technical evaluation of sensor technology (test) program...technical evaluation of sensor technology...
TRANSCRIPT
Technical Evaluation of Sensor Technology (TEST) Program
PurpleAir PA-II Sensor
2019 – 1st Quarter
San Joaquin Valley Air Pollution Control District PurpleAir PA-II Sensor
Technical Evaluation of Sensor Technology (TEST) Program 2019 – 1st Quarter
2
Introduction and Sensor Profile
This analysis report is focused on assessing the performance of the PurpleAir PA-II sensor as a
part of the District’s Technical Evaluation of Sensor Technology (TEST) Program. The PurpleAir
PA-II sensor uses an optical laser-based particle counting methodology to estimate the mass of
varying diameters of particulate matter, including PM1, PM2.5, and PM10. The PA-II sensor
also measures temperature, pressure, and relative humidity.
Background and Approach of Evaluation Test
In November of 2017, NASA began an air quality study to compare the performance of
PurpleAir sensors to regulatory PM2.5 analyzers. The study is focused on the conditions in the
San Joaquin Valley and is based at the ARB air monitoring sites of Fresno-Garland, Visalia-
Church, Modesto-14th St., and Bakersfield-California. The data sets analyzed for this report
compare PM2.5 data collected from PurpleAir sensors and Federal Equivalent Method (FEM)
monitors that are collocated at the ARB air monitoring sites listed above. The scatter plots and
time series graphs below show how the datasets compare for both hourly values and the 24-
hour average.
Overview of Analysis Findings from Current Period
This assessment compares the Purple Air PA-II sensor performance against the MetOne BAM-
1020 continuous particulate monitor. The analysis for this report covers the time period of
January 2019 through March 2019 (2019 – 1st quarter). During this this period, hourly data was
removed from the calculation of bias and average concentrations when either the Purple Air
sensor or regulatory monitor did not have a valid sample. For the 24 hour average line graphs,
all available data is shown for each collocated analyzer and sensor.
The 1st quarter of 2019 was dominated by an active weather pattern where good dispersion
conditions brought lower particulate concentrations. Brief periods of high pressure resided
over the Valley causing stagnation and an increase in PM2.5 concentrations. The strongest
ridge of high pressure over the Valley during these three months was from January 22 until
January 30. PM2.5 concentrations for the Purple Air sensors and regulatory monitors were
highest during this late January period of higher stability. As the plots below show, Purple Air
data was biased higher than the District’s regulatory data during this period for all sites.
San Joaquin Valley Air Pollution Control District PurpleAir PA-II Sensor
Technical Evaluation of Sensor Technology (TEST) Program 2019 – 1st Quarter
3
Site Specific Analysis of PurpleAir PA-II Sensor Performance
Fresno-Garland
For the 24-hour average, PurpleAir data had a 9.0 µg/m3 high bias during the January 1st, 2019
through March 31st, 2019 period. For the hourly average, PurpleAir data had a high bias of
8.5 µg/m3 over the same period.
y = 1.8311x - 0.3831
R² = 0.8724
0
10
20
30
40
50
60
70
80
0 10 20 30 40 50 60 70 80
Pu
rple
Air
MetOne BAM FEM
Fresno 24-hour Average Comparison
y = 1.5605x + 2.6453
R² = 0.8327
0
20
40
60
80
100
120
140
0 20 40 60 80 100 120 140
Pu
rple
Air
MetOne BAM FEM
Fresno Hourly Average Comparison
0
10
20
30
40
50
60
70
80
PM
2.5
(µ
g/m
3)
Fresno 24-hour PM2.5 Average FEM vs PurpleAir Q1 2019
Fresno FEM Fresno Purple Air
San Joaquin Valley Air Pollution Control District PurpleAir PA-II Sensor
Technical Evaluation of Sensor Technology (TEST) Program 2019 – 1st Quarter
4
Visalia-Church
For the 24-hour average, PurpleAir data had an 8.1 µg/m3 high bias during the January 1st, 2019
through March 31st, 2019 period. For the hourly average, PurpleAir data had an 8.2 µg/m3 high
bias over the same period.
y = 1.7873x - 0.8746
R² = 0.9068
0
10
20
30
40
50
60
70
80
0 20 40 60 80
Pu
rple
Air
MetOne BAM FEM
Visalia 24-hour Average Comparison
y = 1.6572x + 0.8836
R² = 0.888
0
20
40
60
80
100
120
140
0 20 40 60 80 100 120 140
Pu
rple
Air
MetOne BAM FEM
Visalia Hourly Average Comparison
0
10
20
30
40
50
60
70
80
PM
2.5
(µ
g/m
3)
Visalia 24-hour PM2.5 Average FEM vs PurpleAir Q1 2019
Visalia FEM Visalia Purple Air
San Joaquin Valley Air Pollution Control District PurpleAir PA-II Sensor
Technical Evaluation of Sensor Technology (TEST) Program 2019 – 1st Quarter
5
Bakersfield-California
For the 24-hour average, PurpleAir data had a 5.2 µg/m3 high bias during the January 1st, 2019
through March 31st, 2019 period. For the hourly average, PurpleAir data also had a 5.2 µg/m3
high bias over the same period.
y = 1.4722x - 0.1737
R² = 0.8929
0
10
20
30
40
50
60
70
80
0 20 40 60 80
Pu
rple
Air
MetOne BAM FEM
Bakersfield 24-hour Average Comparison
y = 1.3383x + 1.5102
R² = 0.8412
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50 60 70 80 90 100
Pu
rple
Air
MetOne BAM FEM
Bakersfield Hourly Average Comparison
0
10
20
30
40
50
60
70
80
PM
2.5
(µ
g/m
3)
Bakersfield 24-hour average FEM vs PurpleAir Q1 2019
Bakersfield FEM Bakersfield Purple Air
San Joaquin Valley Air Pollution Control District PurpleAir PA-II Sensor
Technical Evaluation of Sensor Technology (TEST) Program 2019 – 1st Quarter
6
Modesto-14th St.
For the 24-hour average, PurpleAir data had an 8.6 µg/m3 high bias during the January 1st, 2019
through March 31st, 2019 period. For the hourly average, PurpleAir data had a 9.2 µg/m3 high
bias over the same period.
y = 2.0617x - 1.6213
R² = 0.8436
0
10
20
30
40
50
60
70
80
0 10 20 30 40 50 60 70 80
Pu
rple
Air
MetOne BAM FEM
Modesto 24-hour Average Comparison
y = 1.7959x + 1.7012
R² = 0.8838
0
10
20
30
40
50
60
70
80
90
100
110
0 10 20 30 40 50 60 70 80 90 100 110
Pu
rple
Air
MetOne BAM FEM
Modesto Hourly Average Comparison
0
10
20
30
40
50
60
70
80
PM
2.5
(µ
g/m
3)
Modesto 24-hour average FEM vs PurpleAir Q1 2019
Modesto FEM Modesto Purple Air
San Joaquin Valley Air Pollution Control District PurpleAir PA-II Sensor
Technical Evaluation of Sensor Technology (TEST) Program 2019 – 1st Quarter
7
Statistical Summary
The following table provides a statistical summary of the data collected during the analysis
period of this report.
Statistic Fresno-Garland Visalia-Church Bakersfield-Cal Modesto
FEM Avg 24-hr 11.2 11.8 11.4 8.6
Sensor Avg 24-hr 20.2 19.9 16.6 17.2
FEM Max 1-hr 120.9 83.0 89.0 72.0
Sensor Max 1-hr 117.0 108.3 91.5 105.7
FEM Max 24-hr 47.0 45.0 53.0 34.0
Sensor Max 24-hr 72.6 78.0 73.5 70.3
1-hr R2 0.8327 0.888 0.8412 0.8838
1-hr Slope 1.5605 1.6527 1.3383 1.7595
1-hr Intercept 2.6453 0.8836 1.5102 1.7012
24-hr R2 0.8724 0.9068 0.8929 0.8436
24-hr Slope 1.8311 1.7873 1.4722 2.0617
24-hr Intercept -0.3831 -0.8746 -0.1737 -1.6213