Prediction of Manning's coefficient of roughness for high-gradient streams using M5P

The coefficient of Manning's roughness (n) has been generally implemented in the determination of depth and discharge in open channels and canals. This study unravels the novel idea and potential of Random Forest (RF), M5P, and Random Tree (RT) approaches to evaluate and predict the coefficient of Manning's roughness for hydraulic designing. To achieve this purpose, 42 observations were collected for high-gradient streams in Colorado, USA. All the observations were from boulder-bed, cobble and high gradient (S > 0.002 m/m) streams within bank flows. In order to ascertain the best model, the above-mentioned approaches were evaluated and compared using performance evaluation indices such as mean absolute error (MAE), coefficient of correlation (CC), and root mean square error (RMSE). Outcomes of performance evaluation indices revealed that the proposed pruned M5P approach outperformed other applied models for predicting the coefficient of Manning's roughness for hydraulic designing with CC = 0.7858, 0.7910, RMSE = 0.0195, 0.0195, and MAE = 0.0157, 0.0165 for model development and validation period, correspondingly. Furthermore, Taylor diagram and Box plot also suggest that the M5P based approach works better than RF and RT based approaches for predicting the coefficient of Manning's roughness for high-gradient streams using the given data set. HIGHLIGHTS Three soft computing-based modelling approaches (M5P, RF and RT) were developed in the prediction of Manning's roughness coefficient.; The performance of modelling approaches was compared by mean absolute error (MAE), coefficient of correlation (CC), and root mean square error (RMSE).; The total dataset was divided into training and testing subset in the ratio of 70:30 to perform the modelling approaches.; M5P modelling approach is the best approach in the prediction of the Manning's roughness coefficient.;