1. DEVELOPMENT OF A FLEXIBLE ZONING SYSTEM
TRBAppCon 2017
DEVELOPMENT OF A FLEXIBLE ZONING SYSTEM
APPLIED TO NATIONAL OR STATE-WIDE MODELS
May 2017 – Raleigh

2. SVITM (State of Victoria)
City or State
Model
MITM
(Melbourne)
Land Area
3,860 m2
Population
3.7 m
Transport Zones
2893
Run Time (full)
24 hours
SVITM (State of Victoria)
91,750 m2
5.7 m
5595
7 days
Run Time (aggregated)
24 hours
~8 hrs – 7 days

3. TRANSPORT FOR VICTORIA
Contributors:
Dr Henry Le
AECOM
Craig Somerville
Neville Wood
TRANSPORT FOR VICTORIA 
Oliver Charlesworth
Filippo Contiero
CITILABS

4. One Model…for everything…

5. SVITM: Zone Aggregation Tool (Cube Voyager)

6. SVITM: Structure Multi trip purposes (11)
Matrices, Sectors and Network Aggregation
Commercial Vehicle Module
Trip generation
Urban & Intercity Trips
Mode Choice/ Distribution
Assignment
Convergence
Report
Initialisation
Inputs of network and demographic data at fine detail (all zones)
Model runs at aggregated level from here
Subarea Analysis
Subarea analysis at subarea zones
Loop iteration
Multi trip purposes (11)
Multi modal: car, train, tram, bus
Multi period: AM, IP, PM, OP
Zoning: max zones
Multi-regional: Urban, Intercity

7. SVITM: Zone Aggregation (example)
All zones within 30 minutes of the study area should have no aggregation
All zones within minutes of the study area could have fine aggregation
All zones within >60 minutes of the study area could have fine or broad aggregation depending on local characteristics

8. SVITM: Zone Aggregation (example)
Study Area
700
Fine Aggregation
500
Broad Aggregation
170
Total no. zones
1370
Run time
8 hrs

9. SVITM: Trip Distribution (MITM study area)

10. Citilabs Application:
London Transportation Studies Model (LTS)

11. LTS Application: Aggregation
Greater
London
Great
Britain

12. LTS Application: Aggregation
Normal
Partial

13. LTS Application: Results
Cluster Nodes 4
Variation
Processors
Intel® Xeon® CPU 3.50 GHz
Test case / Number of zones
4106 zones
3237 zones
-21%
Runtime [hours]
02:35
02:00
-23%
Iterations 59 -
Number of selected counts
409
Correlation coefficient
0.9462
0.9430
-0.34%
Determination coefficient (R2)
0.8953
0.8892
-0.68%
RMSE (Root Mean Squared Error)
207.73
190.98
-8.06%
CV (RMSE) (Coeff. of Var. of the RMSE)
0.3871
0.4377
13.07%
NRMSD (Norm. Root Mean Squared Error)
0.0605
0.0556

14. Summary
The inclusion of long distance trips and an aggregated zone system enables the model to be applied practically and consistently to any regional city in Victoria.
The aggregation technique successfully reduces runtime without significant impact to model results
The network and model relationship can be continuously improved as the methodology is progressively applied to other areas.
Any number of aggregation levels can be added
Network simplification can add another level of efficiency (takes some thought)

15. Bibliography
Dr Henry Le, Craig Somerville, Neville Wood, AITPM 2013 National Conference, Development of a Statewide Victorian Integrated Transport Model
Dr Henry Le, Craig Somerville, Neville Wood, AITPM Excellence Awards Category of Transport and Land Use Modelling, Development of a Statewide Victorian Integrated Transport Model
VISTA, Department of Transport, Planning and Local Infrastructure Victoria

16. Dr. Oliver Charlesworth
Global Account Director, Citilabs