Lecture-1: Introduction to Freight Transportation Demand ... · 100 110 120 130 140 150 160 170...

1

Lecture-1: Introduction to Freight

Transportation Demand Modeling

In Today’s Class

Focus is on methods and techniques for freight modelling Short rehearsal of general introduction on modelling We follow the 4 step model architecture – modified for freight

1. Introduction to freight demand issues2. Demand modelling principles

Production / attraction Trade Mode choice Route choice

& integrative forms

What is Globalization?3

(c) International Road Transport Union (IRU) 2010

Source: IRU

Road Transport has become a production tool!

What does it take to have a cup of coffee in a café?

The combined efforts of

29 companies in 18 countries

Q1: what’s behind growth?

Global Container Overslag

Q2: What’s behind structural change?

Growth in freight travel by land modes 2000-2050

0.0

2000.0

4000.0

6000.0

8000.0

10000.0

12000.0

14000.0

2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

ton

ne-k

m

OECD North America

OECD Europe

OECD Pacific

FSU

Eastern Europe

China

Other Asia

India

Middle East

Latin America

Africa

Q2: What’s behind structural change?

Growth in freight travel by land modes 2000-2050

0.0

2000.0

4000.0

6000.0

8000.0

10000.0

12000.0

14000.0

2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050

ton

ne-k

m

OECD North America

OECD Europe

OECD Pacific

FSU

Eastern Europe

China

Other Asia

India

Middle East

Latin America

Africa

100

110

120

130

140

150

160

170

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

ton

tonkm

vehkm

Autos/freight: fundamental differences

One decision maker or many?

Unit of transport = decision maker?

Many interactions between decision makers, or few?

Correspondence between demand and trips: simple or complex?

Heterogeneity in trip purposes: low or high?

NCFRP 25. Holguin Veras et al, 2010

Modelling freight transport demand in the context of public policy

micro: "the firm"

macro: "the region"

normative

“best situation"

descriptive

"as it happens"

OUR

FOCUS

e.g. supply chain design

transport demandforecasting

e.g.transport network design

disaggregatetransport demand models

Freight reorganization responses & their determinants

Producer 1 Consumer 1

Distribution Center

Producer 3

Producer 2

Consumer 2

Production reorganisation

Supply chain reorganisation

Transport reorganisation

Transshipment Hub

There is more than transport costs…

Groothedde, 2004

Gemiddeld aandeel kosten in verkoopprijs

48%

27%

15%

6%

4%

Fabricage kosten

Marketing kosten

Logistieke kosten

Transportkosten

Winst marge

Production costsMarketing costsOther logistics costsTransport costsProfit margin

INVENTORY

POLICY

INVENTORY

COSTS

TOTAL

FLOW

VALUE

DENSITY

INTEREST

RATES

WAREHOUSE

COSTS

PACKAGING

DENSITY

VOLUME TO

WEIGHT

HANDLING

RATES

STOCK

RATES

TRANSPORT

COSTS

TRANSPORT

DISTANCE

CONSIGNM.

SIZE

VALUE OF

SERVICE

TRANSPORT

SERVICES

PHYSICAL DISTRIBUTION COSTS

Total logistics costs: determinants

LOGISTICSFAMILIES

Goods types & critical cost fields

Transport

costs

Value

density

Packaging

density

Handling

costs

Inventory

costs

Ploos van Amstel, 2003

Production Location choice

Type of product

Production volume

Consumer choices

Trading contracts

Shipment sizes, frequencies

Location en volume of inventory

Distribution channels

Mode(s) of transport

DIY or Hire & Reward?

Long term or spot contract?

Means of transport

Routing and scheduling

Time of departure

Capacity planning

Producent / plant mgr.

Consumers

Sales managers

Sourcing managers

Logistics service provider

Logistics manager

Logistics manager

Transport manager

Forwarder

Transport planner

Driver

Decision maker Decision

“ship

per”

“carr

ier”

A layered model of logistics decisions

The conventional 4-stage model

Production/

Consumption

Matrix

(Trade)

pro

du

ctio

n

consumption

1. trip generation models

e.g. as function of GDP or floor area

2. trip distribution models

e.g. gravity type model

as function of

transport costs

origin-

destination

matrix

origin-

destination

matrix

Origin/

Destination

Matrices

by modemodal

split

3. mode choice models

e.g. as function of

transport costs

network

assignment

4. route choice models

e.g. as function of

transport costs

Model taxonomy: 4 stage and beyond

Production/

consumption

Distribution

Mode choice

Route choice

SCGE

Supernetwork

Inventory location

Hypernetwork

Intermediate conclusion

Freight changes caused by: Changes in the economy Changes in number of tons lifted Changes in the transport performance Changes in traffic performance

Supply chain considerations: logistics service & total logistics costs Transport Inventory Handling

4 step transport demand model needs to be extended to accommodate freight specific issues

Production/

consumption

Distribution

Mode choice

Route choice

SCGE

Supernetwork

Hypernetwork

Inventory location

Freight generation

0.0

5.0

10.0

15.0

20.0

25.0

30.0

Me

taal

pro

duc

ten

en m

ach

ines

Che

mie

Rub

ber

en

kun

stst

of

Bas

ism

eta

al

Ele

ctro

tech

nis

che

ind

ustri

e

Ove

rige

indu

stri

e

Voe

din

g e

n ge

not

Tran

spor

tmid

dele

n

Gro

otha

nde

l

ritt

en

/ ha

Simple freight trip generation models

Freight generation vs. freight trip generation

Increases with economic activity (business size, # of consumers)

Depends on sector/ goods type

Mostly simplified into linearmodel

y = 36.741x

0

500

1000

1500

2000

2500

0 10 20 30 40 50 60

bruto bedrijfsoppervlak

ritt

en

/d

ag

Zonnenberg, 1989

Klaver, 2001

Problem:Just

Regression

(why a problem?)

Trip generation & shipment size

Problem Ordering goods from manufacturer: what order size?EOQ - economic order quantity

Total costs = product costs + ordering costs + inventory costs Price (P) * demand (D) Ordering costs /unit (O) * # units (D/Q) Inventory cost / unit (I) * average inventory (Q/2)

TC = P*D + O*D/Q + I*Q/2; minimize for shipment size Q

Solution at OD/Q = IQ/2; Q* = √(2OD/I)OD/Q

IQ/2TC

QQ*

Effect of logistics on freight trip generation –or…?


Trip generation vs.production and consumption

Production Transportation Processing,

storage

Consumption

Shipper,

producer

Carrier

Distribution centers,

warehouses

Intermediate

consumer

End

consumer

To another

PC link

O1

O2

O3

O5

D2

D3

D5

D4


Production and consumption networks

Input/Output analysis allows us to trace demand effects through sectors as pulled by consumer demand (“final demand”)

=> I/O model with fixed relations

More realistic approach through flexible production functions

=> computable general equilibrium models

Price

Volume

Supply

Demand

Input-Output analysis: basic framework

(1) Total production t = Final demand y + Intermediate demand f(t)

(2) Intermediate demand = technical coefficient A * total production t

t = y+ At => t = y(I-A)-1

t = vector of total productiony = vector of final demandA = matrix of technical coefficients

All in Euro per year per sector

I/O origin = estimation of GDP for national accounts

Assumed fixed!

Production/

consumption

Distribution

Mode choice

Route choice

Spatial equilibrium

Supernetwork

Hypernetwork

Inventory location

Distribution models

Trade depends also on costs of interactionLogistics costs in EU and US

0

2

4

6

8

10

12

14

1985 1990 1995 2000 2005 2010

% o

f G

DP

EU

US

sources: ELA/ATKearney and CSCMP

Trade tariffs

0

5

10

15

20

25

30

35

40

45

1940 1950 1960 1970 1980 1990 2000

perc

en

t o

f im

po

rt v

alu

e

source: WTO

Costs of transport

0

50

100

150

200

250

300

350

400

1960 1970 1980 1990 2000

co

st

ind

ex,

1980 =

100 air

land

sea

European Commission, 2004

World GDP, trade and container transport 1990-2006

Understanding the gravity model (1)

Margin = Pricei – Pricej – Costsij

Price

in i

Margin

Region i Region j

Price in j

flow

Understanding the gravity model (2)

Mij = pj – pi - cij

Interregional trade Tij = T * Pr{ij}

Choice probabilty based on logit discrete choice modelPr{ij} = exp (Mij) / Σij exp(Mij)

Then:

Tij = exp(pj – pi - cij)* T/ Σij exp(Mij)

Replace T/ Σij exp(Mij) bij ζ (constant) for convenience

Tij = ζ *exp(-pi)*exp(pj)*exp(-cij) = ζ *Ai * Bj * exp(-cij)

M = margin, p = price,

c = costs of interaction, i & j: regions

T = total tradeA, B, ζ: constants

Economy/transport linkages

LUTI model Trip generation as simple regression function of accessibility Ti*g = f(Ars) ; e.g. 10% change in accessibility means 10% more

trips

Regional (quasi-) production function model Trip generation changes via changes in regional GDP

Ti*g = f(GDPis) and GDPis = f(Li Ki Ri Ais)

SCGE models Trip generation as result of general spatial price equilibrium Tijg = f(Li,j,s Ki,j,s Ri,j,s tijs) where

• Ti*g = Trips from i to all other regions, for good g• A= Accessibility of sector s in region i• L, K, R: Labor, Capital, Land• i,j = short for all ij and sector pairs

Evolution of spatial interaction models

Location of activities

Interaction between activities Intensity of economic activities

Spatial interactions

Sectoral interactions

Land Use models Gravity type models

Input/Output models

Equilibrium models

LUTI models: elasticities for regional trip generation

Multi-regional I/O models (MRIO): stepwise IO & gravity

Computable General Equilibrium models

Spatial Computable General Equilibrium models

Production/

consumption

Distribution

Mode choice

Route choice

Spatial equilibrium

Supernetwork

Hypernetwork

Inventory location

Spatial equilibrium models

Production, consumption and trade combined: spatial general equilibrium

Factor

costs

region A

Factor

costs

region B

Barriers to trade

NEW

INTERREGIONAL

EQUILIBRIUM

Costs of trade

Towards an integrated system model for passengers and freight

Generalized transport costs

migration

commuting

labour marketproduction

income

consumption

trade

Product varieties

Intermediate conclusion

Production and consumption: from I/O to production functions

Spatial interaction well described by the gravity model

Gravity model can be replaced by disaggregate logit

I/O & gravity: MRIO (multiregional IO) models Land Use Transport Interaction models Linkage with production functions: spatial general equilibrium

Also link to integration with passenger transport modelling

Production/

consumption

Distribution

Mode choice

Route choice

SCGE

Supernetwork

Hypernetwork

Inventory location

Modelling inventory chains

Positioning logistics within the 4 step model

Production and Consumption

Trade (Sales and Sourcing)

Logistics Services

Transportation Services

Network Services

I. Trade/

Economy

linkages

II. Supply chains

III. Multimodal

networks

Inventories affect spatial flow patterns

Shipment

size

transport

costs

scale

P/C vs. O/D:inventories

High frequency,

Small

shipments

Low frequency,

Large

shipments

Inventory

ConsumerProducer

level of

centralisation

total costs

service orientation cost orientation

transport costsdepot costs

Frequency

shipment

size

time

Cycle

stock

Mechanism 1: economies of scale

Shipment size

Transport

Costs / ton

Vehicle size

FTL

1 Euro/tripkm

Mechanism 2: inventory policy

Ordering frequency

Shipment

size

time

Cycle

stock

Delivery

Demand

Supply chain changes in Europe

level of

centralisation

total costs

service orientation cost orientation

transport costsdepot costs

Azerbaijan

CENTRALITY

Production/

consumption

Distribution

Mode choice

Route choice

SCGE

Supernetwork

Hypernetwork

Inventory location

Mode choice models

Mode choice: some stats

0% 20% 40% 60% 80% 100%

road (18 bn

tonnes)

rail (1,5 bn

tonnes)

waterways (0,5

bn tonnes)

Agricultural products and live

animals Foodstuffs & animal fodder

Solid mineral fuels

Petroleum products

Ores & metal waste

Metal products

Crude & manuf. minerals, building

materials Fertilisers

Chemicals

Machinery, transp. equipment,

manuf. & misc. articles

Mode choice: some stats

0% 20% 40% 60% 80% 100%

road (18 bn

tonnes)

rail (1,5 bn

tonnes)

waterways (0,5

bn tonnes)

Agricultural products and live

animals Foodstuffs & animal fodder

Solid mineral fuels

Petroleum products

Ores & metal waste

Metal products

Crude & manuf. minerals, building

materials Fertilisers

Chemicals

Machinery, transp. equipment,

manuf. & misc. articles 0% 10% 20% 30% 40% 50%

Mach., transp. eq, manuf. & misc. art.

Chemicals

Fertilisers

Crude & man. minerals, build. mat.

Metal products

Ores & metal w aste

Petroleum products

Solid mineral fuels

Foodstuffs & animal fodder

Agricultural prod. & live animals

tonnes

tonkm

Mode choice models

Mode attributes (=> which ones?)

Commodity attributes (=> which ones?)

Behavioural models

Discrete choice models

Total logistics costs based

Include inventory costs

In transport

At shipper

Q: with all these “-”scores, how come road is so popular?

Mode choice modelling approaches

Inventory (cost based, all-or-nothing) models

Behavioural models:Minimize out-of-pocket costs (K) utility maximization U = -K

V = Km + α Tm

Probabilistic approach discrete choice

Um = Km + α Tm + ε

Deterministic choice & random preferences

Um = Km + α Tm

Um = Km + α Tm + ε

Disaggregate vs. aggregate models

sampling

measurement

estimation

aggregation

measurement

estimation

Firm level Statistics

Statistics

specification

Operational approaches depend on data used

Aggregate Data Land use data Trade statistics Transport

statistics Time and costs

Disaggregate data

• Company surveys

• Shipment records

• Goods attributes

• Time and cost data

Combine

• Split aggregate flow

data using firm size

onto firm level

Or

• Make aggregate

distributions of goods’

attributes

Choice

model

- Mode

- Shipment size

- Inventories

- Routing

Transport costs

Copyright © 1998-2010, Dr. Jean-Paul Rodrigue, Dept. of Global Studies & Geography, Hofstra University. For personal or classroom use ONLY. This material (including graphics) is not public domain and cannot be published, in whole or in part, in ANY form (printed or electronic) and on any media without consent. This includes conference presentations. Permission MUST be requested prior to use.

Distance

Tran

spo

rt c

ost

s p

er u

nit

D1 D2

CRail

D1: break-even point Road/Rail D2: break-even point Rail/Inland navigation

Tx: terminal costs (load/unload costs) Cx: cost function mode x

CRoad

CInlandNav

TRoad

TRail

TInlandNav

Trade-offs through value of service

Umg = αg * Tm + Pm + εm

VOT

PDF

wa

ter b

ike

sh

ip

railw

ays

truck

air c

arg

o

efficient frontier

ship

rail

water bike

car

raft

air cargo

time

price

Value of Time α

helicopter

zeppelin

(Dial, 1972)

raft

Typical VOT switching values between modes

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.000 0.020 0.040 0.060 0.080 0.100 0.120

transport time (inverse speed)

tra

ns

po

rt t

ari

ff

AIR

ROAD

RAILWATERWAYS SEA

0.6 $/ton/hr

0.3 $/ton/hr

1 $/ton/hr

15 $/ton/hr

Determination of market shares

tijd

prijs

1

2

α1

α2

α4

α3

Um = Km + αgTm

α1 α2 α3 α4

Ways to measure the value of time

Accounts based Factor costs or market prices

Behavioural analysis (experimental) Aggregate vs. Disaggregate Revealed and Stated Preferences Between mode or Within mode choice experiments Discrete choice modelling in trade-off situations various alternative choice models other choice situations than mode choice possible

Disaggregate measurements: sampling and aggregation Aggregate approach: based on statistics

Production/

consumption

Distribution

Mode choice

Route choice

SCGE

Supernetwork

Hypernetwork

Inventory location

Route choice models

Route choice for freight

Most freight models apply similar route choice techniques as in passenger transport (e.g. Dijkstra alogorithm)

Specific concerns for freight:

Road: round trips (TSP); restrictions: weight & size regulations

Rail: train paths; restrictions: gauge width; voltage; priorities

Waterways: waterways sizes & ship classes

Sea: shipping line & feeder services; restrictions: port depth

Air: hub & spoke networks; flight level 0 (trucking)

Dynamics in efficiency

A note on the degree of loading

Survey A10-20/RN10 (F): volume and weight of equal importance

(Combes, Univ Paris-Est, 2010)

26%

38%

26%

10%

74%

LTL / empty FTL (m3) FTL (m3 + ton) FTL (ton)

Simple route vs. round trips

Base

1

23

4

5Loaded vehicle-trip

Commodity flow

Notation:

Consumer of cargo

(receiver)

Empty vehicle-trip


Transport reorganization: routing

city A

city B

LTL (1)

LTL (2)

stops

CASE A

city A

city B

LTL (1)

LTL (2)

stops

CASE A

1st tier round trip2nd tier round trip

city A

city B

FTL

1st tier stops

LTL(1)

LTL(2)

CASE B

1st tier round trip2nd tier round trip

city A

city B

FTL

1st tier stops

LTL(1)

LTL(2)

CASE B

Models for empty trips

Classic model Noortman & Van Es (1978)

Empty trips (i,j) ~ laden trips (j,i)

laden trips (i,j) = mij/a

empty trips (i,j) = p*mji/a

But this leads to differences in # of trucks moving (i,j) and (j,i)

(why? why is this a problem?)

Alternative formulation (Hautzinger, 1984)

where:

mij = goods flow (tons)

aij = avg. load (tons/truck)

zij = total flow

p = constant

where:

pij = probability of vehicle from i

to return empty from j

and pij = f(mjij/mij, dij)

Extensions for trip chain models

Production/

consumption

Distribution

Mode choice

Route choice

SCGE

Supernetwork

Hypernetwork

Inventory location

Supernetworks

Combined mode & route choice

Synchromodal network services with dryports & extended gates

waterways

rail

truck

time

price

VOT

PD

wa

terw

ays

railw

ays

truc

k

Waterways

only

Rail only

time

price

VOT

PD

Truck only

zeppelin

T1

T2

T3

T4

T5T6

T1-T6: Routes alongTerminals 1-6

W R T

T7

T8

T9

Production/

consumption

Distribution

Mode choice

Route choice

SCGE

Supernetwork

Hypernetwork

Inventory location

Hypernetwork models (briefly)

Combining choices in hypernetworks

Consumer

Large shipment size networks

Small shipment size networks

Manufacturer

Stock Points:Network SwitchRoute I

Route II

On hypernetwork models

Advantages

Elegant & simple method (all in one)

Close to physical representation, increases first sight acceptance

Behavioural principles aligned between subproblems

But…

Complexity is high; longer calculation times

Difficult to find one good parameter setting for 3 choice problems

Difficult to calibrate due to many degrees of freedom

Concluding remarks

A brief introduction into freight demand modelling

Focus was on descriptive, static, deterministic, aggregate models

Further studying: Static vs. dynamic models

Equilibrium vs. disequilibrium approaches

Supply models: capacity, scale economies, prices and service levels

Aggregate (region) vs. disaggregate (firm level) models

Accounting for heterogeneity and uncertainty

Reading material for exam on the blackboard Introduction in Willumsen & Ortuzar

Review papers

Production Location choice

Type of product

Production volume

Consumer choices

Trading contracts

Shipment sizes, frequencies

Location en volume of inventory

Distribution channels

Mode(s) of transport

DIY or Hire & Reward?

Long term or spot contract?

Means of transport

Routing and scheduling

Time of departure

Capacity planning

Producent / plant mgr.

Consumers

Sales managers

Sourcing managers

Logistics service provider

Logistics manager

Logistics manager

Transport manager

Forwarder

Transport planner

Driver

Decision maker Decision

“ship

per”

“carr

ier”

Summary, questions

Acknowledgement

Prof. Lóránt Tavasszy (TU Delft)

Prof. Frank Southworth (Georgia Tech)

Lecture-1: Introduction to Freight Transportation Demand ... · 100 110 120 130 140 150 160 170...

Documents

Transcript of Lecture-1: Introduction to Freight Transportation Demand ... · 100 110 120 130 140 150 160 170...