Regional hydrological frequency analysis at ungauged sites with random forest regression

Abstract Flood quantile estimation at sites with little or no data is important for the adequate planning and management of water resources. Regional Hydrological Frequency Analysis (RFA) deals hydrological variables ungauged sites. Random Forest (RF) an ensemble learning technique which uses multiple Classification Regression Trees (CART) classification, regression, other tasks. The RF gaining popularity in a number fields because its powerful non-linear non-parametric nature. In present study, we investigate use (RFR) step RFA based on case study represented by collected from 151 hydrometric stations province Quebec, Canada. RFR applied to whole set homogeneous regions delineated canonical correlation analysis (CCA). Using Out-of-bag error rate feature RF, optimal trees dataset calculated. results application CCA model (CCA-RFR) are compared obtained linear models. CCA-RFR leads best performance terms root mean squared error. delineate neighborhoods improves considerably RFR. found be simple apply more efficient than complex models such as Artificial Neural Network-based