Mapping real-time air pollution health risk for environmental management: Combining mobile and stationary air pollution monitoring with neural network models

摘要: Air pollution poses health concerns at the global scale. The challenge of managing air is significant because many pollutants, insufficient funds for monitoring and abatement programs, political social challenges in defining policy to limit emissions. Some governments provide citizens with risk information allow them their exposure. However, regions still have networks real-time mapping. Where available, these mapping systems either absolute concentration data or concentrations are used derive an Quality Index, which provides a mix pollutants single value. When presented as value entire region it does not inform on spatial variation within region. Without understanding local residents can only make partially informed decision when choosing daily activities. typically provided limited number active units area. In our work, we overcome this issue by leveraging mobile techniques, meteorological land use map risks. We propose approach that improved public applying neural network models framework inspired regression. Mobile campaigns were conducted across Hamilton from 2005 2013. These modelled predictor variables included surrounding characteristics, conditions, fixed location monitors, traffic during time collection. Fine particulate matter nitrogen dioxide both modelled. During model fitting process reserved twenty percent validate predictions. models' performances measured coefficient determination 0.78 0.34 PM2.5 NO2, respectively. apply relative importance measure identify each variable black box issues models.