1. Koray Önder, P.Eng., Senior Air Quality Engineer
Greg Unrau, M.Sc., Associate, Senior Air Quality Meteorologist
Rekha Nambiar, M.Eng., Air Quality EIT
Paper #869
CALPUFF Model Sensitivity Analysis
Model Switch Comparison
3. Background
New Alberta Environment (AENV) Modelling
Guideline (May 2009) stipulates EPA default
CALPUFF switches
Any non-default switches have to be justified
Previously:
Most advanced switches turned “on” to use the
increased capability of the model
Some switches were problematic (e.g., MSHEAR)
3
4. Background
EPA recommends CALPUFF only for long range modelling
Alberta allows CALPUFF for near field applications
British Columbia and Ontario modelling guidelines allow
CALPUFF for near field applications
Non-default ‘MDISP’ AND ‘MPDF’ switches are recommended for
near field applications
Alberta Oil Sands Region: modelling efforts involve both
near field and long range (i.e., regional scale where
distances > 50km)
Inconsistency in model switches for near field vs. long range
applications – Challenge!
4
5. Modelling Methodology
CALPUFF version 6.263, level 080827
MODIS (Moderate Resolution Imaging Spectroradiometer)
based geophysical data
One year (2006) of meteorological data using local surface
and mesoscale model version 5 (MM5) output data
5
9. Sensitivity Analysis 2
Regional cumulative contributions from multiple sources
2006 emissions – existing sources in the Oil Sands region
(monthly production/operational variance included)
Three Cases Considered Based on Previous Results
Case 1 – U.S. EPA Defaults
Case 2 – Historic Oil Sands Switches
Case 3 – Alternate Switches (no Puff Split)
Modelling performed for two basic compounds:
Sulphur Dioxide (SO2)
Nitrogen Dioxide (NO2)
9
11. Model Performance Evaluation
Regional cumulative emissions predictions compared with
Wood Buffalo Environmental Association (WBEA)
monitoring station network data
Two model switch combinations
Case 1 – U.S. EPA Defaults
Case 2 – Historical Switches
Case 3 – Alternate Switches (no Puff Split)
Two Statistical Comparison methods
Fractional Bias
Normalized Root Mean Square Error (NRMSE)
11
12. Locations
Emission Sources
272 Point Sources
39 Area Sources
Modelling Domain
392 km by 564 km
12
17. Conclusions
US EPA Default switches (Case 1) results lower – except
area sources
Should not be generalized!
Historic Switches (Case 2) – considerably higher predictions
Alternate Switches (Case 3 - BC and ON guidelines) – in
between
Model Performance Evaluation: No clear winner
Results does not warrant recommending an alternate; however…
Regulatory Compliance Modelling – Oil Sands Region:
Using different switches not practical
Both Near Field / Long Range – primary focus: near field
BC and Ontario recommend alternate switches for near field
Alternate Switches – conservative and newer science
17