The Dissertation Committee for Salvador Arturo González-Ayala certifies that this is the approved version of the following dissertation: Development of a Travel Demand Model for Transborder Commuter Activity Committee: ________________________________ Randy B. Machemehl, Supervisor ________________________________ B. Frank McCullough ________________________________ Susan L. Handy ________________________________ Rob Harrison ________________________________ Leigh Boske ________________________________ Zhanmin Zhang
108
Embed
Development of a Travel Demand Model for Transborder ...
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
The Dissertation Committee for Salvador Arturo González-Ayala
certifies that this is the approved version of the following dissertation:
Development of a Travel Demand Model
for Transborder Commuter Activity
Committee:
________________________________ Randy B. Machemehl, Supervisor
________________________________ B. Frank McCullough
________________________________ Susan L. Handy
________________________________ Rob Harrison
________________________________ Leigh Boske
________________________________ Zhanmin Zhang
Development of a Travel Demand Model
for Transborder Commuter Activity
by
Salvador Arturo González-Ayala, B.S.C.E., M.S.E.
Dissertation
Presented to the Faculty of the Graduate School of
the University of Texas at Austin
in Partial Fulfillment
of the Requirements
for the Degree of
Doctor of Philosophy
The University of Texas at Austin
December 2005
Dedicado a
Anna Paola, Alejandra Renée y Rodrigo Andrés.
Acknowledgments
My special thanks and appreciation to Dr. Randy Machemehl for his dedicated guidance
and support over my last years as a UT student, as well as to Mr. Rob Harrison for his insightful
assistance and interest in the final preparation of this document.
iv
Development of a Travel Demand Model
for Transborder Commuter Activity
Publication No._________
Salvador Arturo González-Ayala, Ph.D.
The University of Texas at Austin, 2005
Supervisor: Randy B. Machemehl
The southern US border is a region of great economic activity. Key port-of-entry
locations on this region usually link twin cities and thus have become facilities were substantial
traffic cross on a daily basis. Seeking to improve the forecast of such flows, the present research
effort focused on development of a new procedure for disaggregate travel modeling of persons in
a bi-national conurbation. This procedure steps away from the conventional approach of studying
separately each side of the international boundary, and thus from modeling ports-of-entry through
the simplistic use of external zones. The new approach extends the model boundaries beyond
international limits, covering the urban areas on both sides, and thus joining the two systems
through the ports-of-entry, which eliminates the need for external zones at these locations.
Developing and validating an international crossing model with mode choice capability is
nevertheless more complex than simply joining together two existing travel demand models
(TDMs). These issues have been considered herein and an initial set of modeling methodologies
have been researched and tested with encouraging findings. The 9/11 events complicated the
border processing element of the commuter trip, however, as the study will show, the model
produced reasonable estimations for 2005, even when originally calibrated with pre-9/11 data.
This study thus represents an unprecedented effort for any border urban area in the
United States or Mexico.
v
Table of Contents
Chapter I. Introduction Background……………………………………………………………………………….….. 1
General objectives of research……………………………………………………..….…. 2
Literature review……………………………………………………………........................ 3
Outline of report……………………………………………………………..………...…...... 6
Chapter II. Model structure Background: the regional model……………………………………………………......... 7
Arrangement of external trips…………………………………………………........ 8
General operation of the regional model………………………………………...... 9
The bi-national OD matrix……………………………………………….………..… 9
Structure of the TTDM…………………….………………………………………...…...…. 10
As the database foundation of the TTDM, the information from the intercept travel
surveys at the ports-of-entry was summarized and electronically coded directly into a simple root
table labeled TripsExp.
In its initial version, the TripsExp table was organized with the following fields:
CLAVE: Record (survey) unique number. ESTACION: Port-of-entry code (Paso del Norte=1, Stanton=2, BOTA=3, Zaragoza=4). SENTIDO: Direction of flow code (northbound=1, southbound=2). HORA: Time of day (hr:min). X_MODE: Code for person border-crossing mode (pedestrian = 1, passenger-vehicle = 2). MODE_ORI: Code for access mode from origin to POE (walk=1, bus=2, taxi=3, auto=4). MODE_DES: Code for access mode from POE to destination (walk=1, bus=2, taxi=3, auto=4). TTAZ_ORI: Origin TAZ number, using the transborder zonal structure. TTAZ_DES: Destination TAZ number, using the transborder zonal structure. PURP: Code for 10 generic transborder trip purposes (HBW=1, HBU=22, HBSc=2,
HBIm=33, HBSh=333, HBO=3, NHSc=4, NHIm=55, NHSh=555, NHO=5). RESID: Place of residency code (Juarez=1, El Paso=2, other MX=3, other US=4). IMM: Dummy variable if trip-end is identified as POE immigration office (yes=1). ExpFACT: Expansion factor specific to POE, direction and crossing mode (refer to Table 1).
For specific modeling processes, some additional fields were later incorporated into
TripsExp table.
Transborder travel characterization: overall volume and modes
Based on the survey information summarized in the TripsExp table, the overall majority of
transborder trips in 1996 had origins and destinations within the El Paso-Juarez region. This is
shown in Table 2, where daily person-crossings are summarized as transborder local (TRBR
local), transborder external-local (TRBR exlo), and transborder through (TRBR thru) trip
categories.
Table 2. 1996 weekday transborder travel by category
The weekday totals shown include both person crossing modes (auto and pedestrian),
both directions of travel, and travel by local area residents as well as non-residents. Figure 8
shows further disaggregation of the weekday transborder travel, by crossing mode, indicating
auto (motorized passenger-vehicles such as automobiles, taxis, vans, pick-ups, motorcycles, etc.)
as the prevailing one.
17
18
pedestrian13%
auto87%
auto100%
pedestrian0%
pedestrian20%
auto80%
a) TRBR local b) TRBR exlo c) TRBR thru
Figure 8. Crossing mode share of transborder person-trips
When analyzing TRBR local trips in particular, the survey shows that 2/3 are done by
Juarez residents, and the rest by El Paso residents as depicted in Figure 9.
El Paso residents
33%
Juarez residents
67%
Figure 9. Proportion of TRBR local trips by traveler place of residence
pedestrian8%
auto92%
pedestrian18%
auto82%
Furthermore, the choice of crossing mode varies somewhat depending on the traveler’s
place of residence as shown in Figure 10. Of the total TRBR local trips by Juarez residents,
nearly 1/5 cross the border as pedestrians, while for El Paso residents this proportion drops to
less than 1/10.
a) Juarez resident b) El Paso residents
Figure 10. Crossing mode of TRBR local trips by traveler’s place of residence
19
Focusing on pedestrian crossings exclusively, four main access modes from-and-to the
ports-of-entry were identified in the 1996 survey: Walk-only (W), Bus (B), Taxi (T), and Auto (A).
From these, up to 16 different access (in-out port-of-entry) combinations are possible regardless
of the direction of travel, which have been summarized as follows:
WW: walk to POE/walk from POE
WB: walk to POE/bus from POE or bus to POE/walk from POE
WT: walk to POE/taxi from POE or taxi to POE/walk from POE
WA: walk to POE/auto from POE or auto to POE/walk from POE
BB: bus to POE/bus from POE
BT: bus to POE/taxi from POE or taxi to POE/bus from POE
BA: bus to POE/auto from POE or auto to POE/bus from POE TT: taxi to POE/taxi from POE TA: taxi to POE/auto from POE or auto to POE/taxi from POE AA: auto to POE/auto from POE
In this regard, the 1996 travel survey shows somewhat different patterns for El Paso and
Juarez residents as depicted in Figures 11 and 12.
0%
10%
20%
30%
40%
50%
60%
WW WB WT WA BB BT BA TT TA AA
EP residentsSB direction
0%
10%
20%
30%
40%
50%
60%
WW WB WT WA BB BT BA TT TA AA
EP residentsNB direction
a) Northbound b) Southbound
Figure 11. Access mode share of pedestrian crossing by El Paso residents
0%
10%
20%
30%
40%
50%
60%
WW WB WT
0%
10%
20%
30%
40%
50%
60%
WW WB WTTT TA AA
JZ residentsNB direction
A TT TA AA
JZ residentsSB direction
WA BB BT BA WA BB BT B a) Northbound b) Southbound
Figure 12. Access mode share of pedestrian crossings by Juarez residents
Bus use in port-of-entry access (BB, BA, WB) is prevalent for Juarez residents that cross
as pedestrians, accounting for close to 70% of pedestrian crossings, compared to 40% for El
Paso residents. El Pasoan pedestrian crossers tend to depend more on the auto and walking
(AA, WA, WW) for access to the ports-of-entry.
Taxi mode, as a means of traveling to the port followed by pedestrian crossing is
practically non-existent for Juarez residents, and very seldom used by El Paso residents; taxi
users rather complete the entire transborder trip in this mode, and thus wait in line and cross
through inspection booths for autos.
In general, access mode patterns appear to show somewhat consistent patterns for both
travel directions once disaggregated by resident type; nevertheless more attention should be
given to the northbound mode selection in addition to the traveler’s place of residence, since
northbound travelers incur the highest delay, and in theory mode could be an important criterion
for selecting round-trip options that reduce travel times for frequent commuters.
Transborder travel characterization: trip purpose
Transborder trip purpose was also characterized by the 1996 travel survey. The
following ten categories were defined, as sub-sets of the generic ones used under model
components 1 and 2:
HBW: home-based-work
HBU: home-based-university
HBSc: home-based-school (non-university)
HBIm: home-based-immigration business (POE immigration offices)
HBSh: home-based-shop
HBO: home-based-other
NHSc: non-home-based-school (includes university)
NHIm: non-home-based-immigration business (POE immigration offices)
NHSh: non-home-based-shop
NHO: non-home-based-other
As depicted in Figure 13, according to the 1996 travel survey, people in the El Paso-
Juarez area cross to the other side of the border to shop and work. The more general purposes
HBO and NHO also account for an important proportion of transborder trips, and these categories
include trips such as visits to family and friends, and site seeing .
20
21
HBW14% HBU
2%
HBSc4%
HBIm1%
HBSh23%
HBO34%
NHSc3%
NHIm1%
NHO12%
NHSh6%
Figure 13. Overall distribution of tranbsorder trip purposes
Similar to transborder mode choice, trip purpose has a notoriously different distribution
depending on the traveler’s place of residence; in addition, significant differences can also be
observed by travel direction. These conditions are depicted in Figures 14 and 15.
HBW10%
HBU0% HBSc
0%HBIm
0%
HBSh20%
HBO59%
NHSc1%
NHIm0%
NHO7%
NHSh3% HBW
7%
HBU0%
HBSc3%
HBIm0%
HBSh16%
HBO33%
NHSc6%
NHIm0%
NHO21%
NHSh14%
a) Southbound (primary) b) Northbound
Figure 14. Distribution of tranbsorder trip purposes by El Paso residents
NHSh1% NHO
2%
NHIm0%
NHSc1%
HBO30%
HBSh32%
HBIm1%
HBSc8%
HBU4%
HBW21%
NHSh10%
NHO21%
NHIm1%
NHSc4%
HBO24%
HBSh20%
HBIm3%
HBSc4%
HBU2%
HBW11%
a) Northbound (primary) b)Southbound
Figure 15. Distribution of tranbsorder trip purposes by Juarez residents
For El Paso residents, the southbound direction is the initial leg of transborder round trips
so southbound tends to characterize the primary trip purposes (Figure 14a), defined as the main
influence driving the need to cross the border. In this case, HBO (visit friends and/or family) and
HBSh (home based shopping) account for the majority (80%) of all trips. For Juarez residents,
the initial leg is northbound so primary trip purposes are those going in the northbound direction
(Figure 15a), and in this case HBO, HBSh, and HBW seem to be the main ones, although HBU
and HBSc (education), account for an important share of trips.
For both El Paso and Juarez’ residents, the return crossing trips seem to have a large
proportion of non-home-based purposes.
Sampled field data: POE queue measurements
Northbound port-of-entry delay, particularly for passenger vehicles, is an important
network attribute and input to the TTDM, therefore the need to characterize it. In this regard,
POE vehicle delay was defined as the average time required to pass through a particular POE. It
would be desirably measured for all vehicles as the total elapsed time from arrival at the end of
the queue to the time of exit from the POE. Direct gathering of this information presented several
challenges, thus a simplified procedure was selected to indirectly estimate this delay during
specific times of the day; this was achieved through the use of the following expression:
di = Nvi*Tp (Eq. 1)
where
di: POE vehicle delay [min], at time i of the
day.
Nvi: Number of vehicles in queue waiting to
cross the POE, at instant i of the day.
Tpi: Inspection processing rate [min/veh], at
time i of the day.
In order to estimate Nvi, the queue length at that instant needed to be establish, therefore
queue lengths were sampled once every hour, during a 12-hour period, and on two weekdays
within the survey period. In addition, the physical layout of the roadway approaches for each of
22
23
the ports-of-entry was studied, focusing on variations in the number of lanes to determine vehicle
capacity under different queue lengths.
Regarding Tpi, processing time at individual primary inspection booths northbound was
sampled randomly on all ports-of-entry, as well as the number of inspection booths opened for
each hour of operation. Thus
Tpi = Nbi * tb (Eq. 2)
where
Nbi: Number of inspection booths opened at
the POE, at time i of the day.
tb: Average processing rate of individual
inspection booths [min/veh].
The result of the described data gathering and its use in Equations 1 and 2 are
summarized in the delay patterns by port-of-entry shown in Figure 16.
0
5
10
15
20
25
30
35
40
07:3
0 a.
m.
08:3
0 a.
m.
09:3
0 a.
m.
10:3
0 a.
m.
11:3
0 a.
m.
12:3
0 p.
m.
01:3
0 p.
m.
02:3
0 p.
m.
03:3
0 p.
m.
04:3
0 p.
m.
05:3
0 p.
m.
06:3
0 p.
m.
PDNBOTAZARA
NB
veh
del
ay (m
in)
BOTA avg = 31.3min
PDN avg = 24.4 min
ZARA avg = 18.8 min
tb = 0.67 min/veh
Figure 16. Northbound crossing delay by time of day for different ports-of-entry
Data gathering was exclusively centered on the northbound direction, since there was
virtually no delay observed for the southbound direction.
Census data: zone level demographics
In order to relate transborder travel patterns to the land use in the region, spatially
disaggregated population and employment information was obtained from sources in both the US
and Mexico. Since the level of detail and the type of fields can vary somewhat when comparing
US and Mexican sources, an effort was made to identify and use data of similar types for both
sides; in addition to simplifying the model, the intention was to facilitate potential TTDM
replication at other border regions.
As previously stated, the survey data used to develop the TTDM model was collected at
POEs, and was developed through application of a brief intercept questionnaire. No socio-
economic characteristics of transborder travelers were obtained, and only information about the
trip was gathered. Trip information included origin-destination locations, and purpose of the
transborder trip; nevertheless, by combining these two elements, it was possible to identify
residence zone location for any HB trip, and thus infer gross socioeconomic information for many
of the surveyed travelers. This characteristic of the survey data related trip making to
demographic information aggregated at the zone level, and thus, the survey information was
related to traffic analysis zones (TAZs), the structure used for travel modeling, as will be
explained in Chapter 4.
US sources
On the US side, population information is available from the US Census Bureau.
Detailed population data is made available by block groups, which represent the basic unit of
area by which information is disaggregated spatially. In addition to population size, population
income reported by household, was selected as an explanatory variable for transborder travel.
Household income was obtained as a block group average from the US Census data, and then
re-estimated for each TAZ. At the time the TTDM model was under development, the latest
population information available was that from the 1990 census, so some extrapolation was
required to estimate 1996 base year conditions.
Employment data on the other hand, was obtained through the Texas Workforce
Commission. This data was summarized for different economic activities, and provided spatially
by employer address. Further refinement of this information by the El Paso Metropolitan Planning
Organization, has produced a GIS where it has been organized into TAZs. At the time the TTDM
model was under development, the employment information available was from 1996.
24
Mexican sources
On the Mexican side, both population and employment information is available from the
Instituto Nacional de Estadística, Geografía e Informática (INEGI), a federal agency in charge of
demographic and economic census planning and implementation. Similar to U.S. Census
practice, INEGI organizes the data in small area units, equivalent to block groups, called Area
Geo-Estadística Básica (AGEB).
Population size and income, as well as total employment is also available by AGEB from
INEGI. Unlike U.S. Census practice though, income is only reported per capita, and thus seeking
consist formats on both sides of the border, special unpublished data from INEGI was requested
to estimate income by household. As will be explained further, per capita income could still be
used on the Mexican side as an explanatory variable for transborder travel models based on
aggregate zonal values. Regarding employment, similar to U.S. Census practice, this data on the
Mexican side is also offered by INEGI as AGEB totals under different economic activities.
All demographic information on the Mexican side of the study area was converted from
AGEBs to TAZ. At the time the TTDM model was under development, the latest demographic
information available from INEGI was that for the 1990 population census, and 1995 economic
census, requiring extrapolation procedures to estimate 1996 base year conditions.
Database adjustment
In order to tie demographic information to the survey data, a second version of the
transborder travel database was constructed as shown in Figure 17, by relating the root table
TripExp to table Demog, that summarizes demographic characteristics by TAZ.
Table Demog was organized with the following fields:
TTAZ_ID: TAZ identification number, under the transborder zonal structure
EMP: Total TAZ employment (96 base year).
POP: Total TAZ population (96 base year).
HHincCAT: Category of average household income for the TAZ (96 base year).
ATYPE: TAZ area type (96 base year).
25
Figure 17. Database design relating POE survey data and zonal demographics
Due to the significant difference in both income levels and development densities
between the two sides of the border, the HHincCAT and ATYPE fields had to be defined under
separate categories for the US and Mexican sides:
In terms of household income, the categories used are the following:
El Paso Juarez
HHincCAT range [US dollars/yr] HHincCAT range [US dollars/yr]
1 $0 - $10,000 1 $0 - $2,171
2 $10,000 - $20,000 3 $2,171 - $6,508
3 $20,000 - $30,000 3 $6,508 - $10,847
4 $30,000 - $40,000 4 $10,847 - $13,018
5 $40,000 - $50,000 5 $13,018 - $19,527
6 >$50,000 6 >$19,527
In 1996, $1 US dollar was equivalent to $7.60 MX pesos.
26
The ranges shown were initially established for optimal trip generation variance when
developing the EPTDM and JZTDM respectively, and have been adopted and tested for the
TTDM component.
Area type is a measure of the activity density, where activity density for a given TAZ i is
defined by:
ActDensi = [POPi + (EMPi * B)] / AREAi (Eq. 3)
where
POPi: Total population at TAZ i.
EMPi: Total employment at TAZ i.
AREAi: Total area (acres) of TAZ i.
B = Total city POP / Total city EMP
The area type categories for each side of the border have been defined as follows:
El Paso Juarez
ATYPE ActDens range ATYPE ActDens range
6 0 - 26 (RURAL)
5 0 - 1 (RURAL) 5 26 - 62
4 1 - 10 4 62 - 100
3 10 - 15 3 100 - 135
2 15 - 50 2 135 - 200
1 > 50 (CBD) 1 > 200 (CBD)
It is estimated that the bi-national region had in 1996 a combined population of 1.8
million, 40% living in El Paso and 60% in Ciudad Juarez. This population is concentrated in an
area of just over 570 square-miles in El Paso, and 95 square-miles in Ciudad Juarez. Overall,
the 1996 average per-capita income in El Paso was $16,500 dollars, while in Juarez was $3,500
dollars. These statistics provide a perspective of the different income and density conditions on
each side of the border.
27
Chapter summary
The current chapter has described the sampled field data necessary to characterize
transborder travel patterns, as well as the demographic information available on both sides of the
US-Mexico border needed to develop the TTDM. This information has been organized into a
preliminary transborder travel database.
A final version of the database was developed by adding multimode transportation
network skims to the root table, including the northbound POE delay estimated from the queue
measurements. This final version of the data base will be discussed in Chapter 4.
28
Chapter IV. Transborder transportation networks
The modeling tasks require, a mathematical representation of the transportation
infrastructure in the transborder study area. This representation of the real networks, is
conventionally constructed as a simplified version of the street layout and other relevant
transportation paths. It is depicted graphically as an interconnected web of links and nodes with
attached attributes, describing specific physical and operational characteristics. Thus, the
information is conveniently stored and organized in a geographic information system (GIS)
environment.
El Paso and Juarez have developed roadway and transit GIS-based networks, as part of
their respective TDMs; in both cases TransCAD has been used as the software platform. Under
the current study an effort was undertaken to join these networks at the ports-of-entry, and to
make network attributes compatible between the cities as well. A description of the joined
networks follows.
The roadway system
The roadway system for the transborder network has been conceived to represent
motorized and non-motorized travel. All current vehicle types were included under motorized
travel, while non-motorized travel for the current version of the TTDM includes only pedestrians.
In this regard, pedestrian mobility has been characterized over the roadway network links,
depending upon the presence or absence of sidewalks and their connectivity. In general, a set of
attributes provided by link, offer different degrees of mobility for each of the modes considered.
Figure 18 shows the joined roadway system for the El Paso-Juarez metropolitan area,
highlighting the location of the POE links created to enable transborder flow modeling. It is
important to point out that as a simplified version of the real system infrastructure, only main
roadways were included in the original JZTDM and EPTDM networks as true on-system links,
while local roads have been aggregated by TAZ through the specification of centroid connector
links (not shown in Figure 18 for clarity). The resulting TTDM network thus carries over this
convention, and represents conditions for base year 1996.
29
GIS link Existing roadway
Figure 18. View of the modeled roadway network between El Paso and Juarez (1996)
Fields and attributes
The roadway system for the transborder network has been summarized under the fields
shown in Table 3. Most of the fields are mode-specific, according to three generic mode
definitions established under the original TDMs:
30
1) Low-occupancy motorized transportation (AUTO). Includes automobiles, motorcycles, vans, and trucks.
2) Non-motorized transportation (NON-MOTORIZED). Only considers walking.
3) High-occupancy public motorized transportation (TRANSIT). Includes public transit, and special bus services (schools and industry).
This differentiation of modes is reflected in the fields defined for the roadway GIS.
Table 3. Fields included in the GIS for the transborder roadway network
Field Description
LINK_ID Link identification number
LENGTH Length in miles
FUNCL Functional class of roadway
ATYPE Area type where roadway is located
DIR Direction of flow: 0=two-way, 1=one-way
SPD_HWY Average 24-hr speed for autos, in mph
TIME_HWY Link's travel time for autos, in minutes
LANES Total number of traffic lanes for autos
CAP_AB Weighted daily capacity [vpd], AB direction
CAP_BA Weighted daily capacity [vpd], BA direction
DIR_WLK Direction of walking flow: 0=two-way, 1=one-way
SPD_WLK Walking speed, in mph
TIME_WLK Link's walking time, in minutes
DIR_BUS Direction of transit flow: 0=two-way, 1=one-way
SPD_BUS Transit speed, in mph
TIME_BUS Link's transit time, in minutes
mod
e
spec
ific
for
mod
em
ode
auto
non-
mot
oriz
e dtra
nsit
c fo
spec
ific
for
spe
ific
r
In the case of the AUTO mode, the link attributes have been summarized according to
the area type where the links are located, and each link’s functional classification. For the El
Paso side, this matrix of attributes is shown in Table 4, while Table 5 shows the matrix of
attributes for the Juarez side.
The suggested values of the attributes come from original TDM development on each
side of the border (Refs 24 and 25).
31
Table 4. Speed and capacity link attributes used for the El Paso side of the network Speed [mph]Capacity [vpl] CBD Fringe Urban E Urban N Urban W Suburb N Suburb W Suburb E Rural Suburb NM Rural NM
For the non-motorized (i.e., pedestrian) mode, speed has a constant value of 3 mph on
any link with continuous sidewalks. In the case of transit, speeds have been obtained for specific
links from the bus schedules in operation. For these two generic modes there are no capacity
constraints.
32
Considerations at the POEs
As previously stated, the original TDM networks have been joined through new links
created at the ports-of-entry, in substitution of previous external station nodes. The time and
corresponding speed values at these new links, have been selected initially to be consistent with
the observed northbound delays (as illustrated in Figure 16, Chapter III).
The transit system
The transit system in a TransCAD environment is developed as a sub-network of the
roadway system, requiring two related GIS coverages built on top of the roadway links: one
depicts the set of transit fixed routes, while the other depicts transit stops for the routes. This
condition allows the use of travel speed and travel time attributes (or any other associated cost)
defined in the underlying roadway system, for both the transit service as well as for pedestrian
access to and from stops.
Figure 19 shows the transit route system for the El Paso-Juarez metropolitan area, as
depicted in centerline format in TransCAD. All routes on each side of the border have been
included: 48 in El Paso and 140 in Juarez.
Fields and attributes
The fields used in the transit system are the following:
Table 6. Fields included under the transborder transit system a) Transit Route GIS
Field Description
ROUTE_ID Route identification number
ROUTE_NAME Route name
HEAD_OFF Headway at off-peak period, in minutes
HEAD_AM Headway at AM peak period, in minutes
HEAD_PM Headway at PM peak period, in minutes
FARE User fare, in dollars
b) Transit Stop GIS
Field Description
STOP_ID Stop identification number
LON Longitud of stop
LAT Latitude of stop
ROUTE_ID Route identification number
PASS_CNT Pass count at the stop location
NODE Roadway system node ID associated to stop
33
roadway link for POE
transit stops
JUAREZ
EL PASO
Figure 19. View of the El Paso-Juarez transit system (base year 1996)
For the 1996 base year transborder network, no formal bi-national transit service was in
operation, yet, as depicted in the zoom-in window in Figure 19, transit stops near the ports-of-
entry were close enough that users on one side of the border could transfer to the transit service
on the other side by walking the length of the POEs. Thus no adjustments to the original transit
system routes were required when building the transborder version.
34
Elements for modeling
Along with the development of the base link-node layout of the transportation networks,
and following state-of-practice, it is necessary to define a compatible traffic analysis zone (TAZ)
structure. From the combination of these GIS elements, optimum travel paths and associated
costs, also referred to as “skims”, can be obtained for all possible TAZ pairs. These skims are
then attached to the trips recorded from the OD field surveys, creating the main data basis for
TDM development.
For the El Paso-Juarez TTDM, the procedure followed was similar in principle, with the
slight difference that skims were only needed for transborder TAZ pairs, in particular TRBR-local,
TRBR-exlo, and TRBR-thru trips as defined in Chapter II, and depicted by the bi-national OD
matrix in Figure 5.
The zone structure
For modeling purposes, TAZs serve as a coarse geographic reference of the origins and
destinations in urban travel. The zoning structure allows the aggregation and simplification of
travel exchange between different geographic locations in the study area. Ideally, the boundaries
of a TAZ should attempt to follow surrounding links of the main roadway system, so travel with
destination to, or originating in the TAZ would use those surrounding links as immediate access
to the transportation system. This was the general premise followed when developing the original
TDMs.
Similar to the context and issues with transportation networks, the zone structures
originally developed for the EPTDM and the JZTDM, were for the most part adopted and carried
over to the TTDM zone structure. The only changes were the deletion of the external TAZs
originally defined at POEs.
Figure 20 shows a view of the TAZ structure established for the El Paso-Juarez
transborder region; the zoom-in window shows as additional detail, the relation maintained
between zone boundaries and the main roadway system. It also exemplifies the convention
followed for centroid and centroid-connector coding. In this regard the TTDM zone structure has
maintained the one centroid per TAZ definition, and therefore the premise of a single point
assignment per TAZ.
In summary, El Paso has 660 internal TAZs, while Juarez has 425, for a total of 1,085
TAZs in the transborder zone structure.
The primary fields and attributes used under the TAZ structure, are those available from
census sources on both sides of border, as explained in Chapter III.
35
centroid connector
TAZ boundary
link
Figure 20. View of the joint El Paso-Juarez TAZ structure (base year 1996)
Development of base skim matrices
Having established a transportation network structure, and a related TAZ structure,
transborder shortest path skims could then be developed for all modes involved. These skims
are intended to portray the most probable paths and associated costs between all transborder
TAZ pairs, and thus help establish average trip information about each OD record obtained with
the POE surveys.
36
As a first step, base skim matrices were developed by only considering a single or
dominant mode for the entire transborder trip. Therefore using the tools available in TransCAD,
base skim matrices were developed for the AUTO, TRANSIT, and NON-MOTORIZED generic
modes. Further rearranging of particular elements of these base skim matrices on specific survey
records (specific transborder OD pairs) would allow proper characterization of their combined
access modes and POE selection, while optimizing data storage requirements (developing skim
matrices for each access mode combination, and for each possible POE would exponentially
increase data storage unnecessarily).
For AUTO generic mode, POE selection was included as an attribute of the path, in addition to
travel time, since the survey showed that these two factors are not fully correlated (i.e., POE
selected is not necessarily the one under the path that yields the least overall transborder travel
time). Therefore, skim matrices have been developed to establish minimum paths separately
through the three POEs available:
SkmA_BOT For this skim matrix, paths were forced to go through the BOTA POE. This
was accomplished by disabling the links for the PDN/STANTON and
ZARAGOZA POEs.
SkmA_PDN For this skim matrix, paths were forced to go through the PDN/STANTON
POE. This was accomplished by disabling the links for the BOTA and
ZARAGOZA POEs.
SkmA_ZAR For this skim matrix, paths were forced to go through the ZARAGOZA POE.
This was accomplished by disabling the connection links for the BOTA and
PDN/STANTON POEs.
Each of these skim matrices for the AUTO generic mode include two fields:
1) TIME: auto travel time (field minimized for optimal path).
2) LENGTH: traveled length.
For TRANSIT generic mode, POE selection was not included as an attribute of the path, since for
this mode the survey showed that POE selection is highly dependent on overall transborder OD
travel time. Instead, the matrices where obtained for different headway configurations at different
periods of the day; even though transit operating speeds remain fairly similar throughout the day,
37
varying headways make a significant difference in overall travel time, and become an important
variable when characterizing trip information of transit users. Thus the developed matrices are:
SkmB_AM This skim matrix has been obtained assuming AM headways only, and
enabling all POE links.
SkmB_OFF This skim matrix has been obtained assuming off-peak headways only, and
enabling all POE links.
SkmB_PM This skim matrix has been obtained assuming PM headways only, and
enabling all POE links.
For each of these skim matrices, the optimal path is selected though the Pathfinder method
(Ref 26) in TransCAD, which makes use of the generalized cost of travel, and thus includes
fare in addition to overall travel time, grouping similar routes and service conditions. The
skim matrices for the TRANSIT generic mode include the following seven fields:
1) FARE: sum of fares charged for the trip
2) IVTT: sum of in-vehicle travel times.
3) WAIT1: initial wait time (1/2 headway).
4) TFER_WAIT: sum of transfer wait times (1/2 headways).
5) TFER_TT: sum of transfer travel times.
6) ACC_TT: access travel time (origin to first transit stop).
7) EGR_TT: egress travel time (last transit stop to destination).
For NON-MOTORIZED generic mode, POE selection was not included as an attribute of the path,
since according to the survey, POE selection in this case is more dependent on total distance
between origin and destination (and travel time). In addition the tolls are considerably lower than
for autos, and thus not a relevant factor for POE selection. Only pedestrian travel has been
included as part of this mode, and only one matrix needed to be developed.
Skm_WLK This matrix includes two fields:
1) TIME_WLK: walking travel time.
2) LENGTH: traveled length (minimized).
38
Database update with combined skims
To be able to portray all transborder transportation possibilities, attributes of the 7 base-
skim matrices needed to be combined as previously suggested; this actually was accomplished
while attaching the skims to the POE survey trip records, a process that allowed the optimization
of data storage needs.
In preparation of the combination step, the observed transborder modes described on
Chapter III, were consolidated according to Table 7.
Table 7. Consolidation of combined modes for transborder trips. OBSERVED CONSOLIDATED TRANSBORDER MODES TRANSBORDER MODEScode description code description
TTAA auto to POE/auto from POE (xing on same auto) AA auto xing/auto access
WB walk to POE/bus from POE or bus to POE/walk from POEBB bus to POE/bus from POE
TT taxi to POE/taxi from POETA taxi to POE/auto from POE or auto to POE/taxi from POEWT walk to POE/taxi from POE or taxi to POE/walk from POE PA ped xing/auto accessWA walk and auto combination accessAA auto to POE/auto from POE
WW walk to POE/walk from POE PW ped xing/walk access
BT bus to POE/taxi from POE or taxi to POE/bus from POE ped xing/bus and autoBA bus to POE/auto from POE or auto to POE/bus from POE combined access
ped xing/bus accessPB
PM
AU
TO x
ing
PED
ESTR
IAN
xin
g
This process resulted in the definition of the following fields:
MNLMODDE: consolidated transborder mode. AA_IVTT: total in-vehicle travel time (minutes) for AA mode. AA_DIST: total in-vehicle distance (miles) for AA mode. PB_FARE: total fare (US dlls) for PB mode. PB_IVTT: total in-vehicle travel time (minutes) for PB mode. PB_INWT: initial wait time (minutes) for PB mode. PB_TRWT: total transfer wait times (minutes) for PB mode. PB_TRTT: total transfer travel time (minutes) for PB mode. PB_ACTT: access travel time (minutes) for PB mode. PB_EGTT: egress travel time (minutes) for PB mode. PA_IVTT: total in-vehicle travel time (minutes) for PA mode. PA_DIST: total in-vehicle distance (miles) for PA mode. PA_OVTT: total out-of-vehicle travel time (minutes) for PA mode.
39
PW_TT: total walk travel time (minutes) for PW mode. PW_DIST: total walk distance (miles) for PW mode. PM_aIVTT: total auto in-vehicle travel time (minutes) for PM mode. PM_aDIST: total auto in-vehicle distance (miles) for PM mode. PM_bFARE: total bus fare (US dlls) for PM mode. PM_bIVTT: total bus in-vehicle travel time (minutes) for PM mode. PM_bINWT: bus initial wait time (minutes) for PM mode. PM_bTRWT: total bus transfer wait time (minutes) for PM modes. PM_bTRTT: total bus transfer travel time (minutes) for PM mode. PM_bACTT: bus access travel time (minutes) for PM mode. PM_bEGTT: bus egress travel time (minutes) for PM mode. A_QT: queue time when crossing by AUTO. A_QC: toll paid when crossing by AUTO. P_QC: toll paid when crossing as PEDESTRIAN.
The combination of the appropriate base-skim attributes, and their attachment to the root
database was done through a computer program, designed in VisualBasic code. Appendix B,
presents a printout of the code. Figure 21 shows the final database design.
Figure 21. Database design relating POE survey data with skims, and zonal demographics
40
Chapter summary
The current chapter has described the characteristics of the transborder GIS-based
transportation networks prepared for the TTDM. In addition, detail explanation has been provided
on skim development to properly characterize the multimodal combination of POE access and
crossing, while optimizing data storage requirements.
A final version of the transborder trip database was developed by adding these
transportation network skims to the root table. With this final database ready, the TTDM
calibration process could then be accomplished.
41
Chapter V. Model development
Overall, the El Paso-Juarez TTDM could be depicted as being in the 4-step-sequential
family, yet its bi-nation transborder condition involves intricacies that have required special
treatment within the component steps; the general logic flow has been previously described on
Chapter II. The present chapter describes the methodology followed in the formulation and
mathematical calibration of each of the TTDM components, as well as in the validation process.
Person trip generation
Trip generation is the initial step in the classical travel demand modeling process,
providing the total number of trip productions and trip attractions for each TAZ in the study area.
In this regard, trip production is conventionally defined as the home end (origin or destination) of
a home-based trip, or the origin of a non-home-based trip; trip attraction on the other hand is
conventionally defined as the non-home end (origin or destination) of a home-based trip, or the
destination of a non-home-based trip. These trip productions and attractions are usually
expressed as person or vehicles trips, and are further categorized by the purpose of the trip.
As an initial approach compatible with the original TDMs on each side of the border, the
trip generation component for the El Paso-Juarez TTDM has been set up to yield daily person-
trips, using a simple cross-classification model structure.
Trip rate estimation
The initial stage in the calibration of the trip generation component required the
development of trip generation rates, based on information from travel surveys; as previously
detailed in Chapter III, under the current project, the travel surveys used were those from the
1996 POE intercepts. Due to the type of survey design and resulting data, an optimal trip-based
approach was to develop the production rates on a per capita basis, under specific categories of
income; in the case of attractions, trip rates could be developed per employee under area type
categories. In both cases, trips were differentiated between northbound (Juarez to El Paso), and
southbound (El Paso to Juarez) flows, as well as between the 10 trip purposes characterized in
Chapter III.
The generation rates have been conceived as follows.
42
Equation 4 defines the estimated daily trip productions per capita, for trip purpose ρ,
crossing direction δ, and income category ι.
(Eq. 4)
Nρδι
Σn=1
popιpρδι =
EFpρδιn
In this expression, EFpρδιn represents the production expansion factor of record n, under
trip purpose ρ, crossing direction δ, and income category ι ; popι represents the total population
under income category ι. Nρδι is the total number of survey records under the specific group ρδι.
Equation 5 defines the estimated daily trip attractions per employee, for trip purpose ρ,
crossing direction δ, and area type category α.
(Eq. 5)
Nρδα
Σn=1
empαaρδα =
EFaρδαn
In this expression, EFaρδαn represents the attraction expansion factor of record n, under
trip purpose ρ, crossing direction δ, and area type category α ; empα represents the total
employment under area type α. Nρδα is the total number of survey records under the specific
group ρδα.
The use of expansion factors directly in the trip rate expressions results from the fact that
the sample number per POE is different and thus, the total daily trips cannot be obtained as a
simple count; each record in the resulting database has an expansion factor attached
corresponding to the POE where the data was collected. The specific values per port-of-entry are
shown on Table 1 (Chapter 3).
Having developed base trip generation rates, the rates where then compared between
adjacent income and area type categories, to establish optimal category aggregation.
The process of grouping category levels for income (in the case of productions), and area
type (in the case of attractions), was based on a statistical comparison of the estimated trip rates
in adjacent categories, in order to identify those that were not significantly different. Categories
with trip rates that were not significantly different, were grouped. These paired comparisons
needed to be established through an iterative process between adjacent categories, and for this
43
purpose a Z-test was used for comparing the estimated trip rates. The mathematical formulation
used was:
Z = (Eq. 6)
µx-µy
sx2 sy
2 Νx Νy
where: Z : test statistic µx, µy : mean trip rate in adjacent categories X and Y sx, sy : trip rate std deviation in adjacent categories X and Y Νx, Νy : Number of records in adjacent categories X and Y
The hypothesis being tested is that the difference between the two mean trip rates is zero
(Ho: µx-µy =0). This hypothesis is rejected at the significance level of 0.05, that is, when |Z|<1.96.
The procedure yielded the final trip rates, and optimal category groups shown on Tables
8 to 11; the rates are expressed as daily person trips, per 1,000 residents in the case of
productions, or 1,000 employees in the case of attractions.
Table 8. Transborder trip production rates in Juarez
Table 9. Transborder trip attraction rates in El Paso
t Juarez to El PasoEl Paso to Juarez
ATYPEAgrRange
High Act 1,2 A 45.92 25.29 NA NA 18.44 7.28 NA NA 18.62 13.95 41.38 37.57Med Ac 3,4 B 43.00 17.23 NA NA 17.29 9.91 NA NA 17.97 13.57 42.52 47.54Low Act 5 C 29.14 29.63 NA NA 9.89 0.00 NA NA 9.73 4.69 11.04 29.27
Table 10. Transborder trip production rates in El Paso
IncomeAgrRange
Low Inc 1,2 A 4.35 3.27 NA NA 1.99 0.05 NA NA 10.62 12.14 19.57 33.58Med Inc 3,4 B 1.17 4.26 NA NA 0.46 0.00 NA NA 3.75 6.08 8.17 16.88High Inc 5,6 C 6.02 7.87 NA NA 0.00 0.00 NA NA 1.68 1.52 15.16 15.25
Table 11. Transborder trip attraction rates in Juarez
Treatment of special generators
In addition to the mean trip rates developed by categories of income and area type, trip
rates were developed for special generators. Special generators are areas of the two cities
(represented at the TAZ level) that show unusual trip generation characteristics; it is better for the
overall predicting accuracy of the model if the trip rates of special generators are developed
separately, since this prevents an outlier effect that might drastically bias otherwise average
values. In the El Paso Juarez TDM these special generator TAZs were the ones with the highest
attractions rates, usually representing large shopping malls or districts, and residential areas
close to the border.
The special generators identified in the El Paso-Juarez study area, according to the
observation frequency on the travel surveys, were:
El Paso Juarez UTEP Thomason Hospital Downtown Juarez PDN POE area El Paso Airport Pronaf area BOTA POE area Waterfill area ZARAGOZA POE area Downtown El Paso Fox Plaza Mall Bassett Center Mall Cielo Vista Mall Sunland Park Mall Zaragoza comercial strip
tt
Juarez to El PasoEl Paso to Juarez
ATYPEAgrRange
High Act 1,2 A 4.64 2.25 NA NA 0.00 0.00 NA NA 4.56 4.17 5.29 4.55Med Act
t3,4 B 8.20 8.72 NA NA 4.37 0.11 NA NA 10.19 10.82 33.93 60.66
Low Ac 5,6 C 5.31 19.96 NA NA 2.61 0.00 NA NA 12.54 6.61 51.32 89.04
(a) Trip generation table (b) Trip distribution matrix
P A
Figure 22. Schematic relationship between trip generation and trip distribution
Do to its simplicity and as an initial approach to test, the trip distribution component of the
TTDM was conceived under the previously described basic concept, using the algorithms of a
doubly constrained gravity model. The limited type of data gathered under the current POE
survey design, restricted the use of other more sophisticated approaches for trip distribution.
Brief review of the gravity model
The doubly constrained version of the classical gravity model has the following form:
Tij = βi*Pi*αj*Aj*f(tij) (Eq. 7) Where:
Tij : Trips produced in zone i and attracted to zone j. Pi : Total trips produced in zone i. βi : Balancing factor for row i (production constraint). Aj : Total trips attracted to zone j. αj : Balancing factor for column j (attraction constraint) f(tij) : Impedance (decreasing) function, based on the travel
time between zone i and zone j.
49
The two constraints that the model is required to meet are that 1) the sum of trips in any
specific row of the matrix should equal the total number of trips produced in that zone, and 2) that
the sum of trips in any specific column should correspond to the number of trips attracted to that
zone. A simplified version of the two conditions are written as follows:
Tij = Pi (Eq. 8) Σ j
Tij = Aj (Eq. 9) Σ i
The expression of both balancing factors βi and αj can thus be derived through simple algebraic
manipulations of Equations 7 to 9. These have the following simplified forms:
βi = (Eq. 10)
1
Σ αj*Aj*f(tij) j
1 αj = (Eq. 11)
Σ βi*Pi*f(tij)
i
As shown here, the balancing factors are interdependent, meaning that the calculation of
one set requires the values of the other set, furthermore suggesting an iterative process until
convergence is achieved. Thus, the practical approach to solving this formulation is to specify
separate singly constrained models to both productions (Equation 12) and to attractions
(Equation 13). The first one is obtained by making αj = 1 since in this case the columns are not
being balanced. Similarly, the second one is obtained by making βi = 1 since in this other case
the rows are the ones not being balanced.
Tij = Pi* (Eq. 12) Aj*f(tij)
Σ Aj*f(tij)
j
Tij = Aj* (Eq. 13) Pi*f(tij)
Σ Pi*f(tij)
i
The solution for the doubly constrained model can then be converged upon by iteratively
applying Equation 12 to balance the productions (rows), and Equation 13 to balance attractions
(columns).
50
Model estimation
For the trip distribution component, the 10 original trip purposes were aggregated into 5
Figure 30. Comparison of NHB gravity-based and travel-survey TLFDs
The previous set of figures show that the modeled TLFDs follow observed patterns to a
fair extent; although some atypical spikes are evident in a couple of trip purposes, differences in
the average travel times were less than 10% in all cases. At this stage no other comparison tests
between model and observed data were considered effective for trip distribution. Validation
results would eventually pinpoint the need to revisit this component.
As previously explained, the fifth OD matrix corresponding to the HBU transborder trip
purpose was obtained directly at the trip generation step, since only one attraction location was
involved. Thus at this point in the model sequence, all matrices were considered ready to be
utilized in the mode choice component.
Crossing mode choice
The objective of this TTDM component is initially to forecast the percent share of
pedestrian and auto crossings at the ports-of-entry, for the different transborder OD pairs. This is
necessarily tied to the type of modes available as access to the ports-of-entry, and their
corresponding networks on each side of the border; therefore for modeling purposes, the
combination of integral crossing modes (including access) have been consolidated into the five
generic ones, as previously defined on Table 7 (Chapter IV):
AA: crossing by AUTO w/ AUTO access (includes walk access on one side)
PB: crossing as PEDESTRIAN w/ BUS access (includes walk access on one side)
PA: crossing as PEDESTRIAN w/ AUTO access (includes walk access on one side)
PW: crossing as PEDESTRIAN w/ WALK access only
PM: crossing as PEDESTRIAN w/ BUS and AUTO combined access
In principle, these modes have been organized in the generalized structure shown in
Figure 31; to model the mode choice process, different MNL and nested-logit configurations have
been explored, which required further refinements to this preliminary tree.
transborder
mode choice
θ
AA PB PA PW PM
Figure 31. General tree structure defined for transborder mode choice
Brief review of the MNL and nested-logit models
The MNL and nested-logit models are two of the mathematical forms used under the
framework of discrete choice analysis, which makes use of the random utility concept:
55
56
Uin = Vin + εin (Eq. 14)
Where: Uin : Utility of alternative i for decision maker n. Vin : Deterministic component of the utility. εin : Random component of the utility (disturbances).
The deterministic component of the utility can be expressed as:
ΣA
a=0 Vin = βina Xina (Eq. 15)
Where:
Xina : Attributes for alternative i. βina : Attribute coefficients. A : Number of attributes for alternative i.
If we assume εin to have a logistic distribution, thus, it has been proven (Ref 27) that for a
set Cn of j alternatives, the probability of choosing alternative i is given by:
(Eq. 16) Pn (i) =
ΣeVin
eVjn
j Cn∈
Equation16 is known as the multinomial version of the logit model or MNL. The nested-
logit version is an extension of the MNL, requiring that each nest η of alternative’s utilities be
included as a weighted term, conventionally known as the logsum:
(Eq. 17) θ logsumη = θ ln Σk C η∈
eVηk
Where:
Cη : Set of alternatives within nest η. k : Number of alternatives in nest. θ : Coefficient of logsum upper nest-level, as exemplified in Figure 31.
The β and θ coefficients in these models are conventionally estimated using maximum-
likelihood theory. BIOGEME (Ref 28) was the particular software used for this purpose.
Model specification and estimation
Having reviewed the mode-choice model types and a preliminary general structure, the
calibration process required the selection of attributes. In this regard, transportation network
attributes such as components of travel time and cost, and aggregate attributes of origin and
destination zones such as income and area type were considered. In this context, multiple
combinations of attributes were tested, under different tree structures; in order to optimize
resources under the current version of the mode choice component, all trip purposes were
aggregated into a single one. In the end, the following structure showed the better fit.
Option 3Route Length boardings transferred Average
[miles] [pax/day] trips boardings/mileBASSETT-BUS ST 11.9 30,269 45% 2,544
Option 4Route Length boardings transferred Average
[miles] [pax/day] trips boardings/mileUTEP-PRONAF 10,687BASSETT-BUS ST 23,315Totals for Op 4 15.6 34,002 30% 2,180
The BRT alternative that yielded the highest demand was Option 4. Yet, Option 1
presented the highest ratio of passengers per mile, although with the highest transfer rate.
77
The previous set of tables show the versatility that the new approach has for evaluating a
variety of alternatives. Such forecasts would not have been possible through the conventional
approach for modeling the border.
Chapter summary
The current chapter has presented a case study where several non-conventional
solutions for transborder transportation have been evaluated. The case study also provided an
opportunity to test and validate the TTDM for a more recent set of conditions on both sides of the
border, updating networks, demographics and land-use. As a result, the exercise helped to
understand the added versatility and precision acquired with this type of approach.
This tool can now be used to asses the influence that one side of the border would
experience in relation to transportation projects and land-use changes on the other side, and thus
represents an important contribution for modeling of bi-national conurbations.
78
Chapter VII. Final analysis
The present research effort focused on development of a new procedure for regional
travel modeling of bi-national conurbations. This procedure steps away from the conventional
approach of studying each side separately, and thus from modeling ports-of-entry through the use
of external zones (as schematically depicted by Figures 44a and 44b). The new approach
extends the model boundaries beyond international limits, covering the urban areas of both sides,
and thus joining the two systems through the ports-of-entry, and eliminating the need for external
zones at these locations (as depicted by Figures 45).
(a) El Paso approach
POE
POE POE
(b) Juarez approach
POE
POE POE
internal zone
external zone
Figure 44. Conventional modeling approaches in the El Paso-Juarez bi-national conurbation
POE
POE
POE
Figure 45. New approach proposed for the bi-national conurbation
79
Crossing at the international ports-of-entry is nevertheless qualitatively different from
travel within a cohesive region. Among other issues, inspection times differ depending upon
direction of crossing, and trip purposes may be different for travelers from Mexico and the US.
Developing and validating an international crossing model with mode choice capability is
therefore more complex than simply joining together two existing TDMs. These issues have been
considered herein and an initial set of modeling methodologies have been researched and tested
with encouraging findings and improvement recommendations that are summarized below.
This study thus represents an unprecedented effort for any border urban area in the
United States or Mexico.
Preliminary conclusions
• The TTDM significantly improves POE forecast accuracy, with a multimodal dimension.
On average, the TTDM reduced the forecasting error on POE links by more than two-
thirds compared to the conventional approach, to less than 10%, including that for
pedestrian-based crossings, an ability not previously available. Although some POEs
under specific modes still showed high error levels (e.g., pedestrian movements at
Zaragosa), the TTDM was able to improve crossing volume forecasts in two different
years (with different land-use conditions), and for an array of access modes, without
having to rely on unrealistic adjustments of crossing delay or capacity.
Regarding the accumulated volume loading at specific network links beyond the POE
areas, the difference between the TTDM and the conventional approach was minimal.
This was due to the larger influence of purely internal travel on each side of the border,
relative to transborder travel.
Therefore the main TTDM contribution centers on POE forecasting accuracy with
multimodal capabilities, and with POE-attribute sensitivity that can be taken full
advantage to properly account for air quality impacts (a significant POE issue, due to the
context of inspection delay and thus motor-vehicle idling), and to study a wider range of
cost-efficient means of moving people across the international border such as adding
inspections booths on existing POEs versus reducing inspections times, or. adding a new
POE location versus implementing transborder transit, to mention a few possibilities.
80
• TTDM is sensitive to POE delay for mode choice estimation.
The TTDM shows appropriate sensitivity to crossing time but although crossing time is
included in the mode choice component, trip generation forecasts are unaffected by
changes in POE delay. This has the potential of yielding unrealistic travel volumes,
although with correct mode shares (proportions). So far the overall error levels shown
are within acceptable range, but it should be noted that this error increased from the 1996
to the 2005 version of the TTDM, over predicting auto-based crossings from less than 1%
to 7%; more importantly this increase in over-prediction error took place when increased
auto-based delay was coded for the 2005 version of the TTDM, which is counter intuitive.
Caution should therefore be exercised for future year forecasts if delay is significantly
different than current trends.
• POE intercept questionnaires are a cost-effective survey instrument for a TTDM.
The POE cordon intercept questionnaire seems to be an ideal survey tool; brief and
apparently non-intrusive to the subject, allows for fast data collection with minimum cost,
both for instrument preparation and application (minimum deployment costs). With a
notoriously high proportion of non-telephone households on the Mexican side, household
surveys end up being more expensive, and yet, could not seem to capture enough
transborder travelers to develop a multimodal, multi-POE bi-directional TTDM (a sample
of close to 50,000 trip records between the 94-El Paso, and the 96-Juarez household
surveys did not yield enough transborder records). Through the POE intercept
questionnaires close to 3,400 trip records were obtained and used to develop the TTDM.
Although sacrificing information from the subject’s household and that on trips beyond the
crossing OD pair, the POE questionnaires yielded enough robustness to validate a
preliminary model. Adding carefully designed questions, could nevertheless incorporate
trip-chaining patterns and more socio-economic attributes of the subject, with little added
cost.
• Current census data fields can be used to develop a TTDM.
To expand the surveyed travel patterns to the entire urban area, disaggregate census
data at the zone level was used under the current format offered by both INEGI and the
US Census Bureau. Although income data for the Mexican side was initially transformed
(per capita to household), further model development showed that this was not
necessary; moreover, there is no need to complement census information with other
81
sources. This is a promising prospect for replicating the effort at bi-national conurbations
along the US-Mexico border. This means following a similar approach with locally
gathered data; direct parameter transferability is not recommended.
Recommendations on further work
• TTDM refinements in trip-generation.
As previously concluded, crossing delay only plays a role in mode share estimation, but
not on actual transborder trip making. The over-estimation of trips for the 1996 and then
for the 2005 version, suggests that the trip-generation TTDM component should be
sensitive to delay, at least for choice trip purposes (e.g., shopping). A simple initial
approach could be the development of adjustment factors based on 1996 and 2005
forecasting errors, correlated to observed delay. A more sophisticated approach can
include discrete choice trip-generation models based on stated-preference (SP) surveys
that could yield trips by time-of-day, or on a time-series evaluation of crossing delay
behavior and crossing counts.
• TTDM refinements in auto-occupancy estimation.
Apparently not as significant as a delay sensitive trip-generation refinement, future
research should at least explore auto-occupancy sensitivity to delay (currently occupancy
is set constant at 1.88pax/veh for TRBR local trips). For this, auto-occupancy needs to
be surveyed at different crossing-delay scenarios to establish the level of correlation.
This was not reviewed for the 2005 validation, so there is a possibility that the auto-based
crossing over-estimation could have been answered at least partially by this factor.
• Refinements in TTDM accuracy by POE location.
As previously concluded, the current version of the TTDM showed a higher than
expected forecasting error for pedestrian crossings at the Zaragosa POE; in addition the
Downtown POEs could benefit from separate treatment, thus requiring more accuracy by
crossing direction on specific POE locations. This would suggest further refinements,
perhaps by disaggregation of the Mode Choice model, specifically exploring separate
models by user-residence location, and by trip purpose. This was somewhat evident by
the initial review of transborder travel patterns from the survey data, as described in
Chapter II. Nevertheless since the initial model calibration tests yielded acceptable
82
parameter criteria, the model was considered appropriate in a preliminary instance. If the
mode choice model is being refined, it would be convenient to have a separate variable
for pedestrian crossing time (P_QT). In 1996 this was not an issue due to insignificant
pedestrian crossing delay, but currently it is, and thus for the 2005 version, this delay had
to be incorporated in the overall pedestrian travel time.
• Refinements on delay estimation procedure.
As described in Chapter III, crossing delay has been estimated trough a combination of
simple processing-rate equations. For long time periods (daily averages) the resulting
delay seem to be consistent, showing little variability; yet, on short periods, the oscillation
of delay has not been characterized and could be significant due to special inspection
procedures, and thus, demand estimation might not be correctly modeled. With the
availability of micro-simulation software, the delay estimation could be enhanced, by
introducing various distributions of speeds, arrival volumes and processing times (as to
emulate variable inspection procedures taking place), in addition to improved reaction
behavior simulation and queue forming patterns for both motorized vehicles and
pedestrians. This has the potential of improving demand estimation precision, in an
iterative loop between macro and micro levels. New software is available that allows for
seamless interaction between these levels
• Refinements on survey design.
Finally, future POE survey efforts should specifically ask for place of residence, and
possibly even household income, household size, and pre/post-crossing trip-chains. This
would require careful wording as to avoid intimidating the subject, and to keep the brief
quality and thus effectiveness identified in this instrument. Related to the trip-generation
improvements, a small set of SP questions could be added. Due to the relative low-cost
of the POE intercept questionnaire, it is recommended as well that consideration be given
for doubling the sample number on any future survey efforts.
83
Appendix A
POE survey questionnaire Sample forms for BOTA (Córdova) POE
84
Form for PEDESTRIANS crossing in the northbound direction
85
ENC
UES
TA D
E VI
AJE
S EX
TER
NO
SR
etén
:__
b6
Para
pea
tone
sSe
ntid
o:de
Juá
rez
a El
Pas
o
vehí
culo
1ve
hícu
lo 2
vehí
culo
31.
¿Aho
rita
de
[1] J
uáre
z1.
¿Aho
rita
de
[1] J
uáre
z1.
¿Aho
rita
de
[1] J
uáre
zdo
nde
vien
e?Lu
gar:
____
____
____
____
____
dond
e vi
ene?
Luga
r:__
____
____
____
____
__do
nde
vien
e?Lu
gar:
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
Cru
cero
:___
____
____
____
____
_C
ruce
ro:_
____
____
____
____
___
Cru
cero
:__
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
__
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
¿Cuá
l?__
____
____
____
____
__¿C
uál?
____
____
____
____
____
¿Cuá
l?__
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
__
2.¿Q
ue a
ctiv
idad
[1] H
ogar
2.¿Q
ue a
ctiv
idad
[1] H
ogar
2.¿Q
ue a
ctiv
idad
[1] H
ogar
real
izó
en e
se[2
] Tra
bajo
o re
laci
onad
ore
aliz
ó en
ese
[2] T
raba
jo o
rela
cion
ado
real
izó
en e
se[2
] Tra
bajo
o re
laci
onad
olu
gar?
[3] E
stud
iar
luga
r?[3
] Est
udia
rlu
gar?
[3] E
stud
iar
[4] C
omer
/soc
ial/d
iver
sión
[4] C
omer
/soc
ial/d
iver
sión
[4] C
omer
/soc
ial/d
iver
sión
[5] C
ompr
as/p
oner
gas
olin
a[5
] Com
pras
/pon
er g
asol
ina
[5] C
ompr
as/p
oner
gas
olin
a[6
] Lle
var o
reco
ger a
una
per
sona
[6] L
leva
r o re
coge
r a u
na p
erso
na[6
] Lle
var o
reco
ger a
una
per
sona
[7] O
tro__
____
____
____
____
____
__[7
] Otro
____
____
____
____
____
____
[7] O
tro__
____
____
____
____
____
__
3.¿C
ómo
llegó
[1] C
amin
ando
exc
lusi
vam
ente
3.¿C
ómo
llegó
[1] C
amin
ando
exc
lusi
vam
ente
3.¿C
ómo
llegó
[1] C
amin
ando
exc
lusi
vam
ente
a es
te c
ruce
?[2
] En
auto
bús
o ru
tera
a es
te c
ruce
?[2
] En
auto
bús
o ru
tera
a es
te c
ruce
?[2
] En
auto
bús
o ru
tera
[3] E
n ta
xi[3
] En
taxi
[3] E
n ta
xi[4
] En
auto
móv
il, v
an o
pic
k-up
[4] E
n au
tom
óvil,
van
o p
ick-
up[4
] En
auto
móv
il, v
an o
pic
k-up
4.¿A
don
de s
e[1
] El P
aso
4.¿A
don
de s
e[1
] El P
aso
4.¿A
don
de s
e[1
] El P
aso
dirig
e ah
orita
?Lu
gar:
____
____
____
____
____
dirig
e ah
orita
?Lu
gar:
____
____
____
____
____
dirig
e ah
orita
?Lu
gar:
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
Cru
cero
:___
____
____
____
____
_C
ruce
ro:_
____
____
____
____
___
Cru
cero
:__
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
__
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
¿Cuá
l?__
____
____
____
____
__¿C
uál?
____
____
____
____
____
¿Cuá
l?__
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
__
5.¿C
uál e
s el
[1
] Reg
resa
r al h
ogar
5.¿C
uál e
s el
[1
] Reg
resa
r al h
ogar
5.¿C
uál e
s el
[1
] Reg
resa
r al h
ogar
mot
ivo
de[2
] Tra
bajo
o re
laci
onad
om
otiv
o de
[2] T
raba
jo o
rela
cion
ado
mot
ivo
de[2
] Tra
bajo
o re
laci
onad
oes
te v
iaje
?[3
] Est
udia
res
te v
iaje
?[3
] Est
udia
res
te v
iaje
?[3
] Est
udia
r[4
] Com
er/s
ocia
l/div
ersi
ón[4
] Com
er/s
ocia
l/div
ersi
ón[4
] Com
er/s
ocia
l/div
ersi
ón[5
] Com
pras
/pon
er g
asol
ina
[5] C
ompr
as/p
oner
gas
olin
a[5
] Com
pras
/pon
er g
asol
ina
[6] L
leva
r o re
coge
r a u
na p
erso
na[6
] Lle
var o
reco
ger a
una
per
sona
[6] L
leva
r o re
coge
r a u
na p
erso
na[7
] Otro
____
____
____
____
____
____
[7] O
tro__
____
____
____
____
____
__[7
] Otro
____
____
____
____
____
____
6.¿C
ómo
se ir
á[1
] Cam
inan
do e
xclu
siva
men
te6.
¿Cóm
o se
irá
[1] C
amin
ando
exc
lusi
vam
ente
6.¿C
ómo
se ir
á[1
] Cam
inan
do e
xclu
siva
men
tede
aqu
í a e
se
[2] E
n au
tobú
s o
rute
rade
aqu
í a e
se
[2] E
n au
tobú
s o
rute
rade
aqu
í a e
se
[2] E
n au
tobú
s o
rute
ralu
gar?
[3] E
n ta
xilu
gar?
[3] E
n ta
xilu
gar?
[3] E
n ta
xi[4
] En
auto
móv
il, v
an o
pic
k-up
[4] E
n au
tom
óvil,
van
o p
ick-
up[4
] En
auto
móv
il, v
an o
pic
k-up
7.H
ora
apro
x__
__:_
____
[ ] A
M ó
[ ]P
M
7.H
ora
apro
x__
__:_
____
[ ] A
M ó
[ ]P
M
7.H
ora
apro
x__
__:_
____
[ ] A
M ó
[ ]P
M
re/9
Puen
te d
e C
órdo
vaFe
cha:
__/N
ovie
m
Form for AUTOMOBILES crossing in the northbound direction
86
ENC
UES
TA D
E VI
AJES
EXT
ERN
OS
eov
Para
veh
ícul
os m
otor
izad
osSe
ntid
o:de
Juá
rez
a El
Pas
o
vehí
culo
1ve
hícu
lo 2
vehí
culo
31.
¿Aho
rita
de
[1] J
uáre
z1.
¿Aho
rita
de
[1] J
uáre
z1.
¿Aho
rita
de
[1] J
uáre
zdo
nde
vien
e?Lu
gar:
____
____
____
____
____
dond
e vi
ene?
Luga
r:__
____
____
____
____
__do
nde
vien
e?Lu
gar:
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
(Tac
har s
ólo
Cru
cero
:___
____
____
____
____
_(T
acha
r sól
oC
ruce
ro:_
____
____
____
____
___
(Tac
har s
ólo
Cru
cero
:___
____
____
____
____
_un
a op
ción
)__
____
____
____
____
__un
a op
ción
)__
____
____
____
____
__un
a op
ción
)__
____
____
____
____
__
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
¿Cuá
l?__
____
____
____
____
__¿C
uál?
____
____
____
____
____
¿Cuá
l?__
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
__
2.¿Q
ue a
ctiv
idad
[1] H
o gar
2.¿Q
ue a
ctiv
idad
[1] H
ogar
2.¿Q
ue a
ctiv
idad
[1] H
ogar
real
izó
en e
se[2
] Tra
bajo
o re
laci
onad
ore
aliz
ó en
ese
[2] T
raba
jo o
rela
cion
ado
real
izó
en e
se[2
] Tra
bajo
o re
laci
onad
olu
gar?
[3] E
stud
iar
luga
r?[3
] Est
udia
rlu
gar?
[3] E
stud
iar
[4] C
omer
/soc
ial/d
iver
sión
[4] C
omer
/soc
ial/d
iver
sión
[4] C
omer
/soc
ial/d
iver
sión
(Tac
har s
ólo
[5] C
ompr
as/p
oner
gas
olin
a(T
acha
r sól
o[5
] Com
pras
/pon
er g
asol
ina
(Tac
har s
ólo
[5] C
ompr
as/p
oner
gas
olin
aun
a op
ción
)[6
] Lle
var o
reco
ger a
una
per
sona
una
opci
ón)
[6] L
leva
r o re
coge
r a u
na p
erso
naun
a op
ción
)[6
] Lle
var o
reco
ger a
una
per
sona
[7] O
tro__
____
____
____
____
____
__[7
] Otro
____
____
____
____
____
____
[7] O
tro__
____
____
____
____
____
__
3.¿A
don
de s
e[1
] El P
aso
3.¿A
don
de s
e[1
] El P
aso
3.¿A
don
de s
e[1
] El P
aso
dirig
e ah
orita
?Lu
gar:
____
____
____
____
____
dirig
e ah
orita
?Lu
gar:
____
____
____
____
____
dirig
e ah
orita
?Lu
gar:
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
(Tac
har s
ólo
Cru
cero
:___
____
____
____
____
_(T
acha
r sól
oC
ruce
ro:_
____
____
____
____
___
(Tac
har s
ólo
Cru
cero
:___
____
____
____
____
_un
a op
ción
)__
____
____
____
____
__un
a op
ción
)__
____
____
____
____
__un
a op
ción
)__
____
____
____
____
__
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
[2] O
tra c
iuda
d o
pobl
ado
¿Cuá
l?__
____
____
____
____
__¿C
uál?
____
____
____
____
____
¿Cuá
l?__
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
____
__
4.¿C
uál e
s el
[1
] Reg
resa
r al h
ogar
4.¿C
uál e
s el
[1
] Reg
resa
r al h
ogar
4.¿C
uál e
s el
[1
] Reg
resa
r al h
ogar
mot
ivo
de[2
] Tra
bajo
o re
laci
onad
om
otiv
o de
[2] T
raba
jo o
rela
cion
ado
mot
ivo
de[2
] Tra
bajo
o re
laci
onad
oes
te v
iaje
?[3
] Est
udia
res
te v
iaje
?[3
] Est
udia
res
te v
iaje
?[3
] Est
udia
r[4
] Com
er/s
ocia
l/div
ersi
ón[4
] Com
er/s
ocia
l/div
ersi
ón[4
] Com
er/s
ocia
l/div
ersi
ón(T
acha
r sól
o[5
] Com
pras
/pon
er g
asol
ina
(Tac
har s
ólo
[5] C
ompr
as/p
oner
gas
olin
a(T
acha
r sól
o[5
] Com
pras
/pon
er g
asol
ina
una
opci
ón)
[6] L
leva
r o re
coge
r a u
na p
erso
naun
a op
ción
)[6
] Lle
var o
reco
ger a
una
per
sona
una
opci
ón)
[6] L
leva
r o re
coge
r a u
na p
erso
na[7
] Otro
____
____
____
____
____
____
[7] O
tro__
____
____
____
____
____
__[7
] Otro
____
____
____
____
____
____
5.Ti
po d
eañ
o:__
__
mar
ca:_
____
____
____
___/
5.
Tipo
de
año:
____
m
arca
:___
____
____
____
_/
5.Ti
po d
eañ
o:__
__
mar
ca:_
____
____
____
___/
ve
hícu
lo
____
____
____
____
__ve
hícu
lo
____
____
____
____
__ve
hícu
lo
____
____
____
____
__
com
bust
ible
:[1
] gas
olin
aco
mbu
stib
le:
[1] g
asol
ina
com
bust
ible
:[1
] gas
olin
a[2
] die
sel
[2] d
iese
l[2
] die
sel
[3] o
tro__
____
____
____
[3] o
tro__
____
____
____
[3] o
tro__
____
____
____
6.#
Ocu
pant
es(in
cluí
r al c
hofe
r):_
____
__6.
# O
cupa
ntes
(incl
uír a
l cho
fer)
:___
____
6.#
Ocu
pant
es(in
cluí
r al c
hofe
r):_
____
__7.
Tipo
de
plac
as[1
] Nac
iona
les
7.Ti
po d
e pl
acas
[1] N
acio
nale
s7.
Tipo
de
plac
as[1
] Nac
iona
les
[2] F
ront
eriz
as[2
] Fro
nter
izas
[2] F
ront
eriz
as[3
] Ext
ranj
eras
, E
stad
o___
____
___
[3] E
xtra
njer
as,
Est
ado_
____
____
_[3
] Ext
ranj
eras
, E
stad
o___
____
___
8.H
ora
apro
x__
__:_
____
[ ] A
M ó
[ ]P
M
8.H
ora
apro
x__
__:_
____
[ ] A
M ó
[ ]P
M
8.H
ora
apro
x__
__:_
____
[ ] A
M ó
[ ]P
M
bre/
96ov
iem
____
/Nha
:Fe
ca
Ret
én:
Punt
e de
Cór
d
Appendix B
VB code to attach skims to survey trip records Combines basic mode skims into multimodal skims
Incorporates attributes into root database
87
Option Compare Database Option Explicit Private Sub Command0_Click() Dim TAB1, TAB2, TAB3, TAB4, TAB5, TAB6, TAB7, TAB20, TAB30 As Recordset Dim MyBASE1 As Database Dim n, mode, mnlmode, est, dir, tazO, tazD, O, D, purp, resid As Integer Dim inc, atyp, m_O, m_D, O_atyp, D_atyp, brEP, brJZ As Integer Dim cl As Long Dim hora, AT, WT, L As Double Dim osgen, dsgen As Variant 'fields for A xing / A access (MNLMODE 1) Dim aa_ivtt, aa_dist As Double 'fields for P xing / B access (MNLMODE 2) Dim pb_fare, pb_ivtt, pb_inwt, pb_trwt, pb_trtt, pb_actt, pb_egtt As Double 'fields for P xing / A access (MNLMODE 3) Dim pa_ivtt, pa_dist, pa_ovtt As Double 'fields for P xing / W access (MNLMODE 4) Dim pw_tt, pw_dist As Double 'fields for P xing / BA access (MNLMODE 5) Dim p5a_ivtt, p5a_dist As Double Dim p5b_fare, p5b_ivtt, p5b_inwt, p5b_trwt, p5b_trtt, p5b_actt, p5b_egtt As Double 'fields for Queue time at POE inspections Dim A_qt, P_qt, A_qc, P_qc As Double Set MyBASE1 = CurrentDb() Set TAB1 = MyBASE1.OpenRecordset("SkmA_BOT", DB_OPEN_TABLE) Set TAB2 = MyBASE1.OpenRecordset("SkmA_PDN", DB_OPEN_TABLE) Set TAB3 = MyBASE1.OpenRecordset("SkmA_ZAR", DB_OPEN_TABLE) Set TAB4 = MyBASE1.OpenRecordset("SkmB_A", DB_OPEN_TABLE) Set TAB5 = MyBASE1.OpenRecordset("SkmB_O", DB_OPEN_TABLE) Set TAB6 = MyBASE1.OpenRecordset("SkmB_P", DB_OPEN_TABLE) Set TAB7 = MyBASE1.OpenRecordset("SkmWLK", DB_OPEN_TABLE) Set TAB20 = MyBASE1.OpenRecordset("MNLext", DB_OPEN_TABLE) Set TAB30 = MyBASE1.OpenRecordset("MNLextSK", DB_OPEN_TABLE) DoCmd.Hourglass False DoCmd.Hourglass True TAB1.Index = "skm1_idx" TAB2.Index = "skm2_idx" TAB3.Index = "skm3_idx" TAB4.Index = "skm4_idx" TAB5.Index = "skm5_idx" TAB6.Index = "skm6_idx" TAB7.Index = "skM7_idx" TAB20.Index = "mnl_idx" TAB20.MoveFirst n = 0 Do While Not TAB20.EOF cl = TAB20!clave mode = TAB20!mode est = TAB20!ESTACION dir = TAB20!SENTIDO hora = TAB20!hora tazO = TAB20!EPTAZ_ORI tazD = TAB20!EPTAZ_DES purp = TAB20!purp resid = Nz(TAB20!resid) inc = TAB20!AToIN_PRD atyp = TAB20!ATYPE_ATT osgen = Nz(TAB20!PRD_SGEN) dsgen = Nz(TAB20!ATT_SGEN) m_O = TAB20!m_O m_D = TAB20!m_D mnlmode = TAB20!mnlmode O_atyp = TAB20!O_atyp D_atyp = TAB20!D_atyp
88
If dir = 1 Then ''''''''''''''''''''''''''''''''''''''''''P XING/W ACCESS SKIMS TAB7.Seek "=", tazO, tazD pw_tt = TAB7!wlk_time pw_dist = TAB7!length_ski ''''''''''''''''''''''''''''''''''''''''''P XING/B ACCESS SKIMS TAB5.Seek "=", tazO, tazD pb_fare = TAB5!fare pb_ivtt = TAB5!invehicle_ pb_inwt = TAB5!initial_wa pb_trwt = TAB5!transfer_w pb_trtt = TAB5!transfer_t pb_actt = TAB5!access_tim pb_egtt = TAB5!egress_tim If hora >= 0.292 And hora <= 0.375 Then TAB4.Seek "=", tazO, tazD pb_fare = TAB4!fare pb_ivtt = TAB4!invehicle_ pb_inwt = TAB4!initial_wa pb_trwt = TAB4!transfer_w pb_trtt = TAB4!transfer_t pb_actt = TAB4!access_tim pb_egtt = TAB4!egress_tim End If If hora >= 0.667 And hora <= 0.75 Then TAB6.Seek "=", tazO, tazD pb_fare = TAB6!fare pb_ivtt = TAB6!invehicle_ pb_inwt = TAB6!initial_wa pb_trwt = TAB6!transfer_w pb_trtt = TAB6!transfer_t pb_actt = TAB6!access_tim pb_egtt = TAB6!egress_tim End If ''''''''''''''''''''''''''''''''''''''''''A and P XING/A ACCESS SKIMS If est = 3 Then 'BOTA TAB1.Seek "=", tazO, tazD aa_ivtt = TAB1!Time aa_dist = TAB1!length_ski If purp = 55 Then AT = 1.27 WT = 13.11 L = 0.627 Else AT = 2.54 WT = 19.66 L = 0.94 End If pa_ivtt = aa_ivtt - AT pa_dist = aa_dist - L pa_ovtt = WT End If If est = 1 Or est = 2 Then 'PDN/STANTON TAB2.Seek "=", tazO, tazD aa_ivtt = TAB2!Time aa_dist = TAB2!length_ski If purp = 55 Then AT = 2.358 WT = 10.15 L = 0.471 Else AT = 3.55 WT = 15.22
89
L = 0.71 End If pa_ivtt = aa_ivtt - AT pa_dist = aa_dist - L pa_ovtt = WT End If If est = 4 Then 'ZARAGOZA TAB3.Seek "=", tazO, tazD aa_ivtt = TAB3!Time aa_dist = TAB3!length_ski If purp = 55 Then AT = 1.12 WT = 11.96 L = 0.567 Else AT = 1.12 WT = 17.94 L = 0.85 End If pa_ivtt = aa_ivtt - AT pa_dist = aa_dist - L pa_ovtt = WT End If ''''''''''''''''''''''''''''''''''''''''''P XING/BA ACCESS SKIMS If dir = 1 Then O = tazO D = tazD End If If dir = 2 Then O = tazD D = tazO End If If est = 3 Then brEP = 124 brJZ = 747 AT = 1.4 L = 0.7 WT = 0 End If If est = 1 Or est = 2 Then brEP = 2 brJZ = 690 AT = 2.5 L = 0.5 WT = 0 End If If est = 4 Then brEP = 340 brJZ = 1075 End If If mnlmode = 5 Then If (m_O = 2 And dir = 1) Or (m_D = 2 And dir = 2) Then 'BUS on JZ side If est = 4 Then AT = 2 L = 1 WT = 24 End If TAB5.Seek "=", O, brEP p5b_fare = TAB5!fare p5b_ivtt = TAB5!invehicle_ p5b_inwt = TAB5!initial_wa p5b_trwt = TAB5!transfer_w p5b_trtt = TAB5!transfer_t p5b_actt = TAB5!access_tim p5b_egtt = TAB5!egress_tim - WT If hora >= 0.292 And hora <= 0.375 Then
90
TAB4.Seek "=", O, brEP p5b_fare = TAB4!fare p5b_ivtt = TAB4!invehicle_ p5b_inwt = TAB4!initial_wa p5b_trwt = TAB4!transfer_w p5b_trtt = TAB4!transfer_t p5b_actt = TAB4!access_tim p5b_egtt = TAB4!egress_tim - WT End If If hora >= 0.667 And hora <= 0.75 Then TAB6.Seek "=", O, brEP p5b_fare = TAB6!fare p5b_ivtt = TAB6!invehicle_ p5b_inwt = TAB6!initial_wa p5b_trwt = TAB6!transfer_w p5b_trtt = TAB6!transfer_t p5b_actt = TAB6!access_tim p5b_egtt = TAB6!egress_tim - WT End If TAB1.Seek "=", brJZ, D p5a_ivtt = TAB1!Time - AT p5a_dist = TAB1!length_ski - L End If If (m_O = 2 And dir = 2) Or (m_D = 2 And dir = 1) Then 'BUS on EP side If est = 4 Then AT = 3.4 L = 1.7 WT = 10 End If TAB5.Seek "=", brJZ, D p5b_fare = TAB5!fare p5b_ivtt = TAB5!invehicle_ p5b_inwt = TAB5!initial_wa p5b_trwt = TAB5!transfer_w p5b_trtt = TAB5!transfer_t p5b_actt = TAB5!access_tim - WT p5b_egtt = TAB5!egress_tim If hora >= 0.292 And hora <= 0.375 Then TAB4.Seek "=", brJZ, D p5b_fare = TAB4!fare p5b_ivtt = TAB4!invehicle_ p5b_inwt = TAB4!initial_wa p5b_trwt = TAB4!transfer_w p5b_trtt = TAB4!transfer_t p5b_actt = TAB4!access_tim - WT p5b_egtt = TAB4!egress_tim End If If hora >= 0.667 And hora <= 0.75 Then TAB6.Seek "=", brJZ, D p5b_fare = TAB6!fare p5b_ivtt = TAB6!invehicle_ p5b_inwt = TAB6!initial_wa p5b_trwt = TAB6!transfer_w p5b_trtt = TAB6!transfer_t p5b_actt = TAB6!access_tim - WT p5b_egtt = TAB6!egress_tim End If TAB1.Seek "=", O, brEP p5a_ivtt = TAB1!Time - AT p5a_dist = TAB1!length_ski - L End If End If If mnlmode <> 5 Then If est = 4 Then AT = 2 L = 1
91
WT = 24 End If TAB5.Seek "=", O, brEP p5b_fare = TAB5!fare p5b_ivtt = TAB5!invehicle_ p5b_inwt = TAB5!initial_wa p5b_trwt = TAB5!transfer_w p5b_trtt = TAB5!transfer_t p5b_actt = TAB5!access_tim p5b_egtt = TAB5!egress_tim - WT If hora >= 0.292 And hora <= 0.375 Then TAB4.Seek "=", O, brEP p5b_fare = TAB4!fare p5b_ivtt = TAB4!invehicle_ p5b_inwt = TAB4!initial_wa p5b_trwt = TAB4!transfer_w p5b_trtt = TAB4!transfer_t p5b_actt = TAB4!access_tim p5b_egtt = TAB4!egress_tim - WT End If If hora >= 0.667 And hora <= 0.75 Then TAB6.Seek "=", O, brEP p5b_fare = TAB6!fare p5b_ivtt = TAB6!invehicle_ p5b_inwt = TAB6!initial_wa p5b_trwt = TAB6!transfer_w p5b_trtt = TAB6!transfer_t p5b_actt = TAB6!access_tim p5b_egtt = TAB6!egress_tim - WT End If TAB1.Seek "=", brJZ, D p5a_ivtt = TAB1!Time - AT p5a_dist = TAB1!length_ski - L End If 'wait times and tolls If est = 3 Then 'BOTA A_qc = 0 P_qc = 0 If dir = 1 Then A_qt = 26 Else A_qt = 0 P_qt = 0 If (hora >= 0.292 And hora <= 0.375) Then '7-9 If dir = 1 Then A_qt = 41 Else A_qt = 0 P_qt = 0 End If If (hora > 0.375 And hora <= 0.5) Then '9-12 If dir = 1 Then A_qt = 26 Else A_qt = 0 P_qt = 0 End If If (hora > 0.5 And hora <= 0.75) Then '12-6PM If dir = 1 Then A_qt = 28 Else A_qt = 0 P_qt = 0 End If If (hora > 0.75 And hora < 0.9167) Then '6-10PM If dir = 1 Then A_qt = 26 Else A_qt = 0 P_qt = 0 End If End If If est = 1 Or est = 2 Then 'PDN A_qc = 1.25 P_qc = 0.25 If dir = 1 Then A_qt = 8 Else A_qt = 0 P_qt = 0 If (hora >= 0.292 And hora <= 0.375) Then '7-9 If dir = 1 Then A_qt = 10 Else A_qt = 0 P_qt = 0
92
End If If (hora > 0.375 And hora <= 0.5) Then '9-12 If dir = 1 Then A_qt = 8 Else A_qt = 0 P_qt = 0 End If If (hora > 0.5 And hora <= 0.75) Then '12-6PM If dir = 1 Then A_qt = 10 Else A_qt = 0 P_qt = 0 End If If (hora > 0.75 And hora < 0.9167) Then '6-10PM If dir = 1 Then A_qt = 8 Else A_qt = 0 P_qt = 0 End If End If If est = 4 Then A_qc = 1.25 P_qc = 0.25 If dir = 1 Then A_qt = 5 Else A_qt = 0 P_qt = 0 If (hora >= 0.292 And hora <= 0.375) Then '7-9 If dir = 1 Then A_qt = 10 Else A_qt = 0 P_qt = 0 End If If (hora > 0.375 And hora <= 0.5) Then '9-12 If dir = 1 Then A_qt = 5 Else A_qt = 0 P_qt = 0 End If If (hora > 0.5 And hora < 0.667) Then '12-4PM If dir = 1 Then A_qt = 6 Else A_qt = 0 P_qt = 0 End If If (hora >= 0.667 And hora <= 0.75) Then '4-6PM If dir = 1 Then A_qt = 8 Else A_qt = 0 P_qt = 0 End If If (hora > 0.75 And hora < 0.9167) Then '6-10PM If dir = 1 Then A_qt = 6 Else A_qt = 0 P_qt = 0 End If End If End If With TAB30 .AddNew !clave = cl !mode = mode !ESTACION = est !SENTIDO = dir !hora = hora !EPTAZ_ORI = tazO !EPTAZ_DES = tazD !purp = purp !resid = resid !AToIN_PRD = inc !ATYPE_ATT = atyp !PRD_SGEN = osgen !ATT_SGEN = dsgen !m_O = m_O !m_D = m_D !mnlmode = mnlmode !O_atyp = O_atyp !D_atyp = D_atyp If dir = 1 Then !aa_ivtt = aa_ivtt !aa_dist = aa_dist !pb_fare = pb_fare !pb_ivtt = pb_ivtt
93
!pb_inwt = pb_inwt !pb_trwt = pb_trwt !pb_trtt = pb_trtt !pb_actt = pb_actt !pb_egtt = pb_egtt !pa_ivtt = pa_ivtt !pa_dist = pa_dist !pa_ovtt = pa_ovtt !pw_tt = pw_tt !pw_dist = pw_dist !p5a_ivtt = p5a_ivtt !p5a_dist = p5a_dist !p5b_fare = p5b_fare !p5b_ivtt = p5b_ivtt !p5b_inwt = p5b_inwt !p5b_trwt = p5b_trwt !p5b_trtt = p5b_trtt !p5b_actt = p5b_actt !p5b_egtt = p5b_egtt !A_qt = A_qt !A_qc = A_qc !P_qt = P_qt !P_qc = P_qc End If .Update End With n = n + 1 TAB20.MoveNext If n = 1000 Then MsgBox "van 1,000" If n = 10000 Then MsgBox "van 10,000" Loop TAB1.Close TAB2.Close TAB3.Close TAB4.Close TAB5.Close TAB6.Close TAB7.Close TAB20.Close TAB30.Close DoCmd.Hourglass False MsgBox "Terminó!...Viajes Totales:" & n End Sub
INEGI: Instituto Nacional de Estadística, Geografía e Informática
JZTDM Juarez Travel Demand Model
MNL: Multinomial logit
NHB: Non-home-based trip purpose
OD: Origin-destination
POE: Port-of-entry
SP: Stated preference
TDM: Travel demand model
TLFD: Travel-length frequency distribution
THRU: Through-travel trip purpose
TTDM: Transborder travel demand model
TRBR: Transborder trip category
UE: User-Equilibrium
VMT: Vehicle-miles traveled
VHT: Vehicle-hours traveled
WESML: Weighted Exogenous Sample Maximum Likelihood
97
References
1. Traffic maps for year 2000; El Paso and Laredo Districts. Texas Department of Transportation, Transportation Planning and Programming Division.
2. US Department of Transportation, Bureau of Transportation Statistics based on data from US Customs Service, Office of Field Operations, Operations (March of 2004), www.bts.programs/international/border_crossing_entry_data/us_mexico/index.html
3. “Forecasting Freight Traffic Between the US and Mexico”, Y. Fang, R. Harrison, and H. Mahmassani. University of Texas at Austin, Center for Transportation Research 1996 (CTR 1319-2).
4. “A Methodology for Determining the Freight Border Transportation Impact of the North American Free Trade Agreement”, C. Strong, R. Harrison, and H. Mahmassani. University of Texas at Austin, Center for Transportation Research 1996 (CTR 1319-4).
5. “Transportation Issues and the US-Mexico Free Trade Agreement”, R. Harrison, L. Boske, C. Lee, J. McCray. University of Texas at Austin, Center for Transportation Research 1997 (CTR 1319-6F).
6. “Texas’ Role as a U.S.-Mexico Trade Gateway”, A. Weissman, R. Harrison, and M. Trevino. University of Texas at Austin, Center for Transportation Research 1995 (CTR 2932-3F).
7. “Analysis of US-Mexico Traffic within Texas”, R. Harrison, and A. Weissman. University of Texas at Austin, Center for Transportation Research 1995 (CTR 2932-2).
8. “Overview of the Texas-Mexico Border: Background”, J. Hanania, A. Weissman, and R. Harrison. University of Texas at Austin, Center for Transportation Research 1994 (CTR 1976-1).
9. “US-Mexico Trade and Transportation: Corridors, Logistics Practices, and Multimodal Partnerships”, L. Boske, and R. Harrison. University of Texas at Austin, Lyndon Baynes Johnson School of Public Affairs 1995 (LBJ Report 113).
10. “Transborder Traffic and Infrastructure Impacts on the City of Laredo, Texas”, C. Said, R. Harrison, and R. Hudson. University of Texas at Austin, Center for Transportation Research 1993 (CTR 1312-1).
11. “Executive Summary – Truck Traffic in Laredo, Texas: A Case Study of Issues and Remedies”, R. Harrison. University of Texas at Austin, Center for Transportation Research 1993 (CTR 1312-3F).
12. “Effect of the North American Free Trade Agreement on the Transportation Infrastructure in the Laredo-Nuevo Laredo Area”, R. Espinosa, R. Harrison, F. McCullough. University of Texas at Austin, Center for Transportation Research 1993 (CTR 1312-2).
13. “Commercial Surface Transportation Efficiency at the Texas/Mexico Border: A Look at Laredo”, M. Sassin, R. Harrison, and L. Boske. Southwest Region University Transportation Center, College Station 1995 (SWUTC-95-465640-1).
14. “Texas Model Border Crossing Project: Retro-Fit Report”, C. Quiroga, E. Kraus, B. Stockton, and R. Harrison. University of Texas at Austin, Center for Transportation Research 2003 (CTR 9014-1).
15. “A Prototype Southern Border Facility to Expedite NAFTA Trucks Entering the United States”, B. Bockner, B. Stockton, D. Burke, and R. Harrison. Texas A & M University, Texas Transportation Institute 1996 (MS-7796).
16. “Briefing Document on Texas Model Border Crossing Project”, R. Harrison, and B. Stockton. University of Texas at Austin, Center for Transportation Research 2002 (CTR 9014-P5).
17. “Texas-Mexico Toll Bridge Study: Summary Report”, B. F. McCullough, R. Harrison, and A. Weissman. University of Texas at Austin, Center for Transportation Research 1994 (CTR 1976-6F).
18. “Overview of the Texas-Mexico Border: Assessment of Traffic Flow Patterns”, A. Weissman, M. Martello, J. Hanania, M. Shamieh, C. Said, B. F. McCullough, and R. Harrison. University of Texas at Austin, Center for Transportation Research 1994 (CTR 1976-3).
19. “Overview of the Texas-Mexico Border: Database”, J. Hanania, A. Weissman, and R. Harrison. University of Texas at Austin, Center for Transportation Research 1993 (CTR 1976-2).
20. “San Diego Region-Baja California Cross-Border Transportation Study”, Final Report prepared for the San Diego Association of Governments, by Parsons Transportation Group, in association with CIC Research and Colegio de la Frontera Norte (November 2000).
21. “Northern Border Crossing Noteworthy Practices Reference Guide >NP-19”, online resource guide prepared by the Federal Highway Administration (page last modified on July 20, 2004), www.fhwa.dot.gov/uscanada/index.htm
22. “El Paso/Dona Ana Urban Transportation Study, 1994 Base Year Travel Demand Model Calibration Summary for Non-Attainment Air Quality Conformity Analysis”, Report prepared by the Texas Department of Transportation, Transportation Planning and Programming Division (October 1998)
23. “Port-of-entry statistics for El Paso”, Report Prepared by the City of El Paso Planning Department (July of 2005)/
24. “El Paso 1997 Base Year Model”, Report prepared by the Texas Department of
Transportation, Transportation Planning and Programming Division (November 2000)
25. “Estudio Integral de Transporte/Juarez Data Analysis and Model Development”, Report series prepared by Instituto Municipal de Investigación y Planeación (1998)
26. “Travel Demand Modeling with TransCAD”, Caliper Corporation (February 2004)
27. “Discrete Choice Analysis: Theory and Application to Travel Demand”, M. Ben-Akiva, S. R.
Lerman. MIT Press (1985)
28. “An introduction to BIOGEME (version 1.0)”, M. Bierlaire, D. Bolduc, M.-H. Godbout (July 2004), http://roso.epfl.ch/biogeme