A Mobile Sensor Network to Map CO2 in Urban Environments

Joey Lee, Andreas Christen, Zoran Nesic, & Rick Ketler
Department of Geography
The University of British Columbia


9th International Conference on Urban Climate July 24, 2015
Toulouse, France, 2015


  • We present a pilot study to show the potential for a mobile sensor network to monitor greenhouse gas concentrations and to derive emissions in cities.

Research Question:

Can we map greenhouse gases, specifically CO2, at a spatial resolution of neighborhoods / blocks across the city with a network of mobile sensors ?

Motivations: Why Mobile Sensing?

  • Cities are a major source of GHGs into the atmosphere (Estimated 70%, Rosenzweig 2010).
    • Currently there are a lack of tools to monitor CO2 emissions in cities; existing tools are limited by funding, time, expertise, spatial & temporal resolution.
    • There’s a need to validate fine scale emissions inventories.
  • Emerging Opportunities
    • Rise of flexible (open source), compact technologies.
    • Enhanced access to mobility services/platforms

But how do you go from concentrations to emissions ?

Proposed Approach: Using the aerodynamic resistance (with a number of assumptions!)

Sensible heat flux & temperature are used to calculate the aerodynamic resistance for heat. Surface temperature is calculated with a radiometer at the surface. Assuming that that the aerodynamic resistance of CO2 and heat are the same, the flux is computed.

The Mobile Sensor System

System Setup

1. IRGA – Licor LI-820 (Licor Inc, Lincoln, NB, USA), 2. Adafruit GPS (Adafruit Industries, Manhattan, NY, USA), 3. Arduino Mega (Arduino CC, Ivrea, Italy) . *Not Shown: OneWire Digital Temperature Thermometer (Maxim Integrated One Wire Digital Temperature Sensor - DS18B20, San Jose, CA, USA )

System Specifications

  • Runs on 12 V Power supply
  • Measures CO2 mixing ratios (ppm), geoposition and speed, and air temperature at 1-second intervals
  • Size: 35.8 cm x 27.8 cm x 11.8 cm, weight: 2.6 kg.
  • Precision: +/- 1.1 ppm
  • Accuracy: +/- 3 ppm, with maximum drift of +/- 4 ppm over seven days.
  • Total delayed response time: 13 s with 3 m sample tube at flow rate of 700 cc/min.

Built & Tested: 5 mobile systems

Image: In total 5 sensors were built – the image shows the full setup including the sample inlet tube and the temperature probes.

Mapping Emissions – Methods: Measurement Campaign

May 25th, 2015

Study Area: 12.7 km2 transect, Vancouver, BC

Image: 12 km2 transect study area in Vancouver, BC. The transect is 1km x 12.7km covering the major land cover types in the city. Sunset Urban Climate tower is shown in orange.

Study Area: Tour of Vancouver

Traverse through Vancouver from Joey Lee on Vimeo.

Study Area: Meteorology - May 25th, 2015

  • Measurement period: 10:30 – 14:00
  • With convective and steady weather:
    • Temperature: 20° - 22° C.
    • Weak winds: 2.5 m/s.
    • Cloudless
    * data measured at Urban Climate Tower "Vancouver Sunset" (SE section of Transect)

Sampling Methods: Vehicle Installation

Image A: Shows the temperature probe covered by PVC tube and reflective tape and sample inlet tube at 2m height - ±0.5°C Accuracy from -10°C to +85°C ; Image B: Shows sensor installation in vehicle.

Sampling Methods: Bike Installation

Image: Shows the installation of the sensor on a bike rack – the inlet is at approximately 2m height.

Sampling Methods: Deployment Transects

Image: The image shows 5 planned transect routes for the measurement campaign. Goal: to cover (almost) all navigable roads (and some trails) along the transect in 3.5 hours.

Pilot Study - Mapping CO2 Emissions: Results

Raw Data: Visualized in Google Earth

Average CO2 Mixing Ratios per 100m Grid Cell

Calculated Emissions: Concentrations to Emissions

Image: Calculated emissions map generated using the aerodynamic resistance approach using CO2 concentration measurements.

Mapping Emissions – Methods: Emissions Inventory

Traffic Emissions Inventory

Image: The traffic emissions inventory derived from Vancouver’s traffic count data and calculated per grid cell using fuel consumption and emissions factors.

Building Emissions Inventory

Image: Building emissions inventory generated by combining factors of building morphology, urban context, and population density derived from LiDAR, building topology, and census data.

CO2 Mixing Ratios Vs. Traffic Emissions Inventory:

CO2 Mixing Ratios Vs. Total Emissions Inventory:

Calculated CO2 emissions vs. Traffic Emissions Inventory

Calculated CO2 emissions vs. Building Emissions Inventory

Calculated CO2 emissions vs. Total Emissions Inventory

Conclusions & Future Directions

  • A mobile CO2 sensing system was developed using open source components.
  • The system was tested in Vancouver and evaluated against traffic and building emissions inventories.
  • The study shows the potential for pervasive mobile sensing in GHG and pollutant monitoring in cities, but relies on specific conditions and assumptions.
  • Currently exploring visualization & feedback opportunities for planning and open science.
  • Collaboration potential with local mobility providers for long term & spatially extensive/intensive monitoring

Acknowledgements & Funding

  • NSERC Discovery Grant
  • NSERC CREATE - Terrestrial Ecosystems Research and World Wide Education & Broadcast Program (TerreWEB).
  • UBC Geography Department
  • Benedikt Groß, 47Nord GmBH, & Moovel Labs
  • Measurement Campaign Drivers: Alex McMahon, Andreas Christen, Mark Richardson, & Thea Rodgers.
  • Data: OpenStreetMap.org, Vancouver Open Data, EPiCC Project, UBC Micrometeorology
  • Tools: Python, R, QGIS, GDAL/OGR
  • References
    • Van der Laan, "Scaling Urban Energy Use and Greenhouse Gas Emissions through LiDAR", MSc Thesis, 2011.
    • Rosenzweig, C., Solecki, W., and Hammer, S. A. (2010). Cities lead the way inclimate-change action. Nature, pages 1–3.

Special Thanks to Andreas, Zoran, and Rick

The slides & links can be found on github:



Questions? Comments?

Many thanks to the ICUC & IAUC Community!