integrating transportation network and regional economic models to estimate the costs of a large...
TRANSCRIPT
INTEGRATING TRANSPORTATION NETWORK AND REGIONALECONOMIC MODELS TO ESTIMATE THE COSTS OF
A LARGE EARTHQUAKE
Sungbin Cho, Research Associate, School of Policy, Planning, & DevelopmentPeter Gordon, Professor of Policy, Planning, & Development -and- Economics
James E. Moore II, Associate Professor of Civil Engineering -and- Policy, Planning, &Development; -and- Visiting Scholar, California Research Bureau, California State Library
Harry W. Richardson, Professor of Policy, Planning, & Development -and- EconomicsMasanobu Shinozuka, Fred Champion Professor of Civil Engineering
University of Southern California
Stephanie E. Chang, Research Assistant Professor of Geography
University of Washington -and- EQE, International
with the assistance of
Seongkil Cho, Kyoung Lee, Shin Lee, Junghoon Ki, and Gang YuSchool of Policy, Planning, & Development
Shashank Agrawal, Anupam Bordia, Xue Dong, Yue Yue Fan,Uma Kompella, and Heng Liu
Department of Civil & Environmental Engineering
University of Southern California
School of Policy, Planning, & Department of Civil &Development Environmental Engineering
Ralph and Goldy Lewis Hall, 330, MC-0626 Kaprielian Hall, 210, MC-2531
University of Southern CaliforniaLos Angeles, California 90089-0626
Volume II of a Final Report to the National Science Foundation,Award CMS 9633386 (EHM)
ABSTRACTWe summarize an integrated, operational model of losses due to earthquake
impacts on transportation and industrial capacity, and how these losses affect themetropolitan economy. The procedure advances the information provided bytransportation and activity system analysis techniques in ways that help capture the mostimportant economic implications of earthquakes. Network costs and origin-destinationrequirements are modeled endogenously and consistently. Indirect and induced lossesassociated with direct impacts on transportation and industrial capacity are distributedacross zones and economic sectors. Preliminary results are summarized for a magnitude7.1 earthquake on the Elysian Park blind thrust fault in Los Angeles. Work on incidenceand distributional consequences continues, including evaluation of post-eventreinvestments.
August, 1999
NSF Final Report CMS 9633386
1
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
INTRODUCTION
The seismic sensitivity of the Los Angeles metropolitan region is well known.
Much of the previous social science based research on earthquakes has focused on
measuring the total economic impacts of structure and contents damage and, more
recently, of business interruption.
The core purpose of applied research in the earthquake field is to assist policy
makers. Three research questions motivate this work. First, we want to integrate regional
economic, transportation, bridge performance, and other structural response models in a
way that respects feedback relationships between land use and transportation. Second, we
want to apply such integrated, operational models to the problem of estimating the full
costs of a large earthquake, and the benefits of proposed mitigation measures. Further,
because "all politics are local," we want to analyze these costs and benefits at the
submetropolitan level.
The large expenditures that are involved in many proposed mitigation programs
suggest that a careful analysis of trade-offs is required. This means that the full costs and
benefits of each mitigation should be studied. The benefits provided by earthquake
mitigation measures consist of the costs avoided if the measures are applied (Gordon,
Richardson, and Davis 1998). Thus a discussion of costs avoided depends on analysts'
ability to determine full costs, but what are the full costs of an earthquake? We cannot
know unless all of the interactions between infrastructure and the economy are
understood. Our own work on the business interruption effects of the 1994 Northridge
earthquake indicates that an exclusive focus on structural damage ignores 25-30 percent
NSF Final Report CMS 9633386
2
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
of the full costs (Gordon, Richardson and Davis 1996, 1998). In 1994, business
interruption job losses were estimated to be 69,000 person-years of employment. About
half of these were outside the area that experienced structural damage (see Table 1).
Disregarding values of such magnitude results in a serious underestimate of the full costs.
Social science research can, therefore, make a substantial contribution by identifying
expected full costs with and without various proposed mitigations.
This research is an effort to trace the effects of an earthquake on the Los Angeles
economy, including its impact on the transportation services delivered by the highway
network. To do this, we must develop an innovative, integrated framework and
methodologies for evaluating the effects of earthquakes on the services delivered by the
transportation network.
To estimate these impacts, we must integrate:
• bridge and other structure performance models,
• transportation network models,
• spatial allocation models, and
• inter-industry (input-output) models.
Highway transportation systems are complex structural systems with many
components. Bridge performance models have been investigated widely. Bridges are only
one kind of transportation structure, but they are also typically among the most complex.
Their seismic performance is difficult to model because it is a function of design,
component performance, and local factors such as peak ground acceleration (PGA) during
an earthquake (see Figure 1 [Shinozuka ,1998] and Volume I).
NSF Final Report CMS 9633386
3
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Work to support this effort includes classifying bridges by local soil type,
characterizing ground motion with pseudo-absolute acceleration response spectra for
NSF Final Report CMS 9633386
4
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 1: Business Interruption Losses from the 1994 Northridge Earthquake(Jobs in person-years, Output in thousands of 1994 $)
Direct Indirect and Induced TotalArea
Jobs Output Jobs Output Jobs Output
Impact Zone Total 34605.4 3,117,528 1,904.9 209,591.1 36,510.1 3,327.119.4
Rest of LA City 2,119.9 232,021.2 2,119.9 232,021.2
Rest of LA County 10,668.2 1,067914.1 10,668.2 1,067914.1
Rest of Region 8,260.7 877,532.0 8,260.7 877,532.0
Regional Total 34,605.4 3,117,528 22,953.7 2,387,058.5 57,559.1 5,504,586.9
Rest of the World 11,454.4 1,031,901.9 Not Computable 11,454.4 1,031,901.9
Total 46,059.8 4,149,430.3 22,953.7 2,387,058.5 69,013.5 6,536,488.8
Source: Gordon, Richardson, and Davis 1996.
Retaining Walls Bridges Connectors
and RampsPavement
and Foundation
Embankment TunnelsAdministration
and Control Buildings
Decks
Permanent Ground
Displacement
Highway Transportation System
Piers and Foundation
Bearings and Joints
Transient Ground Motion = and gate
Figure 1: Design and Environmental Factors Affecting the Seismic Performance ofHighway Bridges
NSF Final Report CMS 9633386
5
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
different soil types, modeling bridge performance by predicting the probability of
exceeding a given damage state as a function of peak ground acceleration (PGA), and
calibrating these models against ground motions and structural failures recorded during
the Northridge earthquake.
The Five-County Area
The Los Angeles metropolitan area is most frequently described as the five-county
area that surrounds Los Angeles. These five counties are Los Angeles, Orange, Riverside,
San Bernardino and Ventura. The population of this area was 15,608,886 in July 1997
(US Bureau of the Census 1998). Intercensal (1980-90) population growth for the five-
county area was 26.4 percent. The core county, Los Angeles, had 9,145,219 inhabitants
in 1997. The five-counties included 5,165,900 households as of January 1997.
The five-county area extends to the Arizona border and covers more than 35,000
square miles, much of it uninhabited. The Bureau of the Census provides some data for
functional Urbanized Areas for Census years. In 1990, the urbanized portion of the five-
county area was 1,965.7 square miles, less than 6 percent of the region’s land area. The
population of this area was 11.403 million. Its gross population density was 5,801 people
per square mile, the highest in the U.S. The same area included 4.069 million housing
units.
There were 347,903 nonfarm business establishments in the five-county area as of
1995. Total personal income in 1994 was $329.6 billion. Per capita personal income was
$21,542. California’s total personal income in the same year was $702.6 billion. If the
NSF Final Report CMS 9633386
6
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
same proportion is applied to California’s 1994 gross state product of $875.7 billion, the
local area’s equivalent of GDP would be $410.8 billion. To put the Southern Californian
economy in some perspective, the 1994 GNP (a slightly different measure) of Canada was
$530 billion (US Bureau of the Census 1997). Southern California’s personal income
consists of 21.6 percent from goods-related activities and 78.0 percent from service-
related activities. The distribution across the major economic sectors was
• 16.4 percent manufacturing,
• 9.6 percent retail,
• 7.3 percent, finance, insurance, and real estate (FIRE),
• 34.3 percent services, and
• 13.3 percent government (US Bureau of the Census 1998).
The area’s civilian labor force in 1996 was 7,454,300, of which 4,415,400 were
based in Los Angeles County. Nonfarm employment in the same year was 6,033,400,
1,214,700 in goods-related work and 4,818,900 in service-related work. Average annual
pay across all sectors was $31,897 in the same year (US Bureau of the Census 1998). See
Appendix A for further details, including regional trade flows.
Previous Research
Reporting less than two months after the Northridge earthquake, Kimbell and
Bolton (1994) relied upon a "historical analogies approach." The nature of that approach
is not made clear in their report except for the fact that they used data on the effects of
prior earthquakes and disasters, i.e. the Loma Prieta Prieta quake, the Whittier quake, the
NSF Final Report CMS 9633386
7
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Oakland fires, and the Los Angeles riots. They find immediate job losses for Los Angeles
County of 29,300 with an additional 6,400 jobs lost outside the County. The authors
report net positive impacts because of reconstruction later in 1994 but there is,
nevertheless, a long-term negative impact of 18,500 jobs lost.
Using a survey approach, Boarnet (1995) sought information on the impacts of
freeway damage from the Northridge earthquake. He found that 43 percent of all firms
reporting any losses mentioned that some of these were because of transportation
problems. Eguchi et al (1996) report on their application of EQE International's Early
Post-Earthquake Damage Assessment Tool (EPEDAT), a GIS-based model, to the
problem of estimating Northridge losses. They calculate that these were in excess of $44
billion.
Chang (1995) introduces multivariate techniques for post-event assessments of
lifeline-related losses vs. those resulting from non-lifeline factors. She applies these
methods to an assessment of the economic effects of lifeline disruptions in the Hanshin
earthquake. Railroad capacity losses were higher than highway losses.
Rose and Benavides (1998) also applied inter-industry models as a method of
measuring regional economic impact analysis. The authors trace and record all the inter-
sectoral ripple effects associated with the full impacts of electricity disruptions expected
from a hypothetical 7.5 M earthquake in the Memphis area. A seven percent loss of gross
regional product over the first 15 weeks after the event was forecast. Rose and Lim
(1997) applied the same model to an analysis of the Northridge earthquake's effects.
NSF Final Report CMS 9633386
8
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Cochrane (1997) elaborates the nature of indirect economic damages, including
problems with backward and forward linkages. He also points out that the receipt of
disaster assistance matters in a full accounting of regional impacts -- even though these are
simply transfers within the larger national context. In addition, any resulting indebtedness
merely shifts the burden to future generations. Cochrane also introduces the NIBS
(National Institute of Building Standards) model to account for net regional losses and
gains after all transfer payments and possible debt payments have been included. Among
other results, he finds that indirect (non-structure) losses are inversely proportional to the
size of the sector shocked.
Okuyama and his colleagues (1997) develop a closed interregional input-output
model that emphasizes distributional effects. The approach is also sequential and
applicable to earthquake-type events where there may be drastic quarter-to-quarter
changes in demand and capacity. The model is applied to the Kobe earthquake. Four
types of model coefficients are manipulated to simulate the disaster.
Boyce (1998) suggests in a working paper how the Leontief and Strout (1963)
multiregional input-output model can be combined with Wilson's (1970) entropy function.
Boyce suggests an approach that focuses on the incentives to return an interregional
(substate) transportation system to pre-earthquake conditions. He discusses a procedure
for matching post-earthquake flows to pre-earthquake flows as a mechanism for imputing
changes in final demands.
As this brief summary indicates, there has been limited attention given to the
socioeconomic impacts of earthquakes. Most of the research on earthquakes has been in
NSF Final Report CMS 9633386
9
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
the engineering and geological fields. Progress in economic impact research is more
recent. Earthquake engineering is a challenging field, but exploring and integrating the
economic impacts of earthquakes with engineering models is especially challenging.
Impact Models and the Southern California Planning Model 1 (SCPM1)
The most widely used models of regional economic impacts are versions of inter-
industry models. These attempt to trace all intra- and interregional shipments, usually at a
high level of industrial disaggregation. Being demand driven, they only account for losses
via backward linkages. The Southern California Planning Model-I (SCPM1) was
developed for the five-county Los Angeles metropolitan region, and has the unique
capability to allocate all impacts, in terms of jobs or the dollar value of output, to 308 sub-
regional zones, mainly municipalities but also including unincorporated areas.
Previous work on the Northridge earthquake utilized SCPM1. In that case, the
model was driven by reduced demands on the part of damaged businesses, as ascertained
via a survey of businesses. In the present work, we focus on a hypothetical earthquake, a
M 7.1 event on the Elysian Park blind thrust fault. In this case, results of structure
damage to businesses, as developed by EQE's EPEDAT model, are used to drive a version
of SCPM that has been elaborated to include the regional transportation network.
EPEDAT predicts, among other impacts, the length of time for which firms throughout
the region will be non-operational. This allows the calculation of exogenously prompted
reductions in demand by these businesses. These are introduced into the inter-industry
model as declines in final demand. Simultaneously, explicit treatment of the transportation
NSF Final Report CMS 9633386
10
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
network makes it possible to model the impact of transportation cost changes on the
activity system, including reductions in regional network capacity resulting from large
numbers of bridge failures.
APPROACH
Figure 2 summarizes our approach. Implementing this approach is a data intensive
effort. We have available
• baseline 1990 census data describing the spatial distribution (by traffic analysis
zone) of residences and workplaces by major economic sector;
• a 515 sector Regional Science Research Institute (RSRI) input-output model
(Stevens, Treyz, and Lahr 1983) of the Los Angeles metropolitan (five-county)
economy based on 1992 economic census data (U.S. Bureau of the Census
1993a);
• the 1994 transportation planning network based on data from the Southern
California Association of Governments (SCAG) and California Department of
Transportation (Caltrans) Headquarters;
• 1991 SCAG Origin-Destination Survey data for 1,527 traffic analysis zones
(Southern California Association of Governments 1993);
• interregional and international trade flows from a variety of sources, including
studies by DRI-McGraw (1994, 1995); software provided to Caltrans by Booz,
Allen and Hamilton (1996), Caltrans (1996); the Southern California Association
of Governments (1992, 1993, 1994), and the Port of Los Angeles; and
NSF Final Report CMS 9633386
11
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
• a spatial allocation model of the same area, the Southern California Planning
Model (SCPM1; Richardson, Gordon, Jun, and Kim 1993) that has been used to
distribute aggregate input-output results (in $ and/or jobs) over urban space.
NSF Final Report CMS 9633386
12
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Spa
tial I
nter
actio
n (G
ravi
ty) M
odel
s
Gen
eral
ized
C
ost o
f Tra
vel,
Link
Flo
ws
Dem
and
for
Des
tinat
ions
(In
terz
onal
Fl
ows)
Inpu
t-Out
put I
dent
ity:
Dire
ct O
utpu
t =
Fina
l Dem
and
Clo
sed
Reg
iona
l In
put-O
utpu
t Mod
el
To
tal
Out
put
(Clo
sed
Loss
)
Ope
n R
egio
nal
Inpu
t-Out
put M
odel
Fi
nal D
eman
d
D
irect
O
utpu
t (L
oss)
Indu
ced
Out
put
(Los
s)
To
tal O
utpu
t (O
pen)
Bas
elin
e A
ctiv
ity a
nd
Tran
spor
tatio
n S
yste
m
Li
nk
Cap
aciti
es
In
dire
ct
Out
put
(Los
s)
Sou
ther
n C
alifo
rnia
P
lann
ing
Mod
el v
1.1
Spa
tial A
lloca
tions
of
O
utpu
t (Lo
ss)
Hou
seho
ld
Con
sum
ptio
n Fu
nctio
ns
Land
U
seN
etw
ork
Brid
ge
Inve
ntor
y
Trip
End
sR
egio
nal I
nput
- O
utpu
t Tab
le
Loss
of I
ndus
trial
Fu
nctio
n C
urve
sLo
sses
of I
ndus
trial
C
apac
ities
Net
wor
k E
quili
briu
m
Mod
el (B
asel
ine)
Ely
sian
Par
k E
arth
quak
eSce
nario
Ear
thqu
ake
Sce
nario
1
Ear
thqu
ake
Sce
nario
N
Ear
thqu
ake
Sce
nario
N
-1E
arth
quak
e S
cena
rio
Num
eric
al In
tera
tion
Cyl
cles
U
sed
to C
ompu
te S
imul
tane
ous
Eco
nom
ic E
quili
bria
Dat
a
Com
puta
tion
D
irect
Indu
stria
l Out
put
O
rigin
s an
d D
estin
atio
ns
Bas
elin
e Li
nk C
osts
Par
amet
eriz
ed D
ista
nce
Dec
ay F
unct
ions
Cal
ibra
ted
Spa
tial
Inte
ract
ion
Mod
els
Gen
eral
ized
C
ost o
f Tra
vel,
Link
Flo
ws
Dem
and
for
Des
tinat
ions
(In
terz
onal
Fl
ows)
Net
wor
k E
quili
briu
m
Mod
el (D
amag
ed)
O
rigin
s an
d D
estin
atio
ns
Itera
tion
Itera
tion
Sub
- Ite
ratio
n
JHW
, JH
S, F
Brid
ge P
erfo
rman
ce
Mod
els
(Fra
gilit
y C
urve
s)E
PE
DA
TLo
sses
of
Brid
ges
Mon
te C
arlo
S
imul
atio
n
Figure 2: Summary of the Southern California Planning Model 2 (SCPM2)
NSF Final Report CMS 9633386
13
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
The Southern California Planning Model 1 aggregates SCAG’s 1,527 traffic
analysis zones into 308 political jurisdictions, and aggregates the 515 sectors represented
in the RSRI model to 17. Each sector uses the other sectors' outputs as inputs, and each
sector supplies the other sectors with inputs. Knowledge of inter-industry transactions
also makes it possible to calculate corresponding supply driven input-output coefficients
(Miller and Blair 1985; Shoven and Whalley 1992). Both sets of coefficients are used for
freight accounting purposes. However, input-output results are computed using the
demand-driven model.
This research extends the Southern California Planning Model 1 in an important
way by treating the transportation network explicitly, endogenizing otherwise exogenous
matrices describing the travel behavior of households, achieving consistency across
network costs and origin-destination requirements, and improving the allocation of
indirect and induced economic losses over zones in response to direct earthquake losses to
both industrial and transportation capacity. Making distance decay relationships and
congestion endogenous also endogenizes the spatial allocation of indirect and induced
economic losses.
Establishing a Baseline
Our goal is to model the effects of earthquakes on industrial capacity and system-
wide transportation demand and supply. We also aim to measure as fully as possible the
economic impacts associated with both of these effects. Our first step is to compute a
NSF Final Report CMS 9633386
14
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
pre-earthquake baseline that is consistent with respect to equilibrium network costs,
network flows, and inter-zonal flows and origin-destination requirements.
SCPM1 includes work and shopping (including service) trips, but ignores other
non-work travel and freight flows. The SCAG origin-destination data include work and
non-work trips, but neglects freight flows. We map the five-county, 1,527-zone SCAG
transportation network to a five-county, 308-zone SCPM activity system. This
compresses the scaled inter-zonal flows associated with the regional transportation
network to a more manageable level among the 308 SCPM zones.
Each element in the SCPM1 journey-from home to-work (JHW) matrix describes
the proportion of workers residing in zone i who work in zone j relative to the total
employment in zone j. Each element of the SCPM1 journey-from home to-shop [service]
(JHS) matrix describes the proportion of residents in zone i who shop (and purchase
services) in zone j relative to total to the total number of residents in zone j. The SCPM1
JHW matrix is based on spatial distributions extracted from 1990 census data. The
SCPM1 version of the JHS matrix is the result of a gravity model estimation. In the
SCPM2 extension developed in this research, the elements of the JHW and JHS matrices
are endogenized as a simultaneous function of network costs and estimated gravity model
parameters.
Some of the model's 17 economic sectors involve freight flows. Freight flows
include intermediate flows to production facilities, as well as flows to final demand sites
inside and outside the region. This includes import and export flows, but not flows to and
from residential sites. Most of these latter flows correspond to shopping and home-based
NSF Final Report CMS 9633386
15
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
service trips.1 Export flows satisfy final demand outside the region. Some import flows
satisfy final demand within the region, and some are inputs to production processes.
Some import and export flows also appear as throughputs. Los Angeles trade flows are
summarized in Appendix A.
Given the SCPM input-output relationships describing input requirements per unit
of output, and given baseline jobs by economic sector and zone from the Census
Transportation Planning Package (CTPP) made available to SCAG by the U.S. Bureau of
Transportation Statistics (1990), the next step is to compute the total requirements of
output i in zone z,
Dzi =Σj aij•Xzj = sector i shipments to zone z from transshipment
zones (imports) and from other zones to accom-
modate local final demand not associated with
households; (1)
where Xzj = the total output of commodity j in zone z given base year employ-
ment in sector j and zone z, and
aij = is the ij th element of A, the matrix of value demand
coefficients
for the (open) input-output model. This is the flow from i to j per
unit output of j.
The first term on the right hand side of equation (1) accounts for inter-industry shipments
out of all zones by aggregate freight sector i. Because this summation applies to the open
input-output model, Dzi excludes most shipments to households. In the open model,
1 A limitation of our database is that we have no information about service trips to residencesfrom work sites (e.g. utility company visits to homes or trades persons repair visits). On the
NSF Final Report CMS 9633386
16
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
households generate local final demands, but no intermediate demands. Most shipments
associated with this final demand are treated as shopping or service purchase trips. Dzi is
the total flow of commodity i supplied from everywhere to all non-final demand activities
in zone z.
Similarly, we compute total supply of output i furnished by zone z,
Ozi =Σj bij•Xzi = sector i shipments to transshipment zones from
zone z to accommodate nonlocal final demand
(exports) and to other zones to accommodate local
final demand not associated with house-
holds; (2)
where Xzi =the total output of commodity i in zone z given base year employ-
ment in sector i and zone z, and
bij = is the ijth element of B, the matrix of value supply coefficients for
for the (open) input-output model. This is the flow from i to j per
unit output of i.
The first term on the right hand side of equation (2) accounts for inter-industry shipments
out of zone z by aggregate freight sector i. Like Dzi, Ozj excludes most shipments to
households. As in the case of (1), these shipments consist of shopping and service trips.
Ozj is the total flow of aggregate freight commodity i supplied from zone z to all activities
everywhere.
Value flows Ozi supplied by activity i and originating in zone z and value flows Dzi
demanded from activity i and terminating in zone z must be translated into freight trip
other hand, a modest proportion of household consumption is converted into freight shipmentsto capture the delivery of household appliances directly to residences.
NSF Final Report CMS 9633386
17
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
production (supply) Pri and attraction (demand) Asi associated with activity i in zone z.
Using conversion factors constructed from the 1993 Commodity Flow Survey (CFS, U.S.
Department of Transportation 1997), we convert all value flows Dzi and Ozj $ values to
truckload equivalents. The CFS describes freight flows in terms of $/ton for the major
industrial sectors. The 1992 census of transportation (U.S. Bureau of the Census 1993b)
describes tons/truck. This permits calculation of a coefficient, ηi, relating the value of
shipments to zonal transportation requirements, typically measured in terms of passenger
car units (PCU).
Pri =ηi•Ozi
=trip production of commodity i in origin zone z = r, (3)
and
Asi =ηi•Dzi
=trip attraction of commodity i to destination zone z = s (4)
Details are available in Appendix A.
Based on SCAG's network equilibrium costs cSCAGr,s and the trip production and
attraction vectors determined in the steps above, we calibrate nine separate spatial
interaction models. These include nine flows involving people,
• home-to-work,
• work-to-home,
• home-to-shop,
• shop-to-home,
• home-to-other,
NSF Final Report CMS 9633386
18
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
• other-to-home, and
• other-to-other;
and four classes of commodity flows (durable manufacturing, non-durable manufacturing,
mining [mainly petroleum deliveries] and wholesale trade). We estimate each of these
thirteen matrices of inter-zonal flows separately, but via using a common measure of
network equilibrium costs. The structure of inter-zonal flows in each of these matrices
influences network equilibrium costs. Thus, this baseline calibration requires iteration
between the network assignment model and the set of gravity models. The objective of
these baseline gravity model calibrations is to estimate distance decay parameters (Wilson
1970). These distance decay parameters are used to predict travel demand following an
earthquake. Also, once estimated, the home-to-work and home-to-shop (service) matrices
are converted to the JHW and JHS matrices by striking proportions in columns, i.e.,
relative to the total number of trips terminating in zone j.
We rely on a singly-constrained gravity model formulation in the case of freight
because we do not have trip interchange matrices for freight sectors. The parameters of
the singly-constrained formulation are calibrated based on the following criterion (Putnam
1983),
Minimize Σr |Pri(βi)•ln(Pri) - Σr Pri(βi)•ln(Pri(βi))|, (5)
βi
where βi = distance decay coefficient for sector i;
Pri(βi) = estimated trip production of commodity i in origin zone r
= Σs Asi•[Bir•exp(-βi•crs) / Σr Bir•exp(-βi•crs)]; (6)
NSF Final Report CMS 9633386
19
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
crs = generalized cost of transportation from origin zone r to
destination zone s;
Pri = trip production of commodity i in origin zone r;
Asi = trip attraction of commodity i to destination zone s; and
Bir = a constant specific to sector i and origin zone r, the square
root of the number of total employees in origin zone r.
We construct production and attraction vectors for each freight sector using equations (1),
(2), (3), and (4). Given initial values for transportation costs and gravity model
parameters, we proceed by estimating inter-zonal flows for sector i and calculating trip
productions implied by these flows. Trip attractions are fixed. For each sector, the value
of βi is adjusted to move the estimated values Pri(βi) toward the target values Pri.
We have more information about flows involving people, because of the
availability of SCAG's empirically estimated trip interchange tables for the nine classes of
flows described above. The availability of these interchange matrices makes it possible to
estimate distance decay parameters for a doubly-constrained gravity model.
trsi(β) = Pri•Asi•[Bir•His •β0,i•exp(-β1,i•crs)•crs(β2,i)]. (7)
where
β0,i, β1,i, and β2,i = elements in a vector of distance decay coefficients for sector i;
crs = generalized cost of transportation from origin zone r to
destination zone s;
Pri = trip production of flow i in origin zone r;
Asi = trip attraction of flow i to destination zone s;
NSF Final Report CMS 9633386
20
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Bir = a constant specific to sector i and origin zone r
= [Σs Asi•His •β0,i•exp(-β1,i•crs)•crs(-β2,i)]-1; and (8)
His = a constant specific to sector i and origin zone rs
= [Σr Pri•Bir •β0,i•exp(-β1,i•crs)• crs(-β2,i)]-1. (9)
The vector β is is adjusted to match the observed travel distribution, which depends on the
observed flows trsi and the equilibrium network costs crs.
In all cases, equilibrium transportation costs crs are initialized as cSCAGrs, based on
estimated link flows and costs provided by the Southern California Association of
Governments (SCAG). The parameters that minimize (5) and match the travel time
distributions for observed flows also imply a set of 13 trip interchange matrices. Summing
the 13 trip interchange matrices provides a new set of flows, expressed in PCUs, and their
associated equilibrium network costs crs. These costs are fed back into each of the gravity
models. The matrix of equilibrium network costs c and the vector of distance decay
parameters β are iteratively adjusted until consistent travel demands and travel costs are
computed. The end result is a matrix of equilibrium link costs c*rs consistent with a
corresponding set of equilibrium trip interchange matrices consisting of elements t*rsi.
The Status Quo: Earthquake Impacts without Mitigation
The information needed to model the baseline with internal consistency is also
sufficient to treat changes in configuration of the network and the activity system.
Following an earthquake, there will be losses in both network and industrial capacity.
NSF Final Report CMS 9633386
21
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
The former reduces transportation capacity and raises transportation and production costs.
The latter reduces the demands imposed on the network. The bridge performance models
and building fragility curve analysis (EPEDAT) assign consistent losses of both types to
particular earthquake scenarios. The spatial interaction elements of our approach make it
possible to capture the changes in transportation requirements associated with changes in
network performance. These changes and changes resulting from earthquake damage to
industrial facilities are treated simultaneously and consistently.
Inputs for the Elysian Park Scenario
SCPM2 has been applied to the Los Angeles metropolitan area for the scenario
defined by a maximum credible earthquake (magnitude 7.1) on the Elysian Park thrust
ramp. This Elysian Park scenario was selected because of its potential to cause major
damage and casualties. Like the 1994 Northridge earthquake, the Elysian Park scenario
occurs on a blind thrust fault. While the maximum size earthquakes that seismologists
believe are possible on these blind thrust faults are lower than those on, for example, the
San Andreas Fault, their location gives them the potential to cause severe damage because
of their proximity to metropolitan Los Angeles. The planar earthquake source
representation for the Elysian Park event varies in depth from 11.0 to 16.0 km below the
surface. The surface projection of this source includes a broad, densely populated area of
central Los Angeles County, including downtown Los Angeles.
Figure 3 displays the freeway and State highway network for the Los Angeles
metropolitan area used in this research. Further details are provided in Table 2. Los
Angeles soil classes are related to Los Angeles County bridges by census tract in
NSF Final Report CMS 9633386
22
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Appendix B. These soil data make it possible estimate the ranges for peak ground
accelerations incident to Los Angeles bridges under the Elysian Park scenario (see Table
Figure 3: The Los Angeles Metropolitan Area Transportation Network, Consistingof 19,601 Links and 1,534 Traffic Analysis Zones
3). These inputs make it possible to use the improved bridge performance models
developed in Volume I (Shinozuka,1998) to simulate changes in the Los Angeles network
resulting fromdamage to bridge structures. These bridge performance models have been
calibrated against structural failures experienced during the Northridge earthquake (see
Appendix B). The work described here makes it possible to place these bridge impacts in
the context of the transportation network, and to combine these network level results with
estimated damage to other structures.
NSF Final Report CMS 9633386
23
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
As described in Volume I, bridge fragility curves associate the peak ground
acceleration (PGA) to which a bridge is subjected to the probability of achieving a
Table 2: Analytical Description of the Southern California Association of Governments(SCAG) Five-County Area 1992 Freeway Network and California Department of
Transportation (Caltrans) District 7 Bridge Data
Freeway Links
Number of DirectionalFreeway Links
Number of Directional LaneMiles
Baseline Passenger Car Unit(PCU) Miles (Representative
Three Hour Peak Period)
1,914 7,774.3 42,195,364
Other Physical Network Links
Directional MajorArterial Links Directional Minor Arterial Links Directional Minor Road Links
3,396 9,738 1,898
Bridges
Number of BridgesNumber of Directional LinksIncorporating One or More
Bridges
Number of Directional LinksWith No Bridges
2,810 1,590 15,356
Traffic Analysis Zones
Number of InternalTraffic Analysis Zones Number of External Zones Number of Zone Centroid
Connectors (Virtual Links)
1,527 7 2,655
Table 3: Bridge Type and Peak Ground Acceleration by Los Angeles Census Tract
PGA (g) CensusTract
CaltransBridges
City/Co.Bridges
Rangemina
Rangemaxa Number Percent Number Percent Number Percent
0.00 0.05 3 0.2 % 24 1.0 % 15 1.0 %0.05 0.10 54 3.3 % 137 5.9 % 122 7.9 %0.10 0.15 156 9.4 % 184 7.9 % 240 15.5 %0.15 0.20 240 14.5 % 349 15.1 % 248 16.0 %0.20 0.25 314 19.0 % 358 15.4 % 156 10.1 %0.25 0.30 255 15.4 % 293 12.6 % 191 12.3 %0.30 0.35 399 24.2 % 547 23.6 % 406 26.2 %
NSF Final Report CMS 9633386
24
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
0.35 0.40 231 14.0 % 426 18.4 % 172 11.1 %
Totals 1,652 100.0 % 2,318 100.0 % 1,550 100.0 %
Source: Shinozuka (1998). Bridge data are provided by courtesy of Caltrans. See Volume I for details.
specific damage state. The maximum observable PGA is in the neighborhood of 1.0
gravity (g), and certainly less than 1.5 g. A Bridge Damage Index (BDI) of 1.0
corresponds to a bridge damage state of collapse. A BDI of 0.0 indicates no measurable
damage to the structure. . Table 4 summarizes the relationship between BDI values and
bridge damage states. The BDI value associated with each state is a representative value
near the midpoint of the values associated with each state.
The bridge fragility curves developed in Volume I generate a distribution of
network-wide results associated with the Elysian Park scenario. The fragility curves
applied here are preliminary versions of the work summarized in Volume I. These
preliminary fragility curves assume that all the bridges in the SCAG region have
homogenous structural characteristics. Subsequent work reported in Volume I relaxes
this assumption, further disaggregating conditioning information about bridges to account
for single versus multiple spans, deck skew, and soil conditions.
How bridge damage states relate to bridge capacity, or how to represent the
capacity of a network link that includes several damaged bridges, are questions on which
both the transportation engineering and earthquake engineering literatures are mostly
silent. How does moderate damage to a bridge translate into traffic carrying capacity?
There are a number of ways traffic demands on a moderately damaged bridge might be
controlled. Traffic could directed into a single lane defined by the sequence of locations
providing the best available support. Speed limits might be introduced to minimize
NSF Final Report CMS 9633386
25
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
vibrations from heavy vehicles. Heavy vehicles could be excluded. Access could be
metered to control traffic density and total load.
NSF Final Report CMS 9633386
26
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 4: Bridge Damage Indices and Bridge Damage States
Bridge DamageState
Representative BridgeDamage Index (BDI)
Approximate Interval ofBridge Damage Index
Values
No Damage 0.000 0.000 - 0.050Minor Damage 0.100 0.050 - 0.200Moderate Damage 0.300 0.200 - 0.525Major Damage 0.750 0.525 - 0.875Collapse 1.000 0.875 - 1.000
Note: See Volume I for details.
In the absence of empirical information, mapping qualitative bridge damage states
(e.g. minor, moderate, and major damage) to quantitative measures of lane capacity
remains subjective. One approach is to assume a functional relationship between the
values of BDI and capacity parameter(s) that appear in engineering models of traffic
congestion. A BDI of 1.0 corresponds to zero capacity, and a BDI value of 0.0
corresponds to full capacity. But how these parameters relate to each other inside this
interval is conjectural. However, if bridge damage can be mapped to capacity, a link
capacity measure can be constructed for a link consisting of a sequence of damaged
bridges (Cho and Moore 1999). This concept corresponds to the link operationality index
described in Volume I.
Interviews with State Department of Transportation officials at transportation
workshops convened by the Pacific Earthquake Engineering Research (PEER) Center in
June 1998 and the Mid-America Earthquake (MAE) Center in January 1999 provide a
more pragmatic approach that is closer to bridge management practice. Bridges damaged
NSF Final Report CMS 9633386
27
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
by an earthquake are either open or closed. Any residual capacity associated with these
structures may eventually be returned to the system prior to repair, but State Department
of Transportation personnel typically believe the circumstances under which this residual
capacity would be relevant are very narrow. Thus, the key operational question is not:
"How much residual capacity does a bridge have in each damage state?" Rather, the key
question is: "At what bridge damage index value should the bridge be closed?" Our
approach makes it possible to systematically investigate the cost implications of alternative
bridge closure criteria.
The approximate midpoints of the bridge damage index intervals associated with
the moderate and severe damage states given in Table 4 are 0.3 and 0.75, respectively.
We treat these values as the most conservative and riskiest BDI thresholds that
transportation authorities are likely to accept as bridge closure criteria. This is reasonable.
A conservative, safety oriented policy would close moderately damaged structures to
traffic, even though this would increase delay and other transportation costs. A less risk-
averse policy emphasizing the maintenance of regional economic function would leave
moderately damaged structures open, while closing the most dangerous structures. The
risk of further catastrophic failure would further disrupt the regional economy. The
simulations described below are defined in terms of these alternative bridge closure
criteria.
Modeling Losses
Earthquakes induce changes in industrial production because of their effects on
fixed capital stocks, particularly factories, warehouses, and office buildings. Damage to
NSF Final Report CMS 9633386
28
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
production facilities is translated into an exogenous change in final demand (Isard and
Kuenne, 1953). EQE's EPEDAT is a GIS-based earthquake loss estimation program that
estimates ground motion, structural damage, and direct business interruption losses
associated with a specific earthquake (Eguchi et. al., 1997; Campbell, 1997).
Building damage leads to direct losses in industrial production. EPEDAT's loss-
of-function curves convert damage to buildings to loss of production by zone and sector.
Specifically, the model relates structural damage states to business closure times and direct
business interruption (production) losses. Inputs are commercial and industrial building
damage estimates from EPEDAT, expressed as the percent of structures in each of four
damage states by use class and by each of the 308 SCPM zones. Outputs are estimates of
direct business interruption loss for the region by industry, month (over the first year
following the earthquake), and SCPM zone. The model is adapted from research
completed for the Multidisciplinary Center for Earthquake Engineering Research
(Shinozuka et al., 1997). Parameter estimates are based on data from the 1994
Northridge earthquake in Los Angeles.
EPEDAT projects structure losses in the five-county Los Angeles metropolitan
region of between $21.7 billion and $36.2 billion for the Elysian Park event (see Table 5).
If building contents are included, property damage increases to a range of $33.9 to $56.6
billion. Residential damage accounts for approximately two-thirds of the total. About 72
percent of the structural damage is estimated to occur in Los Angeles County.
NSF Final Report CMS 9633386
29
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
A corresponding change in final demand drives SCPM2, ultimately providing
changes in output and employment for 17 sectors across 308 zones. This is an iterative
calculation. Direct changes are exogenous, and already spatially identified. SCPM2
NSF Final Report CMS 9633386
30
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 5: Direct Losses Resulting From Structure Damage ($Billions): Elysian ParkMagnitude 7.1 Earthquake
Structure Type Lower Bound Upper Bound
Residential $ 14.5 billion $ 24.2 billionNonresidential
Commercial 4.1 6.9Industrial 2.7 4.5Other 0.4 0.6
Nonresidential Subtotal 7.2 12.0Structure Subtotal $ 21.7 billion $ 36.2 billionContent Losses 12.2 20.4
Total $ 33.9 billion $ 56.6 billion
allocates indirect and induced changes in a way that respects both observed travel
behavior and new network costs. This process is initialized by allocating indirect impacts
to zones in proportion to baseline data by applying a modified version of SCPM1 (see
Appendix C). SCPM1 uses the proportion of workers in each traffic analysis zone to
establish the spatial distribution of economic activities. The modified version of SCPM1
applied here relies instead on a 17 x 1527 matrix of indices constructed from economic
flows into and out of each traffic analysis zone, the elements of which are initialized as
Fzi = [Ozi + Dzi / Σz (Ozi + Dzi)]
(10)
where Ozi and Dzi are the baseline values given by equations (1) and (2).
Given an initial matrix F, a matrix of baseline equilibrium path costs c, baseline
interzonal shipments, baseline JHW and JHS matrices, a matrix V(d) of direct impacts by
sector and municipality from EPEDAT, disaggregated via GIS over traffic analysis zones,
NSF Final Report CMS 9633386
31
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
and vectors of v(i) and v(u) of indirect and induced impacts by sector from the RSRI
input-output model, we establish an interative sequence that spatially allocates the vectors
v(i) and v(u) over the traffic analysis zones. This creates matrices V(i) and V(u). Set
k=0Dzi = Dzi, (11)
k=0Ozi = Ozi, (12)
k=0Fzi = Fzi, (13)
k=0V(i) = k=0F • diag[v(i)], (14)
and
k=0V(u) = k=0JHST • k=0JHW • k=0F • diag[v(u)] (15.
where k is an iteration counter. This initialization associates indirect and induced impacts
with employment locations. The induced impacts are then distributed across residential
and then commercial locations via the journey-from-home-to-work (JHW), and journey-
from-home-to-shop (JHS) matrices. The total impact by zone at any iteration k is
kVz = kV(d)z + kV(i)z + kV(u)z. (16)
Unlike the corresponding elements of SCPM1, the matrices JHS, JHW, and F are
endogenous. They are updated iteratively to search for a spatial allocation of indirect and
induced impacts that produces mutually consistent travel demand and network costs given
simultaneous reductions in transportation demand and transportation supply. Define
∆kDzi = Σj aij• kVzj, (17)
∆kOzi = Σj bij• kVzi. (18)
NSF Final Report CMS 9633386
32
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
These changes represent decrements in economic activity due to impact of the earthquake.
They are subtracted from baseline values. Update kDzi and kOzi by defining
k+1Ozi = kOzi - ∆kOzi (19)
and
k+1Dzi = kDzi - ∆kDzi (20)
Update kF by defining
k+1Fzi = [k+1Ozi + k+1Dzi / Σz (k+1Ozi + k+1Dzi)]. (21)
Convert these updated values k+1Dzi and k+1Ozi to marginal distributions of trip
productions and attractions in PCUs,k+1Pri = ηi• k+1Ozi (22)
and
k+1Asi = ηi• k+1Dzi (23)
In SCPM1, the network is not explicit and trip making is exogenous. In SCPM2,
the trip interchange matrices are adjusted sub-iteratively. The entries of the thirteen
interchange matrices describing four classes of freight flows and nine passenger trip types
are determined by applying the baseline gravity model coefficients and network costs to
the set of updated trip production and attraction elements k+1Pri and k+1Asi. In the case of
freight flows,
tr,si = Asi•[Bir•exp(-βi•crs) / Σr Bir•exp(-βi•crs)]. (24)
In the case of labor, shopping (and services), and other flows involving people,
trsi = Pri•Asi•[Bir•His•β0i•exp(-β1i•crs)•crs(β2i)]. (25)
NSF Final Report CMS 9633386
33
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Collectively, these inter-zonal flows combine with the earthquake damaged configuration
of the transportation network to generate new endogenous network flows and costs
different from the values of crs. This provides an opportunity for further iteration. The
value of the feedback between network costs and the trip interchange matrices attenuates
given fixed trip production and attraction vectors and fixed gravity model parameters.
Trip distribution and network flows converge to consistent values.
The resulting trip interchange matrices imply new values for the matrices k+1JHW
and k+1JHS, which, along with k+1F, update kV(i) to k+1V(i) via equation (14) and kV(u) to
k+1V(u) via equation (15).
The economic consequences of residential structure damage are not yet fully
accounted for in this analysis. In the results reported here, labor flows are only diminished
as a result of damage to employment sites. This question requires further work.
Aggregate Results for the Elysian Park Scenario
The preliminary fragility curves estimated in Volume I were used to execute 200
Monte Carlo simulations of the Elysian Park scenario earthquake. A sample of simulated
bridge failures associated with the Elysian Park scenario are displayed in Table 6 and
Figure 4. The damage state and bridge damage index achieved by any specific structure
varies across each simulation, but each outcome is drawn from the fixed stochastic process
corresponding to the Elysian Park scenario. Collectively, these simulations correspond to
a distribution of damaged transportation networks. Each network is characterized (in
part) by a vector of dimension 2,810 bridges, each assigned a BDI value. The alternative
bridge closure criteria (BDI > 0.30, BDI > 0.75) are applied to every
NSF Final Report CMS 9633386
34
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 6: Partial Simulation Results for a Hypothetical Elysian Park Earthquake Scenario:Bridge Damage Indices by Bridge Number
Simulation 1 Simulation 2 Simulation 3 Simulation 41.0< BDI
≤1.5BDI >1.5 1.0< BDI
≤1.5BDI >1.5 1.0< BDI
≤1.5BDI >1.5 1.0< BDI
≤1.5BDI >1.5
064LA 034LA 032LA 033LA085LA 085LA 055LA 057LA112LA 090LA 058LA113LA 113LA 118LA
bridge in every network in this set, producing two new distributions. The transportation
networks in these distributions are still characterized by a vector of 1,500 bridges, but
each bridge is now open (1), or closed (0). These results are then used to convert closed
bridges to closed links in the transportation network.
Our model of the Los Angeles economy is convergent, but it is computationally
infeasible to investigate exhaustively each network state represented in these distributions.
Instead, we select representative members of each. The 200 simulations are rank ordered
in terms of the baseline vehicle-miles that would otherwise be traveled across the damaged
links. This rank ordering makes it possible to identify those simulations that are
• maximally disruptive with respect to baseline transportation flows, and
• representative in a median sense.
This approach is summarized in Figure 5. We identify four simulations that are the subject
of more intensive investigation (see Table 7). Maps summarizing bridge closures and loss
of network capacity associated with each scenario appear in Figures 6 through 9.
NSF Final Report CMS 9633386
35
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Figure 4: Simulated Bridge Failures Given Conditions Consistent with the ElysianPark Blind Thrust Fault Scenario: Darker Links Have More Failures
NSF Final Report CMS 9633386
36
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Baseline Activity and Transportation System
Elysian Park Earthquake
Scenario
Bridge Fragility Curves for
2,810 Bridges
Monte Carlo Simulation of
200 cases
Damage State Probabilities for 2,810 Bridges
2,810 X 200
Moderate DamageMinor Damage
CollapseMajor Damage
BDI = 1.00BDI = 0.75
BDI = 0.30BDI = 0.10
Qualitative Definitions of Damage State Midpoints in terms of the Bridge Damage Index (BDI):
Basline Vehicle Miles Traveled
(VMT) by Impacted Link
Median Disruption
Case | Criteria of BDI > 0.75
Maximum Disruption
Case | Criteria of BDI > 0.75
Median Disruption
Case | Criteria of BDI > 0.30
795 Open (1) or Closed (0) Links for 200
Cases | Criteria of BDI > 0.75.
2,810 Open (1) or Closed (0)
Bridges for 200 Cases | Criteria of BDI > 0.75.
Damage States Ranging From 0 to 1 for 200
Cases by 2,810 Bridges
2,810 Open (1) or Closed (0)
Bridges for 200 Cases | Criteria of BDI > 0.30.
795 Open (1) or Closed (0) Links for 200
Cases | Criteria of BDI > 0.30.
2,810 X 200
2,810 X 200
1590 X 200 (links are directional)
1 X 1,590
Conversion of Caltrans Brid- ges to SCAG Network Links
GIS
2,810 X 16,946 (exludes virtual links)
Alternative Closure Criteria for Bridges:
Maximum Disruption
Case | Criteria of BDI > 0.30
BDI > 0.30, BDI > 0.75
1590 X 200 (links are directional)
Figure 5: System Simulations Selected for the Elysian Park Scenario
Tables 8 through 11 correspond to Figures 8 through 11. These tables summarize
these preliminary simulation results describing the full costs of a magnitude 7.1 Elysian
Park earthquake. For example, Row A in Table 8 reflects the midpoint of the range of
NSF Final Report CMS 9633386
37
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 7: Summarizing Damage States for 200 Monte Carlo Simulations of the ElysianPark Scenario
Bridge DamageIndex (BDI)Threshold Value forBridge Closure
Number ofDamaged
DirectionalLinks
Number ofDamaged
DirectionalLane Miles
Baseline PassengerCar Unit (PCU)
Miles Associatedw/Damaged Links
QualitativeDescription
BDI Threshold = .30a
Maximumb of 200Cases
326 1,305.7 10,653,932
Very disruptivebridge impacts,conservationbridge closurecriteria
BDI Threshold = .30Median of 200 Cases 277 1,020.2 7,733,999
Representativebridge impacts,conservationbridge closurecriteria
BDI Threshold = .75c
Maximum of 200Cases
122 479.6 3,877,952Very disruptivebridge impacts,risky bridgeclosure criteria
BDI Threshold = .75Median of 200 Cases
84 309.0 2,257,160Representativebridge impacts,risky bridgeclosure criteria
Notes: a Bridges achieving a Bridge Damage Index of 0.30 (moderate damage) or greater are closed totraffic. This is a conservative criterion that emphasizes post-event public safety, but greatlydiminishes the capacity of the transportation system.
b Each of the Monte Carlo simulations for the Elysian Park scenario is ranked according to thebaseline Passenger Car Unit (PCU) Hours associated with damaged links. The scenario withthe maximum score corresponds to the maximum displacement in terms of baseline vehiclemiles traveled. The simulation with the median score is more representative of the stochasticprocess parameterized by the bridge fragility curves.
c Bridges achieving a Bridge Damage Index of 0.75 (major damage) or greater are closed totraffic. This is a risky criterion that emphasizes post-event transportation supply, butincreases risks to the public.
structure damage shown in Table 5, $45.25 billion, including $29 billion in structure loss.
Row B is the sum of direct, indirect, and induced losses computed by the input-output
model of the five-county, Los Angeles metropolitan area. This sum is $46.7 billion.
These values are identical across Tables 8 through 11. Row C summarizes the post-
earthquake network equilibrium transportation costs in light of reduced production and
NSF Final Report CMS 9633386
38
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Figure 6: Closed Freeway Segments, BDI > 0.30, Maximum of 200 Simulations
NSF Final Report CMS 9633386
39
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Figure 7: Closed Freeway Segments, BDI > 0.30, Median of 200 Simulations
Figure 8: Closed Freeway Segments, BDI > 0.75, Maximum of 200 Simulations
NSF Final Report CMS 9633386
40
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Figure 9: Closed Freeway Segments, BDI > 0.75, Median of 200 Simulations
Table 8: Total Loss ($Billions): Elysian Park Magnitude 7.1 Earthquake,Maximum Simulated Disruption to Baseline Transportation (Closure at BDI > 0.30)
Loss Type BaselineElysian Park Scenario:Conservative Bridge
Closure Criterion
A Structure Lossa $ 45.250 billion(33.5% of total)
Business LossDirect Lossb 28.155Indirect Lossc 9.627Induced Lossd 8.955
B Business Loss Subtotal 46.737 billion(34.6% of total)
Network Costse PCU Minutes $ Billions PCU Minutes $ BillionsPersonal Travel Cost 85,396,813. 21.290 225,830,486. 56.300Freight Cost 10,298,781. 4.550 28,285,954. 12.495Total Travel Cost 95,695,594. 25.839 254,116,440. 68.795
Network Loss = ∆ Network Costs PCU Minutes $ Billions
∆ Personal TravelCost
140,433,673. 35.010
∆ Freight Cost 17,987,173 7.946
C ∆ Total Travel Cost 158,420,846. 42.956(31.8% of total)
Loss Total = A + B + C $ 134.943 billion
Notes: a. Midpoint of interval in Table 3.b. EPEDAT, EQE International.c. RSRI Model.d. Difference between the RSRI solution with the processing sector closed with respect to labor
and the RSRI solution with the processing sector open with respect to labor.e. Network cost is the generalized total transportation cost associated with a simultaneous
equilibrium across choice of destinations and routes. These estimates reflect 365 travel daysper year, an average vehicle occupancy of 1.42 for passenger cars, 2.14 passenger car unitsper truck, a value of time for individuals of $6.5/hour, and $35/hr for freight.
NSF Final Report CMS 9633386
41
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 9: Total Loss ($Billions): Elysian Park Magnitude 7.1 Earthquake,Median Simulated Disruption to Baseline Transportation (Closure at BDI > 0.30)
Loss Type BaselineElysian Park Scenario:Conservative Bridge
Closure Criterion
A Structure Lossa $ 45.250 billion(44.2% of total)
Business LossDirect Lossb 28.155Indirect Lossc 9.627Induced Lossd 8.955
B Business Loss Subtotal 46.737 billion(45.7% of total)
Network Costse PCU Minutes $ Billions PCU Minutes $ BillionsPersonal Travel Cost 85,396,813. 21.290 117,493,842. 29.291Freight Cost 10,298,781. 4.550 15,602,872. 6.893Total Travel Cost 95,695,594. 25.839 133,096,713. 36.184
Network Loss = ∆ Network Costs PCU Minutes $ Billions
∆ Personal TravelCost
32,097,029. 8.002
∆ Freight Cost 5,304,091 2.343
C ∆ Total Travel Cost 37,401,119. 10.345(10.1% of total)
Loss Total = A + B + C $ 102.332 billion
Notes: a. Midpoint of interval in Table 3.b. EPEDAT, EQE International.c. RSRI Model.d. Difference between the RSRI solution with the processing sector closed with respect to labor
and the RSRI solution with the processing sector open with respect to labor.e. Network cost is the generalized total transportation cost associated with a simultaneous
equilibrium across choice of destinations and routes. These estimates reflect 365 travel daysper year, an average vehicle occupancy of 1.42 for passenger cars, 2.14 passenger car unitsper truck, a value of time for individuals of $6.5/hour, and $35/hr for freight.
NSF Final Report CMS 9633386
42
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 10: Total Loss ($Billions): Elysian Park Magnitude 7.1 Earthquake,Maximum Simulated Disruption to Baseline Transportation (Closure at BDI > 0.75)
Loss Type BaselineElysian Park Scenario:Risky Bridge Closure
Criterion
A Structure Lossa $ 45.250 billion(47.7% of total)
Business LossDirect Lossb 28.155Indirect Lossc 9.627Induced Lossd 8.955
B Business Loss Subtotal 46.737 billion(49.3% of total)
Network Costse PCU Minutes $ Billions PCU Minutes $ BillionsPersonal Travel Cost 85,396,813. 21.290 94,349,424. 23.522Freight Cost 10,298,781. 4.550 11,581,677. 5.116Total Travel Cost 95,695,594. 25.839 105,931,101. 28.638
Network Loss = ∆ Network Costs PCU Minutes $ Billions
∆ Personal TravelCost
8,952,611. 2.232
∆ Freight Cost 1,282,896. 0.567
C ∆ Total Travel Cost 10,235,507. 2.799(3.0% of total)
Loss Total = A + B + C $ 94.786 billion
Notes: a. Midpoint of interval in Table 3.b. EPEDAT, EQE International.c. RSRI Model.d. Difference between the RSRI solution with the processing sector closed with respect to labor
and the RSRI solution with the processing sector open with respect to labor.e. Network cost is the generalized total transportation cost associated with a simultaneous
equilibrium across choice of destinations and routes. These estimates reflect 365 travel daysper year, an average vehicle occupancy of 1.42 for passenger cars, 2.14 passenger car unitsper truck, a value of time for individuals of $6.5/hour, and $35/hr for freight.
NSF Final Report CMS 9633386
43
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table 11: Total Loss ($Billions): Elysian Park Magnitude 7.1 Earthquake,Median Simulated Disruption to Baseline Transportation (Closure at BDI > 0.75)
Loss Type BaselineElysian Park Scenario:Risky Bridge Closure
Criterion
A Structure Lossa $ 45.250 billion(48.4% of total)
Business LossDirect Lossb 28.155Indirect Lossc 9.627Induced Lossd 8.955
B Business Loss Subtotal 46.737 billion(50.1% of total)
Network Costse PCU Minutes $ Billions PCU Minutes $ BillionsPersonal Travel Cost 85,396,813. 21.290 89,945,131. 22.424Freight Cost 10,298,781. 4.550 10,966,123. 4.844Total Travel Cost 95,695,594. 25.839 100,911,255. 27.268
Network Loss = ∆ Network Costs PCU Minutes $ Billions
∆ Personal TravelCost
4,548,318. 1.134
∆ Freight Cost 667,343 0.295
C ∆ Total Travel Cost 5,215,661. 1.429(1.5% of total)
Loss Total = A + B + C $ 93.416 billion
Notes: a. Midpoint of interval in Table 3.b. EPEDAT, EQE International.c. RSRI Model.d. Difference between the RSRI solution with the processing sector closed with respect to labor
and the RSRI solution with the processing sector open with respect to labor.e. Network cost is the generalized total transportation cost associated with a simultaneous
equilibrium across choice of destinations and routes. These estimates reflect 365 travel daysper year, an average vehicle occupancy of 1.42 for passenger cars, 2.14 passenger car unitsper truck, a value of time for individuals of $6.5/hour, and $35/hr for freight.
NSF Final Report CMS 9633386
44
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
reduced network capacity. These values vary across the four tables. Table 8 corresponds
to a maximum simulated disruption of baseline transportation combined with conservative,
safety-oriented bridge closure criteria. This results in a substantial reduction in
transportation network capacity, and an associated increase in transportation costs of
almost $43 billion. The full costs of the earthquake are estimated to be almost $135
billion. In this case, transportation costs account for a little less than one third of the full
cost of the earthquake. Table 11 summarizes costs for a median simulated disruption of
baseline transportation combined with higher risk bridge closure criteria. This results in a
much less dramatic reduction in transportation network capacity, and an associated
increase in transportation costs of only about $1.4 billion, or about 1.5 percent of
estimated full costs.
The loss-of-function curves provided by EQE International describe
production capacity over a one-year period following the earthquake. Production capacity
is predicted to approach pre-earthquake levels within six months. Restoration of
transportation network capacity is less well accounted for at this point. Bridges are
assumed to remain closed for one year following the earthquake. During this period they
are repaired or replaced. Other assumptions or empirical relationships can be certainly be
accommodated to refine these preliminary results. State DOT officials provide very
different expert estimates of the time required for repair following extensive damage. For
example, California officials are typically much more optimistic than authorities in the
Midwest.
NSF Final Report CMS 9633386
45
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Spatially Disaggregated Results for the Elysian Park Scenario
SCPM2 provides a disaggregation of economic impacts over metropolitan space
to, we believe, an unprecedented level Tabular results for this research are presented in
Appendix D. Corresponding maps are provided in Appendix E.
• Table D5 shows that nonstructural direct economic (business interruption) losses
were $28.2 billion, or 3.65 percent of the input-output model Gross Regional
Product (GRP) baseline. Most of these losses were in Los Angeles County, where
the FIRE sector accounted for 15 percent of the total losses.
• Table D2 reports that the region's five largest cities (Los Angeles, Long Beach,
Anaheim, Santa Ana and Irvine) also suffered the largest business interruption
losses in absolute terms. This is expected: the largest cities are likely to accrue the
largest losses.
• Table D3 ranks the twenty cities and, in the case of Los Angeles, subareas that
suffered the greatest proportionate business interruption impacts. The five cities
and subareas most heavily impacted lost slighly less than six percent of their GRP
(5.46 percent to 5.83 percent). The twenty proportionally hardest hit cities and
subregions were located mainly in the central and east-central areas of the region.
Nine of the top twenty were sub-areas of the city of Los Angeles, mostly located
toward the east. The only westside cities among the top twenty were Beverly Hills
and West Hollywood.
The region has a highly redundant road and highway system. As a result, loss of
direct access routes is more likely to affect the geographic distribution of activities than its
NSF Final Report CMS 9633386
46
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
total magnitude. Even the distributional effects may be small. Earthquakes reduce both
the supply of and demand for transportation services. SCPM2 endogenizes the
relationship between destination choice, route choice, and network costs. SCPM2 does
not endogenize the relationship between network costs and the propensity to travel.
Propensity to travel remains an exogenous impact of the earthquake that is not subject to
further reductions because of increased transportation costs, although flows do include
adjustments in routes destinations consistent with new network costs and baseline travel
behavior.
• SCPM2 makes it possible to calculate indirect, induced and total effects with or
without highway damage. As noted above, SCPM2 is a spatial model in which
distance decay relationships, and therefore the spatial distribution of shipments, are
endogenous. Tables D6 through D20 summarize the indirect, induced and total
economic impacts and their spatial distribution resulting from the Elysian Park
earthquake, given no network damage. The transportation network is held fixed at
baseline conditions. This is done for purposes of comparison.
• Table D10 shows that regional, nonstructural indirect economic losses were $9.6
billion, accruing mostly in Los Angeles County. Table D15 shows that regional
induced economic losses were $8.9 billion, again largely in Los Angeles County.
Total economic losses without network damage were $46.8 billion, or 6.05
percent of GRP. This implies an overall regional multiplier of 1.66. Most regional
indirect economic losses were in the manufacturing (nondurable) and FIRE
NSF Final Report CMS 9633386
47
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
sectors. Most regional induced economic losses were in the same sectors, but the
FIRE sector is more heavily impacted than the nondurable manufacturing sector.
Corresponding results have been calculated for each of the four bridge-closure
simulations enumerated above (see Tables D21 through D80). These results introduce the
new network costs associated with reductions in supply of transportation services. These
tables summarize the way that accounting for network costs influences the spatial
distribution of economic activities. The redistributions of economic activity is merely one
source of local (city level) losses. Increases in network transportation costs are another
significant source of local impacts. These costs are more difficult to disaggregate. There
is insufficient information to allocate these transportation costs to economic sectors
reliably, but these costs can be geographically distributed to traffic origins and destinations
(see Appendix E for details). Appendix E provides the spatial distributions of the
transportation costs aggregated in Tables 8 through 11. Changes in network costs
corresponding to the four simulations are allocated to zones in Figures E6, E9, E12 and
E15. Network costs are allocated to zones as the average of total delay associated with
all trips originating in each zone and the total network delay associated with trips
terminating in the zone. Delays are calculated using the network equilibrium costs. The
transportation cost computed for any zone r is thus
cr =Σi (Σr c*r,s • t*r,si + (Σs c*r,s • t*r,si) / 2, (26)
where c*r,s = post-earthquake equilibrium cost of transportation from origin
zone r to destination zone s, and
t*r,si =an element of the post-earthquake equilibrium trip interchange
NSF Final Report CMS 9633386
48
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
matrix for sector i.
Total costs for these simulations are summarized in Figures E7, E10, E13 and E16.
These new network costs may also influence the distribution of indirect and
induced economic losses via the distance decay relationship between travel cost and
destination choice. But in all four simulations, the overall GRP changes associated with
indirect and induced economic losses remain minor. Differences in spatially distributed
impacts relative to the no-network-damage exercise are also minor.
This finding corroborates the economic importance of the regional transportation
network's high levels of redundancy. The high level of travel endogeneity associated with
the travel choices represented in SCPM2 is dominated by the redundancy of the Los
Angeles regional transportation network. The four bridge closure simulations affect
between 84 and 326 directional network links, including freeway and arterial links. The
representation of the network contained in SCPM2 includes 16,946 links. Bridge closures
do impact total travel cost and route choice. As Tables 8 through 11 show, the
cumulative value of increased network cost can be significant, but the day-to-day increase
does not induce profound changes in destination choice, and thus does not have a
pronounced impact on the spatial distribution of economic losses.
These results suggest several hypotheses relating to the relationships accounted for
by SCPM2 and the way these relationships are parameterized.
• This application of SCPM2 remains incomplete. The loss-of-function curves apply
only to production activities. The impact on households, i.e., on the production of
labor, has not yet been accounted for, and changes in the spatial distribution of
NSF Final Report CMS 9633386
49
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
activities and losses do not reflect the impact of changes in household
consumption.
• Destination choice may be more sensitive to post-earthquake travel costs than to
pre-earthquake costs. The distance decay functions in SCPM2 are estimated with
pre-earthquake data. Post-earthquake responses to travel cost may be different.
Travelers may be more risk averse than the distance decay functions in SCPM2
imply.
• Travelers may also diminish trip frequencies in response to the cost of travel in
addition to decreases driven by the exogenous impact of an earthquake. If so,
some longer trips would be removed from these results, and this would increase
changes in the geographic distribution of activities and losses.
Policy Tests: Earthquake Impacts with Alternative Mitigation Measures
We can apply the same method to any relevant earthquake or mitigation scenario.
The baseline exercise describes pre-earthquake conditions. The simulations described
above summarize post-earthquake outcomes conditioned on present levels of mitigation.
These results should be contrasted with results that include mitigation measures. The
difference between these full-cost results measures the benefits of the mitigation, to be
compared against the costs of implementing the mitigation. Importantly, the benefits
measured in this manner are provided at the local submetropolitan level. This includes
municipalities, and in the case of the City of Los Angeles, Council districts. If all politics
are indeed local, then results like these are critically important to policy implementation.
NSF Final Report CMS 9633386
50
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
CONCLUSIONS
These research results permit us to assess the earthquake risk to the transportation
system and the urban economy by accounting for a wide range of outcomes associated
with damage to bridges and production facilities. This includes the regional productivity
impacts associated with the damage to the transportation system. Our integration of
seismic, transportation network, spatial allocation, and input-output models permits the
study of how the economic impacts of industrial and transportation structure loss are
distributed over metropolitan space. Some of this loss is produced directly by the
earthquake, which destroys industrial capacity. The procedure accounts for the impact of
industrial structure losses and associated direct losses. The model computes further
indirect and induced losses, and makes the spatial distribution of these losses sensitive to
increases in network costs resulting from transportation structure losses.
In addition to the obvious data difficulties, there are a variety of inevitable
omissions at this stage of our research. The methodolgy does not account for the impact
of transportation structure losses on final demand. The employment consequences of
residential structure losses are not considered. Input-output approaches emphasize
forward linkages, but ignore backward linkages. The reduced demand associated with
damaged industrial facilities is included, but the consequences of constraints on industrial
capacity are overlooked. We do not attempt to account for the many nonmaterial costs
inflicted on the victims of earthquakes. However, we hope to add a feedback from
increased freight costs to reduced household final demand.
NSF Final Report CMS 9633386
51
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Perhaps most importantly, the interindustry core of SCPM2 does not account for
the shock-absorbing role of price changes. It is well known that price adjustments are
crucial resource allocation mechanisms, especially in areas afflicted by natural disasters.
Our next research activity, in addition to addressing the points listed above, aims to
extend SCPM2 to account for endogenous price adjustments.
NSF Final Report CMS 9633386
52
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
APPENDIX A:LOS ANGELES METROPOLITAN (FIVE-COUNTY AREA) TRADE FLOWS
Background
Key descriptors of the Southern California economic were gathered, enumerated,
and compared as part of this research. The objective is to better understand the
commodity flows that pass through the region and utilize its highway infrastructure.
Passenger travel has been well documented by local, state and federal agencies. But,
commodity flows on local highways are not well known. Commodity shipments are
complex. They are international, interregional, and intraregional. Some have local origins
and destinations. Others simply pass through the region. Data on all of these must be
built up from a number of sources. International shipments in and out of the five-county
area flow through two major seaports, Los Angeles and Long Beach and through two
major airports, LAX and Ontario; as well as by rail, highway, pipelines, and various multi-
modal combinations. The same is true for interregional shipments. Intraregional
shipments occur mainly via trucks and highways.
A large number of data sources were reviewed. Unfortunately, these do not all
refer to the same year, although most are post-1990. Data sources and the year that the
data refer to are noted below. Dollar-denominated data are in current dollars unless
otherwise indicated. The California Department of Transportation’s Intermodal
Transportation Management System (ITMS) data have recently become available (Booz,
Allen, and Hamilton, Inc. 1996) but their usefulness is not yet clear.
The Los Angeles region is often described as the five-county area that surrounds
NSF Final Report CMS 9633386
53
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Los Angeles. These five counties are Los Angeles, Orange, Riverside, San Bernardino
and Ventura. The population of this area was 15,608,886 on July 1, 1997. Intercensal
(1980-90) population growth for the five-county area was 26.4 percent. The core county,
Los Angeles, had 9,145,219 inhabitants on the same date in 1997. The five-counties
included 5,165,900 households as of January 1, 1997 (US Bureau of the Census 1998).
The five-county area extends to the Arizona border and covers more than 35,000
square miles, most of it essentially uninhabited. The Bureau of the Census provides some
data for functional Urbanized Areas for the census years. In 1990, the urbanized portion
of the five-county area included 1,965.7 square miles. The population of this area was
11,402,946. The gross population density was 5,800.8 people per square mile, the highest
in the US. The same area included 4,069,116 housing units in the same year.
There were 347,903 nonfarm business establishments in the five-county area as of
1995. Total personal income was $329.6 billion as of 1994. Per capita personal income
was $21,542. California’s personal income in the same year was $702.6 billion. If the
same proportion is applied to California’s 1994 Gross State Product of $875.7 billion, the
local area’s equivalent of GDP would be $410.8 billion. The 1994 GNP (a slightly
different measure) of Canada was $530 billion (US Bureau of the Census 1997).
Southern California’s distribution of personal incomes was 21.6 percent from
goods-related activities and 78.0 percent from service-related activities. Across selected
industries, the distribution was 16.4 percent in manufacturing, 9.6 percent in retail, 7.3
percent in FIRE, 34.3 percent in services, and 13.3 percent government (US Bureau of the
Census 1998).
NSF Final Report CMS 9633386
54
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
The area’s civilian labor force in 1996 was 7,454,300 (4,415,400 in Los Angeles
county). Nonfarm employment in the same year was 6,033,400. The sectoral distribution
was 1,214,700 in goods-related work and 4,818,900 in service-related work. Average
annual pay (all sectors) was $31,897 in the same year (US Bureau of the Census 1998).
International export sales from the five-county area in 1996 were $35,696.5
million (US Bureau of the Census 1997). These data are tabulated in the Metro Area
Exporter Location (EL) file (www.ita.doc.gov/) and differ from Customs District (CD)
data. CD data include all international shipments that clear customs at all airports and
seaports in a region that extends beyond the five-county area. Of course, not all of these
are consumed in the region where they are cleared. CD exports in 1996 were $68,932.2
million. CD imports were $101,184.8 million. EL data result from matching the five-digit
zip codes entered on US export declarations. These locations are assumed to be the
addresses of the original producers. The EL data include country of destination. By this
measure, 19.8 percent of the local region’s total exports went to Canada, Mexico, Central
American and Carribean nations. Canadian destinations accounted for 54.4 percent of
North American exports.
Using an approach that apportions state-level data to the five-county area, Grobar
(1998) finds that 1995 exports were higher than EL reports, $54, 464 million, including
$45,721 in manufacturing goods and $8,743 in services. All of these regional magnitudes
are useful as control totals.
Regional Data
NSF Final Report CMS 9633386
55
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Seaborne Cargo
The total value of 1996 seaborne imports and exports for each of the seaports, by
major sector, are available for 1996 from the US Bureau of the Census’ SA780 reports.
The totals are in Table A.1. The average value of shipments computed by Port of Los
Angeles researchers were $4,471 per ton for imports and $1,665 per ton for exports. This
converts the data from A.1 into those found in Table A2.
Port of Los Angeles researchers have also estimated what proportions of these
shipments have origins and destinations in the five-county area: 31 percent of all
waterborne imports remain in the local region, and 24 percent of all waterborne exports
are from the local region. We assume that the same proportions hold for both ports. The
rest of imports and exports are pass-through to and from the rest of the U.S. These
proportions are also available by major industrial sector.
The Port of Los Angeles data are also divided into foreign vs domestic imports and
exports. Foreign imports were 85.4 percent. Foreign exports were 87.0 percent.
Applying the same ratios to both ports makes it possible to summarize the five-county
area’s waterborne trade in Table A.3. The Port of Los Angeles reports assigning the same
proportions for flows originating-in or destined-for the local region to domestic and to
foreign imports. Applying 31 percent (imports to the region) and 24 percent (exports
from the region) to Table A.3 provides the summary shown in Table A.4. The Southern
California Association of Governments’ study of Interregional goods movement
(DRI/McGraw Hill 1994) disaggregates imports and exports by major world regions.
NSF Final Report CMS 9633386
56
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
These proportions have been applied to the data in Table A.3 to determine the amounts
listed in Table A.5.
Airborne Freight
SCAG (1992) notes that 1,413,851 tons of air cargo volumes occurred in 1991,
incoming and outgoing. Mail only adds a small amount to this total and is omitted from
NSF Final Report CMS 9633386
57
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.1: Seaborne Imports and Exports, Major Seaports, 1996 (millions of dollars)
LosAngeles
Long Beach TOTAL (Origin/destinationwithin region)
Imports $57,017 $63,660 $120,677 (31% = $37,410)Exports $15,808 $23,286 $39,093 (24% = $9,382)
Sources: U.S. Bureau of the Census SA780 Reports.
Table A.2: Seaborne Imports and Exports, Major Seaports, 1996 (millions of tons)
LosAngeles
Long Beach TOTAL (Origin/destinationwithin region)
Imports 12.75 14.24 26.99 (31% = 8.37)Exports 9.49 13.99 23.48 (24% = 5.63)
Sources: Calculations by the authors.
Table A.3: Seaborne Trade Summary, 1996
Value TonsTOTAL U.S. World TOTAL U.S. World
Imports $120,677m $17,619m $103,058m 26.99m 3.94m 23.05mExports $39,093m $5,082m $34,011m 23.48m 3.05m 20.43m
Sources: Calculations by the authors.
Table A.4: Seaborne Trade Summary, Local Area vs US
Value TonsTOTAL U.S. World TOTAL U.S. World
Imports $120,677m $17,619m $103,058m 26.99m 3.94m 23.05m5-counties $37,419m $5,462m $31,948m 8.37m 1.22m 7.15mrest of U.S. $83,258m $12,157m $71,110m 18.62m 2.72m 15.90m
Exports $39,093m $5,082m $34,011m 23.48m 3.05m 20.43m5-counties $9,382m $1,220m $8,163m 5.64m 0.73m 4.90mrest of U.S. $29,711m $3,862m $25,848m 17.84m 2.32m 15.53m
Sources: Calculations by the authors.
NSF Final Report CMS 9633386
58
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.5: Origins, Destinations of Southern California Seaborne Trade
IMPORTS EXPORTSValue Tons Value Tons
World (not incl pass-through)
$31,948m 7.15m $8,163m 4.90m
North America (19.8%) $6,325.7m 1.42m $1,616.3m 0.97mCanada (10.8%) $3,450.4m 0.77m $881.6m 0.53mMexico (8.3%) $2,651.7m 0.59m $677.5m 0.41mCentral AmericaAnd Carribean (0.8%) $255.6m 0.057m $65.3m 0.039m
Sources: DRI/McGraw-Hill Inter-Regional Goods Movement Study.
further discussion. Of the freight total, 958,731 (67.8 percent) were domestic and
455,120 (32.2 percent) were international. The report also notes the division of cargo
between the region’s five major airports as LAX (79.1 percent), Ontario International
(17.8 percent), Long Beach (1.7 percent), Burbank (1.2 percent), and John Wayne (0.2
percent). For the purposes of this study, only LAX and Ontario will be included.
H.-S. Jacob Tsao of UC Berkeley’s Institute of Transportation Studies (1998)
reports that 1996 LAX total air cargo tonnage was 1,719,449. He reports that the value
of these shipments was $24.4 billion. Dollars per ton were $14,191. Tsao notes how little
is known about air cargo shipments and argues for an effort to improve these data. Via a
rough interpolation, 1993 LAX tonnage was 1,318,700. This is useful because Tsao
reports 1993 enplaned LAX tonnage was 442,812. This suggests that one-third of LAX
tonnage was enplaned while two-thirds deplaned. We will apply the same ratios to the
1991 data and to Ontario airport.
NSF Final Report CMS 9633386
59
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
The SCAG report cites a 1988 analysis of the international origins and destinations
of LAX shipments (Table III-4). These carge shares are reported as Asia 51 percent,
Europe 29 percent, Latin American 8 percent, North America 4 percent, and other 8
percent. We assume that North America does not exclude the US. Unfortunately, given
the quality of these data, these proportions must be applied to enplanements as well as
deplanements. Because of the value of these shipments, we will also assume that all
domestic shipments are to and from the local region. All these results are summarized for
1991 in Table A.6.
Other Modes
Another important source of interregional and international trade data is the
DRI/McGraw-Hill Inter-Regional Goods Movement Study, prepared for the Southern
California Association of Governments (1995). All modes are ostensibly covered.
These data are only available in terms of tons. Oddly, they list “imports” that have
a local origin and “exports” that have a local destination. Discussions with DRI/McGraw-
Hill staff revealed that these were actually pass-through designations. 1995 control totals
(millions of tons) are shown in Table A.7. Another advantage of these data is that they are
also tabulated by mode. The totals are shown in Table A.8. Because “other” mode is
omitted below, sums may be slightly less than values identified as “Totals. ”
DRI/McGraw-Hill Inter-Regional Goods Movement Study, prepared for the Southern
California Association of Governments (December, 1995). All modes are ostensibly
included.
NSF Final Report CMS 9633386
60
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
These data are only available in terms of tons. Oddly, the report lists “imports”
that have a local origin and “exports” that have a local destination. Discussions with
NSF Final Report CMS 9633386
61
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.6: Airborne Trade, Southern California
LAX Ontario TOTALValue Tons Value Tons Value Tons
InboundForeign $3,409m 240,194 $770m 54,294 $4,179 294,488Domestic $7,209m 508,024 $1,622m 114,321 $8,871 622,345
OutboundForeign $1,702m 119,917 $379m 26,742 $2,081 146,659Domestic $3,551m 250,221 $799m 56,308 $4,350 306,529
Sources: Calculations by the authors.
Table A.7: International and Interregional Trade Summary, 1995
OriginSCAG
DestinationSCAG
Totals 294.2 352.7Exports (pass through) 9.8 Exports (pass through) 19.0Imports (int’l) 23.7 Imports (int’l) 13.6From rest of U.S. 257.2 From rest of U.S. 314.3
5-county area 175.0 5-county area 175.0To rest of N.A. 3.6 From rest of N.A. 5.8
Sources: DRI/McGraw-Hill Inter-Regional Goods Movement Study.Note: The intraregional shipments are from a separate table of the report.
DRI/McGraw-Hill staff revealed that these were actually pass-through designations.
Control totals for 1995 (millions of tons) are shown in Table A.7. Another advantage of
these data is that they are also tabulated by mode. The totals are shown in Table A.8. As
before, because the “other” mode is below, sums may be slightly less than values labeled
“Totals.”
McGraw-Hill’s cross tabulations of these tables are a key to this part of the
research. Unfortunately, the 1995 intraregional shipments are not shown by mode. We
NSF Final Report CMS 9633386
62
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.8: International and Interregional Trade Summary, 1995
Origin SCAG Destination SCAGTotals 294.2 352.7 Railcar 11.7 33.4 Rail Intermodal 29.7 29.0 Rail subtotal 41.4 62.4 Truck Load 191.4 195.1 LTL 4.0 3.9 Private 39.9 42.1 Truck subtotal 253.3 241.1 Air 0.3 0.3 Water 17.1 48.9
Sources: DRI/McGraw-Hill Inter-Regional Goods Movement Study.
work, we have aggregated these into four sectors: mining, wholesale, nondurable
manufacturing, and durable manufacturing.
The data in Table A.9 are troubling because they do not match those in Tables
A.4-A.6. The former, for example, reports waterborne exports that are not to the US nor
the rest of North America at zero. Waterborne imports from the rest of the world are also
zero. International airborne imports are zero, etc. We use the findings of the previous
tables to augment Table A.9. The results are summarized in Table A.10.
Generation of Origin-Destination (O/D) Trip Tables
The O/D table generation procedure in this study is divided into three categories:
estimating freight O/D, modifying the 1991 surveyed SCAG person trip O/D, and freight
trip distribution and total O/D calibration. no existing interzonal freight trip data were
available For the freight O/D estimation procedure. Because of this, we estimated
NSF Final Report CMS 9633386
63
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.9: International and Interregional Trade by Mode and Major Origins andDestinations, 1995
OriginSCAG
DestinationSCAG
Totals 294,209.1 352,738.9
Exports (international) 9,751.9 Exports (pass through) 19,019.4Rail subtotal 5,849.2 Rail subtotal 12,747.6Truck subtotal 3,898.5 Truck subtotal 6,267.9Air 2.0 Air 0.0Water 0.0 Water 0.0
Imports (pass through) 23,667.1 Imports (international) 13,604.6Rail subtotal 12,251.4 Rail subtotal 3,892.1Truck subtotal 11,413.7 Truck subtotal 9,711.1Air 0.0 Air 0.0Water 0.0 Water 0.0
To rest of U.S. 257,155.9 From rest of U.S. 314,348.6Rail subtotal 23,025.2 Rail subtotal 44,912.2Truck subtotal 216,856.1 Truck subtotal 221,327.2Air 229.5 Air 235.1Water 17,045.1 Water 47,874.1
To rest of N.A. 3,634.2 From rest of N.A. 5,766.3Rail subtotal 247.2 Rail subtotal 848.2Truck subtotal 3,182.6 Truck subtotal 3,837.8Air 89.6 Air 40.1Water 63.1 Water 1,022.7
Sources: DRI/McGraw-Hill Inter-Regional Goods Movement Study.
Interzonal flows from data summarized in the previous sections. The base O/D trip table
for person trips was obtained from the 1991 SCAG O/D survey. The SCAG person trip
survey results were slightly modified to accommodate minor definitional changes
NSF Final Report CMS 9633386
64
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.10: International and Interregional Trade by Mode and Major Origins andDestinations, 1995 (Augmented)
OriginSCAG
DestinationSCAG
Totals 317,825.9 378,133.3
Exports (international) 13,818.5 Exports (pass through) 37,635.5Rail subtotal 5,849.2 Rail subtotal 12,747.6Truck subtotal 3,898.5 Truck subtotal 6,267.9Air 140.8 Air 0.0Water 3,930.0 Water 18,620.0
Imports (pass through) 42,285.1 Imports (international) 19,615.9Rail subtotal 12,251.4 Rail subtotal 3,892.1Truck subtotal 11,413.7 Truck subtotal 9,711.1Air 0.0 Air 282.7Water 18,620.0 Water 5,730.0
To rest of U.S. 257,232.9 From rest of U.S. 314,735.8Rail subtotal 23,025.2 Rail subtotal 44,912.2Truck subtotal 216,856.1 Truck subtotal 221,327.2Air 306.5 Air 622.3Water 17,045.1 Water 47,874.1
To rest of N.A. 4,489.4 From rest of N.A. 6,146.1Rail subtotal 247.2 Rail subtotal 848.2Truck subtotal 3,182.6 Truck subtotal 3,837.8Air 89.6* Air 40.1*Water 970.0 Water 1,420.0
Sources: DRI/McGraw-Hill Inter-Regional Goods Movement Study.
required for this research. This calibration procedure makes it possible to estimate the
coefficients of distance decay functions and obtain a set of calibrated O/D freight trip
tables.
NSF Final Report CMS 9633386
65
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Freight Trip Origin-Destination Estimates
We estimated freight flow requirements in dollars across four aggregate industrial
sectors. Based on tonnage, unit dollar per ton, and truck composition, we converted the
estimated freight flow requirements into trip production and attraction values expressed in
passenger car equivalents (PCE).
The freight trip table was divided into three groups of sub-trip tables. These
describe SCAG region - SCAG region (S–S) trips, SCAG region – Other (S–O) trips, and
Other – Other (O–O) trip. S-S trips are those with both trip ends belonging to SCAG
region internal zones. S-O trips are originated from or terminated in the SCAG region
internal zones, but for which the remaining trip end is located in an external zone. The
origin and destination zones of the O-O trips are both external zones. These are pass-
through trips.
We selected the following zones as sources of external flows.
• Freeways: 12 SCAG region boundary Zones
I-40 east end, I-62 east end, I-10 east end, I-111 south end,
I-86 south end, I-15 north and south end, I-5 north and south ends,
I-101 north end, I-58 west end, I-395 north end.
• Seaports: Ports of Los Angeles, Port of Long Beach
• Airports: Los Angeles International Airport, Ontario International Airport
• Railheads: Los Angeles Union Station, Long Beach Carlson
Despite the existence of the external zones, we assumed that pass-through freight
trips among the external zones are all zero except the trips from freeway to freeway. Thus
NSF Final Report CMS 9633386
66
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
the O-O freight trips refer only on the freeway to freeway freight trips. For this reason,
our final freight trip table contains several zero values.
NSF Final Report CMS 9633386
67
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Freight Trips Beginning and Ending in the SCAG Region
We started with the RSRI transactions table to estimate the S-S freight trips. It
includes the zonal demand and supply across sectors in dollar terms. Table A.12 shows
the RSRI output data across industrial sectors.
Since the demand and supply of each sector are not identical to each other, we
chose the smaller value to represent total transactions for each sector. In addition, we
assume that most of the freight movement is associated with mining, manufacturing
(durable and non-durable), and wholesale sectors. We considered only the commodities
from these four sectors to estimate the intra-SCAG region freight movement. Table A.13
shows the selected sectors and their freight in dollars.
We referred to the Inter-Regional Goods Movement Study by DRI/McGraw-Hill
(1995) and the Commodity Flow Survey (CFS: California Bureau of Transportation
Statistics) to estimate sectoral freight weight (tons). The Inter-Regional Goods
Movement Study provided the estimated total tonnage for the SCAG region, which was
174,971 (1000 tons) in 1995. We assumed that all goods movement in the SCAG region
is by truck. From CFS, we calculated the sectoral ratio of freight tonnage, which is
transported by truck in California. This ratio was used to estimate sectoral goods
movement in the SCAG region. Because we referred to disparate sources for estimating
goods movement of SCAG region-Other, Other-Other from SCAG region - SCAG region
calculation, these estimated flows include inevitable but minor inconsistencies. Dollar
values per ton
NSF Final Report CMS 9633386
68
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.12: Transactions in the SCAG Region for the Year 1995 ($1000/Yr)
Sector Demand SupplyAgriculture 4,467,288 4,906,089Mining 7,831,277 *43,201,568Construction 23,364,148 33,329,596Manufacturing (non-durable) 56,561,663 *63,415,358Manufacturing (durable) 46,980,400 *52,247,072Transportation 9,603,055 12,897,536Communications and Utilities 34,060,808 63,828,151Wholesale Trade 19,776,434 *13,533,541Retail 15,076,372 8,130,632F.I.R.E. 42,220,023 47,487,597Business Services 32,748,919 41,470,194Personal Services 2,426,240 2,718,335Entertainment and Recreation 7,414,337 6,940,382Health 827,066 731,240Educational Services 218,252 107,131Professional and Related Services 32,685,802 26,035,307Government 4,822,712 4,021,742Sum 341,144,769 425,001,471
Source: Regional Science Research Institute, 1996.Calculated based on an updated transactions table to 1996 for Southern California Five county-region.
Table A.13: Transactions of the Selected Sectors in the SCAG Region for theYear 1996 ($1000/Yr)
Sector Freight MovementMining 7,831,277Durable 46,980,400Non-Durable 56,561,663Wholesale 13,533,541
Sources: Table A.12 (* items)
were used to cope with potential inconsistencies. Table A.15 shows unit dollar values by
sector for the SCAG region. In the following section, these flows are combined with
NSF Final Report CMS 9633386
69
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.14: Proportion and Tonnage of Goods Movement in SCAG Region by Sector
Sector Proportion(%)1 1000 Tons/YrMining 19.953 34,905Durable 16.65 29,140Non-Durable 44.84 78,452Wholesale 18.56 32,473Sum 100.00 174,9712
Notes: 1. BTS, 1992 Census of Transportation, Communications, and Utilities – 1993 CommodityFlow Survey for California.
2. SCAG, The Inter-Regional Goods Movement Study (Dec., 1995).3. Weighted average by shipment distance was used.
those of SCAG region – Other and Other – Other freight movements. For trip assignment
modeling purposes, the combined freight flows were converted to the number of truck
trips and, eventually, to Passenger Car Equivalent (PCE) units.
Freight Imported Into, Exported From, and Passing Through the SCAG Region
SCAG region – Other and Other – Other freight movements were computed from
Table A.10, and other documents. Combining these freight movements with those of
SCAG region-SCAG region in Table A.15, we obtained a total of 458,749,000 tons of
annual freight shipped by trucks in the SCAG region which is equivalent to 1,256,847 tons
per day. Table A.17 shows that the overall unit dollar value per ton was $1,212.
Generally, durable goods and freight transported by air are the most expensive. Freight
transacted within the SCAG region is less expensive than the overall average. This
reflects short and more frequent shipments.
The numeric values in Table A.16 are summed from many sub-trip tables. For
example, we still had not distributed the SCAG region to Rail movement of 19,122 tons
NSF Final Report CMS 9633386
70
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.15: Dollar/Ton Ratios for SCAG Region in 1995
Sector Dollars/TonMining 224Durable 1,612Non-Durable 721Wholesale 417Average 714
Sources: Tables A.15 and A.14
Table A.16: Total Freight Movement by Truck ($Million/Yr, 1000 Tons/Yr)
External ZonesO/D SCAGRegion Rail Freeway Air Port
$Million 124,907 12,055 59,617 34,338 35,861SCAG Region1000ton 174,971 29,122 48,966 537 21,945$Million 20,461Rail1000ton 49,653$Million 86,218 21,012Freeway1000ton 59,903 17,682$Million 60,444Air1000ton 945$Million 49,035E
xter
nal Z
ones
Sea1000ton 55,024$Million 503,948Total1000ton 458,749
Source: Caltrans, The Intermodal Transportation Management System for California (April, 1996).SCAG, Air Cargo in the SCAG Region (Nov., 1992).SCAG, Inter-Regional Goods Movement Study (Dec., 1995).BTS, 1992 Census of Transportation, Communications, and Utilities – 1993 Commodity FlowSurvey for California.
to the individual zone level. Because the detailed distribution procedure required distance
decay functions for the entire trip table, which also required the final trip table including
person trips, we did this in the final stage. Meanwhile, we kept using the aggregated
values as in Table A.16 for the trip unit conversion procedure.
NSF Final Report CMS 9633386
71
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.17: Dollar/Ton Ratios for SCAG Related Freight Movement in 1995 ($)
External ZonesSector SCAG
Region Rail
Freeway Air Sea Average
Mining 224 216 368 1,751 383Durable 1,612 489 2,287 75,598 6,227 2,718Non-Dura 721 332 1,444 6,902 810 689Whole 417 825 2,081 35,484 1,937 1,086
SCAG Region
Average 714 414 1,218 63,955 1,634 968Mining 216 216Durable 489 489No-Dura 332 332Whole 825 825
Rail
Average 412 413Mining 1,126 282 988Durable 1,751 1,652 1,711No-Dura 871 761 850Whole 2,183 1,039 2,040
Freeway
Average 1,439 1,188 1,382MiningDurable 75,598 75,598No-Dura 6,902 6,902Whole 35,484 35,484
Air
Average 63,955 63,955Mining 853 591Durable 4,797 5,410No-Dura 551 1,052Whole 1,869 1,895
Ext
erna
l Z
ones
Sea
Average 891 1,799Mining 351 216 366 798 430Durable 2,841 489 1,771 75,598 5,658 3,467No-Dura 646 332 1,036 6,902 1,898 724Whole 1,040 825 1,961 35,484 1,920 1,304
Average
Average 1,002 414 1,210 63,955 2,582 1,212
Source: Tables A.14, A.15, and A.16.
NSF Final Report CMS 9633386
72
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Estimation of Freight Origin-Destination Proportions
Freight movements in tons had to be converted into numbers of truck trips. For
this conversion, due to limited data sources, several assumptions and approximations were
applied. Instead of finding the number of truck trips directly, we initially estimated the
proportion of truck trips out of total trips (person trips and truck trips combined).
Two sources were utilized for the estimation of the proportion of truck trips: the
Annual Daily Truck (AADT) Traffic Survey conducted by Caltrans in the SCAG region in
1996, and the Intermodal Transportation Management System (ITMS). The former
reference was used for estimating the proportion of truck trips to freeway vehicle trips in
the SCAG area and the latter reference was used to adjust the proportions.
The AADT Traffic Survey, which counted the number of trucks and non-trucks at
selected points on the SCAG region freeways, shows that the proportion of truck trips in
the SCAG area is 7.76% of all trips. See Table A.18. Because the average trip distance
of freight is usually much longer than that of person trips, freight trips are expected to be
counted more frequently. Therefore, the freight trip proportion of 7.78% is an
overestimation of the true proportion of the truck trips.
We assumed that the probability of a vehicle being observed in the Caltrans Survey
is linearly proportional to the trip length of the vehicle. Thus, by comparing average trip
lengths of trucks and non-trucks, the counting of frequencies in Table A18 can be
adjusted.
NSF Final Report CMS 9633386
73
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
We referred to the national values provided by the Inter-modal Transportation
Management because we do not have the trip length data for trucks in the SCAG region,
It identifies the average truck trip distance as 59.82 miles. SCAG’s Technical
NSF Final Report CMS 9633386
74
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.18: Vehicle Counts on Selected SCAG Freeways
Vehicle Type Number of Count(Vehicles/Day)
%
Trucks devoted toFreight Transport 6,951,601 7.76
Other Vehicles 82,599.599 92.24Total 89,549,200 100.00
Source: Caltrans, 1996, Annual Average Daily Truck Traffic on the California State Highway System.
Documentation and Findings in 1994 identifies the average non-truck trip distance as 7.64
miles. The truck trip distance is 7.83 times the non-truck trip distance, or a truck is
counted 7.83 times more frequently than a non-truck vehicle is. Thus, adjusted truck
counts can be computed as
6,951,601 / 7.83 = 887,816 (trucks/day). (A.1)
Table A.19 shows the adjusted vehicle counts and proportions. The number of
truck trips is adjusted to 1.06% of total vehicle trips in the SCAG region. This estimated
truck trip proportion is validated below by reassigning the total baseline trip table, which is
estimated here and by tracing the truck trip proportions on freeways. We used only the
proportions in Table A.19 to estimate the number of truck trips because the vehicle counts
in the table were not used for trip estimation.
Converting Trucks to Passenger Car Equivalents
The number of non-truck trips counted in the 1992 SCAG O/D survey was
34,032,386 vehicles per day. Assuming the truck trips amounted to 1.06% of the total,
the number of truck trips is computed as
NSF Final Report CMS 9633386
75
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.19: Trip Counts
Vehicle Type Number of Count(Vehicles/day)
%
Trucks devoted toFreight Transport 887,816 1.06
Other Vehicles 82,599,599 98.94Total 83,478,415 100.00
Source: Table A.18 and calculation by the authors.
Table A.20: Proportion of Trucks by Size
Truck Size (Number of Axles) %2 45.033 10.034 4.045+ 40.91
Source: Caltrans, 1996, Annual Average Daily Truck Traffic on the CaliforniaState Highway System.
(Truck Trips) / (Truck Trips + 34,032,386) = 0.0106, (A.2)
which estimates the number of truck trips as 365,924 (trips/day) and, as a result, the total
number of trips is 34,398,310 (trips/day).
To estimate the Passenger Car Equivalents for trucks, we used the number of
trucks used for each sector, and the truck size distributions. We assumed that the overall
truck size distribution follows that of the counted truck size distribution on the freeways
(Table A.20).
NSF Final Report CMS 9633386
76
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.21: Proportion of Registered Truck Vehicles by Sector
Sector MiningManufacturing(Durable/Non-
Durable)Wholesale Total
% 11.8 39.6 48.6 100.0
Source: BTS, 1992, Census of Transportation, Truck Inventory and Use Survey-California.
Table A.22: Proportions of Truck Size by Sector (1)
Axle Mining Durable Non-Durable Wholesale Total %2 P11 P12 P13 P14 45.03 %3 P21 P22 P23 P24 10.03 %4 P31 P32 P33 P34 4.04 %5+ P41 P42 P43 P44 40.91 %
Total % 11.84 % 19.78 % 19.78 % 48.59 % 100.00 %
Source: Tables A.20 and A.21.
It was also assumed that the use of trucks in each sector is proportional to the
registered trucks for each sector in California as shown in Table A.21.
Tables A.20 and A.21, respectively, provided the column sums and the row sums
of Table A.22, of which cell values were estimated.
The problem in estimating Pijs in Table A.22 is that the unknowns are
underspecified, i.e., there are many sets of solutions that satisfy the sums of rows and
columns. We selected a set of solution that satisfies the following equations:
P11 = (r) * (0.4503 * 0.1184), (A.3a)
P12 = (r) * (0.4503 * 0.1978), (A.3b)
P13 = (r) * (0.4503 * 0.1978), (A.3c)
NSF Final Report CMS 9633386
77
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
P14 = (r) * (0.4503 * 0.4859), (A.3d)
P44 = (r) * (0.4091 * 0.4859), and (A.3e)
Piji j,∑ = 1.0 ( i = 1,2,3,4; j = 1,2,3,4). (A.3f)
NSF Final Report CMS 9633386
78
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.23: Proportions of Truck Size by Sector (2)
Axle Mining Durable Non-Durable Wholesale Sum2 4.29 8.04 8.04 24.66 45.03 %3 2.07 1.70 1.70 4.56 10.03 %4 0.47 0.96 0.96 1.65 4.04 %5+ 5.02 9.08 9.08 17.73 40.91 %
Total % 11.84 % 19.78 % 19.78 % 48.59 % 100.00 %
Source: Calculations by the authors
Table A.24: PCE by Truck Size
Truck Size (Number of Axles) Truck PCE2 1.53 1.54 3.0
5+ 3.0
Source: 1) TRB, 1994, Highway Capacity Manual,2) Adjustment by the authors.
Table A.23 shows the estimated proportions of truck size by sector.
We chose PCE values for each truck size from the Highway Capacity Manual
(TRB, 1994).
Applying the PCEs in Table A.24 to Table A.23, we obtained the average PCEs by
sector (Table A.25).
NSF Final Report CMS 9633386
79
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.25: Average PCE by Sector
Sector Truck PCEMining 2.19Durable 2.26Non-Durable 2.26Wholesale 2.10Weighted Average 2.14
Source: Tables A.23 and A.24.
Table A.26: Sector Allocation of Trucks in California (1992)
Sector %Mining 2.4Durable 18.3Non-Durable 18.3Wholesale 61.0
Source: BTS, 1992, Census of Transportation, Truck Inventory and Use Survey-California.
Total PCEs for all freight O/Ds is
365,924 (Trucks/Day) * 2.14 (PCEs/Truck) = 781,954 (PCEs/Day). (A.4)
The 1992 Truck Inventory and Use Survey also provided the sector allocations of
trucks by major usage in California as in Table A.26.
NSF Final Report CMS 9633386
80
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.27: Conversion of Truck Trips/Day to PCE Trips/Day
Sector Truck Trips / Day PCE Trips / DayMining 10,978 24,094Durable 40,252 91,032Non-Durable 40,252 91,032Wholesale 274,443 575,794Total 365,924 781,952
Source: Tables A.25 and A.26.
Table A.28: Freight Trip Distribution (PCEs/Day)
External ZonesO \ D SCAGRegion Rail Freeway Airport Port
Sum
SCAG Region 129,304 22,029 214,510 17,707 50,011 433,561Rail 40,250 40,250
Freeway 219,696 22,170 241,867Airport 31,169 31,169
ExternalZones
Port 35,105 35,105Sum 455,525 22,029 236,680 17,707 50,011 781,952
Source: Tables A.16 and A.27.
Applying the proportions of Table A.26 to the total freight truck trips of 365,924,
we obtained daily number of truck trips by sector. Then, by applying the PCEs in Table
A.25, the number of truck trips were converted to PCEs (Table A.27). Freight trips in
PCE/day for internal transfers within, imports to, and exports from the SCAG region are
described by regional exit point in Table A.28.
NSF Final Report CMS 9633386
81
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Person Trip Origin-Destination Estimates
SCAG Origin-Destination Survey Data and Modifications
Due to the differences among some trip purpose definitions, SCAG’s O/D survey
data were slightly modified. In our case, a “home to work” (HW) trip is defined as a
“one-way trip from home to work regardless of the intermediate trip chains and/or
activities”. This definition is broader than SCAG’s, which distinguished the HW trips
from any home-based work trips with possible intermediate activities. Thus, some “home
to other” (HO) trips chained with “other to work” (OW) trips of SCAG’s survey data
were reclassified as the HW trips in this research.
Another different trip definition involves “home to shopping” (HS) trips. In this
research, HS trips contain additional trips, which include trips involving any money-
spending activities. Due to the expansion of the HS definition, some of the HO, OS, and
WO trips in the SCAG data were classified as HS trips. Any remaining trips, such as HO,
OW, and “other to other” (OO) trips, were unchanged except for the proportion that were
transferred to either HW or HS trips.
Converting Person Trips to Vehicle Trips
To assign trips to the network, person trips were converted to vehicle trips. The
conversion procedures were based on the average vehicle occupancy by trip purpose
observed in the SCAG O/D Survey.
NSF Final Report CMS 9633386
82
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.29: Average Vehicle Occupancy by Trip Purpose
Trip Purpose Average Occupancy (persons / vehicle)H to W 1.10O to W 1.25Other Trips 1.63
Source: SCAG. Summary Findings: 1991 Southern California Origin-DestinationSurvey, February, 1993. p. 33.
Table A.30: Trip Directional Factors (6:00 am to 3:00 pm)
Direction Home to Work Other to work Other TripsP to A 0.4189 0.3691 0.3843A to B 0.0746 0.2365 0.1218
Source: SCAG. Technical Documentation and Findings: Conformity Analysis forthe Fiscal Year 1993 - 1999 Regional Transportation Improvement ProgramAmendment, June 16, 1994. p. 31.
There was a small portion (less than 3 % of total person trips) of transit users, and
considering the average transit vehicle occupancy, the small number of transit vehicles
during the assignment phase was discarded.
Assignment Period
Unlike most other conventional transportation planning analyses, our study is not
concerned with the analysis of peak hour periods. Instead, traffic analysis for an average
weekday was the main interest of this study in order to compute the annual transportation
cost changes and their impacts on the local economy. For this reason, the time period for
trip assignment to the model network was chosen to reflect the daily average instead of a
particular peak period of the day. Thus, three-hour time periods were used for the
NSF Final Report CMS 9633386
83
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
assignment procedure where the hourly trips were the average of trips between 6:00 a.m.
and 3:00 p.m. of the modified SCAG O/D table. Table 13 shows the trip-directional
factors for the selected time period used for the computation of average hourly trip
demand.
Estimating Trip Distribution Parameters
The freight O/D trips estimated above are crudely aggregated. Hence, it is
necessary to distribute them into each zone pair (1542 TAZs). Because we do not have
coefficients for the decay function of freight trips, we initially distributed the freight
demand at each zone using a singly constrained gravity model with the inter-zonal travel
costs of the person trips and with unknown distance decay coefficients. Then, the
unknown coefficients were estimated in a way that minimizes the differences of the two
sets of supplies: one from the distributions of the singly constraint gravity model and the
other one estimated in above.
In case of the person trips, since the SCAG survey provided the distributed trip
table, the decay function could be computed directly. When the person trips were
calibrated, however, the effect of the added freight trips should have been considered in
the later stage. Due to the reciprocal effects of the decay functions of freight and person
trips, we repeated the computation procedures until they approached constant values.
Calibrating the Trip Distribution for Freight
NSF Final Report CMS 9633386
84
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Putman (1983) suggested a method that computes the decay function and, as a
result, a trip distribution table based on the supply and demand for each zone. Putman’s
method computes the coefficient in the decay function in Eq. (1) by minimizing the
objective function (Eq. 2), which is a measure of the differences between observed (fixed)
and estimated supplies.
f(cij) = EXP((β)(cij)) (A.5)
NSF Final Report CMS 9633386
85
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
The procedure is described as follows:
min L(β) = ∑ {S(β) ⋅ log(S(β)) } - ∑ {S(β) ⋅ log(Sobs) } (A.6)
where S(β) = supply (trip production) calculated by a singly constrained
gravity model given β and observed demand, i.e. computed
supply
Si(β) = ∑ j [ Dj ⋅ Ai ⋅ exp(β cij) / ∑ j (Ai ⋅ exp(βcij)) ] (A.7)
β = distance decay coefficient
Sobs = observed supply, i.e. real observed supply from survey
Ai = attractiveness of zone i, in this study we applied total
employment.
The computational steps are:
1) Compute inter-zonal travel costs from the user equilibrium assignment of the person
trip O/Ds.
2) "Search" distance decay coefficients for each of the four sectors of freight by applying
equation (2). Increase the value of β from -5 to 2 with step size of 0.001, and find β
that minimizes L. Compute the freight trip tables based on the new β.
3) Compute the updated inter-zonal travel costs from the user equilibrium assignment of
the person trip O/Ds combined with the freight O/Ds generated in step 2)
4) Repeat steps 2) and 3) until stabilized distance decay and object function values
of the user equilibrium assignment model are obtained.
NSF Final Report CMS 9633386
86
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
After several iterations, a stable optimal objective value, L(β), and the corresponding
value of β was found.
Total Trip Table Calibration
Once we obtained stabilized freight distance decay coefficients, it was necessary to
calibrate the original coefficients of the distance decay functions for person trip O/Ds.
This adjustment would, in turn, change the inter-zonal travel costs again. Thus we had to
re-calibrate the decay function for freight trips. To accommodate this reciprocal effect of
freight and person trips, we combined both trips into one (total trip). Thus we calibrated
the decay functions of both trips simultaneously.
After the trips were combined, we distributed them using a doubly constrained
gravity model. Based on the given trip distribution over travel costs of person trips, we
found the parameters of K0, K1, and K2 for the distance decay function for person trips by
applying a curve fitting technique.
f(cij) = (K0)*EXP((K1)(cij))*(cijK2) (A.8)
This procedure is similar to that used in the case of freight trips, except that the combined
trips now substitute for the freight trips. Thus, the remaining steps are:
5) "Estimate" distance decay of person trip OD's based on revised travel cost.
6) "Search" distance decay coefficients for each of the four sectors of freight by applying
equation (2). Increase the value β from -5 to 2 with step size of 0.001, and find β
that minimizes L. Compute the freight trip tables based on the new β.
NSF Final Report CMS 9633386
87
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
7) Compute the updated inter-zonal travel costs from the user equilibrium assignment of
the person trip O/Ds combined with the freight O/Ds generated in step 6)
8) Repeat steps 5) to 7) until a stabilized distance decay and object function
values for the user equilibrium assignment model are obtained.
Table A.31 shows the results of β in each iteration of the above procedure.
Iterations 1 through 5 were obtained in steps 1) to 4) for freight trips only. Iterations 6
through 16 were generated by steps 5) to 8) for person trips and freight movements
simultaneously.
NSF Final Report CMS 9633386
88
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table A.31: Convergence of the Freight Trip Distance Decay Coefficients
Iteration Mining Non-durable Durable Wholesale1 -0.230 -1.611 -1.022 -1.1122 -0.217 -1.493 -0.961 -1.0423 -0.055 -0.777 -0.459 -0.4844 -0.217 -1.493 -0.961 -1.0425 -0.217 -1.493 -0.961 -1.0426 -0.169 -1.278 -0.813 -0.8837 -0.155 -1.220 0.787 -0.8538 -0.170 -1.274 -0.816 -0.8869 -0.162 -1.244 -0.800 -0.86810 -0.165 -1.261 -0.809 -0.87811 -0.165 -1.249 -0.803 -0.87112 -0.164 -1.257 -0.807 -0.87613 -0.164 -1.254 -0.805 -0.87414 -0.164 -1.255 -0.806 -0.87415 -0.164 -1.252 -0.805 -0.87316 -0.166 -1.255 -0.806 -0.874
Source: Calculated by the authors.
Iterations 1-5 were performed via steps 1-4 for freight trips only. Iterations 6-16 were
performed via steps 5-8 for person trips and freight movements simultaneously.
The final results of three parameters K0, K1, K2 for the nine purpose person trips
are as follows:
Table A.32: Person Trip Distance Decay Coefficients
Trip purpose K0 K1 K2 R2
Home-to-work 0.7170831 -0.0340905 -0.9692865 0.89962Work-to-home 0.6949387 -0.0392511 -0.9222881 0.90586Home-to-shop 10.0242356 -0.0340491 -2.1721108 0.98557Shop-to-home 13.4756787 -0.0231131 -2.3671670 0.98291Home-to-other 3.7132381 -0.0415411 -1.6496039 0.97734Other-to-home 3.9174139 -0.0414954 -1.6737497 0.97780Work-to-other 1.4212996 -0.0351261 -1.2452184 0.94945
NSF Final Report CMS 9633386
89
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Other-to-work 1.4423045 -0.0333827 -1.2640305 0.94849Other-to-other 4.2845710 -0.0435343 -1.7084766 0.97913
Source: Calculated by the authors.
Table A.33: Vehicle Counts on SCAG Freeways from the Baseline Trip TableAssignment
Vehicle Type Number of Count (Vehicles/3 Hrs.) %Trucks devoted to Freight Transport 2,911,090 7.26Other Vehicles 37,187,848 92.74Total 40,098,938 100.00
Source: Estimation by the authors.
We distributed total trips using the coefficients in Tables A.31 and A.32. We obtained an
O/D trip table, which is supposed to replicate the pre-calibration total O/D trip table with
a reasonable error. We then used the calibrated O/D trip table as the baseline trip table.
For the estimation of any seismic influences in the study area, we applied the calibrated
coefficients for the distance decay functions.
Validating Estimated Truck Trips
This section validates the estimated number of truck trips in SCAG area. We
previously estimated the proportion of truck trips based on CALTRANS freeway traffic
counts and on the average trip distances of trucks and other vehicles. This resulted in an
estimate that the proportion of truck trips is 1.06 % of total trips. The baseline trip table
reflects this proportion of truck trips in terms of PCEs. Therefore, we expect that the
1.06 % of truck trips, when assigned to the network, should result in about 7.76 % of
truck counts on the SCAG freeways. The 7.76 % of truck counting proportion on SCAG
NSF Final Report CMS 9633386
90
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
area freeways is the result of CALTRANS freeway traffic counts. By comparing the
proportions of truck counts, one from the CALTRANS survey and the other one from the
assignment of the baseline trip table, we can measure the overall accuracy of the estimated
proportion of truck trips.
We initially assigned the baseline trip table to the network under user equilibrium
conditions. Then the number of trucks and other vehicles on each link were computed in
PCE units from the final assignment. Then the truck counts in PCE units was converted
to truck units according to the weighted average truck PCEs shown in Table A.25. Table
A.33 shows the proportions of trucks and other vehicles of the assignment result.
The assignment resulted in 7.26 % of trucks as a proportion total traffic on the
SCAG freeways. This is very close to the actual surveyed amount, 7.76 %, as shown in
Table A.18. This closeness of fit suggests that the number of truck trips used to estimate
origin destination proportions is sufficiently accurate.
NSF Final Report CMS 9633386
91
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
APPENDIX B:CALIBRATING BRIDGE FRAGILITY CURVESAGAINST THE NORTHRIDGE EARTHQUAKE
Table B1: Soil Class and Bridge Type by Los Angeles Census Tract
SoilClass
UBCSoil
Class
CensusTract
CaltransBridges
City/Co.Bridges
Number Percent Number Percent Number PercentHard rock S1 52 3.1 % 141 6.1 % 153 9.9 %Soft rock S1 138 8.4 % 253 10.9 % 153 9.9 %Soil S2 1,462 88.5 % 1,924 83.0 % 1,244 80.3 %
Totals 1,652 100.0 % 2,318 100.0 % 1,550 100.0 %
Source: The bridge data are provided by courtesy of California Department of Transportation (Caltrans).The soil data are from Evernden and Thompson (1988). The data have been merged by EQE,International.
Note: Soil data are from Evernden and Thomson (1988).
NSF Final Report CMS 9633386
92
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Figure B-1: Peak Acceleration Contour Map for the Northridge Earthquake.
NSF Final Report CMS 9633386
93
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
4.03.53.02.52.01.51.00.50.00.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Period (Sec)
Spe
ctra
l A
ccel
erat
ion
(g)
14.3 Km
27.0 Km
54.6 Km
4.03.53.02.52.01.51.00.50.00.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Period (Sec)
Spe
ctra
l A
ccel
erat
ion
(g)
12.6 Km
26.5 Km
54.6 Km
(a) Selected Census Tracts in Los Angeles (b) Selected Census Tracts in Los AngelesCounty on Soil: County on Soft Rock:
4.03.53.02.52.01.51.00.50.00.0
0.1
0.2
0.3
0.4
0.5
0.6
Period (Sec)
Spe
ctra
l A
ccel
erat
ion
(g)
16.0 Km
29.1 Km
54.7 Km
(c) Selected Census Tracts in Los AngelesCounty on Hard Rock.
Figure B2: Pseudo-Absolute Acceleration Response Spectra (5% damping)
Source: EQE, International. The ground motion attenuation relationship is from Campbell (1997).
NSF Final Report CMS 9633386
94
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table B2: Distribution of State and Local Bridges and the Total Numberof Damaged State Bridges in the Los Angeles Metropolitan Area
County Number ofState Bridges
Number ofLocal Bridges
Total Numberof Bridges
Total Numberof DamagedBridges
Los Angeles 2,097 1,553 3,560 228Riverside 64 338 982 0Orange 463 505 986 0Ventura 324 175 504 5Totals 2,948 2,571 5,519 233
Source: EQE, International. The bridge data are provided by courtesy of California Department ofTransportation (Caltrans).
Figure B3: Empirical Fagility Curves Developed for Caltrans Bridges Subjected to theNorthridge Earthquake.
NSF Final Report CMS 9633386
95
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Table B3: Simulation Results for the Empircal Northridge Earthquake Scenario:Damage Indices by Link Number
Simulation 1 Simulation 2 Simulation 3 Simulation 41.0< DI
≤1.5DI >1.5 1.0< DI
≤1.5DI >1.5 1.0< DI
≤1.5DI >1.5 1.0< DI
≤1.5DI >1.5
009LA 005LA 006LA 071LA 005LA 008LA 006LA 005LA059LA 071LA 059LA 103LA 006LA 109LA 008LA 071LA064LA 072LA 062LA 110LA 058LA 110LA 057LA 072LA073LA 106LA 072LA 062LA 059LA 106LA101LA 110LA 073LA 071LA 087LA 108LA102LA 112LA 078LA 073LA 103LA 109LA105LA 090LA 103LA 117LA 111LA109LA 101LA 106LA 118LA 112LA111LA 105LA 107LA113LA 109LA 112LA
111LA 117LA112LA113LA
NSF Final Report CMS 9633386
96
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Figure B4: Simulated Bridge Failures Given Conditions Consistent with theNorthridge Earthquake: Darker Links Have More Failures
NSF Final Report CMS 9633386
97
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
APPENDIX C:SOUTHERN CALIFORNIA PLANNING MODEL 1 (SCMP1)
The Southern California Planning Model version 1 (SCPM1) is an integrated modeling
approach that incorporates two basic components: input-output and spatial allocation. This
approach allows the representation of estimated impacts corresponding to any vector of changes
in final demand both spatially and sectorally. Changes in final demand are fed through an input-
output model to generate sectoral impacts that are then introduced into the spatial allocation
model.
The first model component is built upon the Regional Science Research Institute (RSRI)
input – output model. This model has several advantages. These include
• a high degree of sectoral disaggregation (515 sectors);
• anticipated adjustments in production technology;
• an embedded occupation-industry matrix enabling employment impacts to be identified
across ninety-three occupational groups (This is particularly useful for disaggregating
consumption effects by income class and facilitates the estimation of job impacts by race.);
• an efficient mechanism for differentiating local from out-of-region input-output
transactions via the use of Regional Purchase Coefficients (RPC); and
• and the identification of state and local tax impacts.
The second basic model component is used for allocating sectoral impacts across 308
geographic zones in southern California. The key is to adapt a Garin-Lowry-type model for
spatially allocating the economic impacts generated by the input-output model. An initial version
of this model was developed to analyze the spatial-sectoral impacts of the South Coast Air
NSF Final Report CMS 9633386
98
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Quality Management District’s Air Quality Management Plan and has been applied to other Los
Angeles metropolitan-area policy problems. The building blocks of the SCPM1 are the
metropolitan input-output model, a journey-to-work matrix, and a journey-to-nonwork-
destinations matrix. This is a journey-to-services matrix that in the Garin-Lowry model is more
restrictively described as a ‘journey-to-shop’ matrix).
Because Garin-Lowry models already include the concept of an economic multiplier, the
key innovation associated with the SCPM1 is to incorporate the full range of multipliers obtained
via input-output techniques to obtain detailed economic impacts by sector and by submetropolitan
zone. The SCPM1 follows the principles of the Garin-Lowry model by allocating sectoral output
(or employment) to zones via a loop that relies on the trip matrices. Induced consumption
expenditures are traced back from the workplace to the residential site via a journey-to-work
matrix and from the residential site to the place of purchase and/or consumption via a journey-to-
services matrix.
A limitation on the conjoining of the input-output and the spatial allocation models is that
the degree of sectoral disaggregation by zones is not as fine as that of the input-output sectors. In
the case of the SCPM, the 494 input-output sectors are collapsed into twelve sectors to allocate
impacts over 219 zones. The zones consist of nineteen identified subcenters (including the Los
Angeles core area, an extended downtown), municipalities, and other intrametropolitan
jurisdictions.
The generic structure of the SCPM may be summarized as follows. First, beginning with a
vector of final demands, v(d), total outputs from the open and closed input – output (I/O) models
are calculated as
NSF Final Report CMS 9633386
99
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
v(o) = (I – Ao) –1 • v(d), (C.1.)
and
v(c’) = (I – Ac)-1 • v(d); (C.2.)
where Ao and Ac are matrices of technical coefficients for the open and closed I/O models
respectively; and where v(o) and v(c’) are the corresponding vectors of total outputs. The
notation c’ indicates that the household sector is included. We use v(c) to represent the vector of
total output from the closed model for all but the household sector. By definition, v(c) may then
be re-expressed as the sum of three types of output; direction (d), indirect (i), and induced (u).
v(c) = v(d) + v(i) + v(u), (C.3.)
v(i) = v(o) – v(d), (17x1) (C.4.)
and
v(u) = v(c) – v(o). (C.5.)
Equation (A.3.6) is the spatial counterpart to equation (A3.3.),
Z(c) = Z(d) + Z(i) + Z(u), (17 x 308) (C.6.)
where in each case Z(•) is a matrix of impacts both by spatial unit (zone) and by sector. The
matrices Z(d), Z(i), and Z(u) are all specified or derived in different ways. The most
straightforward of these is Z(d), which is defined exogenously, such as by an earthquake.
SCPM1 allocates indirect outputs according to the proportion of employees in each sector by
zone. Specifically,
Z(i) = P • diag[v(i)] (C.7.)
}
NSF Final Report CMS 9633386
100
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
where P is a (308x17) matrix indicating the proportion of employees in each zone. The ‘diag’
operator diagonalizes the indicated vector into a (17x17) matrix.
The spatial allocation of induced impacts is somewhat more involved because the induced
output must be traced via household expenditure patterns. Two distinct origin – destination
matrixes are employed, JSH (journey from services to home) and JHW (journey from home to work)
based on 1980 census data. Essentially, employees are traced home from work through JHW and
then from home we take them back further to their shopping destinations, thereby indirectly
accounting for the spatial allocation of that increment of sectoral output satisfying induced
household expenditures. This may be expressed more succinctly in terms of matrix notation as
Z(u) = JSH • JHW • P • diag[v(u)]. (308x308)2 (308x17) (17x17) (C.8.)
The output from SCPM1 is a (308 x 17) matrix of impacts by 17 economic sectors and 308
geographic zones.
NSF Final Report CMS 9633386
101
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
APPENDIX D:TABULAR SUMMARY OF FOUR SETS OF SIMULATION RESULTS FOR THE
ELYSIAN PARK SCENARIO
NSF Final Report CMS 9633386
102
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
APPENDIX E:MAP SUMMARY OF FOUR SETS OF SIMULATION RESULTS FOR THE
ELYSIAN PARK SCENARIO
NSF Final Report CMS 9633386
103
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
REFERENCES
Arnott, Richard, Andre Palma, and Robin Lindsey. 1993. Properties of Dynamic TrafficEquilibrium Involving Bottlenecks, Including a Paradox and Metering.Transportation Science 27(2): 148-160.
Blunden, W. R. 1971. The Land-Use/Transportation System Analysis and Synthesis.Oxford: Pergamon Press.
Boarnet, Marlon. 1998. Business Losses, Transportation Damage and the NorthridgeEarthquake. Journal of Transportation and Statistics, 1(2), 49-64.
Booz, Allen, and, Hamilton, Inc., TEAM. 1996. ITMS (Intermodal TransportationManagement System) User's Guide. Sacramento: California Department ofTransportation.
Boyce, David E. 1998. Formulation of a Transportation Network Model for EarthquakeImpact Analysis. Discussion Paper, Department of Civil and EnvironmentalEngineering, University of Illinois at Chicago (May 3).
Campbell, K. W. 1997. Empirical Near Source Attenuation Relationships for Horizontaland Vertical Components of Peak Ground Acceleration, Peak Ground Velocity,and Pseudo-Absolute Acceleration Response Spectra. Seismological ResearchLetters 68(1), January/February.
Chang, Stephanie E. 1995. Urban Economic Impact of Transportation Disruption:Evidence from the January 17, 1995 Hyogo-ken Nambu Earthquake, presented atthe 14th Conference of the Pacific Regional Science Association (PRSA, Taipei,July 25).
Charney, Alberta and Julie Leones. 1997. Implan’s Induced Effects Identified ThroughMultiplier Decomposition. Journal of Regional Science 37(3): 503-517.
Cochrane, Harold C. 1997. Forecasting the Economic Impact of a Midwest Earthquake,in Barclay G. Jones (ed.) Economic Consequences of Earthquakes: Preparing forthe Unexpected. Buffalo: NCEER.
Conrad, Klaus. 1997. Traffic, Transportation, Infrastructure and Externalities. TheAnnals of Regional Science 31: 369-389.
Dietzenbacher, Eric. 1997. In Vindication of the Ghosh Model: A Reinterpretation as aPrice Model. Journal of Regional Science 37(4): 629-651.
DRI/McGraw Hill. 1994. Interregional Goods Movement Database Final Report. LosAngeles: Southern California Association of Governments.
NSF Final Report CMS 9633386
104
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
DRI/McGraw Hill. December, 1995. Interregional Goods Movement Study. LosAngeles: Southern California Association of Governments.
Eguchi, Ronald T. et al. 1996. Analyzing Economic Impacts and Recovery from UrbanEarthquakes, presented the Earthquake Engineering Research Institute (EERI)Conference on Analyzing Economic Impacts and Recovery From UrbanEarthquakes (Pasadena, October 10).
Eguchi, R. T., J.D. Goltz, H.A. Seligson, P.J. Flores, N.C. Blais, T.H. Heaton, and E.Bortugno. 1997. Real-Time Loss Estimation as an Emergency Response DecisionSupport System: The Early Post-Earthquake Damage Assessment Tool.Earthquake Spectra 13(4): 815-832.
Evernden, J. F. and J. M. Thomson. 1988. Predictive Model for Important GroundMotion Parameters Associated with Large and Great Earthquakes. United StatesGeological Survey. Bulletin 1838.
Florian, Micheal. 1984. An Introduction to Network Models Used in TransportationPlanning, Transportation Planning Models. Amsterdam: North-Holland; NewYork: Elsevier Science.
Grobar, L.M. 1998. Estimates of Export-Related Employment for Southern California.Office of Economic Research, Department of Economics, California StateUniversity, Long Beach.
Gordon, Peter, Harry W. Richardson, and Bill Davis. 1996. The Business InterruptionEffects of the Northridge Earthquake, Report to the National Science Foundation,Grant No. CMS-9416758, the University of Southern California, Los Angeles,CA.
Gordon, Peter, Harry W. Richardson, and Bill Davis. 1998. Transport-Related Impactsof the Northridge Earthquake. Journal of Transportation and Statistics 1(2): 21-36.
Isard, Walter, and Robert E. Kuenne. 1953. The Impact of Steel Upon the Greater NewYork-Philadelphia Industrial Region. Review of Economics and Statistics 35 (4):289-301.
Israilevich,Philip, Geoffrey Hewings, Michael Sonis, and Graham Schindler. 1997.Forecasting Structural Change with a Regional Econometric Input-Output Model.Journal of Regional Science 37(4): 565-590.
Jackson, Randall, Moss Madden, and Harry Bowman. 1997. Closure in Cole’sReformulated Leontif Model. Papers in Regional Science 76(1): 21-28.
NSF Final Report CMS 9633386
105
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Kimbell, Larry J. and Nancy Bolton. 1994. The Impact of the Northridge Earthquake onthe Economies of Los Angeles and California, presented to the Seismic SafetyCommission of the State of California (Burbank, March 2).
Leontif, Wassily. 1951. The Structure of the American Economy, 1919-1939. 2nd ed.New York: Oxford University Press.
Mesnard, Louis. 1997. A Biproportional Filter to Compare Technical and AllocationCoefficient Variations. Journal of Regional Science 37(4): 541-564.
Miller, Ronald E., and Peter D. Blair. 1985. Input-Output Analysis: Foundations andExtensions. New Jersey: Prentice-Hall.
Okuyama, Yasuhide and Geoffrey H.D. Hewings and Michael Sonis. 1997. InterregionalAnalysis of an Unscheduled Event, presented at the 44th North AmericanMeetings of the Regional Science Association International (RSAI, Buffalo,October 31).
Putnam, Steven H. 1983. Integrated Urban Models: Policy Analysis of Transportationand Land Use. London: Pion Ltd.
Richardson, Harry W., Peter Gordon, M-J Jun and M. H. Kim. 1993. PRIDE andPrejudice: The Economic Impacts of Growth Controls in Pasadena, Environmentand Planning A, 25A: 987-1002.
Rose, Adam, Julian Benavides, Stephanie Chang, Philip Szczesniak, and Dongsoon Lim.1997. The Regional Economic Impact of an Earthquake: Direct and IndirectEffects of Electricity Lifeline Disruptions. Journal of Regional Science 37(3):437-458.
Rose, Adam and Julian Benavides. 1998. Regional Economic Impacts, in M. Shinozuka,et al (eds.) Engineering and Socioeconomic Impacts of Earthquakes: An Analysisof Electricity Lifeline Disruptions in the New Madrid Area. Buffalo: MCEER.
Rose, Adam and Dongsoon Lim. 1997. The Economic Impact of Electricity LifelineDisruptions Stemming from the Northridge Earthquake. Final Report to theNational Science Foundation, Department of Energy, Environmental, and MineralEconomics, The Pennsylvania State University, University Park, PA.
Sheffi, Yosef. 1985. Urban Transportation Networks-Equilibrium Analysis withMathematical Programming Methods. New Jersey: Prentice-Hall.
NSF Final Report CMS 9633386
106
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
Shinozuka, M. 1998. Development of Bridge Fragility Curves. Proceedings of the JointUS-Italy Workshop on Seismic Protective Systems for Bridges, New York:Columbia University, April 26-28.
Shinozuka, M., S.E. Chang, R.T. Eguchi, D.P. Abrams, H.H.M. Hwang, and A. Rose.1997. Advances in Earthquake Loss Estimation and Application to Memphis,Tennessee. Earthquake Spectra 13(4): 739-748.
Shoven, John and John Whalley. 1992. Applying General Equilibrium. Cambridge:Cambridge University Press.
Southern California Association of Governments. 1993. 1991 Origin-Destination Survey.Los Angeles, California.
Stevens, Benjamin H., G. Treyz and M. Lahr. 1983. A New Technique for theConstruction of Non-Survey Regional Input-Output Models. InternationalRegional Science Review 8,3: 271-286.
Stopher, Peter R. 1975. Urban Transportation Modeling and Planning, D.C. Healthand Company. Lexington: Lexington Books.
Thomas, Roy. 1991. Traffic Assignment Techniques. England: Avebury Technical.
Tsao, H.-S. Jacob. March 31, 1998. The Role of Air Cargo Movements in California’sGoods Movement. Working paper, Institute of Transportation Studies, Universityof California at Berkeley.
US Bureau of the Census. 1993a. 1992 Economic Census CD-ROM. Washington, DC:Economics and Statistics Administration, US Department of Commerce.
US Bureau of the Census. 1993b. 1992 Economic Census: Census of Transportation,Communication, and Utilities--1993 Commodity Flow Survey. TC92-CF-52.Washington, DC: Economics and Statistics Administration, US Department ofCommerce.
US Bureau of the Census. 1997. Statistical Abstract of the United States: 1997, 117th
Edition. Washington, DC: US Department of Commerce.
US Bureau of the Census. 1998. State and Metropolitan Area Data Book 1997-98: AStatistical Abstract Supplement, 5th Edition. Washington DC: US Department ofCommerce.
US Bureau of Transportation Statistics. 1990. Census Transportation PlanningPackage. Washington, DC: US Department of Transportation.
NSF Final Report CMS 9633386
107
S. Cho, P. Gordon, J. Moore, H.W. University of Southern California Research Bureau, CSL;Richardardson, M. Shinozuka, S.E. Chang California EQE Intl.; University of Washington
US Bureau of Transportation Statistics. 1992. Census of Transportation, TruckInventory and Use Survey - California. Washington, DC: US Department ofTransportation.
US Bureau of Transportation Statistics. 1997. 1993 Commodity Flow Survey: UnitedStates Highlights. Washington, DC: US Department of Transportation.
US Department of Transportation. 1990. National Transportation Statistics AnnualReport, by J. Kelly. Report DOT-TSC-RSPA-90-2, Research and SpecialPrograms Administration. Washington, DC: U.S. Government Printing Office.
Wilson, Alan G. 1970. Entropy in Urban and Regional Modeling. London: Pion.
PASSABLE, BUT CHECK*****
Southern California Association of Governments, Aviation Program. November 1992.Air Cargo in the SCAG Region. Los Angeles.*****
Southern California Association of Governments. February, 1993. Summary Findings:1991 Southern California Origin-Destination Survey. Los Angeles.******
Southern California Association of Governments. June 16, 1994. TechnicalDocumentation and Findings: Conformity Analysis for Fiscal year 1993 – 1999 RegionalTransportation Improvement Program Amendment. Los Angeles.*****
NEEDS WORK*****
California Department of Transportation (Caltrans). 1996. Annual Average Daily Truck(AADT) Traffic on the California State Highway System.**********
Regional Science Research Institute. 1996. ******