Intercomparison and Assessment of Stand-Alone and Wavelet-Coupled Machine Learning Models for Simulating Rainfall-Runoff Process in Four Basins of Pothohar Region, Pakistan

The science of hydrological modeling has continuously evolved under the influence rapid advancements in software and hardware technologies. Starting from simple rational formulae for estimating peak discharge developing into sophisticated univariate predictive models, accurate conversion rainfall runoff assessment inherent uncertainty been a prime focus researchers. Therefore, alternative data-driven methods have gained widespread attention hydrology. Moreover, scientists often couple conventional machine learning models with data pre-processing techniques, i.e., wavelet transformation (WT), to enhance modelling accuracy. In this context, research work attempts explore latent linkage between Pothohar region Pakistan by novel five streamline techniques learning, including single decision tree (SDT), forest (DTF), boost (TB), multilayer perceptron (MLP), gene expression (GEP), more variant WT, maximal overlap discrete (MODWT), boundary correction transformed components timeseries data. This study also implements these stand-alone mode comprehensive comparative analysis performances. Furthermore, uses combined-basin approach that divides two basins compensate complex topographic division area. results indicate MODWT-based DTF outperformed other hybrid terms first scenario, considering Bunha-Kahan River basin, MODWT-DTF yielded highest NSE (0.86) lowest RMSE (220.45 mm) R2 (0.92 at lag order 3 (Lo3)) when daubechies4 (db4) level three. While Soan-Haro produced accuracy 4 (Lo4) (NSE = 0.88, 21.72 m3/s, 0.91). highly performance 3- 4-days lagged reflects temporal consistency response comparison model indicates up 55% increase due transformation.