydrological processes, and climatic factors affecting river flows makes streamflow estimation a challenging problem. These reasons, combined with the increasing prevalence of data on streamflow and precipitation, often lead to data-driven models being preferred over physically-based or conceptual forecasting models. The goal of this study is to predict daily river streamflow data with high accuracy using bagging and boosting approaches, which are ensemble learning methods. In addition, the effect of tributary streamflow on the forecast performance was analyzed in the estimation of the streamflow data. According to the results obtained, it has been shown that ensemble learning models are successful in estimating daily streamflow data, and if the tributary streamflow data is also used as input in the estimation of the streamflow, the determination and correlation performance parameters are improved, and the streamflow data can be estimated using tributary streamflow data.