Extreme heat exposure poses a serious risk to human health and wellbeing, especially in densely populated urban areas due to Urban Heat Island (UHI) effects. This presentation proposes a framework of integrating cyberGIS (i.e., cyber-based geographic information science and systems) and machine learning to better understand urban heat dynamics. This framework is designed to harness complex and heterogeneous geospatial data for capturing fine-scale and real-time characteristics of urban heat dynamics.
Two case studies have been conducted to demonstrate the capability and utility of the framework. One study focuses on predicting UHI in Chicago, Illinois at fine spatiotemporal scales (e.g., 1 km*1km, 10-minutes) using high-frequency urban sensor network data combined with satellite remote sensing data. Humidity, distance to geographic center, and PM2.5 concentration are identified as key factors affecting prediction performance. The other study uses location-based social media data to analyze people’s sentiments on heat exposure in a near real-time fashion by applying a natural language processing algorithm to geotagged Tweets. Dynamic mapping and visualization based on such sentimental analysis help identify vulnerable populations.