Chemical Speciation Network (CSN)
Annual Quality Report
Samples Collected January 1, 2017 through December 31, 2017
Prepared for:
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
EPA Contract No. EP-D-15-020
Prepared by:
Air Quality Research Center
University of California, Davis
One Shields Avenue
Davis, CA 95616
January 20, 2020
UCDAVIS
AIR QUALITY RESEARCH CENTER
-------
Table of Contents
1. Executive Summary 3
1.1 Introduction 3
1.2 Data Quality Overview and Issues 3
2. Summary of Laboratory Operation Issues 4
2.1 DRI Ion Analysis Laboratory 4
2.1.1 Analysis Delays 4
2.2 UC Davis X-Ray Fluorescence Laboratory 4
2.2.1 Vanadium 4
2.2.2 Zinc 4
2.2.3 Calcium 4
2.2.4 Trace Element MDLs 4
2.3 DRI Thermal/Optical Analysis Laboratory 5
2.3.1 Analysis Delays 5
2.3.2 QC Criteria Failures 5
3. Quality Issues and Corrective Actions 5
3.1 Data Quality 5
3.1.1 Completeness 5
3.1.2 Comparability and Analytical Precision 7
3.1.3 Blanks 10
3.2 Corrective Actions 19
3.2.1 Elemental Analysis 20
3.2.2 Ion Analysis 23
3.2.3 Carbon Analysis 23
3.2.4 Data Processing 23
4. Laboratory Quality Control Summaries 24
4.1 DRI Ion Analysis Laboratory 24
4.1.1 Summary of QC Checks and Statistics 24
4.1.2 Summary of QC Results 25
4.1.3 Determination of Uncertainties and Method Detection Limits 31
4.1.4 Audits, Performance Evaluations, Training, and Accreditations 31
4.1.5 Summary of Filter Field Blanks 32
4.2 UC Davis X-Ray Fluorescence (XRF) Laboratory 32
4.2.1 Summary of QC Checks and Statistics 33
4.2.2 Summary of QC Results 34
4.2.3 Determination of Uncertainties and Method Detection Limits 50
4.2.4 Audits, Performance Evaluations, Training, and Accreditations 50
4.2.5 Summary of Filter Field Blanks 50
4.3 DRI Carbon Laboratory 52
4.3.1 Summary of QC Checks and Statistics 52
4.3.2 Summary of QC Results 53
4.3.3 Determination of Uncertainties and Method Detection Limits 67
4.3.4 Audits, Performance Evaluations, Training, and Accreditations 67
4.3.5 Summary of Filter Blanks 67
Page | 1
-------
5. Data Management and Reporting 68
5.1 Number of Events Posted to AQS 68
6. Quality Assurance and Data Validation 69
6.1 QAPP Revisions 69
6.2 SOP Revisions 69
6.3 Summary of Internal Q A Activities 69
6.4 Data Validation and Review 70
6.4.1 Summary of Monthly Data Validation Review Results 70
6.5 Uncertainty Estimates and Collocated Precision Summary Statistics 79
7. References 88
Page | 2
-------
1. Executive Summary
1.1 Introduction
The University of California—Davis (UC Davis) Air Quality Research Center summarizes
quality assurance (QA) annually in this report as a contract deliverable for the Chemical
Speciation Network (CSN) program (contract #EP-D-15-020). The primary objectives of this
report are:
1. Provide the U.S. Environmental Protection Agency (EPA) and other potential users with
graphical and tabular illustrations of quality control (QC) for species measured within the
network.
2. Identify and highlight observations of interest that may have short- or long-term impact
on data quality across the network or at particular sites.
3. Serve as a record and tool for ongoing UC Davis QA efforts.
Each network site includes two samplers: (1) URG 3000N carbon sampler (URG Corporation;
Chapel Hill, NC) for collection of particulate matter on quartz filters; and (2) Met One SASS or
SuperSASS (Met One Instruments, Inc; Grants Pass, OR) for collection of particulate matter on
polytetrafluoroethylene (PTFE) filters and nylon filters. The following analyses are performed:
• PTFE filters: analyzed at UC Davis using energy dispersive X-ray fluorescence (EDXRF)
for a suite of 33 elements.
• Nylon filters: analyzed at the Desert Research Institute (DRI) using ion chromatography
(IC) for a suite of six ions.
• Quartz filters: analyzed at the Desert Research Institute (DRI) for organic and elemental
carbon, including carbon fractions, using Thermal Optical Analysis (TOA).
Unless otherwise noted, data and discussions included in this report cover samples collected
during the time period January 1, 2017 through December 31, 2017.
1.2 Data Quality Overview and Issues
Section 4 of this report provides laboratory performance details for each of the analytical
measurement techniques. The laboratory performance is detailed in Section 4.1 (DRI Ion
Analysis Laboratory), Section 4.2 (UC Davis X-Ray Fluorescence Laboratory), and Section 4.3
(DRI Thermal/Optical Analysis Laboratory).
Across the network, completeness - determined by the total number of valid samples relative to
the total number of scheduled samples - was 96.3% for PTFE filters, 96.2% for nylon filters, and
93.1% for quartz filters. As detailed in Section 3.1.1, there were nine sites with less than 75%
completeness.
No Technical Systems Audit (TSA) of UC Davis was performed by the EPA in 2017.
Page | 3
-------
2. Summary of Laboratory Operation Issues
2.1 DRI Ion Analysis Laboratory
2.1.1 Analysis Delays
Deliveries of analysis data from DRI to UC Davis were delayed, contributing to noncompliance
with 120 days requirement for delivery of data to AQS following receipt of filters by analytical
laboratories. See Section 5.1.
2.2 UC Davis X-Ray Fluorescence Laboratory
2.2.1 Vanadium
During a portion of this reporting period XRF analyses of vanadium were overestimated by
about 30%. Results from an inter-laboratory comparison, confirmed by further comparison with
ICP-MS analysis, revealed that vanadium calibrations based on commercial standards for
samples collected from January 2017 through October 2017 resulted in erroneously high
measurements. XRF calibrations performed in January 2018 utilized newly purchased, re-
certified, and UCD-produced vanadium standards thus eliminating the overestimation beginning
with samples collected November 2017.
For further detail and corrective actions see Section 3.2.1.1.
2.2.2 Zinc
For analyses performed during this reporting period, periodic zinc contamination was observed
on the daily QC laboratory blank and daily QC multi-elemental reference sample on the EDXRF
instruments, XRF-1 and XRF-4. The cause of these random contamination events was
determined to be related to the instrument design, specifically operation of the sample changer.
Samples analyzed during this period were checked for unusually high zinc mass loadings
compared to site specific and network wide historical values. Ten samples in 2017 with unusual
Zn mass loadings were reanalyzed and for cases where the original result had contamination the
reanalysis results were reported.
For further detail see Sections 3.2.1.3 and Section 4.2.2.1.
2.2.3 Calcium
During this reporting period, both XRF-1 and XRF-4 showed gradual increase in calcium mass
loadings of their QC samples. The calcium buildup is likely caused by atmospheric deposition or
instrument wear on these filters, which are analyzed daily and remain in the instruments' sample
changers indefinitely. This gradual buildup of calcium is not expected on actual samples which
are loaded and analyzed once. However, samples are monitored for unusually high calcium
values and reanalyzed as necessary.
For further detail see Section 3.2.1.2 and Section 4.2.2.1.
2.2.4 Trace Element MDLs
Quantification of trace elements is an ongoing challenge using EDXRF. In some cases, for
example in the case of lead (Pb), the MDL increased corresponding with the November 2015
contract transition, due in part to analytical method changes. UC Davis is exploring ways to
Page | 4
-------
optimize the XRF application, including the potential for utilizing alternative secondary targets,
in an effort to improve MDLs for trace elements.
2.3 DRI Thermal/Optical Analysis Laboratory
2.3.1 Analysis Delays
Deliveries of analysis data from DRI to UC Davis were delayed, contributing to noncompliance
with 120 days requirement for delivery of data to AQS following receipt of filters by analytical
laboratories. See Section 5.1.
2.3.2 QC Criteria Failures
In some cases, DRI analyzed samples while instruments were operating outside of the defined
QC criteria. There were instances of impacted data for samples collected during 2017.
Per direction from the EPA, these data were redelivered to AQS with QX (Does Not Meet QC
Criteria) qualifier flag applied.
For further detail see Section 3.2.3.1 and Section 4.3.2.
3. Quality Issues and Corrective Actions
3.1 Data Quality
3.1.1 Completeness
Completeness is evaluated network wide by filter type and determined by the total number of
valid samples relative to the total number of collected and scheduled samples (Table 3.1-1). The
completeness is comparable for PTFE and nylon filters, which are both collected by the Met One
SASS / Super SASS sampler; however, the number of invalid samples is higher for quartz filters,
which are collected by the URG sampler. Quartz filters flagged with the QX qualifier, as detailed
in Section 2.3.2, were not invalidated and are included in the count of valid samples.
Table 3.1-1: Network sample completeness by filter type, January 2017 through December 2017. The total number
of scheduled samples is calculated from the sampling schedule (does not include field blanks). The total number of
collected samples is the actual number of samples collected in the field.
Filter
Type
Total Number
of Scheduled
Samples
Total Number
of Collected
Samples
Number
of Valid
Samples
Number
of Invalid
Samples
% Valid
(relative to #
collected samples)
% Valid
(relative to # of
scheduled samples)
PTFE
13,329
13,336
12,844
492
96.3
96.4
Nylon
13,329
13,336
12,828
508
96.2
96.2
Quartz
13,329
13,320
12,398
922
93.1
93.0
Across the network there were nine sites with sample completeness less than 75% for at least one
filter type (Table 3.1-2). Seven of the nine cases had low completeness resulting from invalid
quartz filters.
Page|5
-------
Table 3.1-2: Network sites with less than 75% sample completeness (relative to the number of collected samples)
for at least one filter type, January 2017 through December 2017. For each filter type, the percentage of different
null codes is listed relative to the total number of null codes per site. For null code definitions, see Table 3.1-3.
AQS ID #
Location
Completeness (%)
Null Codes
PTFE
Nylon
Quartz
PTFE
Nylon
Quartz
12-011-0034-5
(Region 4)
Broward County, FL
(NCore/STN)
65.6
58.2
54.9
BA (79%)
Other (21%)
BA (63%)
AN (16%)
Other (21%)
BA (64%)
AN (27%)
Other (9%)
37-067-0022-5
(Region 4)
Winston-Salem, NC
95.1
95.1
57.3
AL (67%)
AV (33%)
AL (67%)
AV (33%)
AN (58%)
AH (23%)
Other (19%)
72-021-0010-5
(Region 2)
Bayamon Puerto Rico
' (NCore/STN)
87.9
86.8
59.3
BI (36%)
AF (27%)
Other (37%)
BI (33%)
AF (25%)
Other (42%)
AH (59%)
AN (11%)
Other (30%)
44-007-1010-5
(Region 1)
East Providence, RI
(NCore/STN)
96.7
97.5
64.2
AB (25%)
AG (25%)
AH (25%)
AJ (25%)
AB (33%)
AG (33%)
AH (33%)
AN (39%)
AH (36%)
Other (25%)
45-079-0007-5
(Region 4)
Parklane, SC
(NCore/STN)
95.9
95.9
67.2
AF (60%)
AH (20%)
AO (20%)
AF (60%)
AH (20%)
AO (20%)
AN (55%)
AH (25%)
Other (20%)
39-153-0023-5
(Region 5)
Akron, OH
67.9
77.4
88.9
AN (29%)
BA (18%)
BJ (18%)
Other (35%)
BA (25%)
BJ (25%)
Other (50%)
BJ (50%)
Other (50%)
53-077-0009-5
(Region 10)
Yakima, WA
100
100
70.5
...
...
AH (67%)
AN (33%)
30-093-0005-5
(Region 8)
Butte, MT
95.1
98.4
72.1
AQ (67%)
BJ (33%)
BJ (100%)
AN (59%)
AH (24%)
Other (17%)
08-031-0026-5
(Region 8)
La Casa, CO
(NCore/STN)
73.8
75.4
88.5
AV (84%)
Other (16%)
AV (90%)
AQ (7%)
AP (3%)
AN (57%)
AV (21%)
Other (22%)
Samples can be invalidated for a variety of reasons as detailed in the UCD CSN TI801C: CSN
Data Validation and the Data Validation for the Chemical Speciation Network guide. Null codes
indicate the reasons for invalidation (Table 3.1-3).
Page|6
-------
Table 3.1-3: Number and type of null codes applied to SASS and URG samples from January 2017 through
December 2017. Codes are ordered by frequency of occurrence.
Null
Code
SASS
PTFE
SASS
Nylon
URG
Quartz
Null Code Description
AP
1
1
0
Vandalism
TS
1
1
0
Holding time or transport temperature is out of specs
AZ
1
1
1
QC Audit
BE
1
1
1
Building/site repair
DA
0
0
1
Aberrant data (corrupt files, aberrant chromatography, spikes, shifts)
AK
0
0
2
Filter leak
AM
1
1
4
Miscellaneous void
BB
6
6
5
Unable to reach site
SA
5
5
5
Storm approaching
AL
5
5
7
Voided by operator
AR
8
6
7
Lab error
AC
9
9
8
Construction/repairs in area
AI
2
2
8
Insufficient data (cannot calculate)
AQ
14
15
8
Collection error
AJ
12
6
10
Filter damage
BI
10
10
11
Lost or damaged in transit
AO
12
12
12
Bad weather
sv
6
7
16
Sample volume out of limits
AG
9
9
21
Sample time out of limits
AB
33
33
35
Technician unavailable
BJ
44
39
39
Operator error
BA
44
43
53
Maintenance/routine repairs
AV
97
97
65
Power failure
AF*
71
72
119
Scheduled but not collected
AH
40
53
221
Sample flow rate or CV out of limits
AN
112
126
341
Machine malfunction
* Filters that receive this flag were intended for sampling and shipped to the site, but were not sampled.
3.1.2 Comparability and Analytical Precision
Analytical precision is evaluated by comparing data from repeat analyses, where two analyses
are performed on the same sample using either the same instrument (duplicate) or different
instruments (replicate). Reliable laboratory measurements should be repeatable with good
precision. Analytical precision includes only the uncertainties associated with the laboratory
handling and analysis, whereas collocated precision (Section 6.5) also includes the uncertainties
associated with sample preparation, field handling, and sample collection. Analytical precision is
used internally as a QC tool.
Comparisons of ion mass loadings from repeat analyses (replicates and/or duplicates) on nylon
filters analyzed by IC show generally good agreement (Figure 3.1-1).
Page|7
-------
Figure 3.1-1: Ion repeat analysis (replicates and/or duplicates) results; data from samples collected during 2017.
Ammonium
Chloride
10
2 .
cn
C/>
>»
03
C
CD
1 .
£ 10
m a r\0 -
10u
ro
cu
Cl
CD
cu
10
¦1.
10
10
-2
¦3 .
y = -0.09 + 1.01x
r2 = 1
/
/
/
/
'
s
*
J*
-r
10
2 .
10
1.
10
o.
10
-1.
y = -0 + 1x
r2 = 1
. /
~
/
/
CD
Q.
CD
i
CT)
C
"D
TO
O
_l
cr)
CO
cc
10~3 10~2 10~1 10° 101 102
Potassium Ion
10
1.
10
o.
10
•1.
10"
10
-3 .
y = 0.01 + 1x
9 - « .
r =o.y4
s
~
s
*
s
,t
Oo
•
•
10
2 .
10
1.
10
o.
10
-1 _
10
-2 .
Nitrate
o -
o -
o
—X
-
102
Sodium Ion
y = -0 + 0.99x
r2 = 0.99
/
r
/
'
s
/ ®
/
JT
10
2.
10
1 .
10
o.
10
-1_
y = -0.04 + 1x
r2 = 0.99
y
*
10~3 10~2 10~1 10° 10'
10~2 10"' 10u 10' 10' 10"'
Mass Loading (jLtg per filter): routine filter
10u
10
Comparison of carbon mass loadings from repeat analyses (replicates and/or duplicates) on
quartz filters analyzed by TOA generally show agreement (Figure 3.1-2).
Page|8
-------
Figure 3.1-2: Carbon repeat analysis (replicates and/or duplicates) results; data from samples collected during 2017.
Elemental carbon (EC) fractions are indicated as (1) through (3). organic carbon (OC) fractions are indicated as (1)
through (4). Organic pyrolyzed (OP), elemental carbon (EC), and organic carbon (OC) are shown by reflectance (R)
and transmittance (T).
EC1
EC2
EC3
ECR
10
10
10'
10
,2-
1 .
0 .
1 .
~~i r~
10~1 10° 101 102
y = 0.02 + 0.89X
r2 = 0.82:
'8 *•
©
10
10
10'
10
2. y = 0.33 + 0.97X
1 .
,0.
1 .
r = 0.97
10~2 10~1 10° 10'
OC1
10~2 10~1 10° 101
OC2
10"210"1 10° 101 102
102-
y = 0.5 + 0.88x
r2 = 0.9
101 -
o
O
10"1-
^ .. ":'v
- ¦ •
nr-
® ~, •
a)
Q_
D5
3
CD
C
10~1 10° 101 102
OC4
10~2 10~1 10° 101 102
OCR
10~1 10° 101 102
OCT
OC3
y = -0.15 + 1.01* *5
2" r2 = 0.98 '
10"1 10° 101 102
OPR
10
10
10'
,2 J
1 _
0 .
"O
CD
° 10'1J
to
CO
03
y = 0.32 + 0.99x
r2 = 0.96
10°
10z
10'
10°
10
10
10'
10"
10
-2 .
y =-0.04 + 1.03x
r2 = 0.94
•"8 4- " *
• •
10"1 10° 101 102
OPT
10° 101 102 103
10° 101 102 103 10"210"1 10° 101 102
T I V
10"210"110° 101 102
Mass Loading (jug per filter): routine filter
Repeat analyses (replicates and/or duplicates) are not performed by XRF for the routine CSN
samples. Rather, reanalysis is performed on the same set of filters on a monthly basis to assess
both the short- and long-term stability of the XRF m easurements as described in UCD CSN SOP
§$02: XRF Analysis. See Section 4.2.2.4.
Page|9
-------
3.1.3 Blanks
Field blanks are an integral part of the QC process and field blank analysis results are used to
artifact correct the sampled filters as part of the concentration calculation (see Section 3.1.3.1).
Artifacts can result from initial contamination in the filter material, contamination during
handling and analysis, and adsorption of gases during sampling and handling. Additionally, field
blanks are used to calculate method detection limits (MDLs; see Section 3.1.3.2)
There is some variability in field blank mass loadings by species and month, as shown in Figure
3.1-3 through 3.1-8 for ions measured from nylon filters, and Figure 3.1-9 and 3.1-10 for organic
carbon and elemental carbon, respectively, measured from quartz filters. The 10th percentile of
network sample concentrations is indicated in Figure 3.1-3 through Figure 3.1-10 to facilitate
understanding of field blank concentrations in context of network sample concentrations; 90% of
network sample concentrations fall above the indicated 10th percentile. As part of the validation
process (see Section 6), field blank outliers are investigated but are only invalidated if there is
cause to do so. Artifact correction (Section 3.1.3.1) and MDL (Section 3.1.3.2) calculation
methods are robust to accommodate occasional outliers.
Figure 3.1-3: Time series of ammonium measured on nylon filter field blanks (FB), for field blanks collected
January 2016 through December 2017. Gaps in time series are present when no nylon filter field blanks were
collected. The colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper and lower limits of the
boxes represent 75th and 25th percentile, respectively. The whiskers extend to the most extreme data point which is
no more than 1.5 x the length of the box away from the box. The dots are all of the points that lay outside the
whiskers. The black horizontal dashes indicate the 10th percentile of network samples.
a>
O)
£
Ol
£ 2
¦o
(0
o
-I
(/)
(A
(0
0
10th %-ile of
network samples
Field blanks
(year)
$ 2016
$ 2017
-
2016 01 2016 07 2017 01 2017 07
Time (Year Month)
Page|10
-------
Figure 3.1-4: Time series of chloride measured on nylon filter field blanks (FB) for field blanks collected January
2016 through December 2017. Gaps in time series are present when no nylon filter field blanks were collected. The
colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper and lower limits of the boxes
represent 75th and 25th percentile, respectively. The whiskers extend to the most extreme data point which is no more
than 1.5 x the length of the box away from the box. The dots are all of the points that lay outside the whiskers. The
black horizontal dashes indicate the 10th percentile of network samples.
4-
10th %-ile of
network samples
• • •
T
Field blanks
(year)
$ 2016
2017
i • i i • : :
2016 01
2016 07 2017 01
Time (Year Month)
2017 07
Page|11
-------
Figure 3.1-5: Time series of nitrate measured on nylon filter field blanks (FB), for field blanks collected January
2016 through December 2017. Gaps in time series are present when no nylon filter field blanks were collected. The
colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper and lower limits of the boxes
represent 75th and 25th percentile, respectively. The whiskers extend to the most extreme data point which is no more
than 1.5 x the length of the box away from the box. The dots are all of the points that lay outside the whiskers. The
black horizontal dashes indicate the 10th percentile of network samples.
4-
l;
2016 01
2016 07 2017 01
Time (Year Month)
2017 07
10th %-ile of
network samples
Field blanks
(year)
$ 2016
$ 2017
Page|12
-------
Figure 3.1-6: Time series of potassium ion measured on nylon filter field blanks (FB), for field blanks collected
January 2016 through December 2017. Gaps in time series are present when no nylon filter field blanks were
collected. The colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper and lower limits of the
boxes represent 75th and 25th percentile, respectively. The whiskers extend to the most extreme data point which is
no more than 1.5 x the length of the box away from the box. The dots are all of the points that lay outside the
whiskers. The black horizontal dashes indicate the 10th percentile of network samples.
4-
10th %-ile of
network samples
Field blanks
(year)
$ 2016
2017
2016 01 2016 07 2017 01 2017 07
Time (Year Month)
Page| 13
-------
Figure 3.1-7: Time series of sodium ion measured on nylon filter field blanks (FB), for field blanks collected
January 2016 through December 2017. Gaps in time series are present when no nylon filter field blanks were
collected. The colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper and lower limits of the
boxes represent 75th and 25th percentile, respectively. The whiskers extend to the most extreme data point which is
no more than 1.5 x the length of the box away from the box. The dots are all of the points that lay outside the
whiskers. The black horizontal dashes indicate the 10th percentile of network samples.
4-
I I • ' • I
-iiii 1 i iJ
• 5*1
i i i s±s A
2016 01
2016 07 2017 01
Time (Year Month)
2017 07
10th %-ile of
network samples
Field blanks
(year)
$ 2016
$ 2017
Page|14
-------
Figure 3.1-8: Time series of sulfate measured on nylon filter field blanks (FB), for field blanks collected January
2016 through December 2017. Gaps in time series are present when no nylon filter field blanks were collected. The
colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper and lower limits of the boxes
represent 75th and 25th percentile, respectively. The whiskers extend to the most extreme data point which is no more
than 1.5 x the length of the box away from the box. The dots are all of the points that lay outside the whiskers. The
black horizontal dashes indicate the 10th percentile of network samples.
4-
10th %-ile of
network samples
Field blanks
(year)
$ 2016
$ 2017
2016 01
2016 07 2017 01
Time (Year Month)
2017 07
Page|15
-------
Figure 3.1-9: Time series of organic carbon by reflectance (OCR) measured on quartz filter field blanks (FB), for
field blanks collected January 2016 through December 2017. Gaps in time series are present when no quartz filter
field blanks were collected. The colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper and
lower limits of the boxes represent 75th and 25th percentile, respectively. The whiskers extend to the most extreme
data point which is no more than 1.5 x the length of the box away from the box. The dots are all of the points that lay
outside the whiskers. The black horizontal dashes indicate the 10th percentile of network samples.
40
-------
Figure 3.1-10: Time series of elemental carbon by reflectance (ECR) measured on quartz filter field blanks (FB),
for field blanks collected January 2016 through December 2017. Gaps in time series are present when no quartz
filter field blanks were collected. The colored (red, 2016; blue, 2017) horizontal lines indicate median, and the upper
and lower limits of the boxes represent 75th and 25th percentile, respectively. The whiskers extend to the most
extreme data point which is no more than 1.5 x the length of the box away from the box. The dots are all of the
points that lay outside the whiskers. The black horizontal dashes indicate the 10th percentile of network samples.
15-
10
0
10th %-ile of
network samples
Field blanks
(year)
$ 2016
S 2017
—
- -
_ —
•
•
•
1
•
•
•
• •
. • • •
lili
2016 01 2016 07 2017 01 2017 07
Time (Year Month)
3.1.3.1 Blank Correction
Blank correction is performed on data from all filter types (quartz, nylon, and PTFE) by
subtracting a rolling median value from at least 50 field blanks collected in and closest to the
sample month.
3.1.3.2 Method Detection Limits
Network wide method detection limits (MDLs) are updated monthly and delivered to AQS for
each species. A sufficient number of field and/or laboratory blanks must be available in order to
calculate MDLs representative of the network. For samples collected November 2015 through
January 2017, MDLs were calculated at follows:
Page|17
-------
• Elements: Calculated for each species as 3 x standard deviation of lab blanks.
Recalculated for each new lot of PTFE filters.
• Ions and carbon: Calculated monthly for each species as 3x standard deviation of field
blanks, using 50 nylon (for ions) or quartz (for carbon) field blanks collected in and
closest to the sampling month.
The method used for calculating MDLs has evolved as availability of field blanks has increased.
Beginning in March 2017, field blank collection increased to one field blank for each filter type
per site per month, allowing for a more robust MDL calculation method. For data from samples
collected February 2017 onward, the MDL calculation is harmonized for all analysis pathways,
calculated as 95th percentile minus median of field blanks, using 50 field blanks collected in or
closest to the sampling month for each respective filter type. For most cases, the MDLs
calculated using this method are higher than analytical MDLs calculated by the laboratories
using laboratory blanks, which are assigned as the MDL floor values. Field blanks capture
artifacts from both field and laboratory processes, with expectations that field blank mass
loadings are generally higher than lab blanks which have only been handled in a laboratory
environment and have less opportunity for mishandling and contamination. However, when the
MDL determined from field blanks is lower than the analytical MDL, the analytical MDL floor
value is assigned.
The average MDLs calculated using the updated method during this reporting period are
compared to the average MDLs calculated using the old method from the previous reporting
period (Table 3.1-4). MDLs calculated using the old and updated methods are similar for most
species; cases with differences >50% between the methods are highlighted in Table 3.1.4. The
updated MDL calculation results in MDL values that are more stable over time.
Table 3.1-4: Average method detection limits (MDLs) and percentage of reported data above the MDLs for all
species, calculated for data from samples collected November 2015 through December 2016 (previous reporting
period) and February 2017 through December 2017 (current reporting period). Elemental carbon (EC) fractions are
indicated as (1) through (3), organic carbon (OC) fractions are indicated as (1) through (4). Organic pyrolyzed (OP),
elemental carbon (EC), and organic carbon (OC) are shown by reflectance (R) and transmittance (T). Species shown
in bold have differences >50% between the old (November 2015 - December 2016) and updated (February 2017 -
December 2017) MDL calculation methods.
Species
November 2015 - December 2016
February 2017 - December 2017
Average MDL, jig/m3
% Above MDL
Average MDL, jig/m3
% Above MDL
Ag
0.019
1.4
0.017
2.7
A1
0.038
32.4
0.038
30.1
As
0.003
7.2
0.003
6.2
Ba
0.086
1.9
0.081
1.5
Br
0.005
17.7
0.005
15.8
Ca
0.027
65.1
0.034
56.8
Cd
0.024
0.7
0.016
2.9
Ce
0.116
0.9
0.096
1.3
CI
0.005
42.9
0.007
31.9
Co
0.003
1.5
0.003
0.9
Cr
0.004
14.4
0.004
20.1
Cs
0.078
0.5
0.056
2.5
Cu
0.009
18.9
0.011
10.9
Page|18
-------
Species
November 2015 - December 2016
February 2017 - December 2017
Average MDL, jig/m3
% Above MDL
Average MDL, jig/m3
% Above MDL
Fe
0.023
85.1
0.027
79.8
In
0.031
0.2
0.037
0.0
K
0.016
95.8
0.012
98.4
Mg
0.055
9.0
0.042
12.5
Mn
0.007
7.2
0.006
8.3
Na
0.070
27.3
0.088
20.9
Ni
0.002
11.1
0.002
14.7
P
0.002
9.9
0.002
8.0
Pb
0.015
4.7
0.012
7.4
Rb
0.008
1.1
0.009
0.3
S
0.009
99.4
0.005
99.5
Sb
0.047
1.1
0.040
2.3
Se
0.006
1.3
0.005
1.6
Si
0.015
90.3
0.020
83.7
Sn
0.046
0.9
0.050
0.5
Sr
0.007
2.7
0.007
2.2
Ti
0.003
45.8
0.003
41.4
V
0.002
5.5
0.002
7.9
Zn
0.004
78.0
0.003
78.9
Zr
0.037
0.9
0.036
1.0
Ammonium
0.015
80.7
0.006
81.6
Chloride*
—
—
0.047
60.1
Nitrate
0.095
89.7
0.036
98.5
Potassium Ion
0.008
90.5
0.047
29.5
Sodium Ion
0.043
53.2
0.016
66.6
Sulfate
0.144
96.1
0.047
99.4
Elemental Carbon (EC1)
0.011
99.5
0.007
99.5
Elemental Carbon (EC2)
0.010
95.7
0.009
95.5
Elemental Carbon (EC3)
0.002
3.6
0.002
3.6
Elemental Carbon (ECR)
0.017
99.1
0.013
99.4
Elemental Carbon (ECT)
0.014
98.6
0.012
98.9
Organic Carbon (OC1)
0.024
60.6
0.019
76.8
Organic Carbon (OC2)
0.050
98.9
0.036
99.5
Organic Carbon (OC3)
0.151
94.8
0.053
98.7
Organic Carbon (OC4)
0.031
99.3
0.012
99.7
Organic Carbon (OCR)
0.213
98.9
0.081
99.6
Organic Carbon (OCT)
0.216
99.0
0.083
99.6
Organic Pyrolyzed (OPR)
0.010
79.2
0.008
72.4
Organic Pyrolyzed (OPT)
0.013
95.8
0.010
93.9
* Chloride results were not reported until February 2017.
Page|19
-------
3.2 Corrective Actions
To ensure ongoing quality work, UC Davis reacts as quickly and decisively as possible to
unacceptable changes in data quality. These reactions are usually in the form of investigations,
and, if necessary, corrective actions. The following subsections describe significant corrective
actions undertaken for data from samples collected during 2017.
3.2.1 Elemental Analysis
3.2.1.1 Vanadium
As discussed in Section 2.2.1, during a portion of this reporting period (samples collected from
January 2017 through October 2017) XRF analyses of vanadium (V) were overestimated by
about 30%. This issue and resulting corrective actions are detailed here.
Reported elemental concentrations rest on linear calibrations of the Panalytical Epsilon 5
instruments since their implementation for EDXRF analysis at UC Davis. The calibration factors
are derived from observed instrumental responses to a variety of certified standards and
reference materials of known composition. UC Davis certifies and calibrates with some
standards created in their own laboratory, aerosolizing known materials and collecting them on
PTFE filters using IMPROVE (Interagency Monitoring of PROtected Visual Environments)
samplers and/or Met One samplers utilized for CSN. The resulting deposits better mimic actual
ambient samples than the vacuum-deposited thin-film membranes traditionally obtained from
commercial vendors. Such in-house standards are certified for 18 of the elements reported for
CSN, including Na, Al, Si, S, CI, K, Ca, Ti, V, Mn, Fe, Co, Ni, Cu, Zn, Se, Sr, and Pb. However,
XRF calibrations for vanadium were based solely on two commercial standards for analysis of
samples collected through October 2017.
During inter-laboratory comparison studies of novel multi-element reference materials (ME-RM)
under development, it was discovered that UC Davis XRF results for vanadium were higher than
expected by about 30-50%, while results from other laboratories (including analyses by XRF,
PIXE, and ICP-MS) were within 20% of expected values (Figure 3.2-1).
Figure 3.2-1: Inter-laboratory comparison of multi-element reference materials for vanadium where the UC Davis
results are shown as filled red circles and results from other laboratories are shown as circles and triangles.
1.8
1.6
1.4
M-
0)
tc
(U
1
0.8
0.6
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Reference, ng/cm2
OUCD-XRF OXRF1 XRF2 OXRF3 XRF4 OXRF5 OXRF6 OXRF7 O PIXE ~ ICP1 A ICP2 A ICP3
o
O
2) oo
O
9)
So®
oO
W ^
I 9
U
0
0
cP
0 0
o°
— ^
O 4
1
©
g> ^
•k
o_
2
0
i~
A
i
Page|20
-------
Overestimation by UC Davis XRF analysis was confirmed by further comparisons with ICP-MS
analysis by a collaborating laboratory (Figure 3.2-2).
Figure 3.2-2: Comparison of multi-element reference materials for vanadium at UC Davis using EDXRF and a
collaborating laboratory using ICP-MS.
1.6
1.5
1.4
C 13
0) 1,3
•4-12
QJ
0C
LL 1.1
ce
X
1
0.9
0.8
o
o
o
o
o
OC
9> °
OUCD-EC
oirp-MS
XRF
Q
o
o
c
>
o
o
O o
o
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
Reference, ng/cm2
UC Davis' ability to design and generate custom reference materials provided further
confirmation with single-compound (vanadyl sulfate) standards of known hydration, whose
loadings could be gravimetrically certified (Figure 3.2-3).
Figure 3.2-3: UC Davis XRF results for single vanadyl sulfate standards that conform to expectations for sulfur (red
circle) but are high for vanadium (blue circle).
1.5
1.4
a>
1.3
u
c
CD
L_
1.2
HI
cc
1.1
Li-
ar
X
1
0.9
0.8
n
o
O O
0
0
o
0
0
O
O
O
O
O r> (
D
8
0
0
o°
°o 00
O
0
0
Reference, ug/cm2
Inter-laboratory comparison studies of UC Davis ME-RMs, together with UC Davis' custom
single-compound standards, converged to indicate that the existing calibration of the UC
Davis Panalytical Epsilon-5 instruments for vanadium was about 30% high. Continuity of the
historical vanadium record was already tested, when the newer Epsilon 5 (E5) instruments
were used to reanalyze the 15-year archive of samples collected from 1995 to 2009 at Great
Smoky Mountains NP (GRSM) as part of the IMPROVE program. These had previously been
analyzed and reported from the UC Davis-built copper- and molybdenum-anode XRF
systems, which had been calibrated using a different set of standards. The new measurements
Page | 21
-------
were about 30% higher than those previously reported (Figure 3.2-4).
Figure 3.2-4: Reanalyses by Panalytical Epsilon 5 (E5) of 1995 - 2009 samples from Great Smoky Mountains
National Park (GRSM, IMPROVE) previously analyzed versus earlier Cu-Mo XRF system.
40
IN
E
u.
1.1
DC
X
1
0.9
•
•
~
o
~£, o
20 , 40
Certified, |ig/cmz
60
3.2.1.2
Calcium
As discussed in Section 2.2.3 and Section 4.2.2.1, laboratory QC filters that are exposed to the
environment for prolonged periods for repeat analysis show a general increase in calcium mass
loadings. These increases are not observed if the QC filter is cleaned with air or replaced with a
new filter. The contamination appears to occur mostly on filters that are analyzed multiple times
Page|22
-------
and therefore should not impact routine samples or field blanks. Even so, CSN samples and field
blank filters were monitored during QA checks for calcium contamination. Six filters identified
as having potential contamination were reanalyzed and reanalysis results were reported
accordingly.
3.2.1.3 Zinc
As discussed in Section 2.2.2 and Section 4.2.2.1, the design of the sample changer arm on the
EDXRF instruments results in sporadic cases of zinc contamination. Ten filters identified as
having potential contamination were reanalyzed and reanalysis results were reported
accordingly.
3.2.2 Ion Analysis
3.2.2.1 Chloride
As discussed in the UC Davis Chemical Speciation Network 2016 Annual Report
(https://aqrc.ucdavis.edu/csn-documentation), a chloride contamination issue was discovered in
the network beginning in November 2015; the contamination was traced to cleaning wipes used
in the filter handling laboratory. Following resolution of the issue, chloride data were reported to
AQS beginning with data from samples collected during February 2017.
See Section 4.1 for further details.
3.2.3 Carbon Analysis
3.2.3.1 QC Criteria Failures
As discussed in Section 2.3.2, in some instances DRI analyzed samples while instruments were
operating outside of the defined QC criteria. For these cases, an internal QC flag was applied.
Per direction from the EPA, these data were redelivered to AQS with QX (Does Not Meet QC
Criteria) qualifier flag applied.
For further detail see Section 2.3.2 and Section 4.3.2.
3.2.3.2 Carbon Data Reprocessing
All reportable CSN carbon analyses were performed by DRI, a subcontractor to UC Davis on the
CSN contract. After examination of data from the IMPROVE and CSN programs, DRI
determined the Model 2015 (used for analyses from January 1, 2016 onward) carbon signal
integrations threshold differed from that of the Model 2001 (used for analysis prior to January 1,
2016). DRI reprocessed CSN carbon data from January 1, 2016 through September 13, 2017, and
analysis of the differences between the original and reprocessed data was prepared by UC Davis.
The EPA determined that reprocessed carbon results would not be delivered to AQS because the
impact on the data was minor. UC Davis prepared a data advisory: https://aqrc.ucdavis.edu/csn-
documentation.
3.2.4 Data Processing
3.2.4.1 Data Flagging Modifications
Data are flagged as part of the CSN data validation process as detailed in the UCD CSN TI801C:
CSN Data Validation and the Data Validation for the Chemical Speciation Network guide. Flags
are applied throughout the sampling, filter handling, analysis, and validation processes, using
Page|23
-------
automated checks or on a case-by-case basis. The use and application of flags evolves as
problems are identified and remedied, and also in response to process improvements that are
implemented to improve the quality and consistency of data for the end user.
4. Laboratory Quality Control Summaries
4.1 DRI Ion Analysis Laboratory
The DRI Ion Analysis Laboratory, as a subcontractor to UC Davis, received and analyzed nylon
filters from batches 22 through 38 for samples collected January 1, 2017 through December 31,
2017. Analysis of these samples was performed April 11, 2017 through June 13, 2018. DRI
performed analyses of both anions (chloride [CI"], nitrate [NO3"], and sulfate [SO42"]) and cations
(sodium [Na+], ammonium [NH4+], and potassium[K+]) on these nylon filter samples using three
DIONEXICS-5000+ Systems (Chow and Watson, 2017) and reported the results of those
analyses to UC Davis. Chloride was reported to AQS beginning with data from samples
collected during February 2017; reasoning for not delivering chloride prior to February 2017 is
discussed in the UC Davis Chemical Speciation Network 2016 Annual Report
(https://aqrc.ucdavis.edu/csn-documentation).
4.1.1 Summary of QC Checks and Statistics
Samples received at the DRI Ion Analysis Laboratory were logged in following the chain-of-
custody procedure specified in DRI CSN SOP #2-117. Samples were analyzed using DIONEX
ICS-5000+ or ICS-6000 Systems following DRI CSN SOP #2-228 for anions and DRI CSN SOP
#2-229 for cations. QC measures for the DRI ion analysis are summarized in Table 4.1-1. The
table indicates the frequency and standards required for the specified checks, along with the
acceptance criteria and corrective actions.
During daily startup, an eight-point calibration is performed over the range from 0.02 to 3 |ag/m L
(e.g., 0.02, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0, and 3.0 |ig/mL) before analysis starts. Then two
deionized-distilled water (DDW) samples and a method blank are analyzed, followed by two
types of QC control standards: (1) 1-2.5 |ig/mL QC standards diluted from NIST certified
Dionex standard solutions; and (2) DRI-made control standards (e.g., 1.00 |ig/mL CI", 1.00
|ig/mL NO3", 1.00 |ig/mL SO42" for anions and 0.39 |ag/mL NH4+ and 1.03 |ag/mL Na+ for
cations). During routine analysis, after every 10 samples, one duplicate, one DDW, and a
selected QC check standard (same as calibration solution concentrations; diluted from certified
Environmental Research Associates (ERA) stock solutions) at various concentrations (0.1-3
|ig/mL) are analyzed.
Page|24
-------
Table 4.1-1: QC measures for ion (anion and cation) analysis by ion chromatography.
Requirement
Frequency
Calibration Standard
Acceptance Criteria
Corrective Action
Multipoint
Calibration
Daily or every batch of-100,
whichever comes first
NIST certified ERA
± 10% of certified
value
Identify and correct problem before
analyzing samples; and recalibrate
Minimum
Detection Limit
(MDL)a
Initially, then annually or after
major instrument maintenance
Nylon filter lab blanks
(7 or more)
Within ± 10% of
previous instrument
limit
Troubleshoot instrument and check
filter lots
DDW
Four initially to establish
background, followed by one
every 10 samples
DDW with resistance >
18 MQ
Within 3 standard
deviations of MDLsa
Verify instrument response to DDW
without extraction
Method blank b
One for every 40 samples
DDW with resistance >
18 MQ
Within 3 standard
deviations of MDLsa
Check instrument response for DDW
with extraction
QC Control
Standards
Daily or every run
DRI-made or Dionex
NIST-certified multi-
component standard
solution
± 10% of listed value
Rerun the QC standard and reanalyze
samples between this standard and
previous QC standard
QC Check
Standards
Every 10 samples
NIST-certified multi-
component standard
solution from ERA
± 10% of listed value
Reanalyze samples between this
standard and previous check standard
Duplicates0
10% of samples
N/A
± 10% when value >
lOx MDLa
Reanalysis of duplicate sample
Laboratory
Validation
Every sample
N/A
See note d
Reanalysis of problem sample or
flagging per SOP
a MDL indicated here is an internal laboratory QA indicator, distinct from the MDL reported to AQS.
b 15 mL DDW solution that follows the same extraction procedure as the sample extraction.
0 Duplicate indicates analysis results obtained from two different aliquots of the same filter sample extract analyzed
on the same instrument.
d Per Section 5.1 in Dili SOP #2-228 and DRI SOP #2-229. Non-quantitative criteria such as baseline position and
noise, identification of peaks, shape of peak and integration with respect to baseline.
4.1.2 Summary of QC Results
Table 4.1-1 outlines corrective actions for failed QC checks. For failed method blanks,
instrument malfunction was ruled out first. Next, the blank was reanalyzed to rule out
contamination during the extraction process and within the IC system. For the cases of failed
method blanks in Table 4.1-2, reanalysis of the blanks resulted in concentrations below QC
threshold and sample data were not affected. When the Dionex and DRI-made QC control
standards (Tables 4.1-3 and 4.1-4) that were run after multipoint calibration and before sample
analysis failed to pass the acceptance criteria, the multipoint calibration, the QC control standard,
and any samples that were analyzed were rerun to ensure that the QC standards passed
acceptance criteria. Failed ERA QC check standards that were analyzed every 10th sample (Table
4.1-5) resulted in reanalysis of all samples between the failed standard and the nearest previous
passing QC standard. Reported sample data all passed acceptance criteria for the QC check
standards. Duplicate analyses (Table 4.1-6) that exceeded acceptance criteria were reanalyzed
and compared to the original analysis. If the second duplicate met acceptable tolerance, the first
duplicate data point was considered spurious and was replaced. If the second duplicate analysis
Page|25
-------
did not meet tolerance standards, all ten samples in the set were reanalyzed. Sample data are not
affected by reanalyzing duplicates.
4.1.2.1 Method Blanks
Table 4.1-2 lists the number of method blanks analyzed during the report period and their
concentration statistics. Both median and average concentrations are near or below the MDLs
(MDL indicated here is an internal laboratory QA indicator, distinct from the MDL reported to
AQS).
Table 4.1-2: Method blank counts and concentrations for all reported ions.
Ions
CI
NOs
so;
Na
NH4
k:
Count
386
386
386
386
386
386
Median (jig/mL)
0.000
0.000
0.000
0.000
0.000
0.000
Average (jig/mL)
0.001
0.003
0.001
0.001
0.000
0.000
St. Dev. (jig/mL)
0.003
0.006
0.004
0.002
0.001
0.001
Min (jig/mL)
0.000
0.000
0.000
0.000
0.000
0.000
Max (jig/mL)
0.036
0.039
0.034
0.032
0.005
0.015
# Exceed 3xMDLa
0
10
1
0
1
0
a MDL indicated here is an internal laboratory QA indicator, distinct from the MDL reported to AQS.
4.1.2.2 QC Control Standards
Table 4.1-3 and Table 4.1-4 list the analysis statistics for Dionex and DRI-made ion QC control
standards, respectively. The control charts of these analyses are shown in Figure 4.1-la and
Figure 4.1-lb. The average difference between the measured and nominal concentrations are
within the ±10% limit (Table 4.1-1), although a few individual checks failed the 10% acceptance
criteria (see Figure 4.1-lb). Corrective actions for failed analyses are shown in Table 4.1-1.
Table 4.1-5 summarizes analysis statistics for the ERA QC check standards at different
concentration levels. Some individual standards failed QC criteria, but were reanalyzed
following the procedure outlined in Table 4.1-1. All reported CSN sample ion concentrations
passed the QC control and check standard verification.
Page|26
-------
Table 4.1-3: Statistics for Dionex ion QC control standards.
Ions
Nominal
(jig/mL)
Count
Median
(jig/mL)
Average
(jig/mL)
Min
(jig/mL)
Max
(jig/mL)
# Fail
Ave %
Recovery
% St. Dev.
CI
1.000
130
1.017
1.011
0.929
1.075
0
101.1%
3.2%
NOs
1.000
130
0.922
0.932
0.900
1.033
0
93.2%
3.2%
S(V
1.000
130
1.067
1.064
0.984
1.099
0
106.4%
2.2%
Na+
1.000
129
0.997
1.002
0.937
1.096
0
100.2%
2.9%
NH4+
1.250
129
1.289
1.292
1.222
1.369
0
103.3%
2.9%
K
2.500
129
2.606
2.609
2.384
2.749
0
104.38%
5.6%
Table 4.1-4: Statistics for DRI-inade ion QC control standards.
Ions
Nominal
(jig/mL)
Count
Median
(jig/mL)
Average
(jig/mL)
Min
(jig/mL)
Max
(jig/mL)
# Fail
Ave %
Recovery
% St. Dev.
CI
1.000
130
0.986
0.992
0.897
1.078
4
99.2%
4.3%
NOs
1.000
130
0.921
0.932
0.860
1.099
6
93.2%
4.0%
S(V
1.000
130
0.998
1.010
0.904
1.137
3
101.0%
6.0%
Na+
1.030
133
1.030
1.029
0.930
1.190
1
99.9%
4.2%
NH4+
0.390
133
0.331
0.390
0.331
0.444
10
99.9%
2.4%
K
0.000
133
0.001
0.001
0.000
0.011
0
NA
0.2%
aNA=Not applicable
Page|27
-------
Table 4.1-5: Statistics for ERA QC control standards.
Ion
Nominal
(fig/mL)
Count
Median
(jig/mL)
Average
(jig/mL)
Min
(fig/mL)
Max
(fig/mL)
Ave%
Recovery
% St. Dev.
CI
0.2
64
0.187
0.189
0.179
0.213
94.5%
0.9%
0.5
387
0.480
0.482
0.415
0.541
96.3%
2.0%
1
375
0.979
0.984
0.844
1.097
98.4%
3.9%
2
339
2.015
2.028
1.669
2.327
101.4%
7.6%
3
294
3.072
3.096
2.922
3.483
103.2%
10.0%
NOs
0.2
64
0.187
0.189
0.177
0.208
94.7%
0.8%
0.5
387
0.472
0.477
0.414
0.546
95.4%
2.1%
1
375
0.962
0.971
0.864
1.155
97.1%
4.0%
2
339
2.001
2.020
1.623
2.359
101.0%
7.9%
3
294
3.080
3.098
2.889
3.495
103.3%
10.4%
S(V
0.2
64
0.193
0.195
0.180
0.220
97.3%
1.1%
0.5
387
0.484
0.485
0.425
0.550
96.9%
2.2%
1
375
0.975
0.977
0.879
1.158
97.7%
3.7%
2
339
2.003
2.016
1.651
2.356
100.8%
7.5%
3
294
3.052
3.075
2.771
3.471
102.5%
11.0%
Na+
0.2
62
0.187
0.188
0.170
0.201
94.0%
0.7%
0.5
401
0.503
0.499
0.426
0.562
99.9%
2.1%
1
373
1.004
1.001
0.907
1.082
100.1%
2.9%
2
330
1.982
1.987
1.715
2.161
99.3%
5.9%
3
312
3.017
3.017
2.727
3.358
100.6%
8.4%
NH4+
0.2
62
0.195
0.193
0.169
0.218
96.7%
0.9%
0.5
401
0.472
0.476
0.407
0.580
95.1%
1.9%
1
373
0.983
0.982
0.877
1.124
98.2%
3.2%
2
330
2.001
2.002
1.780
3.297
100.1%
4.7%
3
312
3.023
3.024
2.738
3.297
100.8%
7.8%
K
0.2
62
0.193
0.192
0.162
0.211
95.9%
1.1%
0.5
401
0.500
0.497
0.420
0.576
99.4%
3.1%
1
373
0.962
0.976
0.842
1.163
97.6%
5.2%
2
330
2.185
2.533
1.737
3.550
126.7%
53.6%
3
312
3.058
3.076
2.770
3.550
102.5%
11.3%
Page|28
-------
Figure 4.1-la: Control charts for Dionex ion QC control standards. The limits are ±10% of the nominal
concentrations (red dashed lines).
CI- (Nominal 1.000 jjg/mL)
N03" (Nominal 1.000 jjg/mL)
1.15
_l
£
1.10
O)
3.
1.05
c
0
1.00
5
c
0.95
a)
0
c
0
0.90
O
c
0.85
0
0.80
. •
•a a #
1.15
_i
£
1.10
a)
ZL
1.05
c.
0
1.00
TO
C
0.95
a)
u
c
0.90
0
0
c
0.85
0
0.80
•••
Analysis Date
Analysis Date
1.15
1.10
1.05
1.00
0.95
0.90
0.85
0.80
S042- (Nominal 1.000 Hg/mL)
•_ • • if
Na+ (Nominal 1.000 jjg/mL)
j 1.10
E
g 1.05
.1 1.00
5
¦£ 0.95
J
Analysis Date
Analysis Date
1.50
1.45
1.40
1.35
1.30
1.25
1.20
1.15
1.10
1.05
1.00
NH4+ (Nominal 1.25 Hg/ml_)
K+ (Nominal 2.5 jjg/mL)
V*^ • v?1*
3.0
2.9
2.8
2.7
2.6
2.5
2.4
2.3
2.2
2.1
2.0
Analysis Date
Analysis Date
Page|29
-------
Figure 4.1-lb: Statistics for DRI-made ion QC control standards. The limits are ±10% of the nominal
concentrations (red dashed lines), except for K+ which is 3 xMDLa (red dashed lines).
1.15
1.10
1.05
1.00
0.95
0.90
0.85
0.80
CI- (Nominal 1.000 |jg/mL)
•
H*tV
• . * •*
* ' % ^
• • * +
'i? .• •
•
1.15
1.10
1.05
1.00
0.95
0.90
0.85
0.80
N03- (Nominal 1.000 |jg/mL)
• •• *•
1.20
1.15
_i
£
1.10
O)
3.
C
1.05
o
TO
1.00
C
0.95
Q)
U
C
o
0.90
o
c
0.85
o
0.80
Analysis Date
S042- (Nominal 1.000 ng/mL)
.*»
V fVv'
. \ I'll ~
1.25
J
1.20
?-
U)
1.1b
=L
C
1.10
o
1 Oh
c
1.00
a)
0.95
o
O
0.90
c
O
0.85
0.80
Analysis Date
Na+ (Nominal 1.030 (jg/mL)
- t*l
I •
J '
E 0.45
O
O
Analysis Date
NH4+ (Nominal 0.390 jjg/mL)
-
•1
0.020
0.015
zr
£
0.010
ai
ZL
c
0.005
O
CO
0.000
c
a>
o
-0.005
c
o
O
-0.010
c
O
-0.015
-0.020
Analysis Date
K+ (Nominal 0.000 ng/mL)
Si[\ ,*»
Analysis Date
Analysis Date
1MDL indicated here is an internal laboratory QA indicator, distinct from the MDL reported to AQS.
Page|30
-------
4.1.2.3
Duplicate Analyses
Table 4.1-6 gives the criteria and summary statistics for duplicate analysis results. Duplicate
analysis results are obtained from two different aliquots of the same filter sample extract run on
the same instrument. The criteria used for each ion were that 1) if the average concentration was
less than 10 times the lower quantifiable limit (LQL), the absolute value of the average
difference should be less than ten times the LQL, and 2) if the average concentration was greater
than or equal to ten times the LQL, then the relative percent difference (RPD) should be less than
10%. LQLs are given in Tables 4.1-7a and b. The LQLs are used as internal QA indicators,
distinct from the MDLs reported to AQS.
Table 4.1-6: Ion duplicate analysis criteria and statistics.
Range
Criteria
Statistic
Na
nh4
k:
CI
NO3
SO4'
Units
Count
1346
770
1409
1426
907
600
No. Fail
0
13
0
0
0
0
%Fail
0
1.7
0
0
0
0
%
Ion< 10 x
10 x
RPDa
Mean
1.0%
1.8%
2.3%
0.6%
0.7%
0.9%
RPD
LQL
<10%
St. Dev.
1.3%
2.1%
2.5%
0.8%
0.8%
1.0%
RPD
Max
5.9%
14.3%
9.7%
3.7%
6.8%
7.9%
RPD
Min
0.0%
0.0%
0.1%
0.0%
0.0%
0.0%
RPD
Median
0.5%
1.0%
1.3%
0.4%
0.4%
0.5%
RPD
aRPD= 100 x absolute value [original sample - duplicate sample] / [(original sample + duplicate sample) / 2]
4.1.3 Determination of Uncertainties and Method Detection Limits
For discussion of Method Detection Limits (MDLs) see Section 3.1.3.2.
For discussion of analytical uncertainty and total uncertainty see Section 3.1.2 and Section 6.5,
respectively.
4.1.4 Audits, Performance Evaluations, Training, and Accreditations
4.1.4.1 System Audits
The prime contractor (UC Davis) did not conduct any audits of the DRI Ion Analysis laboratory
during 2017.
4.1.4.2 Performance Evaluations
The EPA provided five nylon samples for anion and cation analysis as part of the EPA Office of
Air Quality Planning and Standards (OAQPS) interlaboratory performance evaluation. This
Page|31
-------
evaluation was carried out during the timeframe when DRI also analyzed CSN samples collected
during 2017. Reported z-scores were all below 2, indicating satisfactory interlaboratory
comparison results.
4.1.4.3 Training
All new laboratory staff receive training in performing tasks in the SOPs for their assigned work.
4.1.4.4 A ccreditations
There are no accreditations for analysis of ions on aerosol filters by Ion Chromatography.
4.1.5 Summary of Filter Field Blanks
Over the sampling period (January 1, 2017 through December 31, 2017) there were 1,368 valid
nylon filter field blanks. Table 4.1-7a and Table 4.1-7b summarize the field blank statistics. The
lower quantifiable limits (LQLs) are defined as three times the standard deviation of field blanks
and are used an internal laboratory QA indicators, distinct from the MDLs reported to AQS.
Table 4.1-7a: Nylon filter field blank statistics in |ig/mL.
Ions
Count
Median
(jig/mL)
Average
(jig/mL)
Min
(jig/mL)
Max
(jig/mL)
St. Dev.
(jig/mL)
LQL
(jig/mL)
CI
1349
0.011
0.020
0.000
1.625
0.061
0.184
no3
1349
0.012
0.020
0.000
2.975
0.090
0.269
SC>42-
1349
0.005
0.018
0.000
1.119
0.081
0.243
Na+
1349
0.003
0.007
0.000
1.331
0.049
0.146
NH4
1349
0.000
0.001
0.000
0.683
0.019
0.058
K+
1349
0.000
0.002
0.000
0.363
0.012
0.036
Table 4.1-7b: Nylon filter field blank statistics in |ig/filtcr (extraction volume 15 mL).
Ions
Count
Median
(jig/filter)
Average
(jug/filter)
Min
(jug/t'ilter)
Max
(jug/t'ilter)
St. Dev.
(jug/t'ilter)
LQL
(jug/t'ilter)
CI
1349
0.169
0.305
0.000
24.378
0.921
2.764
no3
1349
0.181
0.293
0.000
44.625
1.347
4.041
SC>42-
1349
0.070
0.275
0.000
16.779
1.215
3.646
Na+
1349
0.042
0.109
0.007
19.969
0.730
2.189
NH4
1349
0.000
0.022
0.000
10.243
0.288
0.863
K+
1349
0.000
0.024
0.000
5.450
0.182
0.545
4.2 UC Davis X-Ray Fluorescence (XRF) Laboratory
The UC Davis XRF Laboratory received and analyzed PTFE filters from batches 1 through 38,
which includes samples collected January through December 2017. UC Davis performed
analysis for 33 elements using energy dispersive X-ray fluorescence (EDXRF) instruments.
These analyses were performed during an analysis period from April 6, 2017 to September 6,
2018. Two EDXRF instruments, XRF-1 and XRF-4, performed all analyses during this period;
see Table 4.2-1 for details.
Page|32
-------
Table 4.2-1: Sampling dates and corresponding EDXRF analysis dates covered in this report. Analysis dates include
reanalysis - as requested during QA Level 1 validation - of any samples within the sampling year and month.
Sampling Year
Sampling Month
XRF-1 Analysis Dates
XRF-4 Analysis Dates
2017
January
04/06/2017-05/01/2017
04/06/2017-05/01/2017
2017
February
05/01/2017-05/21/2017
05/01/2017-09/22/2017
2017
March
05/21/2017-06/18/2017
05/21/2017-06/17/2017
2017
April
06/23/2017-07/24/2017
06/10/2017- 10/17/2017
2017
May
07/24/2017-08/21/2017
07/23/2017-08/25/2017
2017
June
08/21/2017-09/21/2017
08/22/2017-09/19/2017
2017
July
09/21/2017-01/03/2018
09/19/2017- 10/20/2017
2017
August
10/19/2017-02/08/2018
10/19/2017-02/16/2018
2017
September
11/15/2017-02/08/2018
11/20/2017-04/02/2018
2017
October
12/12/2017-04/27/2018
12/16/2017-01/24/2018
2017
November
01/27/2018-02/27/2018
01/26/2018-02/25/2018
2017
December
02/24/2018-09/06/2018
02/26/2018-03/25/2018
2017
All Months
04/06/2017-09/06/2018
04/06/2017-04/02/2018
4.2.1 Summary of QC Checks and Statistics
Samples are received by the UC Davis XRF Laboratory following the chain-of-custody
procedures detailed in the UCD CSN TI302B. Samples are analyzed using Malvern-Panalytical
Epsilon 5 EDXRF instruments following UCD CSN SOP 302. Calibration of the EDXRF
instruments is performed annually and as needed to address maintenance or performance issues
(e.g. an X-ray tube or detector is replaced). Quality control procedures are described in UCD
CSN TI 302D and are summarized in Table 4.2-2.
Table 4.2-2: Frequency and types of checks performed and associated criteria and corrective actions for analysis by
EDXRF.
Analysis
Frequency
Criterion
Corrective Action
Detector
Calibration
Weekly
None (An automated process done
by XRF software)
• XRF software automatically adjusts
the energy channels
PTFE Blank
Daily
< acceptance limits with
exceedance of a single element
allowed for a maximum of two
consecutive days
• Change/clean blank if
contaminated/damaged
• Clean the diaphragm, if necessary
• Further cross-instrumental testing
UC Davis Multi-
element sample
Daily
±10% of reference mass loadings
for Al, Si, S, K, Ca, Ti, Cr, Mn,
Fe, Ni, Cu, Zn and Pb
• Check sample for
damage/contamination
• Further cross-instrumental testing
• Replace sample if necessary
Micromatter
Al&Si sample
Weekly
±10% of reference mass loadings
UC Davis Multi-
element sample
Weekly
±10% of reference mass loadings
for Al, Si, S, K, Ca, Ti, Cr, Mn,
Fe, Ni, Cu, Zn and Pb
Reanalysis
samples
Monthly
z-score between ±1 for Al, Si, S,
K, Ca, Ti, Mn, Fe, Zn, Se and Sr
SRM 2783
Monthly
Bias between ±1 for Al, Si, S, K,
Ca, Ti, Cr, Mn Fe, Ni, Cu, Zn and
Pb
Daily QC checks include a laboratory blank (PTFE blank) and a multi-elemental reference
material (ME-RM) to monitor contamination and stability/performance of the instruments. A
Micromatter Al&Si ME-RM and a UC Davis-made ME-RM are also analyzed weekly to check
Page|33
-------
instalment performance. Inter-instrumental comparability is monitored by analyzing the bias and
precision between instruments of the weekly UC Davis ME-RM. Long-term inter-instrumental
comparability is monitored using a set of reanalysis filters which are reanalyzed monthly on each
instrument. Long-term reproducibility is monitored using the reanalysis filters and by analyzing
a NIST SRM 2783 standard monthly and comparing the EDXRF error from the
certified/reference mass loadings to acceptance limits.
4.2.2 Summary of QC Results
QC tests conducted over the course of the analysis period showed good overall control of the
instruments and process. There were sporadic failures of the QC criteria, which were
investigated promptly and corrected with minimal impact on sample analysis. The following
summarizes the QC issues which occurred during the analysis period reported here.
Random occasional zinc contamination was observed on QC filters for both XRF-1 and XRF-4.
This sporadic zinc contamination appears to be related to the design of the instrument and is
unavoidable. Samples analyzed during this period were monitored closely for any contamination
and were reanalyzed if there was any question of contamination. The reported data are not
impacted. See Section 2.2.2, Section 3.2.1.3, and Section 4.2.2.1 for further detail.
Both XRF-1 and XRF-4 also exhibited some failures of the acceptance criteria for all QC checks
of Ca. Investigation is ongoing, with initial findings suggesting gradual increase in Ca
concentrations on QC filters may be caused by environmental deposition during extended
residence in the instruments. Samples are only exposed to the environment for a day or two
during routine analysis, thus are not susceptible to gradual Ca contamination. However, samples
are carefully monitored for atypical and abrupt calcium contamination events and reanalyzed as
necessary. The reported data are not impacted. See Section 2.2.3, Section 3.2.1.2, and Section
4.2.2.1 for further detail.
4.2.2.1 Results of Daily QC Checks
Possible contamination and instability issues are monitored by analyzing a PTFE blank daily.
The EDXRF results are compared to acceptance limits, which are calculated as three times the
standard deviation plus the mean of a set of laboratory PTFE blanks. Figures 4.2-la and 4.2-lb
show the results of daily analyses of laboratory blanks on both instruments. If the mass loading
exceeds the limit for more than two consecutive days, the blank is replaced to distinguish
between blank contamination and instrument contamination. Some occasional exceedance of the
acceptance limits is expected but not continuous or repeated exceedances. In all cases of
exceedance, the other QC filters are checked to determine if the problem is instrumental or
strictly contamination of a blank. Analysis results are reviewed during QA Level 1 validation
(UCD CSN TI801C), and elements associated with occasional contamination (Zn and Ca; see
Section 4.2.2) are monitored closely. When contamination is suspected, filters are reanalyzed and
the reanalysis result is reported if contamination was present in the original analysis. A total of
10 samples from 2017 were reanalyzed for suspected Zn contamination (six from XRF-1, four
from XRF-4). Of those, three were found to have Zn contamination and the reanalysis result was
reported (two from XRF-1 and one from XRF-4). Six samples were reanalyzed for suspected Ca
contamination (two from XRF-1 and four from XRF-4). Only one sample was found to have Ca
contamination and the reanalysis result was reported.
Page|34
-------
Both XRF-1 and XRF-4 had sporadic elevated measurements of Zn on laboratory blanks
throughout the analysis period (as discussed in Section 2.2.2 and Section 3.2.1.3). These elevated
levels were not measured over consecutive days thus did not fail the acceptance criteria;
however, these occurrences are monitored closely. Zn contamination likely comes from wear on
the sample changer; Zn is a common contaminant in elemental analysis systems.
Both XRF-1 and XRF-4 show gradual increases in Ca (as discussed in Section 2.2.3 and Section
3.2.1.2), which is reduced immediately after the blank filter is changed. This indicates
contamination of the blank filter likely from atmospheric deposition and/or instrument wear.
This situation worsened on XRF-4 for analyses performed during 2018. The cause of Ca increase
on QC filters with long, multi-day, residences in the instrument is being investigated.
In June and November 2017 for XRF-1 and XRF-4, respectively, there were slightly elevated
signals for Fe, Ni, and Cr for laboratory blanks. These were isolated events due to small stainless
steel contamination on the blank filters from instrument wear. The exceedances did not occur on
consecutive days thus did not fail the acceptance criteria. Replacing the contaminated blank
filters resolved the issue and no samples were affected. The reported data are not impacted.
Lastly, CI had a few exceedances on both XRF-1 and XRF-4 during the analysis period. For the
larger exceedances laboratory blanks were replaced which corrected the exceedance; for others
the signal decreased without correction. The cause of the CI exceedances is unknown; as a
volatile element it has a highly variable signal from QC filters.
Page|35
-------
Figure 4.2-la: Results of daily analyzed PTFE laboratory blanks during the analysis period for samples collected
January through December 2017. Elements Na through Zn shown.
Accpeptance Limit XRF-1 — XRF-4
Na
Si
CI
Ti
Mn
Ni
0.080-
0.060
0.040
0.020
o.ooo-
0.004-
0.003
0.002
0.001
o.ooo-
0.020-
0.015-
0.010-
0.005
o.ooo-
0.003
0.002
0.001-
0.000
0.040
0.030-
0.020-
0.010
Mg
Fe
0.100
0.080 H
0.060
Al
0.005-
0.004
0.003 i
0.002
0.001 ¦
o.ooo-
0.060
0.040-
0.020
o.ooo-
0.010
0.008 i
0.005
0.002
0.004
0.002 H
0.000
Ca
Cr
Co
Zn
-i 1 1 1 1 1 1 1 r
1^
r-
00
00
CO
00
00
h-
h-
h-
h-
00
CO
CO
00
00
h-
h-
00
00
00
00
CO
T—
r—
r—
t—
T—
r—
T—
T—
o
o
o
o
O
O
o
o
o
o
O
o
O
o
o
O
O
o
O
o
O
o
O
O
O
O
o
CM
CM
CM
CM
CM
—
Q_
>
£Z
i—
CO
>¦.
—
Q.
>
—
Q.
>
C
CD
—
Q-
>%
—
Q_
>
a
CD
>
—
CL
CD
—j
:>
CO
CO
2
D
CO
CO
2
—3
1
CO
XRF Analysis Date
Page|36
-------
Figure 4.2-lb: Results of daily analyzed PTFE laboratory blanks during the analysis period for samples collected
January through December 2017. Elements As through Pb shown.
Accpeptance Limit
XRF-1
XRF-4
r--
h-
00
00
00
00
CO
Is-
h-
1^
N-
00
CO
CO
00
CO
h-
h-
r-
CO
CO
00
00
CO
v—
T—
r—
T—
x—
T—
T—
o
o
O
o
o
o
O
O
O
O
o
O
o
O
O
O
o
O
O
O
o
o
o
o
o
O
o
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
CM
—
Q_
>
c
CO
>%
—
Cl
>%
—
Q.
>
c
CO
>>
—
CL
>
—
Q_
>
c
CO
>%
—
Q-
CO
—j
0)
o
co
CO
—j
CO
CO
2
CO
XRF Analysis Date
Page|37
-------
Daily operational performance of the instruments is monitored by a multi-element reference
material (ME-RM). Each instrument had its own daily ME-RM produced by UC Davis. The
acceptance limits are set to +/- 10% RSD of the reference values for the relevant elements, as
listed in Table 4.2-2. When more than two consecutive measurements exceed these limits the
results are marked unacceptable. Corrective actions for unacceptable QC results include
checking the sample for damage or contamination, checking the results for the affected element
on other QC samples, cross-instrumental testing if necessary to determine if the unacceptable
result is due to the instrument or the QC sample, and further investigations as necessary. Sample
analysis is halted or samples analyzed after the unacceptable QC result are noted for possible
reanalysis depending on the outcome of the investigation. When a problem with the instrument is
found the affected samples are reanalyzed on a different instrument or the same instrument after
the issue is corrected and once it has been demonstrated to be within control again. QC samples
which have been found to be damaged or contaminated are replaced (UCD CSN TI302D).
Tables 4.2-3 and 4.2-4 show the results of the UC Davis ME-RMs. A small number of criteria
exceedances are expected statistically, but this should be no more than 3% of the total number of
measurements. Investigations of other QC filters and laboratory blanks, following these
exceedances, did not show any contamination or instrumental issues, so no corrective actions
were taken. Unacceptable QC results for Ca are expected to be from the same source as
discussed for laboratory blank contamination (see Section 2.2.3, Section 3.2.1.3, and Section
4.2.2). The laboratory blanks were replaced when contamination occurred; however, the ME-RM
samples were not replaced in response to contamination.
Table 4.2-3: Descriptive statistics of XRF-1 results (pg/cnr) of the daily UC Davis ME-RM from 04/06/2017 to
09/06/2018, N = 537 (see Table 4.2-1 for corresponding sampling dates).
Element
Average
Lower Limit
Upper Limit
% Exceedance
% Unacceptable
RSD %
A1
1.646
1.463
1.789
0.0
0
1.9
Si
2.799
2.504
3.061
0.0
0
1.0
S
10.959
9.826
12.010
0.0
0
0.9
K
1.669
1.497
1.829
0.0
0
0.7
Ca
1.941
1.688
2.063
7.1
4.3
2.1
Ti
0.129
0.114
0.139
2.0
0
3.5
Cr
0.701
0.629
0.769
0.0
0
0.9
Mn
0.337
0.304
0.372
0.0
0
2.1
Fe
1.992
1.777
2.171
0.0
0
1.4
Ni
0.117
0.105
0.129
0.0
0
2.3
Cu
0.546
0.488
0.596
0.0
0
1.4
Zn
0.406
0.357
0.436
0.2
0
1.6
Pb
0.644
0.578
0.707
0.0
0
2.4
Limits are +/-10% of the reference loading (TI 302D).
Page|38
-------
Table 4.2-4: Descriptive statistics of XRF-4 results (pg/cnr) of the daily UC Davis ME-RM from 04/06/2017 to
04/02/2018, N = 350 (see Table 4.2-1 for corresponding sampling dates).
Element
Average
Lower Limit
Upper Limit
% Exceedance
% Unacceptable
RSD %
A1
1.469
1.285
1.571
0.6
0
2.5
Si
2.921
2.597
3.174
0
0
2.0
S
10.673
9.588
11.718
0
0
1.0
K
1.662
1.493
1.825
0
0
1.1
Ca
1.989
1.701
2.079
0.9
0
1.8
Ti
0.133
0.118
0.144
0.6
0
3.2
Cr
0.706
0.633
0.774
0
0
1.2
Mn
0.335
0.300
0.367
0
0
2.3
Fe
1.997
1.784
2.181
0
0
2.6
Ni
0.116
0.103
0.126
0
0
2.6
Cu
0.544
0.488
0.597
0
0
1.5
Zn
0.384
0.338
0.413
0
0
2.5
Pb
0.635
0.573
0.700
0
0
2.5
Limits are +/-10% of the reference loading (TI 302D).
4.2.2.2 Results of Weekly QC Checks
Weekly QC checks include analysis of a UC Davis produced ME-RM (different than the daily
ME-RM) and a ME-RM purchased from Micromatter containing only A1 and Si. The UC Davis
weekly ME-RM was replaced in September 2017. Weekly results are compared to acceptance
limits of +/- 10% of the reference values for the relevant elements, as listed in Table 4.2-2. When
more than two consecutive measurements exceed these limits the results are marked
unacceptable. Corrective actions for unacceptable results are described in Section 4.2.2.1 of this
report and can be found in UCD CSNSOP #302 and UCD CSN TI 302D. A weekly QC report is
generated internally, which includes checks of the laboratory blanks and the daily and weekly
ME-RMs.
Table 4.2-5 and Table 4.2-6 show the EDXRF statistics of the weekly UC Davis ME-RM run
until September 2017.
Page|39
-------
Table 4.2-5: Descriptive statistics of XRF-1 results (|ig/cnr) of the weekly UC Davis ME-RM from 04/10/2017 to
09/27/2017, N = 20 (see Table 4.2-1 for corresponding sampling dates).
Element
Average
Lower Limit
Upper Limit
% Exceedance
% Unacceptable
RSD %
A1
0.898
0.803
0.981
0
0
2.6
Si
1.441
1.288
1.574
0
0
1.2
S
5.430
4.835
5.910
0
0
0.9
K
0.790
0.711
0.869
0
0
0.4
Ca
0.863
0.773
0.944
0
0
1.2
Ti
0.059
0.054
0.065
0
0
4.4
Cr
0.318
0.287
0.351
0
0
1.1
Mn
0.154
0.139
0.169
0
0
2.4
Fe
0.895
0.806
0.985
0
0
0.8
Ni
0.052
0.049
0.060
5.0
0
4.2
Cu
0.115
0.105
0.128
0
0
2.4
Zn
0.108
0.095
0.116
0
0
2.1
Pb
0.279
0.249
0.305
0
0
3.7
Limits are +/-10% of the reference loading (TI 302D).
Table 4.2-6: Descriptive statistics of XRF-4 results (|ig/cm2) of the weekly UC Davis ME-RM from 04/10/2017 to
09/27/2017, N = 24 (see Table 4.2-1 for corresponding sampling dates).
Element
Average
Lower Limit
Upper Limit
% Exceedance
% Unacceptable
RSD %
A1
0.786
0.705
0.862
0
0
3.2
Si
1.470
1.323
1.617
0
0
2.0
S
5.307
4.730
5.781
0
0
1.3
K
0.793
0.714
0.873
0
0
0.8
Ca
0.868
0.776
0.948
0
0
1.6
Ti
0.061
0.057
0.070
0
0
4.1
Cr
0.320
0.287
0.351
0
0
1.1
Mn
0.153
0.137
0.168
0
0
3.0
Fe
0.901
0.808
0.988
0
0
0.8
Ni
0.052
0.046
0.056
0
0
3.3
Cu
0.117
0.105
0.128
0
0
2.0
Zn
0.108
0.094
0.115
0
0
2.8
Pb
0.284
0.248
0.303
4.2
0
3.9
Limits are +/-10% of the reference loading (TI 302D).
Page|40
-------
Table 4.2-7 and Table 4.2-8 show results of the new weekly UC Davis ME-RM, used beginning
September 2017.
Table 4.2-7: Descriptive statistics of XRF-1 results (|ig/cnr) of the new weekly UC Davis ME-RM from
09/29/2017 to 09/05/2018, N = 43 (see Table 4.2-1 for corresponding sampling dates).
Element
Average
Lower Limit
Upper Limit
% Exceedance
% Unacceptable
RSD %
A1
1.573
1.400
1.711
0
0
2.0
Si
2.572
2.311
2.825
0
0
1.1
S
9.680
8.686
10.616
0
0
0.9
K
1.470
1.314
1.606
0
0
1.1
Ca
1.594
1.435
1.754
0
0
1.6
Ti
0.120
0.108
0.132
0
0
2.6
Cr
0.603
0.542
0.662
0
0
0.9
Mn
0.290
0.258
0.316
0
0
2.7
Fe
1.675
1.501
1.834
0
0
1.6
Ni
0.100
0.090
0.110
0
0
2.4
Cu
0.285
0.255
0.312
0
0
1.5
Zn
0.226
0.200
0.244
0
0
2.3
Pb
0.539
0.480
0.586
0
0
2.4
Limits are +/-10% of the reference loading (TI 302D).
Table 4.2-8: Descriptive statistics of XRF-4 results (|ig/cnr) of the new weekly UC Davis ME-RM from
09/29/2017 to 04/02/2018, N = 51 (see Table 4.2-1 for corresponding sampling dates).
Element
Average
Lower Limit
Upper Limit
%Exceedance
% Unacceptable
RSD %
A1
1.405
1.251
1.529
0
0
2.6
Si
2.659
2.386
2.917
0
0
2.6
S
9.663
8.677
10.606
0
0
0.9
K
1.474
1.325
1.619
0
0
0.8
Ca
1.593
1.424
1.740
0
0
0.9
Ti
0.123
0.111
0.135
0
0
3.2
Cr
0.613
0.552
0.675
0
0
1.4
Mn
0.293
0.263
0.322
0
0
2.0
Fe
1.695
1.525
1.864
0
0
3.1
Ni
0.101
0.090
0.110
0
0
2.6
Cu
0.290
0.261
0.319
0
0
1.3
Zn
0.228
0.205
0.250
0
0
1.9
Pb
0.549
0.502
0.614
0
0
2.9
Limits are +/-10% of the reference loading (TI 302D).
Page | 41
-------
A Micromatter ME-RM containing A1 and Si is also run weekly. The results from this sample are
plotted in Figure 4.2-2. The acceptance limits are set as +/- 10% of the average of the first five
measurement results from each XRF. No issues were observed.
Figure 4.2-2: EDXRF results of the weekly Micromatter ME-RM containing A1 and Si. Limits are +/-10% of the
reference loading.
E
o
CD
O)
C
'~o
03
O
_l
03
C/5
CC
10.50
10.00
9.50
9.00
10.00
9.50
9.00
8.50
Al
XRF-1
Si
XRF-1
1 1 1 i 1 1 1 r-
—i 1 i—r-
o >> <= '
9- 0 C= JZJ )= >C5 CD Q.
=3(D>sO(1)C0(D^9-CDZJ-^3(D
9-=3 -
<5-1
j D> Q-t5 > O CJ3 I; f: >>C q Oia
=50>>;O,g9-ca=3-^=S!D
XRF-1
XRF-4
+/-10%
XRF Analysis Date
4.2.2.3 Reproducibility and Inter-instrument Performance Tests
The weekly ME-RM is also used as an inter-instrument comparison with the same sample
analyzed by both EDXRF instalments. Figures 4.2-3 and 4.2-4 plot the elemental concentrations
for both of the weekly UC Davis ME-RM samples used during this analysis. As mentioned in
Section 4.2.2.2, the UC Davis weekly ME-RM was replaced in September 2017. The following
approach is used to quantify the differences observed in the plots. The scaled relative difference
(SRD) between the two instalments is calculated for each element each week as:
SRDi =
(XRF 1 j -XRF4 ,) / V2
(XRF1 j + XRF4,) / 2
where XRFli and XRI'A, are the mass loadings of the ith element measured by each instrument.
For each element, ith random error (precision) of each instrument is estimated as the standard
deviation of the weekly results SRDW, w = 1, , N:
Page|42
-------
Precision= 1^^ {SRDLw - SRDt)2
Precision
Where is the mean scaled relative difference for element i over the analysis period.
The bias between instruments is the mean value of the unsealed relative differences,
1 V / XRFlf - XRF4,. \
mSi nZjw\(X RF1, + XRF4,) / 2/w
( XRFlf - XRF4,-
iw \(XRF1,- + XRF4, ) / 2
The precision acceptance limit for the ith element is calculated from the variation in the response
of each instrument,
where o is the standard deviation in the mass loading measured by the instrument. The bias
acceptance limit is calculated as the sum of the error of both instruments to a mean reference
mass loading for the ith element of the ME-RM,
where G,re/is the reference elemental mass loading and k is a coverage factor which is set to a
value of two to account for distribution of uncertainties possible in a given measurement. The
acceptance limits are based on the mean mass loading for both instruments and provide a
historical bias from which to compare the weekly bias of each instrument.
The results from this analysis, for elements greater than ten times the detection limit, averaged
over both UC Davis ME-RM samples are presented in Table 4.2-9. The results of the inter-
instrument comparison show a larger bias for Na, Mg, and P. These elements are difficult to
quantify using the EDXRF method and some differences are expected.
1
Bias Acceptance Limits = k * —
Page|43
-------
Table 4.2-9: Precision and bias between XRF-1 and XRF-4 from the weekly UC Davis ME-RM calculated from
04/06/2017 to 09/06/2018 (see Table 4.2-1 for corresponding sampling dates). Only elements with mass loadings >
lOxMDL are reported. CI and Br are not reported because they are volatile and mass loadings degrade over time.
Element
Bias %
Bias Acceptance
Limit %
Precision %
Precision Acceptance
Limit %
Na
-5.0
±17.4
4.4
6.5
Mg
43.7
±79.3
10.7
13.3
A1
12.0
±24.6
2.9
3.9
Si
-2.0
±6.5
2.0
2.7
P
-85.5
±159.5
12.4
16.3
S
1.1
±4.5
1.0
1.5
K
-0.1
±2.7
0.9
1.1
Ca
0.3
±4.5
1.2
1.9
Ti
-2.5
±12.6
3.2
5.1
V
-1.2
±28.8
2.2
11.3
Cr
-0.5
±3.8
1.0
1.7
Mn
-0.1
±8.1
2.3
3.7
Fe
0.7
±5.7
1.4
2.3
Co
0.1
±11.5
3.3
5.2
Ni
0.2
±10.1
3.1
4.5
Cu
-0.9
±6.2
1.9
2.6
Zn
-0.3
±7.3
1.6
3.4
As
-3.3
±8.9
1.9
2.7
Se
-1.3
±7.4
2.2
3.2
Rb
-0.5
±11.1
3.2
5.2
Sr
-0.4
±11.9
3.7
5.0
Pb
-1.1
±10.8
3.2
4.6
Page|44
-------
Figure 4.2-3: Instrumental comparison using the weekly UC Davis ME-RM. XRF-1: 4/12/2017 to 9/21/2017, N
19. XRF-4: 4/10/2017 through 9/21/2017, N = 19. (See Table 4.2-1 for corresponding sampling dates.)
XRF-1 Ej3 XRF-4
Na, Bias= -8.2%
S, Bias= 2.3%
0.500
0.400
0.300
0.200
0.110
0.100
0.090
0.080
0.070
Mg, Bias= 49.3%
Al, Bias= 12
CI, Bias= 9.6%
•
•
1—1—1
i
0.900- F^l
0.850-
0.800-
0.750-
Mn, Bias
= 0.1%
1
XRF-1 XRF-4
XRF-1 XRF-4
0.300-
0.290-
0.280-
0.270-
0.260 ¦
Zr, Bias= -9.8%
•
;T
Sb, Bias=
-25.3%
•
I
V
0.020
0.010
0.075-
0.050-
0.025-
Ag, Bias= -18.4%
•
•
I
I
Cs, Bias= -33.7%
I
I
Pb, Bias= -2.2%
, J
L
[
I
XRF-1 XRF-4
XRF-1 XRF-4
XRF-1 XRF-4
Page|45
-------
Figure 4.2-4: Instrumental comparison using the new weekly UC Davis ME-RM. XRF-1: 10/2/2017 to 8/28/2018, N -
41. XRF-4: 10/05/2017 to 8/27/2018, N = 41. (See Table 4.2-1 for corresponding sampling dates.)
E^3 XRF-1 Ej3 XRF-4
P. Bias= -94.7%
E
o
D5
CT>
C
x>
03
O
_l
w
C/>
03
9.900
9.800
9.700
9.600
9.500
9.400
0.150
0.140
0.130
0.120
0.110
0.104
0.100
0.096
0.006-
0.004-
0.002-
o.ooo-
0.140-
0.130-
0.120-
0.110-
0.100-
0.400
0.360
0.320
Ti, Bias= -2.2%
V
Ni, Bias= 0.3%
i
=^= —
!
Br, Bias= 8.8%
Cd, Bias= -0.1%
*
Ba, Bias= 0.5%
4-
0.050-
0.040-
0.030-
0.020-
0.150
0.120
0.090
0.060
In, Bias= 5.0%
1
I 1
1
Sn, Bias= -16.7%
0.060-
0.050-
0.040 -
0.030 -
0.020-
0.010-1
0.050 -
0.040 -
0.030-
0.020 -
0.010-
Sb, Bias=
-6.2%
-
|
-
-
1
Ce, Bias= -1.7%
I
Pb, Bias= -0.6%
XRF-1 XRF-4
XRF-1 XRF-4
0.570 -
0.550-
0.530-
0.510-
XRF-1 XRF-4
0.100-
0.080-
0.060-
0.040-
0.020-
Cs, Bias= -6.7%
XRF-1 XRF-4
XRF-1 XRF-4
4.2.2.4
Long-term Stability, Reproducibility, and Inter-instrument Performance
A set of filters are reanalyzed monthly to monitor the long-term instrument performance; the set
was changed once during 2018. For analyses performed April 2017 through May 2018, the set
Page|46
-------
consisted of 16 CSN samples and one UC Davis produced ME-RM. The samples were on MTL
47 mm PTFE filters and covered a range of mass loadings representative of the CSN. The second
set of 16 filters, used beginning June 2018, were UC Davis ME-RMs and covered a range of
mass loadings simulating the CSN and higher for trace elements. In addition to these filters, a
NIST SRM 2783 standard was included in the new set. In order to compare multiple filters with
different mass loadings, the results of reanalysis are first converted to z-scores. For a given
month, the z-score for the ith element and jtb filter is
where x,-y is that month's result, X'Jis the reference value for element i in filter /, and U(x,,) and
U{xij )are the uncertainty of that month's result and the reference uncertainty respectively. The
instrument-specific reference values for the samples of the reanalysis set are determined as the
mean and standard deviation of five initial measurements, while the values for SRM 2783 are the
certified or reference loadings. Monthly z-scores for each element are then summarized across
the N filters in terms of
Every month, two different reference values are used to calculate z-scores: (1) one reference
value is only based on the average response from the one instrument for which the z-score is
being calculated, and (2) the other reference value is based on the average response from both
instruments. The first z-score serves as long-term reproducibility of each instrument while the
second z-score is an inter-instrumental comparison. These two z-scores are plotted and checked
to be within -1 to 1 for elements which have mass loadings well above the MDL (Al, Si, S, K,
Ca, Ti, Mn, Fe, Zn, Se, and Sr). For further detail see UCD CSN TI302D.
Figure 4.2-5 shows the mean z-score plots over the analysis period. Issues observed include
increasing mean z-scores for Ca on both instruments and low XRF-4 mean z-scores for Al. The
increasing Ca z-scores relate to the previously mentioned Ca contamination on QC filters (see
Section 2.2.3, Section 3.2.1.2, Section 4.2.2.1), and are observed for both instruments on both
sets of reanalysis filters, occasionally resulting in acceptance criteria exceedances. The XRF-4
low mean z-score for Al is from bias between the XRF-4 and XRF-1 Al values (Table 4.2-9),
which drives the XRF-4 mean z-score down with respect to the mean reference. However, the
XRF-4 mean z-score with respect to its own reference remains constant with only a slight
decrease in September 2018. This indicates that the low z-score values are from an inherent bias
in the XRF-4 Al measurement, and are not indicative of instrument change during the analysis
period. Changes in the XRF analysis protocol are being investigated to decrease the inter-
instrument bias.
' and
y '
Page|47
-------
Figure 4.2-5: Inter-instrument comparison by z-score of reanalysis sample set. Vertical red line denotes change in
re-analysis set. Multiple measures of the new re-analysis set during the month of June 2018 (denoted by A, B, and
C) were made for determination of reference values.
• XRF-1 vs. mean Ref O XRF-1 vs. XRF-1 Ref • XRF-4 vs. mean Ref 0 XRF-4 vs. XRF-4 Ref
1.00
0.00
-1.00
1.00
0.00
Al
-©¦ ® -®- o Vi © "8" § i"3, fi
•••••
Si
Sjo.p-gi-g.54-*
J O o
~-*»
~ • •
rcr®«--°-?r
° 0 O n
§ § -»-• o-cr 0-6- e -§-• •*#
Ol
f|«*»
(D
O
O
-1.00-
C/J
N
C
03
0)
1.00-
S 8 8 S * s S S § S . . * 2
»?J» « » 4 «-* « _
• • •
#-? *-
• •
Ca
to-0#
@2°
t?!
Ti
°-8 2-® 8-°-
•8 •
0.00
-1.00
Mn
Fe
Zn
L
M
K
1"!
w
288
»o*
nil
lliii'gi
»*
Hi
H
^-1
o ^ o
1 1 1 1
i i i i i i i i
i i
i i
1.00-
-1.00
Se
0-oo§§:<-• i
lulu
Sr
#*t •"•"It,
~i—i—i—i—i—i—i—i—i—i—i—i—i—r-h—i—i—i—i—H H—i—i—i—i—i—i—i—i—i—i—r~
r^h-h-r^r^r^r^r^r^coOTcocococQOO000000 r^r^r^h-r^-h-h-h-h-cococxD
ooooooooooooooS^^ooo oooooooooooo
r^r^r^^i^r^r^r^r^cococococo O C .Q jz >>(\| (\| r\i —j O) Q.
O C -O £r >>f\| CM CM Z3 uj Q. C —j
Q-(0 3 0)X o 0) (n Cc^ ^ 9-CD
<"5-5^
-------
The results for the analysis period are shown in Figure 4.2-6. Both XRF-1 and XRF-4 underwent
routine calibrations in January 2017 and January 2018. XRF-4 also underwent calibration in June
2018 due to a replacement of the CaF2 target in the secondary target wheel. The results from the
monthly NIST SRM 2783 analyses indicate that calibrations for both instruments are stable over
the one-year calibration period. The overall error for most elements is less than 20%. However,
the error in Zn is around 30%. Per Yatkin et al. (2016b), an XRF inter-laboratory comparison
reported SRM 2783 Zn error varying from -15% to 30%; the results shown here fit within that
range. The only acceptance criteria exceedance was for XRF-4 Al during August 2017 and June
2018 (prior to the new calibration). The error for Al on XRF-4 was near the acceptance limit and
had two measures outside of the limits. These two exceedances were within the normal variance
of the errors for this instrument (i.e. not outliers) and considering other QC filter results for Al
during these time periods, no instrument issues were suspected. After the June 2018 calibration,
the XRF-4 Al error was reduced and within the acceptance criteria.
Figure 4.2-6: Error of each XRF instrument from the NIST SRM standard run monthly.
Acceptance Limit ° XRF-1 ° XRF-4
25%
-25% -
Mn
o o
o°° o2e so
o o J
© -© 1
30
o
8
o o ° 0
©
© 9
Fe
10%-
5% -
o%-
-5% -
-10%-
ooo
' o ® 0
' o „ S> X
Cu
0
oo
oo ,
00 '
8 8 0 o 0 §
0 O
0 o 0
Pb
H
O O | o
o o v
i-~r-~-h~r^r^f^i^i^r--oococoaocococoooco
100% ~
ooco coco coco
; >c"5 o>Q-tj >oc:xjJ=|r>>c:-50)Q.
J-CD=3^:3(D>CO(Dcoaj^5-cn3-^=5 o £Z .Q )= >C5 O) a.
,D(D^O(D(Oa)® 9-CO 3
-------
4.2.3 Determination of Uncertainties and Method Detection Limits
For discussion of Method Detection Limits (MDLs) see Section 3.1.3.2.
For discussion of analytical uncertainty and total uncertainty see Section 3.1.2 and Section 6.5,
respectively.
4.2.4 Audits, Performance Evaluations, Training, and Accreditations
4.2.4.1 System Audits
The EPA did not conduct any audits or performance evaluations of the UC Davis XRF
laboratory during 2017.
4.2.4.2 Performance Evaluations
The UC Davis XRF laboratory actively participates in interlaboratory comparisons.
In 2018 (during the analysis period for samples collected during 2017), UC Davis participated in
an interlaboratory comparison with Environment and Climate Change Canada. Q1SO4 and CuO
reference materials, generated at UC Davis, were analyzed by XRF, IC, and ICP-MS. Results
indicate good agreement between the laboratories with less than 5% absolute difference.
Additionally, the EPA provided five PTFE samples for elemental analysis as part of the EPA
Office of Air Quality Planning and Standards (OAQPS) interlaboratory performance evaluation.
This evaluation was carried out during the timeframe when UC Davis was also analyzing CSN
samples collected during 2017. Reported z-scores were all below 2, indicating satisfactory
interlaboratory comparison results.
4.2.4.3 Training
Training of all personnel who assist with or operate the XRF instruments is mandatory through
UC Davis. Personnel in the XRF laboratory are required to take the following UC Davis safety
trainings: UC Laboratory Safety Fundamentals, Radiation Safety for Users of Radiation
Producing Machines, Analytical X-ray Quiz, and Cryogen Safety.
Only personnel listed in UC Davis CSN Quality Assurance Project Plan (QAPP), trained on the
appropriate SOPs and Technical Instructions (UCD CSN SOP #302 and CSN TI 302A-D\ and
authorized by the Laboratory Manager can perform XRF analysis on CSN samples.
4.2.4.4 A ccreditations
There are no accreditations for elemental analysis on aerosol filters by XRF.
4.2.5 Summary of Filter Field Blanks
Over the sampling period (January 1, 2017 through December 31, 2017) there were 1,367 valid
PTFE filter field blanks. Table 4.2-10 summarizes the field blank statistics.
Page|50
-------
Table 4.2-10: PTFE filter field blank statistics.
Species
Count
Median
(Ug/cm2)
Average
(Ug/cm2)
Min
(Ug/cm2)
Max
(Ug/cm2)
St. Dev.
(Ug/cm2)
Na
1367
0.000
0.014
-0.002
0.529
0.030
Mg
1367
0.013
0.016
0.000
0.188
0.015
A1
1367
0.081
0.083
0.051
0.431
0.022
Si
1367
0.010
0.016
0.000
0.691
0.041
P
1367
0.000
0.000
0.000
0.012
0.001
S
1367
0.000
0.004
0.000
0.845
0.041
CI
1367
0.003
0.004
0.000
0.271
0.009
K
1367
0.007
0.007
0.000
0.140
0.011
Ca
1367
0.003
0.010
0.000
0.587
0.033
Ti
1367
0.001
0.001
0.000
0.033
0.002
V
1367
0.000
0.000
0.000
0.003
0.001
Cr
1367
0.005
0.005
0.002
0.104
0.005
Mn
1367
0.007
0.007
0.000
0.036
0.002
Fe
1367
0.024
0.029
0.000
0.461
0.033
Co
1367
0.002
0.002
0.000
0.006
0.001
Ni
1367
0.001
0.002
0.000
0.031
0.002
Cu
1367
0.006
0.007
0.000
0.030
0.003
Zn
1367
0.003
0.003
0.000
0.061
0.003
As
1367
0.000
0.000
0.000
0.006
0.001
Se
1367
0.003
0.003
0.000
0.011
0.001
Br
1367
0.003
0.003
0.000
0.012
0.001
Rb
1367
0.004
0.005
0.001
0.013
0.002
Sr
1367
0.005
0.005
0.002
0.016
0.002
Zr
1367
0.025
0.026
0.000
0.065
0.011
Ag
1367
0.013
0.014
0.004
0.042
0.005
Cd
1367
0.013
0.014
0.003
0.044
0.006
In
1367
0.035
0.036
0.012
0.073
0.008
Sn
1367
0.044
0.045
0.017
0.097
0.012
Sb
1367
0.043
0.044
0.013
0.107
0.013
Cs
1367
0.065
0.067
0.000
0.139
0.020
Ba
1367
0.101
0.103
0.046
0.198
0.025
Ce
1367
0.118
0.121
0.047
0.240
0.029
Pb
1367
0.014
0.015
0.004
0.030
0.005
Page|51
-------
4.3 DRI Carbon Laboratory
The DRI Carbon Analysis Laboratory, as a subcontractor to UC Davis, received and analyzed
quartz filters from batches 22 through 38 for samples collected January 1, 2017 through
December 31, 2017. Analysis of these samples was performed May 8, 2017 through April 26,
2018. All analyses were performed using the DRI Model 2015 multi-wavelength carbon analyzer
with the IMPROVEA method and analysis results were reported to UC Davis. Twelve DRI
Model 2015 Thermal/Optical Carbon Analyzers (designated as units # 21, 31, 34-38, 40-43, and
47) were used for these CSN IMPROVE_A analyses.
4.3.1 Summary of QC Checks and Statistics
Samples received at the DRI Carbon Laboratory follow the chain-of-custody procedure specified
in DRI CSN SOP #2-231. This SOP is specific for the Chemical Speciation Network. Quality
control (QC) measures for the DRI carbon analysis are included in the SOP and summarized in
Table 4.3-1. The table specifies the frequency and standards required for the checks, along with
the acceptance criteria, and corrective actions for the carbon analyzers. More detail on individual
control measures is provided in specific subsections.
Page|52
-------
Table 4.3-1. DRI carbon analysis QC measures for Model 2015 analyzer.
QA/QC Activity
Calibration Standard and
Range
Calibration
Frequency
Acceptance Criteria
Corrective Action
System Blank Check
NAa
Once per week
<0.2 (ig C/cm2. See Table 4.3-2 and
Figure 4.3-1.
Check instrument.
Laboratory Blank
Check
NA
Beginning of
analysis day
<0.2 (ig C/cm2. See Table 4.3-3 and
Figure 4.3-2.
Check instrument and filter punch and
rebake
Calibration
Peak Area Check
NIST 5% CHi/He gas standard;
20 (ig C (6-port valve injection
loop, 1000 (il)
Every analysis
Typical counts 15,000-25,000 and
95-105% of average calibration peak
area of the day. See Figure 4.3-4.
Void analysis result; check flowrates,
leak, and 6-port valve temperature;
conduct an auto-calibration; and repeat
analysis with second filter punch.
Auto-Calibration
Check
NIST 5% CHi/He gas standard;
20 (ig C (Carle valve injection
loop, 1000 (il)
Alternating
beginning or
end of each
analysis day
Relative standard deviation of the
three injection peaks <5% and
calibration peak area 90-110% of
weekly average. See Table 4.3-4 and
Figure 4.3-3.
Verify if major maintenance has
occurred. Troubleshoot and correct
system before analyzing samples.
Manual Injection
Calibration
NIST 5% CHi/He or NIST 5%
CCb/He gas standards; 20 (ig C
(Certified gas-tight syringe, 1000
Hi)
Four times a
week (Sun.,
Tue., Thu., and
Sat.)
95-105% recovery and calibration
peak area 90-110% of weekly
average. See Figure 4.3-5a.
Troubleshoot and correct system
before analyzing samples.
Sucrose Calibration
Check
10|xL of 1800 ppm C sucrose
standard; 18 (ig C
Thrice per
week (began
March, 2009)
17.1-18.9 (ig C/lilter. See Figure
4.3-5b.
Troubleshoot and correct system
before analyzing samples.
Potassium Hydrogen
Phthalate (KHP)
Calibration Check
10|xL of 1800 ppm C KHP
standard; 18 (ig C
Twice per
week (Tue. and
Thu.)
17.1-18.9 (ig C/lilter. See Figure
4.3-5c.
Troubleshoot and correct system
before analyzing samples.
Multiple Point
Calibrations
1800 ppm C Potassium hydrogen
phthalate (KHP) and sucrose;
NIST 5% CHi/He, and NIST 5%
CCte/He gas standards; 9-36 (ig
C for KHP and sucrose; 2-30 (ig
C for CH4 and CO2
Every six
months or after
major
instrument
repair
All slopes ±5% of average. See
Table 4.3-5.
Troubleshoot instrument and repeat
calibration until results are within
stated tolerances.
Sample Replicates
(on the same or a
different analyzer)
NA
Every 10
analyses
±10% when OCR and TCR >10 (ig
C/cm2
±20% when ECR > 10(ig C/cm2 or
<±1 (ig/cm2 when OCR and TCR
<10 (ig C/cm2
<±2 (ig/cm2 when ECR <10(ig C/cm
See Table 4.3-8 and Figure 4.3-6.
Investigate instrument and sample
anomalies and rerun replicate when
difference is > ±10%.
Temperature
Calibrations
NIST-certilied thermocouple
Every six
months, or
whenever the
thermocouple
is replaced
Linear relationship between analyzer
and NIST thermocouple values with
R2>0.99. See Table 4.3-6.
Troubleshoot instrument and repeat
calibration until results are within
stated tolerances.
Oxygen Level in
Helium Atmosphere
(using GC/MS)b
Certified gas-tight syringe; 0-
100 ppmv
Every six
months, or
whenever leak
is detected
Less than the certified amount of He
cylinder. See Table 4.3-7.
Replace the He cylinder and/or O2
scrubber.
a NA: Not Applicable.
b Gas chromatography/mass spectrometer (Model 5975, Agilent Technology, Palo Alto, CA, USA).
4.3.2 Summary of QC Results
Detailed results of the carbon QC are presented in the subsections below. All system blanks
(Table 4.3-2) or laboratory blanks (Table 4.3-3) that did not meet the acceptance criteria were
reanalyzed and if they did not pass the second analysis, instrument maintenance was performed
and additional blanks were run before the analyzer was placed on-line. Exceedance in multipoint
calibrations (Table 4.3-5) result in verification of individual calibration points, troubleshooting
Page|53
-------
the instrument, and repeating calibrations. Exceedances in auto-calibrations (Table 4.3-4),
internal calibrations (Figure 4.3-4), as well as CO2 (Figure 4.3-5a), sucrose (Figure 4.3-5b), and
KHP (Figure 4.3-5c) calibrations result in reanalysis and/or instrument maintenance. For cases
where CSN samples were analyzed after an exceedance, data were flagged with the QX (Does
Not Meet QC Criteria) qualifier in files delivered to AQS by UC Davis (see Section 2.3.2 and
Section 3.2.3.1). As a corrective action, software tools are being developed to generate QC
control charts and summaries to ensure QC exceedances are captured and corrected immediately.
4.3.2.1 System and Laboratory Blanks
Table 4.3-2 lists the number of system blanks analyzed during the report period and their
concentration statistics. The system blank control chart is shown in Figure 4.3-1. System blanks
are used to ensure that the system is not introducing bias in the carbon analysis. Most system
blanks were below the limit of 0.2 |igC/cm2. When an exceedance is observed, possible
contamination is checked, parts are cleaned, the sample oven is baked, and a second system
blank is rerun to ensure that it passes the criterion. One system blank did not pass the acceptance
criteria and was rerun to pass.
Table 4.3-2: Statistics of system blanks ran on the Model 2015 analyzer between 5/8/2017 and 5/26/2018.
Elemental carbon (EC) fractions are indicated as (1) through (3), organic carbon (OC) fractions are indicated as (1)
through (4). Organic pyrolyzed (OP), elemental carbon (EC), and organic carbon (OC) are shown by reflectance (R)
and transmittance (T). Total carbon is shown be reflectance (TCR).
Parameter
Count
Median
frig/cm2)
Average
frig/cm2)
Min
frig/cm2)
Max
frig/cm2)
St. Dev.
frig/cm2)
# Exceedance
OC1
465
0.000
0.000
0.000
0.006
0.000
0
OC2
465
0.000
0.002
0.000
0.049
0.006
0
OC3
465
0.000
0.010
0.000
0.117
0.018
0
OC4
465
0.000
0.001
0.000
0.084
0.006
0
OCR
465
0.000
0.014
0.000
0.171
0.026
0
OCT
465
0.001
0.015
0.000
0.228
0.029
1
OPR
465
0.000
0.001
0.000
0.074
0.005
0
OPT
465
0.000
0.002
0.000
0.074
0.008
0
EC1
465
0.000
0.000
0.000
0.016
0.001
0
EC2
465
0.000
0.002
0.000
0.064
0.006
0
EC3
465
0.000
0.000
0.000
0.060
0.003
0
ECR
465
0.000
0.001
0.000
0.062
0.006
0
ECT
465
0.000
0.000
0.000
0.020
0.001
0
TCR
465
0.001
0.015
0.000
0.228
0.029
1
Page|54
-------
Figure 4.3-1: Control chart of system blank total carbon by reflectance (TCR) concentrations on the DRI Model
2015 carbon analyzers. The red dash lines indicate the limit of 0.2 |ig C/cnf.
^ nn
System Blank; n=465
— 0.90
| 0.80
O) 0.70
2.
^ 0.60
~ 0.50
(0
tz 0.40
c
o 30
o
O °-20
O °-10
I- 0.00
•
—
A
•
3
ih1!
•
•
I * 9ld
5/7/2017
6/6/2017
7/6/2017
8/5/2017
9/4/2017
§" 10/4/2017
£2.
11/3/2017
(0
a? 12/3/2017
r+
CD
1/2/2018
2/1/2018
3/3/2018
4/2/2018
5/2/2018
Table 4.3-3 lists the number of laboratory blanks analyzed during the report period and their
concentration statistics. The laboratory blank control charts are shown in Figure 4.3-2.
Laboratory blank analyses are performed daily to check for system contamination and evaluate
laser response. Most laboratory blanks were below the limit of 0.2 |igC/cm2. When an
exceedance is observed, the sample oven is baked and a second laboratory blank is run. If the
second blank still exceeds the limit, the analyzer is taken offline for cleaning and maintenance. A
total of 37 CSN samples were run after failed laboratory blanks; these cases were flagged with
the QX (Does Not Meet QC Criteria) qualifier in files delivered to AQS by UC Davis (see
Section 2.3.2 and Section 3.2.3.1). However, the sucrose analysis that immediately followed the
failed laboratory blank indicated acceptable values.
Page|55
-------
Table 4.3-3: Statistics of laboratory blanks run on the Model 2015 analyzer between 5/8/2017 and 5/26/2018.
Elemental carbon (EC) fractions are indicated as (1) through (3), organic carbon (OC) fractions are indicated as (1)
through (4). Organic pyrolyzed (OP), elemental carbon (EC), and organic carbon (OC) are shown by reflectance (R)
and transmittance (T). Total carbon is shown be reflectance (TCR).
Parameter
Count
Median
frig/cm2)
Average
frig/cm2)
Min
frig/cm2)
Max
frig/cm2)
St. Dev.
frig/cm2)
# Exceedance
OC1
4519
0.000
0.000
0.000
0.093
0.003
0
OC2
4519
0.000
0.001
0.000
0.688
0.013
1
OC3
4519
0.000
0.005
0.000
0.693
0.022
9
OC4
4519
0.000
0.001
0.000
1.060
0.020
3
OCR
4519
0.000
0.008
0.000
2.058
0.048
34
OCT
4519
0.000
0.010
0.000
2.450
0.062
39
OPR
4519
0.000
0.001
0.000
0.238
0.007
1
OPT
4519
0.000
0.003
0.000
1.339
0.028
3
EC1
4519
0.000
0.001
0.000
1.090
0.017
2
EC2
4519
0.000
0.003
0.000
0.884
0.018
3
EC3
4519
0.000
0.000
0.000
0.105
0.002
0
ECR
4519
0.000
0.003
0.000
1.339
0.028
5
ECT
4519
0.000
0.001
0.000
0.290
0.008
3
TCR
4519
0.000
0.011
0.000
2.450
0.065
41
Figure 4.3-2: Control chart of daily laboratory blank total carbon by reflectance (TCR) concentrations ran on the
DRI Model 2015 carbon analyzers. The red dash lines indicate the limit of 0.2 |igC7cnr.
-a n
Laboratory Blank;N=4519
CM
E 2.5
O
O)
3 2.0
c
0
1 1.5
+¦»
i i.o
c
o
° 0.5
O
H
0.0
•
•
•
•
•
•
V
•
•
• |
rrc
5/7/2017
6/6/2017
7/6/2017
8/5/2017
9/4/2017
> 10/4/2017
3
5L
^ 11/3/2017
<7S"
a? 12/3/2017
(D
1/2/2018
2/1/2018
3/3/2018
4/2/2018
5/2/2018
4.3.2.2 Auto-Calibration and Internal Calibration Peak Area Check
Once per day each analyzer runs an auto-calibration protocol. Using the Carle valve, an aliquot
of methane standard is injected once in a He-only atmosphere (organic carbon stage), once in a
He/02 atmosphere (elemental carbon stage), and finally as the normal internal calibration peak.
The three peaks should have similar peak areas if the catalysts are in good condition. The
Page|56
-------
similarity of the three peaks are measured by the relative standard deviation (RSD), which is the
standard deviation divided by the average of the three peak areas. The acceptance limit is RSD
<5% and ±10% from weekly average. Table 4.3-4 summarizes the RSD of the three methane
injection peaks during the analysis period and the control chart is shown in Figure 4.3-3. There
were 32 exceedances and most of them occurred when the analyzer was under maintenance and
no samples were run. When an exceedance is observed, the analyzer is checked and the auto-
calibration is rerun. The calibration peak areas of previous runs are examined and/or manual
injections are done to ensure the analyzer is working properly. A total of 106 CSN samples were
analyzed during auto-calibration peak area exceedances; these cases were flagged with the QX
(Does Not Meet QC Criteria) qualifier in files delivered to AQS by UC Davis (see Section 2.3.2
and Section 3.2.3.1).
Table 4.3-4: Statistics of the relative standard deviation (RSD) of the three methane injection peaks from auto-
calibration checks.
Statistic
Auto-Calibration
Count
3989
Median
0.8%
Average
1.1%
Min
0.0%
Max
13.2%
Standard deviation
1.0%
Exceedance
32
Figure 4.3-3: Control chart of the relative standard deviation of the three methane injection peaks from daily auto-
calibration ran on the DRI Model 2015 carbon analyzers. The red dash lines indicate the limit of 5% RSD.
Auto Calibration; N=3989
14%
12%
Analysis Date
At the end of each run, a fixed amount of methane is injected via a Carle valve as an internal
calibration standard. The internal calibration peak area is examined for each sample. Significant
changes in calibration peak area counts are monitored and instruments are checked for
Page|57
-------
performance against daily calibrations. Typical ranges for the internal calibration peaks fall
between 15,000 and 25,000 counts for Model 2015. In addition to peak area ranges, the peak
areas are also compared to the daily averages. Sudden changes or atypical counts result in
instrument maintenance. Metadata concerning QC measures and instrument maintenance are
reported to UC Davis quarterly. Figure 4.3-4 shows the daily internal calibration peak area
during the reporting period for all analyzers. For the Model 2015, 16,030/16,111 (99.5%) passed
both peak area and daily average criteria. A total of 96 original analyses and two replicate
analyses exceeded internal calibration QC limits; these cases were flagged with the QX (Does
Not Meet QC Criteria) qualifier in files delivered to AQS by UC Davis (see Section 2.3.2 and
Section 3.2.3.1). However, other QC analyses (e.g., replicates, auto-calibration, and internal
calibration peak area check) within the time period indicate acceptable values. One sample,
quartz filter F052584, did not meet QC criteria for the calibration peak; the filter was
subsequently invalidated by UCD with the AS (Poor QA Results) null code and updated in AQS.
Figure 4.3-4: Control chart of the internal calibration peak area for the DRI Model 2015 carbon analyzers. The red
dash lines indicate the typical internal calibration peak area between 15,000 and 25,000 for Model 2015. Sample
F052584 (collected 2/24/2017, analyzed 6/14/2017) did not meet QC criteria and was invalidated.
Internal Calibration; N=16,111
«f 35000
C
g 30000
o
^ 25000
0)
20000
rc
0)
a.
c
o
•*->
2
£
ra
O
15000
10000
5000
• I
I
Analysis Date
4.3.2.3 Multipoint Calibration and Manual Injection Check
Multipoint carbon calibrations are performed semi-annually or whenever major repairs or
changes are made to the instruments. The calibration uses four different sources of carbon:
methane (CH4), carbon dioxide (CO2), sucrose (C12H22O11), and potassium hydrogen phthalate
(KHP), each with four injections with different carbon content (except that 15 |iL sucrose and
KHP are injected twice) resulting in a total of 18 calibration points in the set. The calibration
result is plotted as |ig carbon in the calibration standard versus total carbon peak area normalized
by the internal calibration peak area. A regression slope is obtained by fitting the calibration
points with a linear line forced through the origin. The slope relates the measured normalized
peak area to carbon content. It represents the response of the entire analyzer to generic carbon
compounds, including the efficiency of the oxidation oven and sensitivity of the NDIR. If the
ratio of carbon over normalized peak area for individual calibration point differs from the
regression slope by more than 10%, the calibration point is treated as an outlier and redone.
Daily calibration injections and replicate analysis also verify acceptable slopes. Table 4.3-5
Page|58
-------
provides summary statistics for full multipoint calibrations by analyzer for the period during
which the project samples were analyzed. The QC criterion requires the slope to be within ±5%
of average by each analyzer (Table 4.3-1), where the slope is obtained after individual calibration
outlier points are removed and redone. There were five cases where samples were run with
carbon calibration slopes outside of the QC criteria (all on Analyzer #47); these data were
flagged with the QX (Does Not Meet QC Criteria) qualifier in files delivered to AQS by UC
Davis (see Section 2.3.2 and Section 3.2.3.1).
Table 4.3-5: Multipoint calibration statistics (CSN sample dates 1/1/2017-12/31/17). Units for the slope are |ig
carbon per ratio of standard injection peak count/calibration gas peak count. For analyzers 31, 37, 38, 40, 41, and 43,
more than 6 months passed between calibrations. As noted in Sections 2.3.2 and 3.2.3.1, a calibration calendar lias
been established by DRI to avoid recurrence of this issue.
Carbon
Analyzer
Calibration
Date
Slope
Difference from
Analyzer Average
# of Samples
Flagged
Comment
21
6/2/2017
19.963
0.998
3%
0
9/22/2017
18.841
0.994
-3%
0
11/30/2017
19.658
0.995
1%
0
5/30/2018
19.348
0.998
-1%
0
31
6/12/2017
19.703
0.997
0%
0
8/16/2017
19.565
0.993
0%
0
2/21/2018
19.598
0.998
0%
0
32
10/24/2017
18.319
0.996
5%
0
11/20/2017
17.191
0.996
-2%
0
1/9/2018
17.256
0.991
-1%
0
6/13/2018
16.664
0.996
-5%
0
6/29/2018
17.870
0.998
2%
0
34
5/5/2017
19.238
0.999
0%
0
9/12/2017
18.591
0.999
-3%
0
11/14/2017
19.561
0.996
2%
0
11/27/2017
18.326
0.996
-5%
0
12/5/2017
19.668
0.998
2%
0
3/13/2018
19.464
0.995
1%
0
a
3/15/2018
19.486
0.995
2%
0
35
6/8/2017
19.343
0.996
1%
0
10/9/2017
18.62
0.995
-3%
0
12/5/2017
19.421
0.999
1%
0
1/26/2018
19.808
0.995
3%
0
4/9/2018
18.636
0.994
-3%
0
36
4/3/2017
19.37
0.992
2%
0
a
4/28/2017
19.307
0.997
2%
0
6/23/2017
19.065
0.997
0%
0
11/3/2017
18.698
0.995
-2%
0
11/8/2017
19.153
0.996
1%
0
Page|59
-------
Carbon
Analyzer
Calibration
Date
Slope
r2
Difference from
Analyzer Average
# of Samples
Flagged
Comment
1/4/2018
19.274
0.991
1%
0
5/9/2018
19.175
0.996
1%
0
a
5/14/2018
18.389
0.991
-3%
0
a
5/16/2018
18.888
0.998
-1%
0
a
5/24/2018
18.742
0.993
-1%
0
a
5/30/2018
18.853
0.995
-1%
0
37
5/3/2017
18.814
0.994
-1%
0
12/18/2017
19.035
0.993
0%
0
b
1/5/2018
18.830
0.996
-1%
0
6/1/2018
19.180
0.998
1%
0
38
5/10/2017
18.478
0.995
-2%
0
7/2/2017
18.804
0.992
0%
0
8/17/2017
18.912
0.996
0%
0
a
8/18/2017
19.074
0.997
1%
0
11/6/2017
18.858
0.997
0%
0
6/21/2018
18.973
0.991
1%
0
b
6/26/2018
18.964
0.996
1%
0
40
6/21/2017
19.350
0.993
0%
0
8/3/2017
19.178
0.993
-1%
0
2/13/2018
19.391
0.993
0%
0
41
7/5/2017
19.254
0.995
0%
0
8/17/2017
19.480
0.994
1%
0
2/26/2018
18.982
0.996
-1%
0
42
9/28/2017
19.443
0.997
1%
0
1/19/2018
18.897
0.992
-2%
0
4/2/2018
19.399
0.998
1%
0
5/17/2018
19.090
0.997
-1%
0
43
5/19/2017
18.580
0.995
0%
0
9/8/2017
19.125
0.995
3%
0
6/5/2018
18.022
0.971
-3%
0
c
47
7/15/2017
17.520
0.991
-6%
5
8/1/2017
19.013
0.998
2%
0
9/11/2017
18.604
0.995
0%
0
1/26/2018
18.709
0.996
1%
0
4/18/2018
18.876
0.997
2%
0
a Carbon calibration repeated
b Calibration overdue, new QA calendar system implemented to prevent delay
0 Instrument offline from February 2018 to June 2018
Page|60
-------
C02 calibrations are performed on each analyzer four times per week, sucrose calibration checks
are done on each analyzer three times per week, and KHP calibrations are done twice per week.
Calibration control charts for the Model 2015 analyzers are shown in Figures 4.3-5a through 4.3-
5c. For the period during analysis of 2017 CSN samples, 266 out of 1,937 CO2 calibrations, 190
out of 2,208 sucrose calibration, and 127 out of 1,599 KHP calibrations exceeded the criteria.
When an exceedance is observed, the analyzer is checked and the calibration is rerun. No CSN
samples were run after any of the CO2 calibration exceedances. However, there were 30 and 25
CSN samples analyzed after the sucrose and KHP exceedances, respectively; these data were
flagged with the QX (Does Not Meet QC Criteria) qualifier in files delivered to AQS by UC
Davis (see Section 2.3.2 and Section 3.2.3.1). However, for all samples that were run after an
exceedance calibration, other QC analyses (e.g., replicates, auto-calibration, and internal
calibration peak area checks) within the time period indicate acceptable values.
Figure 4.3-5: Control chart of manual calibration checks for: (a) CO2, (b) sucrose, and (c) KHP injections. The red
dash lines indicate the total carbon limits of 17.1 and 18.9 |igC per injection for sucrose and KHP and 19.57 and
21.63 |iC per injection for CO2.
(a)
24
c
23
0
'•*->
0
22
'c
21
o>
3
20
c
0
19
Si
i_
ro
18
O
3
17
0
H
16
C02 Calibration; N=1937
Analysis Date
(b)
Sucrose Calibration; N=2028
22
21
c
O
20
0
0
'E
19
o>
3
18
c
0
17
-Q
i—
ro
16
O
2
15
0
1-
14
Analysis Date
Page | 61
-------
(C)
22
"c 21
O
O 20
o
19
18
O
O
£
(5
o
KHP Calibration; N=1599
17
16
iS 15
* 14
•.-W ¦ •
Analysis Date
4.3.2.4 Temperature Calibrations
Table 4.3-6 provides summary statistics for the multi-point temperature calibrations of each
Model 2015 carbon analyzer. The temperature calibrations are performed every six months or
after a major instrument repair. Criteria for an acceptable calibration is linear regression
coefficient of determination (r2) >0.99. Separate linear regressions are used for the lower
temperatures and higher temperature ranges. These two ranges are separated with a toggle point
typically around 200-300 °C, which is set to the temperature at which the two regression lines
intercept (see Figure 3-6 in Model 2015 SOP). All calibrations met the acceptable r2 criteria
(r2>0.99) during this report period.
Table 4.3-6: Multi-point temperature calibration statistics on the Model 2015 carbon analyzer (CSN sample dates
1/1/2017-12/31/2017). For analyzers 21 and 40, more than 6 months passed between calibrations. As noted in
Sections 2.3.2 and 3.2.3.1, a calibration calendar lias been established by DRI to avoid recurrence of this issue.
Carbon
Calibration
Low T
Low T
Low T
HighT
HighT
HighT
Analyzer
Date*
Slope
Intercept
Slope
Intercept
r2
21
5/25/2017
1.082
2.647
0.996
1.007
21.494
0.999
11/22/2017
0.999
9.900
1.000
0.970
15.804
1.000
5/25/2018
1.038
11.702
0.999
0.989
25.596
0.999
31
8/31/2017
1.003
27.743
1.000
1.024
24.987
0.995
2/8/2018
1.066
1.937
0.999
1.016
12.757
1.000
3/29/2018
1.068
2.815
1.000
1.015
18.449
1.000
32
8/28/2017
1.086
10.185
1.000
1.013
29.379
0.996
Offline
10/2/2017
1.129
7.970
0.999
1.022
35.979
0.993
12/4/17- 7/16/18
5/31/2018
1.071
-0.507
0.999
0.982
24.810
0.999
34
8/1/2017
1.110
4.964
1.000
1.015
31.440
0.995
8/24/2017
1.079
3.448
1.000
1.019
19.294
0.997
10/26/2017
1.125
0.161
0.999
1.027
26.695
0.995
3/12/2018
1.055
2.648
0.999
1.018
14.510
0.999
Page|62
-------
Carbon
Analyzer
Calibration
Date*
Low T
Slope
Low T
Intercept
Low T
r2
HighT
Slope
HighT
Intercept
HighT
r2
7/16/2018
1.069
2.360
1.000
1.001
20.822
1.000
35
8/28/2017
1.053
2.086
1.000
0.996
15.357
0.999
10/5/2017
0.989
15.789
1.000
0.999
13.220
0.999
4/4/2018
1.059
6.988
0.999
0.998
23.598
0.999
5/16/2018
0.996
12.725
0.999
0.976
16.756
0.999
36
5/23/2017
1.046
8.541
0.998
1.010
15.871
1.000
6/20/2017
1.057
8.064
1.000
1.010
7.437
0.999
10/31/2017
1.028
10.745
1.000
0.989
19.705
0.999
2/26/2018
1.041
7.715
1.000
1.028
1.777
0.998
4/26/2018
0.841
19.276
0.997
0.877
7.824
1.000
5/21/2018
0.979
8.527
0.999
1.004
2.841
1.000
6/5/2018
0.998
14.056
1.000
0.986
19.062
1.000
37
12/7/2017
1.034
21.554
1.000
0.979
35.483
0.997
2/7/2018
1.052
0.408
1.000
0.988
14.824
1.000
5/29/2018
1.060
4.289
1.000
0.978
27.330
1.000
38
6/20/2017
1.034
2.173
1.000
1.024
5.939
0.999
10/31/2017
1.054
-2.295
**
0.996
13.166
**
2/21/2018
1.076
0.097
1.000
1.024
14.323
1.000
6/18/2018
1.096
2.330
1.000
1.006
27.547
1.000
40
8/1/2017
1.053
2.086
1.000
0.996
15.357
0.999
2/5/2018
1.034
1.629
1.000
1.020
5.244
1.000
5/11/2018
0.994
10.153
0.999
0.982
14.964
1.000
41
8/22/2017
1.033
4.682
1.000
1.027
7.430
1.000
2/22/2018
1.033
4.682
1.000
1.027
7.430
1.000
7/23/2018
1.025
9.735
1.000
1.005
16.973
1.000
42
9/26/2017
1.086
3.827
1.000
1.020
21.993
0.997
12/26/2017
1.112
-1.727
1.000
1.013
24.237
0.996
2/26/2018
1.022
4.927
0.999
1.026
3.574
1.000
3/30/2018
1.049
6.483
1.000
1.026
14.628
1.000
5/15/2018
0.988
10.325
1.000
0.988
9.498
1.000
43
10/30/2017
1.125
0.161
1.000
1.027
26.695
0.995
2/12/2018
1.146
-4.097
1.000
1.013
24.664
0.996
5/18/2018
1.050
5.926
1.000
1.005
19.394
1.000
5/29/2018
1.060
4.289
1.000
0.978
27.330
1.000
47
7/12/2017
1.094
10.587
1.000
1.016
31.998
1.000
9/20/2017
1.110
14.054
1.000
1.005
41.559
1.000
1/17/2018
1.062
9.538
1.000
1.013
23.398
1.000
2/22/2018
1.085
5.700
1.000
1.018
24.721
1.000
* Includes both regular maintenance and semi-annual calibration data
** Calibration point data were deleted from file, therefore r2 data not available
Page|63
-------
4.3.2.5
Oxygen Level Check
Table 4.3-7 provides a summary of the Model 2015 oxygen leak test results that are performed
every six months or after major instrument repairs. The results are considered acceptable if the
O2 concentration is < 100 ppm. The O2 contents were well below 100 ppm, in the range of 8-74
ppm.
Table 4.3-7: Model 2015 oxygen test statistics (CSN sample dates 1/1/2017-12/31/2017).
Carbon
Analyzer
O2
Statistics
Feb 2017
Aug 2017
Feb 2018
(ppm)
140 (°C)
580 (°C)
140 (°C)
580 (°C)
140 (°C)
580 (°C)
21
Mean O2
28.1
21.3
14.1
10.9
17.9
16.7
Std Dev
8.9
5.0
0.3
0.1
4.7
4.6
31
Mean O2
19.5
18.1
20.0
19.3
19.8
18.3
Std Dev
5.4
5.1
0.0
0.1
4.6
4.6
32
Mean O2
Std Dev
N/A
N/A
18.7
0.7
13.9
0.5
24.8
4.7
26.5
4.8
34
Mean O2
55.3
74.4
12.3
8.7
39.1
50.5
Std Dev
8.1
10.2
0.1
0.0
5.6
5.6
35
Mean O2
28.5
21.4
19.8
19.9
22.6
26.6
Std Dev
5.3
5.2
0.0
0.4
4.7
4.8
36
Mean O2
20.0
21.0
24.3
24.3
20.0
22.7
Std Dev
5.1
5.5
0.0
0.0
4.7
4.7
37
Mean O2
34.8
25.5
21.0
15.7
18.8
16.6
Std Dev
8.6
5.2
0.7
0.1
4.9
4.6
38
Mean O2
22.4
23.5
20.3
18.5
31.2
28.1
Std Dev
5.1
5.3
1.2
1.3
4.9
4.7
40
Mean O2
33.1
24.8
16.7
19.9
24.3
25.3
Std Dev
7.6
5.3
0.2
0.1
4.7
4.8
41
Mean O2
29.0
24.4
14.2
14.5
23.8
20.9
Std Dev
6.7
5.3
3.0
1.4
4.8
4.7
42
Mean O2
21.8
21.0
14.5
14.9
17.5
16.7
Std Dev
6.9
5.2
0.4
0.0
4.8
4.7
43
Mean O2
24.3
19.3
21.7
14.5
26.7
24.8
Std Dev
5.8
5.14
1.3
0.3
5.0
4.8
47
Mean O2
26.1
22.9
19.1
17.0
17.8
16.8
Std Dev
7.6
5.2
0.5
0.6
4.7
4.8
Page|64
-------
4.3.2.6 Replicate and Duplicate Analyses
Replicate analysis results are from two or more punches of the same sample filter analyzed on
different instruments, while duplicate analysis results are from two punches of the same sample
filter analyzed on the same instruments. No valid duplicate analyses are available for this
reporting period. A replicate analysis was performed randomly on one sample from every group
of 10 samples. Table 4.3-8 gives the criteria and summary statistics for replicate IMPROVEA
carbon analyses during the reporting period January 1, 2017 through December 31, 2017.
Control charts for replicate analyses are plotted in Figure 4.3-6.
Replicate analysis results for total carbon (TCR), organic carbon (OCR), and elemental carbon
(ECR) by reflectance agree well, with only 22/5328 data points (0.41%) for OCR, ECR, and
TCR exceeded the criteria. The small size (25 mm) of the filter used in the IMPROVE A carbon
analysis method does not permit more than three punches (each -0.5 cm2) to be taken from the
filter. Samples not meeting replicate criteria (i.e., for TCR, OCR, or ECR < 10 [j,g C/cm2, TCR,
OCR < ± 1.0 [j,g C/cm2 and ECR < ± 2.0 [j,g C/cm2; and for TCR, OCR or ECR > 10 [j,g C/cm2,
TCR or OCR < 10% RPD and ECR < 20% RPD) are re-analyzed, typically on a third analyzer.
Filter inhomogeneity, which is flagged prior to first analysis, is also examined.
Table 4.3-8: Replicate analysis criteria and statistics (CSN sample dates 1/1/2017-12/31/2017). Total carbon (TCR),
organic carbon (OCR), and elemental carbon (ECR) are shown by reflectance.
Range
Criteria
Replicates
Statistic
No. TCR
OCR
ECR
TCR, OCR, & ECR
TCR, OCR < ±1.0 |ig C/cm2
Count
112
406
1528
<10 |ig C/cm2
ECR < ±2.0 |ig C/cm2
No. Fail
1
5
0
%Fail
0.90
1.25
0
Units: |ig C/cm2
Mean
0.34
0.21
0.22
StdDev
0.25
0.22
0.20
Max
1.43
1.84
1.49
Min
0.01
0.00
0.00
Median
0.30
0.14
0.17
TCR, OCR, & ECR
TCR, OCR %RPD < 10%
Count
1664
1370
248
> 10 |ig C/cm2
ECR %RPD < 20%
No. Fail
12
4
0
%Fail
0.73
0.29
0
Units: %
Mean
2.15
2.23
4.64
StdDev
1.86
1.71
3.30
Max
16.64
18.23
16.74
Min
0.00
0.00
0.07
Median
1.73
1.94
3.96
Note: RPD = 100 x absolute value [original sample-duplicate sample]/[(original sample+ duplicate sample)/2]
Page|65
-------
Figure 4.3-6: Replicate (two punches from the same sample filter analyzed on different instruments) analysis
results. The limits are ±1.0 |ig/cm2 for TCR and OCR <10 |ig/cm2. ±2.0 |ig/cm2 for ECR <10 |ig/cm2. ±10% relative
percent difference for TCR and OCR >10 |ig/cm2. and ±20% relative percent difference for ECR >10 ng/cm2.
TC: 0-10 |jg/cm2; N=112
TC>10 |jg/cm2; N=1664
1.5
E
i; i.o
(U
t 0.0
b
o
l- -0.5
0)
ro
u
=5. "10
a>
a:
•
•
•
•• •
*%
c • *
• •
**
• •
•o
• •
•
#•
•
*
•
<• •
• •
. • •
•
• •
• •
•
•••
•
\ •
' \ •
•
•
• •
•
•
0 20 40 60 80 100 120
Sample # with Increasing TC (0-10 jjg/cm2)
£
b
a:
o
Q.
10 jjg/cm2)
OC: 0-10 |jg/cm2; N=406
OC >10 |jg/cm2; N=1370
£
o
O)
1.0
t 0.0
b
o
=5. -10
a>
a:
•
•
•
i*
•
• •• ••
• v •. •
•
1
r
•
«
•
r
0 100 200 300 400
Sample #with Increasing OC (0-10 ^g/cm2)
500
0 200 400 600 800 1000 1200 1400
Sample # with Increasing OC (>10 jjg/cm2)
EC: 0-10 |jg/cm2; N=1528
EC >10 Mg/cm2; N=248
2.5
£
2.0
o
D)
1.5
3.
0.5
a:
-2.0
-2.5
0 400 800 1200
Sample #with Increasing EC (0-10 ^g/cm2)
1600
£
~
¦u
u
Z
0.
aj
EE
.M
Cl
te.
25
20
15
10
5
0
-5
¦10
-15
'20
-25
*
• • • t
**\ * * * *
• ¦
** "•
+*:
C*.V
*
•f •
•
• *
* ••
0 Stt 100 1*0 m !50
Sample # with Increasing EC (>10 |jg./cmJ)
MO
Page|66
-------
4.3.3 Determination of Uncertainties and Method Detection Limits
For discussion of Method Detection Limits (MDLs) see Section 3.1.3.2.
For discussion of analytical uncertainty and total uncertainty see Section 3.1.2 and Section 6.5,
respectively.
4.3.4 Audits, Performance Evaluations, Training, and Accreditations
4.3.4.1 System Audits
The prime contractor (UC Davis) provided 9 quartz audit samples collected during 2017 to DRI
for carbon analysis in March 2018. DRI delivered the carbon analysis data on April 18, 2018.
The average relative percentage difference between results from the DRI and the reference
laboratory (UC Davis) are -2% (negative number meaning DRI lower than reference), -18% and
-6% for OC, EC, and TC.
4.3.4.2 Performance Evaluations
The EPA provided 5 quartz samples for carbon analysis as part of the EPA Office of Air Quality
Planning and Standards (OAQPS) interlaboratory performance evaluation. This evaluation was
carried out during the timeframe when DRI also analyzed CSN samples collected during 2017.
Reported z-scores were all below 2, indicating satisfactory interlaboratory comparison results.
4.3.4.3 Training
All new laboratory staff receive training in performing the tasks in the SOPs for their assigned
work.
4.3.4.4 A ccreditations
There are no accreditation programs for analysis of carbon on aerosol filters by TOA.
4.3.5 Summary of Filter Blanks
Over the sampling period (January 1, 2017 through December 31, 2017) there were 1,298 quartz
filter field blanks. Table 4.3-9 summarizes the field blank statistics. The lower quantifiable limits
(LQLs) are defined as three times the standard deviation of field blanks and are used as internal
QA indicators, distinct from the MDLs reported to AQS.
Page|67
-------
Table 4.3-9: Quartz filter field blank statistics. Elemental carbon (EC) fractions are indicated as (1) through (3),
organic carbon (OC) fractions are indicated as (1) through (4). Organic pyrolyzed (OP), EC, and OC are shown by
reflectance (R) and transmittance (T).
Species
Count
Median
frig/cm2)
Average
frig/cm2)
Min
frig/cm2)
Max
frig/cm2)
St. Dev.
frig/cm2)
LQL
frig/cm2)
EC1
1298
0.000
0.011
0.000
5.435
0.192
0.575
EC2
1298
0.000
0.011
0.000
0.779
0.040
0.121
EC3
1298
0.000
0.000
0.000
0.000
0.000
0.000
ECR
1298
0.000
0.017
0.000
4.942
0.186
0.558
ECT
1298
0.000
0.008
0.000
3.812
0.124
0.372
OC1
1298
0.136
0.162
0.000
1.092
0.158
0.473
OC2
1298
0.262
0.283
0.000
2.736
0.155
0.465
OC3
1298
0.504
0.553
0.043
6.172
0.330
0.989
OC4
1298
0.000
0.035
0.000
4.276
0.148
0.444
OCR
1298
0.979
1.038
0.127
13.184
0.614
1.842
OCT
1298
0.982
1.048
0.127
14.314
0.664
1.993
OPR
1298
0.000
0.006
0.000
1.560
0.050
0.149
OPT
1298
0.000
0.015
0.000
3.705
0.115
0.344
5. Data Management and Reporting
5.1 Number of Events Posted to AQS
Summarized in Table 5.1-1 are dates that data were delivered to AQS for samples collected
January 1, 2017 through December 31, 2017. Data are expected to be delivered to AQS within
120 days of receipt of filters by the analytical laboratories. Laboratory analysis delays resulted in
later deliveries to AQS (see Section 2.1.1 and Section 2.3.1).
Table 5.1-1: Summary of data deliveries to AQS, January 1, 2017 through December 31, 2017.
Data (Month Samples Collected)
Filter Receipt Date
AQS Delivery Date
Days
January 2017
March 15, 2017
October 27, 2017
226
February 2017
April 3, 2017
November 28, 2017
239
March 2017
May 18, 2017
November 28, 2017
194
April 2017
June 29, 2017
January 3, 2018
188
May 2017
July 18, 2017
February 9, 2018
206
June 2017
August 22, 2017
February 9, 2018
171
July 2017
September 8, 2017
March 14, 2018
187
August 2017
October 26, 2017
April 12, 2018
168
September 2017
November 15, 2017
May 23, 2018
189
October 2017
December 14, 2017
June 13, 2018
181
November 2017
January 17, 2018
July 6, 2018
170
December 2017
February 7, 2018
August 2, 2018
176
Page|68
-------
6. Quality Assurance and Data Validation
6.1 QAPP Revisions
The UC Davis Quality Assurance Project Plan (QAPP) for Laboratory Analysis and Data
Processing/Validation for Chemical Speciation of PM2.5 Filter Samples was accepted by the EPA
on November 29, 2017. The QAPP is updated annually.
6.2 SOP Revisions
The UC Davis Standard Operating Procedures (SOPs) for Laboratory Analysis and Data
Processing/Validation for Chemical Speciation of PM2.5 Filter Samples were accepted by the
EPA on November 29, 2017. The SOPs are updated annually.
6.3 Summary of Internal QA Activities
Following laboratory analysis, all analytical results are assembled by UC Davis for processing
and initial validation. Data processing involves calculating ambient concentration, uncertainty,
and MDL for each analyte using the laboratory result plus the sample volume and sampling
duration determined from the field data. The calculated concentrations undergo two levels of
validation at UC Davis: (1) Level 0 validation to examine the fundamental information
associated with each measured variable, such as chain of custody, shipping integrity, sample
identification, and damaged samples, and (2) Level 1 review for technical acceptability and
reasonableness based on information such as routine QC sample results, data quality indicator
calculations, performance evaluation samples, internal and external audits, statistical screening,
internal consistency checks, and range checks. Further detail regarding the UC Davis data
processing and validation can be found in UCD CSN SOP #801: Processing and Validating Raw
Data, and in the associated Technical Information (TI) documents as follows:
1) UCD CSN TI 801A - Data Ingest. Sample event information (including filter IDs,
flow rates, flags, and comments) are received from the Sample Handling Lab via
email and uploaded to the UC Davis CSN database. XRF results are transferred into
the database through an automated service. IC and TOR analysis result files are
received via email from DRI. Results are ingested to the UC Davis CSN database.
2) UCD CSN TI 801C - Level 0 Validation: Data and metadata are reviewed through
several visualizations to identify oddities such as inconsistent dates that appear to be
data transcription and/or data entry errors. These are resolved through communication
with the Sample Handling Lab.
3) UCD CSN TI 80IB - Data Processing: Sample volume and analysis results are
combined to calculate concentrations. Blank values are used to derive MDLs. MDLs
and concentrations are used to estimate uncertainty.
4) UCD CSN TI 801C - Level 1 Data Validation: Several statistical and visual checks
are applied and examined. Reanalyses are requested as needed. Data are flagged with
qualifier or null codes.
5) UCD CSN TI 80ID - Data Posting: Initially validated concentration data and
metadata are posted to DART for SLT (State, Local, and Tribal) review. After the
Page|69
-------
specified 30-day review period, changed or unchanged data are re-ingested to the UC
Davis CSN database.
6) UCD CSN TI80ID - AQS Delivery : SLT initiated changes and comments are
reviewed and resolved. Data are formatted for delivery to AQS and posted.
6.4 Data Validation and Review
The validation graphics shown in this section are a small subset of the many QC evaluations that
UC Davis performs on a routine basis. They are selected to illustrate the nature and use of the
QC tools, and provide an overview of the review process.
Additional information and detail regarding analytical and validation procedures can be found in
the standard operation procedure (SOP) documents, UC Davis CSN Quality Assurance Project
Plan (QAPP), and the Data Validation for the Chemical Speciation Network guide, all available
at the UC Davis CSN site: https://aqrc.ucdavis.edu/csn-documentation.
6.4.1 Summary of Monthly Data Validation Review Results
6.4.1.1 Comparisons Across Years
Multi-year time series plots are used to examine large-scale trends and/or analytical problems.
Comparisons to historical network data provide context for validation and review of more recent
data.
Figures 6.4-1 and 6.4-2 show time series for the network-wide 90th percentile, median (50th
percentile), and 10th percentile concentrations of organic carbon by reflectance (OCR) and
elemental carbon by reflectance (ECR). These figures show raw data without blank correction to
enable comparison across a wider timeframe. The carbon fractions OCR and ECR are
determined by DRI using thermal analysis with a correction for pyrolysis based on optical
monitoring as it is heated. Measurements from 2005 through 2015 were made with DRI Model
2001 analyzers monitoring at the single wavelength 633 nm; starting with January 2016 samples,
DRI switched to Model 2015 analyzers monitoring seven wavelengths centered at 635 nm. OCR
concentrations at the median and 90th percentile were elevated during August and September, but
otherwise trend similarly to previous years. The ECR concentrations are lower at the median and
90th percentile during the first six months of 2017, but trend similarly to previous years during
the latter half of 2017.
Page|70
-------
Figure 6.4-1: Multi-year time series, organic carbon by reflectance (OCR).
90th Percentile
61
D)
=1
~c
o
O 3^
Jan Feb
Mar Apr
May
Jun
Jul
Aug
Sep Oct
Nov Dec
Year
Median
2.5-
O)
=L
2.0-
15-
O 10
o
1.0-
Jan Feb Mar Apr May Jun Jul Aug Sep Oct
Nov Dec
10th Percentile
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov Dec
Page|71
-------
Figure 6.4-2: Multi-year time series of network-wide elemental carbon by reflectance (ECR) concentrations.
90th Percentile
1.8-
O)
3-1-4"
01
u 1 o-
LLJ
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug Sep Oct Nov
Dec
Year
Median
0.8--
cO 0.7-
— 0.6-
O)
w 0.5-
Cd
O 0.4-
LLI
0.3-
Jan
Feb
Mar
Apr May
Jun
Jul
Aug Sep Oct Nov
Dec
10th Percentile
0.25-
E
0.20
cn
^L
0.15
OL
O
LU
0.10
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug Sep Oct Nov
Dec
Page|72
-------
During TOR analysis some of the OC pyrolyzes during the heating phase. The organic pyrolyzed
carbon (OPR) is combusted with the EC collected on the filter, and is accounted for by
monitoring the laser signal and identifying an OC/EC split point based on return of the last signal
to its initial value. To some extent, the split point - and thus the amount of OPR - is
operationally defined based on instalment parameter settings. As seen in Figure 6.4-3,
corresponding with the change in analyzers from DRI Model 2001 to DRI Model 2015 that
occurred on January 1, 2016, the OPR concentrations at the median and 90lh percentile
decreased.
Figure 6.4-3: Multi-year time series of network-wide organic pyrolyzed carbon by reflectance (OPR)
concentrations.
90th Percentile
Jan Feb Mar Apr May
Jun
Jul Aug Sep Oct
Nov Dec
Year
Median
0.4
E
0.3
co
=L
^—¦/
0.2
Dd
Q.
o
0.1
~ ~
^ *—
•
#
#—
.
#
•
~
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
10th Percentile
o>
=L
cr
a.
O
0.10
0.05-
0.00-
Jan
Feb
Mar
Apr May
Jun
Jul
Aug Sep
Oct
Nov
Dec
Page|73
-------
Similar to 2016, the 2017 sulfur concentrations generally continue to be low (Figure 6.4-4) with
reduced seasonal variability.
Figure 6.4-4: Multi-year time series of network-wide sulfur (S) concentrations.
90th Percentile
D)
CO
0.8-
0.6-
0.4-
0.2-_
0.25-
£ 0.20
cn
CO
0.15
0.10-
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Median
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
10th Percentile
Year
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Page|74
-------
The 2016 and 2017 nitrate concentrations show strong seasonality with elevated winter
concentrations; however, 2016 and 2017 concentrations are generally lower relative to previous
years (Figure 6.4-5).
Year
2010
2011
2012
• 2013
2014
2015
2016
• 2017
Year
2010
2011
2012
• 2013
• 2014
2015
• 2016
• 2017
Year
2010
2011
2012
• 2013
~ 2014
~ 2015
2016
• 2017
6.4.1.2 Comparisons Between Modules
The following graphs compare two independent measures of aerosol properties that are expected
to correlate. These graphs highlight cases where the two measurements do not correlate well,
which can result from real atmospheric and anthropogenic events or analytical and sampling
issues.
Sulfur versus Sulfate
PTFE filters are analyzed for elemental sulfur using EDXRF, and nylon filters are analyzed for
sulfate (SO4) using IC. The molecular weight of SO4 (96 g/mol) is three times the atomic weight
of S (32 g/mol), so the concentration ratio (3 S)/S0i should be one if all particulate sulfur is
present as water-soluble sulfate. In practice, real measurements routinely yield a ratio greater
than one (Figure 6.4-6), suggesting the presence of some sulfur in a non-water soluble form of
Page|75
-------
sulfate or in a chemical compound other than sulfate. However, instances are observed where
(3>
-------
concentrations (Figure 6.4-7). A known exception to this expectation is for soil-borne potassium,
which is not water soluble; high soil contributions are thus expected to result in ratios greater
than one.
Figure 6.4-7: Scatter plot of potassium versus potassium ion, January 1, 2017 through December 31, 2017. Number
of observations (complete pairs) is 12,753. Dotted black horizontal and vertical lines indicate MDLs. Solid gray line
indicates 1:1. Solid red line indicates regression.
3 -i
• IV.
y = 0.83x + 0.05
= 0.58
• Birmingham - North Birmingham
• Chamizal
• Deer Park (Collocated)
• Karnack
• Los Angeles - North Main Street
• Macon
• New York - Division Street
» San Jose - Jackson Street
• Washington DC - McMillan Reservoir
Potassium Ion (ng/m3)
PM2.5 versus Reconstructed Mass (RCM)
Gravimetric data are compared to RCM, where the RCM composite variable is estimated from
chemical speciation measurements, to test many different aspects of overall data quality. The
formulas used to estimate the mass contributions from various chemical species are detailed in
UCD CSN TI80IB - CSNData Processing. In the simple case where valid measurements are
available for all needed variables, reconstructed mass is the following sum:
RCM = (4.125 x S) + (1.29 x NO3" ) + (1.4 x OC) + (EC) +
(2.2 x A1 + 2.49 x Si + 1.63 x Ca + 2.42 x Fe + 1.94 x Ti) + (1.8 x chloride)
The parenthesized components represent the mass contributions from, in order, ammonium
sulfate, ammonium nitrate, organic compounds, elemental carbon, soil, and sea salt.
Gravimetric analysis is not routinely performed using CSN filters. Thus, for comparison
purposes 24-hour average gravimetric PM2.5 mass data from AirNow Tech is used as part of the
validation process in DART. The data provided by AirNow Tech is not final, so the data used
here is a snapshot, downloaded at the time the plots were generated.
Page|77
-------
If the RCM completely captures and accurately estimates the different mass components, the
RCM to AirNow Tech mass ratio is expected to be near one. The RCM and AirNow Tech mass
generally correlate (Figure 6.4-8), but RCM tends to underestimate AirNow Tech mass.
Figure 6.4-8: Scatter plot of RCM versus AirNow Tech PM2.5 mass data (Mass), January 1, 2017 through December
31, 2017. Number of observations (complete pairs) is 9,286. Solid gray line indicates 1:1. Solid red line indicates
regression.
Mass (ng/m3)
Page|78
-------
6.5 Uncertainty Estimates and Collocated Precision Summary Statistics
Several network sites are equipped with collocated samplers, where simultaneous samples are
collected on independent samplers and analyzed using the same analytical protocols. Differences
between the resulting data provide a measure of the total uncertainty associated with filter
substrates, sampling and handling in the field, and laboratory analysis.
Scaled relative difference between sample pairs collected at CSN collocated sites is calculated as
shown in Equation 6.5-1 and used to evaluate collocated precision (Figure 6.5.1, elements;
Figure 6.5-2, ions; Figure 6.5-3, carbon).
Scaled Relative Difference = ^ol|°cated—routine)/V2 6.5-1)
11 (collocated + routine) / 2 n
The scaled relative differences are ±V2 when one of the two measurements is zero, and vary
between these limits at concentrations close to the detection limit. They generally decrease with
increasing concentration, and are expected to converge to a distribution representative of
multiplicative measurement error when the concentration is well above the detection limit. Fe, K,
Si, and Zn are examples of elements that are measured at a wide range of concentrations and
display this behavior. S is measured well above the MDL and has good collocated measurement
agreement throughout the range. This convergence is not observed for many elements and carbon
fractions that are rarely measured above the MDL.
Page|79
-------
Figure 6.5-1: Scaled relative difference for element measurements at sites with collocated samplers across the
network (January 1, 2017 through December 31, 2017). Dotted vertical lines indicates MDL.
Ag
Cd
°
°M
°sj§
Cu
°Wo
Sb
o
°rW
V
c
°°$
jS0
o
(3DOO@ O
As
ag© <$$
&
OOO o
B-t)o
DO
o
Ba
°(S^
o Jp£>
m
oWP
In
° ^
Zr
°§J||
J|
0.01 1 100
0.01 1
100
0.01 1 100
0.01 1 100
Average Concentration/MDL
Br
o
o
Co
cPo
°°8§
w
%
K
• o
: o
: o
:6b o
Pb
°w
°w
Sn
°°8Jf^
i
1#:
0.01 1 100
Cr
IT
JL
Mg
°%$jL
° »°
°°w
Rb
o o
: Iff
Co
<|r
<-b
Sr
Ca
oo i
o0? o
Cs
:
Oq&t
m
Mn
o°ofe
SJIP
c
#
o'11
S
GD
O
COD
Ti
1
0.01 1
100
o Bakersfield - California Ave. (06-029-0014)
Site ° Deer Park (48-201-1039)
o G.T. Craig (39-035-0060)
o Riverside - Rubidoux (06-065-8001)
o Dudley Square - Roxbury (Boston) (25-025-0042) o Rutgers (34-023-0011)
Page|80
-------
Figure 6.5-2: Scaled relative difference for ion measurements at sites with collocated samplers across the network
(January 1, 2017 through December 31. 2017). Dotted vertical lines indicates MDL.
Ammonium
o
OfflSO o o
o
o °
o
oo
O
o 08 Qd0 ° o o
o CP Wo § °
° °
o
o
»
3 o °G52) °
0 o o °
8 a
o
o
o $
n°°
o
O 0
Potassium Ion
o :o
°S6 ^ '°
% :
°;
Chloride
: o
° a % 0
o
°*> o
o : u
Nitrate
@°
o.
o <9 °
o
QD
ft I
o
0 o
Sodium Ion
6 o
im^
B
o
o
G ¦ •(
o *
O
Sulfate
o
o
@ °
OH&
Ws
1
cP
0.01
100
0.01 1 100
Average Concentration/MDL
0.01
100
o Bakersfield - California Ave.
Site Name 0 Deer Park
o G.T. Craig
o Riverside - Rubidoux
o Dudley Square - Roxbury (Boston) o Rutgers
Page|81
-------
Figure 6.5-3: Scaled relative difference for carbon measurements at sites with collocated samplers across the
network (January 1, 2017 through December 31, 2017). Dotted vertical lines indicates MDL. Elemental carbon (EC)
fractions are indicated as (1) through (3), organic carbon (OC) fractions are indicated as (1) through (4). Organic
pyrolized (OP), elemental carbon (EC), and organic carbon (OC) are shown by reflectance (R) and transmittance
(T).
EC1
o °
^jf^'
Qdo°
0
EC2
oa§> o
• (£0
O
00
ECT
o o
o° J
o
OC4
Jjt
o
OPT
o
00©
o °
o
o
„
-------
There can be substantial variability in collocated agreement between different pairs of collocated
sites, even in cases where measurements are well above the MDL. To illustrate, potassium
(Figure 6.5-4), sulfate (Figure 6.5-5), and organic carbon by reflectance (Figure 6.5-6) are shown
with scaled relative difference plotted separately for each of the collocated site pairs. For the
nylon and PTFE filters, G.T. Craig, OH (AQS ID #39-035-0060) site shows cases of poor
collocated agreement, even at high concentrations. For the quartz filter, poor agreement is more
apparent at the Riverside, CA (AQS ID #06-065-8001) and Dudley Square, MA (AQS ID #25-
025-0042) sites.
Figure 6.5-4: Scaled relative difference potassium at sites with collocated samplers across the network (January 1,
2017 through December 31, 2017). Dotted vertical lines indicates MDL.
Bakersfield - California Ave.
o
o
oo
o
o
Deer Park
O 8ifer>ao
o
o
G.T. Craig
o J °
o
Dudley Square - Roxbury (Boston)
©
1fw
o
Riverside - Rubidoux
o
difflStov,
cb
Rutgers
: o
cj
(*) °
1
0.01
100
0.01 1 100
Average Concentration/MDL
0.01
100
Page|83
-------
Figure 6.5-5: Scaled relative difference for sulfate at sites with collocated samplers across the network (January 1,
2017 through December 31, 2017). Dotted vertical lines indicates MDL.
0
0
c
CT3
-------
Figure 6.5-6: Scaled relative difference for organic carbon by reflectance (OCR) at sites with collocated samplers
across the network (January 1, 2017 through December 31, 2017). Dotted vertical lines indicates MDL.
Bakersfield -
California Ave.
o
<
o
o
G.T. Craig
o
-ee-
10
Deer Park
o
o
° o
Rutgers
JOo°
10
100
Page|85
-------
Collocated precision is reported with CSN data delivered to AQS as fractional uncertainty.
Fractional uncertainty is calculated from scaled relative differences (Equation 6.5-1) between
sample pairs collected at CSN collocated sites, using the subset of observations with
concentrations at least three times the MDL. To limit uncertainty in determination of the
necessary percentiles, calculations are performed using multiple years of collocated data
(January 1, 2009 through December 31, 2014 for this reporting period) with a minimum of 60
collocated pairs per year. The calculation for fractional uncertainty is documented in UCD CSN
TI801B, and summarized in Equation 6.5-1, Equation 6.5-2, and Equation 6.5-3.
Collocated Precision (cp) = mch SRD>-^ "ercenau, of sro) (£q 6 __2)
Fractional Uncertainty = 100 x
n
i
i=1
(Eq. 6.5-3)
Tables 6.5-1 (elements), 6.5-2 (ions), and 6.5-3 (carbon) list fractional uncertainties calculated
for this reporting period. Since many species are routinely measured at or below the MDL, there
are numerous instances where insufficient pairs were available, in which cases a fractional
uncertainty of 0.25 is assigned. Historical data (2009-2014) are used to calculate fractional
uncertainties for this reporting period because insufficient data were available following the
contract transition (November 20, 2015). As more data becomes available, the fractional
uncertainty will be updated annually and calculated using collocated data from the previous two
years.
The network measurement quality objectives (MQOs) are based on the coefficient of variation
(CV) between collocated measurements, and are defined as CV of 10% for ions, 20% for
elements, and 15% for total carbon. As shown in Equation 6.5-4 and Equation 6.5-5, CV is
calculated from sample pairs collected at CSN collocated sites (Rice and Landis, 2016), using the
subset of observations with concentrations at least three times the MDL.
Relative Percent Difference (RPD) = ^l+^.y2 x 100 (Eq. 6.5-4)
CVt = (Eq. 6.5-5)
where X and Y, are the measurements from routine and collocated sites, respectively, for the ith
pair of measurements. Tables 6.5-1 (elements), 6.5-2 (ions), and 6.5-3 (carbon) list median CV
calculated from collocated samples collected during 2017.
Page|86
-------
Table 6.5-1: Fractional uncertainty (calculated from collocated samples collected 2009 through 2014) and median coefficient
of variation (CV; calculated from samples collected during 2017) for elemental species. Fractional uncertainty and CV values
not reported for species with less than 60 collocated pairs with concentrations at least three times the MDL.
Species
Fractional Uncertainty (%)
2009 - 2014
Pairs
Coefficient of Variation (%)
2017
Pairs
Na
16.4
1,270
—
32
Mg
24.5
365
—
4
A1
25.2
1,209
—
37
Si
15.2
3,897
8.0
151
P
17.3
93
—
3
S
6.2
5,530
3.6
320
CI
34.2
1,740
24.3
79
K
10.6
4,825
4.9
236
Ca
16.8
4,067
7.0
72
Ti
17.4
697
—
51
V
12.8
499
—
0
Cr
38.9
83
—
1
Mn
15.4
623
—
6
Fe
17
5,520
8.5
127
Co
—
10
—
0
Ni
17.8
400
—
0
Cu
26.9
2,313
—
3
Zn
12.3
3,144
8.0
120
As
18.8
155
—
0
Se
—
43
—
0
Br
15
1,610
—
1
Rb
—
0
—
0
Sr
—
58
—
0
Zr
—
3
—
0
Ag
—
1
—
0
Cd
—
0
—
0
In
—
0
—
0
Sn
—
0
—
0
Sb
—
0
—
0
Cs
—
7
—
0
Ba
16.5
123
—
0
Ce
—
21
—
0
Pb
18.5
381
—
0
Page|87
-------
Table 6.5-2: Fractional uncertainty (calculated from collocated samples collected 2009 through 2014) and median coefficient
of variation (CV; calculated from samples collected during 2017) for ions. Fractional uncertainty and CV values not reported
for species with less than 60 collocated pairs with concentrations at least three times the MDL.
Species
Fractional Uncertainty (%)
2009 - 2014
Pairs
Coefficient of Variation (%)
2017
Pairs
Ammonium
7.1
5,466
13.7
248
Chloride*
—
—
7.7
129
Nitrate
7.6
5,767
7.2
311
Potassium Ion
12.6
2,072
—
54
Sodium Ion
24.7
3,562
9.0
178
Sulfate
4.9
5,680
5.7
320
*Collocated chloride results were not available/reported until February 2017.
Table 6.5-3: Fractional uncertainty (calculated from collocated samples collected 2009 through 2014) and median coefficient
of variation (CV; calculated from samples collected during 2017) for carbon fractions. Fractional uncertainty and CV values
not reported for species with less than 60 collocated pairs with concentrations at least three times the MDL. Elemental carbon
(EC) fractions are indicated as (1) through (3), organic carbon (OC) fractions are indicated as (1) through (4). Organic
pyrolyzed (OP), elemental carbon (EC), and organic carbon (OC) are shown by reflectance (R) and transmittance (T).
Species
Fractional Uncertainty (%)
2009 - 2014
Pairs
Coefficient of Variation (%)
2017
Pairs
Elemental Carbon (EC1)
12.9
1,948
9.7
312
Elemental Carbon (EC2)
36.8
992
20.9
205
Elemental Carbon (EC3)
—
4
—
0
Elemental Carbon (ECR)
15.5
1955
10.5
311
Elemental Carbon (ECT)
12.8
1,606
9.9
306
Organic Carbon (OC1)
32.9
1,039
22.2
190
Organic Carbon (OC2)
13.6
1,877
7.7
309
Organic Carbon (OC3)
17.8
1,860
9.5
287
Organic Carbon (OC4)
15.7
1,487
12.3
310
Organic Carbon (OCR)
11.6
2,033
6.5
310
Organic Carbon (OCT)
7.3
1,774
6.1
310
Organic Pyrolyzed (OPR)
25.1
919
25.4
162
Organic Pyrolyzed (OPT)
17.3
1,557
15.9
280
7. References
Chen, L.-W.A.; Chow, J.C.; Wang, X.L.; Robles, J.A.; Sumlin, B.J.; Lowenthal, D.H.; Watson,
J.G. (2015). Multi-wavelength optical measurement to enhance thermal/optical analysis for
carbonaceous aerosol. Atmos. Meas. Tech., 8:451-461. http://www.atmos-meas-
tech. net/8/451/2015/amt-8-451-2015. html.
Chen, L.-W.A.; Chow, J.C.; Watson, J.G.; Schichtel, B.A. (2012). Consistency of long-term
elemental carbon trends from thermal and optical measurements in the IMPROVE network.
Atmos. Meas. Tech., 5:2329-2338. http://www.atmos-meas-tech.net/5/2329/2012/amt-5-2329-
2012.pdf.
Chow, J.C., Watson, J.G. (2017). "Enhanced ion chromatographic speciation of water-soluble
PM2.5 to improve aerosol source apportionment." Aerosol Science and Engineering 1:7-24.
Page|88
-------
Chow, J.C.; Wang, X.L.; Sumlin, B.J.; Gronstal, S.B.; Chen, L.-W.A.; Trimble, D.L.; Kohl,
S.D.; Mayorga, S.R.; Riggio, G.M.; Hurbain, P.R.; Johnson, M.; Zimmermann, R.; Watson, J.G.
(2015). Optical calibration and equivalence of a multiwavelength thermal/optical carbon
analyzer. AAQR, 15(4): 1145-1159. doi:10.4209/aaqr.2015.02.0106.
http://aaqr.org/ArticlesInPress/AAQR-15-02-OA-0106_proof.pdf.
Chow, J.C.; Watson, J.G.; Robles, J.; Wang, X.L.; Chen, L.-W.A.; Trimble, D.L.; Kohl, S.D.;
Tropp, R.J.; Fung, K.K. (2011). Quality assurance and quality control for thermal/optical
analysis of aerosol samples for organic and elemental carbon. Anal. Bioanal. Chem.,
401(10):3141-3152. DOI 10.1007/s00216-011-5103-3.
Chow, J.C.; Watson, J.G.; Chen, L.W.; Chang, M.C..; Robinson, N.F..; Dana Trimble; Steven
Kohl. (2007). The IMPROVEA Temperature Protocol for Thermal/Optical Carbon Analysis:
Maintaining Consistency with a Long-Term Database. J. Air Waste Manage. Assoc., 57:1014-
1023.
EPA 40 CFRPart 58. Available at http://origin.www.gpo.gov/fdsys/pkg/FR-2010-12-
27/pdf/2010-32153.pdf#page=l
Rice, J. and Landis, E. (2016): Chemical Speciation Network (CSN) Summary of Precision from
Six Collocated Sites [Poster presentation], 2016 National Ambient Air Monitoring Conference,
Portland, OR, United States, https://www.epa.gov/sites/production/files/2016-
09/documents/summary_of_preci si on. pdf.
Yatkin, S., Amin, H. S., Trzepla, K., Dillner, A.M. (2016a). Preparation of Lead (Pb) X-ray
Fluorescence Reference Materials for the EPA Pb Monitoring Program and the IMPROVE
Network Using an Aerosol Deposition Method. Aerosol Sci. Technol. 50:309-320.
Yatkin, S., Belis, C.A., Gerboles, M., Calzolai, G., Lucarelli, F., Fabrizia, C., Trzepla, K. (2016b). An
Interlab oratory Comparison Study on the Measurement of Elements in PMio. Atmos. Environ. 125: 61-
68.
Page|89
------- |