local authority case study large scale deployment of...
TRANSCRIPT
Local Authority Case Study – Large Scale
Deployment of Sensors
D. Carruthers1, D. Clarke2, K. J. Dicks3, R. A. Freshwater4, M. Jackson1,
R. L. Jones4, C. Lad1, I. Leslie5, A. J. Lewis3, H. Lloyd4, O. A. M. Popoola4,
A. Randle6, S. Ulrich4
1Cambridge Environmental Research Consultants, UK 2Cambridgeshire County Council, Cambridge, UK3Cambridge City Council, Cambridge, UK4Department of Chemistry, University of Cambridge, UK5Computer Laboratory, University of Cambridge, UK6Environmental Instruments Ltd., Stratford-upon-Avon, UK
AQE 2017, Telford, 25th May 2017 1
AQE 2017, Telford, 25th May 2017
Talk outline
• Premise of inter-comparison
• AQMesh, network deployment
• Comparison with reference instruments
• What can we do with the measurements?
- Model comparison
- Source attribution• Local emissions
• Long range sources
2
AQE 2017, Telford, 25th May 2017
Premise of inter-comparison:
• Test of ‘out of box’ AQMesh performance
• No local calibration/re-scaling
• No pan-network analysis (individual sensors)
• NO, NO2, PM2.5, PM10 only (determined by
available reference instruments)
3
Sensor Calibration Gas sensors
Comparison between Alphasense electrochemical
sensors and local (to AQMesh) reference
instrumentation to determine sensor specific calibration
parameters.
Particle sensor
OPCs co-located with a “gold standard pod” at the
AQMesh outdoor test facility to provide consistent
calibration parameters.
AQE 2017, Telford, 25th May 2017 4
CO, NO, NO2, O3, SO2,
PM1, PM2.5 and PM10
Cross network NO2 performance
(pre-deployment)
Gradients = 0.94 ± 0.07
R2 = 0.8 ± 0.11
Intercepts = 0.34 ± 0.47ppb
AQE 2017, Telford, 25th May 2017 5
Sensor-sensor
comparisons
Cross network PM2.5 performance
(pre-deployment)
Gradients = 0.98 ± 0.07
R^2 = 0.98 ± 0.17
Intercept=0.13 ± 0.3ug/m3
AQE 2017, Telford, 25th May 2017 6
Sensor-sensor
comparisons
AQE 2017, Telford, 25th May 2017 7
Cambridge deployment (20 nodes)
Northwest Cambridge(building development)
Central Cambridge(high traffic density)
South Cambridge(biomedical campus development)
Reference site(Gonville Place)
AQE 2017, Telford, 25th May 2017 8
NO2 city centre comparison (pre-ratified)
Gradient Intercept R2
0.55 (0.006) 10.0 (0.1) 0.50
• Similar features (diurnal
signatures) in both
• AQMesh significantly higher
in absolute amounts than
reference (not consistently)
AQE 2017, Telford, 25th May 2017 9
NO2 city centre comparison (pre-ratified)
Gradient Intercept R2
0.75 (0.01) 4.5 (0.2) 0.64
1.67 (0.02) 6.1 (0.16) 0.64
0.67 (0.007) 4.9 (0.19) 0.78
• Clear calibration changes
(three distinct phases) (in
which instrument……?)
AQE 2017, Telford, 25th May 2017 10
NO2 city centre comparison (ratified)
No reference data!
NO2 Gradient Intercept R2
pre 0.55 (0.01) 10.0 (0.1) 0.50
post 0.82 (0.01) 5.1 (0.13) 0.74*
*Improvement is from AURN ratification
NO Gradient Intercept R2
pre 0.85 (0.01) 4.4 (0.21) 0.49
post 0.87 (0.01) 0.63 (0.3) 0.65*
AQE 2017, Telford, 25th May 2017 11
PM2.5 city centre comparison (ratified)
• PM events captured
• Some overestimation by AQMesh
• Little/no difference on ratification
Gradient Intercept R2
pre 0.92 (0.01) -3.0 (0.15) 0.41
post 0.92 (0.01) -3.0 (0.15) 0.42
AQE 2017, Telford, 25th May 2017 12
PM10 city centre comparison (ratified)
• PM events captured by AQMesh
• Magnitudes significantly
overestimated in AQMesh
• Little/no difference on ratification
Gradient Intercept R2
pre 1.17 (0.02) -8.7 (0.51) 0.21
post 1.17 (0.02) -8.7 (0.51) 0.21
Gonville Place AQMesh reference
comparison statistics summary
Gradient Intercept R2
NO2 pre 1.07 (0.01) 10.0 (0.1) 0.50
NO pre 1.07 (0.01) 4.4 (0.21) 0.49
PM2.5 pre 0.92 (0.01) -3.0 (0.15) 0.41
PM10 pre 1.17 (0.02) -8.7 (0.51) 0.21
AQE 2017, Telford, 25th May 2017 13
Gradient Intercept R2
NO2 post 0.82 (0.01) 5.1 (0.13) 0.74
NO post 1.09 (0.01) 0.63 (0.27) 0.65
PM2.5 post 0.92 (0.01) -3.0 (0.15) 0.42
PM10 post 1.17 (0.02) -8.7 (0.51) 0.21
Pre- AURN ratification
Post- AURN ratification
Improvement is from AURN ratification
AQE 2017, Telford, 25th May 2017 14
NO2 city centre model-AQMesh
comparisons
Gradient R2
AQMesh - ADMS 0.34 0.24
AQMesh - reference 0.82 0.74
• Model captures broad behaviour
• Poorer R2 c.f. AQMesh-reference
• Poorer model-reference R2
AQE 2017, Telford, 25th May 2017 15
PM2.5 city centre model-AQMesh
comparisons
Gradient R2
AQMesh - ADMS 0.265 0.03
AQMesh - reference 0.92 0.42
• Model captures magnitudes of
events but not timing…..
• Significantly poorer R2 c.f. AQMesh -
reference
AQE 2017, Telford, 25th May 2017 16
Average NOx model-AQMesh comparisons
– all stations
• Model ~ captures AQMesh spatial gradients
• Local (spatially heterogeneous) sources
AQE 2017, Telford, 25th May 2017 17
Average PM model-AQMesh comparisons
– all stations
• Model ~ captures (lack of) spatial gradients
• Averages dominated by non-local sources
AQE 2017, Telford, 25th May 2017 18
What does this mean?
Spatial gradient implies local
source in city…..
Local intervention possible
No spatial gradient implies
mainly regional source…..
Regional intervention
required
Polar bivariate plots (© OpenAir!): source
apportionment
19AQE 2017, Telford, 25th May 2017
NO2: significant differences– road sources (1)– local sources (2)– non local sources (3)
1
1
1
2
3
3
3
AQE 2017, Telford, 25th May 2017 20
3
3
3
PM10: broadly similar – longer range transport?– building site influences (1)
1
11
AQE 2017, Telford, 25th May 2017 21
AQE 2017, Telford, 25th May 2017 22
Summary
• AQMesh sensor performance ‘out of box’ performance:• Sensor- sensor reproducibility very good.
• NO/NO2/NOx inter-comparison with ratified measurements good.
• PM measurements capture events, some overestimation (esp. PM10).
• Ratification process produces some rather surprising results……
• AQMesh captures spatial gradients well• Spatial gradients for NOx (local emissions)
• Little spatial gradients for PM (long range transport)
Important for understanding role of policy interventions
• Source attribution
- Hotspot detection
- NO2 short term exceedences
- PM emissions building works
AQMesh performs well: hotspots, A/Q monitoring/control
D. Carruthers1, D. Clarke2, K. J. Dicks3, R. A. Freshwater4, M. Jackson1,
R. L. Jones4, C. Lad1, I. Leslie5, A. J. Lewis3, H. Lloyd4, O. A. M. Popoola4,
A. Randle6, J. Stocker1, T. Townend6, S. Ulrich4
1Cambridge Environmental Research Consultants, UK 2Cambridgeshire County Council, Cambridge, UK3Cambridge City Council, Cambridge, UK4Department of Chemistry, University of Cambridge, UK5Computer Laboratory, University of Cambridge, UK6Environmental Instruments Ltd., Stratford-upon-Avon, UK
AQE 2017, Telford, 25th May 2017 23