STN Carbon Field Blank Analysis, Derived Organic Carbon Analysis and

advertisement
STN Carbon Field Blank Analysis, Derived
Organic Carbon Analysis
and
IMPROVE blank corrected artifact analysis
Bret Schichtel
Incomplete Observations from Analysis
STN field blank carbon concentrations
• Met One, Anderson and R & P 2300 have similar field
blank concentrations
• Urg and R & P 2025 samplers have carbon
concentrations a factor of 2 to 4 smaller than the other
samplers.
• There is little seasonality in the field blank carbon
concentrations.
• Appears to be some spatial variability with the Eastern
U.S. having higher blank values than most of the west.
However, there are plenty of exceptions making any
definitive statement on the spatial variability difficult to
judge.
Incomplete Observations from Analysis
Derived Carbon Artifacts from STN Data
• There is a strong seasonality in the derived artifact with
summer values 50 – 100% higher than winter. Spring and
Fall are between the winter and summer values.
• The winter artifacts are similar to the field blank carbon
concentrations
• Like the field blanks, Met One, Anderson and R & P 2300
have similar carbon artifacts with smaller values for the Urg
samplers
• There is a strong spatial variability with lower artifacts in the
Eastern U.S. (opposite the field blanks) during the winter,
spring and fall. During the summer the lowest artifacts are
generally at the northern sites.
• The highest artifacts occur in southern California
• The site by site difference between the field blanks and
derived carbon artifact are compared in Slide 23. As shown
the field blanks are typically larger than the derived artifact in
the eastern US and smaller in the west and midwest.
Incomplete Observations from Analysis
Derived Carbon Artifacts from IMPROVE Data (OC vs PM)
• On average the derived carbon artifacts are small, but
positive, 0.1 to 0.4 micro-grams/m3
– The IMPROVE artifact correction appears to work!
• Weak seasonality in the carbon artifacts
• Weak spatial variability during winter, spring and fall with
somewhat higher values in the eastern U.S. than west
• Strong spatial variability during the summer with
concentrations of 0.5 to 1 micro-g/m3 in the eastern U.S.
and Southwest but ~0 micro-g/m3 in the northwest and
north great plains.
STN Carbon Artifact Correction
• The analyses clearly shows a large positive organic carbon artifact
on the STN filters.
• On average, the STN field blanks and derived carbon artifacts are
consistent during the winter, spring and fall. However, the summer
field blanks are 50%-100% smaller during the summer. This is
consistent with the IMPROVE data in the eastern and southwestern
U.S.
• It is not apparent if the carbon artifact is actually larger than the field
blank during the summer or if this is an “artifact” of the analysis.
• The field blanks and derived artifacts have different seasonality. It is
not clear what is driving these differences. Does the “true” carbon
artifact have a different seasonal pattern than the field blanks or is
the seasonal variation an “artifact” of the analysis?
• Base on this STN will be carbon artifact corrected similar to
IMPROVE, i.e. estimate a seasonal artifact from the field blanks
and apply to all sites.
Seasonal median of STN Carbon Field Blanks
Met One Sampler, trip blnks used for missing field blnks
Winter
Summer
Spring
Fall
Seasonal median of STN Carbon Field Blanks
Anderson Sampler, trip blnks used for missing field blnks
Winter
Summer
Spring
Fall
Seasonal median of STN Carbon Field Blanks
Urg Sampler, trip blnks used for missing field blnks
Winter
Summer
Spring
Fall
Seasonal median of STN Carbon Field Blanks
R&P 2025 Sampler, trip blnks used for missing field blnks
Winter
Summer
Spring
Fall
Seasonal median of STN Carbon Field Blanks
R&P 2300 Sampler, trip blnks used for missing field blnks
Winter
Summer
Spring
Fall
STN Carbon Blank Stats
Mean Median St Dev # Sites
0.97
0.96
0.17
160
0.96
0.95
0.15
176
1.09
1.06
0.18
160
1.04
1.03
0.15
184
Met One
Met One
Met One
Met One
Win
Spr
Sum
Fall
Anderson
Anderson
Anderson
Anderson
Win
Spr
Sum
Fall
0.83
0.95
0.93
0.92
0.85
0.95
0.89
0.91
0.12
0.18
0.15
0.14
16
23
20
26
Urg
Urg
Urg
Urg
Win
Spr
Sum
Fall
0.24
0.28
0.24
0.27
0.24
0.26
0.23
0.29
0.05
0.06
0.03
0.03
7
7
7
7
Mean
R&P 2025 Win
0.51
R&P 2025 Spr
0.60
R&P 2025 Sum
0.73
R&P 2025 Fall
0.62
R&P 2300
R&P 2300
R&P 2300
R&P 2300
Win
Spr
Sum
Fall
0.91
1.06
1.17
1.01
Median St Dev # Sites
0.49
0.09
21
0.61
0.10
20
0.75
0.14
21
0.62
0.12
22
0.87
1.09
1.14
0.97
0.18
0.15
0.16
0.17
16
15
13
17
Comparison of OC vs PM
• The OC artifact is approximated by regressing the measured OC
against measured PM or PM minus aerosol components
• The Theil regression is used with the slope fitted thought the
medians to find the intercept. To reduce the variability, the data
were first sorted by the PM concentrations and placed into 33 to
100 bins. The data in each bin were averaged together and
regressed against each other.
• The analysis is conducted for each season, for the different
samplers and for PM, PM minus Amm. Sulfate, Fine Soil; and PM
minus Amm. Sulfate, Fine Soil, Amm. Nitrate and EC.
• The analysis was conducted by first aggregating the data from all
sites together, then conducting the regression on each individual
site and aggregating the results together.
• The following slides presents the scatter plots and statistics for
the aggregation of data across all sites.
• Note only the scatter plots for the OC vs PM minus Amm. Sulfate,
Fine Soil
Seasonal OC vs PM2.5: Met One Sampler, All Sites
Data sorted on PM, placed into 100 bins and averaged
White line – Theil Regression
Megenta – Linear Regression
Theil
Theil
OC vs PM2.5
Intercept
Slope
Win
0.65
0.27
Spr
1.51
0.18
Sum
2.71
0.14
Fall
1.61
0.23
OC vs PM2.5-AS-Soil
Win
1.02
0.34
Spr
1.38
0.33
Sum
1.95
0.33
Fall
1.35
0.37
OC vs PM2.5-AS-Soil-AN-EC
Win
1.01
0.56
Spr
1.70
0.43
Sum
2.33
0.34
Fall
1.65
0.48
r
0.99
0.99
0.99
0.99
# bins
99
99
99
99
1.00
0.98
0.99
0.99
99
99
99
99
0.99
0.98
0.98
0.98
99
99
99
99
Seasonal OC vs PM2.5: Anderson Sampler, All Sites
Data sorted on PM, placed into 33 bins and averaged
OC vs PM2.5
Intercept
Slope
Win
0.83
0.29
Spr
1.26
0.22
Sum
2.31
0.17
Fall
1.37
0.26
OC vs PM2.5-AS-Soil
Win
1.07
0.38
Spr
1.06
0.41
Sum
1.56
0.41
Fall
1.17
0.43
OC vs PM2.5-AS-Soil-AN-EC
Win
1.37
0.58
Spr
1.54
0.49
Sum
2.13
0.43
Fall
1.47
0.52
r
0.99
0.99
0.98
0.99
# bins
32
32
32
32
1.00
0.99
0.99
0.99
32
32
32
32
0.99
0.97
0.99
0.98
32
32
32
32
Seasonal OC vs PM2.5: Urg Sampler, All Sites
Data sorted on PM, placed into 33 bins and averaged
OC vs PM2.5
Intercept
Slope
Win
0.63
0.45
Spr
0.74
0.41
Sum
0.92
0.41
Fall
0.76
0.45
OC vs PM2.5-AS-Soil
Win
0.55
0.31
Spr
0.46
0.36
Sum
0.80
0.37
Fall
0.53
0.38
OC vs PM2.5-AS-Soil-AN-EC
Win
0.64
0.22
Spr
0.74
0.18
Sum
1.06
0.16
Fall
0.82
0.23
r
0.99
0.98
0.99
0.99
# bins
32
32
32
32
0.99
0.99
0.99
0.99
32
32
32
32
0.99
0.98
0.98
0.98
32
32
32
32
Seasonal OC vs PM2.5: R&P - 2025 Seq. Quartz, All Sites
Data sorted on PM, placed into 33 bins and averaged
OC vs PM2.5
Intercept
Slope
Win
1.84
0.44
Spr
1.94
0.31
Sum
1.40
0.44
Fall
1.43
0.48
OC vs PM2.5-AS-Soil
Win
1.44
0.35
Spr
1.67
0.27
Sum
1.06
0.43
Fall
1.14
0.46
OC vs PM2.5-AS-Soil-AN-EC
Win
1.38
0.18
Spr
1.35
0.16
Sum
1.33
0.13
Fall
0.17
0.27
r
0.42
0.62
0.91
0.92
# bins
32
32
32
32
0.64
0.66
0.91
0.94
32
32
32
32
0.82
0.82
0.68
0.97
32
32
32
32
Seasonal OC vs PM2.5: R&P - 2025 Seq. Quartz VSCC, All Sites
Data sorted on PM, placed into 33 bins and averaged
OC vs PM2.5
Intercept Slope
Win
0.94
0.18
Spr
0.83
0.18
Sum
1.01
0.17
Fall
0.54
0.25
OC vs PM2.5-AS-Soil
Win
0.77
0.33
Spr
0.75
0.39
Sum
1.11
0.40
Fall
0.43
0.46
OC vs PM2.5-AS-Soil-AN-EC
Win
Spr
Sum
Fall
r
# bins
0.93
0.97
0.95
0.97
32
32
32
32
0.95
0.97
0.97
0.99
32
32
32
32
Seasonal OC vs PM2.5: R&P 2300 Seq Quartz, All Sites
Data sorted on PM, placed into 33 bins and averaged
OC vs PM2.5
Intercept
Slope
Win
1.19
0.38
Spr
1.56
0.35
Sum
2.19
0.27
Fall
1.38
0.38
OC vs PM2.5-AS-Soil
Win
1.15
0.16
Spr
1.58
0.20
Sum
1.93
0.25
Fall
1.45
0.25
OC vs PM2.5-AS-Soil-AN-EC
Win
1.09
0.13
Spr
1.51
0.12
Sum
2.16
0.11
Fall
1.44
0.14
r
0.95
0.96
0.96
0.98
# bins
32
32
32
32
0.97
0.96
0.98
0.97
32
32
32
32
0.98
0.96
0.97
0.98
32
32
32
32
OC vs PM at individual sites - Spatial Variation in the
Carbon Artifact
• Spatial analysis of derived organic artifact
– Used Theil regression comparing measured OC to
reconstructed OC where reconstructed OC, OCrec =
(PM2.5 – Ammonium Sulfate – Fine Soil)/1.4 At each
monitoring site.
– Data were sorted on OCrec and placed into 33 bin.
The OC and OCrec in each bin was averaged
together.
– All figures are for only the Met One sampler.
Theil Intercept from OC vs. (PM2.5 – Amm SO4 – Fine Soil)
Met One Sampler for sites with r>0.7
Winter
Spring
Summer
Fall
Median of the Regression Line Intercept, Slope, correlation across
all Sites with r > 0.7, for each season and each sampler
Compared OC to PM2.5 – Amm. SO4 - Soil
Met One Sampler
OC vs PM2.5-AS-Soil
OC vs PM2.5
Intercept
Slope
r
Win
1.06
0.46
0.89
Spr
1.17
0.49
0.86
Sum
1.64
0.49
0.89
Fall
1.26
0.50
0.88
URG Sampler
OC vs PM2.5-AS-Soil
Intercept
Win
0.56
Spr
0.12
Sum
0.73
Fall
0.66
Anderson Sampler
OC vs PM2.5-AS-Soil
Win
0.98
Spr
1.02
Sum
1.55
Fall
0.99
853 R&P MDL2300 PM2.5 Seq Spec Quartz
OC vs PM2.5-AS-Soil
Win
1.20
0.29
0.86
Spr
1.49
0.33
0.83
Sum
1.59
0.38
0.82
Fall
1.22
0.43
0.83
0.56
0.63
0.58
0.60
0.92
0.92
0.91
0.89
Slope
0.47
0.49
0.53
0.57
r
0.93
0.94
0.94
0.93
Compared OC to PM2.5
Met One Sampler OC vs PM2.5
OC vs PM2.5
Intercept
Slope
Win
0.79
0.35
Spr
1.05
0.30
Sum
1.85
0.25
Fall
1.34
0.31
Anderson Sampler
OC vs PM2.5
Win
0.52
Spr
1.06
Sum
1.84
Fall
1.03
0.43
0.36
0.29
0.39
r
0.87
0.83
0.86
0.85
URG Sampler
OC vs PM2.5
Intercept
Win
0.51
Spr
0.41
Sum
1.12
Fall
0.68
0.87
0.84
0.87
0.84
853 R&P MDL2300 PM2.5 Seq Spec Quartz
OC vs PM2.5
Win
1.01
0.22
0.84
Spr
1.48
0.18
0.82
Sum
2.20
0.18
0.83
Fall
1.29
0.23
0.83
Slope
0.31
0.28
0.25
0.36
r
0.90
0.90
0.88
0.85
Median of the Regression Line Intercept, Slope, correlation across
all Sites with r > 0.7, for each season and each sampler
Compared OC to PM2.5 – Amm. SO4 - Soil
Met One Sampler
OC vs PM2.5-AS-Soil-AN-EC
OC vs PM2.5-AS-Soil-AN-EC
Intercept
Slope
r
Win
1.38
0.70
0.88
Spr
1.62
0.63
0.87
Sum
2.04
0.53
0.87
Fall
1.62
0.63
0.87
URG Sampler
OC vs PM2.5-AS-Soil-AN-EC
Intercept
Slope
Win
0.49
0.66
Spr
0.66
0.57
Sum
0.92
0.61
Fall
0.68
0.64
Anderson Sampler
OC vs PM2.5-AS-Soil-AN-EC
Win
1.26
0.83
Spr
1.56
0.67
Sum
1.83
0.66
Fall
1.42
0.76
853 R&P MDL2300 PM2.5 Seq Spec Quartz
OC vs PM2.5-AS-Soil-AN-EC
Win
1.20
0.61
0.78
Spr
1.73
0.48
0.81
Sum
2.20
0.44
0.80
Fall
1.18
0.60
0.81
0.94
0.91
0.88
0.85
r
0.90
0.90
0.93
0.92
Difference Plots: Derived Artifact - Field Blank at each site then contoured
This is for the Met One Sampler, derived used PM-AS-Soil
Similar to the Slide 26 – IMPROVE derived artifact from blank corrected data
Winter: Avg=0.11; median=0.04 ug/m3
Spring: Avg=0.2; median=0.24 ug/m3
Summer: Avg=0.57; median=0.61 ug/m3
Fall : Avg=0.24; median=0.2 ug/m3
IMPROVE data
Data from all sites 2001-04 were
used in the following analyses
Seasonal OC vs PM2.5: IMPROVE, All Sites (2001-04)
Data sorted on PM, placed into 150 bins and averaged
Theil
Theil
OC vs PM2.5 Intercept
Slope
Win
0.18
0.36
Spr
0.11
0.31
Sum
0.36
0.38
Fall
0.13
0.42
OC vs PM2.5-AS-Soil
Win
0.19
0.65
Spr
0.17
0.70
Sum
0.21
0.79
Fall
0.14
0.80
OC vs PM2.5-AS-Soil-AN-EC
Win
0.34
1.08
Spr
0.34
0.87
Sum
0.42
0.87
Fall
0.22
1.02
r
1.00
1.00
0.97
0.99
# bins
149
149
149
149
0.99
1.00
1.00
1.00
149
149
149
149
0.97
0.97
0.99
0.98
149
149
149
149
Seasonal theil Intercept from OC vs. (PM2.5 – Amm
SO4 – Fine Soil) with r>0.7, IMPROVE 2001-04 data
Winter
Summer
Spring
Fall
Median of the Regression Line Intercept, Slope,
correlation across all Sites with r > 0.7, for each season
Theil
Theil
OC vs PM2.5 Intercept
Slope
Win
0.16
0.35
Spr
0.09
0.31
Sum
0.02
0.46
Fall
0.07
0.41
OC vs PM2.5-AS-Soil
Win
0.22
0.65
Spr
0.18
0.65
Sum
0.20
0.81
Fall
0.12
0.80
OC vs PM2.5-AS-Soil-AN-EC
Win
0.50
0.95
Spr
0.32
0.74
Sum
0.34
0.86
Fall
0.30
0.92
r
0.86
0.84
0.86
0.88
# sites
115
137
132
129
0.87
0.87
0.91
0.91
116
146
139
138
0.86
0.82
0.91
0.88
58
97
131
103
Others
• Spatial patterns of derived carbon artifact
from STN data for each season, sampler
and different PM minus aerosol
components
81300.0
81310.0
81320.0
81330.0
Monitor (813=Met one), Season (0=win), OCrec (0 = (PM25- AS-Soil-AN-EC)/1.4)
81301.0
81311.0
81321.0
81331.0
81302.0
81312.0
81322.0
81332.0
82300.0
82310.0
82320.0
82330.0
82301.0
82311.0
82321.0
82331.0
82302.0
82312.0
82322.0
82332.0
83300.0
83310.0
83320.0
83330.0
83301.0
83311.0
83321.0
83331.0
83302.0
83312.0
83322.0
83332.0
85300.0
85310.0
85320.0
85330.0
85301.0
85311.0
85321.0
85331.0
85302.0
85312.0
85322.0
85332.0
Download