Assessment of the hydrological and coupled soft computing models, based on different satellite precipitation datasets, to simulate streamflow and sediment load in a mountainous catchment. Issue 2 (3rd February 2023)

• Record Type: Journal Article
• Authors: Khan, Muhammad Adnan; Stamm, Jürgen
• Is Part Of: Journal of water and climate change. Volume 14:Issue 2(2023)
• Page Start: 610
• Page End: 632
• Publication Date: 2023-02-03
• Subjects: artificial neural networks -- random forest -- satellite precipitation products -- support vector regression -- SWAT; Water -- Periodicals; Hydrology -- Periodicals; Climatic changes -- Periodicals; Climatic changes; Hydrology; Water; Electronic journals; Periodicals; 333.9116
• DOI: 10.2166/wcc.2023.470
• Languages: English
• ISSNs: 2040-2244

Abstract

Abstract: This study evaluated the performance and hydrologic utility of four different satellite precipitation datasets (SPDs), including GPM (IMERG_F), PERSIANN_CDR, CHIRPS, and CMORPH, to predict daily streamflow and SL using the SWAT hydrological model as well as SWAT coupled soft computing models (SCMs) such as artificial neural networks (SWAT-ANNs), random forests (SWAT-RFs), and support vector regression (SWAT-SVR), in the mountainous Upper Jhelum River Basin (UJRB), Pakistan. SCMs were developed using the outputs of un-calibrated SWAT models to improve the predictions. Overall, the GPM shows the highest performance for the entire simulation with R 2 and PBIAS varying from 0.71 to 0.96 and −13.1 to 0.01%, respectively. For the best GPM-based models, SWAT-RF showed a superior ability to simulate the entire streamflow with R 2 of 0.96, compared with the SWAT-ANN ( R 2 = 0.90), SWAT-SVR ( R 2 = 0.87), and SWAT-CUP ( R 2 = 0.71). Similarly, SWAT-ANN presented the best performance capability to simulate the SL with an R 2 of 0.71, compared with the SWAT-RF ( R 2 = 0.66), SWAT-SVR ( R 2 = 0.52), and SWAT-CUP ( R 2 = 0.42). Hence, hydrological coupled SCMs based on SPDs could be an effective technique for simulating hydrological parameters, particularly in complex terrain where gauge network density is low or uneven. HIGHLIGHTS: Soft computing models development using the outputs of un-calibrated SWAT models to improve the prediction of daily streamflow and sediment load in Rivers. Effectiveness of the hydrological coupled soft computing models based on satellite precipitation datasets for simulating hydrological parameters. Auto-optimization of different sensitive parameters of the soft computing models to improve predictions. Graphical Abstract