An exploration of urban air health navigation system based on dynamic exposure risk forecast of ambient PM2.5

Abstract

Under international advocacy for a low-carbon and healthy lifestyle, ambient PM2.5 pollution poses a dilemma for urban residents who wish to engage in outdoor exercise and adopt active low-carbon commuting. In this study, an Urban Air Health Navigation System (UAHNS) was designed and proposed to assist users by recommending routes with the least PM2.5 exposure and dynamically issuing early risk warnings based on topologized digital maps, an application programming interface (API), an eXtreme Gradient Boosting (XGBoost) model, and two-step spatial interpolation. A test of the UAHNS’s functions and applications was carried out in Wuhan city. The results showed that, compared with trained random forest (RF), LightGBM, Adaboost models, etc., the XGBoost model performed better, with an R2 exceeding 0.90 and an RMSE of approximately 15.74 μg/m3, based on data from national air and meteorological monitoring stations. Further, the two-step spatial interpolation model was adopted to dynamically generate pollution distribution at a spatial resolution of 300 m*300 m. Then, an exposure comparison was performed under randomly selected commuting routes and times in Wuhan, showing the recommended routes for lower PM2.5 exposure made effectively help. And the route difference ratios of about 14.9 % and 16.9 % for riding and walking, respectively. Finally, the UAHNS platform was integrally realized for Wuhan, consisting of a real-time PM2.5 query, a one-hour PM2.5 prediction function at any location, health navigation on city map, and a personalized health information query.

Keywords

• PM2.5
• Air Health
• Dynamic Exposure Risk
• Machine learning model
• Route planning