Droughts are increasingly recognized as environmental crises with far-reaching consequences, not just on water availability, but on agriculture, the economy, public health, and society. While current drought monitoring systems primarily focus on assessing drought severity using quantitative measurements, such as meteorological and hydrological data or economic losses, they often miss what matters most: how societies and communities are affected.
A new study led by Professor Dong Wang introduces SIDE (Socially Informed Drought Estimation), a novel socially informed AI-driven drought estimation framework that estimates both drought severity and its societal impact using data from social and news media.
"The lack of human-centric perspectives in drought severity assessment can lead to an incomplete understanding of the societal impact and hinder the development of effective mitigation strategies that address the diverse needs and concerns of affected populations," Wang said.
SIDE is the first tool of its kind to capture what the researchers call the "social-physical interdependence of droughts," the way human behavior and environmental conditions influence each other. For example, as water becomes scarce, communities might respond by increasing usage in fear of shortage, worsening the situation. These behavioral patterns, often captured in local reporting or social media, could offer valuable insights into the real-world consequences of environmental crises.
The researchers evaluated SIDE using the publicly available SocialDrought dataset, focusing on California and Texas as the primary geographical areas in their study due to their significant drought vulnerability, population diversity, and agricultural importance. Their study analyzed data from January 2017 to April 2023, during which both states experienced significant drought events.
Compared to five leading time-series forecasting models, SIDE substantially outperformed them in accurately estimating drought severity and societal impact. Notably, SIDE observed distinct patterns between California and Texas: In California, societal concern focused more on agriculture and wildfire management, whereas in Texas, ecosystem and public health were more prominent.
Designed for real-world deployment, SIDE could be integrated into national systems like the U.S. Drought Monitor (USDM 2024) or DIP-Drought Monitor (DIP-Drought 2024), a collaborative initiative co-led by Wang and Ximing Cai, a professor of civil and environmental engineering at Illinois, who also serves as the domain expert from hydrology for SIDE. The team presented their work at the 39th AAAI Conference on Artificial Intelligence earlier this year.
By offering timely, human-centric insights, SIDE can support faster and smarter decision-making by government agencies, water resource managers, agricultural organizations, and community leaders and members. The tool could also be adapted to other environmental crises, such as floods, wildfires, and extreme weather events.