Short-term River streamflow modeling using Ensemble-based additive learner approach

Accurate streamflow (Qt) prediction can provide critical information for urban hydrological management strategies such as flood mitigation, long-term water resources management, land use planning and agricultural irrigation operations. Since the mid-20th century, Artificial Intelligence (AI) models have been used in a wide range of engineering scientific fields, their application has increased last few years. In this study, predictive capabilities reduced error pruning tree (REPT) model, both standalone model within five ensemble-approaches, were evaluated to predict Kurkursar basin Iran. The ensemble-approaches combined REPT with bootstrap aggregation (BA), random committee (RC), subspace (RS), additive regression (AR) disjoint aggregating (DA) (i.e. BA-REPT, RC-REPT, RS-REPT, AR-REPT DA-REPT). developed using 15 years daily rainfall data period 23 September 1997 22 2012. A set eight different input scenarios was constructed combinations variables find most effective scenario based on linear correlation coefficient. comprehensive suite graphical (time-variation graph, scatter-plot, violin plot Taylor diagram) quantitative metrics (root mean square (RMSE), absolute (MAE), Nash-Sutcliff efficiency (NSE), Percent BIAS (PBIAS) ratio RMSE standard deviation observation (RSR)) applied evaluate accuracy six developed. outcomes indicated that all performed well but outperformed other by rendering lower errors higher precision across number statistical measures. BA, RC, RS, AR DA enhanced performance about 26.82%, 18.91%, 7.69%, 28.99% 28.05% respectively.