Innovative Sensors and Models for City-Level Air Pollution Exposure Monitoring
conference paper
Results: The exploratory analysis showed that the variability in the ILM PM data increased from 10% to 16% (coefficient of variation) between 2013 and 2015 The ILM data tended to record lower values than the LML data, although the peaks and troughs were still observed. The 10-minute data were typically too noisy to allow the identification of spatial correlation, although this was clear when averaged to 1-hour averages and for longer time periods. Using BME we were able to integrate the dispersion model and ILM data, yielding an RMSE of 1 µg m-3 for daily PM2.5. The SOA was effective to combine different datasets (e.g., ILM and LML) and to implement simple geostatistical models in an automated fashion. Conclusions: Using low-cost sensors allowed us to identify spatial and temporal patterns that are valuable for spatial prediction at sub-daily time resolutions. These patterns can be identified and modelled using space-time geostatistics e.g., BME. Standards-based tools should be used for organizing, archiving and disseminating data and are essential for future automated analysis.
-
TNO Identifier
849728
Publisher
ECN
Collation
5 p.
Place of publication
Petten
Files