This paper proposes runoff models based on machine learning to estimate daily streamflows in Oued Sebaou watershed, a Mediterranean coastal basin located in northern Algeria. Therefore, we applied random forest (RF), artificial neural networks (ANN – under different training algorithms), and locally weighted linear regression (LWLR) using as input combinations of current and past rainfall amounts and previous values of streamflow. We selected streamflow and rainfall records to calibrate and validate the stated approaches. We used root mean square error (RMSE) and correlation coefficient (R) to evaluate the accuracy of the models. Analyses of the results show that RF provided the best outcomes for both training (RMSE = 4.7458 and R = 0.9834) and validation (RMSE = 2.3617 and R = 0.9719). The ANN calibrated with the Levenberg-Marquardt algorithm presented the second-best result, outperforming its counterparts and LWLR. Graphical Abstract

Characterized by their high spatiotemporal variability, rainfalls are difficult to predict, especially under climate change. This study proposes a multilayer perceptron (MLP) network optimized by Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Firefly Algorithm (FFA), and Teleconnection Pattern Indices - such as North Atlantic Oscillation (NAO), Southern Oscillations (SOI), Western Mediterranean Oscillation (WeMO), and Mediterranean Oscillation (MO) - to model monthly rainfalls at the Sebaou River basin (Northern Algeria). Afterward, we compared the best-optimized MLP to the application of the Extreme Learning Machine optimized by the Bat algorithm (Bat-ELM). Assessment of the various input combinations revealed that the NAO index was the most influential parameter in improving the modeling accuracy. The results indicated that the MLP-FFA model was superior to MLP-GA and MLP-PSO for the testing phase, presenting RMSE values equal to 33.36, 30.50, and 29.92 mm, respectively. The comparison between the best MLP model and Bat-ELM revealed the high performance of Bat-ELM for rainfall modeling at the Sebaou River basin, with RMSE reducing from 29.92 to 11.89 mm and NSE value from 0.902 to 0.985 during the testing phase. This study shows that incorporating the North Atlantic Oscillation (NAO) as a predictor improved the accuracy of artificial intelligence systems optimized by metaheuristic algorithms, specifically Bat-ELM, for rainfall modeling tasks such as filling in missing data of rainfall time series.

In this paper, the Innovative Trend Methodology (ITM) and their inspired approaches, i.e., Double (D-ITM) and Triple (T-ITM), were combined with Hilbert Huang transform (HHT) time frequency-based method. The new hybrid methods (i.e., ITM-HHT, D-ITM-HHT, and T-ITM-HHT) were proposed and compared to the DWT-based methods in order to recommend the best method. Three total annual rainfall time series from 1920 to 2011 were selected from three hydrological basins in Northern Algeria. The new combined models (ITM-HHT, D-ITM-HHT, and T-ITM-HHT) revealed that the 1950–1975 period has significant wet episodes followed by a long-term drought observed in the western region of Northern Algeria, while Northeastern Algeria presented a wet period since 2001. The proposed approaches successfully detected, in a visible manner, hidden trends presented in the signals, which proves that the removal of some modes of variability from the original rainfall signals can increase the accuracy of the used approaches.