*•	*
®!
"V o<"°
PRO"^
Review of Sunset Organic and Elemental
Carbon (OC and EC) Measurements During
EPA's Sunset Carbon Evaluation Project

-------

-------
EPA-454/R-19-005
May 2019
Review of Sunset Organic and Elemental Carbon (OC and EC) Measurements During EPA's
Sunset Carbon Evaluation Project
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Air Quality Assessment Division
Research Triangle Park, NC

-------
This document contains blank pages to accommodate two-sided printing.

-------
Review of Sunset Organic and Elemental
Carbon (OC and EC) Measurements
During EPA's Sunset Carbon Evaluation
Project
Prepared by
Sonoma Technology, Inc.
1450 N. McDowell Blvd., Suite 200
Petaluma, CA 94954-6515
Prepared for
U.S. Environmental Protection Agency
Office of Air Quality Planning and
Standards
Research Triangle Park, NC 27711
STI Final
Report
STI-915313-6843
November 12, 2018

-------

-------
Contents
Contents
Figures	iv
Tables	iv
Acknowledgments	v
1,	Executive Summary[[[ 1
2.	Introduction	3
3.	Methods [[[5
3.1	Monitoring Site Locations	5
3.2	Sunset OC/EC	6
3.3	CSN URG 3000N Sampler and Lab Analysis	7
3.4	Aethalometer	8
3.5	Data Processing and Quality Control	8
3.6	Comparison of Sunset Data to CSN and Aethalometer Data	10
4,	Results	11
4.1	Sunset OC Bias and Detection Limit Calculations	11
4.2	Sunset and CSN OC	11
4.3	Sunset and CSN Thermal EC and Sunset Optical EC	16
4.4	Sunset OptEC and Aethalometer BC	19
4.5	Precision from Collocated Measurements in St. Louis	22
4.6	Diurnal Patterns	23
elusions	25

-------
Figures and Tables

Figure 1. Box plot of OC concentrations at each site via Sunset and CSN measurements	13
Figure 2. Scatter plot of Sunset and CSN OC concentrations, colored by site; the linear
regression equation written in black is for all data at all sites	14
Figure 3. Box plot of Sunset EC, Sunset OptEC, and CSN EC concentrations at each site	18
Figure 4. Scatter plot of CSN EC concentrations with Sunset EC (left) and Sunset OptEC (right),
colored by site; the linear regression equation written in black is for all data at all sites	18
Figure 5, Box plot of 24-hour average Sunset OptEC and Aethalometer BC	21
Figure 6. 24-hour averaged Sunset OptEC and Aethalometer BC, colored by site; the linear
regression equation written in black is for all data at all sites	21
Figure 7. Collocated 24-hour OC (left) and OptEC (right) measurements at St. Louis during
August 11, 2016, through January 11, 2017	22
Figure 8. Average hourly OC and EC concentrations on weekdays and weekends at each site	24
Tables
Table 1. Summary of measurements by site; date range indicates the time frame when Sunset
data were available in AQS	5
Table 2, Available collocated 24-hour Sunset and CSN measurements by site	10
Table 3 Calculations of CV (%), bias (%), and detection limit (ug/m3) based on sucrose injection
results for two Sunset OC/EC instruments at EPA OAQPS	11
Table 4, Summary of Sunset and CSN OC measurements and comparison statistics	15
Table 5. Summary of Sunset and CSN EC measurements and comparison statistics	17
Table fa. Summary of Sunset OptEC and Aethalometer BC measurements and comparison
statistics	20

-------
Acknowledgments
Acknowledgments
This project was made possible by the support and data collection from staff at multiple agencies:
Dustin Kuebler and Will Wetherell from the Missouri Department of Natural Resources (DNR); Holly
Landuyt and staff at the Texas Commission on Environmental Quality (TCEQ); Yousef Hameed,
Kristopher Simonian and staff at Clark County Department of Air Quality in Nevada; Pinaki Banerjee
and staff at Cook County, Illinois; Richard Tun, Bob Day and staff at Washington D.C. Department of
Energy and Environment; and Andrea Polidori and staff at California's South Coast Air Quality
Management District (AQMD). Josh Dixon from Sunset Labs helped set up instrumentation, answered
questions from staff running instrumentation, and provided troubleshooting and replacement parts
during operations. Josh Dixon and Bob Cary from Sunset Labs provided comments on the draft
report and data analysis.
V

-------

-------
1. Executive Summary
Mr- .rW^y
As part of an EPA project to evaluate the feasibility of the Sunset Semi-Continuous Organic and
Elemental Carbon (OC/EC) monitor, EPA sponsored the deployment of this monitor by local air
quality agencies at Chicago, Illinois; Houston, Texas; Las Vegas, Nevada; St. Louis, Missouri; Rubidoux,
California; and Washington, D.C. Sunset monitors were collocated with existing 24-hr measurements
of OC and EC via filter sampling as part of the Chemical Speciation Network (CSN); at Houston, St.
Louis, and Washington, they were also collocated with Aethalometer instruments, which measure
black carbon (BC). Sunset data were compared to CSN and Aethalometer data to assess whether the
Sunset monitors could be deployed in lieu of making filter measurements at routine monitoring
stations. In addition, two Sunset monitors were collocated at EPA OAQPS' on-site monitoring station
to assess detection limits and precision via injections of known sucrose standards.
Using injections of known sucrose standards, the coefficient of variation (CV) and bias values of the
Sunset were measured; the values ranged between 5%-6% for bias, and 6%-8% for CV, which met
the data quality objectives of 15%. The calculated volumetric detection limit was between 1.4 to
1.5 fig/m3, using a one-hour cycle with 47 minute collection at 8 Ipm. Agencies encountered
significant operational issues with the Sunset monitors, substantially reducing the number of valid
data points for the multi-year deployment. After screening for these issues, and excluding Las Vegas
results that were suspect, we found that Sunset OC generally compared well with the CSN OC
(r2=0.73 across five sites); the Sunset/CSN OC ratio was, on average, 1.06, with a range among sites
of 0.96 to 1.12. The Sunset instrument measures thermal EC (referred to as "EC") and optical EC
("OptEC"). CSN measurements are thermal EC. Sunset thermal EC and CSN EC did not compare as
well, with an overall r2 of 0.22, in part because 26% of the hourly Sunset EC measurements were
below the detection limit. Sunset Optical EC had a much better correlation to CSN EC (r2=0.67 across
all sites), with an average Sunset/CSN ratio of 0.90 (range of 0.7 to 1.08). There was also a high
correlation of Sunset Optical EC with Aethalometer BC (r2=0.77 across all sites), though with a larger
bias (average Sunset/Aethalometer ratio of 0.56). There was no systematic difference among the
Sunset Optical EC and Aethalometer BC measurements by site location, i.e., Sunset Optical EC was
consistently lower than Aethalometer BC at all three sites, with no significant seasonal variation.
Somewhat surprisingly, the diurnal pattern of OC was fairly invariant, while EC had a morning peak
across all sites. Overall, operational issues with the Sunset monitors were persistent at all sites, but
when the instruments were operating well, collected data were comparable to results from CSN.
Implications:
•	The Sunset instrument and its software was not robust enough for routine deployment and
operation by state/local air quality agencies at the beginning of this study, but a number of
improvements have since been made to address the issues encountered in this study.
•	When the Sunset instrument was working well, OC and OptEC data were comparable to CSN
OC and EC.
1

-------

-------
2 Introduction
Carbonaceous aerosol is a significant, and often the largest, component of fine particulate matter
less than 2.5 microns in diameter (PM25) in many areas of the United States. It is composed of
organic and elemental carbon (OC, EC) (Jacobson et al., 2000), but its composition, sources, and
spatiotemporal variations are not well characterized (Jimenez et al., 2009). OC comprises thousands
of individual molecules that can be directly emitted as primary emissions or can be formed in the
atmosphere from semi-volatile and gaseous precursors over the course of minutes to days. EC is
directly emitted from combustion processes, such as from mobile sources or from biomass burning.
While it is well established that elevated PM25 levels are associated with many health effects, such as
respiratory and cardiac disease, the complex interaction of specific health effects from individual
compounds or PM25 components such as OC and EC is not well understood.
EPA monitors OC and EC in urban areas as part of the Chemical Speciation Network (CSN), where
over 100 monitors across the United States collect filters that are subsequently analyzed for OC and
EC on a routine basis. Such measurements have been collected for over 15 years, offering an
opportunity to evaluate long-term temporal and spatial trends. As continuous monitoring
technology has advanced, EPA and other air monitoring agencies have begun to assess whether
continuous monitoring technologies could feasibly be used to reduce the frequency and amount of
filter-based measurements. If continuous monitors were used to continue the long-term monitoring,
they could provide a significant improvement to the data collected in three main ways: (1) provide
data every day, rather than on the l-in-3 or l-in-6 day schedule typical for filter measurements; (2)
provide hourly data, so that data analyses such as wind direction and diurnal analysis would become
feasible; and (3) significantly reduce the cost of sample preparation, shipping, and laboratory
analysis. The Sunset OC/EC instrument provides integrated measurements of OC and EC on a
customizable sampling time (such as hourly or 2-hour intervals) and flow rate (2-9 Ipm) via a
thermal method similar to that used in CSN, as well as an optical EC (OptEC) measurement that is
based on transmission of 660 nm wavelength light through the filter.
The Sunset OC/EC instrument has been widely used in the United States and throughout the world
(Snyder and Schauer, 2007; Bae et al., 2004a; Jeong et al., 2004). OC measurements have typically
been comparable to other measurements of carbonaceous aerosol, such as from the Aerodyne
Aerosol Mass Spectrometer (AMS). At a near-road site in Las Vegas, Nevada, Brown et al. (2013)
found that AMS-derived OC and Sunset OC were very consistent, with small bias (r2 of 0.89, slope of
0.91). In Hong Kong, Lee et al. (2013) also found good agreement between Sunset and AMS
measurements (r2 of 0.87 and slope of 0.88). Other studies had more variation between AMS and
Sunset measurements, for example in Riverside (r2 of 0.53) (Docherty et al., 2011), Tokyo (r2 range of
0.67-0.83 in two seasons) (Takegawa et al., 2005), and Pittsburgh (r2 of 0.88) (Zhang et al., 2005). In
Riverside, Snyder and Schauer (2007) found the Sunset measurements compared well with filter
measurements (r2 of 0.90 and slope of 1.11). At Atlanta, lvalues between the Sunset and the Aerosol
3

-------
2 Introduction
Chemical Speciation Monitor (ACSM) were between 0.86-0.92 in summer and fall (Budisulistiorini et
al., 2014).
EC from Sunset and BC from Aethalometer instruments have also been compared. In Prague, Zfkova
et al. (2016) found that Sunset OptEC and BC were fairly comparable (slope of 0.77 and r2 of 0.99) in
winter. In New York, Rattigan et al. (2010) found a consistent seasonal difference in BC/EC ratio over
the course of three years of measurements, with a ratio of 1.4 in October-March and ratio of 2.0 in
April-September. They also found an average OptEC/EC ratio of 0.88 in October-March and 1.04 in
April-September. Throughout the year, there was a high correlation of BC with EC, with a monthly
range of 0.82-0.96. In Ontario, collocated EC and BC measurements also had high correlation (r2 of
0.85 and 0.77 at two sites), with a BC/EC ratio of 1.7 at both sites (Healy et al., 2017).
To evaluate the utility of the Sunset OC/EC instrument as part of the CSN, EPA sponsored the
deployment of this monitor by local air quality agencies at CSN sites in Chicago, Houston, Las Vegas,
St. Louis, Rubidoux, and Washington, D.C. Monitors were operated at these locations, as well as at
EPA in Raleigh, North Carolina, for varying lengths of time during 2012-2017. The primary objectives
of the study were to evaluate Sunset instrument performance in various locations and conditions;
determine how well the Sunset measurements compare with the CSN and Aethalometer
measurements, where available; assess precision and detection limits via injections of a known
standard amount of sucrose solution; and determine whether integration of the Sunset OC/EC
instrument across a larger number of sites is appropriate for long-term monitoring in the CSN.
Results from the study are presented in this report. Appendix A documents the operational issues
encountered by agencies operating the instrument, and the actions Sunset Labs has taken to address
these issues. Appendix B provides additional statistics comparing the Sunset data to CSN and
Aethalometer data, plus figures showing the ratio between Sunset and CSN or Aethalometer data,
diurnal patterns, and time series of data as they exist in the EPA's Air Quality System (AQS) and after
additional quality control was done.
4

-------
3. Methods
3. Methods
3.1 Monitoring Site Locations
Six locations at existing CSN sites were used in this project: Chicago (Com Ed site in Lawndale,
AQS ID 17-031-0076); Houston (Deer Park, AQS ID 48-201-1039); Las Vegas (East Las Vegas, AQS
ID 32-003-0540); Rubidoux (Rubidoux, AQS ID 06-065-8001); St Louis (Blair Street, AQS ID
29-510-0085); and Washington, DC. (McMillan Reservoir, AQS ID 11-001-0043). Two Sunset
instruments were operated at St. Louis from August 11, 2016, through January 11, 2017. Table 1
summarizes the site locations and measurements. CSN measurements were collected every third day.
Sunset, Aethalometer, and CSN data were acquired from EPA's AQS in summer 2017.
Table 1. Summary of measurements by site; date range indicates the time frame when Sunset
data were available in AQS.
City
AQS ID
Site
Operator
Measurements
Dates with
Sunset Data
. Cook County Dept.
Com Ed, rn • I ciki 5/1/14-
Chicago 17-031-0076 , . . of Environmental Sunset, CSN
Lawndale _ x . 12/31/15
Control
Houston
48-201-1039
Deer Park
Texas Commission
on Environmental
Quality (TCEQ)
Sunset, CSN,
Aethalometer AE21
8/2/13-
12/31/16
		 East Las 8/15/12-
Las Vegas 32-003-0540 .. Clark County Sunset, CSN
3 Vegas 1 12/31/14
Los Angeles
06-065-8001
Rubidoux
South Coast Air
Quality
Management
District
Sunset, CSN
12/17/13-
10/14/15
C+ 1 -mnnnnoir di- c+ + Missouri Dept. of Sunset, CSN,
St. Louis 29-210-0085 Blair Street K1 , „ » , , 5/7/13-3/30/17
Natural Resources Aethalometer AE33
Washington,
D.C.
11-001-0043
McMillan
Reservoir
District Dept. of the
Environment
Sunset, CSN,
Aethalometer AE21
10/7/12-
8/13/16
5

-------
3. Methods
3.2
In this application, the Sunset OC/EC instrument used a thermal optical method similar to NIOSH
5040 (Chow et al., 2001; 2007; Bauer et al., 2009; Park et al., 2005; Bae et al., 2004b; Sin et al., 2004;
Birch and Cary, 1996). Other methods, such as IMPROVE-A by TOT, could also be used. Aerosol is
drawn through a PM2.5 cyclone inlet with a carbon denuder and deposited for 47 minutes at a flow
rate of 8 Ipm on a quartz fiber filter located in an oven chamber. The collected aerosol is then heated
off of the filter during an 8-minute cycle by heating the filter to 850°C for 5 minutes to quantify OC.
As the evolved carbon flows through the manganese oxide (Mn02) oven, it is converted to carbon
dioxide (C02) gas, which is carried in a helium stream and measured directly by a self-contained non-
dispersive infrared (NDIR) detector system. Next, an oxidizing carrier gas (helium with 2% oxygen
[02]) is introduced at 850°C for 3 minutes to quantify EC, where the EC is detected (similar to the way
OC was detected). The remaining 5 minutes is used for cooling down the oven. During the filter
heating, carbonaceous material evolves off the filter as C02, which is quantified using an NDIR
detector. EC is determined as any carbon evolved off the filter after the introduction of He/02 once
the laser-monitored filter absorbance matches the initial absorbance measured when the filter was
first heated. After each hourly analytical cycle, calibration gas of 5% CH4 with He flushes the system.
Manufacturer-specified detection limits are 0.4 |jg C/m3 for OC and 0.2 |jgC/m3 for EC.
Where reported by the monitoring agency, both thermal EC (referred to as "EC") and optical EC
(OptEC) comparisons are provided here. The OptEC is a measurement of transmittance through the
filter at a wavelength of 660 nm prior to the thermal analysis, measuring the amount of absorbance
in the sp2 bonds of graphitic carbon. Since the measurements of both OptEC from the Sunset and BC
from the Aethalometer are based on optical absorbance methods, we compared how consistent
measurements from these techniques were to each other and to the thermal EC from CSN. At
Chicago, no OptEC was reported. At St. Louis, thermal EC was not reported after 2014 because the
instrument needed very frequent filter replacements; this is likely due to high loadings of metal
oxides at the monitoring site. Once only OptEC was measured, the instrument filter did not have to
be replaced as often, so only OptEC was reported for the majority of the study.
Two Sunset OC/EC instruments were operated at the EPA site to test instrument setup, and quantify
bias, precision, and detection limits using injections of a sucrose standard; the equations used to
quantify bias, following EPA guidance, are shown below (Camalier et al., 2007). A known amount
(10 uL or 5 uL) of 99.5% sucrose from Sigma Aldrich (product #S9378) was injected into each
instrument intermittently over the course of two years. The absolute percent difference (d) between
the observed response from the instrument and the injected amount of carbon was then calculated.
The coefficient of variation upper bound (90th percentile) was calculated as the precision estimate:
CV =
M
X
n(n — 1)
n = 1
X2o.l,n-l
6

-------
3. Methods
Where X20.i,n-i is the highest 10th percentile of a chi-squared distribution with n-1 degrees of
freedom. Bias is calculated as the upper bound of the mean absolute value of the percent differences
d across all d/s, from the mean of absolute values of all ds (AB) and the standard deviation of the
absolute values of all ds (AS):
AS
\bias\ = AB + t095n_1 x —
yn
n
AB = —xV Id/1
n Z_i
i=1
•J	u(n - 1)
In addition, the instrument response to clean, blank quartz fiber filters was used to calculate the
detection limit The detection is calculated following 40 CFR Part 136, Appendix B:
MDL = X + t(>i-l,l-«=0.99) x S
Where X is the mean of replicate method blank results, t^n_i,i_k=o.99) shows the Student's t value at a
99% confidence level, with n-1 degrees of freedom, and S is the standard deviation of the blank
samples.
33 CSN URG 3000N Sampler and Lab Analysis
As part of routine measurements in the CSN, quartz fiber filters are prepared and shipped to
monitoring sites. Filters are pre-baked to remove organic vapor and residue. A URG 3000N sampler is
used to collect aerosol on filters, but unlike the Sunset instrument, no denuder is used. Aerosol is
sampled at a flow rate of 22 Ipm through a PM2.5 inlet for 24 hours, every third or sixth day. OC and
EC are then determined via the IMPROVE_A temperature protocol (Chow et al., 2007) by Desert
Research Institute (DRI) using a DRI Model 2001 carbon analyzer. In this protocol, a 0.5 cm2 circular
segment of the filter is removed, and aerosol are thermally evolved off of the filter (similar to the
process for the Sunset instrument), where OC is determined under a non-oxidizing atmosphere with
He gas, and then EC is found using a mix of 98% He and 2% 02. Carbonaceous aerosol is volatilized
off the filter and converted to C02 in an Mn02 oxidizer, and then reduced to methane via a nickel
catalyst and quantified as methane with a flame-ionization detector (FID). For OC, the temperature is
ramped to four temperature plateaus at 140°C, 280°C, 480°C, and 580°C, where the temperature is
held constant at each plateau until the response in the FID has returned to baseline for 30 seconds
(i.e., until there is no more carbonaceous material being volatilized from the filter at that
temperature). The He/02 atmosphere is then introduced while the temperature is held at 580°C in
order to initially quantify pyrolyzed organic carbon (OP), and then the temperature is increased to
7

-------
3. Methods
740°C and 840°C. The sum of the carbon evolved in the He atmosphere plus the OP is equal to total
OC, while the sum of the carbon evolved under the He/02 atmosphere minus the OP is equal to total
EC. As reported in EPA's 2014 Environmental Technology Verification Report EPA/600/R-14/308, the
precision of this instrument based on replicate analyses is greater than 15%, and indicates "a lower
degree of data quality than desired."
3.4
A Magee Scientific Aethalometer was operated at Washington, D.C. (AE21 instrument), St. Louis (AE33
instrument), and Houston (AE21 instrument). The Aethalometer measures BC via an optical method,
instead of the thermal method used by the Sunset and CSN (Allen et al., 1999; Weingartner et al.,
2003). Aerosol is sampled through a BGI model SCC PM2.5 cyclone inlet at 5 Ipm and deposited on a
filter tape. Every 5 minutes, the Aethalometer measures the light attenuation at 880 nm through the
filter tape, and is converted into a BC concentration by assuming an attenuation cross-section of
16.6 m2/g. The measured BC is subtracted from the prior measurement of BC to determine the BC
collected during the 5 minutes of sampling. No post-processing of the raw data was done. For
example, when the tape on which aerosol is deposited reaches a given saturation point, the tape
advances, so that aerosol is now deposited on a new section of tape. When this occurs, there can be
an artifact in the data stream that is not automatically accounted for or corrected without post-
processing (Drinovec et al., 2015; Jimenez et al., 2007; Weingartner et al., 2003). The AE33 has two
built-in light sources to automatically correct for this (Drinovec et al., 2015), but no correction was
made for the AE21 data.
3.5 Data Processing and Quality Control
Sunset and Aethalometer data were reported in both local conditions (LC) and standard temperature
and pressure (STP). STP data were converted to LC using local meteorological data; all data reported
here are in LC. Daily 24-hr averages were calculated from hourly Aethalometer and Sunset data
where at least 75% of the hourly data were available.
During the project, the agencies operating the Sunset instrument encountered instrument
component malfunctions such as cracked ovens, NDIR detector failure, heating element failure, and
pump failure. These issues were not easily diagnosed during operations and led to shifts in baselines
and other data issues that made the data unusable for this analysis. The oven and NDIR problems
were typically not found early on, since at the time there was no routine output from the instrument
alerting users to these issues, or readily available data from CSN for comparison; this resulted in
multiple weeks of data being removed prior to analysis. Data were visually inspected on time series
to identify periods where there were sudden shifts in concentration, small quantities of data between
data gaps, and unusual outliers.
8

-------
3. Methods
At St. Louis, starting in January 2015, a filter was stuck, and then during March 2015-January 2016,
operators suspected contamination, adjusted the thermocouple, and installed a new photodetector.
However, the new photodetector was not working correctly, and data did not return to "normal" until
after the oven was replaced in January 2016. There were additional issues with keeping the flow
steady in June through July 2016. At Washington, D.C., there were periods where OC or EC
concentrations were greater than 100 ng/m3, even though collocated PM2 5 concentrations were low;
these data were excluded from analysis here. Prior to May 2014, OC was not reported at this site, so
no data were included here for analysis. Only data starting June 2014 were included for analysis,
since there were operational issues prior to this time. Data in June 2015 and February-March 2016
were also excluded from analysis because of operational issues associated with a software update in
June and a heating coil malfunction at the end of January 2016, which was not fixed until the end of
March 2016. Time series graphics of all measurements at each site are provided in Appendix B, and
completeness for each parameter is detailed in Table Bl.
At Chicago, there was a significant shift in the lowest reported OC values beginning at the end of
December 2014, so only data prior to this shift are included here, and only when OC and EC are both
reported. Data after January 2015 were excluded from analysis since there was a clear gradual rise in
baseline of OC due to degradation of the NDIR. In Houston, there were multiple gaps in the data as
NDIR detectors and ovens had to be replaced. Data prior to December 2014 were excluded since
older software was used to determine OC/EC and OptEC, the NDIR malfunctioned and was replaced
twice, the oven thermocouple malfunctioned and was replaced, there were leaks, and the instrument
was sent back to Sunset twice for maintenance. Data during May-August 2015 and July-August 2016
had an unusual shift in OC, and EC was near zero; both of these issues occurred when there were
leaks in the sampling line, and neither was seen in the collocated CSN measurements.
Data in Las Vegas were intermittent during the course of operations, resulting in many anomalous
data points and shifts in data. Only data with multiple weeks of consistent measurements were
included for analysis. For example, in November 2012, OC was consistently reported as less than
0.5 jjgC/m3, and in July and October 2014, the NDIR and heater coils broke and needed to be
replaced multiple times, there was vandalism at the site so the shelter air conditioning unit was not
working, and instrument software was not routinely updated. The period of December 2012 to May
2013 was the most consistent and complete period of data, and is used here. Given the operational
issues at this site, results are not expected to be representative of optimal instrument operations or
of other locations, but they are included for completeness. At Rubidoux, there were two periods
where there was a significant shift in the lowest reported OC values (May-September 2014 and
March-October 2015), when operators found leaks in the sampling line and the oven had to be
replaced twice. These data were screened out from further analysis; time series graphs showing the
data as reported in AQS and after the subsequent QC described above are provided in Appendix B.
This QC process substantially reduced the number of valid Sunset data points compared to the
number reported in AQS. Data availability and summary statistics after data processing and
validation is available in Appendix B. After QC, there was a range of coincident, collocated 24-hr
9

-------
3. Methods
Sunset and CSN values available for comparison, which is detailed in Table 2. Since there were a
number of operational issues throughout the project, the quality of data varies by site. For example,
data recovery was low at Las Vegas, Chicago, and Rubidoux, so results for these sites are likely less
representative than results for Houston, St Louis, and Washington, D.C. While these latter three sites
also had operational issues—in particular, problems with broken ovens and NDIRs not being
detected—sufficient data were collected for comparison to CSN data.
Table 2. Available collocated 24-hour Sunset and CSN measurements by site.
Site
N Collocated OC
N Collocated EC
Date Range
Measurements
Measurements
Chicago
57
60
5/2/2014-12/31/2014
Houston
154
154 OptEC, 152 EC
12/13/2014 -
10/15/2016
Las Vegas
53
53
12/11/2012 - 9/20/2014
Rubidoux
75
75
12/18/2013 - 3/10/2015
St. Louis
198
198 OptEC, 63 EC
9/22/2013 - 1/10/2017
Washington, D.C.
208
211 (OptEC), 208 EC
6/1/2014 - 8/10/2016
3.6 Comparison of Sunset Data to CSN and
Aethalometer Data
Detailed measurement quality objectives (MQOs) for comparing Sunset data to CSN and
Aethalometer data were discussed in the Project QAPP (U.S. Environmental Protection Agency, 2011).
These MQOs include comparison via linear least squares regression, comparison of means including
variability, and ratio of the means. In addition, collocated measurements at St. Louis were used to
estimate precision, which is the measure of agreement among repeated measurements of the same
property under identical, or substantially similar, conditions. The MQO for ratio-of-means in CSN
measurements was set as 1 ±0.15, where the coefficient of variation (CV) is used as the measure of
variability. A recent assessment of collocated CSN data at six sites results in a CV of 8.8% (Rice and
Landis, 2016). For all statistics, we report values when the two measurements we are comparing
occurred on the same day; e.g., for Sunset and CSN OC, only those days with measurements of both,
and for Sunset and Aethalometer, only those days with measurements of both. Thus, there will be
some differences in reported values, especially when comparing Sunset EC to either Aethalometer BC
or CSN EC, since there are many more days with Aethalometer data than with CSN data. In addition
to comparing means within the CV, we also report whether concentrations between two
measurements were statistically significant based on a Student's t-test.
10

-------
4. Results
4. Results
4.1 Sunset OC Bias and Detection Limit Calculations
Results of CV, bias and detection limit calculations using data from sucrose injections for the two
Sunset OC/EC instruments at EPA site are shown in Table 3. The CV and bias values meet the data
quality objectives of 15%, ranging between 5% to 6% between the two instruments for bias, and 6%-
8% for CV, which is similar to the 8.8% CV across six collocated CSN OC TOR measurements (Gantt et
al., 2017). The bias estimates are similar to prior estimates from collocated Sunset OC/EC instrument
data, where Bauer et al. (2009) estimated bias of 5.3%-5.6% for OC. The calculated detection limit was
between 1.4 to 1.5 ng/m3, which was higher than the estimate of 0.2 ng/m3 in Bauer et al. and Sciare
et al. (2011), and higher than the estimated MDL from CSN of 0.2 ng/m3 in Sciare et al. The difference
in detection limit calculation methodologies may explain part of the differences among results. The
CSN results are calculated as three times the standard deviation of 50 field blanks, while Bauer et al.
used a limit of detection calculation as the 95th percentile of the standard deviation across zero air
measurements, and Sciare et al. took the average value across 7 blank filter samples. The detection
limit found here is similar to an estimated detection limit of 2.0 ng/m3 from Zheng et al. (2014), who
evaluated how results varied under different operational protocols.
Table 3. Calculations of CV (%), bias (%), and detection limit (ug/rrf) based on sucrose
injection results for two Sunset OC/EC instruments at EPA OAQPS.
Metric
Sunset 1

Sunset 2
N valid
68
85

Coefficient of variance, CV (%)
7.6%
5.8%

Bias
6.3%
5.4%

Detection limit ug/m3
1.4
1.5

4.2 Sunset and CSN OC
Figure 1 shows box plots of 24-hr OC concentrations via Sunset and CSN, and Figure 2 shows Sunset
versus CSN on a scatter plot. Summary statistics of the Sunset-to-OC comparison are provided in
Table 4; only days where both Sunset OC and CSN OC data were available are included. Average
Sunset OC concentrations ranged from 2.1 ng/m3 at Houston to 3.2 ng/m3 at Rubidoux. Overall, OC
concentrations were higher when measured with the Sunset than in CSN, with an average ratio of
means (ROM) of 1.13. However, this was largely driven by differences between Sunset and CSN at Las
11

-------
4. Results
Vegas, where the means between the two methods are not comparable. At the other five sites, the
ROM was 1.06, indicating that on average there was good agreement between the two methods and
that MQOs were generally met. As noted earlier, there were significant and frequent operational
problems at Las Vegas that likely biases the results there. When the 8.8% precision (CV) of CSN OC is
considered, all sites have comparable means between Sunset and CSN except at Las Vegas. Where
sufficient samples were available, we also found that there was no significant change in the ROM
among seasons, i.e., Sunset and CSN means were comparable in all seasons at each site that had at
least 10 24-hr values, except at Las Vegas.
The correlation (r2) between Sunset and CSN with all measurements was 0.67, and nearly meets the
MQO of R=0.90 if Las Vegas is excluded (r2 = 0.73, R = 0.85). The slope is close to 1 at Rubidoux,
Chicago, and St. Louis (0.87 to 0.93), and lower at Las Vegas and Houston (0.62 to 0.66). Grouping all
measurements together yields a slope of 0.77, with a bias towards Sunset OC being higher than CSN
OC. The scatter plot shows a number of outliers, in particular at Las Vegas and Houston, where both
CSN and Sunset measurements initially appeared to be valid and were not removed after initial
investigation. Without these outliers, the correlation improves marginally, but the bias between the
two measurements would remain relatively unchanged. In fact, even with the multiple operational
issues that occurred, the bias between Sunset and CSN measurements is fairly consistent across sites.
Overall, the Sunset and CSN OC concentrations compared fairly well across the sites, with an r2 of
0.67 and comparable means at all sites except Las Vegas, though with variations in the degree of
scatter depending on the frequency of operational issues. There is consistently a bias toward Sunset
OC being higher than CSN OC, though this varies by site; however, only at Las Vegas and Houston
are the Sunset OC values significantly higher than the CSN OC values. At St. Louis, where there are
nearly 200 measurements included in the analysis, the ratio between Sunset and CSN switches from a
Sunset/CSN ratio of 1.06 during the early period of operations of 2013 through early-2015 to 0.91 in
2016 and 2017. The differences between the two periods is that new software, a new oven, and a new
NDIR detector were installed, so it is unclear which of these specific actions led to a change in Sunset
OC readings. At Las Vegas, there were frequent operational issues, and the sample size is relatively
small compared to other sites (n = 53), so these results cannot be weighted as heavily as those from
other sites. With a somewhat broad range of results, Sunset operations likely play a large role in how
well the instrument compares to CSN OC data.
In addition, there are differences in how the two thermal-optical methods determine OC and EC;
these differences may play a role in how comparable the Sunset (which used NIOSH) and CSN (which
used IMPROVE_A) measurements are, even though the total carbon (OC + EC) typically compares
well between the two methods (Chow et al., 2001; 2007). A main difference between the two
methods is the temperature regime used to determine OC and EC: in NIOSH; the temperature is
ramped to 870°C for determining OC while in IMPROVE_A, it is ramped to 550°C. This means that
some carbon that is quantified as OC in NIOSH may be quantified as EC in IMPROVE_A; therefore, for
a given sample, the NIOSH OC would be higher than the IMPROVE_A OC, and the NIOSH EC would
be lower than the IMPROVE_A EC. In a direct comparison of these different maximum temperature
12

-------
4. Results
regimes, Piazzalunga confirmed that a "significant amount" of weakly light-absorbing carbonaceous
aerosols were evolved off under S70°C (Piazzalunga et al., 2011). In Hong Kong, Wu et al. compared
NIOSH and IMPROVE measurements across urban, roadside, and suburban sites over three years, and
found that differences between the two methods are mostly from the way the OC and EC split is
determined, such that in Hong Kong EC from IMPROVE_A was roughly 2.2 times higher than from
NIOSH, with more minor differences for OC between the methods (Wu et al., 2016). They also found
that the amounts of biomass burning and metal oxides such as iron and zinc also impacted how well
the two methods compared, where higher metal oxide concentrations led to an increase in the
difference between IMPROVE_A and NIOSH EC. Similar results were found in the Southeastern United
States during 2003-2005, with total carbon comparable between the two methods but with lower EC
from the NIOSH method (Cheng et al., 2011).
10.0
o>
ZL
c
O
ro
i—
c
CD
O
£=
O
O
a>
a>
cp
o
>
<
CN
7.5
5.0
2.5
0.0
5te
I#
-Q °
or
i
Co
T-*
<\/
^ Sunset OC
$ CSN OC
*
; • : i	: !
O co
Qo
-
C II
£ <=•
0>CO
-c o
CO ^
cd
/V
„ <¦o
0	II
CD C
.££•
o-
1
V
r-.
O CQ
.S OJ
^ 		
.CO «
COC
t—
/

&
o>
^ <¦o
- II
co cr
co
ctr O)

V—
c\/
Site
Figure 1. Box plot of OC concentrations at each site via Sunset and CSN measurements.
13

-------
4. Results
10
8
Q)
O
o
CD
If)
d
=s
en
CD
CD
CD
1_
CD
>
<
OJ
cm 2
•	Rubidoux, CA
Washington, DC
•	Chicago, IL
St. Louis, MO
•	Las Vegas, NV
•	Houston, TX
y = 0.70 + 0.88 x, /T = 0.71
0
y = 0.37+ 0.93
y = 0.33 + 0.87
y = 1.63 + 0.62
y = 0.86+ 0.66
y = 0.67 + 0.77
x.
x,
X,
x,
X,
R' = 0.89
R = 0.70
R2 = 0.32
Rz = 0.67
R = 0.67
0
CSN OC (|Lig/m3)
8
10
Figure 2. Scatter plot of Sunset and CSN OC concentrations, colored by site; the linear
regression equation written in black is for all data at all sites.
14

-------
4. Results
Table 4. Summary of Sunset and CSN OC measurements and comparison statistics.
Site
N
Mean
Sunset
OC
StDev
Sunset
OC
Mean
CSN
OC
StDev
CSN
OC
Ratio of
the
Means
Comparable
Means?
Slope
Intercept
r2
Rubidoux
75
3.2
1.7
2.8
1.6
1.12
Yes
0.88
0.70
0.71
Washington
D.C.
208
2.3
0.8
2.4
1.1
0.96
Yes
0.70
0.64
0.85
Chicago
57
2.5
1.1
2.3
1.2
1.09
Yes
0.93
0.37
0.89
St. Louis
198
2.4
1.2
2.4
1.1
1.01
Yes
0.87
0.33
0.70
Las Vegas
53
2.8
1.2
1.8
1.1
1.52
No
0.62
1.63
0.32
Houston
154
2.1
0.9
1.9
1.2
1.10
Yes
0.66
0.86
0.67
15

-------
4 Results
43 Sunset and CSN Thermal EC and Sunset Optical EC
Summary statistics of the Sunset-to-CSN EC comparison are provided in Table 5; only days where
both Sunset EC or OptEC data plus CSN EC were available are included. Figure 3 shows box plots of
24-hr EC concentrations at each site via Sunset and CSN, and Figure 4 shows scatter plots comparing
CSN EC concentrations with Sunset EC and Sunset OptEC. (For Sunset EC, both thermal EC and
optical EC results are shown.) Mean CSN EC concentrations varied between 0.26 ng/m3 (Houston)
and 0.83 ng/m3 (Rubidoux). Sunset thermal EC was similar to CSN EC on average (1.03 ROM) when
excluding Las Vegas and Houston; these latter two had much higher ROM (1.76 and 3.55,
respectively) and recorded significantly higher Sunset EC compared to CSN EC. OptEC was
consistently lower than CSN EC except at Houston; the average Sunset OptEC/CSN EC ratio when
excluding Las Vegas and Houston was 0.90. Houston OptEC was much closer to CSN EC than the
thermal EC was (ratio of 1.08 instead of 3.55). Thus, except for OptEC at Las Vegas and the thermal EC
at Houston, MQOs were met.
While there is good agreement between the overall EC means at all sites, with the exception of
thermal EC at Las Vegas and Houston, the slope and correlation between Sunset and CSN
measurements vary widely. For thermal EC, there is relatively high correlation at St. Louis, Rubidoux,
and Chicago (r2 of 0.76 to 0.89 for Sunset EC to CSN EC), but there is poor correlation for the other
sites (r2 of 0.33 to 0.41). Correlations and slopes are more comparable for OptEC to CSN EC, with an
r2 value of 0.67 when all measurements are pooled together, although with a bias toward CSN EC
being higher (slope of 0.65).
Overall, OptEC measurements appear to be more in line with CSN EC than the thermal EC
measurements are. It is unclear what operational differences occurred at Houston to result in such a
large disparity between the site's EC and OptEC results, which was consistent throughout the study.
In addition, having fairly consistent results across all five sites with OptEC, despite numerous
operational issues, is significant: the OptEC measurement is fairly consistent when compared to CSN
EC despite different locations and operations. The difference between OptEC and thermal EC is
partially due to differences in detection limits; 26% of hourly concentrations were below the
detection limit of 0.2 |jg/m3 for thermal EC. Bauer et al. (2009) estimated that the detection limit for
OptEC is lower than for thermal EC, at between 0.02 to 0.1 |jgC/m3; having so many of the
observations near or below the detection limit for thermal EC likely impacts these results. Potential
interferences in the thermal method from metal oxides may also play a role.
16

-------
4. Results
Table 5. Summary of Sunset and CSN EC measurements and comparison statistics.
Site Code
Site Name
N
Mean
Sunset
EC
StDev
Sunset
EC
Mean
CSN
EC
StDev
CSN
EC
Ratio
of the
Means
Comparable
Means?
Slope
Intercept
r2
Sunset EC
vs. CSN EC











06 065 8001
Rubidoux, CA
75
0.94
0.57
0.83
0.56
1.13
Yes
0.88
0.21
0.76
11 001 0043
Washington,
DC
208
0.40
0.23
0.52
0.42
0.75
Yes
0.35
0.21
0.41
17 031 0076
Chicago, IL
60
0.50
0.26
0.40
0.21
1.24
Yes
1.18
0.02
0.89
29 510 0085
St. Louis, MO
63
0.41
0.35
0.42
0.24
0.99
Yes
1.28
-0.13
0.76
32 003 0540
Las Vegas, NV
53
0.93
0.77
0.53
0.45
1.76
No
0.97
0.42
0.33
48 201 1039
Houston, TX
154
0.93
0.37
0.26
0.14
3.55
No
1.50
0.54
0.33
Sunset
OptEC vs.
CSN EC











06 065 8001
Rubidoux, CA
75
0.75
0.52
0.83
0.56
0.90
Yes
0.81
0.07
0.77
11 001 0043
Washington,
DC
211
0.44
0.26
0.52
0.42
0.83
Yes
0.45
0.20
0.50
29 510 0085
St. Louis, MO
198
0.42
0.24
0.43
0.25
0.98
Yes
0.89
0.04
0.88
32 003 0540
Las Vegas, NV
53
0.37
0.26
0.53
0.45
0.70
No
0.42
0.15
0.53
48 201 1039
Houston, TX
154
0.28
0.18
0.26
0.14
1.08
Yes
1.06
0.01
0.69
17

-------
4. Results
231
^Sunset EC
* Sunset OptEC
^CSN EC
o
03
0 '
O '
C
o
O

<
C\J

S*
£§
~o o
or<^
i
-------
4.4 Sunse
d Aethalometer BC
4. Results
Summary statistics of 24-hour Sunset OptEC to Aethalometer BC are provided in Table 6, and
comparisons between the two measurements are shown in Figures 5 and 6. The Sunset OptEC was
consistently lower than the Aethalometer BC at all three of the sites that had data for both
measurements (Washington, D.C., St. Louis, and Houston), with a mean ROM of 0.57. This ratio was
fairly consistent at each site, with little seasonal variation. For example, the OptEC/BC ratio at
Washington, D.C. varied between 0.65 in the winter to 0.70 in the summer. At all three sites, the
differences in OptEC and BC measurements were statistically significant, and the measurements were
not comparable even when accounting for the precision of the Aethalometer (3.5% for 24-hr
measurements).
This consistent offset between the two measurements is clearly seen in the scatter plot at Figure 6,
where the r2 value is higher than 0.82 and the intercept is near zero at all three sites. The relationship
between Sunset OptEC and Aethalometer BC is consistent at a range of concentrations, which in this
study is up to 2 ng/m3 OptEC. However, the slope of the regression, and the ratio of the OptEC/BC
means, varies across sites. At St. Louis, the OptEC/BC ratio is 0.47, but at Washington, D.C., it is 0.67.
In a multi-year study in New York, similar results were found, with a high correlation between EC and
BC, and with BC higher by nearly a factor of 2 during summertime (ratios of 1.3 in winter and 1.8 in
summer) (Rattigan et al., 2010; 2013). They also report variation in the BC/EC ratio that we did not
see at the sites in this study, with a higher ratio in summer than in winter. In addition, Rattigan et al.
(2013) found variation in the BC/EC ratio between the Bronx and Rochester, i.e., between a major
urban area and a smaller one. They ascribe part of this variation to changes in optical properties of
the ambient aerosol due to emissions from residential wood burning or fuel oil in the wintertime. We
do not see the large seasonality in the BC/EC ratio that was observed in Rattigan et al. However, the
OptEC/BC ratio is consistently different at each of the three sites here, either due to operational
differences among the sites and the sites in Rattigan et al., and/or due to differences in aerosol
sources between New York and the sites here.
19
-------
4. Results
Table 6. Summary of Sunset OptEC and Aethalometer BC measurements and comparison statistics.
Site
N
Mean
Sunset
OptEC
StDev
Sunset
OptEC
Mean
Aeth
BC
StDev
Aeth
BC
Ratio of
the Means
Comparable
Means?
Slope
Intercept
r2
Washington D.C.
618
0.45
0.27
0.67
0.43
0.67
No
0.59
0.05
0.85
St. Louis
544
0.40
0.22
0.85
0.47
0.47
No
0.44
0.02
0.87
Houston
415
0.31
0.20
0.57
0.26
0.54
No
0.71
-0.10
0.82
20
-------
4. Results
¦2 2
O
O
oS
. CO
c //
¦5 c
0)C0
v-
s?

CO
V
C\/
c\/
Figure 5. Box plot of 24-hour average Sunset OptEC and Aethalometer BC.
o
LU
03
O
Q.
O
0 2
in
c
13
CO
-------
4. Results
4.5 Precision from Collocated Measurements in St Louis
Collocated Sunset instruments were operated at St. Louis during August 11, 2016, through January
11, 2017, and offer a way to gauge precision between two relatively well-operating instruments. A
scatter plot of the two instruments, termed POC 1 and POC 2, is shown in Figure 7 for 24-hr average
OC and Optical EC (n=102); only OptEC was reported at St Louis during this time period. POC 1,
which had been operated during the course of the study, used He as a carrier gas; POC 2 was set up
in the summer of 2016 and used zero air as a carrier gas.
OC concentrations varied between 0.6 and 6.7 |jgC/m3, and OptEC between 0.1 and 1.7 |jgC/m3.
There is consistent agreement between both OC and OptEC measurements from the two
instruments, with r2 values of 0.93 for OC and 0.91 for OptEC. There is a bias in the slope (1.04 for OC,
1.12 for OptEC), but with an offset (y-intercept) of 0.7 |jgC/m3 for OC and no offset for OptEC (y-
intercept of zero). This offset for OC may be due to differences in carrier gas, with small impurities in
either the He or zero air carrier gas influencing the OC concentrations but not the OptEC. Results are
similar to some of the first series of collocated measurements reported by Bauer et al., which found
high correlation (0.97 and 0.98) for OC and OptEC when an ambient sample stream was split and
routed to two collocated Sunset instruments (Bauer et al., 2009). They found a lower slope for OptEC
(0.82) and magnitude similar to our results for OC (0.95). Their interpretation of results similar to the
ones found here at St. Louis was that the instrument produces reliable and reproducible
measurements when mass loadings are higher than detection limits. They also note that the
instrument needs to be working properly for obtaining such reliable data, similar to the experiences
found at multiple sites in this study.
7
• OC
linear fit y=0.71+1.04x
1:1 line
6
5
E
o
O)
Zi.
O 4
O
c
-------
4. Results
4.6 Diurnal Patterns
A strong diurnal pattern was seen at Rubidoux and Las Vegas for both OC and EC, but the average
diurnal pattern at other sites was more muted (Figure 8). At Rubidoux, OC peaked in the evening
while EC peaked in the morning. At the other sites, an overnight peak in OC was also seen, though
this overnight peak was only modestly higher than the morning or midday concentrations. EC
peaked in the morning at all sites, and was clearly higher on weekdays compared to weekends at all
sites. OC was slightly higher on weekends compared to weekdays for nearly all hours at each site.
This suggests that while ambient EC concentrations may be more driven by changes in traffic in the
morning and on weekdays compared to weekend, ambient OC concentrations are a complex mixture
of aerosol and semi-volatile material that varies due to changes in photochemistry, ambient
particulate matter concentration levels, and emissions (Jimenez et al., 2009; Donahue et al., 2012;
Robinson et al., 2007).
23
-------
4. Results
Chicago, IL
£ 4
Sunset OC Weekday
- Sunset OC Weekend
• Sunset EC Weekday
Sunset EC Weekend
012345678 9 1011121314151617181920212223
Hour
Houston, TX
E 4
Sunset OC Weekday
Sunset OC Weekend
• Sunset EC Weekday
Sunset EC Weekend
0123456789 1011 121314151617181920212223
Hour
Las Vegas, NV
Rubidoux, CA
E 4
Sunset OC Weekday
Sunset OC Weekend
• Sunset EC Weekday
Sunset EC Weekend
E 4
Sunset OC Weekday
Sunset OC Weekend
• Sunset EC Weekday
Sunset EC Weekend
10 12 14 16 18 20 22
Hour
0123456789 1011121314151617181920212223
Hour
St. Louis, MO
Washington, DC
E 4
Sunset OC Weekday
-	Sunset OC Weekend
-	Sunset EC Weekday
Sunset EC Weekend
E 4
Sunset OC Weekday
¦ Sunset OC Weekend
Sunset EC Weekday
Sunset EC Weekend
012345678 9 1011 121314151617181920212223
Hour
0123456789 1011 121314151617181920212223
Hour
Figure 8. Average hourly OC and EC concentrations on weekdays and weekends at each site.
24
-------
5. Conclusions
	¦
Sunset OC/EC instruments were operated at six sites collocated with CSN measurements, with
Aethalometer measurements collocated at three of these sites. Operations were quite variable
among the sites, with multiple operational issues at all the sites. Critically, when components of the
instrument broke, e.g., the oven or NDIR, it was not clear in the data output that there was a
problem. This led to multiple weeks to months of operations with a broken component, resulting in
large gaps in quality data. Many of these components have since been upgraded or redesigned by
Sunset Labs, including changes to the software that will better alert users when a component is
damaged.
Despite the operational problems at all sites, overall Sunset OC and OptEC compared well with CSN
and Aethalometer measurements. Sunset OC was consistently higher than CSN OC, with a
Sunset/CSN ratio of 1.06. This ratio is well within the precision of the CSN measurements, so the
Sunset and CSN OC are comparable. While, on average, the Sunset EC and CSN EC were similar at
four of the six sites, there was large scatter and varying biases between these two thermal EC
measurements across all sites. This indicates that the thermal EC measurements are not as
comparable as the OC measurements are, though part of this may be due to having 26% of Sunset
EC measurements below the detection limit. Sunset OptEC data had a much better agreement with
CSN EC data, as well as with Aethalometer BC data; Sunset OptEC also has a lower detection limit
than Sunset EC, which likely accounts for its improved comparison to CSN EC. The OptEC was
consistently lower than the BC, similar to what has been seen previously in the literature, though we
did not see seasonal fluctuations in the OptEC/BC ratio. That OptEC is quite comparable to both CSN
EC and BC indicates that it is a robust and consistent measurement. Overall, with improvements to
the NDIR, oven, and software, the Sunset instrument is a viable instrument for field deployment,
though not as "plug and play" as other particulate instruments used in routine monitoring networks.
25
-------
-------
6. References
6. References
Allen G.A., Lawrence J., and Koutrakis P. (1999) Field validation of a semi-continuous method for aerosol
black carbon (Aethalometer) and termporal patterns of summertime hourly black carbon
measurements in Southwestern Pennsylvania. Atmos. Environ., (33), 817-823.
Bae M.S., Schauer J.J., DeMinter J.T., and Turner J.R. (2004a) Hourly and daily patterns of particle-phase
organic and elemental carbon concentrations in the urban atmosphere. J. Air Waste Manage.,
54(7), 823-833, Jul.
Bae M.S., Schauer J.J., DeMinter J.T., Turner J.R., Smith D., and Cary R.A. (2004b) Validation of a semi-
continuous instrument for elemental carbon and organic carbon using a thermal-optical method.
Atmos. Environ., 38(18), 2885-2893, Jun.
Bauer J.J., Yu X.-Y., Cary R., Laulainen N., and Berkowitz C. (2009) Characterization of the Sunset semi-
continuous carbon aerosol analyzer. J. Air Waste Manage., 59(7), 826-833, doi: 10.3155/1047-
3289.59.7.826.
Birch M.E. and Cary R.A. (1996) Elemental carbon-based method for monitoring occupational exposures to
diesel exhaust. Aerosol Science and Technoiogy, 25, 221-241.
Brown S.G., Lee T., Roberts P.T., and Collett J.L., Jr. (2013) Variations in the OM/OC ratio of urban organic
aerosol next to a major roadway. J. Air Waste Manage., 63(12), 1422-1433, doi:
10.1080/10962247.2013.826602 (STI-5421), December. Available at
http://www.ncbi.nIm.nih.gov/pubmed/24558705.
Budisulistiorini S.H., Canagaratna M.R., Croteau P.L., Baumann K., Edgerton E.S., Kollman M.S., Ng N.L.,
Verma V., Shaw S.L., Knipping E.M., Worsnop D.R., Jayne J.T., Weber R.J., and Surratt J.D. (2014)
Intercomparison of an Aerosol Chemical Speciation Monitor (ACSM) with ambient fine aerosol
measurements in downtown Atlanta, Georgia. Atmospheric Measurement Techniques, 7,1929-
1941, doi: 10.5194/amt-7-1929-2014.
Camalier L., Eberly S., Miller J., and Papp M. (2007) Guideline on the meaning and the use of precision and
bias data required by 40 CFR Part 58 Appendix A. Report by U.S. Environmental Protection
Agency, Office of Air Quality Planning and Standards, Research Triangle Park, NC, EPA-454/B-07-
001, October. Available at
https://www3.epa.gov/ttn/amtic/files/ambient/qaqc/PBGuidancel01007.pdf.
Cheng Y., Zheng M., He K.-b., Chen Y., Yan B., Russell A.G., Shi W., Jiao Z., Sheng G., Fu J., and Edgerton E.S.
(2011) Comparison of two thermal-optical methods for the determination of organic carbon and
elemental carbon: results from the southeastern United States. Atmos. Environ., 45(11), 1913-1918,
doi: 10.1016/j.atmosenv.2011.01.036. Available at
http://www.sciencedirect.com/science/article/pii/S13522310110Q0641.
Chow J.C., Watson J.G., Crow D., Lowenthal D.H., and Merrifield T. (2001) Comparison of IMPROVE and
NIOSH carbon measurements. Aerosol Science and Technology, 34, 23-34.
27
-------
6. References
Chow J.C., Watson J.G., Chen L.-W.A., Chang M.O., Robinson N.F., Trimble D., and Kohl S. (2007) The
IMPROVE_A temperature protocol for thermal/optical carbon analysis: maintaining consistency
with a long-term database. J. Air Waste Manage., 57(9), 1014-1023.
Docherty K.S., Aiken A.C., Huffman J.A., Ulbrich I.M., DeCarlo P.F., Sueper D., Worsnop D.R., Snyder D.C.,
Peltier R.E., Weber R.J., Grover B.D., Eatough D.J., Williams B.J., Goldstein A.H., Ziemann P.J., and
Jimenez J.L. (2011) The 2005 study of organic aerosols at Riverside (SOAR-1): instrumental
intercomparisons and fine particle composition. Atmos. Chem. Phys., 11(23), 12387-12420, doi:
10.5194/acp-ll-12387-2011. Available at http://vvww.atrnos-chern-phys.net/ll/123S7/2011/acp-
ll-l.2387-2011.pclt
Donahue N.M., Kroll J.H., Pandis S.N., and Robinson A.L. (2012) A two-dimensional volatility basis set - Part
2: Diagnostics of organic-aerosol evolution. Atmos. Chem. Phys., 12, 615-634, doi: 10.5194/acp-12-
615-2012.
Drinovec L., Mocnik G., Zotter P., Prevot A.S.H., Ruckstuhl C., Coz E., Rupakheti M., Sciare J., Muller T.,
Wiedensohler A., and Hansen A.D.A. (2015) The "dual-spot" Aethalometer: an improved
measurement of aerosol black carbon with real-time loading compensation. Atmospheric
Measurement Techniques, 8,1965-1979, doi: 10.5194/amt-8-1965-2015. Available at
http://www.atmos-meas-tech.net/8/1965/2015/amt-8-1965-2015.pdf.
Gantt B., Landis B., and Rice J. (2017) Sunset OC-EC evaluation project: preliminary results. Information
poster by the U.S. Environmental Protection Agency, Office of Air Quality Planning and Standards,
Research Triangle Park, NC.
Healy R.M., Sofowote U., Su Y., Debosz J., Noble M., Jeong C.-H., Wang J.M., Hilker N., Evans G.J., and
Doerksen G. (2017) Ambient measurements and source apportionment of fossil fuel and biomass
burning black carbon in Ontario. Atmos. Environ., 161, 34-47.
Jacobson M.C., Hansson H.C., Noone K.J., and Charlson RJ. (2000) Organic atmospheric aerosols: Review
and state of the science. Reviews of Geophysics, 38(2), 267-294, May.
Jeong C.H., Hopke P.K., Kim E., and Lee D.W. (2004) The comparison between thermal-optical transmittance
elemental carbon and Aethalometer black carbon measured at multiple monitoring sites. Atmos.
Environ., 38(31), 5193-5204, October 2004.
Jimenez J., Claiborn C., Larson T., Gould T., Kirchstetter T.W., and Gundel L. (2007) Loading effect correction
for real-time aethalometer measurements of fresh diesel soot. J Air Waste Manag Assoc, 57(7),
868-873, doi: 10.3155/1047-3289.57.7.868, July. Available at
http://wvmrKbi.nIm.nih.gov/pubmed/17688QQ2.
Jimenez J.L., Canagaratna M.R., Donahue N.M., Prevot A.S.H., Zhang Q., Kroll J.H., DeCarlo P.F., Allan J.D.,
Coe H., Ng N.L., Aiken A.C., Docherty K.S., Ulbrich I.M., Grieshop A.P., Robinson A.L., Duplissy J.,
Smith J.D., Wilson K.R., Lanz V.A., Hueglin C., Sun Y.L., Tian J., Laaksonen A., Raatikainen T.,
Rautiainen J., Vaattovaara P., Ehn M., Kulmala M., Tomlinson J.M., Collins D.R., Cubison M.J., Dunlea
E.J.,	Huffman J.A., Onasch T.B., Alfarra M.R., Williams P.I., Bower K., Kondo Y., Schneider J., Drewnick
F.,	Borrmann S., Weimer S., Demerjian K., Salcedo D., Cottrell L., Griffin R., Takami A., Miyoshi T.,
Hatakeyama S., Shimono A., Sun J.Y., Zhang Y.M., Dzepina K., Kimmel J.R., Sueper D., Jayne J.T.,
Herndon S.C., Trimborn A.M., Williams L.R., Wood E.C., Middlebrook A.M., Kolb C.E., Baltensperger
28
-------
6. References
U., and Worsnop D.R. (2009) Evolution of organic aerosols in the atmosphere. Science, 326,1525-
1529, doi: 10.1126/science.ll80353, December 11.
Lee B.P., Li Y.J., Yu J.Z., Louie P.K.K., and Chan C.K. (2013) Physical and chemical characterization of ambient
aerosol by HR-ToF-AMS at a suburban site in Hong Kong during springtime 2011. Journal of
Geophysical Research: Atmospheres, 118(15), 8625-8639, doi: 10.1002/jgrd.50658.
ParkS.S., Harrison D., Pancras J.P., and OndovJ.M. (2005) Highly time-resolved organic and elemental
carbon measurements at the Baltimore supersite in 2002. Journal of Geophysical Research,
110(D07S06), doi: 10.1029/2004JD004610, April 1.
Piazzalunga A., Bernardoni V., Fermo P., Valli G., and Vecchi R. (2011) Technical note: on the effect of water-
soluble compounds removal on EC quantification by TOT analysis in urban aerosol samples.
Atmos. Chem. Phys., 11(19), 10193-10203, doi: 10.5194/acp-ll-10193-2011. Available at
https;//wvw.atmos-chem-phys.net/ll/10193/2011/.
Rattigan O.V., Felton H.D., Bae M.-S., Schwab J.J., and Demerjian K.L. (2010) Multi-year hourly PM2.5 carbon
measurements in New York: diurnal, day of week and seasonal patterns. Atmos. Environ., 44, 2043-
2053, doi: 10.1016/j.atmosenv.2010.01.019, January.
Rattigan O.V., Civerolo K., Doraiswamy P., Felton H.D., and Hopke P.K. (2013) Long term black carbon
measurements at two urban locations in New York. Aerosol Air Qual. Res, 13(4), 1181-1196.
Rice J. and Landis E. (2016) Chemical speciation network (CSN): summary of precision from six collocated
sites. Poster presented at the 2016 National Ambient Air Monitoring Conference, St. Louis, Missouri,
August 8-11, by the U.S. Environmental Protection Agency, Office of Air Quality Planning and
Standards, Research Triangle Park, NC. Available at
https://www.epa.gov/sites/productiorVflIes/2016-09/documents/summary_of_precisioapdf.
Robinson A.L., Donahue N.M., Shrivastava M.K., Weitkamp E.A., Sage A.M., Grieshop A.P., Lane T.E., Pierce
J.R., and Pandis S.N. (2007) Rethinking organic aerosols: semivolatile emissions and photochemical
aging. Science, 315,1259.
Sciare J., d'Argouges O., Sarda-Esteve R., Gaimoz C., Dolgorouky C., Bonnaire N., Favez O., Bonsang B., and
Gros V. (2011) Large contribution of water-insoluble secondary organic aerosols in the region of
Paris (France) during wintertime. Journal of Geophysical Research (Accepted), 116(D22203), doi:
10.1029/2011JD015756, November 17.
Sin D.W.M., Fung W.H., Choi Y.Y., Lam C.H., and Wong Y.C. (2004) Measurement of carbonaceous aerosols:
validation of a thermal gravimetric method and its comparison with a thermal optical
transmittance method. Microchemical Journal, 77(1), 63-70, May.
Snyder D.C. and Schauer J. (2007) An inter-comparison of two black carbon aerosol instruments and a
semi-continuous elemental carbon instrument in the urban environment. Aerosol Science &
Technology, 41(5), 463-474, May.
Takegawa N., Miyazaki Y., Kondo Y., Komazaki Y., Miyakawa T., Jimenez J.L., Jayne J.T., Worsnop D.R., Allan
J.D., and Weber RJ. (2005) Characterization of an Aerodyne Aerosol Mass Spectrometer (AMS):
29
-------
6. References
intercomparison with other aerosol instruments. Aerosol Science & Technology, 39, 760-770, doi:
10.1080/02786820500243404.
U.S. Environmental Protection Agency (2011) Sunset Quality Assurance Project Plan (QAPP). September 27.
Weingartner E., Saathoff H., Schnaiter M., Streit N., Bitnar B., and Baltensperger U. (2003) Absorption of
light by soot particles: determination of the absorption coefficient by means of aethalometers.
Journal of Aerosol Science, 34(10), 1445-1463, Oct.
Wu C., Huang X.H.H., Ng W.M., Griffith S.M., and Yu J.Z. (2016) Inter-comparison of NIOSH and IMPROVE
protocols for OC and EC determination: implications for inter-protocol data conversion. Atmos.
Meas. Tech., 9(9), 4547-4560, doi: 10.5194/amt-9-4547-2016. Available at https://www.atmos~
meas-tech.net/9/4547/2016/.
Zhang Q., Canagaratna M.R., Jayne J.T., Worsnop D.R., and Jimenez J.-L. (2005) Time- and size-resolved
chemical composition of submicron particles in Pittsburgh: Implications for aerosol sources and
processes. Journal of Geophysical Research, 110, D07S09, doi: 10.1029/2004JD004649.
Zheng G.J., Cheng Y., He K.B., Duan F.K., and Ma Y.L. (2014) A newly identified calculation discrepancy of
the Sunset semi-continuous carbon analyzer. Atmospheric Measurement Techniques (Accepted), 7,
1969-1977, doi: 10.5194/amt-7-1969-2014, July 3.
Zfkova N., Vodicka P., Ludwig W., Hitzenberger R., and Schwarz J. (2016) On the use of the field Sunset
semi-continuous analyzer to measure equivalent black carbon concentrations. Aerosol Science and
Technology, 50(3), 284-296, doi: 10.1080/02786826.2016.1146819, 2016/03/03.
30
-------
Appendix A
Appendix A. Summary of Instrument
Operations During the Sunset Carbon
:	K: ^ w
A.1
-------
-------
Appendix A
Sonoma Technology, Inc.
Innovative Environmental Solutions
Technical Memorandum
October 9, 2017
STI-915313-6804-TM
To: Beth Landis, EPA OAQPS
From: Hilary Minor, Steven Brown
Re: Summary of instrument operations during the Sunset carbon evaluation project
Study Overview
The Sunset Carbon Evaluation Project, led by EPA's Office of Air Quality Planning and Standards
(OAQPS) Ambient Air Monitoring Group (AAMG), assessed the performance and feasibility of the
Sunset Semi-Continuous Organic and Elemental Carbon (OC/EC) instrument at six sites throughout
the United States (Chicago, St Louis, Washington D.C., Houston, Las Vegas, and Los Angeles) during
2012-2017. State and local monitoring agencies operated the Sunset instrument to evaluate
instrument performance, gain proficiency in the proper operation and maintenance of the
instrument, compare the Sunset thermal OC/EC with the URG 3000N thermal OC/EC sampler, and
compare the Sunset optical EC with the Aethalometer optical black carbon (BC) instrument After the
study ended, Sonoma Technology, Inc. (STI) worked with the monitoring agencies that operated the
Sunset carbon analyzer, representatives from Sunset, and EPA staff involved in the evaluation project
to understand the operational issues encountered during the Sunset Carbon Evaluation Project. This
report summarizes the issues experienced during the evaluation study and, if applicable, details of
Sunset's improvements to the instrument since the study has ended.
Operational Issues, Analysis, and Resolution
Issue: Users struggled with the sucrose injection process and found it complicated and time intensive.
Sucrose injections were recommended every two weeks, in order to determine the calibration
constant (the total mass [jjg] of carbon in the instrument's calibration loop at any given time). Users
requested guidance on the type of pipette or syringe for easiest application, application technique,
and acquisition and storage of the sucrose solution. A system that is more user-friendly would
improve their experience. Initially, there was high variability among sucrose injections, so the type of
syringe used was changed to an autopipette, and Sunset implemented a drying step to remove
variability in the sucrose injections.
A. 3
-------
Appendix A
•	As part of the review of the project, Sunset's representative commented that users were
performing sucrose injections more frequently than needed. Sucrose injections are used to
determine the calibration constant, which should only change as the calibration gas tank is
changed, approximately every 1 to 1.5 years.1 After the evaluation study, Sunset developed
instrument performance filters that can be used instead of the sucrose injections. However,
the Sunset manual also indicates that calibration checks should be performed on a weekly
basis, which conflicts with the information presented above.
•	The non-dispersive infrared (NDIR) detector was improved, negating an interference issue
that required the sucrose solution to dry before the sucrose standard analysis was begun.
•	The preparation and storage of carbon standard stock solution is detailed in the
manufacturer's instrument guide.2
Resolution: The NDIR detector was improved and performance filters are now available to users, so
sucrose injections do not appear necessary except when calibration gas tanks are changed. However,
additional clarification on the frequency of and the procedure for performing sucrose injections is
needed in the manual.
Issue: Users experienced hardware issues throughout the evaluation study.
Hardware issues included cracked ovens, NDIR detector failure, heating element replacement, and
pump failure. Users noted that the instrument component malfunctions were not easily diagnosed
and led to shifts in baselines and other data issues that make the data unusable. These hardware
failures were also expensive and time-consuming to fix.
•	Sunset improved the NDIR detector, increasing the lifespan of the device from 2 years to 5-7
years.
•	The instrument now features extended-life heating coils, so these should not have to be
replaced as frequently as the coils in this study.
•	The error message screens are designed to minimize errors and report specific hardware or
performance-related problems to users. To monitor the functionality of the hardware, Sunset
recommends plotting and tracking instrument diagnostic metrics continuously over two- to
four-week periods to monitor instrument performance. Specifically, Sunset recommends
plotting and tracking the calibration area value, read from the _Res.txt results file, for
changes of more than 10% that would indicate a leak. Users can also track the Laser-Temp
Correction value, available on the main screen and in the _LCRes.txt results file, for any result
below 0.90, which indicates that the filter needs to be changed. The results file, Local
Conditions Results (_LCRes.txt), also contains error flags (1, "review", and 2, "fail") that alert
users to problems.
1	Sunset Laboratory Inc., Organic Carbon and Elemental Carbon Field Instrument: Model 4 User's Manual (M4-Rev 9), page 43.
2	Sunset Laboratory Inc. Semi-Continuous OCEC Carbon Aerosol Analyzer: A Guide to Running and Maintaining the Sunset
Laboratory Semi-Continuous OCEC Analyser (M5-Rev3), page 83. Available at https://vmw3.epa.gov/ttnaiTitil/files/
ambient/pm25/spec/Sunset_Manual.pdf.
A.4
-------
Appendix A
Resolution: Sunset improved the NDIR and heating coil components and provided guidance on what
metrics can be used to track instrument performance.
Issue: Users noted that software updates during the study did not detail the changes made to the
software since the last version,
•	Sunset's software improvement process first deploys software updates to a laboratory
instrument and then, after testing, deploys the software updates to instruments in the field.
Software updates during the evaluation study improved the OC/EC split determination and
baseline corrections, and added the LC results files and the data validation columns.
•	Sunset provided a list of major software changes in an undated manual.3 Continuing to list
major software changes would benefit users in the future.
Resolution: After the study was complete, Sunset provided information on software changes deployed
during the evaluation study. Additionally, detailed software update information was provided in an
older, undated version of the manual. Sunset should continue to inform users of the details of new
software changes when the software changes are made.
Issue: Helium, a carrier gas used to purge the front and back ovens of ambient air, was in short supply
and expensive during the evaluation study.
Resolution: A zero air tank, instead of a helium tank, can be used as the carrier gas. Staff at the
Missouri Department of Natural Resources (DNR) were successful in running a Sunset with zero air
instead of helium.
Issue: Users struggled with filter replacement technique and frequency
Users noted that there were issues with the filter staying in place: if not tightened enough, the filter
flips; if tightened too much, the filter or quartz insert breaks. Furthermore, Missouri DNR staff noted
that filters needed to be replaced weekly when the instrument was running in thermal mode,
compared to every two weeks when the instrument was running in optical mode.
Resolution: There was no resolution found for the filter replacement technique, except that users can
run the instrument in optical mode to reduce the frequency of filter changes. The Laser-Temp
Correction parameter, displayed on the instrument's main screen, can be used to determine when the
filter needs to be replaced. A value below 0.90 indicates that the filter needs to be changed.
Issue: In general, operating the Sunset analyzer was relatively time-consuming and challenging.
Users noted that, in general, operating the Sunset analyzer is more time-consuming and challenging
than operating other air monitoring instruments.
Resolution: Improved guidance on filter changes, sucrose injection frequency, and on what instrument
metrics to monitor will help to reduce the time needed for running the instrument, as will improved
NDIR detectors and heating coils.
3 Sunset Laboratory Inc. Semi-Continuous OCEC Carbon Aerosol Analyzer: A Guide to Running and Maintaining the Sunset
Laboratory Semi-Continuous OCEC Analyser (M5-Rev3), beginning on page 25. Available at
httpR//vww3.ep&gw/ttnamtil/fi!es/ambient/pm25/spec/ Sunset_Manual.pdf.
A, 5
-------
-------
Appendix B
Appendix B. Supplemental Information
•	Summary statistics of parameter by site
•	Time series of Sunset data available in AQS by site, annotated where data were excluded from
the analysis presented here
•	Box plots of Sunset to CSN or Aethalometer ratios by site
•	Box plots of Sunset and Aethalometer data by hour by site
B.l
-------
Appendix B
Table B-l. Summary of 24-hr average valid data by site and variable. Completeness percentage is calculated for Sunset data
only; SD is standard deviation.
Site Name
Variable
Count
Mean
Min
Max
Sd
Start
Date
End Date
Expected
Count
Completeness
Rubidoux, CA
CSN EC
469
0.7
0.0
3.0
0.5
8/1/2012
7/29/2016
1458

Rubidoux, CA
CSN OC
469
2.7
0.4
9.5
1.3
8/1/2012
7/29/2016
1458

Rubidoux, CA
Sunset EC
237
1.0
0.1
4.1
0.6
12/17/2013
10/14/2015
666
32%
Rubidoux, CA
Sunset OC
237
3.3
0.7
15.8
1.9
12/17/2013
10/14/2015
666
32%
Rubidoux, CA
Sunset OptEC
237
0.8
0.1
3.5
0.6
12/17/2013
10/14/2015
666
36%
Washington, DC
Aeth BC
1466
0.7
0.1
3.4
0.4
8/1/2012
12/22/2016
1604

Washington, DC
CSN EC
544
0.5
0.0
3.8
0.3
8/4/2012
3/29/2017
1698

Washington, DC
CSN OC
544
2.4
0.3
10.4
1.3
8/4/2012
3/29/2017
1698

Washington, DC
Sunset EC
644
0.4
0.0
1.6
0.2
10/7/2012
8/13/2016
1406
32%
Washington, DC
Sunset OC
644
2.3
1.0
7.1
0.9
8/20/2013
8/13/2016
1089
32%
Washington, DC
Sunset OptEC
649
0.4
0.1
1.9
0.3
5/28/2013
8/13/2016
1173
46%
Washington, DC
Sunset TC
647
2.6
1.0
8.3
1.0
1/1/2013
8/13/2016
1320
59%
Chicago, IL
CSN EC
429
0.4
0.0
1.9
0.2
8/1/2012
7/29/2016
1458

Chicago, IL
CSN OC
429
2.3
0.1
10.5
1.2
8/1/2012
7/29/2016
1458

Chicago, IL
Sunset EC
191
0.5
0.1
1.6
0.3
5/1/2014
12/31/2015
609
29%
Chicago, IL
Sunset OC
181
2.5
0.6
7.9
1.1
5/1/2014
12/31/2015
609
29%
Chicago, IL
Sunset TC
182
3.0
0.7
9.2
1.3
5/1/2014
12/31/2015
609
31%
St. Louis, MO
Aeth BC
1501
0.8
0.1
4.7
0.5
8/1/2012
3/30/2017
1702

St. Louis, MO
CSN EC
539
0.4
0.0
1.5
0.2
8/1/2012
3/29/2017
1701

B.2
-------
Appendix B
Site Name
Variable
Count
Mean
Min
Max
Sd
Start
Date
End Date
Expected
Count
Completeness
St. Louis, MO
CSN OC
539
2.5
0.5
9.7
1.2
8/1/2012
3/29/2017
1701

St. Louis, MO
Sunset EC
202
0.4
0.0
1.6
0.4
5/7/2013
4/22/2014
350
32%
St. Louis, MO
Sunset OC
658
2.4
0.5
9.9
1.1
5/7/2013
3/30/2017
1423
32%
St. Louis, MO
Sunset OptEC
658
0.4
0.1
1.5
0.2
1/1/2013
3/30/2017
1549
58%
St. Louis, MO
Sunset TC
658
2.8
0.6
10.2
1.2
1/1/2013
3/30/2017
1549
46%
Las Vegas, NV
CSN EC
378
0.6
0.0
2.9
0.6
8/1/2012
7/26/2016
1455

Las Vegas, NV
CSN OC
378
2.3
0.0
9.8
1.6
8/1/2012
7/26/2016
1455

Las Vegas, NV
Sunset EC
207
1.1
0.0
7.6
1.1
8/15/2012
12/31/2014
868
26%
Las Vegas, NV
Sunset OC
211
2.9
1.1
12.6
1.5
8/15/2012
12/31/2014
868
26%
Las Vegas, NV
Sunset OptEC
210
0.4
0.0
1.7
0.3
8/15/2012
12/31/2014
868
24%
Las Vegas, NV
Sunset TC
211
3.9
1.2
14.6
2.2
8/15/2012
12/31/2014
868
24%
Houston, TX
Aeth BC
1398
0.5
0.1
1.9
0.2
8/1/2012
6/11/2016
1410

Houston, TX
CSN EC
547
0.3
0.0
0.9
0.1
8/1/2012
3/29/2017
1701

Houston, TX
CSN OC
547
1.9
0.0
10.4
1.3
8/1/2012
3/29/2017
1701

Houston, TX
Sunset EC
697
0.7
0.0
2.6
0.5
8/1/2013
12/31/2016
1248
32%
Houston, TX
Sunset OC
697
2.4
0.1
8.2
1.1
8/1/2013
12/31/2016
1248
32%
Houston, TX
Sunset OptEC
696
0.3
-0.2
1.6
0.2
8/2/2013
12/31/2016
1247
56%
Houston, TX
Sunset TC
697
3.1
0.1
10.6
1.3
8/1/2013
12/31/2016
1248
56%
B.3
-------
Appendix B
o Sunset OC
SunsetEC
o Sunset TC
o Sunset OptEC
o Sunset OC
Sunset EC
o Sunset TC
Sunset OptEC
Figure B-l. Time series of Sunset data available in AQS (top) and data used in this work (bottom) for
Washington, D.C. Prior to May 2014, OC was not reported, so no data were included here for analysis.
Data in June 2015 and February-March 2016 were also excluded from analysis because of
operational issues associated with a software update in June and a heating coil malfunction at the end
of January 2016, which was not fixed until the end of March 2016.
B.4
-------
Appendix B
o Sunset OC
Sunset EC
o Sunset TC
03 ° OO
° ® JTl
o <9
o
, QD
O
O
CO
CO
b§°&
oO
o
o
o
o
o
$ «&.
o O
a°°
,0
b % !
° Q~.
O ^n0(9
u °<^p
^oo

o
OOo
o o
?o
gp ^
o o
o
o
o
<9 o
° °n
O O
o <§>o
%
oM i % * §8
oho O ® ° "
o
<9 J>
CD
OO oQ
«S no
cP„orLcO
?>
o o
§ o
o
o
%
OGD
s,
@ °
o
cv
o
o
rv
o
¦I

o
o
o
-------
Appendix B
Q) O ° °DO
o o
<9 Q
o o
Pi
Q. ^ cr
_c> OJ
CO ^ ^
i #
s1
Co

c
-?


s1
Q. i C
,
c
o
CO
CD
o
c
o
o
5-
Oo o

o
©
o
'1?
°o°
X °



So
o
§8
° to
o Sunset OC
Sunset EC
o Sunset TC
o Sunset OptEC
o
o
-------
Appendix B
o Sunset OC
o Sunset EC
o Sunset OptEC
15-
1? 10-
£=
O
£=
0)
O
£=
O
O
5-
o
o

©
o

o _
o

o
Q °0
o O o ~ o(
O 0
5®°C
o
OO '
. o
&po0oi
9 °rf3®@00
pooSt
o
o
i
a o
o Sunset OC
o Sunset EC
o Sunset OptEC
o
o
CV
o
o
o
o
o
CV
o
c
-?
#

8-
CO
£
c
Figure B-4. Time series of Sunset data available in AQS (top) and data used in this work
(bottom) for Rubidoux. In May 2014, there was a clear shift in OC upward and a shift of EC
downward, and these data were excluded from analysis. During these periods, operators found
leaks in the sampling line and the oven was replaced twice.
B.7
-------
Appendix B
8 H9
c\/ ro co co 05 co CO
(O^lOtOlOlOCOCOCOCOCO
o	Sunset OC
o	Sunset EC
o	Sunset TC
o	Sunset OptEC
oooooooooooooooooooooooooooo
C\I C\j C\j C\j C\j C\j C\j C\j c\f C\j C\j C\j C\j C\j C\j C\j C\j C\j C\j C\j C\j C\j C\j C\j C\i C\j C\j C\j
O ~Q

o
o
o
o
o
o
CV
'V
tv
IV
'V
IV
IV
tV
IV
tv
IV
IV
c V
tv
tv
tv
tv
lV
c V
tv
tv
tv
C Jb
£ "? ^ £
^ Q.
SP	a>
^ C/D
c
-?
£
^ Q.
^ CO
c
£ £
¦=> Q.	C
£• ^ 8- g ¦» g
o	Sunset OC
o	Sunset EC
o	Sunset TC
o	Sunset OptEC
Figure B-5. Time series of Sunset data available in AQS (top) and data used in this work (bottom) for
St. Louis. Data prior to September 2013 were excluded from analysis as this was a "warm-up" period
when operations were getting settled. There was a sudden shift in OC concentrations starting in
January 2015 when the filter was stuck and new calibration calculations were put into place. During
March 2015-January 2016, operators suspected contamination, adjusted the thermocouple, and
installed a new photodetector. However, data did not return to "normal" until after the oven was
replaced in January 2016. There were additional issues with keeping the flow steady in June-July 2016.
B.8
-------
Appendix B
o Sunset OC
SunsetEC
o Sunset TC
o Sunset OptEC
O Sunset OC
Sunset EC
o Sunset TC
Sunset OptEC
Figure B-6. Time series of Sunset data available in AQS (top) and data used in this work
(bottom) for Houston. Data prior to December 2014 were excluded since older software was
used to determine OC/EC and OptEC, the NDIR malfunctioned and was replaced twice, the
oven thermocouple malfunctioned and was replaced, there were leaks, and the instrument was
sent back to Sunset twice for maintenance. Data during May-August 2015 and in July-August
2016 had an unusual shift in OC, and EC was near zero, which were not seen in collocated
measurements, and which occurred when there were leaks in the sampling line.
B.9
-------
Appendix B
£3 Sunset OC/CSN OC
to
c 4
.2 1
03
b
O
O
"si- O
Csl ^1
M—
O
03
q:

3*2
$§
-Q O
CJ r-
Q:
i

r»-
O Q3
Q ;?
c II
£r<=
0><0
-c o
CO 1—
cd
. to
2 "
Co c
8 v
.o
CJ-

-W"/i
s *
0
CO ^
T-.
1
<0
v-.
£?
OS
> F3
¦? <0
- II
CO c
i5!"
co sy
(0 CD
-'oj
?5"
/» to
„ y*»
c n
¦8 c
i? co
^ V*-
T"«
1
V
T"»~
£v
-------
0
9
8
7
6
5
4
3
2
1
0
Appendix B
^ Sunset EC/CSN EC

3' *
&
§¦-

X—
qd
v--
O CD
- ?V
C '/
5 c-
Q><£>
¦£¦ \
-cr o
CO V-
v
T-—
o
» 1^
^ 5"
CO ^
OJ oj
-'oj
CV
r*Z Cv
A-. /
Site
Figure B-8. Box plot of daily Sunset EC/CSN EC ratios by site.
B.ll
-------
^ Sunset OptEC/CSN EC
5T (o	O	Oco	^ <~o	V-
O /v.	Q *-	,S 05	5	T-	,
-------
Appendix B
$ Sunset OptEC/Aeth BC
O co
Q T-
- 5°
c a
£ c
Q)CO
-c rv
CO r-
V
V-»
C5
C?
6^
-'C
Co ^
T-.
Y—.
&
O)
g /T
£ c
-------
Appendix B

16

14

12

10

8

6

4

2

0

16

14

12

10

8

6

4

2

0

16
„	„
14
m
12
E
10

8

C
=L
O

4
c
2
o
0
CO
16
1—
14
c
12
QJ
10
u
8
c
6
o
4
u
2

0

16

14

12

10

8

6

4

2

0

16

14

12

10

8

6

4

2

0
Rubidoux, CA
Washington, DC
: i
Chicago, IL
St. Louis, MO
Houston, TX
0 1 2 3 4 5 6 7	8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
; • :	Las Vegas, NV
¦ • t *	• .........
•	§ •	.j........j
Ul	fill!! 11


_L
10 12 14 16 18 20 22
Hour
Figure B-ll. Box plot of hourly Sunset OC by site.
B.14
-------
Appendix B
DO
C
o
'a-i
fO
i—
4->
c
QJ
u
c
O
o
Rubidoux, CA
Washington, DC
i i • i

Chicago, IL
St. Louis, MO
i * ' •
• . • •
~ 88
Houston, TX .
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
; La§ Vegas, NV

0 2 4 6 8 10 12 14 16 18 20 22
Hour
Figure B-12. Box plot of hourly Sunset tC by site
B.15
-------
Appendix B

6

5

4

3

2

1

0

6

5

4

3

2

1


m
0
E


6
Qfl

=L
5
-—*

c
4
o
3


CO
1—
2


c
1
QJ

u
0
c

o
6
u


5

4

3

2

1

0

6

5

4

3

2

1

0
Rubidoux, CA
I Washington, DC
St. Louis, MO
ill!
E±j$E±3
Houston, TX
• ¦	w	w
m *	m	*	#	*
s # .	i!"	.	1 : •
UUUUUaUuiUiUU
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Las Vegas, NV
i
a a
±ii
10 12 14
Hour
16 18 20 22
Figure B-13. Box plot of hourly Sunset OptEC by site.
B.16
-------
Appendix B
!
' » !
it.*!
• •
Washington, DC
* •
. !
; * !
• : i
¦ • • *
i ! i ! I
E
DO
C
.2 3
4_j
CO
c
QJ
o
c
o
U
•	i
•	i
: . §t. Louis, MO
: : : i
i
: i
•
¦	1 : i
I ! •
• 1 «

*	i : !
:	i ! i !
*	I
*	-
t
Houston, TX
. •
, j ; » . .

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hour
Figure B-14. Box plot of hourly Aethalometer BC by site.
B.17
-------
-------
United States	Office of Air Quality Planning and Standards	Publication No. EPA-454/R-19-005
Environmental Protection	Air Quality Assessment Division	May 2019
Agency	Research Triangle Park, NC
-------