Unit-IV; Professional Sales Representative (PSR).pptx
Methodology and Learnings from Calculating the Cost of the Causes of Congestion
1. Methodology and Learnings from
Calculating the Cost of the
Causes of Congestion
David Johnston, Intelligent Transport Services
Kath Johnston, QLD Transport and Main Roads
27 July 2016
2. Project Objective
To produce a congestion pie
for TMR similar to the FHWA
example, but with the
following causes of excessive
congestion:
• Recurring congestion
• Traffic Incidents
• Roadworks
• Inclement Weather
• Special Events/Other
3. Steps in Methodology
A) Import data
B) Generate benchmarks of
link performance (i.e. ‘Normal’)
C) Generate congestion cost components
(delay, fuel use, pollutants)
D) Generate abnormal congestion footprints
E) Map causes onto abnormal congestion footprints
F) Produce reports
Start
(A) Import data
for processing
(B) Generate
benchmark link
performance
profiles
(C) Generate
Congestion Cost
measures
End
(E) Map causes
onto abnormal
congestion
footprints
(F) Produce
reports
(D) Generate
abnormal
congestion
footprints
4. Import Data
• NPI Link Data – Speed & Volume
• STREAMS Transport Network model – links, intersections,
movements, NPI Links
• Weather data (30 minute rainfall observations)
• SIMS data – incidents, roadworks, planned events.
• 131940 data (traffic information line)
• Fleet data - % by vehicle type, % business / private use
• Unit cost data – delay (ABS wages), fuel, pollution
5. Step B: Benchmark ‘Normal’ Traffic
METHODOLOGY AND LEARNINGS
FROM CALCULATING THE COST
OF THE CAUSES OF CONGESTION
6. ‘Normal’ Profiles
A profile defines what is ‘normal’ for an NPI Link and each 15-
minute period
• Profile holds mean & standard deviation of volume & speed
across days selected for profile
• Multiple profiles across the days in a data set
• Key question: How do you select which days to include in a
profile?
7. Day Types in the Calendar
The following attributes are identified in the calendar for each day:
• Weekday (Sat and Sun will normally be different to Mon – Fri)
• Season (More travel to & from the beach during summer)
• Public Holidays
• School Holidays
• School Fringe (e.g. November when grade 10-12 out, private
schools)
• Late night shopping (Thursdays plus week before Christmas)
8. Break types associated with Days
Further intelligence required for public holidays near weekends
• Each weekend is a 2-day “break”
• If Friday is a public holiday, Thursday traffic will be more like a
normal Friday
• If Thursday is a public holiday, Wednesday will be like a normal
Friday and Friday will be much quieter than normal.
• To ‘learn’ these, the calendar identifies each day as one of:
a) Day not in “break”;
b) Day before “break”;
c) First day of “break”;
d) Day inside “break”;
e) ‘Normal’ day during “break”;
f) Last day of break; or
g) Day after “break”
9. Step C: Generate Congestion Cost Components
METHODOLOGY AND LEARNINGS
FROM CALCULATING THE COST
OF THE CAUSES OF CONGESTION
10. 10 |
Daily cost of congestion for Brisbane state-controlled roads
(Network & Performance Team E&T Road Operations Feb 2016)
11. Allocation of Costs
Excessively Congested
(as per ARRB formula)
Not Excessively Congested
(as per ARRB formula)
Less than Normal
Congestion All congestion cost attributed to
Recurring Excessive Congestion.
No cost of excessive congestion
to allocate.
Normal
Congestion
Greater than
Normal
Congestion
Any ‘normal’ congestion cost
attributed to Recurring Excessive
Congestion.
All excessive congestion cost
attributed to one or more
causes.
17. Merging Abnormal Congestion Footprints
• Where separating link is excessively congested and this is
normal, merge the abnormal congestion footprints.
• NPI Link X meets this condition. NPI Link Y does not.
18. Step E: Map Causes onto
Abnormal Congestion Footprints
METHODOLOGY AND LEARNINGS
FROM CALCULATING THE COST
OF THE CAUSES OF CONGESTION
19. 19 |
Separating the causes of excessive congestion
Excessive
Congestion
Normal
Infrastructure
bottlenecks
Abnormal
Incidents Weather Roadworks
Special
events
26. Additional Opportunities Arising
• ‘Normal’ profiles could be used to:
– improve detector monitoring, improve incident detection
– input to traffic models, better understanding of what is ‘normal’ when
– calculate actual operational capacity of each link in real time & where
there is spare capacity
• Calculate impact of individual weather or incident events: cost, VKT
affected, VKT lost, actual start time & duration, etc. and save with
SIMS or 131940 record
• Improve traffic management methods by analysis of cost data to
target specific causes
• Visualisation of congestion events (see example)
27. The authors wish to acknowledge the support of
QLD Transport and Main Roads and thank Kelvin Marrett, Miranda Blogg
and Frans Dekker for their contributions to this project.
METHODOLOGY AND LEARNINGS
FROM CALCULATING THE COST
OF THE CAUSES OF CONGESTION
Include example of weather data with stats by months
In terms of our daily benchmarking, you can see on this slide the ave daily cost (y axis) for 2012 to 2015 (x axis)
you can see it fluctuates from $0.6 and 1.7M$ per day for weekdays, less than half of that on Saturdays and half of that again for Sundays
Note the lower cost for Mondays (green) and progressively builds up over the week
Also note the seasonal fluctuations
The slow uptrend is also quite evident
Other analysis of the data concludes that most of the cost is associated with passenger vehicles as opposed to other vehicle types
Also a substantial portion of the cost of congestion comes from the delay cost - more than 90 % - time that people could be spending doing other activities such as working
The rest of the costs are from fuel and emissions
I am now going to run you through step-by-step how we split normal and abnormal in our model
Produced a pie chart for Brisbane state controlled roads using 12 months in 2014
Showing you today a couple of examples to show you how we have developed the pie chart
Graph showing 12 months of data for 1 section of the network – Pacific motorway southbound before the Gateway merge at 5.30 – 5.45pm in the afternoon
Shows the number of days (y axis) that this section operates at various speeds (x axis) e.g. weekend and public holidays close to 100 km/h speed limit but since this is a bottleneck there are a substantial number of days with low speeds
Speed limit is 100 km/h - 70% of the speed limit is considered excessive – 70 km/h is considered the benchmark
Any speed drop below this is considered excessive and therefore included in our cost of congestion calculation
53 km is the overall average speed
Use this to define a range of normal
This cost is classified into normal recurring in the congestion pie chart
FHWA in the US produced a sources of congestion pie chart that has been referred to internationally for decades. The purpose of this pilot study was to replicate the analysis for Brisbane, produce a pie chart and develop a methodology to be implemented in Queensland and useful for decision-making at strategic, tactical and operational levels
Conducting a multi-variable analysis, building on the cost of excessive congestion model
Uses STREAMS data to build normal traffic profiles and separate abnormal from normal congestion
This normal or recurring congestion information informs our infrastructure investment choices
Then uses incident data from 131940 and weather data from Bureau of Meteorology (BoM) to identify the causes of abnormal
This information assists us to plan our response and manage network operations
As an example an incident occurs at this lower speed drop
Excessive congestion cost associated with this incident can be determined below the benchmark
When an incident occurs, you can’t blame all of the speed reduction on the incident
If the incident had not occurred, a vehicle would have experienced normal congestion and a speed decrease
The portion of this excessive congestion that is normal is categorised into recurring (in abnormal)
The rest less than normal is categorised into incidents
We now know what is causing congestion on our state-controlled roads in Brisbane
Most of Brisbane’s congestion is recurring (78%)
If we better manage abnormal events we can reduce cost of excessive congestion by up to $30.5M per annum
The majority of this is due incidents
The pie chart is showing us network-level results, however the analysis has been conducted on a link-by-link basis so we are able to drill down to more detailed answers as required
For example we could rank our incident related congestion hot spots, place our response vehicles closer to reduce response times thereby reducing congestion.
Opportunity to improve data for example, long term roadworks to reduce the unknown section of the pie
We are now automating the methodology into an intuitive system to provide to TMR decision-makers
This is where we will see our Congestion Management Approach and the Australian Infrastructure Plan recommendations realised.
This is where strategy hits the road.
This is where we will prepare ourselves for the rapidly changing transport sector and make smarter decisions to improve congestion for the benefit of the Queensland community.
If we better manage abnormal events we can reduce cost of excessive congestion by up to $30.5M
The majority of this is incidents
Many links with recurring congestion
Greater focus on quality of detector data as impact of bad data can be significant. We can filter bad data, but the less bad data the better.
There are limitations with the data for example we don’t pick up data when detectors have failed, volumes when vehicles are static or specific modes like buses
So therefore the proper maintenance of detectors is important for our data and information
We are working with ARRB and TransLink to develop a methodology to incorporate multi-modal components - on-road bus excessive congestion costs and costing reliability
We are also starting to work with emerging data sources for example, speed data from mobile phones and in-car navigation systems