Development of Long Short-Term Memory-Based Water-Energy-Food Nexus for Assessing Resources vulnerability

Recent climate changes, including rising temperatures and altered precipitation patterns, pose a significant threat to global water, food, and energy security. In particular, the acceleration of global warming has resulted in water scarcity, reduced crop yields, and unstable energy supplies, impacting not only human livelihoods but also various industries. Although water, energy, and food have traditionally been managed independently, they are closely interconnected, and changes in one sector can directly influence the others.

It is vital to manage the water-energy-food (WEF) nexus in an integrated manner and anticipate potential resource shortages in advance. In this context, introducing the concept of a water footprint enables the development of efficient, water-centric management strategies that systematically measure and manage water usage across energy and food production.

This study applies a water footprint approach to factors related to water, energy, and food, calculates the Water Stress Index (WSI), Energy Stress Index (ESI), and Food Stress Index (FSI), thereby assessing each resource’s vulnerability. Specifically, an Long Short-Term Memory(LSTM)-based WEF nexus model was developed for Chungcheongnam-do in South Korea to evaluate nexus interactions using historical observation data. Furthermore, Shared Socioeconomic Pathway (SSP) scenarios were employed to project resource fluctuations under future climate change.

To validate the model, a Walk-Forward Validation Fold method was used, yielding an R² of 0.88 and a Nash-Sutcliffe efficiency (NSE) of 0.67—indicating satisfactory predictive accuracy. By setting 2030 as the target year, the model showed that WSI could range from -168% to +16%, ESI from -690% to +69%, and FSI was projected to decrease by -113% to -87%. Notably, reduced precipitation was identified as having a significant impact on energy production, underscoring the need for strategies to ensure a stable energy supply.

 

Acknowledgement: This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (NRF-RS-2023-00280330).