Supplement of Atmos. Meas. Tech., 8, 2555–2576, 2015 http://www.atmos-meas-tech.net/8/2555/2015/ doi:10.5194/amt-8-2555-2015-supplement © Author(s) 2015. CC Attribution 3.0 License.
Supplement of ACTRIS ACSM intercomparison – Part 2: Intercomparison of ME-2 organic source apportionment results from 15 individual, co-located aerosol mass spectrometers R. Fröhlich et al. Correspondence to: A. S. H. Prévôt (
[email protected])
The copyright of individual parts of the supplement might differ from the CC-BY 3.0 licence.
2 1
Roman Fr¨ohlich : Supplement Supplementary figures, tables and equations
z-score: z=
(xi,j − Xj ) σj
(Eq. S1)
with xi,j : contribution of source j for instrument i Xj : median contribution of source j σj : standard deviation
Roman Fr¨ohlich : Supplement
3
Tab. S 1. Participating ACSM instruments sorted by serial number. The order of instruments in this table purposely does not coincide with the order #1 - #13 used in the text to avoid an unintentional “rating” of individual instruments.
ACSM serial #
Operating institution(s)
Country
Monitoring location during ACTRIS
A140-104 A140-110 A140-113 A140-133 A140-134 A140-142 A140-143 A140-144 A140-145 A140-151 A140-152 A140-153 A140-156 ToF A003 HTOF AMS 215-081
University of Helsinki IDAEA-CSIC LSCE National University of Ireland, Galway DWD LSCE TROPOS NILU PSI EC-JRC & ENEA ISAC-CNR CIEMAT King’s College London PSI Ecole Nationale Sup´erieure des Mines de Douai
Finland Spain France Ireland Germany France Germany Norway Switzerland Italy & Italy Italy Spain United Kingdom Switzerland France
Hyyti¨al¨a, FIN Montseny, ESP Gif-Sur-Yvette (SIRTA), FRA Mace Head, IRL Hohenpeißenberg, DEU Cap Corse, FRA Melpitz, DEU Birkenes, NOR Cabauw, NLD Ispra, ITA Bologna, ITA Madrid, ESP North Kensington, GBR Jungfraujoch, CHE Revin, FRA (EMEP)
Tab. S 2. Calibrated values of the Q-ACSM response factor (ions / pg value for the ToF-ACSM).
ACSM serial #
Response factor (RF) in A/(µg/m3 )
A140-104 A140-110 A140-113 A140-133 A140-134 A140-142 A140-143 A140-144 A140-145 A140-151 A140-152 A140-153 A140-156 ToF A003
3.0 × 10−11 5.3 × 10−11 4.4 × 10−11 4.0 × 10−11 3.1 × 10−11 3.7 × 10−11 5.5 × 10−11 2.8 × 10−11 2.4 × 10−11 3.0 × 10−11 2.7 × 10−11 3.1 × 10−11 3.3 × 10−11 8.6 ions / pg
4
Roman Fr¨ohlich : Supplement
Tab. S 3. Modifications of calculated ME2 input matrices before analysis. m/z 12 was removed in all Q-ACSM instruments due to nonphysical negative signals.
Instrument
m/z downweighted by factor 3
ToF #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13 HR-AMS
none m/z 13 & 24 m/z 13 & 15 m/z 13 m/z 13, 15, 24 & 37 none m/z 15 m/z 13 m/z 13, 15, & 37 m/z 13 & 15 m/z 13 & 15 none none m/z 13 & 15 CH+ 2 (14.0156) & C3 H+ 3 (39.0235) by factor 3 CH3 O+ 3 (62.9905) by factor 6
additional info
m/z 13 removed (neg. S/N)
7 entries removed: CH+ 4 (16.0313), CH4 O+ (32.0262), CH2 O+ 2 (46.0055), CH4 NO+ (46.0293), C2 H6 O+ (46.0419), C2 HCl+ 2 (94.9455), C3 H3 O+ 5 (118.998)
Tab. S 4. Correlations (R2 ) between factors from HR PMF analysis with external measurements.
R2
BCwb
BCff
NOx
SO4
#1 (HOA) #2 (COA) #3 (OOA) #4 (BBOA)
0.39 0.64 0.20 0.90
0.68 0.40 0.15 0.27
0.65 0.29 0.06 0.22
0.02 0.15 0.50 0.08
Roman Fr¨ohlich : Supplement
4x10
5
-2
m/z 130 m/z 105
m/z 132 m/z 182
m/z 221
Signal (a.u.)
3
2
1
0
-1 12:00 17.11.2013
00:00 20.11.2013
00:00 25.11.2013
12:00 27.11.2013
00:00 30.11.2013
correction applied to ToF-ACSM data
-3
20x10
summed signal (a.u.)
12:00 22.11.2013
Reference value: IE/AB from calibration at the end of campaign
15
10
5
12:00 17.11.2013
00:00 20.11.2013
12:00 22.11.2013
00:00 25.11.2013
12:00 27.11.2013
00:00 30.11.2013
Fig. S 1. Top panel: signals of largest chamber background ions during the investigated period. All ions show a similar decrease, however concentrations of the chamber background ions are not expected to change over time (e.g. tungsten is released by the ioniser filament). Bottom panel: summed signal of all ions shown above. An exponential fit (black line) was calculated and used to correct the ambient data. The IE/AB value determined in a calibration after the campaign was used as reference.
6
Roman Fr¨ohlich : Supplement
fraction of total org (44 related peaks removed)
a)
0.14
IQR / median (percent)
0.12 10
0.10
20
30
40
50
0.08 0.06 0.04 0.02 0.00 15
fraction of total organic signal
b)
20
25
30
35
40
45
50
55 m/z
0.20
f29
f43
f55
f57
0.15
60
65
70
75
80
85
0.10
0.112
0.097
0.092
0.121
0.142
0.05
0.044 0.03
0.00
0.038
0.013
0.011
#1
HR
0.053 0.042
0.051
0.052
0.038
0.039
0.016
0.016
0.017
#2
#3
#4
0.093
0.089
0.087
0.079 0.058
0.102
0.099
0.096
0.125
0.121
0.112
0.108 0.096
100
0.156
0.153 0.143
0.133 0.116
95
f60 0.146
0.12
90
0.051
0.052
0.034
0.036
0.013
0.015
0.015
#5
#6
#7
0.052
0.048
0.048
0.05
0.047
0.033
0.031
0.029
0.016
0.015
0.014
0.014
#8
#9
0.038
0.031
0.114 0.092
0.087
#10
0.09
0.051
#11
0.095
0.049
0.034
0.035
0.015
0.017
#12
0.09
0.058 0.032 0.012
#13
TOF
Ratio f44#12(high f44) / f44#2 (low f44)
Ratio f44#12(high f44) / f44#2 (low f44)
Fig. S 2. (a) Median organic mass spectrum of the 13 Q-ACSMs. The signals at m/z = 44 and all related sticks (m/z = 16, 17 and 18) were removed and the mass spectrum was then renormalised. The boxes represent the interquartile range (IQR) (25-75 percentile) for each m/z stick and the whiskers represent the corresponding full range over all instruments. The colour bar represents the ratio of the width of the individual boxes in relation to the corresponding median in percent. (b) Fractions of the total organic signal at single m/z channels for all 15 participating instruments after all CO+ 2 related peaks were removed from the spectrum. The instruments were sorted like in Fig. 2a (by fraction of m/z 44 before removing). Grey: f29, blue: f43, purple: f55, pink: f57, red: f60. The respective fractions are given as numbers in the same colours.
3.0 2.5 2.0 1.5 1.0
orthogonal linear fit: slope: 0.007
0.5 0.0 5
10
15
20
25
30
35
TotalToF-ACSM [ug/m3] with CE = 0.5 3.0 2.0 1.0
orthogonal linear fit: slope: 0.072
0.0 0
1
2
3
AmmoniumToF-ACSM [ug/m3] with CE = 0.5
orthogonal linear fit: slope: 0.021
orthogonal linear fit: slope: 0.010 5
10
15
20
25
OrganicToF-ACSM [ug/m3] with CE = 0.5
0
2
4
6
8
10
12
14
NitrateToF-ACSM [ug/m3] with CE = 0.5
orthogonal linear fit: slope: 0.048 1
2
3
4
SulphateToF-ACSM [ug/m3] with CE = 0.5
Fig. S 3. Ratio of f44 between a Q-ACSM with high f44 (#12) and a Q-ACSM with low f44 (#2, cf. Fig. 2b) plotted against concentration (black: total, yellow: ammonium, green: organics, blue: nitrate, red: sulphate). Slopes of an orthogonal linear regression are given in each plot. Slopes are all positive but very small, indicating only a very minor dependency of f44 on concentration.
Roman Fr¨ohlich : Supplement
f44 average
b) Filters z1 z22 z49
400
500
600
400
500
600
40 20
ratio f44vsf43
0 4 3 2 1 0
300
300
400 500 600 vaporiser T (°C)
0.20 0.15 0.10 0.05 0.00
-3
f43 average
0.20 0.15 0.10 0.05 0.00 -3 x10 300 60
x10 60 40 20 0
ratio f44vsf43 avg
f43
f44
a)
7
4 3 2 1 0
z1
z22
z49
z1
z22
z49
z1
z22 filter name
z49
Fig. S 4. Fractions of m/z 43 and 44 in the organic spectra of three different filter samples (z1, z22 and z49). All three filters contain ambient aerosol particles sampled in Zurich, CH. Samples z1 and z22 were taken in winter and sample z49 was taken in summer. The filters were extracted into a liquid solution and subsequently nebulised to be able to measure them with an ACSM. For method description refer to Daellenbach et al. (2014). (a) f44, f43 and ratio of f44 vs f43 at different vaporiser temperature settings for the three different filters. Each spectrum was averaged for 3 × 10 min. (b) Average f44, f43 and ratio of f44 vs f43 over the measurements with vaporiser temperatures between 550 °C and 660 °C (possible error margin of the ACSM temperature reading). The error bars show the first standard deviation.
Tab. S 5. Correlations (R2 ) between factors from three factor PMF analysis with external measurements.
R2
HOA BCff
HOA NOx
BBOA BCwb
OOA SO4
ToF AMS (UMR) #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12 #13
0.41 0.56 0.67 0.58 0.44 0.63 0.53 0.53 0.43 0.61 0.49 0.62 0.64 0.52 0.53
0.49 0.60 0.62 0.60 0.44 0.66 0.60 0.52 0.46 0.54 0.48 0.70 0.61 0.54 0.53
0.88 0.86 0.93 0.93 0.91 0.93 0.86 0.87 0.87 0.88 0.85 0.91 0.84 0.91 0.86
0.51 0.56 0.72 0.52 0.71 0.54 0.76 0.73 0.72 0.72 0.75 0.75 0.74 0.78 0.59
Roman Fr¨ohlich : Supplement
3
3.0
NOx (R2 = 0.65) HOA factor
BCff (R2 = 0.68)
100
2.0
50
1.0 0.0
0 17.11.2013
4.0
25.11.2013
0
0.30 0.20
2.0 0.10
1.0 0.0
4
3
10 15 10 5
2 0 17.11.2013
10
NO3 (R2 = 0.47)
8
30
0
20 15 10
3
0
3
2
2
25
NO3 (µg/m )
4
25.11.2013
2 0
4
21.11.2013
4
0
6
6
17.11.2013
6
29.11.2013
3
8
SO4 (R2 = 0.43) OOA factor
25.11.2013
8
SO4 (µg/m )
OOA (µg/m )
d)
21.11.2013
5
methanol (ppb) acetonitrile x 15 (ppb)
6
10
29.11.2013
BCwb (R2 = 0.90) methanol (R2 = 0.76) acetonitrile (R2 = 0.48) BBOA factor
8
15
0
BCwb (µg/m3)
BBOA (µg/m )
10
25.11.2013
20x10
PTR-MS
c)
-3
0.00 21.11.2013
4
29.11.2013
C3H3O (55.0184) R2 = 0.80 (R2 with HOA = 0.38) C6H10O (98.0732) R2 = 0.36 (R2 with HOA = 0.23) COA-like factor
17.11.2013
8
C6H10O (µg/m3)
3.0
21.11.2013
C3H3O (µg/m3)
COA (µg/m3)
b)
12
150
BCff (µg/m3)
4.0
NOx (ppb)
a) HOA (µg/m )
8
5 0
29.11.2013
Fig. S 5. Time series correlations between factors from HR PMF analysis with external measurements during the whole two weeks of experiment. Corresponding R2 values are given in the plot. (a) HOA factor (grey) plotted together with NOx (black) measured with the NOx analyser and BCff (red) time trace from the BC source apportionment of the Aethalometer data. (b) COA-like factor (yellow) plotted together with time trace of C3 H3 O+ ion (pink) and C6 H10 O+ ion (grey) from the AMS HR analysis. Coefficients of determination (R2 ) with COA-like and HOA factors are given in the plot. (c) BBOA factor (brown) plotted together with time traces of gaseous methanol (violet) and acetonitrile (red), both measured by the PTR-MS, and BCwb (black) from the BC source apportionment of the Aethalometer data. (d) OOA factor (green) plotted together with inorganic nitrate (blue) and sulphate (red) measured by the HR-AMS.
Roman Fr¨ohlich : Supplement
Fraction of total COA-like factor
0.12
9
Reference COA spectrum from Crippa et al. 2013
Cx CH CHO1 CHOgt1
0.08
CHN CHOgt1N CS HO
0.04 0.00 0.12
20
40
60
80
100
COA spectrum from HR-ToF-AMS ME2 analysis
0.08 0.04 0.00 15
20
25
30
35
40
45
50
55 m/z
60
65
70
75
80
85
90
95
100
Fig. S 6. High-resolution COA factor spectrum from Crippa et al. (2013) (top) and from this study (bottom).
relative factor contribution
HOA: 19.3 ± 5.4
OOA: 50.3 ± 7.9
BBOA: 32.4 ± 6.3
1.0 0.8 0.6 0.4 0.2 0.0
#1
AMS
#2
#3
#4
#5
#6
#7
#8
#9
#10
#11
#12
#13
ToF
instrument (sorted by f44)
Fig. S 7. Relative contributions of HOA (grey), OOA (green) and BBOA (brown) factors from a three factor PMF solution without constraints for all 15 instruments (sorted by total f44). The shown AMS solution is from the analysis of the UMR AMS data set. Average contributions and the first standard deviation are given in the legend.
10
Roman Fr¨ohlich : Supplement
ME-2 1 factor constrained a₁
2 factors constrained a₁, a₂
a₁ = a ± �a a₁ ∈ [0,1] �a = 0.05
a₁ = a₂ = a ± �a a₁, a₂ ∈ [0,1] �a = 0.05
Maximise R² => afinal
Maximise R² => a₁ = a₂ = amax
a₂ = afinal,2 a₁ = amax ± �a a₁, a₂ ∈ [0,1] �a = 0.05 Maximise R² => a1 = afinal,1 => a2 = afinal,2
a₁ = amax a₂ = amax ± �a a₁, a₂ ∈ [0,1] �a = 0.05 Maximise R² => a₁ = amax => a₂ = afinal,2
Fig. S 8. Flowchart showing the procedure applied to find the ideal a value for ME-2 runs with one or two constrained anchor profiles.
factor conc. (ug/m3)
8
factor conc. (ug/m3)
Roman Fr¨ohlich : Supplement
8
6 4
13 x Q-ACSM ToF-ACSM HR-AMS
HOA
0 19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
01.12.2013
COA-like
6 4 2 0 17.11.2013
factor conc. (ug/m3)
all concentrations normalised to total HR-AMS organics
2
17.11.2013
19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
01.12.2013
20
BBOA
15 10 5 0 17.11.2013
factor conc. (ug/m3)
11
19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
8
01.12.2013
OOA
6 4 2 0 17.11.2013
19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
01.12.2013
local time
Fig. S 9. Full absolute time series of all four ME2 factors in µg/m3 (from top to bottom: HOA, COA-like, BBOA and OOA) normalised to the same total organic concentrations (reference: HR-ToF-AMS). Pink: HR-ToF-AMS, green: ToF-ACSM, grey: all 13 Q-ACSMs.
12
Roman Fr¨ohlich : Supplement
2
3
'1 to 1 line'
1.5 1.0 0.5 0.0
'1 to 1 line'
1.5 1.0 0.5
1.0
1.5 3
2.0
0.5
2.5
-0.01 + 1.01x
2
R = 0.90
3
2.0
'1 to 1 line'
1.5 1.0 0.5
1.0
1.5 3
2.0
2.5
y = 0.00 + 1.16x
2
1.0 0.5
0.5
2.5
1.0
1.5
3
1.0 0.5
3
1.0
1.5 3
'1 to 1 line'
1.0 0.5
2
0.5
2.5
y = -0.02 + 0.94x
2
3
HOA ACSM #11 (µg/m )
1.0 0.5 0.0
1.0
1.5 3
1.0
1.5 3
'1 to 1 line'
1.0 0.5
2.5 2
3
'1 to 1 line'
1.5 1.0 0.5 0.0
0.5
1.0
1.5
0.5
1.0
1.5 3
HOA Median (µg/m )
2.0
3
2.0
y = 0.05 + 1.01x
'1 to 1 line'
1.5 1.0 0.5
3
2.0
0.0
0.5
2.5
y = -0.11 + 1.16x
2
R = 0.84
2.0
1.0
1.5 3
2.0
HOA Median (µg/m )
'1 to 1 line'
1.5 1.0 0.5
y = 0.03 + 0.80x
R = 0.82
2.0
'1 to 1 line'
1.5 1.0 0.5 0.0
0.0 0.0
1.5
R = 0.68
2.0
HOA Median (µg/m )
y = -0.04 + 1.15x
1.0
0.0 0.0
HOA TOF-ACSM (µg/m )
2.0
0.5
2
3
2
R = 0.91
y = 0.03 + 1.05x
0.5
2.5
1.5
HOA Median (µg/m ) 2.5
2.0
HOA Median (µg/m )
y = -0.07 + 1.18x
R = 0.92
2.0
2.0
3
1.0
0.0
HOA HR-AMS (µg/m )
0.5
1.5
'1 to 1 line'
1.5
2.0
0.0 0.0
1.0
R = 0.94
2.0
HOA Median (µg/m )
'1 to 1 line'
1.5
0.5
0.0 0.0
2.0
R = 0.94
2.0
0.5
2
1.5
HOA Median (µg/m ) 2.5
1.0
2.5
R = 0.63
2.0
3
0.5
y = -0.02 + 0.93x
HOA Median (µg/m )
HOA ACSM #12 (µg/m )
0.0
2.0
'1 to 1 line'
0.0
0.0
0.0
3
1.5
2.0
y = 0.06 + 0.73x
2
'1 to 1 line'
1.5
1.5
R = 0.87
2.0
HOA Median (µg/m )
y = -0.05 + 1.27x
1.0
0.0 0.0
HOA ACSM #8 (µg/m )
3
HOA ACSM #7 (µg/m )
2
0.5
HOA Median (µg/m )
'1 to 1 line
1.5
2.0
R = 0.91
2.0
0.5
0.0
3
2.5
3
3
HOA ACSM #9 (µg/m )
0.5
1.0
2.0
R = 0.94
HOA Median (µg/m )
HOA ACSM #10 (µg/m )
1.5
0.0
0.0 0.0
'1 to 1 line'
1.5
HOA Median (µg/m )
HOA ACSM #5 (µg/m )
3
HOA ACSM #4 (µg/m )
2
1.0
3
2.5
y = -0.00 + 1.26x
R = 0.89
2.0
0.0 0.0
HOA ACSM #6 (µg/m )
0.5
HOA Median (µg/m )
3
2
0.0 0.0
HOA ACSM #13 (µg/m )
2.5
y = -0.04 + 1.20x
R = 0.93
2.0
3
R = 0.92
HOA ACSM #2 (µg/m )
3
HOA ACSM #1 (µg/m )
2.5
y = -0.02 + 0.85x
2
2.0
HOA ACSM #3 (µg/m )
2.5
0.0
0.5
1.0
1.5 3
HOA Median (µg/m )
2.0
0.0
0.5
1.0
1.5 3
2.0
HOA Median (µg/m )
Fig. S 10. Correlations of normalised HOA factor time series (reference: total organic mass of HR-ToF-AMS) of all instruments (name indicated in the y label) to the median of all instruments. Slopes and coefficients of correlation (R2 ) are given in the plots. The black dashed line indicates the 1 : 1 line.
Roman Fr¨ohlich : Supplement
1.0 0.5
'1 to 1 line'
1.5 1.0 0.5
1.0
1.5 3
0.0
2.0
0.5
R = 0.91 '1 to 1 line'
1.5 1.0 0.5
1.0
1.5 3
1.0 0.5
0.5
2.5 2
3
COA ACSM #8 (µg/m )
'1 to 1 line'
1.5 1.0 0.5
3
1.0
1.5 3
1.0 0.5
0.5
2.5 2
3
COA ACSM #11 (µg/m )
'1 to 1 line'
1.5 1.0 0.5 0.0
1.0
1.5 3
1.0
1.5 3
2
'1 to 1 line'
1.0 0.5
2.5 3
COA ToF-ACSM (µg/m )
1.0 0.5 0.0
0.5
1.0
1.5
1.0
1.5 3
COA Median (µg/m )
2.0
1.0
1.5 3
2.0
y = -0.08 + 1.38x (ordinary linear regression) y = -0.30 + 1.88x (orthogonal linear regression)
2.0
2
R = 0.67
1.5 1.0 0.5
3
0.0
2.0
0.5
2.5
y = 0.04 + 1.14x (ordinary linear regression) y = -0.11 + 1.47x (orthogonal linear regression)
2.0
2
2
R = 0.69
1.5
1.0
1.5 3
2.0
COA Median (µg/m )
1.0 0.5
y = -0.06 + 1.18x
R = 0.77
2.0
'1 to 1 line'
1.5 1.0 0.5 0.0
0.0 0.5
0.5
0.0 0.0
y = -0.00 + 0.72x
'1 to 1 line'
0.0
0.5
COA Median (µg/m )
R = 0.80
1.5
1.0
2.5
y = 0.03 + 0.62x
1.5
COA Median (µg/m ) 2.5
2.0
COA Median (µg/m )
R = 0.72
2.0
2.0
3
'1 to 1 line'
0.0
3
0.5
1.5
y = -0.05 + 1.07x
2
1.5
2.0
0.0 0.0
1.0
R = 0.88
2.0
COA Median (µg/m )
R = 0.89
2.0
0.5
0.0 0.0
y = -0.04 + 0.86x
2
0.5
2.5
y = -0.07 + 1.33x
1.5
COA Median (µg/m ) 2.5
1.0
COA Median (µg/m )
R = 0.63
2.0
2.0
2.0
'1 to 1 line'
0.0
3
0.5
3
y = -0.02 + 1.15x
2
1.5
2.0
COA ACSM #12 (µg/m )
0.0
3
1.5
0.0
0.0
COA ACSM #10 (µg/m )
1.0
COA HR-AMS (µg/m )
3
COA ACSM #7 (µg/m )
2.0
1.5
R = 0.91
2.0
COA Median (µg/m )
R = 0.85
1.0
0.0 0.0
2.0
y = 0.06 + 0.77x
2
0.5
2.5
3
2.5
3
0.0
COA ACSM #9 (µg/m )
0.5
0.5
COA Median (µg/m )
'1 to 1 line'
COA Median (µg/m )
COA ACSM #13 (µg/m )
3
0.0 0.0
1.0
2.0
y = 0.00 + 1.04x
2
1.5
0.0
2.0
1.5
R = 0.91
2.0
3
2.0
COA ACSM #5 (µg/m )
3
COA ACSM #4 (µg/m )
2.5
y = -0.04 + 1.43x
2
'1 to 1 line'
1.5
COA Median (µg/m )
COA Median (µg/m ) 2.5
1.0
3
0.5
COA ACSM #6 (µg/m )
0.0
R = 0.87
2.0
0.0
0.0
0.0
y = -0.22 + 1.39x
2
3
3
'1 to 1 line'
1.5
2.5
y = -0.02 + 1.35x
2
R = 0.89
2.0
COA ACSM #2 (µg/m )
3
COA ACSM #1 (µg/m )
2.5
y = 0.10 + 0.76x
2
R = 0.55
2.0
COA ACSM #3 (µg/m )
2.5
13
0.0
0.5
1.0
1.5 3
COA Median (µg/m )
2.0
0.0
0.5
1.0
1.5 3
2.0
COA Median (µg/m )
Fig. S 11. Correlations of normalised COA factor time series (reference: total organic mass of HR-ToF-AMS) of all instruments (name indicated in the y label) to the median of all instruments. Slopes and coefficients of correlation (R2 ) are given in the plots. The black dashed line indicates the 1 : 1 line.
14
Roman Fr¨ohlich : Supplement
6
3 2 1
'1 to 1 line'
4 3 2 1
0
0 2
3
4
3
5
0
1
2
BBOA Median (µg/m ) y = 0.06 + 1.01x
2
6
R = 0.94
3
5
'1 to 1 line'
4 3 2 1 0
2
3
4
3
5
3
6
3 2 1
3
2 1 0
2
3
4
5
3
4
3
5
'1 to 1 line'
3 2 1
6
1
2
3
4
3
5
3 2 1
y = -0.09 + 1.22x
2
5
2
3
4
3
5
'1 to 1 line'
4 3 2 1
1
BBOA ToF-ACSM (µg/m )
y = -0.07 + 0.93x
2
5
'1 to 1 line'
4
6
2
3
4
3
5
3 2 1
0
1
2
3
4
3
BBOA Median (µg/m )
5
6
6
5
6
5
6
3 2 1
1
2
3
4
3
y = -0.06 + 0.83x
2
R = 0.98
5
'1 to 1 line'
4 3 2 1
0
1
y = 0.14 + 1.26x
2
5
6
'1 to 1 line'
4
2
3
4
3
BBOA Median (µg/m )
3 2 1 0
0
5
'1 to 1 line'
4
6
R = 0.90
3
3
R = 0.97
3
y = -0.10 + 1.20x
2
BBOA Median (µg/m )
BBOA Median (µg/m ) 6
4
0 0
6
3
R = 0.93
5
6
3
1
2
BBOA Median (µg/m )
BBOA HR-AMS (µg/m )
0
1
0
0
0
6
1
6
R = 0.94
3
'1 to 1 line'
4
5
2
BBOA Median (µg/m )
BBOA ACSM #11 (µg/m )
3
5
6
0 0
y = -0.05 + 1.05x
2
5
3
6
y = -0.10 + 1.16x
2
4
6
R = 0.98
3
BBOA Median (µg/m )
R = 0.96
5
BBOA Median (µg/m ) 6
4
'1 to 1 line'
4
3
3
3
y = -0.03 + 1.07x
2
0
BBOA ACSM #12 (µg/m )
2
2
R = 0.95
5
6
0 1
1
0 1
6
3
0
1
BBOA Median (µg/m )
'1 to 1 line'
4
2
BBOA Median (µg/m )
'1 to 1 line'
4
0
y = 0.08 + 1.23x
2
3
0
y = 0.03 + 1.02x
2
6
R = 0.92
5
'1 to 1 line'
4
6
3
1
BBOA ACSM #8 (µg/m )
3
5
BBOA ACSM #9 (µg/m )
6
BBOA ACSM #7 (µg/m )
4
R = 0.92
5
BBOA Median (µg/m )
BBOA ACSM #10 (µg/m )
3
0 0
5
BBOA Median (µg/m )
BBOA ACSM #5 (µg/m )
3
BBOA ACSM #4 (µg/m )
6
y = -0.11 + 0.81x
2
R = 0.98
0
6
3
1
BBOA ACSM #6 (µg/m )
0
BBOA ACSM #13 (µg/m )
6 3
3
'1 to 1 line'
4
y = 0.01 + 0.77x
2
R = 0.93
5
BBOA ACSM #3 (µg/m )
y = 0.11 + 0.92x
2
R = 0.96
5
BBOA ACSM #2 (µg/m )
3
BBOA ACSM #1 (µg/m )
6
y = 0.13 + 0.99x
2
R = 0.91
5
'1 to 1 line'
4 3 2 1 0
0
1
2
3
4
3
BBOA Median (µg/m )
5
6
0
1
2
3
4
3
BBOA Median (µg/m )
Fig. S 12. Correlations of normalised BBOA factor time series (reference: total organic mass of HR-ToF-AMS) of all instruments (name indicated in the y label) to the median of all instruments. Slopes and coefficients of correlation (R2 ) are given in the plots. The black dashed line indicates the 1 : 1 line.
Roman Fr¨ohlich : Supplement
4 3
3
'1 to 1 line'
2 1
3
'1 to 1 line'
2 1
2
3
3
1
4
y = 0.04 + 0.81x 3
3
'1 to 1 line'
2 1
1
0
4
3
4
'1 to 1 line'
2 1
3 3
0
2
1
4
y = 0.04 + 0.82x 3
'1 to 1 line'
2 1 0
2
3
4
3
4
R = 0.95
3
3 3
'1 to 1 line'
2 1
2
2
0
1
4
y = -0.18 + 1.28x
2
2
3
'1 to 1 line'
2 1 0
3
4
y = -0.24 + 1.25x
3
'1 to 1 line'
2 1
3
3
4
1
OOA Median (µg/m ) 4 2
4 2
2
3
'1 to 1 line'
2 1 0
3
1
2
3
3
OOA Median (µg/m )
4
3
4
y = -0.13 + 1.14x
3
'1 to 1 line'
2 1
0
1
3
2
3
3
4
OOA Median (µg/m ) 4
y = 0.12 + 0.66x
2
'1 to 1 line'
2 1
y = 0.25 + 0.74x
R = 0.78
3
'1 to 1 line'
2 1 0
0 0
2
4
R = 0.91
3
OOA ToF-ACSM (µg/m )
3
R = 0.87
3
3
R = 0.94
OOA Median (µg/m )
y = 0.03 + 1.27x
2
0 0
3
2
1
OOA Median (µg/m )
OOA HR-AMS (µg/m )
1
y = 0.00 + 1.15x
1
0
0 0
4
'1 to 1 line'
4
R = 0.88
3
OOA ACSM #11 (µg/m )
3
3
3
2
OOA Median (µg/m )
R = 0.93
3
0
4
OOA Median (µg/m ) 4
2
R = 0.92
3
3
2
1
OOA Median (µg/m )
OOA ACSM #12 (µg/m )
1
y = 0.05 + 0.96x
'1 to 1 line'
0
0 0
4
1
y = -0.04 + 1.14x
2
OOA ACSM #8 (µg/m )
3
3
2
OOA Median (µg/m )
R = 0.91
3
0
4
OOA Median (µg/m ) 4
2
3
2
2
R = 0.92
3
OOA ACSM #9 (µg/m )
1
1
OOA Median (µg/m )
0 0
3
3
R = 0.95
3
0
OOA ACSM #7 (µg/m )
2
y = -0.11 + 1.07x
2
R = 0.85
OOA ACSM #5 (µg/m )
3
OOA ACSM #4 (µg/m )
2
'1 to 1 line'
2
OOA Median (µg/m )
OOA Median (µg/m ) 4
y = -0.10 + 1.23x
0 0
4
3
1
OOA ACSM #6 (µg/m )
0
OOA ACSM #10 (µg/m )
2
R = 0.93
3
0
0
OOA ACSM #13 (µg/m )
4
y = 0.08 + 0.99x
2
R = 0.72
3
y = -0.05 + 0.98x
OOA ACSM #3 (µg/m )
2
R = 0.78
OOA ACSM #2 (µg/m )
3
OOA ACSM #1 (µg/m )
4
15
0
1
2
3
3
OOA Median (µg/m )
4
0
1
2
3
3
4
OOA Median (µg/m )
Fig. S 13. Correlations of normalised OOA factor time series (reference: total organic mass of HR-ToF-AMS) of all instruments (name indicated in the y label) to the median of all instruments. Slopes and coefficients of correlation (R2 ) are given in the plots. The black dashed line indicates the 1 : 1 line.
16
Roman Fr¨ohlich : Supplement
fraction of total apportioned organics
0.12
sticks: ToF-ACSM boxes and whiskers: Q-ACSM
HOA (indv.)
0.08 0.04 0.00 20
fraction of total apportioned organics
0.12
30
40
50
60
70
80
90
100
COA-like (constrained)
0.08 0.04 0.00 20
fraction of total apportioned organics
0.30 0.25
30
40
50
60
70
80
90
100
30
40
50
60
70
80
90
100
40
50
60
70
80
90
100
BBOA (free)
0.20 0.15 0.10 0.05 0.00
fraction of total apportioned organics
20 0.30
OOA (free)
0.25 0.20 0.15 0.10 0.05 0.00 20
30
m/z
Fig. S 14. ME2 source spectra of the ToF-ACSM solution (sticks). The boxes and whiskers show the IQR and the full range of the Q-ACSM spectra, respectively. The line in the box indicates the median of all Q-ACSM. From top to bottom: HOA, COA-like, BBOA and OOA.
Tab. S 6. Average factor concentrations with first standard deviation in µg/m3 . This combines the uncertainties of the source apportionment with the uncertainties of the total organic mass discussed in part 1 of this study by Crenn et al. (2015).
factor
average concentration (µg/m3 )
first standard deviation (µg/m3 )
HOA COA-like OOA BBOA
1.01 1.06 2.90 2.11
0.25 0.35 0.51 0.73
17
COA-like
relative intensity
Roman Fr¨ohlich : Supplement
0.12 0.10
0.12 0.10
0.12
0.08
0.08
0.08
0.06 0.04
0.06 0.04
0.06
0.02 0.00
0.02 0.00
relative intensity
20 0.12
relative intensity relative intensity
100
0.02 0.00 20
40
60
80
100
0.12
0.12 0.10
0.08
0.10
0.08
0.08
0.08
0.06
0.06 0.04
0.06
0.06 0.04
0.02 0.00
0.02
0.04 0.02 0.00 40
60
80
100
0.04
40
60
80
100
0.10 0.08 0.06 0.04 0.02 40
60
80
100
20
40
60
80
100
20
40
60
80
100
20
40
60
80
100
40
60
80
100
20
40
60
80
100
20
40
60
80
100
0.12 0.10
0.12 0.10 0.08 0.06 0.04 0.02 0.00
0.00
20
0.02 0.00
0.00 20
0.12 0.10 0.08 0.06 0.04 0.02
0.12
0.12 0.10
0.08 0.06 0.04 0.02 0.00 20
40
60
80
100
0.12 0.10 0.08 0.06 0.04 0.02 0.00
0.08 0.06 0.04 0.02 0.00 20
40
60
80
100
relative intensity
HOA
0.12
0.12
0.12
0.08
0.08
0.08
0.04
0.04
0.04
0.00
0.00 20
relative intensity
80
0.12 0.10
20
40
60
80
0.00 20
100
40
60
80
100
0.12
0.12
0.12
0.12
0.08
0.08
0.08
0.08
0.04
0.04
0.04
0.04
0.00
0.00
0.00
20
relative intensity
60
0.04
0.10
20
40
60
80
100
20
40
60
80
100
40
60
80
100
0.12
0.12
0.12
0.08
0.08
0.08
0.08
0.04
0.04
0.04
0.04
0.00
0.00
0.00
40
60
80
100
0.12
0.12
0.08
0.08
0.04
0.04
20
40
60
80
100
20
40
m/z 60
80
100
20
40
60
80
100
20
40
60
80
100
20
40
m/z 60
80
100
0.00 20
0.12
20
relative intensity
40
0.10
0.00 20
40
m/z
60
80
100
0.00
0.00 20
40
m/z
60
80
100
Fig. S 15. Source profiles from ME-2 ”best solutions” of all 13 Q-ACSMs. All normalised to 1. COA-like (yellow) and HOA (grey). Organised from left to right and top to bottom according to the instrument numbers #1 - #13.
18
Roman Fr¨ohlich : Supplement
0.30 0.25
0.30 0.25
0.20
0.20
0.20
0.15
0.15
0.15
0.10
0.10
0.10
0.05
0.05
0.00
0.00
relative intensity
0.30 0.25
OOA
20
relative intensity
0.20 0.15 0.10 0.05 0.00 40
60
80
40
60
80
0.00 40
60
80
100
0.20 0.15 0.10 0.05 0.00 20
40
60
80
100
20
40
60
80
100
0.30 0.25 0.20 0.15 0.10 0.05 0.00
100
20
40
60
80
100
20
40
60
80
100
20
40
60
80
100
20
40
60
80
100
20
40
60
80
100
20
40
60
80
100
20
40
60
80
100
20
40
m/z 60
80
100
0.30 0.25 0.20 0.15 0.10 0.05 0.00
0.30 0.25 0.20 0.15 0.10 0.05 0.00 20
40
60
80
100
0.30 0.25 0.20 0.15 0.10 0.05 0.00 20
40
60
80
100
BBOA
0.25
0.25
0.25
0.20
0.20
0.20
0.15
0.15
0.15
0.10
0.10
0.10
0.05
0.05
0.00
0.00 20
relative intensity
0.05 20
100
0.30 0.25 0.20 0.15 0.10 0.05 0.00 20
40
60
80
0.05 0.00 20
100
40
60
80
100
0.25
0.25
0.25
0.25
0.20
0.20
0.20
0.20
0.15
0.15
0.15
0.15
0.10
0.10
0.10
0.10
0.05
0.05
0.05
0.00
0.00
0.00
20
relative intensity
80
0.30 0.25
100
relative intensity
relative intensity relative intensity
20
0.30 0.25 0.20 0.15 0.10 0.05 0.00
60
0.30 0.25 0.20 0.15 0.10 0.05 0.00
0.30 0.25
0.30 0.25 0.20 0.15 0.10 0.05 0.00
40
40
60
80
20
100
40
60
80
100
0.05 0.00 20
40
60
80
100
0.25
0.25
0.25
0.20
0.20
0.20
0.20
0.15
0.15
0.15
0.15
0.10
0.10
0.10
0.10
0.05
0.05
0.05
0.00
0.00
0.00
20
0.25
40
60
80
100
20
40
20
40
60
80
100
60
80
100
0.05 0.00 20
40
relative intensity
0.25
0.20
0.20
0.15
0.15
0.10
0.10
0.05
0.05
0.25
m/z
60
80
100
0.00
0.00 20
40
m/z
60
80
100
m/z
Fig. S 16. Source profiles from ME-2 ”best solutions” of all 13 Q-ACSMs. All normalised to 1. OOA (green) and BBOA (brown). Organised from left to right and top to bottom according to the instrument numbers #1 - #13.
4
HOA
3
COA-like
BBOA
OOA
z-score
2 1 0 -1 -2 -3 -4
HR
#1
#2
#3
#4
#5
#6
#7
#8
#9
#10
#11
#12
#13
TOF
Fig. S 17. z-score values for all factors and all instruments. z-score values were calculated with median contribution as reference and standard deviations determined with the robust algorithm A of ISO 13528 according to Eq. S1. |z| ≤ 1: “ok”, 1 < |z| ≤ 2: “acceptable”, 2 < |z| ≤ 3: “warning”, |z| > 3: “action” (see §7.4.2, ISO13528 (2005)).
Roman Fr¨ohlich : Supplement 2
Update of the Q-ACSM PMF export function in recent analysis software update (version 1.5.5.0) and its influence on the presented dataset
After the presented work was performed an update of the analysis software was released introducing changes of the PMF matrix calculation procedure. While in the former version the same relative ion transmission (RIT) curve (red curve in Fig. S18a) was applied to all datasets, the updated version introduces an individual correction (blue curve in Fig. S18a) for each instrument with the goal to reduce spectral differences between instruments. All data discussed in this manuscript, however, was calculated with version 1.5.3.2, i.e. with the same RIT applied. Fig. S18a shows both RITs of the instrument (#9) with the highest deviation between “old” and “new” calculation of all 13 Q-ACSMs. Note the error range for the calculated curve. In almost all cases (except #10 & #13) the “new” calculation leads to lower intensities of the m/z channels >55. The error matrix calculation is equally affected, however there the change is about a factor two lower. This means the S/N of the m/z >55 was slightly overestimated in the data presented in the manuscript compared to if the data was calculated with the “new” procedures. Figure S18b shows relative changes of PMF organic matrix and error matrix and Fig. S18c shows the absolute change of the spectrum for the most severely affected instrument #9. What are possible consequences on the presented work? On the one hand the presented uncertainties may be slightly lower if the “new” calculation was applied because input spectra are more similar, but on the other hand a lower S/N at higher m/z may lead to slightly larger residuals and hence to slightly larger deviations between solutions. A comparison of source apportionment results of one instrument with matrices calculated with both software versions showed no significant differences in neither factor contributions, factor time series nor factor profiles. All Q-ACSM source apportionment results published so far used the “old” calculation method.
19
1.0
a)
0.8 0.6 0.4 Average Calculted RIT Default Correction Calculated correction (fit to average)
0.2 0.0 0
50
100 m/z
150
0.2 b) 0.0 -0.2 Q-ACSM #9
-0.4
PMF org matrix PMF error matrix
-0.6
200
20
40
60 m/z
80
100
Q-ACSM #9 "old" V1532 "new" V1550
c)
0.4
intensity (a.u.)
org signal calculated with v1550 minus org signal calculated with v1532 in %
Roman Fr¨ohlich : Supplement
Relative Ion Transmission
20
0.3 0.2 0.1 0.0 0
10
20
30
40
50 m/z
60
70
80
90
100
Fig. S 18. (a) Relative ion transmission (RIT) curves of instrument #9 with the largest deviation between default and calculated correction (red and blue curve). Black dots and error bars were calculated from naphthalene fragments in the background spectra. Red curve: default correction used in analysis version 1.5.3.2, i.e. for all data presented in this manuscript. Blue curve: calculated correction from fit to the measured data. This correction is applied in version 1.5.5.0 of the analysis software. (b) Relative change of PMF organic matrix (red) and PMF error matrix (black) as function of m/z for the two software versions. Note that the relative change of the error is smaller than the relative change of the signal, leading to decreased S/N in this case. (c) Spectrum (one data point) of instrument #9 calculated with the two versions. Red was calculated with version 1.5.5.0, black with 1.5.3.2. Intensities from the newer software version are slightly lower.
Roman Fr¨ohlich : Supplement 3
HR-AMS PMF analysis diagnostics
A four factor solution (HOA,COA,BBOA,OOA) is presented as best solution in the manuscript. The applied HOA and COA anchor profiles were taken from a higher order PMF solution (8 factors) and constrained with a = 0 in the 4 factor constrained run of the ME-2 package. In the following more detail about how and why the 4 presented solution was chosen is presented. 3.1
4 factor unconstrained PMF
Mean profile and diurnal cycle of 10 seed runs with 4 unconstrained factors are shown in Fig. S19 including first standard deviations (error bars). While factor 1 and factor 2 are relatively stable over the 10 seeds factor 3 and factor 4 vary significantly. Both show similar profiles (with elevated f60) and diurnal trends indicating a split of a BBOA factor and possibly some mixing with factor 2 (OOA) or other not yet resolved factors. The overview about the factor contributions in the seed runs (Fig. S20) shows a similar picture: stable contributions of the HOA factor, some variation of the OOA factor and large variations of the rest. All solutions show only little difference in Q/Qexp values (top right of Fig. S20). Comparison with external data yields good correlations of factor 1 with NOx and BCff , factor 2 and sulphate and between BCwb and factors 3 and 4. However, in any case correlations are less good than the ones shown for the solution presented in the manuscript in Tab. 2.
21
-3
conc. / µg·m
0
1.0 0.8 0.6 0.4 0.2 0.0
1.0 0.8 0.6 0.4 0.2 0.0
1.2 0.8 0.4 0.0 0 2 4 6 8 10 12 14 16 18 20 22 24
0 2 4 6 8 10 12 14 16 18 20 22 24
0 2 4 6 8 10 12 14 16 18 20 22 24
14
16
18
20
22
24
2
4
6
8
10 CH3O2 CH5NO C4 CH4O2 C4H CH5O2 C3N C4H2 C3HN C4H3 C3O C3H2N C4H4 C3HO C3H3N C4H5 C3H2O C3H4N C4H6 C3H3O C3H5N C4H7 C2O2 C3H4O C3H6N C4H8 C2HO2 C3H5O C3H7N C4H9 C2H2O2 C2H4NO C3H6O C3H8N C4H10 COP C2H3S C2H3O2 C3H7O C3H9N C2HCl CSO C5 C2H4O2 C3H8O CHSO C5H C2H5O2 CH2SO C5H2 C2H6O2 CH3SO CH3O3 C5H3 C4O C4H2N C5H4 C4HO CH5O3 C5H5 C4H2O C4H4N C5H6 C4H3O C5H7 C3O2 C3H2NO C4H4O C4H6N C5H8 C3HO2 C3H3NO C4H5O C5H9 C3H2O2 C4H6O C4H8N C5H10 C3H3O2 C4H7O C4H9N C5H11 C6 C3H4O2 C4H8O C4H10N C5H12 C6H C3H5O2 C4H9O C4H11N C6H2 C3H6O2 C4H10O C6H3 C3H7O2 CS2 C5O C6H4 C3H8O2 C5HO C6H5 C3H9O2 CH2SO2 C5H2O C6H6 C5H3O C6H7 C4O2 C5H4O C6H8 C4HO2 C5H5O C6H9 C4H2O2 C5H6O C6H10 C4H3O2 C5H7O C6H11 C7 C4H4O2 C5H8O C6H12 C7H C4H5O2 C5H9O C6H13 C7H2 C4H6O2 C5H10O C6H14 C3H3O3 C7H3 C4H7O2 C5H11O C2SO2 C6O C7H4 C4H8O2 C5H12O C2HSO2 C6HO C7H5 C4H9O2 C2H2SO2 C6H2O C7H6 C4H10O2 C6H3O C7H7 C5O2 C6H4O C7H8 C5HO2 C6H5O C7H9 C2Cl2 C5H2O2 C6H6O C7H10 C5H3O2 C6H7O C7H11 C8 C5H4O2 C6H8O C7H12 C8H C5H5O2 C6H9O C7H13 C8H2 C5H6O2 C6H10O C7H14 C8H3 C5H7O2 C6H11O C7H15 C7O
C3 C3H C2N C3H2 C2HN C3H3 C2H2N C3H4 C2HO C2H3N C3H5 CNO C2H2O C2H4N C3H6 CHNO C2H3O C2H5N C3H7 CO2 CH2NO C2H4O C2H6N C3H8 CHO2 CH3NO C2H5O C2H7N
C2 C2H C2H2 CN CHN C2H3 CO CHO CH3N C2H5 CH2O CH4N C2H6 CH3O
x10
-3
80 60 40 20 0
CO2plus2
C CH CH2 CH3 O HO H2O
rel. contribution
22 Roman Fr¨ohlich : Supplement
0.15 0.10 0.05 0.00
0.10 0.08 0.06 0.04 0.02 0.00
0.12 0.08 0.04 0.00 variables
0.8 0.6 0.4 0.2 0.0 mean values error bar = 1std dev.
hours
12
Fig. S 19. Top panel: mean spectra of the 4 factor unconstrained PMF (10 seeds). Bottom panel: mean diurnal cycles of the 4 factor unconstrained PMF (10 seeds)
Roman Fr¨ohlich : Supplement
23
Q / Qexp value
HOA OOA slightly mixed mixed mixed
1.0 0.8 0.6 0.4 0.2
f-
10
8
9 4
4
f-
7
f-
4
4
f-
6
5
f-
4
4
f-
4
f-
2
f-
4
1
f-
4
f-
4
4
2.0
seed
Sulphate
f5 f4 6 f4 7 f4 8 f 4 -9 f10
4
4
4
3
f-
f-
1 f4
4
seed
seed
BCff NOx BCwb
f-
4 f 4 -9 f10
0.716
8
7 4
f-
6 4
f-
5 4
f-
4
f-
f-
4
4
2
3
f-
4
f-
4
4
f-
1
0.0
0.720
4
0.5
0.724
2
1.0
4
1.5
Q / Qexp value
conc. / µg·m
-3
2.5
3
0.0
3.0
F1
F2 0.67 0.60 0.39 0.03
F4
F3 0.00 0.00 0.01 0.47
0.25 0.21 0.88 0.07
0.33 0.23 0.69 0.16
Fig. S 20. Top left: factor contributions of 10 seeds, 4 factor unconstrained PMF. Top right: Q/Qexp for the 10 different seeds. Bottom panel: R2 coefficients of the four factors with external data.
24 3.2
Roman Fr¨ohlich : Supplement 5 factor unconstrained PMF
An increase of the factor number to five still shows a stable factor 1 (HOA) and a factor 2 (OOA) with only small variability in all 10 seeds shown in the top left panel of Fig. S21. Factor contributions of all other factors also seem to be more stable than in the 4 factor unconstrained PMF but a look at profiles (middle panel of Fig. S21) and diurnal cycles (top right panel of Fig. S21) still points to factor mixing. The profile of factor 3 looks like a mixture of typical BBOA (high f60) and COA (high f55) and also the diurnal cycle shows peaks at lunch time and at dinner time overlayed by a large afternoon increase (possibly due to residential heating by wood combustion). Profile and diurnal cycle of factors 4 and 5 point towards a splitted BBOA with some contribution of OOA according to a comparison to external data. Coefficients of correlation are shown in the bottom panel of Fig. S21. The values are comparable to the 4 factor solution with the exception of a lower R2 of the OOA with SO4 .
Roman Fr¨ohlich : Supplement
25
3.0 1.0
-3
mean conc./ µg·m
conc. / µg·m
-3
2.5 2.0 1.5 1.0
0.8
0.6
0.4
0.5
0.2
0.0
0.0
factor1 factor2 factor3 factor4 factor5
time / hours
5
5
f1 f5 2 f5 3 f5 4 f5 5 f5 6 f5 7 f5 8 f 5 -9 f10
0 1 2 3 4 5 6 7 8 9 101112131415161718192021222324
seed Q / Qexp value -3
Q / Qexp value
x10
-3
x10
Relative Intensity
-3
100x1080 60 40 20 0 0.20 0.15 0.10 0.05 0.00 80 60 40 20 0 60 40 20 0 0.25 0.20 0.15 0.10 0.05 0.00
15
20
25
30
35
40
45
50
55
60
65
70
75
1.0 0.8 0.6 0.4 0.2 0.0
0.7050 0.7040 0.7030 0.7020
80
1 2 3 4 5 6 7 8 9 0 f- f- f- f- f- f- f- f- f- -1 5 5 5 5 5 5 5 5 55f seed
variables
BCff BCwb NOx Sulphate
F1
F2 0.68 0.33 0.63 0.01
F3 0.15 0.32 0.06 0.34
F4 0.32 0.75 0.23 0.18
F5 0.25 0.88 0.22 0.05
0.25 0.71 0.16 0.27
Fig. S 21. Top left: factor contributions of 10 seeds, 5 factor unconstrained PMF. Top right: diurnal cycles of the 5 unconstrained factors. Middle left: profiles of an unconstrained 5 factor solution. Middle right: Q/Qexp for the 10 different seeds. Bottom panel: R2 coefficients of the five factors with external data.
26 3.3
Roman Fr¨ohlich : Supplement 8 factor unconstrained PMF
An investigation of unconstrained PMF solutions with higher numbers of factors resulted in a separation of a clean COA factor in addition to the HOA factor which already was resolved well in the 4 factor unconstrained solutions. The top right panel of Fig. S22 shows the profiles of such an 8 factor unconstrained PMF run. Here, factor 4 represents the COA factor which shows great similarities to previously published COA profiles (cf. Fig S6). In addition to the diurnal cycle of this COA factor (top right panel of Fig. S22), which shows clear lunch and dinner time peaks, an investigation of the residuals of an OA fragment typical for COA (C3 H3 O+ ) shows strong evidence for the validity of this factor. Residuals are discussed in Sect. 3.6 of the Supplementary material. The remaining factors in this 8 factor solution are nonphysical splits of BBOA (factors 5,6 & 7) and OOA (factors 2 & 8) or a mix of primary OA with OOA which only has very low contributions to total OA (factor 1). These mixed and split factors vary a lot when the model is run several times with varying seeds (see Fig. S23) while for HOA and COA only two types of solutions exist. Only in the solutions denoted with a B the COA is well separated from the rest. In the other, less frequently appearing type part of the COA is still mixed with BBOA (see profiles in fig. S23). Figure S24 shows the mean COA (bottom) and HOA (top) profiles extracted from the 8 seed runs belonging to A in Fig. S23. The first standard deviation is indicated by the vertical error bars. Note that these are not stacked HR spectra which can contain more than one important fragment at a given m/z.
Roman Fr¨ohlich : Supplement
-3
0 0.10 0.00 0.10
extracted HOA
0.00 0.10
-3
x10
0.6
0.4
0.2
0.00 0.20 0.10 0.00
0.0
40
50
60
70
0 1 2 3 4 5 6 7 8 9 101112131415161718192021222324
80
time / hours
variables
F5
0.18 0.12 0.73 0.08
seed
F8 0.26 0.22 0.88 0.06
0.08 0.02 0.06 0.41
7
8
f-
f-
8
8
8
5
4
3
1
6
f-
8
f-
8
f-
f-
8
8
f-
f-
2
0.6792 0.6788 0.6784 0.6780
f-
Q / Qexp value 9
10
f-
F7
0.22 0.65 0.71 0.18
8
f-
7
6
5
4
8 f-
8
8
f-
8
f-
f-
8
f-
8
2
3 f-
8
1
f-
f-
8
8
F6
0.39 0.28 0.53 0.12
1.0 0.8 0.6 0.4 0.2 0.0
8
Q / Qexp value
F4 0.67 0.59 0.25 0.01
8
F3 0.27 0.16 0.55 0.40
8
F2 0.28 0.18 0.49 0.05
8
F1 BCff NOx BCwb Sulphate
10
30
9
20
f-
0 0.10
mean conc./ µg·m
0.00 0.5 0.0 60
factor3 factor4 error bar = 1 std.dev of mean
0.8
extracted COA -3
Relative Intensity
50x10
27
seed
Fig. S 22. Top left: profiles of an unconstrained 8 factor solution. Top right: Diurnal cycles of the HOA and COA factor of the 8 factor unconstrained PMF. Middle panel: R2 coefficients of the eight factors with external data. Bottom: Q/Qexp for the 10 different seeds.
28
Roman Fr¨ohlich : Supplement
0.10
0
-3
x10
50
A
20
30
40
50
60
70
80
variables
HOA COA
3.0
conc. / µg·m
-3
2.5 2.0 1.5 1.0 0.5 0.0
0.10
B
0.00 20
30
40
50
60
70
80
variables
Fig. S 23. Two types of COA separation in the 8 factor unconstrained PMF. Type A shows an insufficient separation of COA and BBOA and is therefore not considered in the further analysis. Type B shows a good separation of COA. Variability of factor contributions of the 8 factor unconstrained PMF solution is shown in the center of the plot (10 seeds).
Roman Fr¨ohlich : Supplement
29
HOA_8F
relative signal intensity
0.10
(avg 8 seeds) 0.08 0.06 0.04 0.02 0.00 20
40
60 m/z (fractional)
100
COA_8F
-3
80x10
relative signal intensity
80
(avg 8 seeds)
60 40 20 0 20
40
60 m/z (fractional)
80
100
Fig. S 24. Average HR spectra of the extracted HOA and COA profiles from the 8 seeds belonging to type B (cf. Fig S23).
30 3.4
Roman Fr¨ohlich : Supplement 4 factor constrained PMF
The 4 factor solution with constrained HOA and COA anchors taken from the 8 factor solutions are presented in the main text as best solution. Figure S25 shows the variability over 10 seeds. Some ME-2 runs show lower BBOA concentrations (e.g. seed t8). For these runs the covariance of the BBOA and OOA time series is increased and worse correlations of the OOA factor with inorganic secondaries but also of the BBOA factor with BCwb are observed. See bottom panel of Fig. S25 for the R2 coefficients. Consequently, in the solutions with lower BBOA contributions, BBOA and OOA were considered not well separated.
31
1.0
3.0 HOA COA BBOA OOA
2.0 1.0 0.0
Q / Qexp value
conc. / µg·m
-3
Roman Fr¨ohlich : Supplement
0.740
0.8
0.738
0.6 0.4
0.736
0.2
0.734
0.0 t1 t2 t3 t4 t5 t6 t7 t8 t9 t10
t1
t2
t3
t4
t5
t6
t7
t8
t9
t10
t1
t2
t3
t4
t5
t6
t7
t8
t9
t10
R2 factor 3 (t8) factor 4 (t8) factor 3 (t3) factor 4 (t3) BCwb 0.66 0.69 0.9 0.19 sulphate 0.32 0.5 factor 3 (t8) 0.55 factor 3 (t3) 0.11
Fig. S 25. Top left: factor contributions of 10 seeds, 4 factor constrained PMF (ME-2). Top middle and right: Q/Qexp for the 10 different seeds. Bottom panel: R2 coefficients of factors 3 (BBOA) and 4 (OOA) with external tracers and with each other for seed “t3” and seed “t8”. A comparison of the coefficients shaded with the same colours shows better model performance for seeds of type “t3”.
32 3.5
Roman Fr¨ohlich : Supplement 5 factor constrained PMF
The 5 factor constrained solution (anchors: COA and HOA from the 8 factor solution) shows a splitting of the BBOA factor with one of the splits containing most of the signals around m/z 29 (diurnal cycles, profiles and seed variations are shown in Fig S26). Additionally the correlations of OOA with sulphate decreased significantly compared to the 4 factor constrained solution (R2 (SO4 –OOA) = 0.26 – 0.30 depending on the seed). Contributions and time traces of HOA and COA are very similar to the 4 factor constrained solution.
Roman Fr¨ohlich : Supplement
33
COA (a = 0.1)
HOA (a = 0.1) legend Cx CH CHO1 CHOgt1 CHN CHO1N CHOgt1N
OOA (free)
BBOA split I (free)
-3
BBOA split II (free)
x10
Relative Intensity
-3
80x10 60 40 20 0 0.10 0.08 0.06 0.04 0.02 0.00 0.20 0.15 0.10 0.05 0.00 0.20 0.15 0.10 0.05 0.00 50 40 30 20 10 0
15
20
25
30
35
40
45
50
55
60
65
70
75
80
variables 3.0
1.4
1.0 0.8 0.6
2.5 -3
mean conc./ µg·m
-3
COA HOA OOA BBOA split I BBOA split II
conc. / µg·m
1.2
2.0 1.5 1.0 0.5
0.4
0.0
0.2
5
0.0 0 1 2 3 4 5 6 7 8 9 101112131415161718192021222324
time / hours
f-
1 5
f-
2 5
f-
3 5
f-
4 5
f-
5 5
f-
6 5
f-
7 5
f-
8 5
9 0 f- f-1 5
seed
Fig. S 26. Bottom right: factor contributions of 10 seeds, 5 factor constrained PMF (ME-2). Bottom left: diurnal cycles of the 5 constrained factors. Top panel: profile of a 5 factor constrained PMF (ME-2) solution. Profile names and a values are given in the plot.
34 3.6
Roman Fr¨ohlich : Supplement Residuals
The total residuals over time are shown in Fig. S27 for the 4, 5 and 8 factor unconstrained solutions and for the 4 factor constrained solution. Only little remaining variability can be observed. Also the increase to 8 factors did not significantly change the overall residual structure which indicates that already 4 factors do a “good job” explaining the variability of the dataset. A more detailed look into the residuals of specific frag+ ments (C3 H3 O+ , C4 H+ 9 and C4 H7 ) is presented in Fig. S28. + The C3 H3 O fragment is typical for cooking related OA emissions while the C4 H+ 9 fragment usually is more pronounced in HOA profiles. C4 H+ 7 is found in both profiles to larger amounts. The median diurnal cycles of the 4 factor unconstrained (red), the 8 factor unconstrained (green) and the 4 factor constrained solution using the HOA and COA profiles found in the 8F solution are shown. C3 H3 O+ clearly shows increased residuals during mealtimes (11 a.m. to 2 p.m. and 7 – 8 p.m.) while C4 H+ 9 shows increased residuals during the traffic rush hour times (7 a.m. and 5 p.m.). C4 H+ 7 does not show any clear peaks. An increase to 8 factors decreases the median residual of C3 H3 O+ significantly during mealtimes (e.g. by 59 % at 1 p.m.) and the median residuals of C4 H+ 9 during rush hours (e.g. by 52 % at 7 a.m.) while C4 H+ , 7 which is important for both COA and HOA does not show large differences. By constraining the extracted profiles, the residuals increase again a bit but still are significantly lower than in the unconstrained case with the same number of factors (difference C3 H3 O+ : 32 %, difference C4 H+ 9 : 37 %). This behaviour of the residuals strongly points towards a previously mixed factor which explained well the variation of the shared signals (e.g. C4 H+ 7 ) but not taking into account the differences at the more typical fragments of COA and HOA (C3 H3 O+ and C4 H+ 9 ). Consequently, extracting and constraining COA and HOA helped to better separate these OA sources.
Roman Fr¨ohlich : Supplement
2
35
4 factor unconstrained
res.
1
0
-1
n1 n5 n9
n2 n6 n10
n3 n7
n4 n8
-2 17.11.2013 2
19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
5 factor unconstrained
res.
1
0
-1
-2 17.11.2013
19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
2
8 factor unconstrained
res.
1
0
-1 n2 n6
n3 n8
n4 n9
n5 n10
-2 17.11.2013
19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
date
2
4 factor constrained
0
res.
1
-1 t1 t6
t2 t7
t3 t8
t4 t9
t5 t10
-2 17.11.2013
19.11.2013
21.11.2013
23.11.2013
25.11.2013
27.11.2013
29.11.2013
date
Fig. S 27. Total residual time series during the whole period for (from top to bottom): 4 factor unconstrained, 5 factor unconstrained, 8 factor unconstrained, 4 factor constrained PMF. Different colours in a plot mean different seeds.
36
Roman Fr¨ohlich : Supplement
4F free PMF (median 10 runs) 8F free PMF (source of COA profile) 4F incl. COA & HOA -3
32%
5x10
C3H3O+
residual
4
(55.0184)
3 2 1
59%
0 -1 0
5
-3
5x10
C4H9+
4
10 37%
hour
15
20
52%
(57.0704)
residual
3 2 1 0 -1 0
5
10
5
10
hour
15
20
15
20
C4H7+
(55.0548)
-3
4x10
residual
2 0 -2 0
hour
+ Fig. S 28. Diurnal cycle of residuals of specific fragments (from top to bottom): C3 H3 O+ , C4 H+ 9 , C4 H7 . The different traces show different PMF runs: 4 factor unconstrained PMF (red), 8 factor unconstrained PMF (green), 4 factor constrained PMF/ME-2 (black).
Roman Fr¨ohlich : Supplement 4
ACSM PMF analysis diagnostics
The variability of 10 seed runs for each ACSM is shown in Fig. S29. The appearance of two groups of solutions: type I (similar to “t8”) and type II (similar to “t3”) which was already discussed in Sect. 3.4 of the Supplementary also was observed in all ACSM except # 1. The frequency of type II which exhibits the better correlations with external data seems to decrease with instrument number and hence total f44. The solutions of type II are shown in full colours and type I solutions in light colours. Within the type II which were considered in the final results discussed in the manuscript only little variation is observed within an instrument.
37
4
f-
f-
f-
f-
f-
f-
8
7
6
5
4
3
2
1
f-
f-
f-
f-
3
7
6
5
4
TOF
3
2
1
0
8
9 10
f-
4
seed
f-
4
f-
6 7
f-
f-
f-
f-
8
7
6
5
4
3
f- 9 10
4
4
4
4
f-
f-
f-
f-
f-
f-
f-
f-
f-
f-
f-
2
1
0
4
4
4
4
4
4
4
f-
f-
f-
f-
f-
f-
f-
f-
8
7
6
5
4
3
2
1
f 4 -9 f10
4
3
8
4
4
6
7
6
5
4
3
2
1
f 4 -9 f10
4
4
4
4
4
4
4
4
seed
4
seed
4
f-
5
seed
f-
4
4
seed
4
4
4
0
f-
2
4
4
3
0
f-
1
0
4
1
4
5
f-
2
4
0
2
3
1
4
f-
2
f-
5
4
6
2
4
1
0
f-
4 -3
-3
conc. / µg·m
2
f-
#6 7
conc. / µg·m
f-
4
4
5 -3
6
#9 6
conc. / µg·m
seed
1 f2 f4 3 f4 4 f4 5 f4 6 f4 7 f4 8 f 4 -9 f10 4
4
conc. / µg·m
-3
6
4
5
-3
f-
2
1
3 f4 f4 5 f4 6 f4 7 f4 8 f 4 -9 f10 4
4
4
f-
f-
4
4
4
-3
#3
4
#12
6
conc. / µg·m
f-
3 f4 f4 5 f4 6 f4 7 f4 8 f 4 -9 f10 4
4
4
4
4
3
2
1
conc. / µg·m
#2 8
4
f4 8 f 4 -9 f10
seed
4
4
4
4
4
seed
4
4
5 f6 f4 7 f4 8 f 4 -9 f10
f-
3
4
0
f-
2
4
6
f-
0
4
1
2
seed
4
#13 1
0
f-
2
f-
#8
4
3
f-
0
4
1
2
1
1
2
f-
4
f-
2
4
#5 8
-3
#4 5
conc. / µg·m
3
4
f-
f-
f-
f-
f-
0
2
-3
seed
5 f6 f4 7 f4 8 f 4 -9 f10
4
4
4
4
4 1
1
3
conc. / µg·m
4
3
2
1
-3
2
4
4
f-
f-
f-
f-
f5 f6 f4 7 f4 8 f 4 -9 f10 4
4
4
4
4
4
4
conc. / µg·m 3
f-
7
6
5
4
3
2
1
-3
4
f-
#11
4 -3
#105
conc. / µg·m
f-
f-
f-
f-
f-
f-
f-
f-
8 f 4 -9 f10
4
4
4
4
4
4
4
4
conc. / µg·m 5
4
8
7
6
5
4
3
2
1
-3
6
4
-3
f-
f-
f-
f-
f-
f-
f-
f-
conc. / µg·m
#1 7
4
4
conc. / µg·m
4
4
4
4
4
4
4
4
4
f 4 -9 f10
seed
f 4 -9 f10
seed
4
4
4
4
4
4
f-
f-
-3
#7
4
4
conc. / µg·m
38 Roman Fr¨ohlich : Supplement
7
6
5
4
3
2
1
0
seed
4
3
2
6
4
3
2
1
0
dark colours: type II -> considered in the final result light colours: type I -> not considered in the final result cf. Fig S23 for types I + II
seed
Fig. S 29. Factor contributions of 10 seeds for ACSM # 1 to # 13 and for the ToF-ACSM. Solutions are separated into type I and type II, cf. Sects 3.4 and 4 of the Supplementary.
Roman Fr¨ohlich : Supplement References Crenn, V., Sciare, J., Croteau, P. L., Favez, O., Verlhac, S., Belis, ¨ al¨a, M., Alastuey, A., Arti˜nano, C. A., Fr¨ohlich, R., Aas, W., Aij¨ B., Baisn´ee, D., Baltensperger, U., Bonnaire, N., Bressi, M., Canagaratna, M., Canonaco, F., Carbone, C., Cavalli, F., Coz, E., Cubison, M. J., Gietl, J. K., Green, D. C., Heikkinen, L., Lunder, C., Minguill´on, M. C., Moˇcnik, G., O’Dowd, C. D., Ovadnevaite, J., Petit, J.-E., Petralia, E., Poulain, L., Pr´evˆot, A. S. H., Priestman, M., Riffault, V., Ripoll, A., Sarda-Est`eve, R., Slowik, J., Setyan, A., and Jayne, J. T.: ACTRIS ACSM Intercomparison: part I - Intercomparison of concentration and fragment results from 13 individual co-located aerosol chemical speciation monitors (ACSM), submitted to Atmos. Meas. Tech. Disc., 2015. Crippa, M., DeCarlo, P. F., Slowik, J. G., Mohr, C., Heringa, M. F., Chirico, R., Poulain, L., Freutel, F., Sciare, J., Cozic, J., Di Marco, C. F., Elsasser, M., Nicolas, J. B., Marchand, N., Abidi, E., Wiedensohler, A., Drewnick, F., Schneider, J., Borrmann, S., Nemitz, E., Zimmermann, R., Jaffrezo, J.-L., Pr´evˆot, A. S. H., and Baltensperger, U.: Wintertime aerosol chemical composition and source apportionment of the organic fraction in the metropolitan area of Paris, Atmos. Chem. Phys., 13, 961– 981, 2013. Daellenbach, K. R., Bozzetti, C., Kˇrepelov´a, A., Canonaco, F., Wolf, R., Huang, R.-J., Zotter, P., Crippa, M., Slowik, J. G., Zhang, Y., Szidat, S., Baltensperger, U., Pr´evˆot, A. S. H., and El Haddad, I.: Characterization and source apportionment of organic aerosol using offline aerosol mass spectrometry, in prep. Atmos. Meas. Tech., 2014. ISO13528: Statistical methods for use in proficiency testing by interlaboratory comparisons, ISO 13528, International Organization for Standardization, Geneva, Switzerland, 2005.
39