-
Trip Length Distributions in Commodity-Based and Trip-Based
Freight Demand ModelingInvestigation of Relationships
Jos Holgun-Veras and Ellen Thorson
cast methodologies have been developed for passenger trips,
notfreight trips. This methodological void usually is filled by
simplis-tic approaches such as assuming that freight trips follow
the samebehavioral mechanisms as passenger trips, which is an
implicitassumption when truck traffic is estimated as a function of
passenger-car traffic. Although the error introduced by this
assumption maynot have major consequences for small urban areas
where the num-ber of freight trips is relatively small, it cannot
be used in large met-ropolitan areas such as New York City where
freight-related tripsare a major contributor to urban congestion,
and freight-specifictransportation policies are warranted.
The complexity of modeling freight demand arises from a
combi-nation of factors. First and foremost, multiple dimensions
are to beconsidered (6). Whereas in passenger transportation there
is only oneunit of demandthat is, the passenger, who for the most
part happensto be the decision makerin freight transportation there
are multipledimensions (volume, weight, and trips) under the
control of a numberof decision makers (drivers, dispatchers,
freight forwarders) whointeract in a rather dynamic environment.
Also, a significant portionof freight demand is discretionary in
nature. In this context, a rela-tively small number of companies
have control over a significantnumber of freight movements.
Integrating their behavior into plan-ning models is rather
challenging because the dynamics of theirdecision-making process,
marked by their commercially sensitivenature, are not part of the
public domain.
The significant differences in time value, or opportunity
costs,exhibited by cargoes also are cited as a major factor in
determiningthe complexity of modeling freight (7). Whereas the
passengers timevalue ranges within the same order of magnitude,
cargo time valuedetermined by opportunity costsexhibits a much
wider range.Cargoes opportunity costs are determined by a
combination of theintrinsic cargo value (determined by market value
and replacementcosts) and the logistic cargo value (a function of
the importance ofthe cargo for the production system at a given
moment in time andinventory levels). At one end, low-priority
cargoes may have intrin-sic cargo values as low as $9/ton (gypsum);
and at the other, high-priority cargoes have intrinsic cargo values
that frequently exceed$500,000/ton (e.g., computer chips) (8).
These figures wouldincrease significantly once the logistic cargo
value is factored in.
The multiple dimensions that could be used for freight
demandmodeling (i.e., weight, trips, and volume) have given rise to
twomajor modeling platforms: commodity-based and trip-based
mod-eling. Different modeling approaches can be used on each of
theseplatforms. The most widely used options include (a) variants
of the
The City College of New York, 135th Street and Convent Avenue,
Building Y-220,New York, NY 10031.
Transportation Research Record 1707 37Paper No. 00-0910
Commodity-based and vehicle-trip-based freight demand modeling
isdiscussed. The characteristics of the trip length distributions
(TLDs) areexamined, defined in terms of tons, as required in
commodity-basedmodeling, and in vehicle trips, as required in
trip-based modeling. Withdata used from a major transportation
study in Guatemala, the TLDsare estimated for both tons and vehicle
trips. The analysis revealed that(a) the shape of the TLDs depends
upon the type of movements beingconsidered; (b) TLDs defined in
terms of tonnage differ significantlyfrom those defined in terms of
vehicle trips; (c) TLDs for different typesof vehicles,
transporting similar commodities, reflect the range of use ofeach
type of vehicle; (d ) though tons TLDs and vehicle TLDs are
differ-ent, the relationship between them seems to follow a
systematic patternthat, if successfully identified, would enable
transportation planners toestimate one type of TLD given the other;
and (e) major freight gener-ators affect the shape of the TLDs, so
complementary models may beneeded to provide meaningful depictions
of freight movements.
The transportation modeling process uses demand models to
forecast,in combination with network models to analyze supply. For
the mostpart, the evolving transportation modeling paradigms have
focused onpassenger transportation while paying little or no
attention to freight.This is because passenger issues traditionally
have been assigned thehighest priorities, effectively reducing the
amount of resources andattention allocated to freight
transportation research and education.
An appropriate consideration of freight transportation issues
isimportant, both in modeling and policy making, because in
additionto making significant contributions to the economy, truck
freighttransportation generates externalitiesthat is, pollution
(1). For thatreason, an increasing number of planning studies are
focusing on(a) defining the role of advanced technologies to
enhance system pro-ductivity and efficiency (e.g., 2,3); (b)
defining traffic control strate-gies aimed at ameliorating the
negative environmental impacts oftruck traffic upon local
communities (e.g., 4); and/or (c) estimatingfuture freight supply
and demand to estimate future needs (e.g., 5).The latter type of
analysis faces significant limitations because (a) thebulk of
transportation demand models have been developed for pas-senger
transportation, not freight; and (b) transportation
planningagencies usually do not have the specialized staff required
to deal withfreight issues nor do they assign freight
transportation a high priority.
The most significant hurdle to including freight transportation
inthe transportation modeling process is that most of the demand
fore-
Thiago
Thiago
Thiago
Thiago
-
38 Paper No. 00-0910 Transportation Research Record 1707
Four Steps Model, (b) direct demand models, and (c)
input-outputmodels. Table 1 summarizes the major combinations,
according tothe principal authors experience. As can be seen,
approaches suchas developing regression models of truck traffic as
a function ofpassenger-car traffic have been purposely left out of
the figure becausethey have very little behavioral and economic
support, regardless oftheir statistical significance.
In addition to the modeling approaches depicted in Table 1,
thereare methodologies that attempt to synthesize (estimate)
freight origin-destination matrices from secondary data, such as
traffic counts, andscreen counts (9, 10). However promising, these
methodologies areout of the scope of this paper because they do not
attempt to explainthe fundamental mechanisms of freight demand,
their main focusbeing on the estimation of origin-destination
matrices consistent withtraffic counts or other secondary data.
This paper analyzes variants of the Four Steps Model applied
toboth commodity-based and trip-based platforms. More
specifically,the paper analyzes the characteristics of the trip
length distributions(TLDs) used in these platforms. The paper has
three major sections,in addition to the introduction. In the
section on commodity-basedversus trip-based models, the major
modeling platforms are brieflydescribed, discussing advantages and
disadvantages and the potentialbenefits that could be derived from
their integration. The case-studysection discusses the TLDs
obtained from a recent origin-destinationsurvey conducted as part
of a major modeling project in GuatemalaCity. This project was
selected as the case study because (a) the datawere collected using
state-of-the-art questionnaires; (b) the samplewas expanded with
proper consideration of double counting of trips;and (c) the sample
contains a mix of intercity and urban trips, whichenables the
comparison of the TLDs from these different environ-ments. The
findings extracted from this analysis are, for the mostpart, of
general applicability and are presented in the conclusionssection.
These conclusions will assist freight modelers in
producingmeaningful models of freight demand.
COMMODITY-BASED MODELS VERSUS TRIP-BASED MODELS
Commodity-Based Models
Commodity-based models focus on modeling the amount of
freightmeasured in tons, or any comparable unit of weight. It is
widelyaccepted that the focus on the cargoes enables
commodity-based
models to capture more accurately the fundamental economic
mech-anisms driving freight movements, which largely are determined
bythe cargoes attributes (e.g., shape, unit weight). In general
terms, thecomponents of the modeling process are those depicted in
Figure 1(built upon Ogden, 6).
The commodity generation models estimate the total number oftons
produced and attracted by each of the individual zones com-prising
the study area. The commodity distribution phase estimatesthe
number of tons moving between each origin-destination pair, andit
usually is undertaken with the help of gravity models (simply
ordoubly constrained) or any other form of spatial interaction
models,such as intervening opportunity models. The mode-split
component,intended to estimate the number of tons moved by each of
the avail-able modes, usually is done with discrete choice models
or paneldata from a group of business representatives (as in Cross
HarborFreight Movement Major Investment Study, 5), or both. Once
theorigin-destination matrices for each mode have been
estimated,vehicle-loading models estimate the corresponding number
of vehi-cle trips. Finally, the vehicle trips estimated in the
previous stepare assigned to the different networks, thus
completing the demandestimation process.
The aforementioned process is believed to have the potential
forcapturing the fundamental mechanisms of the freight demand
process,though some issues deserve further discussion:
Empty trips. Since commodity-based models focus on the
actualcargoes being transported, there is no clear way to model
emptytrips, which by different estimates may represent between 15
and50 percent of the total trips in specific corridors (11).
Modelingempty trips is quite challenging because they are
determined by thelogistics of the freight movements in the area
(something the trans-portation modelers do not have access to), and
for that reason, it usu-ally is very difficult to establish a
cause-effect relationship betweenempty trips and commodity flows or
any other attributes of the trans-portation zones. In some cases,
practitioners have opted to considerempty trips as another
commodity. However pragmatic, this approachneglects the
interrelationship among empty trips, commodity flows,and the
logistics of freight movements, and it does not ensure
com-patibility between the total number of loaded trips and the
totalnumber of empties.
Needed commodity flows. Another obvious disadvantage is
thatcommodity-based approaches require commodity flows,
estimatedthrough expensive and time-consuming origin-destination
surveys,such as the commodity flow surveys conducted by the U.S.
Cen-
TABLE 1 Modeling Platforms and Approaches Most Frequently Used
(I-O = Input-Output)
Thiago
Thiago
Thiago
Thiago
Thiago
Thiago
-
Holgun-Veras and Thorson Paper No. 00 -0910 39
sus Bureau (12); although in the United States an increasing
num-ber of transportation planning agencies are relying on
proprietaryfreight demand databases. These databases are assembled
by pri-vate companies from waybill samples and complemented
withsmall origin-destination surveys.
Trip-Based Models
Trip-based models, as the name implies, focus on modeling
vehicletrips. As can be seen in Figure 2, they only have three
components: tripgeneration (to estimate the number of vehicle trips
produced andattracted by each zone), trip distribution (to estimate
the number ofvehicle trips between each origin-destination pair),
and traffic assign-ment (to estimate the traffic in the network).
Since the focus is onvehicle trips, which presupposes that the mode
selection and thevehicle selections already were done, trip-based
models do not needmode-split or vehicle-loading models.
Trip-based models have some advantages. First and foremost,
theyfocus on a unit (the vehicle trip) for which there is a
significantamount of data in the form of traffic counts, screen
counts, and soforth. Furthermore, an increasing number of
intelligent transporta-tion systems applications are able to track
the movements of vehiclesthrough, at least, segments of the highway
networks, thus increas-ingly becoming an important source of
traffic data. Second, since thefocus is on the vehicle trip,
considering empty trips does not presentany major problem.
Trip-based models have some disadvantages, however. First,
theyare of questionable applicability to situations in which
multiple freighttransportation modes are to be considered. Second,
since the vehicletripthe focus of trip-based modelsis in itself the
result of a modechoice and vehicle selection processes (which have
not been takenexplicitly into account), the identification and
modeling of the eco-nomic and behavioral mechanisms determining
freight demandbecome more difficult, because those mechanisms are
associatedwith the actual commodities being transported.
As can be seen, neither commodity-based nor trip-based
modelingis able to capture the full complexity of freight
movements. The for-mer is expected to capture more accurately the
fundamental mecha-nisms driving freight demand, though it fails to
properly considerempty trips; whereas the latter is able to use
readily available data,including empty trips, but it may not be
able to fully capture themechanisms driving freight demand (which
are conditioned by thecommodities attributes).
This is because by focusing on either the commodities or the
vehi-cle trips, the analysis takes into account only one dimension
of freightdemand. In an ideal world, the ultimate freight demand
model wouldbe able to properly take into account the three main
dimensions:weight, volume, and vehicle trips. An enhanced
representation offreight movements could be achieved if, at least,
two of these dimen-sions were jointly modeled. This quest provides
the rationale forthe study of the relationship between vehicle TLDs
and tons TLDs.Should it be feasible to develop approximation
functions to estimatevehicle TLDs from tons TLDs, or more
appropriately to develop
FIGURE 1 Model components of commodity-based Four Steps
approach.
FIGURE 2 Model components of trip-based Four Steps approach.
-
40 Paper No. 00-0910 Transportation Research Record 1707
models of empty trips as a function of commodity flows, it would
bepossible to create freight demand models able to take advantage
ofthe best features of both commodity-based and trip-based
modeling.
ROLE OF THE TRIP LENGTH DISTRIBUTION
As indicated in Table 1, regardless of the platform being used
(i.e.,commodity- or trip-based), the bulk of areawide studies rely
on spa-tial interaction models, either gravity models or
intervening oppor-tunity models. A fundamental input to these
models is the trip lengthdistribution, or related parameters, which
describe the way in whichfreight demand decreases along a given
impedance variable (e.g., dis-tance, travel time, generalized
cost). The negative relationship be-tween demand and the impedance
variable is required by postulatesof economic rationalitythat is,
the consumption of normal goodsdecreases with price.
The TLD, as traditionally defined in passenger transportation
mod-eling, is nothing more than a representation of the frequency
of trips(or tons) made for various intervals of the impedance
variable. Thecalibration of the trip distribution models entails
ensuring that themodeled TLD resembles the observed TLD (13).
In freight demand modeling, the TLD can be defined in terms
of(a) the weight of the commodities being transported, referred to
astonnage (or tons) TLDs; and (b) the number of vehicle trips,
referredto as vehicle TLDs. Commodity-based models use tonnage
TLDs(usually by commodity group), whereas trip-based models use
vehi-cle TLDs (usually by type of vehicle) in their respective trip
distri-bution models. This ensures consistency with the previous
step oftrip generation modeling. Should an origin-destination
sample beavailable (as in this project), tons TLDs and vehicle TLDs
could be estimated for various commodity groups and/or types of
vehicle,enabling comparison and analysis. The distinction between
vehicleTLDs and tons TLDs is important because, as shall be seen
later, theyare significantly different.
The main objectives of this paper are (a) to examine the
character-istics of vehicle and tonnage TLDs, (b) to identify the
typical prob-lems found when dealing with TLDs, and (c) to study
the relationshipsbetween commodity-based TLDs and trip-based TLDs.
This analysissheds light into the nature of real-life TLDs and
provides invaluablelessons for modeling purposes.
CASE STUDY
The data used in this paper were part of a freight
origin-destinationsurvey conducted in the demand-modeling process
for a major high-way project in Guatemala City. Roadside interviews
were comple-mented by classified traffic counts to expand the
sample accordingto time of day and type of vehicle. The
origin-destination question-naire included questions about the time
of the interview, vehicle type,origin, destination, commodity type,
and shipment size, amongothers. The questionnaire was administered
by the staff of IngenierosConsultores de Centro Amrica. The sample,
comprised of 5,276observations, was expanded by time of day and
type of vehicle andprocessed to eliminate double counting of trips.
The overall expansionfactor was 6.476.
Due to the nature of the project under analysis, a bypass road
on theoutskirts of Guatemala City, and the location of the survey
stations,the sample includes a mix of intercity and urban freight
movements.Out of a total of 34,986 trips/day, pickup trucks carried
59.47 percent;large two-axle trucks, 23.77 percent; semitrailers,
10.40 percent; andthe other truck types captured the rest. The
total tonnage is distributedas follows: pickups, 9.54 percent;
large two-axle trucks, 33.79 percent;semitrailers, 46.41 percent;
with the other types capturing the rest.
The TLDs for the entire data set, shown in Figure 3, were
analyzedfirst and found to have some interesting features: (a) the
vehicle TLDfor total trips is a weighted combination of the
empty-vehicles TLDand the loaded-vehicles TLD, as expected; (b) the
tons TLD is sig-nificantly different from the vehicle TLDs; (c) the
differences bet-
FIGURE 3 Vehicle and tonnage TLDs.
-
Holgun-Veras and Thorson Paper No. 00 -0910 41
ween the TLDs disappear with increasing values of distance;
and(d) the TLDs are (statistically speaking) multimodal. As shown
later,the first mode (around 30 km) corresponds to the mode of
internaltrips, while the second mode (around 60 km) corresponds to
the modeof intercity trips.
The statistical multimodality of the TLDs reflects the nature of
thetrips being considered and the associated land use
characteristics. To
examine this hypothesis, the TLDs were obtained for two trip
types:(a) internal-internal trips, that is, origin and destination
inside thestudy area, shown in Figure 4a; and (b) their complement
(external-external, external-internal, and internal-external),
mostly intercitytrips, shown in Figure 4b.
As can be seen, the TLDs depicted in Figure 4a (internal trips)
aresignificantly different from those depicted in Figure 4b. First,
the
FIGURE 4 TLDs for the entire sample: (a) internal trips; (b)
intercity trips.
-
42 Paper No. 00-0910 Transportation Research Record 1707
internal TLDs are smoother than the intercity TLDs. This
arisesbecause origins and destinations in the study area
(predominantlyurban and suburban) are located in a continuum of
distances, whereasorigins and destinations for intercity trips do
not occupy the full rangeof distances because of terrain and the
location of other cities. Second,intercity TLDs are more likely to
be affected by the existence ofspecial land uses such as marine
ports and major agricultural areas,which tend to produce spikes
(modes) on the TLDs. For example, inFigure 4b, the relatively high
percentage of movements taking placebetween 300 and 350 km is
explained by the freight flows betweenPuerto Barrios (the main
Guatemalan port) and Guatemala City. Thisphenomenon is discussed
later.
The statistical multimodality of the TLDs is more than a
statisti-cal curiosity. First, systematic (not random)
multimodality is notconsistent with rational economic behavior
because it implies thatdemand increases with cost (the exception is
the ascending branchof the first mode, which usually is explained
by intermodal or inter-vehicle competition, or both). Thus, when
systematic multimodal-ity is identified (e.g., port flows in Figure
4b), complementarymodels should be used to represent those flows
separately. Second,from the practical standpoint, unimodal
impedance functions, suchas the ones in the commercial demand
modeling software, are notable to adequately represent TLDs such as
the ones depicted inFigure 4b. This situation poses a problem to
practitioners calibratingtrip distribution models.
In the following sections, both tons TLDs and vehicle TLDs
areestimated according to types of vehicle and to commodity
groups.This will provide the basis for the analysis and
identification of thecharacteristics associated with the
corresponding TLDs.
Observed Characteristics of Vehicle TLDs and Tons TLDs by Type
of Vehicle
This section analyzes the TLDs for different types of vehicles.
For thesake of brevity, the analysis focuses only on two cases:
pickup trucksand semitrailers, shown in Figures 5 and 6. As can be
seen, the TLDsare significantly different with shapes and ranges
that are conditionedby the vehicle size, reflecting various ranges
of use. These resultsconfirm long-held assumptions that the TLDs
depend upon the typeof vehicle and the corresponding carrying
capacity. This clearly indi-cates that trucking companies perceive
the different types of vehiclesas having a range of optimal use,
which, in turn, is reflected in thecorresponding TLDs. This
assumption is supported by Figures 7a, b,and c, which show the
percentage of tons transported, at variousdistance intervals, by
specific groups of vehicles.
The TLDs for semitrailers shown in Figure 6b exhibit a
peakaround 330 km, which corresponds to flows to and from Puerto
Bar-rios. However, the TLDs for pickup trucks do not exhibit
suchbehavior (see Figure 5b), because they are not used to serve
the port.This indicates that major generators produce differential
impactsupon vehicle classes, the magnitude of which depends upon
the typeof vehicle serving the generators.
Observed Characteristics of Vehicle TLDs and Tons TLDs by
Commodity Group
This section analyzes the TLDs for the main commodity groups
cap-tured in the origin-destination survey. The tonnage
distribution trans-ported by the different types of vehicles
includes (a) construction
materials, 30.37 percent; (b) manufactured products, 15.03
percent;(c) mineral fuels, 6.67 percent; (d) fruits and vegetables,
5.96 percent;(e) cereals, 4.79 percent; ( f ) others, 4.40 percent;
(g) electrical equip-ment, 3.92 percent; (h) beverages, 3.24
percent; (i) miscellaneousmetal articles, 2.27 percent; and ( j)
textiles, 2.02 percent. The analy-sis focuses on a selected number
of commodity groups. Figure 8shows the TLDs for mineral fuels, and
Figure 9 shows them for fruitsand vegetables. The TLDs shown
correspond to the types of vehicleused to transport these cargoes
(vehicles with less than 2 percent werenot considered) as well as
the total tons TLD.
Interestingly, while the total tons TLD for the entire
commoditygroup reflects the geographic distribution of economic
activities(e.g., production, consumption), the tons TLDs for each
of the vehi-cle types reflect the intervehicle competition
discussed in the pre-vious section. The vehicle TLDs are similar to
the total tons TLDonly in cases, such as the one depicted in Figure
8b, in which onetype of vehicle dominates the market.
As discussed before, major generators have an impact upon
theTLDs. The peak in Figure 8b, taking place around 310 km,
corre-sponds to the flow of fuel being transported by semitrailers
betweenPuerto Barrios and Guatemala City. The uniqueness of the
economicmechanisms explaining this relationship necessitates the
use ofcomplementary models able to capture the fundamental elements
ofthis process.
Major Freight Generators
As indicated before, the existence of major freight
generatorsinthis case, a major porthas an impact upon both tons
TLDs andvehicle TLDs. In order to illustrate the nature of this
impact, gammafunctions were estimated for two different cases: with
port flows (allflows) and without port flows. Gamma functions were
selectedbecause they frequently are used in trip distribution
modeling. Theoriginal tons TLDs and the corresponding gamma
functions aredepicted in Figure 10. As can be seen, the inclusion
of the port flowsin the tons TLD skews the gamma function to the
right, shifting itsmode from 100 km to 180 km.
CONCLUSIONS
This research conducted an in-depth review of the major
modelingplatforms for freight movements: commodity-based and
vehicle-trip-based modeling. It is found that these platforms
represent unidimen-sional views of a phenomenon that, in essence,
is multidimensionalin naturea full depiction of freight flows
entails joint considerationof cargoes weight and volume as well as
vehicle trips.
Both commodity-based and vehicle-trip-based models
encounterchallenges that are difficult to overcome. Commodity-based
model-ing is not able to adequately model empty trips, although it
hasthe potential to capture the mechanisms driving freight
demand.Trip-based modeling, though able to consider empty trips, is
limited inits ability to fully capture the fundamental mechanisms
conditioningfreight demand, which are determined by the cargoes
attributes(which are not explicitly modeled). This situation
suggests thatintegrating both commodity-based and trip-based
modeling wouldprovide an enhanced understanding of freight
movements. Suchintegration, at the trip distribution stage,
requires an understand-ing of the relationship between tons TLDs
and vehicle TLDs. Thisquest provided the rationale for this
paper.
Thiago
-
Holgun-Veras and Thorson Paper No. 00 -0910 43
FIGURE 5 TLDs for pickup trucks: (a) internal trips; (b)
intercity trips.
This paper focused on the relationship between tons TLDs
andvehicle TLDs. It was found that the shape of the TLDs is
condi-tioned, to a great extent, by the environment in which the
freightmovements take place. Freight movements taking place in
urbanand suburban areas, as a consequence of the relatively
homoge-neous land use (from the standpoint of freight demand), lead
toTLDs that are smooth and relatively unimodal. At the other end
ofthe spectrum, freight movements in intercity travel lead to
TLDsthat are (statistically) multimodal and conditioned by major
freight
generators, such as ports. These major generators have the
potentialto significantly alter the shape of the TLDs, and they may
requirecomplementary models to fully represent their fundamental
economicrelationships.
The TLDs of different vehicles reflect the range of use of each
typeof vehicle. This means that each type of vehicle tends to
dominate therange of distance at which its relative performance is
better.
The TLDs, regardless of being defined in terms of tons or
vehicletrips, are significantly different. Nevertheless, the tons
TLDs and
-
FIGURE 6 TLDs for semitrailers: (a) internal trips; (b)
intercity trips.
-
FIGURE 7 Vehicle shares by distance intervals: (a) pickups and
small trucks; (b) large two- and three-axle trucks; (c)
semitrailers.
-
FIGURE 8 Tons TLDs for mineral fuels: (a) internal trips; (b)
intercity trips.
-
Holgun-Veras and Thorson Paper No. 00 -0910 47
FIGURE 9 Tons TLDs for fruits and vegetables: (a) internal
trips; (b) intercity trips.
-
48 Paper No. 00-0910 Transportation Research Record 1707
vehicle TLDs tend to exhibit a fairly consistent behavior with
respectto each other. In all cases, the difference between them
tends to in-crease with distance up to a point, and then it
consistently decreasesuntil becoming indistinguishable in the upper
range of distances. Thisfeature may be exploited by developing
approximation functionsbetween vehicle TLDs and tons TLDs. If such
approximation func-tions are successfully calibrated, this will
enable the practitionerto exploit the best features of
commodity-based and truck tripmodeling.
ACKNOWLEDGMENTS
The authors acknowledge the cooperation and assistance of
Inge-nieros Consultores de Centro Amrica and its president, Jorge
Erdg-menger, during the data collection component of this research.
Thisresearch was partially supported by the University
TransportationResearch Center and its director, Robert Paaswell.
This support isboth acknowledged and appreciated.
REFERENCES
1. Small, K., C. Winston, and C. Evans. Road Work: A New Highway
Pric-ing and Investment Policy. The Brookings Institution,
Washington,D.C., 1989.
2. Holgun-Veras, J., and C. M. Walton. The Role of Information
Technol-ogy on the Implementation of Priority Systems for
Intermodal Contain-ers. Proc., Third Annual World Congress on
Intelligent TransportationSystems, Orlando, Fla., 1996.
3. Regan, A. C., H. S. Mahmassani, and P. Jaillet. Improving
Efficiencyof Commercial Vehicle Operations Using Real-Time
Information:Potential Uses and Assignment Strategies. In
Transportation ResearchRecord 1493, TRB, National Research Council,
Washington, D.C.,1995, pp. 188198.
4. Holgun-Veras, J., V. Ochieng, P. Kinney, and T. S. Lena.
ExploratoryAnalysis of the Impacts of Truck Activity upon Local
Communities.Research in progress funded by the New York City
Department ofHealth.
5. Cross Harbor Freight Movement Major Investment Study. New
YorkCity Economic Development Corporation, 1998.
6. Ogden, K. W. Urban Goods Movement. ISBN 1-85742-029-2.
AshgatePublishing Limited, England, 1992.
7. NCHRP Report 388: A Guidebook for Forecasting Freight
Trans-portation Demand. TRB, National Research Council,
Washington,D.C., 1997.
8. Holgun-Veras, J., and S. Jara-Daz. Optimal Space Allocation
and Pric-ing for Priority Service at Container Ports.
Transportation Research B,Vol. 33, No. 2, 1999, pp. 81106.
9. Tamin, O. Z., and L. G. Willumsen. Freight Demand Model
Estimationfrom Traffic Counts. In Simplified Transport Demand
Modeling (J. deDios Ortzar, ed.), PTRC, London, 1992.
10. List, G. F., and M. A. Turnquist. Estimating Truck Travel
Patterns inUrban Areas. In Transportation Research Record 1430,
TRB, NationalResearch Council, Washington, D.C., 1994, pp. 19.
11. Holgun-Veras, J. Desarrollo de un Modelo para Cuantificar la
OfertaVehicular del Transporte de Carga. Master thesis. Universidad
Centralde Venezuela, July 1984.
12. Commodity Flow Survey. U.S. Census Bureau, U.S. Department
of Com-merce, 1993 and 1997.
13. Ortzar, J. D., and L. G. Willumsen. Modelling Transport, 2nd
ed. JohnWiley & Sons, New York, 1994.
Publication of this paper sponsored by Committee on Freight
TransportationPlanning and Logistics.
FIGURE 10 Tons TLDs with and without port flows.