AN AI-BASED FRAMEWORK FOR MODELING STREAMFLOWS UNDER THE WATER-ENERGY-FOOD-ECOSYSTEM NEXUS APPROACH: A COMPREHENSIVE ASSESSMENT OF THE WATERSHEDS OF TÜRKİYE

Date

2025-10-17

Author

Habash, Abdelrahman

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

334
views

0
downloads

Cite This

A well-functioning Water–Energy–Food–Ecosystem (WEFE) Nexus is critical to support sustainable development under increasing climatic and anthropogenic pressures. Yet, most hydrological modelling frameworks still treat human activities only implicitly, limiting their ability to represent the feedbacks between water resources, energy demand, food production, and ecosystem functioning. This thesis develops an AI-based modelling framework for predicting streamflow dynamics across Türkiye while explicitly integrating WEFE indicators. The framework couples a recurrent branch, which ingests short-term climatic sequences, with a feed-forward branch representing static physiographic descriptors and annual WEFE indicators, and is trained on daily streamflow records from 258 DSİ gauging stations spanning the country’s major river basins. Model outputs are evaluated at the monthly scale using Nash–Sutcliffe Efficiency (NSE), RSR, and percent bias (PBIAS), and two configurations are compared: a Baseline model driven only by hydroclimatic and physiographic predictors, and a WEF-enhanced model that additionally incorporates socio-economic WEFE indicators. Across all stations, 149 (58%) achieve satisfactory performance in at least one configuration, with an average NSE of 0.77, RSR of 0.46, and |PBIAS| of 9.5%, and 93 stations (62% of the satisfactory subset) exceeding NSE = 0.75. Within this satisfactory group, approximately 60% of stations exhibit improved predictive skill when WEFE indicators are included, with around one third showing moderate to substantial gains, particularly in hydrologically constrained, low-snow basins where human water management plays a stronger role. An Accumulated Local Effects (ALE) analysis is used to interpret the contribution of WEFE predictors, highlighting variables such as pumped irrigation capacity, municipal water abstraction, livestock numbers, and electricity consumption as key drivers of anthropogenically modulated flow regimes. Overall, the results demonstrate that WEFE-aware, explainable AI models can complement traditional climate-based approaches by both enhancing streamflow prediction and providing physically meaningful insight into human–water–energy–food interactions to support integrated water resources planning in Türkiye and similar data-scarce regions.

Subject Keywords

AI in Hydrology, Hydrological Modeling, Streamflow Prediction, Water-Energy-Food-Ecosystem (WEFE) Nexus, Sustainable Water Resource Management

URI

https://hdl.handle.net/11511/117366

Collections

Graduate School of Natural and Applied Sciences, Thesis