This presentation is on how to create optimized receptors for air dispersion modelling applications

1. Koray Önder, Geoff Scott and Rekha Nambiar
Self Optimizing Receptor Grid:
An Iterative Process for Generating Strategically
Placed R
Pl d Receptors f Di
t for Dispersion M d lli
i Modelling
Applications

2. Outline
 Background
 Approach
pp
 Local Scale Example - AERMOD
 Regional Scale Example - CALPUFF
 Summary

3. Background – Regulatory Modelling
 Ontario Modelling
Guideline:
(a) 20 m or less rectangle at least 200 m
less, rectangle,
from every source of contaminant;
(b) 50 m or less, area described in clause
(a); bounded by a rectangle, at least
300 m from the area described in
clause (a);
(c) 100 m or less, 800 m from the area
described in (a);
(d) 200 m or less, 1,800 m from the area
described in (a);
(e) 500 m or less, 4,800 m from the area
described in (a);
(f) 1,000 m or less, in the area that
surrounds the area described in(e).
(g) 10 m spaced at the fenceline

4. Background – Regulatory Modelling
 Alberta Modelling
Guideline:
 20 m receptor spacing i th
t i in the
general area of maximum
impact and the property
boundary,
 50 m receptor spacing within
500 m from the source,
 250 m receptor spacing
within 2 km from the sources
of interest,
 500 m spacing within 5 km
from the sources of interest,
 1000 m spacing beyond 5
km.

5. Background – Regional Modelling?
 Runtime linearly proportional to
# of receptors!
 Example: Athabasca Oil Sands
Region:
o Thousands of sources
o Scattered developments
o ~200×300 km area
o Terrain
o ~10,000 receptors

6. Approach – first iteration
 Model entire domain – coarsest resolution - resolution (a)
 Model entire domain – additional complement grid
 Increase overall resolution
 no removal of receptors!

7. Triangular/Hexagon Grid
 Receptors are exactly the same space from each other in every direction
 Resolution (b) = (a)/2

8. Approach – first iteration – and move on…
 Grid the coarse layer – SURFER®
 natural neighborhood method
 higher resolution than original ( ) - a/4 or finer resolution
g g (a)
 Compare gridded (a) resolution with modelled (b) resolution
 % difference [(b)-(a)]/(b)
 Plot the difference map
 classed post - highlight higher differences
 Decide on the areas to focus for next layer
 Generate next complement grid within the area of focus
 Resolution (c) = (b)/2
 Repeat same sampling b t
R t li between (b) and ( ) l
d (c) layers
 Keep moving on until all layers are generated

9. Approach – Alternative Resolutions…
 25  20
 50  40
 100  80
 200  160
 400  320
 800  640
 1,600
1 600  1,280
1 280
 3,200  2,560
 6,400  5,120
,
 12,800  10,240
 25,600  20,480

10. Example – AERMOD
 40 × 40 km area
 Two tall sources
 85 and 65 m
 Complex Terrain
 Grid resolution
 3,200
3 200 m
 1,600 m
 800 m
 400 m

11. Example: AERMOD – layer difference only
1-hour
 1-hour predictions: ±10% difference
 Blue - negative difference – overestimation by interpolation
 Red – positive difference – underestimation by interpolation
p y p
 picking up terrain effects
 not all places are important!

12. Example: AERMOD – layer difference only
24-hour
 Slight difference between 1-hour & 24-hour
 In general, highlighting similar areas

13. Example: AERMOD – introduce magnitude
1-hour
 Need more emphasis on the magnitude of predictions
 [((b)-(a))/(b)] × [(a)/max(a1-n)] : ±2%
 Focus on important areas - regulatory compliance purposes

14. Example: AERMOD – introduce magnitude
1-hour – zoom in…

15. Example: AERMOD – introduce magnitude
24-hour
 similar story…

16. Example: CALPUFF
 Annual SO2
 ~300×500 km area
 >1000 sources
 tall/short point sources
 mining areas – area sources

17. Example: CALPUFF – layer difference only

18. Example: CALPUFF – introduce magnitude

19. Summary
1. New Geometry- Triangular/Hexagonal
 distance between receptors same at each direction
 easy to add refined layers – consistent geometry – better blending
2. Iterative approach to add new refined layers
 introducing the magnitude focuses on where it really matters
 identifies places where higher resolution receptors are needed
 shows where they are not needed
3.
3 Streamlined d i i making process
St li d decision ki
 models/algorithms can make the decision on where to place denser
receptors

20. Questions?
THANK YOU!
Koray_Onder@golder.com
Geoff_Scott@golder.com
Rekha_Nambiar@golder.com