Improving obesogenic environmental assessments with advanced geospatial methods
Shaoqing Dai is a PhD student in the Department of Earth Observation Science. (Co)Promotors are prof.dr.ir. A. Stein and dr. F.B. Osei from the Faculty ITC and prof. P. Jia from the Wuhan University.
This thesis explores the intricate connections between the environment and obesity. It develops and applies advanced geospatial methods to enhance the assessment of obesogenic environments and obesity risks. Its primary objective is to evaluate obesogenic environments and explore potential associations between environmental factors and obesity, crucial for effective obesity prevention. The thesis is structured around four key objectives.
The first sub-objective involves an investigation into the current literature on the measurement of the built environment. Street View Imagery (SVI) and advanced urban visual intelligence technologies have transformed Built Environment Auditing (BEA) substantially, enabling large-scale auditing at a detailed geographical level. A meticulous review of 96 articles published before September 15, 2023, reveals key areas for improvement in SVI-based BEA. Recommendations include standardized datasets for more accurate audits, the integration of multi-source SVI for comprehensive assessments, and the design of auditing tools tailored to developing countries. Addressing these areas enhances the potential of SVI in environmental auditing, as they contribute to a better understanding of the built environment’s health impact and facilitate informed decision-making in urban planning and public health initiatives.
The second sub-objective focuses on analyzing exposure to increasing PM2.5 pollution, associated with rising morbidity and mortality. An ensemble machine learning model, integrating multi-source geospatial data, is presented to map hourly street-level PM2.5 concentrations in the city of Nanjing, China, at a 100 m spatial resolution. The study concludes that mapping these concentrations reveals spatiotemporal trends, supporting the establishment of exposome studies.
The third sub-objective addresses the development of a framework to evaluate Physical Activity (PA) opportunities (bikeability) in urban environments, aiming to enhance sustainable urban transportation planning. A framework is proposed that comprises safety, comfort, accessibility, and vitality sub-indices. It uses open-source data, advanced deep neural networks, and GIS spatial analysis, to eliminate subjective evaluations and enhance efficiency. Experimental results in the city of Xiamen, China, demonstrate the framework’s effectiveness in identifying areas for improvement and enhancing cycling mobility.
The fourth sub-objective investigates the associations between PA opportunities, specifically walkability, and obesity. Using a cross-sectional cohort from Nanjing, China. A Logistic regression model with a double robust estimator estimates the effects of walkability on obesity risks. A newly developed walkability index shows a significant negative association with obesity, particularly when using a data-based-buffer derived from web-mapping navigation that better represents individual activity spaces. These findings provide evidence for developing explicit strategies for obesity prevention.
In summary, this thesis contributes to addressing the knowledge gap in health geography between obesogenic environments and obesity risks, employing advanced geospatial methods. The integration of multi-source geospatial data, machine learning methods like deep learning in a GIS environment, and spatial statistics presents a major step forward.
