The document compares two methods for predicting nitrogen dioxide (NO2) concentrations from oil sands operations in Alberta, Canada: the Ambient Ratio Method (ARM) and the Ozone Limiting Method (OLM). It finds that both methods over-predict NO2 concentrations compared to observed values, with CALPUFF performing better for 24-hour and monthly predictions than 1-hour predictions. Fractional bias values were within acceptable ranges. Overall, ARM and OLM showed similar performance in modeling NO2 concentrations.
4. Background – Emissions in the Region
Emission Sources
272 Point Sources
39 Area Sources
Modelling Domain
392 km by 564 km
Sources NOX Emissions (t/d)
Shell Muskeg River Mine 14.15
Syncrude Mildred Lake 34.81
Suncor Firebag 1.64
Suncor Base Plant 63.28
Syncrude Aurora North 12.81
Conoco Phillips Surmont Pilot 0.08
Other Regional Sources 86.00
Total 212.77
5. Background – Emissions in the Region
Table 1. Predicted NO2 concentrations at WBEA monitoring stations using ARM
Station 1-hour 24-hour Monthly
Albian Mine 215.0 141.7 61.4
Fort McKay 86.6 59.3 28.0
Athabasca Valley 87.3 48.0 22.5
Patricia McInnes 87.3 61.2 32.0
Millennium 148.9 106.7 73.0
Syncrude UE-1 86.1 55.1 26.9
Table 2. Predicted NO2 concentrations at WBEA monitoring stations using OLM
Station 1-hour 24-hour Monthly
Albian Mine 288.4 127.8 66.9
Fort McKay 79.1 41.6 24.9
Athabasca Valley 87.4 43.3 24.4
Patricia McInnes 84.3 47.7 26.4
Millennium 140.7 110.3 83.1
Syncrude UE-1 72.9 41.4 24.2
7. Methodology
Alberta Environment (AENV) MM5 data at 12 km for
2006
Surface and upper air observations from Environment
Canada stations
CALMET
horizontal grid resolution – 4 km
MODIS – Land use, monthly LAI and monthly Albedo
Emissions scenario:
existing oil sands sources in 2006
emissions prorated based on production
8. Methodology
NOX and NO2 concentrations predicted at ten Wood
Buffalo Environmental Agency (WBEA) monitoring
stations
2006 predictions compared to 2006 observations
CALPUFF performance measured using
Fractional bias (FB)
Normalized root mean square error (NRMSE)
10. Methodology
ARM
Digitized points along the NO2/NOX curve
CALPOST postprocessor used to predict concentrations at
WBEA monitoring stations
OLM
Ozone concentration determined based on WBEA
monitoring data
Hourly ozone concentrations from Athabasca Valley
External program used to generate NO2 concentrations
11. Methodology – Fractional Bias
Fractional Bias
FB = 2 [OB – PR]/[OB + PR]
Based on highest 25 values
FB of -0.67 indicates over prediction by factor of 2
FB of +0.67 indicates under prediction by factor of 2
FB values between -0.67 and +0.67 are considered
acceptable model performance
FB values closer to 0 indicate better model performance
12. FB – NOX 4 km Results
Hourly
Hourly NOX - 4 km Daily
Daily NOX - 4 km Monthly
Monthly NOX - 4 km
Bias of Standard Deviation
2.00 2.00 2.00
1.00 1.00 1.00
n n n
o
i o
i o
i
t t
a t
a
i i a
i
v v
e v
e e
D D D
d 0.00 d
r 0.00 d
r 0.00
r
a-2.00 a a
d -1.00 0.00 1.00 2.00 d -2.00 -1.00 0.00 1.00 2.00 d -2.00 -1.00 0.00 1.00 2.00
n n
a n
a t a
t
t
S S S
f f f
o o o
s s
a s
a i a
i
i B B
B -1.00 -1.00 -1.00
-2.00 -2.00 -2.00
Bias of Mean Bias of Mean Bias of Mean
Bias of Mean
13. FB – NO2 ARM and OLM Results
Hourly
Hourly NO2 - 4 km ARM
Daily
Daily NO2 - 4 km ARM
Monthly
Monthly NO2 - 4 km ARM
2.00 2.00 2.00
1.00 1.00 1.00
n n n
o
i o
i o
i
t t
a t
a
a
i i i
Bias of Standard Deviation
v v v
e e e
D 0.00 D 0.00 D 0.00
d d
r-2.00 d
r
r-2.00
a -1.00 0.00 1.00 2.00 a -1.00 0.00 1.00 2.00 a
-2.00 -1.00 0.00 1.00 2.00
d d d
n n
a n
a
a
t t t
S S
f S
f
f o o
o s s
s
a
-1.00 a
i
-1.00
a
i
-1.00
i B B
B
-2.00 -2.00 -2.00
Hourly NO2 of 4 km OLM
Bias - Mean Daily NO2 -of Mean
Bias 4 km OLM
Monthly Bias of MeanOLM
NO2 - 4 km
2.00 2.00 2.00
1.00 1.00 1.00
n n
o n
o
i i o
i
t t
a t
a
i i a
i
v v v
e e e
D 0.00
D 0.00 D 0.00
d d
r -2.00 d
r-2.00
a -1.00 0.00 1.00 2.00 a -1.00 0.00 1.00 2.00 r-2.00
a -1.00 0.00 1.00 2.00
d d d
n n
a n
a
t t
S a
t
S
f f S
f
o
o s
a
o
s
a -1.00 i -1.00 s
a -1.00
i B i
B B
-2.00 -2.00 -2.00
Bias of Mean Bias of Mean Bias of Mean
Bias of Mean
14. FB – Results Discussion
NOX
CALPUFF performed well against observations
FB values within a factor of 2 for most WBEA stations
Emissions are modelled reasonably well
NO2
Overall the concentrations are over predicted
CALPUFF performance better for 24-hour and annual
averaging periods
ARM and OLM show similar results
15. Quartile-Quartile – Fort McKay
1 - hr 24 - hr
Fort McKay - 4 km ARM
Monthly- 4 km ARM
Fort McKay
1,000
1000 100
100 100
100
ARM ARM ARM
ARM
m
m
g
100
100
g
µ
µ
]
³
2[
]
³
2[
/
/
10 10
O
-H
N
10
u
o
10
r
NO2 [µg/m3]
10
10
M
O
N
P
P
d
h
n
o
d
4
2
c
e
y
c
e
r
r
Ranked Comparison
t
t
i
l
i
Unbiased Line
Factor of 2
1 Factor of 4
1 11
1,000 1
1
11 10
10 100
100 1000
1,000 1
1 10
10 100
100 1
1 10
10 100
100
Fort McKay - 4 km OLM Monitored 24-Hour km OLM
Fort McKay - 4 NO2 [µg/m³] Monitored Monthly NO2 km OLM
Fort McKay - 4 [µg/m³]
-H Predicted
1000
1,000 100
100 100
100
OLM OLM OLM
100
m
m
g
100
µ
g
µ
]
³
2[
]
³
2[
/
100
/
m
O
N
10 10
µ
o
u
O
2
-H
N
g
10
/
10
u
o
r
³
]
[
r
10
10
d
P
1
e
c
r
t
i
10
M
O
N
P
P
h
n
o
d
d
y
c
e
4
2
c
e
r
r
t
t
l
i
i
11 1
1 1
1
1
1 10
10 100
100 1000
1,000 1
1 10
10 100
100 1
1 10
10 100
100
Monitored 1-Hour NO2 [µg/m³] Monitored 24-Hour NO2 [µg/m³] Monitored Monthly NO2 [µg/m³]
Monitored 1 2 [µg/m3]
NO
1 10 100 1,000
16. Quartile-Quartile - Millenium
1 - hr
Millennium - 4 km ARM
24 - hr4 km ARM
Millennium -
Monthly- 4 km ARM
Millennium
1000
1,000 100
100 100
100
ARM ARM ARM
m
m
m
u µ
N g
O /
100
g
³
g
o [
]
µ
µ
-H 2
]
³
2[
100
]
³
2[
/
/
r
10 10
O
-H
N
10 10
u
o
r
Predicted NO2 [µg/mP ]
d
1
e
c
r
t
3
10
i
10
M
O
N
P
P
d
h
n
o
d
4
2
c
e
y
c
e
r
Ranked Comparison
r
t
t
i
l
i
Unbiased Line
Factor of 2
Factor of 4
11 1
1 1,000 1
1
1
1 10
10 100
100 1000
1,000 1
1 10
10 100
100 1
1 10
10 100
100
Millennium - 4 km OLM
Monitored 1-Hour NO2 [µg/m³] Monitored 24-Hour NO2 km OLM
Millennium - 4 [µg/m³] Monitored Monthlykm 2 [µg/m³]
Millennium - 4 NO OLM
1000
1,000 100
100 100
100
OLM OLM OLM
100
m
m
m
g
g
g
µ
µ
100
µ
100
]
³
2[
]
³
2[
]
³
2[
/
/
/
10 10
O
O
-H
N
-H
N
10 10
u
o
u
o
r
r
10
10
10
P
d
1
c
e
M
O
r
N
P
t
P
d
i
h
n
o
d
2
c
4
e
y
c
e
r
r
t
t
i
l
i
1
1 1
1 1
1
1
1 10
10 100
100 1000
1,000 1
1 10
10 100
100 1
1 10
10 100
100
Monitored 1-Hour NO2 [µg/m³] Monitored 24-Hour NO2 [µg/m³] Monitored Monthly NO2 [µg/m³]
Monitored 1 2 [µg/m3]
NO
1 10 100 1,000
17. Quartile-Quartile – Syncrude UE-1
1 - hr
Syncrude UE-1 - 4 km ARM
24 - hr
Syncrude UE-1 - 4 km ARM
Monthly - 4 km ARM
Syncrude UE-1
1000
1,000 100
100 100
100
ARM ARM ARM
m
m
m
100
g
g
µ
µ
µ
g
/
³
]
2[
2
]
³
2[
]
³
[
100
/
/
O
N
-H
O
-H
N
o
u
10 10
u
o
r
10 10
r
Predicted NO2 [µg/m3]
10
10
d
P
1
e
c
r
t
i
P
M
d
O
4
2
c
e
N
P
h
n
o
d
r
y
c
e
t
i
Ranked Comparison
r
t
l
i
Unbiased Line
Factor of 2
Factor of 4
11 11
1,000 1
1
1
1 10
10 100
100 1000
1,000 1
1 10
10 100
100 1
1 10
10 100
100
Syncrude UE-1 - 4 km 2OLM
Monitored 1-Hour NO [µg/m³] Syncrude UE-1 - 4 km
Monitored 24-Hour NO2[µg/m³] OLM Syncrude UE-1 - 4 km OLM
Monitored Monthly NO2 [µg/m³]
1000
1,000
100
100 100
100
OLM OLM OLM
100
m
m
m
g
g
g
µ
µ
100
µ
]
³
2[
]
³
2[
]
³
2[
100
/
/
/
10 10
O
O
-H
N
-H
N
10 10
u
o
u
o
r
r
10
10
10
P
d
1
c
e
M
O
r
N
P
t
P
d
i
h
n
o
d
2
c
4
e
y
c
e
r
r
t
t
i
l
i
11 1
1 1
1
1
1 10
10 100
100 1000
1,000 1
1 10
10 100
100 1
1 10
10 100
100
Monitored 1-Hour NO2 [µg/m³] Monitored 24-Hour NO2 [µg/m³] Monitored Monthly NO2 [µg/m³]
Monitored 1 2 [µg/m3]
NO
1 10 100 1,000
18. Methodology – NRMSE
Normalized Root Mean Square Equation
x1,i represents the modelled concentration
x2,i represents the monitored concentration
xmax represents the maximum observed concentration
xmin represents the minimum observed concentration
Measure of model performance
Smaller NRMSE values means better model
performance
19. NRMSE – Results
Table 1. NRMSE values for ARM NO2 versus observations
Station 1-hour 24-hour Monthly Total
Albian Mine 0.90 3.21 0.80 4.9
Fort McKay 1.18 3.09 0.16 4.4
Athabasca Valley 0.85 0.59 0.38 1.8
Patricia McInnes 1.61 7.08 0.83 9.5
Millennium 2.20 7.84 1.23 11.3
Syncrude UE-1 1.82 3.81 0.38 6.0
Total 8.6 25.6 3.8
Table 2. NRMSE values for OLM NO2 versus observations
Station 1-hour 24-hour Monthly Total
Albian Mine 1.55 2.58 1.46 5.6
Fort McKay 1.13 0.46 0.32 1.9
Athabasca Valley 0.76 0.99 0.39 2.1
Patricia McInnes 1.84 5.18 0.83 7.9
Millennium 1.88 9.17 1.21 12.3
Syncrude UE-1 1.10 1.62 0.46 3.2
Total 8.3 20.0 4.7
20. NRMSE – Results Discussion
1-hour and 24-hour values slightly better than monthly
values for OLM
ARM performed better at some stations while OLM
performed better at others
No clear winner !
21. Conclusion
NO2/NOX relationship is non-linear
FB results for NOX show emissions are modelled reasonably
well
Performance of ARM and OLM is similar
No clear winner !!
CALPOST can do ARM so logistically easier to use