Low cost sensor networks for measuring urban air quality

Low cost sensor networks for measuring urban air quality

Rod Jones (and others)Centre for Atmospheric Science

University of [email protected]

mailto:[email protected]

Measuring Environmental Exposure, Montreal 2011

Granularity in urban air quality on all measured scales

• Substantial inhomogeneity exists in urban A/Q (sources, meteorology, chemistry)

• Non-linearity in chemical processes –main atmospheric science driver

• Precision (e.g. UK: AURN) fixed monitoring sites costly and sparse (capture real variability?)

• Low cost solutions (e.g. NO2 diffusion tubes) give time (>bi-week) averages (precision?)

® Alternative solutions?

2mm

Basis for sensor network system approach:

Low cost miniature gas Sensor technologies

• Electrochemical• Non Dispersive IR• Photo-ionisation Detection• Metal Oxide• SAW• ……..

>factor of 100 cheaper (and smaller) than traditional methods- if they can be made sensitive enough


Can (e.g.) electrochemical sensors be sensitive/selective

enough for urban air quality applications?

(traditionally used at >>ppm levels)

(ppm = parts per million)


Enhancements to electrochemical sensors:

• Hardware- Improved sensitivity and stability- Improved selectivity & related methodologies- Quantification of ambient temperature effects

• Control electronics- Improved stability/noise characteristics

• Software - Algorithms for baseline correction- Performance monitoring- Calibration methodologies


70ppb

35ppb

17ppb

8ppb3ppb

0ppb

(20 ppb of NO2 ~ 40⎧g/m3)

Performance of (enhanced) electrochemical sensors NO2 sensitivity (laboratory)


160 ppb

80 ppb

38.4 ppb

9.6 ppb0 ppb

Performance of (enhanced) electrochemical sensors CO sensitivity (laboratory)


130 ppb

31.2 ppb

15.6 ppb7.8 ppb0 ppb 2.6 ppb

Performance of (enhanced) electrochemical sensors NO sensitivity (laboratory)


Sensitivity at the ppb level, (repeatability + selectivity*) in laboratory

Current UK/EU threshold

Electrochemical sensor CO/NO/NO2sensitivity (laboratory)

* Some exceptions, but…


Electrochemical sensor performance (laboratory)

Typical sensor sensitivities/LoD are < 5ppb (< 7⎧g/m3) for CO, 1-2 ppb (~2-4 ⎧g/m3) for NO and NO2.

SO2, O3 have comparable performance to NOx.

Typical sensor T90 ~ 10-20s (determined by diffusion)

Very low power consumption (⎧W)

Noise characteristics:


Electrochemical sensor performance:

Is laboratory performance replicated in the field?

•Ambient temperature corrections…•Cross interferences…..•…….•…•..


Electrochemical sensor baseline temperature dependence correction

Raw data (electrochemical, chemiluminescence)

Sensor temperature and baseline correction

Baseline temperature corrected(comparison with ratified data)


Cross interference (NO2/O3) + comparison with ratified site (hourly average)


‘Real world’ comparison of NO2 and NO with ratified AURN site

Performance replicated in the field….* Corrected for O3 interference

* *


Reproduced in field? – single sensor

NO2


Co-located sensors (NO2) – real structure

(Uncorrected for O3)


Co-located sensors (NO)


Co-located sensors (CO)

Combine technologies to provide sensitive low cost sensor network system

• Real time location and data transmission by coupling sensor technology to GPS and GPRS

• Real time data processing including analysis/interpretation and visualisation(Imperial College)


Multi-species real time mobile measurements of air quality in complex environments

Multiple sensors, (NO2)

QuickTime™ and a decompressor

are needed to see this picture.

QuickTime™ and aCinepak decompressor

are needed to see this picture.

Relating traffic and CO, NO, NO2

CO ( ⎟ 50)NO ( ⎟ 5)NO2


Mobile sensor network deployment: Cambridge (UK)

• 4 hour deployment (2009)

• > 40 sensors (CO, NO, NO2, CO2, VOCs)• 3 transport modes • (walkers, cyclists, vehicles)• Inner city, outer loop (A14/M11)• Real time GPRS transmission

• >200,000 measurements


Cambridge deployment September 2009: NOx


Visual determination of pollution hotspots – not possible with static sites


Statistical assessment of mobile A/Q data by transport mode (simplest possible!)


Derive ensemble averaged statistics –distinguish between transport modes:


An example: NO2: car vs bicycle– rural vs motorway

Wind direction

Personal exposure vs activity……..


Quantification of individual exposure (not dosage – how to link to health outcomes?)

Methodology for quantification of peak and average exposures

Individual high pollution events

Ensemble statistics on exposure

vs

Threshold?

Comparison of different environments: (Cambridge vs Valencia (Spain))

Significant differences e.g. CO, NO2

- representativeness??

Compare ‘snapshots’ with Lagos (Nigeria)

Dramatic differences……….

Flexible low cost way of characterising A/Q

20 ppm

Validation of models

CERC ADMS model (David Carruthers)

Structure within street canyons?

Mobile sensor network quantification of personal exposure

Determine personal exposures rather than ‘simple’ spatial

distribution?


Static Sensor Deployment, Cambridge (UK)

• >2 month deployment (May-July 2010)• >40 sensors (CO, NO, NO2) , T, RH• Lamp post mounted, GPRS (GPS)• Inner city, mixed urban, rural• Real time GPRS transmission

• >25,000,000 measurements


1 month

1 week

1 day

4 hours (individual pollution events)

Static deployments (carbon monoxide)

Grid(46 sites)

vs


A) High density statistical evaluations of A/Q

B) Source attribution in urban environment

• Industrial emissions (in this case airports)

BB

AA CC

Sensor network applications:

NO2

CO

NO

UK National funded high density sensor network system at UK Heathrow airport (2011-2013)

•60 sensor nodes, real time data transfer •NO, NO2, CO, CO2, SO2, O3, VOCs and size-speciated PM.•Source attribution/model validation for area. •Novel software tools for calibration,

data-mining, visualisation/interpretation.•Methodology for optimising sensor network design.

(Electrochemical, NDIR, PID, Optical)

One key aim is to assess added information content from sensor network…

High density sensor network system for air quality studies at Heathrow airport

Participants:Institution

InputUniversity of Cambridge (PI) sensors, a/q models

Imperial College London traffic models, visualisation

University of Hertfordshire aerosol measurements

University of Manchester aerosol measurements

CERC Ltd a/q modelling – ADMS

National Physical Lab. metrology, calibration

Alphasense Ltd sensor Monitoring Ambient Air 2011Royal Society of Chemistry


High density sensor network system for air quality studies at Heathrow airport

Timeline:

Year 1 (2011): planning, sensor node manufactureYear 2 (2012, Q2): sensor network deployment Year 3 (2012-13): data mining, source attribution, network

design

watch this space!Monitoring Ambient Air 2011Royal Society of Chemistry


Summary Remarks• Ultra small low cost A/Q sensors viable for relevant gas phase species

(demonstrated at least scientifically) at the ppb level- Quantification of air quality at high “actual” spatial resolution - “Real” personal exposure vs spatial maps- Source attribution

• Added value of dense networks (Heathrow Airport high density network?)

• Complement existing FSM networks

• Combine with air quality models (e.g. exposure quantification)

• Extend low cost philosophy……- Individual VOCs, other….- PM…..- Indoor air quality


AcknowledgementsSensors and Sensor NetworksIq MeadLekan PopoolanGregor StewartJohn Saffell, AlphasenseMark HayesMark CallejaRobin NorthJeremy CohenPaul Kaye and UH teamDavid Carruthers (CERC)

Earlier workImperial CollegePeter Landshoff…….

Documents

Low cost sensor networks for measuring urban air quality