No Detail Found!

Machine Learning-Based Assessment of Hydrological Drought across the Upper, Middle, and Lower Indus River Reaches

Author(s)
M. Irfan Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore,Lahore,Pakistan
M. Atiq Ur Rehman Tariq Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore,Lahore,Pakistan
M. Waseem Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore,Lahore,Pakistan;College of Hydrology and Water Resources, Hohai University, Nanjing, China
Mudassar Iqbal Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore,Lahore,Pakistan
K. Ahmad Centre of Excellence in Water Resources Engineering, University of Engineering and Technology Lahore,Lahore,Pakistan
Abstract
Hydrological drought is a critical event that severely affects surface and groundwater resources, particularly in regions like the Indus River Basin. This study investigates hydrological drought across the upper (Besham Qila), middle (Guddu), and lower (Kotri) reaches of the Indus River using machine learning models, including Support Vector Machines (SVM) and Artificial Neural Networks (ANN). Daily streamflow data were utilized to calculate Standardized Drought Indices (SDI and SDHI) for drought assessment, covering training (1991-2004), evaluation (2005-2013), and prediction periods (2014-2020 for Guddu and Kotri, and 2014-2015 for Besham Qila). The models were evaluated through statistical performance indicators, including RMSE, RSR, PBIAS, R², and NSE. ANN demonstrated strong performance at Besham Qila (R² = 0.79) and Kotri (R² = 0.84), while SVM excelled at Guddu (RMSE = 0.84, R² = 0.89). SVM performs better at Guddu because of localized differences in hydrology and data patterns, even though the overall climate and models are similar to Kotri. Guddu's data showed simpler, more linear relationships that align with SVM's strengths, while Kotri's complex, and non-linear patterns favor ANN. Factors like flow consistency, data quality, and upstream location influence at Guddu create stable conditions, allowing SVM to perform effectively. The study highlights the potential of machine learning in predicting hydrological drought and offers a more accurate understanding of water scarcity patterns. These findings emphasize the need for integrating advanced machine learning techniques into drought management strategies to mitigate the impacts of climate variability on water resources in the Indus River Basin.
Publication Details
Page(s) 1-92
DOI DOI not available
Published Journal: Book of Abstracts: 8th International Conference "Earth Sciences Pakistan 2026" (ESP-2026), June 12-14, 2026, Volume: 1, Issue: 1, Year: 2026

Request the full text from PASTIC

Keywords
Indus river, ANN, Hydrological drought, SVM, machine learning
0

Citations

0

Downloads

8

Views