UNITED REPUBLIC OF TANZANIA Prime Ministers Office for Regional Administration and Local Government The Dar es Salaam City Council ANNEX VOLUME 4 DEMAND MODELING AND FORECASTING CHAPTER 4 Draft Final Report CONSULTANCY SERVICES FOR THE CONCEPTUAL DESIGN OF A LONG TERM INTEGRATED DAR ES SALAAM BRT SYSTEM AND DETAILED DESIGN FOR THE INITIAL CORRIDOR Dar es Salaam April, 2007
36
Embed
Prime Ministers Office for Regional Administration …Prime Ministers Office for Regional Administration and Local Government The Dar es Salaam City Council ANNEX VOLUME 4 DEMAND MODELING
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
UNITED REPUBLIC OF TANZANIA Prime Ministers Office for Regional
Administration and Local Government
The Dar es Salaam City Council
ANNEX VOLUME 4 DEMAND MODELING AND FORECASTING CHAPTER 4 Draft Final Report
CONSULTANCY SERVICES FOR THE CONCEPTUAL DESIGN OF A LONG TERM INTEGRATED DAR ES SALAAM BRT SYSTEM AND DETAILED DESIGN FOR THE INITIAL CORRIDOR
2. DEMAND MODEL CHARACTERISTICS...................................................................2
2.1. GEOGRAPHIC BASE.........................................................................................................5 2.2. ROAD NETWORK.............................................................................................................5 2.3. ZONING AND SOCIO-ECONOMIC INFORMATION ..........................................................9 2.4. TRANSPORTATION MODES AND SERVICES .................................................................11 2.5. FARE STRUCTURE .........................................................................................................12 2.6. TRAVEL TIME SUBJECTIVE VALUE .............................................................................14 2.7. PUBLIC TRANSPORTATION PASSENGER DEMAND......................................................15 2.8. MODEL CALIBRATION..................................................................................................16
3. POPULATION AND DEMAND GROWTH FORECAST.........................................18
3.1. POPULATION EXPANSION.............................................................................................18 3.2. DEMAND EXPANSION....................................................................................................19 3.3. EXPANSION FACTORS ...................................................................................................22
4.1. SCENARIO DEFINITION.................................................................................................24 4.1.1. BASE AND FUTURE YEARS TO BE EVALUATED...........................................................24 4.1.2. DART FARE STRUCTURE ............................................................................................24 4.2. RESULTS ........................................................................................................................25
TABLE 1 DAR ES SALAAM CITY SUBDIVISION SUMMARY ____________________________________9 TABLE 2 FARE STRUCTURE SCENARIO FOR BRT __________________________________________14 TABLE 3 POPULATION INDEX PER INCOME GROUP_________________________________________18 TABLE 4 YEARLY POPULATION GROWTH FORECAST RESULTS _______________________________19 TABLE 5 DSM TRANSPORTATION NETWORK ORIGIN DESTINATION MATRICES FORECASTED (TOTAL
TRIPS IN THE PEAK HOUR) _______________________________________________________22 TABLE 6 YEAR EXPANSION FACTOR CALCULATION _______________________________________23 TABLE 7 FARE STRUCTURE EVALUATED FOR OPERATIONAL DESIGN ___________________________25 TABLE 8 GENERALIZED TRAVEL TIME RESULTS __________________________________________25 TABLE 9 PEAK HOUR RESULTS FARE STRUCTURE SELECTED ________________________________26 TABLE 10 DEMAND RESULTS PER STATION SCENARIO 2009 . PEAK HOUR ______________________27
iii
LIST OF FIGURES
FIGURE 1 EXAMPLE FOR DAR ES SALAAM CITY GEOGRAPHIC BASE-CBD AREA __________________5 FIGURE 2 EXAMPLE FOR ROAD NETWORK SYSTEM- CBD AREA _______________________________6 FIGURE 3 DAR ES SALAAM MODELING NETWORK – GENERAL Y DETAIL VIEWS ___________________8 FIGURE 4 TRANSPORT ZONES DIVISION _________________________________________________10 FIGURE 5 ZONING DETAIL WITH ZONE CODES ____________________________________________11 FIGURE 6 EMME2 MODELING ENVIRONMENT ____________________________________________16 FIGURE 7 LINEAR REGRESSION OF DEMAND MODEL RESULTS AT PEAK HOUR ___________________17
iv
ACRONYMS AND ABBREVIATIONS
DSM: Dar es Salaam
DCC: Dar es Salaam City Council
PMU: Project Management Unit
PMORALG: Prime Minister’s Office for Local Government and Regional
Administration
TTSV: Travel Time Subjective Value
GTC: Generalized Travel Cost
TZS: Tanzanian Shillings
GIS: Geographic Information System
CBD: Central Business District
DART: Dar Rapid Transit
BRT: Bus Rapid Transit
VBASu: Velocity Boarding and Alighting Survey
ODSu: Origin Destination Survey
Pax: Passenger
1 Demand Modeling and Forecasting
1. INTRODUCTION
The process of planning a transportation system is in great part supported on the
availability of a comprehensive and realistic method of depicting the existing
conditions, regarding passenger demand values, travel times, corridor loads and
vehicle fleet, among many others. Giving solution to this requirement, demand
forecasting models have been implemented to approximate into a simulation the
reality of transportation. This simulation is the combination of the preparation of a
city’s simplified road network, its public transportation lines and routes, and an
existing trips distribution between production and attraction regions, the later
mainly provided by a origin destination trip matrix. The validity of the model is
verified and later adjusted by comparing the existing situation with the simulated
on certain control points, containing information obtained from field surveys.
The present volume explains the process undergone for constructing the demand
forecast model for DSM.
2 Demand Modeling and Forecasting
2. DEMAND MODEL CHARACTERISTICS
As aforementioned, the demand forecasting process for a transportation system
can be summarized in three fundamental stages: potential demand calculation
(origin destination matrix), transportation supply simulation (transport and road
networks) and itinerary and services choice (models of modal distribution and
allocation).
A model makes understanding reality easier. By assuming simplifications in a
complex phenomenon, one can select the most relevant aspects for that
observation and assure that the relation among those characteristics is set in a
way that reflects reality. These aspects are not exactly the same as reality, but
correlate well with our understanding of it. One can then use the model for
evaluation and planning – either long-term or short-term planning.
The transport demand modeling is based on the analysis and evaluation of trip
strategies/alternatives between each origin and destination pair of zones. This
strategy or choice for each user in the transportation network, depends as much
on the transport supply (routes and frequencies), as of the costs of each possible
combination of ways from the origin of the trip to the final destiny. For calculating
the trip cost the time spent on each stage of the trip should be considered as well
as the monetary cost of accessing each one of the of public transport vehicles
boarded. The total times of trip can be disturbed in:
Access from the origin to a public transportation stop or station.
Waiting time.
In vehicle travel time.
Access between stops in case of transfers.
Access from last stop to final destination.
For modeling algorithm purposes, the weigh process performed for the different
travel time and monetary cost components is expressed mathematically by an
3 Demand Modeling and Forecasting
equation, known in the transportation engineering as the Generalized Travel
Cost, in the following form:
TTSVFareTwafwaTwfwTvGTC +++= **
Where: GTC = Generalized Travel Cost
Tv = In vehicle time
fw = Walking time factor weighed against in vehicle time
Tw = Walking time
fwa = Waiting time weigh factor
Twa = Waiting time
Fare = Transportation fare
TTSV = Travel time subjective value for each user
The analysis and election is based on the best option available for each trip to
complete the travel desire from origin to destination by comparing the
generalized cost of commuting, expressed in time units and choosing
accordingly.
Before the analysis is done, a crucial stage is the TTSV estimation and the
population structure for which it will be applied, this value represents the
equivalency in money of the travel time unit (e.g. X of TZS per minute traveled).
This value is either obtained from stated preferences surveys discriminating the
different income population or by determining an average population income
based on local standards.
Other elements such as waiting, in vehicle and walking time are calculated by the
model algorithm. Waiting times and boarding probabilities are estimated based
on the public transportation routes or available vehicles frequencies per route
related to the combined routes frequency available on a single stop or boarding
point.
4 Demand Modeling and Forecasting
By bringing together all the results from the modeling algorithm (boarding
probabilities, generalized costs and travel times, modal choice for each trip within
the origin destination matrix) the simulation then produces the results required
such as operational information for registered transportation means and
passenger demand volumes on the entire network for each mode registered.
Structuring the model for optimal results requires the definition and consecution
of the following:
Geographic base map, positioning the city in global coordinates for
accurate model referencing.
Road network updated to existing accessibility conditions along which the
transportation route network will be distributed.
Regional division in transportation zones for travel demand representation.
Existing public transportation routes itineraries, operational frequencies,
vehicle typology and authorized fares applied.
Weigh factors for generalized cost calculation. Estimation of TTSV.
Origin destination trip matrix stating the travel desires between the
transportation zones defined previously within the area of analysis.
The entire modeling process for the DSM transportation network and later public
transportation demand simulation was performed using the software emme2,
developed by INRO Consultants. Further analysis and data information were
simplified by simple calculations on spreadsheets and detailed revision and
understanding of the process done should be completed before manipulating the
simulation model information platform.
TransCAD GIS1 software, developed by Caliper Corporation, served as the
primary GIS analysis tool used.
1 TransCAD GIS, Geographic Information System software, Caliper Corporation.
5 Demand Modeling and Forecasting
2.1. GEOGRAPHIC BASE
For designing a transportation system involving road infrastructure definition and
bus operations details, a thorough geographic base should be available and
prepared as reference tool to the existing conditions. Detailed and updated
cartography is required, for the present project, the PMU/DCC provided a
geographic database, updated from early year 2000 based on digitalization form
aerial photographs of DSM.
The following figure is the geographic base of Dar es Salaam City.
Figure 1 Example for Dar es Salaam City Geographic Base-CBD Area
2.2. ROAD NETWORK
Though the geographic base was updated enough to the existing conditions, the
road network was incomplete and lacked detail given the precision and
6 Demand Modeling and Forecasting
refinement required for structuring the demand forecasting model and the
simulation platform for emme2 to run appropriately. This road network, was
updated and corrected based on the digital cartography (geographic base)
obtained from the client (DCC) (see figure 2). Main characteristics like number of
lanes per road, street names and directions, etc. were added, when possible, as
backup data for the digital geographic information file. No further analysis was
done since the available information such as road condition, hierarchy, and
infrastructure improvement, among others, was either never available or not
included/processed in the database obtained.
Figure 2 Example for Road Network System- CBD Area
7 Demand Modeling and Forecasting
Summarizing, the road network prepared for the simulation model contains (see
figure 3):
271 Centroids. They are virtual connections that represent the transport
zones where the travels is generated or attracted.
8838 links of the road network. Represent the road sections between
intersections.
4443 links used by transportation routes.
Observed Daladala speed flow obtained from VBASu2 introduced in the
model as a link attribute.
2 Please refer to Annex Volume 3 – Data Collection and Calibration
8 Demand Modeling and Forecasting
Figure 3 Dar es Salaam Modeling Network – General y detail Views
9 Demand Modeling and Forecasting
2.3. ZONING AND SOCIO-ECONOMIC INFORMATION
Dar es Salaam is divided into three municipalities. The three municipalities are
Kinondoni on the Northern region, Ilala on the central and southeastern regions
and Temeke on the south and southwestern areas, each one having regional
autonomy and local administration.
For administration and management issues, the municipalities subdivide into
wards on a first stage and then into sub wards (see table 1). The analysis done
used the existing division and adjusted it to the simulation model’s requirements
of representing the areas based on its attractor or generator of trips
characteristics.
Table 1 Dar es Salaam City Subdivision Summary Municipality Wards Subwards Transportation Zones Kinondoni 27 131 115
Ilala 22 101 70 Temeke 24 156 86
Total 73 388 271
The definition of transportation zones was mainly supported on the existing
subward division (figure 4), with participation form the ward level division. The
absence of a street numeration or nomenclature made difficult the definition or
assumption of a new division scheme and considering that this division was done
based on regional characteristics, administrative alikeness and/or land use
similarities, the zoning process was based solely on sub ward level on the
urbanized areas and ward level on the city’s outskirts.
10 Demand Modeling and Forecasting
Figure 4 Transport Zones Division
Nevertheless the advance the sub ward and ward divisions offered to the model
set up process, detailed analysis had to be carried out particularly on zones too
big to be considered a homogeneous demand zone. The process then was
focused to the re-division of these zones, detailing the precision on demand
forecasting and transportation supply and coverage3, particularly on the CBD and
along DART corridors on Morogoro Road and Kawawa Road.
Socio-economic information is classified following ward-subward division and
mainly supported on the national census from 20024, National Bureau of
Statistics and World Bank information, basically on the matters of modal choice
shares, average road conditions, employment levels and activities, poverty,
population growth, etc.
3 See Annex Volume 3 - Field Surveys and Data Calibration Section 4.9 4 http://www.tanzania.go.tz/census/census
11 Demand Modeling and Forecasting
As a result, 271 zones were defined for structuring the simulation model, and are
discriminated as seen in Table 1 and figure 5.
Figure 5 Zoning Detail with Zone Codes
2.4. TRANSPORTATION MODES AND SERVICES
As commonly found in modern urban centers, the citizens commute basically
either on private or public transportation. Private modes are those varying from
pedestrian access to private vehicles, enclosed in between bicycles,
motorcycles, man powered carts, animal powered chariots, and many others.
Public transportation modes comprise the flow of buses, microbuses and taxis
serving the commuters under the charges of fixed and variable fares, depending
on distance traveled or vehicle boarding.
Dar es Salaam transportation network supports the movement of 5 basic means
of transportation or modes, grouped as usual on private and public use. On the
12 Demand Modeling and Forecasting
private modes: pedestrians, bicycles (including tricycles) and private cars and on
the public modes: taxi cars and public daladala buses and microbuses.
The demand simulation and forecast model implemented for Dar es Salaam
considered a public transportation trip matrix. As usual along with this
configuration and considering that in every trip generated there is a portion of it
done by foot, a pedestrian mode was allowed. Likewise, representing the existing
situation the network supports daladala mode and eventually private modes.
Based on the defined available modes, the present public transportation system
network was included in the model represented by 1915 daladala routes. The
itineraries were drawn and adjusted on top of the road network previously
defined, with PMU assistance and support. There is no other major public
transportation mean within DSM so the analysis just focused on those daladala
routes identified by PMU staff and updated during the field surveys.
Summarizing the transportation modes included in the simulation model are:
Public Transportation – Daladala Routes
Public Transportation – DART Services
Public Transportation – Feeder Services
Pedestrian
2.5. FARE STRUCTURE
The system currently considers the charges of a standard and generalized fare
for boarding daladalas of TZS 200 (value of the Base Scenario). In the future
scenarios, fares for the different components of DART system have been
separated into two groups depending on the level of fare integration desired and
which offers better financial and economic scenarios for the correct system’s
operation.
5 191 routes obtained as legally authorized routes by the time of the model elaboration. (April 2005)
13 Demand Modeling and Forecasting
The fare element is considered a complementary and crucial input for simulation
and modeling purposes which allows the valuation of travel generalized costs for
each user through the different modes their modal choice takes them.
Pedestrian modes are not being fare penalized allowing their flow in every
available link of the network except those segregated links along the trunk
corridors, avoiding undesired and non realistic pedestrian movements on this
imaginary links. The trunk system network has been structured and modeled as
segregated parallel network to the existing road network.
Access and transferences between these two components (networks) is done
through special access modes, allowing the identification and quantification of
passenger demand volumes. The fare is charged to the pedestrian user as an
additional time equivalent to the individual passenger fare or additional time
penalization due to connection and/or transfer delays or commute time between
integration stations6, when applicable. Exit movements on pedestrian modes are
not penalized and represent no charge to the user.
Developed in parallel with the financial and economical evaluations, the final fare
structure scenario defined for a future DART operation, and based on the
available and feasible fare structures applicable for the local situation, the
simulation was done for a structure as shown in Table 2.
DART services and the eventual number of boardings done to one or many trunk
services are independent with the fare paid per user, allowing the possibility of
boarding as many services as desired with no additional charges other than the
initial paid fare and transfer and waiting time penalization. Transfer penalty is
weighed based on the inconvenience a vehicle change represents to the user
and always is referred as time consumption charge.
6 Feeder integration at intermediate stations and terminals.
14 Demand Modeling and Forecasting
Table 2 Fare Structure Scenario for BRT
Fare (TZS) Transportation Mode
DART Only User 400 Feeder Only User (1 Route) 400 Feeder Only User (2 Routes) 800
Trunk + Feeder User 500 Daladala User 300
Feeder to Trunk 100 Trunk to Feeder 0
Trunk to Daladala 300 Daladala to Trunk 400
Daladala to Feeder 400
2.6. TRAVEL TIME SUBJECTIVE VALUE
The modal choice induced by the demand modeling process is mainly directed
by this value and its correct estimation. Estimation procedure followed includes
the income level standards identification for the population, which can be done
based on social division by wealth or simply by average income. Dar es Salaam
income distribution is predominantly represented by low income classes7, also
the ones that stand for the highest readership in public transportation. Therefore,
for calculating the TTSV for the city, the analysis was focused in obtaining an
average population income amount.
With the assistance and knowledge of local experts the analysis then assumes
the following labor legislation facts and common practice information for
employers and employees for the final TTSV calculation:
Monthly hours worked: 160
Monthly average salary: TZS 144.0008
7 Please refer to Annex Volume 2 – Background in Public Transportation for Income level estimates and distribution performed for Dar es Salaam population. 8 Done in October 2005.
15 Demand Modeling and Forecasting
Based on general experience and knowledge from other cities with extensive
analysis of stated and revealed preferences, form weighing the value of traveled
time against the worked time, the first one is valued as a third of the last, this will
mean that per three minutes worked, a user would be willing to spend one for
commute or travel.
Travel/Work Time Factor: 1/3
Following on the analysis:
Monthly Average Salary
Working Hours/Month TZS/Hour Travel
Cost/HourTZS 144,000 TZS 160 TZS 900 TZS 300
This value for hour is equivalent to TZS 5 per minute and so, represents the
TTSV for Dar es Salaam transport demand simulation model. Furthermore and
explaining this value and the effect it has, an average user will agree walking 40
minutes to avoid paying one standard daladala fare (TZS 200).
2.7. PUBLIC TRANSPORTATION PASSENGER DEMAND
Following the process of structuring the model, the demand source was
established as an origin destination trip matrix. Surveys were carried out for
approximately two and a half months, one of which was directed to identify the
trip desires, later being basic material to build a trip matrix between the
transportation zones previously defined for the network.
Bearing in mind the necessity of identifying the current system’s critical situation
or period of time where the largest volume of people is simultaneously onboard a
public service and actually traveling, the combination between two different
measurements (all day and morning period surveyed points) enable the
identification of the peak hour to be from 07:00 to 08:00 (see figure 6).
16 Demand Modeling and Forecasting
Figure 6 Emme2 Modeling Environment
Summarizing the process carried out, approximately 33,000 people were
surveyed on 35 points all over the city. For the Origin Destination Survey (ODSu)
completed, and regarding the issue of considering within the passenger demand
analysis the “double counting” of one single trip between different survey points,
the database depuration executed cut off this redundant information through an
adjustment process by identifying this Origin/Destination pairs passing through
several sections along its path in conjunction with the daladala route in which the
trip was surveyed.
After these adjustments and double counting elimination the origin destination
matrix for the peak hour contained 123,047 trips.
2.8. MODEL CALIBRATION
Upon information collected form field surveys (dispatch frequencies, passenger
volumes, boarding and alighting passengers at stations, etc.) the model’s
calibration process was executed.
17 Demand Modeling and Forecasting
Having the survey points as control points within the model to monitor the gap
between the modeled and existing situations, the model was adjusted to match
the existing conditions by internal trip matrix internal adjustments procedures.
Indicating the quality level reached by the calibrated model, a linear regression is
performed to compare statistically the accuracy and approximation achieved by
the demand simulation offered by the model (see figure 7). The regression
quality and data relation (between modeled and observed data) are also
evaluated by measuring the correlation index R2 and angle coefficient. For this
process the values reached are 0.98 and 1.0 respectively. Again, the analysis
was carried out by comparing the control point information with modeled one.
Figure 7 Linear Regression of Demand Model Results at Peak Hour
18 Demand Modeling and Forecasting
3. POPULATION AND DEMAND GROWTH FORECAST
3.1. POPULATION EXPANSION
With a transportation simulation model calibrated for the existing stage at hand
and relevant information regarding present conditions already processed, the
course was then directed to establish a future scenario for which DART system
will be up un running and from then forecast the operation by expanding the trip
matrix based on population travel patterns and inner city growth and expansion.
The date suggested by the Client was 2009 to hold the inaugural day for DART
being operational. From year 2002, Tanzania census information was available
and served as a reference year to begin a forecasting analysis, to be based on
regional growth, along with existing and maximum (critical) population density
values per transportation zone.
The maximum density was determined according to area income group as
follows:
Table 3 Population Index per Income Group Income group inhabitants/ hectare
1 630 2 630 3 490 4 350 5 210 6 70
The model adopted assumed a population expansion made for every zone i
starting from year 2002. Using a logistic model:
( ) ( ) ( )( )
( )⎟⎠⎞
⎜⎝⎛ −∗⎟⎟
⎠
⎞⎜⎜⎝
⎛⎟⎟⎠
⎞⎜⎜⎝
⎛∗+∗=+
totPtPtot
iPtPiktPitPi
max1
max11
With:
Pmax(i)= Maximum population for zone i
Pmax(i)= max(1.2*Pi(2002), area x maximum density)
19 Demand Modeling and Forecasting
Pi(2002)= Population on 2002 year for zone i
K= Constant for all zones= 0.0691
Pmaxtot= Maximum admitted total Dar population= 20 million
Following, K and Pmaxtot were adjusted in order to obtain an average of 0.043
yearly growth index (n) for the period 2002 – 2005, and a total increase ratio of
2.745 on the 2002 – 2032 period. 745.2)2002()2032(=
PtotPtot (This is the growth starting
with 0.043 and slowing down to 0.025 in 2032).
The yearly estimatives of Dar es Salaam population are: