Urban Pollution Monitoring

1
Urban Pollution Monitoring
Carbon monoxide monitoring on Marylebone Road,
May 2004
2
System Diagram
3
Software

• PollutionReader logs GPS and CO data to files on
  the PDA.
• PollutionSender sends PDA log files to server
  via http.

• PollutionModeller CGI program that serves VRML
  models to http clients.
• DataViewer 2D viewing program for analysing CO
  and GPS

4
Carbon Monoxide Data
24 January 2005 1430, Clerkenwell Road
5
Studies to Date
Marylebone Road, May 2004

• Collaboration with the DAPPLE project.
• Walking and cycling.
• 227,496 data points (63 hours).

Camden, April to May 2004

• Cycling only.
• 109,272 data points (30 hours).

Clerkenwell, November 2004 to February 2005

• Vivacity2020 fixed sensors.
• Walking only.
• 287,690 data points (80 hours).

6
Results

• Marylebone Road and the DAPPLE Project

7
Mean CO levels for the Marylebone Road area.5
metre grid.
Scale
8
Dorset Square

• 4ppm CO levels along Gloucester Place, reducing
  to 1ppm on the west side.
• Results match the traffic flow patterns and agree
  with other DAPPLE scientists results.
• Baker Street and Dorset Road features are visible
  in the raw data.
• Wind affecting CO levels on different sides of
  the road is visible with processing.

9
Dorset Square 17 May 2004, 5 metre grid
10
Results

• Camden

11
Left Route from south Camden across Euston
railway bridge to Euston Road junction with Gower
Street. Right A close-up of the same bridge
showing characteristic peaks of CO.
12
Results

• Clerkenwell

13

• Much larger area 1km x 1.5km.
• 4 fixed sensors as part of Vivacity2020.
• Soundwalk data.

Data from 17 November 2005
14
Clerkenwell
15
Unexpected Features

• High CO in low traffic areas Malet Street,
  Guildford Street and Exmouth Market.

7 December 1430, high peaks near Russell Square
and the corner of Malet Street caused by portable
generators.
16
Plotting a combined variance and mean chart
should show areas of consistently high or low
carbon monoxide. In practice, this highlights the
GPS positioning errors.
Scale Variancegt1.0 GREEN Variancelt1.0 colour
according to mean CO
Marylebone Road, all data.
17
GPS Position Correction

• There is useful data in the shape of the trace.
• Stationary points give an absolute position.
• Position offsets can be removed.
• Knowledge of the route taken provides the most
  important information.
• Additional information in the NMEA logs has yet
  to be included (number of satellites and DOP).
• Correction of ad-hoc data collection is
  difficult.

18
Using simple path correction, differences in CO
level on opposite sides of the street can be seen.
GPS traces being fitted to the correct side of
the road.
The resulting data plotted after correction.
5 May 1000 to 1300
19
Conclusions

• There is good correspondence between our data and
  the data of other scientists working in this
  area.
• The accuracy of the sample position is the main
  limiting factor.

20
Future Work

• Analysis of the Clerkenwell data and comparisons
  with the fixed sensors.
• Fully automatic correction of positions.
• Post processing of GPS?
• Increased positioning accuracy.

• Comparison of sound data with CO.

21
(No Transcript)