As climate change intensifies the frequency and severity of droughts worldwide, effective water resource management has become a critical priority. A team of researchers at POSTECH, led by Professor Jonghun Kam, has leveraged explainable artificial intelligence (XAI) to quantify how media coverage of drought influences household water conservation behaviors during the severe 2022-2023 drought in southwestern South Korea. Their groundbreaking study, published in Water Resources Research, bridges physical drought metrics with social responses, providing new insight into the human dimension of drought mitigation.
Traditional drought models primarily focus on hydrological factors such as precipitation deficits, reservoir levels, and temperature anomalies. Yet, water conservation decisions by households are deeply influenced by social variables including public awareness, media exposure, and regional socioeconomic characteristics. To capture these complex dynamics, the research team integrated diverse datasets—household water usage, drought indices, dam storage, temperature data, and counts of drought-related news articles—into sophisticated AI models.
Using the N-BEATS deep learning framework, the study accurately predicted monthly variations in water conservation rates with over 73% precision across both metropolitan and rural settings. Notably, metropolitan regions with higher population density and income demonstrated more proactive voluntary water-saving behavior during drought events. For example, Gwangju Metropolitan City achieved water conservation increases of up to 10%, whereas smaller cities and rural districts in Jeollanam-do experienced more modest gains around 3%.
The AI-driven scenario analysis revealed an impactful relationship between drought news frequency and conservation outcomes. Increased media coverage during the early and caution stages of drought led to a significant 14-percentage point rise in household conservation in Gwangju, translating into approximately $0.82 million in reduced tap water production costs. In contrast, comparable media surges in smaller or rural areas yielded only a four-percentage point increase, equivalent to nearly $0.3 million saved. This disparity suggests that urban populations possess greater flexibility to adjust water use and respond to informational cues, whereas rural households may have less discretionary water to cut back.
Timing also emerged as a critical factor in effective drought communication. Media influence was strongest prior to severe drought conditions, when awareness and voluntary action have room to grow. Once drought conditions escalated to alert stages, public concern and conservation behaviors were already elevated, limiting the additional benefits of news exposure. This underscores the importance of early, well-timed communication strategies to optimize water-saving responses.
Currently, drought policies often adopt uniform, top-down water conservation targets without accounting for regional variation. This study’s AI-driven findings advocate for regionally tailored water-saving goals and communication plans that consider both local socioeconomic profiles and drought severity stages. Such data-driven policies promise more efficient allocation of resources and improved drought resilience.
First author Eunmi Lee emphasized that explainable AI allowed the team to quantitatively unravel the subtle interplay between social dynamics—particularly media influence—and actual household water use behavior. The researchers envision that integrating environmental data with social factors via AI will become vital for designing actionable, adaptive drought management strategies under climate change.
Supported by grants from Korean government agencies, this pioneering research highlights the power of AI to deepen our understanding of human-environment interactions. As droughts grow more frequent globally, the fusion of technology, meteorology, and social science offers a promising path toward sustainable water management.
Subject of Research: Water conservation behavior influenced by media during drought, using AI analysis
Article Title: Spatiotemporal and Economic Impacts of Media on Water Conservation During Drought: An Explainable AI Approach
News Publication Date: 17-Jun-2026
Web References: http://dx.doi.org/10.1029/2025WR042406
Image Credits: POSTECH
Keywords: drought, water conservation, explainable AI, climate change, media influence, hydrology, water resources management
