Forecasting Assumptions with the Wellington Transport Strategy Model

Forecasting Assumptions
with the Wellington
Transport Strategy Model
NZMUGS 2014
Geoffrey Cornelis (GWRC)
Dan Jones (freelance)
Forecasting Assumptions with WTSM
 What are the main input/assumptions
when forecasting with a 4-step strategic
transport model?
 What is their effect on travel demand
(especially PT)?
 How can they be used for forecasting?
Analysis of PT growth in WTSM Forecasts
Why this analysis?
• WTSM baseline forecasts showed
decrease in PT patronage after 2021, at
odds with GWRC policies and global
trends
• Limited response from the model to PT
infrastructure improvements (e.g. PTSS)
WTSM Car and PT Trips Growth
Indexed to 2011
1.20
1.10
1.00
Car
PT
0.90
0.80
2011
2021
2031
2041
 Prior to modelling for Regional Land Transport Plan, important
to understand main drivers for and against PT growth.
Wellington Transport Strategy Model (WTSM)
Fixed equations within
model (internal)
Input into the model (external)
Landuse
Car Ownership
Trip Generation
Economic Parameters
Trip Distribution
Mode Choice
Assignment
•
•
•
•
Vehicle Operating Cost
PT Fares
Parking Cost
Value of Time
Road and
PT networks
When forecasting: internal equations are fixed,
external input based on assumptions about future.
Economic Parameters Forecast Assumptions
• Vehicle operating costs: MED forecasts for fuel price and MoT Fleet
Emission Model for vehicle efficiency.
• Value of time: based on real GDP/capita forecasts from NZ Treasury,
with elasticity of 1 for work and 0.8 non-work purposes.
•
Parking costs: based on real GDP/capita forecasts.
• PT fares: based on real GDP/capita forecasts with elasticity
of 0.25.
Analysis of PT growth in WTSM Forecasts
Annual PT Patronage 2011-2021
Pax
(x1000)
Pax (x1000)
40
+1%
30
30.0
20
-5%
-4%
+15%
+11%
35.0
-1%
• VoT increase eases the pain of
rising costs, and car ownership
rises.
• Impact of network changes
limited but favourable to car.
10
0
40
Pax (x1000)
30
20
10
0
• Demographic drives total trip
growth (+increase in
employment)
• Increasing fuel and parking cost
leads to more PT trips.
Annual PT Patronage 2021-2031
+6%
35.0
+2%
+1%
-5%
-5%
-3%
33.4
• Less demographic-related
growth
• Fuel and parking price increases
have moderated.
• VoT and car ownership continue
to rise.
• Full RoNS impact
on PT usage.
Main Findings
• WTSM is operating as expected!
• Natural ‘drift’ to more car and less PT use in WTSM due to
assumptions and Wellington context:
- Moderate demographic growth  moderate increase in road
congestion.
- Significant additional road capacity – RoNS
- Most exogenous drivers (GDP/income growth, car ownership)
favourable to car use.
 Business as usual in WTSM = more car use
 PT share can increase through infrastructure improvements, but
pricing and policy levers have more impact.
 Economic input variables are the main tools to test
these policy and measures in WTSM.
Application for RLTP
• Regional Land Transport Plan: sets out the region
land transport’s objective, policies and measures for
the next 10 years. WTSM used to help define
achievable targets.
• 2031 ‘Expected Future’ main scenario, includes:
- More PT improvements (BRT, integrated ticketing)
-
Car ownership revised down, based on census 2013
• Range of sensitivity tests for analysis of more voluntary
policies: parking charge, congestion charge, PT fare
discount, etc.
• Same incremental approach used to understand
impact of all parameters
 better communication of results to stakeholders
Looking back... WTSM 2001
• 2001 original version of current WTSM
• Look at the then 2011 forecast, compare with
2001 – 2011 observed trends, how different?
• Economic variables not forecasted
 relative weight of all costs assumed to be
constant in real term
2001-2011 real increase
80%
60%
40%
20%
0%
Fuel
VoT
Parking PT Fares
• Can difference be explained by real increase in economic
variables?
2001 to 2011 Trends – Observed vs Modelled
35%
PT Trips annual
veh-km annual
25%
15%
5%
-5%
40%
30%
Rail boardings annual
20%
AADT SH2 North of Ngauranga
AADT SH1 Newlands
Bus boardings annual
AADT SH1 Terrace Tunnel
10%
20%
10%
0%
0%
-10%
 Model replicates trends better when factoring in real
changes in economic variables.
2001 – 2011 Observed Trends
1.30
Population
Population
1.25
Employment
1.20
GDP
1.15
Annual VKT
1.10
Car ownership
1.05
Bus Pax
1.00
Rail Pax
2001
2006
2011
• Population increases but other indicators flatten out
after 2006.
But with 2011 actual landuse...
35%
•
Population actually grew
faster than forecasted.
PT Trips annual
veh-km annual
25%
15%
•
How do results change with
actual change in landuse?
5%
-5%
40%
30%
20%
Rail boardings annual
20%
Bus boardings annual
AADT SH2 North of Ngauranga
AADT SH1 Newlands
AADT SH1 Terrace Tunnel
10%
10%
0%
0%
-10%
 Changes in behaviours, can not be replicated solely by
varying input parameters.
Conclusion
• Economic/policy assumptions have strong impact on future demand
and mode share. Analysis of past forecast shows that impact seems
realistic.
• In the past, changes in exogenous variables have driven the trends.
 Forecasting assumptions and input must be carefully considered and
their impact understood.
• Now, changes in lifestyle and tastes start to impact on travel behaviour
(peak car? suits on the bus?)  More uncertainty in forecasting.
• Uncertainty is lost if based on a single view of the future  Importance
of scenario planning, using a wider range of assumptions.
• Varying assumptions allows move from simply forecasting to more
aspirational futures and policies  From “predict and provide” to
“desirable equilibrium”.