Home Articles GIS for identification of demand-oriented urban rail transit corridor

GIS for identification of demand-oriented urban rail transit corridor

Ashish Verma
Research Scholar, Transportation Systems Engg., Dept. of Civil Engg.
IIT Bombay, Powai Mumbai-76, India.
Ph: 91-22-25767329
Fax: 91-22-25767302 / 25723480
E-mail: [email protected]

S. L. Dhingra
Professor of Transportation Systems Engineering, Dept. of Civil Engg.
IIT Bombay, Powai Mumbai-76, India.
Ph: 91-22-25767329
Fax: 91-22-25767302 / 25723480
E-mail: [email protected]

Introduction
During the second half of the last century, urban population in India had grown enormously. This has resulted in a steady increase in number of cities with a population of one million and above from 5 in 1951 to 35 in 2001. This level of urbanization has brought in its wake its own problems, especially with regard to its impact on the infrastructure facilities. The urban transportation systems have come under heavy strain affecting the quality of life of urban dwellers. Public transport facilities provided by buses are grossly inadequate to meet the increased travel demand and to provide a good level of service. In such situation, planning of higher order (rail based) mass transit system is essential. The first step in this process is to identify the corridor for the new system. Conventionally, the new rail transit corridors are identified based on land availability, and/or planner’s judgement, and then its ridership estimates are made using various techniques, for instance stated preference (SP) technique. However, in many cases these estimates prove to be much higher than the actual patronage observed resulting in huge revenue losses (Bodell 2002). Moreover, such differences in forecast and actual patronage can also be attributed to many other reasons, like lack of integration with other public transport services, ease of station access/ egress (and connectivity with major land uses generating or attracting trips) etc. Hence, a more rational approach to identify demand-oriented rail transit corridor is needed.

In past, many researchers have developed different approaches to identify rail transit corridors. In the study carried out by Clark and Oxley (1991) the rail transit corridors were identified by assigning the origin-destination (O-D) matrix to an assumed spider network. The drawback with the spider network is that it only gives travel desires between different zones and does not give the actual system alignment. A similar approach was adopted by Moorthy (1997), but in addition to spider network the flows were assigned on to a combined network of highway and track guided system and the corridors were obtained based on the assigned flows. Gipps et al. (2001) identified corridors for a new road or railway using convergence of geospatial imaging, softcopy photogrammetry, regional significance analysis and alignment optimization. They obtained the corridors based on land availability and cost considerations, which could be suitable for identifying road corridors. However, consideration of actual travel demand pattern of the city is also required for rail. Besides the limitations listed above, all these approaches did not address the need for an integrated planing i.e. operational, institutional, and physical integration, to achieve the forecasted demand for the new rail system. This is particularly important for a choice rider, who will shift to a new public transport only if it offers a comparable level-of-service. Although this point has been addressed to some extent in the work carried out by Chien and Schonfeld (1998), they used a pre-determined hypothetical network of one rail corridor and perpendicular feeder routes at equal spacing, which may not offer a realistic view.

Considering the limitations of earlier approaches, this paper presents a model for identifying new rail transit corridors for future public transport demand patterns based on user equilibrium approach. The objective of the study is to identify the new rail corridor using GIS, which is optimum from both users’ and operator’s point of view. The rail corridor is to be aligned to allow the user to travel through shortest path with comparable level-of-service and on high demand corridors to give maximum revenues to the operator. In addition, the type of technology to be used on the newly identified corridor is also to be investigated. Although, identifying the corridor depends not only on demand levels but also on economic and political considerations, these are not within the scope of this paper. GIS has emerged out as a good tool for dealing large and complex data including both spatial and non-spatial, it has promising potential in applying to problems like integrated urban transport planning, rail corridor identification etc. (Verma and Dhingra 2002).

Proposed Model
Considering the above objective, a new model is proposed consisting of four stages: generation of base year O-D person trips matrices, base year travel demand modeling, forecasting of O-D person trips matrices, and the rail transit corridor identification (Fig. 1). These stages are discussed below.


Fig.1- Proposed Four Stage Rail Transit Corridor Identification Model

Base Year O-D Person Trips Matrices Generation
The first stage in the model consists of generating the base year O-D person trips matrices. To forecast the ridership, the model only requires total number of person trips (irrespective of the mode used) from each zone to every other zone, for the planning year. Hence, the base year O-D matrices are to be generated for total person trips only. The freight modes are not to be considered while creating the matrices because the model requires a desired share for public transport, which can be obtained by splitting the person trips made only by the passenger modes. Accordingly, the O-D matrices are generated for base year. In this stage, the model requires home interview survey, screen line, cordon line, network, and O-D survey at passenger terminals data. Following are the steps involved in the procedure: –

Step 1: From the home interview survey data, the expansion factors for each zone are calculated by dividing the total household by the sample households.
Step 2: The O-D matrices are extracted from the home interview travel survey data, by counting each trip from the person’s origin to final destination as one person trip. This is to be done separately for HB, I-E, and E-I trips. The sample flows are then expanded to zonal level by multiplying them with the expansion factor for each zone.
Step 3: The O-D flows obtained from home interview survey data are supplemented by the O-D flows obtained from O-D cordon survey, and O-D survey at passenger terminals. This gives the final O-D matrices with all person trips (HB, NHB, I-E, E-I, and E-E).
Step 4: To validate the matrices, the combined O-D matrix of all person trips is loaded on to the base year network (after converting it to peak hour matrix using the appropriate average daily to peak hour ratio) using user equilibrium approach. While assigning the O-D flows, the capacity of links is used in terms of passengers per hour per direction (pphpd), which can be obtained by multiplying the capacity of each link in PCU per hour by the average car occupancy rate. Also, it is necessary to specify the link performance function and its parameters before the assignment can be done on the network because the user equilibrium process assigns the flows iteratively based on the link performance function. A link performance function is a mathematical description of the relationship between speed or travel time and link volume, a typical example of it is Bureau of public roads (BPR) formulation:


Where, t is congested link travel time, tf is link free-flow travel time, v is the link volume, c is link capacity, and a, b are calibration parameters. Using the above set-up, the assignment can be done. Comparing the assigned and observed link flows across the screenlines then validates the matrices.
Step 5: Finally the validated daily O-D person trips matrices for base year will be obtained, which will be used for the base year travel demand modelling, as described below.

Base Year Travel Demand Modelling
The base year travel patterns are to be modelled as accurately as possible and the models along with the horizon year planning variables and network information can then be used to forecast the trips in horizon year. This stage involves development of trip end and trip distribution models.

Trip End Models
Trip end models are developed only for the intra-city trips made by the residents of the study area. All other trips viz., I-E, E-I, E-E, are modelled by growth factors. The trip ends of internal trips for the base year can be obtained from the validated O-D matrices. Also, the trip end models can be developed using stepwise multiple linear regression technique. In general the multiple linear regression equation has the following form:

Y = a + b1X1 + b2X2+…+bnXn (2)
where Y is the trip production or attraction, X1, X2,…Xn are planning variables, a is the constant error term, and b1, b2,…bn are the regression coefficients. Calibrating a multiple linear regression model involves finding out the values of the constant error term and the regression coefficients.

The values of planning variables for each zone can be obtained through various primary and secondary sources. An inter correlation matrix is prepared to know the extent of correlation of the independent variables to be used in trip end models. The potential variables that can enter the models are decided based on their correlation with the dependent variables. Two variables having high correlation coefficient can not be used together in the same model. Different regression models for trip production and attraction can be tried. Based on the goodness of fit of the models indicated by coefficient of determination R2, standard error SE and t-values, the best models are adopted.

Trip Distribution Model for Intra-city Trips
A doubly constrained gravity trip distribution model of the following form can be calibrated to represent the base year travel pattern for intra-city trips, within the study area:

Tij = ai Pi bj Ajf(cij) ——— (3)
Subject to:


where Tij is the forecast flow produced by zone i and attracted to zone j, Pi is the forecast number of trips produced by zone i, Aj is the forecast number of trips attracted to zone j, cij is the impedance or cost of travel between zone i and zone j, f(cij) is the friction factor between zone i and j, ai is the balancing factor for row i given as


bj is the balancing factor for column j given as


The interzonal travel impedance is taken in terms of travel time. It is to be noted that the generalised cost cannot be used here, as the study deals with only the passenger trips irrespective of the mode used. A suitable friction factor function such as, exponential, inverse power, or gamma, is to be defined, for calibration and application of the gravity model. The exponential function of the form


is generally found to give better results than the others, here, r is the calibration parameter (Caliper 1996).

Calibrating the gravity model consists of evaluating the parameters of the friction factor function so that the gravity model reproduces, as closely as possible, the base year productions and attractions and trip length distribution. The calibration can be done by using the base year P-A matrix, the impedance matrix, and a geographic zone layer in GIS to generate the observed trip length frequency distribution (OTLD). The aim of the calibration is to match OTLD and estimated trip length frequency distribution (ETLD) as closely as possible.

Forecasting of O-D Person Trips Matrices
The intra-city person trips can be forecasted using the gravity model (equation 3). To apply the gravity model, the inputs required are a friction factor matrix, and a geographic view with forecasted production and attraction for each zone. The output will be a zone-to-zone forecasted production-attraction (P-A) person trips matrix. The friction factor matrix can be obtained using the impedance matrix of the base year and the friction factor function (equation 8). The total value of planning variables to be used in trip end equations can be forecasted giving due consideration to the city development plan, the conceptual land use plan for the horizon year, and the population size. They can be distributed among the zones in proportion to their base year levels and appropriately adjusting for the newly developed zones based on their increase in population. To obtain the forecasted production and attraction for each zone, the trip end equations developed are used, along with the forecasted planning variables. Therefore, using the friction factor matrix and the forecasted productions and attractions, the O-D person trips matrix for internal trips for the planning year will be obtained using equation (3). Now, it may be quite reasonable to postulate the suitability of a gravity trip distribution model for internal-to-internal (I-I) trips. Since, a significant proportion of the trips are external, the gravity model may not be suitable for modelling such trips as it depends on variables like trip distance or cost, which are not defined for external trips. In such situations, the common practice is to take these trips outside the synthetic modelling process and to undertake roadside interviews at cordon points. The resulting matrices of E-E, I-E, and E-I trips is then updated and forecast using Furness growth factor methods (Ortuzar and Willumsen 1996). The growth factors for external zones can be obtained using the past traffic and other secondary data.

Finally, the O-D person trips matrices of internal and external zones are merged together to obtain the daily O-D person trips matrix for the planning year. The daily matrix obtained is factored by the average daily to peak hour ratio, to obtain the peak hour O-D person trips matrix. This matrix is used in the next stage for the identification of rail corridor as described below.

Identification of Rail Corridor
The major steps involved in this stage are: obtaining the desired modal share between public transport (PT) and other passenger (OP) modes, estimating the travel demand pattern for PT modes, and selecting the alignment and recommending a suitable system. The first step is to obtain a desired share of public transport for the city in the planning year.

Estimating Desired Modal Share
To obtain the desired share between PT and OP modes, the type of commuters who will use the PT modes need to be understood. In general, the commuters using the PT modes can be classified as captive and choice riders. When a new attractive demand-oriented rail based PT system is introduced in any city, its demand can be estimated by including all the captive riders plus some choice riders who will transfer to the new rail based PT system depending on its level of service. In such cases, a suitable criterion can be defined to obtain the potential share of PT modes against all OP modes. This can be developed in the form of a curve giving desired model share between PT and OP modes based on population or any other relevant variable. A typical example of this is the curve developed in the Report of the study group on alternative systems of urban transport, GOI (1987), which gives the potential share of public transport (PT) modes based on the population of city. Taking the desired modal share from this curve, the forecasted peak hour person trips matrix can be divided into peak hour O-D person trip matrix for PT modes and for OP modes. It is to be noted that estimating the potential demand for the new PT system based on the base year actual modal share will be incorrect because there will not be any consideration of the shift of choice riders from OP to PT modes. Forecasting techniques like stated preference (SP) also can not be used because they are suitable for a proposed rail alignment with a known set of characteristics. This is different from the case where one is trying to find out potential demand on the city network to identify the alignment for the new rail based mass transit system. In the next step, the forecasted matrices are used to obtain the travel demand pattern for PT modes.

Introduction
During the second half of the last century, urban population in India had grown enormously. This has resulted in a steady increase in number of cities with a population of one million and above from 5 in 1951 to 35 in 2001. This level of urbanization has brought in its wake its own problems, especially with regard to its impact on the infrastructure facilities. The urban transportation systems have come under heavy strain affecting the quality of life of urban dwellers. Public transport facilities provided by buses are grossly inadequate to meet the increased travel demand and to provide a good level of service. In such situation, planning of higher order (rail based) mass transit system is essential. The first step in this process is to identify the corridor for the new system. Conventionally, the new rail transit corridors are identified based on land availability, and/or planner’s judgement, and then its ridership estimates are made using various techniques, for instance stated preference (SP) technique. However, in many cases these estimates prove to be much higher than the actual patronage observed resulting in huge revenue losses (Bodell 2002). Moreover, such differences in forecast and actual patronage can also be attributed to many other reasons, like lack of integration with other public transport services, ease of station access/ egress (and connectivity with major land uses generating or attracting trips) etc. Hence, a more rational approach to identify demand-oriented rail transit corridor is needed.

In past, many researchers have developed different approaches to identify rail transit corridors. In the study carried out by Clark and Oxley (1991) the rail transit corridors were identified by assigning the origin-destination (O-D) matrix to an assumed spider network. The drawback with the spider network is that it only gives travel desires between different zones and does not give the actual system alignment. A similar approach was adopted by Moorthy (1997), but in addition to spider network the flows were assigned on to a combined network of highway and track guided system and the corridors were obtained based on the assigned flows. Gipps et al. (2001) identified corridors for a new road or railway using convergence of geospatial imaging, softcopy photogrammetry, regional significance analysis and alignment optimization. They obtained the corridors based on land availability and cost considerations, which could be suitable for identifying road corridors. However, consideration of actual travel demand pattern of the city is also required for rail. Besides the limitations listed above, all these approaches did not address the need for an integrated planing i.e. operational, institutional, and physical integration, to achieve the forecasted demand for the new rail system. This is particularly important for a choice rider, who will shift to a new public transport only if it offers a comparable level-of-service. Although this point has been addressed to some extent in the work carried out by Chien and Schonfeld (1998), they used a pre-determined hypothetical network of one rail corridor and perpendicular feeder routes at equal spacing, which may not offer a realistic view.

Considering the limitations of earlier approaches, this paper presents a model for identifying new rail transit corridors for future public transport demand patterns based on user equilibrium approach. The objective of the study is to identify the new rail corridor using GIS, which is optimum from both users’ and operator’s point of view. The rail corridor is to be aligned to allow the user to travel through shortest path with comparable level-of-service and on high demand corridors to give maximum revenues to the operator. In addition, the type of technology to be used on the newly identified corridor is also to be investigated. Although, identifying the corridor depends not only on demand levels but also on economic and political considerations, these are not within the scope of this paper. GIS has emerged out as a good tool for dealing large and complex data including both spatial and non-spatial, it has promising potential in applying to problems like integrated urban transport planning, rail corridor identification etc. (Verma and Dhingra 2002).

Proposed Model
Considering the above objective, a new model is proposed consisting of four stages: generation of base year O-D person trips matrices, base year travel demand modeling, forecasting of O-D person trips matrices, and the rail transit corridor identification (Fig. 1). These stages are discussed below.


Fig.1- Proposed Four Stage Rail Transit Corridor Identification Model

Base Year O-D Person Trips Matrices Generation
The first stage in the model consists of generating the base year O-D person trips matrices. To forecast the ridership, the model only requires total number of person trips (irrespective of the mode used) from each zone to every other zone, for the planning year. Hence, the base year O-D matrices are to be generated for total person trips only. The freight modes are not to be considered while creating the matrices because the model requires a desired share for public transport, which can be obtained by splitting the person trips made only by the passenger modes. Accordingly, the O-D matrices are generated for base year. In this stage, the model requires home interview survey, screen line, cordon line, network, and O-D survey at passenger terminals data. Following are the steps involved in the procedure: –

Step 1: From the home interview survey data, the expansion factors for each zone are calculated by dividing the total household by the sample households.
Step 2: The O-D matrices are extracted from the home interview travel survey data, by counting each trip from the person’s origin to final destination as one person trip. This is to be done separately for HB, I-E, and E-I trips. The sample flows are then expanded to zonal level by multiplying them with the expansion factor for each zone.
Step 3: The O-D flows obtained from home interview survey data are supplemented by the O-D flows obtained from O-D cordon survey, and O-D survey at passenger terminals. This gives the final O-D matrices with all person trips (HB, NHB, I-E, E-I, and E-E).
Step 4: To validate the matrices, the combined O-D matrix of all person trips is loaded on to the base year network (after converting it to peak hour matrix using the appropriate average daily to peak hour ratio) using user equilibrium approach. While assigning the O-D flows, the capacity of links is used in terms of passengers per hour per direction (pphpd), which can be obtained by multiplying the capacity of each link in PCU per hour by the average car occupancy rate. Also, it is necessary to specify the link performance function and its parameters before the assignment can be done on the network because the user equilibrium process assigns the flows iteratively based on the link performance function. A link performance function is a mathematical description of the relationship between speed or travel time and link volume, a typical example of it is Bureau of public roads (BPR) formulation:


Where, t is congested link travel time, tf is link free-flow travel time, v is the link volume, c is link capacity, and a, b are calibration parameters. Using the above set-up, the assignment can be done. Comparing the assigned and observed link flows across the screenlines then validates the matrices.
Step 5: Finally the validated daily O-D person trips matrices for base year will be obtained, which will be used for the base year travel demand modelling, as described below.

Base Year Travel Demand Modelling
The base year travel patterns are to be modelled as accurately as possible and the models along with the horizon year planning variables and network information can then be used to forecast the trips in horizon year. This stage involves development of trip end and trip distribution models.

Trip End Models
Trip end models are developed only for the intra-city trips made by the residents of the study area. All other trips viz., I-E, E-I, E-E, are modelled by growth factors. The trip ends of internal trips for the base year can be obtained from the validated O-D matrices. Also, the trip end models can be developed using stepwise multiple linear regression technique. In general the multiple linear regression equation has the following form:

Y = a + b1X1 + b2X2+…+bnXn (2)
where Y is the trip production or attraction, X1, X2,…Xn are planning variables, a is the constant error term, and b1, b2,…bn are the regression coefficients. Calibrating a multiple linear regression model involves finding out the values of the constant error term and the regression coefficients.

The values of planning variables for each zone can be obtained through various primary and secondary sources. An inter correlation matrix is prepared to know the extent of correlation of the independent variables to be used in trip end models. The potential variables that can enter the models are decided based on their correlation with the dependent variables. Two variables having high correlation coefficient can not be used together in the same model. Different regression models for trip production and attraction can be tried. Based on the goodness of fit of the models indicated by coefficient of determination R2, standard error SE and t-values, the best models are adopted.

Trip Distribution Model for Intra-city Trips
A doubly constrained gravity trip distribution model of the following form can be calibrated to represent the base year travel pattern for intra-city trips, within the study area:

Tij = ai Pi bj Ajf(cij) ——— (3)
Subject to:


where Tij is the forecast flow produced by zone i and attracted to zone j, Pi is the forecast number of trips produced by zone i, Aj is the forecast number of trips attracted to zone j, cij is the impedance or cost of travel between zone i and zone j, f(cij) is the friction factor between zone i and j, ai is the balancing factor for row i given as


bj is the balancing factor for column j given as


The interzonal travel impedance is taken in terms of travel time. It is to be noted that the generalised cost cannot be used here, as the study deals with only the passenger trips irrespective of the mode used. A suitable friction factor function such as, exponential, inverse power, or gamma, is to be defined, for calibration and application of the gravity model. The exponential function of the form


is generally found to give better results than the others, here, r is the calibration parameter (Caliper 1996).

Calibrating the gravity model consists of evaluating the parameters of the friction factor function so that the gravity model reproduces, as closely as possible, the base year productions and attractions and trip length distribution. The calibration can be done by using the base year P-A matrix, the impedance matrix, and a geographic zone layer in GIS to generate the observed trip length frequency distribution (OTLD). The aim of the calibration is to match OTLD and estimated trip length frequency distribution (ETLD) as closely as possible.

Forecasting of O-D Person Trips Matrices
The intra-city person trips can be forecasted using the gravity model (equation 3). To apply the gravity model, the inputs required are a friction factor matrix, and a geographic view with forecasted production and attraction for each zone. The output will be a zone-to-zone forecasted production-attraction (P-A) person trips matrix. The friction factor matrix can be obtained using the impedance matrix of the base year and the friction factor function (equation 8). The total value of planning variables to be used in trip end equations can be forecasted giving due consideration to the city development plan, the conceptual land use plan for the horizon year, and the population size. They can be distributed among the zones in proportion to their base year levels and appropriately adjusting for the newly developed zones based on their increase in population. To obtain the forecasted production and attraction for each zone, the trip end equations developed are used, along with the forecasted planning variables. Therefore, using the friction factor matrix and the forecasted productions and attractions, the O-D person trips matrix for internal trips for the planning year will be obtained using equation (3). Now, it may be quite reasonable to postulate the suitability of a gravity trip distribution model for internal-to-internal (I-I) trips. Since, a significant proportion of the trips are external, the gravity model may not be suitable for modelling such trips as it depends on variables like trip distance or cost, which are not defined for external trips. In such situations, the common practice is to take these trips outside the synthetic modelling process and to undertake roadside interviews at cordon points. The resulting matrices of E-E, I-E, and E-I trips is then updated and forecast using Furness growth factor methods (Ortuzar and Willumsen 1996). The growth factors for external zones can be obtained using the past traffic and other secondary data.

Finally, the O-D person trips matrices of internal and external zones are merged together to obtain the daily O-D person trips matrix for the planning year. The daily matrix obtained is factored by the average daily to peak hour ratio, to obtain the peak hour O-D person trips matrix. This matrix is used in the next stage for the identification of rail corridor as described below.

Identification of Rail Corridor
The major steps involved in this stage are: obtaining the desired modal share between public transport (PT) and other passenger (OP) modes, estimating the travel demand pattern for PT modes, and selecting the alignment and recommending a suitable system. The first step is to obtain a desired share of public transport for the city in the planning year.

Estimating Desired Modal Share
To obtain the desired share between PT and OP modes, the type of commuters who will use the PT modes need to be understood. In general, the commuters using the PT modes can be classified as captive and choice riders. When a new attractive demand-oriented rail based PT system is introduced in any city, its demand can be estimated by including all the captive riders plus some choice riders who will transfer to the new rail based PT system depending on its level of service. In such cases, a suitable criterion can be defined to obtain the potential share of PT modes against all OP modes. This can be developed in the form of a curve giving desired model share between PT and OP modes based on population or any other relevant variable. A typical example of this is the curve developed in the Report of the study group on alternative systems of urban transport, GOI (1987), which gives the potential share of public transport (PT) modes based on the population of city. Taking the desired modal share from this curve, the forecasted peak hour person trips matrix can be divided into peak hour O-D person trip matrix for PT modes and for OP modes. It is to be noted that estimating the potential demand for the new PT system based on the base year actual modal share will be incorrect because there will not be any consideration of the shift of choice riders from OP to PT modes. Forecasting techniques like stated preference (SP) also can not be used because they are suitable for a proposed rail alignment with a known set of characteristics. This is different from the case where one is trying to find out potential demand on the city network to identify the alignment for the new rail based mass transit system. In the next step, the forecasted matrices are used to obtain the travel demand pattern for PT modes.

Travel Demand Pattern for Public Transport (PT) Modes
Before assigning the forecasted flows on to the network to obtain the travel demand pattern for PT modes, the existing rail links (if any) are converted into equivalent road links in the road layer. Then the O-D person trips matrix for all OP modes are assigned to the network, disabling the equivalent road links. These assigned flows for OP modes are then taken as preload on to the network. This is done to assign the O-D person trips matrix for all PT modes on to the complete network (i.e. enabling the equivalent road links). This will give the future ridership pattern in terms of passengers per hour per direction (pphpd) for all PT modes on every link of the network. It is to be noted here that while assigning the matrix for PT modes although the complete PT demand is taken into consideration, but, the existing bus transport network is not considered as the same may not be optimum for the planning year flows and also the buses can be re-routed without incurring any extra cost, any existing rail network is considered as it is, since, any re-routing requires heavy expenditure.

Alignment Selection and System Recommendation
The final step is to obtain the alignment for the new rail transit system. For this, the forecasted ridership patterns are compared with the capacity ranges for various mass transit systems based on different levels of service in terms of passenger density. The detailed capacity assessment of different mass transit technologies is done by calculating the passenger capacity in the peak direction during peak hour. For bus transit system the passenger capacity is found out by:


and for the rail transit system the passenger capacity is given by:


The suitability range in terms of capacity (pphpd) for each transit system can be taken as between that obtained for normal load and crush load. The values of normal and crush load can be decided suitably. The capacity ranges obtained are then compared with the forecasted ridership estimate on every link of the network. All the links with ridership level greater than the maximum capacity of the street transit should be sorted and clubbed together to obtain the newly identified rail transit alignment. Moreover, the recommendation on most suitable system for operation on the newly identified corridor can be done based on comparing the maximum peak hour ridership obtained and capacity ranges for each type of transit system. Finally, the newly identified corridor can be represented graphically on the study area map.

With this approach, the newly identified rail corridor will be demand-oriented and optimum from both users’ and operator’s point of view because a user will be able to travel from his origin to destination by shortest or nearly shortest path and the operator will earn maximum revenues as the rail system will be aligned along the high passenger ridership routes. However, it is to be noted that this will be completely achieved only when the feeder routes and integrated schedules are introduced having a level-of-service comparable to OP modes.

Role of GIS
The present model requires collection, management, retrieval and analysis of large amount of data. Besides this, it needs a big and complex city network which can be comprehended well only with some good visualization tools. Handling such data is not possible with generic database management systems. Hence, there is a need for a specialized tool to handle such large-scale complex data. Geographical Information System (GIS) is one such promising tool. It is basically a relational database management system (RDBMS), which integrates common database operations such as query and statistical analysis with the unique visualization and geographic analysis benefits offered by maps and layers. In short, GIS can store, analyze, and display large amount of spatial and non-spatial information. It stores data about the world as a collection of thematic layers that can be linked together by geography. This simple but extremely powerful and versatile concept has proven invaluable for solving many real-world problems, from tracking delivery vehicles to modelling global atmospheric pollution (Verma and Dhingra 2002).

Case Study

Study Area Details
The model is applied on Thane municipal corporation (TMC) area, a major urban center of Mumbai metropolitan region (MMR), India (Fig. 2). In the past, the development of various industrial estates and also the supporting residential and service employment has created the city more dynamic in nature. The city has developed in a circular fashion, expanding outwards from the initial CBD adjacent to the main railway station, which is served by the Central Railway’s north-south line providing commuter services to MMR as well as longer distance ones. Buses run by various transport undertakings are the present modes of public transport in Thane. The closeness of Thane to Greater Mumbai, the commercial capital of India, has resulted in a rapid growth both in population and employment. This has resulted in an exponential growth in intra-city travel demand, which is beyond the capacity of bus transit system to handle. Hence, the study area requires an intra-city rail based mass transit system in future planning period, to cater this demand. The population growth in the area from 1951 to 2001 and the forecasted population up to 2031 is shown in Table. 1. The travel demand in the study area is analysed on the basis of total 122 Traffic analysis zones (TAZs) within and outside the TMC area.


Fig.2- Map of the Study Area, Thane

Table.1- Population Growth and Forecasted Population for Thane

Year Population (millions)
1951 0.07
1961 0.1
1971 0.25
1981 0.47
1991 0.80
2001 1.26
2011 2.45*
2021 2.88*
2031 3.04*
*Forecasted population, Source: CES (2001)

Data Collection
Relevant data was collected for the base year 2001 (CES 2001) from various primary and secondary sources. From the data collected, it was found that a total of 17,78,178 trips are generated in the study area, by the residents of TMC, on an average working day. The per capita and per household trip rate was worked out to be 1.44 and 5.2 respectively. Out of the total trips generated, 23.66% were found to be intra-zonal and 76.34% as inter-zonal. While classifying the trips by purpose, work trips accounted for the highest share (46.04%) followed by educational trips (38.06%). The average journey speed during peak period on some of the major arterial of Thane was found to vary from a low of 8.2 km/h to a high of 45.3 km/h. Also, the average household size in TMC area was found to be 3.81, which actually ranges between one to 12 person. All these figures form an important input information in developing transport demand models, and subsequently identifying the new rail corridors, CES (2001).

Database Management in TransCAD
Generic GIS software performs five processes or tasks: input, manipulation, management, query and analysis, and visualisation. However, the study requirements warrants for the software, which can extend the traditional GIS data model to include transportation data objects such as, transportation networks, matrices, routes and route systems, linear-referenced data etc. It should also include transportation application modules like, network analysis, transportation planning and travel demand modeling, vehicle routing and logistics, and districting and location modeling etc. The best such state-of-art software available is TransCAD, (Caliper 1996). The application modules in TransCAD are fully integrated with GIS functions for improved performance and ease of use. Hence, looking to all these capabilities, TransCAD was an obvious choice for use in this study. Its objects and application modules were accordingly used to calibrate and apply the various models, to get faster and effective results, and also for the effective graphical representation of the results.

To use the various objects and application modules, the whole database was created and maintained in TransCAD, which includes both the spatial and non-spatial data. It was maintained in five different geographic layers on the digitized map. All the data pertaining to traffic analysis zones (TAZs) was coded in zone layer, which includes zone ids, zone area, and other planning variables for each zone. The road layer contained the details of all road links and also the centroid connectors, including link ids, link type, link distance in kilometer, capacity in passenger car units (PCU) per hour and in person trips per hour (for both directions), free flow and average speed in km/h, free flow and average time in minutes. The node layer contained all the data for endpoints of every links and also of zone centroids. The suburban rail network passing through study area was coded separately in rail layer and rail node layer, the section between two railway stations was identified as a rail link and the stations were considered as the rail node, consequently the respective data was coded in the rail and rail node layers.

Base Year O-D Person Trips Matrices Generation
To apply the proposed methodology on the case study area, the base year was taken as 2001 and the planning year as 2031. For converting the daily O-D matrices to peak hour matrices, the average daily to peak hour ratio was taken as 8, as observed in the study area. Also, the average car occupancy rate was taken as 2.8 for converting the capacity of links to pphpd. The link performance function (equation 1) parameters alpha and beta were set by link type. Using the above set-up, the assignment was done and the matrix was validated by comparing the assigned and observed link flows across the screenlines.

Base Year Travel Demand Modelling
The application modules for travel demand modelling in TransCAD were used to calibrate the models and forecast the travel demand pattern for the horizon year. Considering the differences in the level of development, separate trip end models were developed for the central business district (CBD) area having 72 zones, and for the south-east and north-west (SE-NW) suburbs of TMC having 43 zones. The planning variables used for development of zonal least square linear regression models for trip ends are presented in Table.2. The trip end model obtained for CBD area is given as:

Table.2- Planning Variables for Trip End Models

Variable Name Description
X1 Zone area in square kilometer
X2 No. of household in base year 2001
X3 Population in base year 2001
X4 Residential workers
X5 No. of private vehicles owned
X6 Service employment
X7 Trade employment
X8 Industrial employment
X9 Total employment
X10 No. of students
Y1 Trip production
Y2 Trip attraction

In the next step, the required inputs for calibration of gravity model were given, and the calibration results were obtained. The estimated and observed mean trip lengths, in minutes, were obtained to be 13.047 and 13.343 respectively. The value of the calibration parameter of the exponential function r was obtained to be 0.1393. Fig. 3 gives the comparison between the observed (OTLD) and estimated trip length frequency distribution (ETLD), it can be seen that they matches very closely.

Forecasting of O-D Person Trips Matrices
The forecasted value of planning variables population, total employment, and residential workers were obtained as, 30,366,58, 10,020,97, and 10,324,64 respectively. To obtain the forecasted production and attraction for each zone, equations (11) to (14) were used. The intra-city (i.e. for 115 internal zones) person trips were forecasted using the gravity model in equation (3). The Furness growth factor method was used for forecasting the external trips. Finally, the O-D person trips matrices of internal and external zones were merged together to obtain the forecasted daily O-D person trips matrix for the year 2031. The daily matrix obtained was factored by the daily to peak hour ratio of 8, to obtain the peak hour O-D person trips matrix.


Fig.3- Trip Length Frequency Distribution

Identification of Rail Corridor
Based on the curve developed in GOI(1987), the potential share of PT modes against OP modes for Thane city with a projected population of about 3.04 million in the year 2031 (Table.1), was taken as 64 %. Before assigning the forecasted flows on to the network to obtain the travel demand pattern for public transport (PT) modes, the existing suburban rail links were converted into equivalent road links in the road layer. To do so, the suburban rail links were digitised into equivalent road links in the road geographic layer and suitable capacity values were assigned to them. Based upon some earlier studies done (GOI 1987, MMRDA 1998), the values of normal and crush load were taken as 5 Standees/m2 and 8 Standees/m2 respectively.

The maximum peak hour ridership was obtained as 34,369 pphpd. Fig. 4 shows the forecasted ridership pattern on the study area map, as generated in TransCAD. To obtain the alignment for the new rail transit system, the mass transit systems considered for detailed capacity assessment were: street transit systems (ST), light rail transit 1 (LRT1), light rail transit 2 (LRT2), Rapid Rail Transit (RRT or Metro), and Regional Rail Transit (Suburban Railway).

The capacity ranges obtained are as shown below:

Street transit system = Up to 12,000 pphpd
(Mini-bus, single decker standard bus, double-decker and articulated)
LRT1 = 12,000 to 36,000 pphpd
LRT2 = 36,000 to 50,000 pphpd


Fig.4- Map Showing the Results of Assignment for Horizon Year Peak Hour PT Matrix, Generated in TransCAD.

RRT (Metro) = 50,000 to 69,000 pphpd
RGR = 59,000 to 89,000 pphpd

More detailed calculations can be referred from Verma and Dhingra (2001). Since the maximum peak hour ridership was found to be 34,369 pphpd, the most suitable system recommended for operation on the newly identified corridor will be Light Rail Transit 1, as the maximum peak hour ridership falls within its capacity range.

Results and Discussion
Fig. 5 graphically represents the newly identified rail corridor on the study area map. Also, Table. 3 give the details of the links, which form the newly identified rail corridor, their length in kilometres, and the corresponding ridership estimates on these links. It can be observed from the table that all the links have ridership level within the capacity range of LRT1, except a few, which were included to give continuity to the rail corridor. The newly identified corridor consists of two routes, the length of 1st route is 9.79 kms, and the length of 2nd route is 11.66 kms. Both the routes start from link no. 85, they move parallel and cross through the CBD area of Thane till link no. 209, from where they change their direction, route 1 move towards the north-west part of the study area, up to link no. 9, and the route 2 move towards the south-east part of the study area up to link no. 277.


Fig.5- Newly Identified Rail Corridor for Thane City

These results shows that besides the requirement for proper connectivity within the CBD area, the higher development in future of north-west and south-east zones also warrants a proper connectivity in the form of a rail based transit system. These routes being obtained using the user equilibrium approach will allow the user to travel from its origin to final destination via. the shortest possible path, provided the feeder bus routes and integrated schedules are also developed. Also, being aligned along the maximum ridership corridors, the routes will ensure maximum revenue returns for the operator. Hence, it can be claimed that by using the proposed methodology the routes will be aligned along the optimum corridors, from both the users’ and operator’s point of view.

While implementing the routes, the chances of getting the required land within the CBD area are generally very less, hence they can be constructed either underground or on elevated tracks, but, as the routes move out of the CBD area, they can be brought on to the surface, as the required land would probably be available.

Table.3- Details of the Links Forming the Newly Identified Rail Corridors

Route 1 Route 2
Link No. Length (Kms.) Ridership (pphpd) Link No. Length (Kms.) Ridership (pphpd)
85 0.250 34,369 85 0.250 34,369
78 1.0 26,129 78 1.0 26,129
83 0.6 14,204 83 0.6 14,204
109 0.15 16,296 109 0.15 16,296
110 0.15 15,646 110 0.15 15,646
117 0.30 16,176 117 0.30 16,176
121 0.20 16,105 121 0.20 16,105
209 0.10 20,074 209 0.10 20,074
225 0.40 8,002 221 0.35 15,071
227 0.70 4,280 212 0.25 13,227
234 0.20 5,118 213 0.30 24,217
243 0.750 13,845 324 0.30 26,959
244 0.10 27,261 252 0.72 18,769
48 0.53 21,314 257 0.77 15,774
42 0.83 17,378 258 1.02 3621
23 0.90 18,386 262 0.68 1169
14 0.70 18,302 273 2.77 3522
11 1.03 19,159 274 0.69 14,580
9 0.90 12,974 275 0.270 13,771
Total Length = 9.79 Kms 276 0.22 12,940
      277 0.57 12,155
      Total Length = 11.66 Kms

In the proposed model, by assigning the O-D flows using user equilibrium approach, it is ensured that the higher flow levels are obtained on the major roads (with multiple lanes) only, because the flows are assigned iteratively based on the capacity measurements of each link. This way it is ensured that the rail alignment will pass through the major roads only, giving ample space to construct elevated tracks (or even surface tracks) for the rail system within CBD area, which will be economically cheaper than the underground system.

Summary & Conclusion
This paper discusses a model for identifying demand-oriented urban rail transit corridor on a city road network, using geographical information system (GIS) tools. The proposed model is able to identify new rail corridors, which are optimum from both the users’ and operator’s point of view. The corridors were identified based on the demand obtained using user equilibrium approach, and suitable recommendations were made on the technology that can meet this demand.

The conclusions from the study are as follows:

  • Since the ridership estimates are obtained using the user equilibrium approach, the routes identified will allow the user to travel from its origin to final destination via. the shortest path, thus achieving the user optimum.
  • While comparing the ridership estimates with the capacity, the suitability range of capacity (pphpd) for each transit system is between normal load and crush load. The lower bound ensures that the peak hour flow on most of the links of the identified route is more than the normal load capacity of the rail transit system. This way steady revenue returns are guaranteed for the operator. On the other hand, the upper bound makes certain that the desired level-of-service for the user is achieved.
  • The routes identified using the proposed model effectively covers the study area i.e. both central business district CBD and outer areas. Since, most of the links in the identified routes have ridership level above 12,000 pphpd, the routes will be demand oriented and will give steady revenue returns to the operator.
  • Since, the maximum peak hour ridership was found to be 34,369 pphpd. The maximum peak hour load in LRT1 will always remain less than its crush load.
  • GIS turns out to be an excellent tool for transportation related applications. The various application and generic modules of TransCAD helps immensely in calibration and application of transport models at various stages of the study. It also serves as an excellent tool for graphical representation of the results. To provide reasonably direct access to the commuter from its origin to final destination, and also to achieve the estimated ridership on the newly identified rail corridor, the feeder bus routes and integrated schedules for feeder bus and rail operation can be developed. The feeder routes can be designed taking into consideration the estimated ridership pattern on every link within the feeder area, and also the access and egress demand of the station. The integrated schedules can be developed by optimizing the combined objective function of operator cost and transfer time and comfort for the user.

References

  • Bodell G. (2002). “Urban rail demand forecasts – where do models go wrong ?” Smart Urban Transport, 1(2), 6-9.
  • Caliper. (1996). Travel demand modeling with TransCAD 3.2, Massachusetts, U.S.A. Consulting Engineering Services India Ltd. (CES). (2001). Proposed Mass Rapid Transit System for Thane City, Draft Final Report, MSRDC, Mumbai, India.
  • Chien S., Schonfeld P. (1998). “Joint Optimization of a rail transit line and its feeder bus system.” J. of Adv. Transp., 32(3), 253-284.
  • Clark J., Oxley P. (1991). “Strategic public transport planning in Baghdad.” Urban transport in developing countries, Heraty M., eds., PTRC Education Research Services Ltd., London, 224-228.
  • Gipps P.G., Gu K.Q., Held A., Barnett G. (2001). “New technologies for transport route selection” Transp. Res., 3(9), 135-154. Government of India (GOI). (1987). Report of the study group on alternative systems of urban transport, New Delhi, India.
  • Mumbai Metropolitan Region Development Authority (MMRDA). (1998). Working paper: review of various mass transit systems, Mumbai, India.
  • Moorthy N.V.R. (1997). “Planning of integrated transit network for bus and LRT.” J. of Adv. Transp., 31(3), 283-309.
  • Ortuzar J.de.D., Willumson L.G. (1996). “Modelling transport.” Wiley, England, pp.180.
  • Verma A., Dhingra S.L. (2001). “Suitability of alternative systems for urban mass transport for Indian cities.” Trasporti Europei, VII(18), 4-15. Verma A., Dhingra S.L. (2002), “GIS Technology for Transportation and its Trends”, Proc. All India Seminar on Remote Sensing & Geographical Information System Techniques, The Institution of Engineers (India), Indore, India.

Acknowledgements
The authors would like to thank Maharashtra State Road Development Corporation (MSRDC) and M/s Consulting Engineering Services (I) Ltd., for providing all the necessary data for this study.

Case Study

Study Area Details
The model is applied on Thane municipal corporation (TMC) area, a major urban center of Mumbai metropolitan region (MMR), India (Fig. 2). In the past, the development of various industrial estates and also the supporting residential and service employment has created the city more dynamic in nature. The city has developed in a circular fashion, expanding outwards from the initial CBD adjacent to the main railway station, which is served by the Central Railway’s north-south line providing commuter services to MMR as well as longer distance ones. Buses run by various transport undertakings are the present modes of public transport in Thane. The closeness of Thane to Greater Mumbai, the commercial capital of India, has resulted in a rapid growth both in population and employment. This has resulted in an exponential growth in intra-city travel demand, which is beyond the capacity of bus transit system to handle. Hence, the study area requires an intra-city rail based mass transit system in future planning period, to cater this demand. The population growth in the area from 1951 to 2001 and the forecasted population up to 2031 is shown in Table. 1. The travel demand in the study area is analysed on the basis of total 122 Traffic analysis zones (TAZs) within and outside the TMC area.


Fig.2- Map of the Study Area, Thane

Table.1- Population Growth and Forecasted Population for Thane

Year Population (millions)
1951 0.07
1961 0.1
1971 0.25
1981 0.47
1991 0.80
2001 1.26
2011 2.45*
2021 2.88*
2031 3.04*
*Forecasted population, Source: CES (2001)

Data Collection
Relevant data was collected for the base year 2001 (CES 2001) from various primary and secondary sources. From the data collected, it was found that a total of 17,78,178 trips are generated in the study area, by the residents of TMC, on an average working day. The per capita and per household trip rate was worked out to be 1.44 and 5.2 respectively. Out of the total trips generated, 23.66% were found to be intra-zonal and 76.34% as inter-zonal. While classifying the trips by purpose, work trips accounted for the highest share (46.04%) followed by educational trips (38.06%). The average journey speed during peak period on some of the major arterial of Thane was found to vary from a low of 8.2 km/h to a high of 45.3 km/h. Also, the average household size in TMC area was found to be 3.81, which actually ranges between one to 12 person. All these figures form an important input information in developing transport demand models, and subsequently identifying the new rail corridors, CES (2001).

Database Management in TransCAD
Generic GIS software performs five processes or tasks: input, manipulation, management, query and analysis, and visualisation. However, the study requirements warrants for the software, which can extend the traditional GIS data model to include transportation data objects such as, transportation networks, matrices, routes and route systems, linear-referenced data etc. It should also include transportation application modules like, network analysis, transportation planning and travel demand modeling, vehicle routing and logistics, and districting and location modeling etc. The best such state-of-art software available is TransCAD, (Caliper 1996). The application modules in TransCAD are fully integrated with GIS functions for improved performance and ease of use. Hence, looking to all these capabilities, TransCAD was an obvious choice for use in this study. Its objects and application modules were accordingly used to calibrate and apply the various models, to get faster and effective results, and also for the effective graphical representation of the results.

To use the various objects and application modules, the whole database was created and maintained in TransCAD, which includes both the spatial and non-spatial data. It was maintained in five different geographic layers on the digitized map. All the data pertaining to traffic analysis zones (TAZs) was coded in zone layer, which includes zone ids, zone area, and other planning variables for each zone. The road layer contained the details of all road links and also the centroid connectors, including link ids, link type, link distance in kilometer, capacity in passenger car units (PCU) per hour and in person trips per hour (for both directions), free flow and average speed in km/h, free flow and average time in minutes. The node layer contained all the data for endpoints of every links and also of zone centroids. The suburban rail network passing through study area was coded separately in rail layer and rail node layer, the section between two railway stations was identified as a rail link and the stations were considered as the rail node, consequently the respective data was coded in the rail and rail node layers.

Base Year O-D Person Trips Matrices Generation
To apply the proposed methodology on the case study area, the base year was taken as 2001 and the planning year as 2031. For converting the daily O-D matrices to peak hour matrices, the average daily to peak hour ratio was taken as 8, as observed in the study area. Also, the average car occupancy rate was taken as 2.8 for converting the capacity of links to pphpd. The link performance function (equation 1) parameters alpha and beta were set by link type. Using the above set-up, the assignment was done and the matrix was validated by comparing the assigned and observed link flows across the screenlines.

Base Year Travel Demand Modelling
The application modules for travel demand modelling in TransCAD were used to calibrate the models and forecast the travel demand pattern for the horizon year. Considering the differences in the level of development, separate trip end models were developed for the central business district (CBD) area having 72 zones, and for the south-east and north-west (SE-NW) suburbs of TMC having 43 zones. The planning variables used for development of zonal least square linear regression models for trip ends are presented in Table.2. The trip end model obtained for CBD area is given as:

Table.2- Planning Variables for Trip End Models

Variable Name Description
X1 Zone area in square kilometer
X2 No. of household in base year 2001
X3 Population in base year 2001
X4 Residential workers
X5 No. of private vehicles owned
X6 Service employment
X7 Trade employment
X8 Industrial employment
X9 Total employment
X10 No. of students
Y1 Trip production
Y2 Trip attraction

In the next step, the required inputs for calibration of gravity model were given, and the calibration results were obtained. The estimated and observed mean trip lengths, in minutes, were obtained to be 13.047 and 13.343 respectively. The value of the calibration parameter of the exponential function r was obtained to be 0.1393. Fig. 3 gives the comparison between the observed (OTLD) and estimated trip length frequency distribution (ETLD), it can be seen that they matches very closely.

Forecasting of O-D Person Trips Matrices
The forecasted value of planning variables population, total employment, and residential workers were obtained as, 30,366,58, 10,020,97, and 10,324,64 respectively. To obtain the forecasted production and attraction for each zone, equations (11) to (14) were used. The intra-city (i.e. for 115 internal zones) person trips were forecasted using the gravity model in equation (3). The Furness growth factor method was used for forecasting the external trips. Finally, the O-D person trips matrices of internal and external zones were merged together to obtain the forecasted daily O-D person trips matrix for the year 2031. The daily matrix obtained was factored by the daily to peak hour ratio of 8, to obtain the peak hour O-D person trips matrix.


Fig.3- Trip Length Frequency Distribution

Identification of Rail Corridor
Based on the curve developed in GOI(1987), the potential share of PT modes against OP modes for Thane city with a projected population of about 3.04 million in the year 2031 (Table.1), was taken as 64 %. Before assigning the forecasted flows on to the network to obtain the travel demand pattern for public transport (PT) modes, the existing suburban rail links were converted into equivalent road links in the road layer. To do so, the suburban rail links were digitised into equivalent road links in the road geographic layer and suitable capacity values were assigned to them. Based upon some earlier studies done (GOI 1987, MMRDA 1998), the values of normal and crush load were taken as 5 Standees/m2 and 8 Standees/m2 respectively.

The maximum peak hour ridership was obtained as 34,369 pphpd. Fig. 4 shows the forecasted ridership pattern on the study area map, as generated in TransCAD. To obtain the alignment for the new rail transit system, the mass transit systems considered for detailed capacity assessment were: street transit systems (ST), light rail transit 1 (LRT1), light rail transit 2 (LRT2), Rapid Rail Transit (RRT or Metro), and Regional Rail Transit (Suburban Railway).

The capacity ranges obtained are as shown below:

Street transit system = Up to 12,000 pphpd
(Mini-bus, single decker standard bus, double-decker and articulated)
LRT1 = 12,000 to 36,000 pphpd
LRT2 = 36,000 to 50,000 pphpd


Fig.4- Map Showing the Results of Assignment for Horizon Year Peak Hour PT Matrix, Generated in TransCAD.

RRT (Metro) = 50,000 to 69,000 pphpd
RGR = 59,000 to 89,000 pphpd

More detailed calculations can be referred from Verma and Dhingra (2001). Since the maximum peak hour ridership was found to be 34,369 pphpd, the most suitable system recommended for operation on the newly identified corridor will be Light Rail Transit 1, as the maximum peak hour ridership falls within its capacity range.

Results and Discussion
Fig. 5 graphically represents the newly identified rail corridor on the study area map. Also, Table. 3 give the details of the links, which form the newly identified rail corridor, their length in kilometres, and the corresponding ridership estimates on these links. It can be observed from the table that all the links have ridership level within the capacity range of LRT1, except a few, which were included to give continuity to the rail corridor. The newly identified corridor consists of two routes, the length of 1st route is 9.79 kms, and the length of 2nd route is 11.66 kms. Both the routes start from link no. 85, they move parallel and cross through the CBD area of Thane till link no. 209, from where they change their direction, route 1 move towards the north-west part of the study area, up to link no. 9, and the route 2 move towards the south-east part of the study area up to link no. 277.


Fig.5- Newly Identified Rail Corridor for Thane City

These results shows that besides the requirement for proper connectivity within the CBD area, the higher development in future of north-west and south-east zones also warrants a proper connectivity in the form of a rail based transit system. These routes being obtained using the user equilibrium approach will allow the user to travel from its origin to final destination via. the shortest possible path, provided the feeder bus routes and integrated schedules are also developed. Also, being aligned along the maximum ridership corridors, the routes will ensure maximum revenue returns for the operator. Hence, it can be claimed that by using the proposed methodology the routes will be aligned along the optimum corridors, from both the users’ and operator’s point of view.

While implementing the routes, the chances of getting the required land within the CBD area are generally very less, hence they can be constructed either underground or on elevated tracks, but, as the routes move out of the CBD area, they can be brought on to the surface, as the required land would probably be available.

Table.3- Details of the Links Forming the Newly Identified Rail Corridors

Route 1 Route 2
Link No. Length (Kms.) Ridership (pphpd) Link No. Length (Kms.) Ridership (pphpd)
85 0.250 34,369 85 0.250 34,369
78 1.0 26,129 78 1.0 26,129
83 0.6 14,204 83 0.6 14,204
109 0.15 16,296 109 0.15 16,296
110 0.15 15,646 110 0.15 15,646
117 0.30 16,176 117 0.30 16,176
121 0.20 16,105 121 0.20 16,105
209 0.10 20,074 209 0.10 20,074
225 0.40 8,002 221 0.35 15,071
227 0.70 4,280 212 0.25 13,227
234 0.20 5,118 213 0.30 24,217
243 0.750 13,845 324 0.30 26,959
244 0.10 27,261 252 0.72 18,769
48 0.53 21,314 257 0.77 15,774
42 0.83 17,378 258 1.02 3621
23 0.90 18,386 262 0.68 1169
14 0.70 18,302 273 2.77 3522
11 1.03 19,159 274 0.69 14,580
9 0.90 12,974 275 0.270 13,771
Total Length = 9.79 Kms 276 0.22 12,940
      277 0.57 12,155
      Total Length = 11.66 Kms

In the proposed model, by assigning the O-D flows using user equilibrium approach, it is ensured that the higher flow levels are obtained on the major roads (with multiple lanes) only, because the flows are assigned iteratively based on the capacity measurements of each link. This way it is ensured that the rail alignment will pass through the major roads only, giving ample space to construct elevated tracks (or even surface tracks) for the rail system within CBD area, which will be economically cheaper than the underground system.

Summary & Conclusion
This paper discusses a model for identifying demand-oriented urban rail transit corridor on a city road network, using geographical information system (GIS) tools. The proposed model is able to identify new rail corridors, which are optimum from both the users’ and operator’s point of view. The corridors were identified based on the demand obtained using user equilibrium approach, and suitable recommendations were made on the technology that can meet this demand.

The conclusions from the study are as follows:

  • Since the ridership estimates are obtained using the user equilibrium approach, the routes identified will allow the user to travel from its origin to final destination via. the shortest path, thus achieving the user optimum.
  • While comparing the ridership estimates with the capacity, the suitability range of capacity (pphpd) for each transit system is between normal load and crush load. The lower bound ensures that the peak hour flow on most of the links of the identified route is more than the normal load capacity of the rail transit system. This way steady revenue returns are guaranteed for the operator. On the other hand, the upper bound makes certain that the desired level-of-service for the user is achieved.
  • The routes identified using the proposed model effectively covers the study area i.e. both central business district CBD and outer areas. Since, most of the links in the identified routes have ridership level above 12,000 pphpd, the routes will be demand oriented and will give steady revenue returns to the operator.
  • Since, the maximum peak hour ridership was found to be 34,369 pphpd. The maximum peak hour load in LRT1 will always remain less than its crush load.
  • GIS turns out to be an excellent tool for transportation related applications. The various application and generic modules of TransCAD helps immensely in calibration and application of transport models at various stages of the study. It also serves as an excellent tool for graphical representation of the results. To provide reasonably direct access to the commuter from its origin to final destination, and also to achieve the estimated ridership on the newly identified rail corridor, the feeder bus routes and integrated schedules for feeder bus and rail operation can be developed. The feeder routes can be designed taking into consideration the estimated ridership pattern on every link within the feeder area, and also the access and egress demand of the station. The integrated schedules can be developed by optimizing the combined objective function of operator cost and transfer time and comfort for the user.

References

  • Bodell G. (2002). “Urban rail demand forecasts – where do models go wrong ?” Smart Urban Transport, 1(2), 6-9.
  • Caliper. (1996). Travel demand modeling with TransCAD 3.2, Massachusetts, U.S.A. Consulting Engineering Services India Ltd. (CES). (2001). Proposed Mass Rapid Transit System for Thane City, Draft Final Report, MSRDC, Mumbai, India.
  • Chien S., Schonfeld P. (1998). “Joint Optimization of a rail transit line and its feeder bus system.” J. of Adv. Transp., 32(3), 253-284.
  • Clark J., Oxley P. (1991). “Strategic public transport planning in Baghdad.” Urban transport in developing countries, Heraty M., eds., PTRC Education Research Services Ltd., London, 224-228.
  • Gipps P.G., Gu K.Q., Held A., Barnett G. (2001). “New technologies for transport route selection” Transp. Res., 3(9), 135-154. Government of India (GOI). (1987). Report of the study group on alternative systems of urban transport, New Delhi, India.
  • Mumbai Metropolitan Region Development Authority (MMRDA). (1998). Working paper: review of various mass transit systems, Mumbai, India.
  • Moorthy N.V.R. (1997). “Planning of integrated transit network for bus and LRT.” J. of Adv. Transp., 31(3), 283-309.
  • Ortuzar J.de.D., Willumson L.G. (1996). “Modelling transport.” Wiley, England, pp.180.
  • Verma A., Dhingra S.L. (2001). “Suitability of alternative systems for urban mass transport for Indian cities.” Trasporti Europei, VII(18), 4-15. Verma A., Dhingra S.L. (2002), “GIS Technology for Transportation and its Trends”, Proc. All India Seminar on Remote Sensing & Geographical Information System Techniques, The Institution of Engineers (India), Indore, India.

Acknowledgements
The authors would like to thank Maharashtra State Road Development Corporation (MSRDC) and M/s Consulting Engineering Services (I) Ltd., for providing all the necessary data for this study.