Extreme heat is deadly and disproportionately affects the elderly and residents of low-income neighborhoods. Extreme heat and humidity events will increase in coming years. Lack of air-conditioning and the urban heat island effect create dangerous indoor conditions. Up to 60% of older, poorly built homes in low-income areas lack AC. Combining residents’ behavior and building characteristics in machine learning (SciML) and agent-based models (ABM) will identify when residents are exposed to overheating risks in their home and connect them with resources to mitigate dangerous conditions. Leveraging collaboration between Iowa State University, the City of Des Moines, Polk County, community organizations, and collaborators at UNI and UTA, this team will use empirical data and participatory processes to develop novel hybrid social- and physics-aware models to increase predictability of extreme heat-related indoor conditions. A community-focused microclimate-informed indoor heat emergency alert (CommHEAT) system will personalize community heat-related emergency management capacity. This will provide societal benefits via improved prediction of indoor conditions in homes for adaptation to extreme heat. The social-biophysical models are broadly transferable to facilitate climate adaptation and improve public health associated with extreme heat. Project completion will provide communities with a framework for microclimate-informed heat alerts in real time. Outcomes will support local heat health action plans to reduce emergency calls, heat illness-related hospitalizations, and mortality from indoor heat exposure.

This research will address knowledge gaps through social and thermal-physical models of adaptation/response to extreme heat indoors. Data will integrate social/behavioral responses to extreme heat with physics-constrained models and develop response strategies across spatial and temporal scales. Intellectual merit includes development of novel data-driven modeling combining validated ABM and physics-constrained SciML models of building features with human behavior within/near buildings. This project contributes to three scientific advances: (1) describing human behavior during extreme events based on human choices; (2) creating localized physics-constrained indoor condition models with real-time parameters; and (3) integrating models in an app using a transferable framework to predict conditions over time. Through participatory design with vulnerable residents in the study area the ABM will be an empirically valid model of the community, enabling prediction of responses to different app-enabled heat mitigation strategies under climate scenarios in a heat alert system that will improve health outcomes. Transferable SciML will account for underlying physics that tie local conditions to building thermal properties. The CommHEAT app will visualize alert mitigation scenarios to guide decision-making at multiple scales for adaptation.