You are on page 1of 19

Traffic and Revenue Forecasting for

the Dhaka-Chittagong Expressway:


Distributed Values of Time and
Vehicle-Specific Volume-Delay Functions
Richard Di Bona
Independent Consultant (Hong Kong)
David Freer
Principal Transport Planner, SMEC Australia & New Zealand Division (Brisbane)

Model City Seminar, Hong Kong, 24th June, 2016


Current Situation
• 250km National Highway 1 connects Dhaka (capital, 15-17m people in metro
area) and Chittagong (2.5m people and Bangladesh’s deep sea port)
• Single-2, being upgraded to Dual-2
• But runs through towns
• Motorised and non-motorised traffic
• Very slow: can take 6-8 hours or more to drive by car
• Average speed around 30-40kph
Current Situation
Heavy truck traffic on National Highway 1 (NH1): Not counting auto-rickshaws (20%
of vehicles), Goods Vehicles are 55% of vehicles and 64% of pcu’s
• Trucks travel more slowly than cars
Trucks also have to park-up and wait:
• For a loading/ unloading slot
at Chittagong Port
• And due to a daytime truck ban
in Dhaka
Current Widening of NH1
NH1 is being widened from Single-2 to Dual-2 standard, and with a flyover being
constructed within Comilla (a city roughly 40% of way from Dhaka to Chittagong)
But NH1 would still run through towns, e.g.:
• Southern Dhaka, Sonargaon,
• Daudkandi, Eliotganj, Chandina,
• Nimsar, Paduar Bazar, Chauddagram,
• Feni, Sitakund, Bhatiari
And national population density higher than many American cities!
Purposes of Transport Modelling
Modelling was intended to investigate a number of questions:
• Decongestion, economic & environmental impacts of constructing an Expressway
• Traffic and revenue forecasts
• Assessing impacts of Dual-2 or Dual-3 National Highway 1
• Help determine preferred alignment: from 3 options:
Key Modelling Issues
Note: I am not discussing engineering, land acquisition, political
or other issues!
Key modelling issues were:
1. A large number of scheme permutations and combinations to
be analysed
2. Accuracy and reliability issues of toll traffic models, in general
3. Experience and knowledge of modelling team
4. Establishing robust data on willingness-to-pay tolls
5. Different speeds for different vehicle types
Full List of Scenarios
Besides 2014 calibration, models were run (on 24-hour basis) for:
• Forecast years 2020, 2025, 2030 and 2040 (4)
• Dual-2 and Dual-3 configuration of National Highway 1 (2)
• Expressway options: “Do Nothing” and: (1+ 6x8 = 49)
– Three possible alignments, at Dual-2 or Dual-3 configuration (2x3=6)
– Toll rates (based on BDT/km for Medium Truck): 0, 2, 2½, 3, 3½, 4, 5 and 6 (8)
• Economic growth scenarios (3)

So a total of 4 x 2 x 49 x 3 = 1,176 model scenarios were run


Analysis spreadsheets automated (VisualBasic) to handle EMME reports to provide:
• Interpolation for non-modelled years
• Transport cost and savings streams (travel time, VOC, emissions, etc)
• Toll revenue streams
Toll Forecast Accuracy/ Reliability
Construction
Delay Willumsen, L. and Russell, C.
(1998) “Reducing Revenue Risk”,
Change Orders European Transport Conference,
Loughborough, 14-18 September
1998
Risk (nominal)

Construction
Costs

Ramp Up

Traffic &
Revenue

O&M
er
-2

-1

10
0

ov
d
an
H

Year
Toll Forecast Accuracy/ Reliability
Average initial year traffic: 70% of forecast
• For lender-commissioned studies: 82%
• For other studies: 66%
– Bain, R. and Wilkins, M. “Credit Implications of Traffic Risk in Start-Up Toll Facilities”,
Standard & Poor’s, September 2002
Due to Ramp Up?
• Can last 3-5 years
– Streeter, W. and McManus, K. “Challenges of Start-Up Toll Roads”, Project Finance
Special Report, Fitch ICBA, 1999
Or optimism bias constant up to year 5?
– Bain, R. and Polakovic, L, “Traffic Forecasting Risk Study Update 2005: Through Ramp-
Up and Beyond”, Standard & Poor’s, August 2005

So substantial care and caution required!


Primary Author’s Experience
• Over 23 years’ experience gained on projects in over 30
countries across 6 continents
• Using EMME since 1995
• Substantial experience in patronage forecasting and
advisory studies covering toll roads, public transport
systems, Value of Time, etc in countries across all stages of
economic development, including for: BOT’s, PPP’s, IPO’s,
Secondary Listings, Bond Issues, M&A’s and Financial
(Re)Structuring
• Experience coupled with economic analysis, including
impacts of business cycles (e.g. price inflation and interest
rate risks on projects)
Willingness-to-Pay: Lack of RP Data
Only tolls in Study Area are at Meghna
and Gumti River bridges:
• Tolls charged one-way on either
bridge
• No alternative routes
So no “choice” data available
This increases “traffic risk” as per
Traffic Risk Index, also:
• Toll road not open
• Runs parallel to untolled route
• Bangladesh “low income” country
– Bain, R. and Wilkins, M. “Credit
Implications of Traffic Risk in Start-
Up Toll Facilities”, Standard &
Poor’s, September 2002
Establishing Willingness-to-Pay
“Triangulation” from three sources for Behavioural Values of
Time (BVOT):
1) Economic resource cost analysis of travel time savings
2) Interviews with major road freight hauliers (operators and
key customers)
3) Stated Preference (SP) interview surveys with drivers

SP data showed that use of a simple mean BVOT by vehicle


class would suffer very badly from the “flaw of averages”:
• Whether logit diversion curves or generalised cost
methodologies were used (both were tested)
Car BVOT (BDT/ Vehicle/ Hour)
1,600

1,400

1,200
(BDT per vehicle per Hour)
Behavioural Value of Time

1,000

800

600

400

200

0
0 1,000 2,000 3,000 4,000 5,000 6,000
Number of Respondents (Sorted by BVOT per Vehicle)

Distributed Segment Mean


GV BVOT (BDT/ Vehicle/ Hour)
1,600

1,400

1,200
(BDT per vehicle per Hour)
Behavioural Value of Time

1,000

800

600

400

200

0
0 2,000 4,000 6,000 8,000 10,000 12,000
Number of Respondents (Sorted by BVOT per Vehicle)

Distributed Segment Mean


Different Vehicle Speeds
• EMME usually has a single set of volume-delay functions
applied to all vehicles in a multi-class assignment
• Preloading was not deemed appropriate as all vehicle
classes affected by one another and all vehicle classes
likely to “choose” routeing
• Different speeds for different vehicles addressed using
multi-class generalised cost assignment
Different Vehicle Speeds
Outline of method:
• Define a vehicle type-specific free-flow speed for each
vehicle class
• Calculate speed differential from “base” vehicle type (car)
• Calculate additional delay for bus and truck
• Include this additional delay in generalised cost term, which
is assignment-class specific
• Overall “shape” of vdf remains unchanged, but delays
“factored up” when considering bus and trucks
Multi-Class Generalised Cost Assignment
• 12 vehicle sub-classes:
– 3 x Car
– 3 x Bus (BVOT’s not shown here)
– 6 x Goods Vehicle
• Behavioural Value of Time for each of 12 sub-classes
• Vehicle Operating Costs for 3 classes (Car, Bus & Truck)
– Also included allowance for road roughness on local roads
• Tolls (as appropriate) for 3 classes
• Volume-Delay functions by link type, incorporating:
– Impacts of frontage interference through towns
– Impacts of non-motorised transport and auto-rickshaws
– Different speeds for different vehicle types
What Might Be Done Differently?
Having selected an alignment (Alignment #1, parallel to NH1),
if more detailed investigation were required, we would:
• Increase the number of sub-classes, to improve definition
of willingness-to-pay segments
– Assignment class limit has now been substantially relaxed
– WTP segments more significant than size of truck or car or bus
(though cross-tabulations could be investigated also)
• Try to increase zoning beyond the 87 zones used here,
especially along proposed alignment
– But getting robust land use & planning forecast data at even
this level is a challenge!
Thank
You!
rfdibona@yahoo.com
david.freer@smec.com

You might also like