Professional Documents
Culture Documents
Calculating Statistics: Concentration Related Performance Goals
Calculating Statistics: Concentration Related Performance Goals
Concentration Related
Performance Goals
James W. Boylan
Georgia Department of Natural Resources
Outline
Performance Statistic
Standard Bias and Error Calculations
Performance Metrics
Equation
1
MB
N
1
ME
N
C
i 1
N
C
i 1
1 N C m Co
MNB
N i 1
Co
N
NMB
C
i 1
1
MFB
N
i 1
Co
1
MNE
N
i 1
C m Co
Co
C
i 1
C
i 1
1
MFE
N
Co
Co C m
NME
C m Co
Co
Co
i 1
C m Co
i 1 Co C m
Example
North Carolina 77
Georgia Tech 88
GT showed a positive bias of 11 points
NB = 14.3%
FB = 13.3%
Performance Metrics
Mean Normalized Bias and Error
Usually associated with observation-based minimum
threshold
Some components of PM can be very small making it
difficult to set a reasonable minimum threshold value
without excluding a majority of the data points
Performance Metrics
Normalized Mean Bias and Error
Biased towards overestimations
Example Calculations
Model
g/m3
)
Obs.
g/m3
)
MB
NMB
(%)
g/m3
)
0.05
1.0
-0.95
-95
-180.95
+0.95
+95
+180.95
1.0
0.05
+0.95
+1900
+180.95
+0.95
+1900
+180.95
1.0
0.01
+0.99
+9900
+196.04
+0.99
+9900
+196.04
0.683
0.353
+0.33
+3901.7
+65.3
0.96
3965.0
186.0
+93.4
MNB
(%)
MFB
(%)
ME
g/m3
)
NME
(%)
272.9
MNE
(%)
MFE
(%)
Speci
es
#
Obs
Mea
n
g/m
MB
g/m
NMB
(%)
MNB
(%)
MFB
(%)
ME
g/m
NME
(%)
MNE
(%)
MFE
(%)
SO4
134
6.71
-1.16
-17.3
1.1
-22.7
2.48
37.0
55.1
50.2
NO3
134
0.63
-0.30
-47.6
6.8
-73.6
0.52
81.8
112.8
107.2
NH4
134
2.70
-1.25
-46.4
-27.4
-57.4
1.43
53.1
61.6
70.0
NH4
Bi
134
1.44
0.01
0.4
34.2
-2.6
0.62
42.9
70.4
44.4
ORG
132
3.41
-0.27
-7.8
15.8
-6.0
1.37
40.4
53.8
43.9
EC
132
0.56
-0.05
-8.6
15.1
-12.7
0.27
48.3
61.9
50.4
Soils
135
0.55
0.25
46.2
171.6
21.9
0.57
102.9
207.4
72.5
PM2.5
130
17.05
-4.79
-28.1
-9.1
-28.8
6.8
39.8
48.9
47.6
PM10
130
23.44
-5.21
-22.2
-6.2
-21.0
9.18
39.1
44.2
43.5
PMC
126
6.98
-0.48
-6.9
43.9
7.8
3.86
55.2
78.7
54.1
bext
132
133.1
-27.91
-21.0
-10.2
-23.7
43.70
32.8
40.0
40.4
MFE 150e
0.75 g / m 3
50
0.5 ( Co C m )
MFB 170e
0.5 g / m 3
30
Example Calculations
Species X
Model
g/m3)
Obs.
g/m3)
FB (%)
FE (%)
Day 1 Site A
2.0
1.0
+66.7
+66.7
Day 1 Site B
1.0
2.0
-66.7
+66.7
Day 2 Site A
1.0
0.4
+85.7
+85.7
Day 2 Site B
0.5
1.5
-100.0%
+100.0%
Average
1.125
1.225
-3.6%
79.8%
200
150
100
50
0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
200
100
0
-100
-200
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
SAMI 6 Episodes
Speciated Fine PM Performance
200
Nitrate
Soils
150
Elem. Carbon
Ammonium (Sulfate)
100
Organics
PMC
PM2.5
PM10
50
Ammonium (Bisulfate)
Sulfate
0
0.0
5.0
10.0
15.0
20.0
25.0
SAMI 6 Episodes
Speciated Fine PM Performance
200
Soils
Elem. Carbon
100
Ammonium (Bisulfate)
PMC
PM2.5
PM10
0
Organics
-100
Sulfate
Ammonium (Sulfate)
Nitrate
-200
0.0
5.0
10.0
15.0
20.0
25.0
200
Nitrate
Soils
150
Organics
Elem. Carbon
100
PM2.5
PM10
PMC
50
Sulfate
Ammonium (Sulfate)
0
0.0
5.0
10.0
15.0
20.0
25.0
200
Elem. Carbon
PM2.5
Soils
100
PM10
Ammonium (Sulfate)
0
Sulfate
-100
PMC
Nitrate
Organics
-200
0.0
5.0
10.0
15.0
20.0
25.0
200
Soils
150
Nitrate
Elem. Carbon
100
PM2.5
PMC
50
PM10
Organics
Ammonium (Sulfate)
0
0.0
2.0
4.0
Sulfate
6.0
8.0
10.0
12.0
200
Soils
PM2.5
Elem. Carbon
100
Ammonium (Sulfate)
PM10
Organics
0
Sulfate
-100
Nitrate
PMC
2.0
4.0
-200
0.0
6.0
8.0
10.0
12.0
1 N
Bias
N i 1
N C
Cm component
o component
1
C
N
Co Total
1
i
m Total
0.2 RP%
(5%)
Example Calculation
Calculating component proportions based on
concentrations averaged over multiple days
can hide poor model performance
Observed RP (%)
Modeled RF (%)
Day 1
50%
95%
Day 2
50%
95%
Day 3
50%
5%
Day 4
50%
5%
Average
50%
50%
4.3%
4.7%
SO4
6.8%
ORG
2.8%
NH4
10.8%
NO3
12.3%
47.1%
50.5%
25.6%
Observed
EC
Soils
26.7%
Simulated
Elem. Carbon
Bias (%)
10
Soils
Nitrate
Organics
Sulfate
Ammonium (Bisulfate)
-5
-10
-15
10
20
30
40
50
60
1
Error
N
Cmcomponent
i 1
Cm
Total
Cocomponent
Co
Total
Error (%)
15
10
Soils
Nitrate
Organics
Sulfate
Ammonium (Bisulfate)
Elem. Carbon
10
20
30
40
50
60
Concluding Remarks
Recommended performance values are
model goals, not model criteria
Failure to meet proposed performance goals
should not prohibit the modeling from being
used for regulatory purposes
Help identify areas that can be improved upon
in future modeling
Questions?