Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Using low-cost sensors to improve NO2 concentration maps derived from physico-chemical models

Emma Thulliez, Camille Coron

Abstract

Urban air quality is a major concern today. Concentrations of pollutants, such as nitrogen dioxide, must be monitored to ensure that they do not exceed hazardous thresholds. For this reason, scarse reference stations, which are generally managed by air quality monitoring associations, are located in major cities. Two recent approaches enable fine-scale mapping of pollutant concentrations. The first relies on deterministic physico-chemical models that incorporate the street network and compute concentration estimates on a grid, producing spatial maps. The second is based on the emergence of low-cost sensors, which enable monitoring organizations to increase the density of their measurement networks. However, these sensors are unreliable and require regular and important calibration. We propose to combine these approaches and improve maps generated by deterministic models by integrating data from multiple sensor networks. Specifically, we model the bias of deterministic models and estimate its parameters using measurements, through a Bayesian nested framework. Our approach simultaneously enables the calibration of low-cost sensors and the correction of deterministic models outputs. This method, although general, is applied to the city of Rouen (France), combining outputs of the physico-chemical model SIRANE (Soulhac et al. 2011) and the measurements provided both by 4 reference monitoring stations and 10 low-cost sensors during December 2022. Results show that the method indeed corrects the concentration maps, reducing the root mean squared error by approximately 12.4%, and that low-cost sensors play an essential role in this correction.

Using low-cost sensors to improve NO2 concentration maps derived from physico-chemical models

Abstract

Urban air quality is a major concern today. Concentrations of pollutants, such as nitrogen dioxide, must be monitored to ensure that they do not exceed hazardous thresholds. For this reason, scarse reference stations, which are generally managed by air quality monitoring associations, are located in major cities. Two recent approaches enable fine-scale mapping of pollutant concentrations. The first relies on deterministic physico-chemical models that incorporate the street network and compute concentration estimates on a grid, producing spatial maps. The second is based on the emergence of low-cost sensors, which enable monitoring organizations to increase the density of their measurement networks. However, these sensors are unreliable and require regular and important calibration. We propose to combine these approaches and improve maps generated by deterministic models by integrating data from multiple sensor networks. Specifically, we model the bias of deterministic models and estimate its parameters using measurements, through a Bayesian nested framework. Our approach simultaneously enables the calibration of low-cost sensors and the correction of deterministic models outputs. This method, although general, is applied to the city of Rouen (France), combining outputs of the physico-chemical model SIRANE (Soulhac et al. 2011) and the measurements provided both by 4 reference monitoring stations and 10 low-cost sensors during December 2022. Results show that the method indeed corrects the concentration maps, reducing the root mean squared error by approximately 12.4%, and that low-cost sensors play an essential role in this correction.

Paper Structure

This paper contains 25 sections, 14 equations, 9 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Physico-chemical models outputs
  4. Air quality measurements
  5. Spatial covariates
  6. Temporal covariates (Meteorological data)
  7. Model
  8. Bias of the physico-chemical model output
  9. Measures
  10. Estimation
  11. Parameters
  12. Estimation procedure
  13. Prior distributions
  14. Validation process
  15. Results
  16. ...and 10 more sections

Figures (9)

  • Figure 1: Map of Rouen provided by OpenStreetMap and locations of low-cost sensors (blue circles) and reference stations (red circles).
  • Figure 2: NO$_{2}$ concentration at stations SUD3. Measurements are represented in red and SIRANE estimations in blue.
  • Figure 3: NO$_{2}$ concentrations at sensors ASE4 (solid lines) and ASE9 (dashed lines). Measurements (in mV) are represented in red (scale on left) and SIRANE estimations in blue (scale on the right).
  • Figure 4: Spatial data in Rouen
  • Figure 5: Temporal data in Rouen during December 2022, part of SIRANE's inputs. Retrieved from a local meteorogical monitoring station.
  • ...and 4 more figures