Journal
ATMOSPHERIC ENVIRONMENT
Volume 126, Issue -, Pages 171-181Publisher
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.atmosenv.2015.11.033
Keywords
Particle number concentration (PNC); Statistical modelling; Mobile sensor network; Geoinformation; Pollution maps; Urban environment
Funding
- VBZ (public transport company of the City of Zurich)
- Nano-Tera.ch with Swiss Confederation financing
Ask authors/readers for more resources
Highly resolved pollution maps are a valuable resource for many issues related to air quality including exposure modelling and urban planning. We present an approach for their generation based on data from a mobile sensor network and statistical modelling. An extensive record of particle number concentrations (PNCs) spanning more than 1.5 years was compiled by the tram-based OpenSense mobile sensor network in the City of Zurich. The sensor network consists of 10 sensor nodes installed on the roof of trams operating on different services according to their regular operation schedules. We developed a statistical modelling approach based on Generalized Additive models (GAMs) utilizing the PNC data obtained along the tram tracks as well as georeferenced information as predictor variables. Our approach includes a variable selection algorithm to ensure that individual models rely on the optimal set of predictor variables. Our models have high temporal and spatial resolutions of 30 min and 10 m by 10 m, respectively, and allow the spatial prediction of PNC in the municipal area of Zurich. We applied our approach to PNC data from two dedicated time periods: July-Sept. 2013 and Dec. 2013-Feb. 2014. The models strongly rely on traffic related predictor variables (vehicle counts) and, due to the hilly topography of Zurich, on elevation. We assessed the model performance by leave-one-out cross validation and by comparing PNC predictions to measurements at fixed reference sites and to PNC measurements obtained by pedestrians. Model predictions reproduce well the main features of the PNC field in environment types similar to those passed by individual trams. Model performance is worse at elevated background locations probably due to the weak coverage of similar spots by the tram network. We end the paper by outlining a route finding algorithm which utilizes the highly resolved PNC maps providing the exposure minimal route for cyclists. (C) 2015 Elsevier Ltd. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available