Jourrrtrl of Trmcporr G‘cqrcrph~ 1991 2( 2) 101-I IO
A spatial decision support system for transportation policy analysis in Bangladesh Bruce A. Ralston Department
of Geography.
George
Tharakan
The World Bunk.
Cheng
Room
University
HIO-123,
of Tennessee,
1818 HSt NW,
KnoxwIle.
Wushingtort,
TN 37996-1420,
USA
DC 20433, USA
Liu
Ouk Ridge Nationrrl Laboratory,
PO Box 2008, 4500N,
MS 6270, Ouk Ridge,
TN 37836270.
USA
Evaluating various alternatives for transportation policy and investments naturally leads one to consider a series of ‘What if’ questions. Merging the technologies of geographic information systems with the analytic power of transportation models can yield a spatial decision support system capable of assessing the impacts of a wide array of planning options. We report here on the development and application of such a system, the Bangladesh Transportation Modeling System. Building the BTMS forced us to resolve several of the issues involved in merging GIS technology with transportation models.
Evaluating various policy and investment alternatives for transport development naturally leads one to consider a series of ‘What if’ questions. Merging the technologies of geographic information systems with the analytic power of mathematical models can yield a spatial decision support system (SDSS) capable of assessing the impacts of a wide array of options. We report here on the development and application of such a system, the Bangladesh Transportation Modeling System (BTMS). The BTMS, which was developed for the World Bank with funding from the United Nations Development Program, is a microcomputer-based software package for analysing transportation policies and investments in Bangladesh. While developed for Bangladesh. many of the issues addressed in its construction and application apply to any SDSS which ties geographic information systems with analytic models.
Project
overview
Development agencies often are faced with many options with respect to transportation investments and policies. Several models are available for estimating the effects of changes in the transport system
on modal splits, route assignment and trip distribution. However, these models often exist as discrete units and are not part of an integrated decision support system. Further, the difficulty in using many software packages and interpreting their results has been a barrier to decision makers using techniques of geographical analysis. The BTMS was developed, in part, to overcome these problems. Specific goals of the project were as follows:
(1)
Develop an SDSS for analysing proposed transportation policies and investments for use by the Bangladesh National Planning Commission and various line agencies such as the Roads and Highways Department (RHD). Bangladesh Railways (BR), the Bangladesh Inland Water Transport Corporation (BIWTC) and the Bangladesh Inland Water Transport Authority (BIWTA). (7 Calibrate the models in the SDSS using secondary data sources. (3) Use the SDSS to evaluate proposed transportation projects and policies for the Fourth Five Year Plan (FFYP). (4) Use the SDSS to support the World Bank in formulating long-term, multimodal transportation policies.
In order to use an SDSS to assess the effects of proposed interventions, planners must be able to simulate the movement of passengers and a variety of goods on various transport modes, over alternative routes. and under conditions which in Bangladesh are subject to significant seasonal variations. Various transportation modelling tasks. including trip distribution, modal split, route assignment and cost savings calculations. need to be performed. Further. a high degree of flexibility is needed so that many possible ‘What if’ questions can be studied. Since transportation is obviously a spatial process, interactive graphics play a key role in the maintenance of network data and the display of model results. Indeed, the use of interactive mapping is crucial in studying the regional impact of transport investments. The strength of an SDSS lies in its ability to project the effects of changes to a system. We constructed the BTMS to support changes in: (1) (2) (3) (4) (5) (6)
network structure; network performance characteristics; the transportation pricing policies; supplies and demands; the commodities studied; the modal split propensities of commodities passengers; (7) trading partners: (8) the seasonal variations in transport flows costs; (9) the characteristics of individual links.
and
and
These changes are accommodated by minimizing the number of assumptions which are ‘hard coded’ into the BTMS software. Instead, there are input files and run time options which can be edited to reflect the changes listed above. We have tried to make the editing of these files and options as simple and straightforward as possible. To simulate the effects of these changes. appropriate analytic models had to be included in the system.
Models in the tool box The merging of spatial analysis tools with GIS databases and mapping technology is crucial if GTS systems are to achieve their full potential. To study the types of changes presented above, we had to determine the types of models which should be in our ‘tool box’. At the beginning of the project it was not clear whether cost and delay functions in Bangladesh were flow dependent. Further. even if flow dependent. nonlinear models were appropriate. it was not clear that the data needed to support them would be available. Therefore. models based on both linear (fixed costs and times for a given arc) and nonlinear functions (cost or times dependent on flow volume) were incorporated into the system. The modelling types based on linear cost and delay 107
functions are referred to as ‘logit-based models’, while those based on nonlinear cost and delays are referred to as ‘equilibrium approaches’. Since both approaches require that costs and times be assigned to network links. we begin with a discussion of how the BTMS makes such assignments. Network
characterization
Each link of a transport network can be classified as either a physical link or a logical link. The physical links, which are normally part of a GTS, arc the roads, rail lines and inland waterway routes. The links connect nodes, which are the towns, villages and zila (district) headquarters of the country. The logical links are constructed at nodes to capture pickups, deliveries and intermodal transfers. Logical network entities are needed to model the effects of making investments in facilities at specific nodes, eg. new container cranes at the country’s major ports. Physical links of the same mode are subdivided into classes. For example. there are four classes of roads in the Bangladesh network (paved, unpaved, minor ferry crossing, major ferry crossing). Each class has cost (expressed in Taka. the Bangladesh unit of currency, 35 Tk = US$l.OO). capacity and speed characteristics. In addition, each link in the network has associated with it unique cost and time multipliers. These allow the user to adjust the cost and time for a specific link. Thus a poorly maintained paved road link may be given a higher cost or time than the average for paved roads. For cost and time assignments, physical links are divided into two types: all links except ferry croscings, and ferry crossings. In the first case, the time to traverse the link is a function of the length of the link and the speed of the mode being used. The cost of a non-ferry physical link is dependent on its length and the per-ton kilometre or per-passenger kilomctre cost of the mode being used. Thus. both costs and times are functions of the length of each link and the link’s mode classification. The values for cost and time for each link arc multiplied by the link specific cost and time multipliers and mode ceasonality multipliers. For linear cost and delay functions. the resulting cost and times per unit for non-ferry physical links are as follows: COST where: L VC LCM SCM TIME where: L s
= I,“‘VC’LCM’“SCM = the length of the link, in kilometres. = the variable cost for the mode being used, in Tk/ton-km, = the link’s unique cost multiplier. = the mode’s seasonality cost multiplier. = LIS”;I>TM”STM = the length of the link, = the speed of the mode type. in kmph.
on links of that
LTM STM
= the link’s unique time multiplier, = the mode’s seasonality time multiplier.
The link and seasonal multipliers are set by the user. The values used in the link multipliers reflect the state of repair on each of the links. This information is supplied by the line agencies responsible for each mode’s network maintenance (the Roads and Highways Department, the Inland Water Transport Authority, and Bangladesh Railways). The seasonal multipliers are based on observed changes in mode performance. For example, in the monsoon season, some inland water transport arcs become deeper and can accommodate larger ships. Each season is defined by the user, but our experience indicates two seasons (dry and monsoon) are enough to capture the changes in the network. For ferry links the cost and time on the link are computed based on fixed costs and delays (ie, independent of link length). There are two types of ferry crossings: those run by BIWTC and those run by RHD. The RHD ferries are the more numerous and smaller of the two. BIWTC crossings are along the major rivers in the country and are serious bottlenecks. Both BTWTC and RHD ferries are assigned fixed costs and times. Variations for individual cases are captured using each link’s unique cost and time multipliers. Thus, the cost and time functions for ferry links are as follows: COST where: FFC LCM SCM TIME where: FFT LTM STM
= FFC*LCM”SCM = the fixed ferry charge, in Tk per ton or per passenger. = the link’s unique cost multiplier, = the mode’s (ferry type) seasonality time multiplier. = FFT*LTM”STM = the fixed ferry time, in hours. = the link’s unique time multiplier. = the mode’s seasonality time multiplier.
Logical links Logical links can be divided into two classes: loading and unloading links, and intermodal transfer links. The costs and delays associated with loading and unloading arcs are as follows. COST where: MFC CFC
NCM TIME
= (MFC
+ CFC)*NCM
= the mode fixed loading or unloading cost, in Tk per ton or per passenger, = the commodity (or passenger) fixed loading or unloading cost (ie, extra handling charges due to special characteristics) in Tk per ton or per passenger. = the node cost multiplier. = (MFT + CFT)*NTM
where: MFT CFT
NTM
= the mode fixed loading or unloading time, in hours, = the commodity fixed loading and unloading time (ie, extra handling time due to special characteristics) in hours, = the node time multiplier.
Intermodal transfer cost and time functions are similar. The only difference is that the fixed cost and time values are for intermodal transfers, not for loading and unloading. Any path between an origin and destination comprises a chain of logical and physical links. The sum of the costs and times over each link in a path determines the path cost and time. The logit-based models The logit-based approach assumes that costs and delays along arcs are fixed (independent of transport volume). although they may vary by season. Mode choice is assumed to be a function of trip cost. trip time and possibly other factors such as reliability or comfort. Destination choice, in a trip distribution model, is assumed to be a function of a ‘friction of distance measure’ and the amount of demand at alternative destination points. Given these assumptions. it is possible to calculate trip distributions, modal splits and route assignments. The modal split programs calculate the modal shares for each origin-destination-commodity (ODC) combination. The modal shares are based on the logit formula (Ben-Akiva and Lerman, 1987): P(mlijk) where: P(mlijk)
= exp(~,,~I,I)~~Nexp(~,t~II)
= the probability of using mode ~1. given that there is to be a shipment of commodity k from i to j. u 1Jhrn= the utility associated with shipping the commodity k from origin i to the destination j via mode m. The utility function is assumed to have the form: u IJj\lll = PIAC,,X,,, + Pzx T,,x,,,V/, + P3~4 + Pu& where: C lJX??l= the cost of shipping commodity k from origin i to destination j via mode m, T r/r( k from origin I?, = the time to ship commodity i to destination j via mode 1~. = the value of commodity k (for passengers VX this is set to I), = 1 if mode 111is Rail. 0 otherwise. 6, = 1 if mode m is Water, 0 otherwise. & The fl values, which are assumed to vary by commodity type, act as weights in determining the utility of travel. Their signs and values are crucial to all mode choice and routeing assignments made in the BTMS. fllx is expected to be negative: the higher IO3
the cost of a mode. the lower its utility. Similarly,/L~ should be non-positive because trips involving longer travel times are to be avoided. The ratio ofP2,! top,x is referred to as the value of time. The remaining Jj values can be of any sign, and reflect the relative advantages of one mode over another. For example, we might expect that the /3 value for rail transport would be positive for passengers owing to the greater comfort afforded by rail. The BTMS assumes up to four modes available for any shipment: road. rail, water and mixed mode. The least disutility path for each mode is chosen as the basis for evaluating the mode choices. The mixed mode is considered only if an intermodal transfer arc is present in the path. The utilities for each of the four modes are then used to make modal split calculations. The trip distribution programs calculate the amount of a commodity (or passengers) shipped between any O-D pair via a doubly constrained spatial interaction (or gravity) model (Wilson, 1970). The formula for calculating that amount is: T rlh = &~,xO,hD,h’P,,x” where: T,,, = the amount of commodity k shipped from i to ;. balancing factor that ensures A,, = a supply c,T,,x = Orht = a demand balancing factor that ensures B,h &T,, = D/h. 0,1, = the supply of commodity k at node 1. D,h = the demand for commodity k at node j, P I/h = the propensity to ship commodity k from i to j. P,,h is calculated as follows: P l/h = &$Xp( u,,hm) a and r are parameters to be estimated. The first is a ‘friction of distance’ parameter, the second a measure of scale effects in attraction. For example, a r value greater than 1.0 would indicate that if a destination’s demand doubled, its attractiveness would more than double. The BTMS treats trip distribution modelling as a nested logit model. First, the utility of shipping one unit of k between i and j by mode m is calculated for each mode m. The resulting utilities are then sent ‘up to’ the trip distribution model for the second round (or outer nested) calculation - the trip distribution stage. The logit-based network assignment program assigns the origin-destinationcommodity-mode-amount combinations (an output of the modal split programs) to the least disutility path by each mode.
Cotlgestiot~ effects We originally included an equilibrium assignment in the BTMS package module as an option (Aashtiani and Magnanti, 1982). However. we were 103
unable to find data suitable for calibrating flow volume-delay functions. The equilibrium model was never used for investment analysis, and so is not reported on here. We believe that applying the logitbased models on a seasonal basis adequately captures changes in the costs and times on the network. Culcrrlatittg savit2g.s The modules described above generate the total cost (in Taka) and the time (in ton-hours or passengerhours) for each network assignment. When this information is generated for two different network specifications (eg. differing network structure. cost functions or modal split propensities), it is possible to calculate the change in total cost and time. Thi\ information can then be used in determining the benefits from changes in the transport system. It should be noted that the function used to ‘cost out’ a network assignment need not be the same one used to generate it. This is a valuable option when considering cases where tariffs have little relation with the economic or financial cost of supplying transport services. In Bangladesh, as in manv countries, passenger tariffs on some modes are subsidized. Thus, some passengers pay less than the cost of supplying passenger transport services. Passengers may make mode choice decisions based on tariffs. However. to evaluate the benefit to society it is necessary to cost out the passenger flows with appropriate economic costs. The BTMS allows the user to use one cost function for assignment of flows, and an alternative cost function for evaluating the economic cost of that assignment. As a group. these models give an analyst the ability to simulate the flow of commodities and passengers over a network under various scenario assumptions.
The geographic
data
The geographic data used in the BTMS consist of information on administrative units (polygons). network links (lines) and towns (points). These GIS primitives were digitized from a series of maps supplied by the Bangladesh Planning Commission, RHD, BIWTC. BIWTA, BR and the World Bank. The maps were digitized into four files. The first file consists of the administrative outlines in Bangladesh. This boundary file contains left and right polygon pointers, which indicate whether the boundary is a country boundary, a division boundary (four administrative divisions) or a zifu (64 districts) boundary. The remaining digitized files relate to network structure. There are three such files: a node file, a link file and a graphics file. The node file contains information on each physical node in the network, such as its name, ID number, location and a flag which indicates whether it is a centroid (supply or
demand point) or intermodal transfer point. These flags serve as keys in the construction of the logical nodes. The link file contains information on the physical links (their mode, from-node and to-node IDS, and cost and time multipliers), while the graphics file contains the points in each arc needed to draw the link on a map accurately. These files make it possible to construct maps of the country, and its transportation network. The polygon potnters in the outline file allow one to display all administrative boundaries, or selected subsets of them. The node flags allow various types of nodes to be turned ‘off’ or ‘on’. The link mode classifications allow the same type of turning ‘off and ‘on’ to be applied to modes in our network.
Relating GIS data structure
to model needs
The data files described above present a relational model of geographic data. The link file is related to the graphics file via link IDS. The nodes are related to links via origin and destination node IDS. The analytic models also need information on links and nodes. However, most transport models perform poorly using the type of data structure described above. Further, the GTS databases do not contain any information on logical nodes and links. These must be constructed. The BTMS uses a program to transform the digital representation of the physical network into a logical network. Building a logical network encompasses several steps. First. for each physical arc, its bidirectional twin arc is constructed. Second. all the logical nodes and arcs needed to capture pickups. deliveries and intermodal transfers are constructed. Once the logical structure at each node is constructed, the network is then rewritten in a forward star data structure. This type of data structure greatly increases execution speeds for path algorithms, which form the core of the analytic models (Van Vuren and Jansen, 1988). Thus, we keep two data structures: the GTS relational structure accessed by the mapping programs; and the forward star data structure accessed by the analytic models. Keeping two different data structures for the same information does involve some overhead. If a network is changed (links or nodes added, deleted or modified), the logical network generator must be run on the new data files. We have found that this extra computation is more than compensated for by the increased speed in analytic model execution.
Managing files and data As with any large modelling system, data must be maintained and files must be passed between programs. Users in Dhaka of an early version of the BTMS indicated that keeping track of the various files used in a scenario analysis and assigning
consistent file names were difficult, time consuming and sometimes confusing tasks. We therefore redesigned data and file management systems to address their concerns.
Data in the BTMS can be classified as belonging to one of two types: data that do not relate the toplogy and characteristics of specific network elements, and those which do. A spreadsheet program with macros is used to manage the first type of data. When retrieved, the spreadsheet displays a macro directory that lists the commands needed to access various parts of the spreadsheet. We chose a spreadsheet program rather than a database management package for two reasons. First, we felt it was easier to teach the use of a spreadsheet program to our users. Second, the ultimate recipient of this package, the Planning Commission of Bangladesh, is also responsible for collecting data on transport flows. We believe that use of the spreadsheet program will be of use to them in their other work. The spatial data are accessed and edited via the interactive graphics program. The BTMS contains options for spatial queries, via point-and-click methods, and spatial data editing. These are operations found in many GIS packages.
The BTMS, like any other decision support system, uses input and output files and has several run-time options. For a given scenario, all files and options set by the user are identified in a catalogue. A set-up program accesses the catalogue which is used to activate these files and set the options. For example, all files used for a scenario with the Jamuna Bridge (a major proposed network investment) in the network are identified in a catalogue called JAMUNA.CAT. The set-up program treats such possible files and options as elements of a catalogue data structure. Each program in the BTMS reads the same data structure to access its input and output files and run-time options. Grouping the files and options in this manner minimizes the chances of mixing data for one scenario with that for another. Accessing
model
results
One of the drawbacks of computer models of spatial processes is that they often produce reams of output that are difficult to interpret. The spatial impact of policies can be lost in the sheer magnitude of data. We have tried to develop reporting systems that communicate a scenario’s results directly through the graphics interface to allow the user to examine the results in the geographical context of the network. For example, the network flow assignments can also be studied with the interactive graphics program. The flows for each commodity can be plotted on the 105
map of Bangladesh. tound by clicking on helpful when trying investment on other regional interactions. gram can also be used cost, time or disutility and origin-destination
and the flow amount can be the links. This is particularly to asses\ the effects of an parts of the network and on The interactive graphics proto calculate shortest distance, paths for specific commodities pairs.
Calibration To use the BTMS, it is necessary to calibrate the various parameters used in the models. In addition to the BTMS, we also developed a calibration package so that members of the Planning Commission can update the model parameters as new data become available. The programs in the calibration package are built around a maximum likelihood estimation program for logit model parameter\. During a visit to Bangladesh. we collected data on cost functions. running speeds and turnaround times from the various line agencies. We also travelled on several transport modes in an effort to understand better each mode’s characteristics. When combined with other reports supplied by the World Bank, enough data were available to estimate cost functions, modal split parameters (the j1 values) and trip distribution parameters (a and r) for the major commodities. Data for freight were taken from the 1978 Bangladesh Transportation Study. Although these are fairly old data. they make the most complete data set which could be found. From this we were able to estimate modal split parameters for various commodities.’ We had a difficult time estimating utility functions for commodities which travel by traditional (non-mechanized) transport. Estimating modal split and trip distribution parameters for passengers has proved even more difficult. Observed trip tables of passenger movements were not available for Bangladesh. We suspect this reflects the bias of many national transport planners who think ‘development’ is synonymous with &freight’ (Owen, lYX7). In a poor country with a high population density (Bangladesh has over I IS million people in an area slightly larger than Wisconsin), passenger movements dominate the transport sector. Savings to passengers are often the largest component of benefits from transport investments. Being able to generate modal split and trip distribution parameters for passengers is clearly important. Faced with the lack of data on passenger movements. we were forced to use value of time and trip generation parameters used in the feasibility study
I06
for the Jamuna Tritton, 1989). Application:
Bridge
project
the analysis
(Rendel
of proposed
Palmer
and
actions
In contemplating the future cvolutmn of the main transport network of Bangladesh. it is necessary to consider a variety of alternative scenarios with respect to major investments or policy options. The scenarios considered below involve those types of strategic investments or policy actions likely to have an impact on the overall functioning or efficienq ot the main transport system. An example of a strategic investment decision is the proposed Jamuna Bridge investment. which would radically transform transport flows in western Bangladesh. Similarly. an example of strategic policy actions is those involved with changes in regulated tariff policies The \cenarias considered are as follows.
A base network against which all other scenarios would be evaluated was first established. The base network selected is that which existed in April IYYO. Road, rail and inland waterway tariffs. speeds and other service characteristics are those estimated by a field mission which visited the country during that month. Transport flows and system costs are cstimated for the year lYY6 based on projected demands for that year. These flow patterns and the resulting aggregate transport system costs serve as bcnchmarks against which all of the other projected scenarios were compared.
The road sector scenarios involve two type\ ot investments - bridging of ferries and improvement of highways under various rehabilitation projects. Two scenarios are considered. Scenurro .?: ittlplcttlenlatiot1 of cdl RHD projrcts proposed for the Fourth Plm IW-199.~. The Roads and Highways Department (RHD) has proposed a core programme of investments to be undertakon within the Fourth Five-Year Plan (FFYP). These investments include rehabilitation to improved standards of parts of the national. regional and feeder road network. and a number of bridges on the main network which together involve an outlay of about Tk 30 000 million. Scetiurlo 3: ittiplettietitcitlt,rl oj N ttiujor hrrtigitig prograrnrnc for nutionul highuwys. The national highway network includes 20 RHD operated ferries in addition to the major ferries operated by BIWTC (Frgwe I). The Bangladesh government has emphasized a programme of bridge construction which has been the fastest growing segment of the road sector investment programme. This scenario seek\ to assess
-
Primary
Roads
...... ..‘.. Secondary -
Figure I.
Major
ferry crossing
Roads
Major Ferry Crosslngs
location\
Figure routes
the likely impact of this strategy on the transport system if in fact the strategy were carried out over the entire national highway network. The scenario includes the Jamuna Bridge in addition to all the minor bridges necessary for eliminating all RHD ferries on the national network. Rail sector
(Scenarios
4 to 6)
The scenarios of interest for the rail sector related to operational improvements and divestiture of uneconomic segments of the rail network. The scenarios examined were as follows. Scenario
2. The
rail
-
Major Water Transport
-
Core Metre Gauge
Rail
Routes
Rail
-
Core Broad Gauge
-
Non-Core
Metre Gauge
Rail
e
Non-Core
Broad Gauge
Rail
system
and
major
water
transport
shutdown of the north-west metre gauge network BR’s overall traffic performance.
on
Scenario
6: reduced wagon ferry delays. Presently. crossing the Jamuna River experience wagons average delays of about 15 days in each direction (see Figure 2). With improved operational controls, it is believed that average delays to trans-Jamuna wagons can be reduced to two days in each direction. This scenario examines the effect on BR’s competitive position of such an improvement in the wagon ferry operations.
4: closure of note-core rail network. The core network for Bangladesh Railways (BR) is shown in Figure 2. These routes account for 1638 km of the BR route network. The remaining I I07 km of noncore routes were assumed closed to freight and passenger traffic.
These scenarios relate to improvements in the inland water transport sector and allocation of import and export cargoes to the two ports of Chittagong and Mongla. The scenarios examined are as below.
Scennrio 5: closure wvrk. In the event
Scerlario 7: core programme for inland water transport urlder the Fourth Plan. This scenario includes
of north-west
metre
gauge
net-
that the Jamuna Bridge is built without a metre gauge rail link, BR believes it will lose a significant portion of its freight traffic to the north-west and as a result would have to shut down the metre gauge network west of the Jamuna River. This scenario seeks to assess the impact of such a
Water transport
(Scenarios
7 to 8)
improved access to the north-east, Baghabari Port improvements and improvements in night navigation on the waterways. Essentially, this programme upgrades the river system serving Sylhet (source of stone materials) and Baghabari Port to Class IT I07
waterways (6 ft draught) and improves tion on all Class I (13 ft draught) waterways - thereby reducing average
night navigaand Class II transit times.
Scenario 8: rational allocation of freight to seaports. A majority of cargoes now arrive at Chittagong Port. and some of this cargo needs to move to the northwest. This scenario examines transport flows when cargo moving through the two ports is rationally allocated based on the inland origin or destination.
taneously. No attempt is made here to examine the joint effects of various scenarios as this would lead to a large number of combinations. The BTMS does. however, provide the analyst with the possibility ol examining such combined scenarios in the process of putting together a package of measures for development of the sector. As an example, a combined scenario which includes Scenarios 2. 6, 7, 8, 9 and IO was analysed. Tuhle I summarizes the results obtained for each scenario.
Trunsport
Road sector scenarios
policy
(Scenarios
9 to II)
Specific Government policies on tariffs and prices of inputs affect the functioning of the transport sector. The policy scenarios analysed include the effect of Government policies on passenger tariffs. rail freight tariffs and fuel prices. Scenario 9: enhanced pussenger tarijJs. Regulated passenger tariffs for road and water transport were set by Government at Tk 0.24 per passenger-km. This level of tariff does not allow full cost recovery at reasonable (80%) average load factors. Rail passenger tariffs average Tk 0. I5 per passenger-km and have resulted in a cost recovery factor for BR passenger operations of around 30%. This scenario examines the effect of raising road tariffs to Tk 0.35 per passenger-km. water transport tariffs to Tk 0.30 per passenger-km and rail tariffs to Tk 0.35 per passenger-km. Scenano 10: reduced rail freight tariffs. BR’s tariffs for freight are at about the same level as those for road transport, thereby giving the shipper (except those located on a rail head) no incentive to use rail when there is an additional cost for accessing rail facilities. BR freight tariffs are at least 50”L above those of other railways in Asia. As a result, practically all BR freight traffic is for Government account. This scenario examines the effect on traffic volumes and BR’s competitive position of reducing the rail freight tariff by 30%. Scenario I I: fuel price increase of50%. As a result of increases in the international price of crude oil, fuel prices in Bangladesh are under strong upward pressures. A 50% increase in the price of fuel is estimated to result in at most a 20% increase in road transport costs. This scenario examines the effect of such an increase on the future competitiveness of road transport as an indicator of the likely influence of fuel prices on road transport development strategy.
Results The 11 foregoing scenarios are intended to provide indications of relative impacts and the directions of change which would result from various actions. These scenarios. however, are not mutually exclusive, and clearly many of these could occur simul108
Scenarios analysed for the road sector include the package of investments proposed by RHD for the Fourth Plan period, and an extreme bridge construction scenario where all RHD ferries on the national highway network are replaced by bridges. The last i\ clearly not a likely scenario (involving construction of some 10 bridges) but is of interest in that it indicates the potential for expansion of the role of road transport. The entire RHD Fourth Plarl package (Scenario 3) results in a reduction in system costs of about 7% or Tk 0.6 billion. This indicates that the package. which entails a total expenditure of Tk 30 billion. should be judiciously reduced to delete low-return investments. The major bridging programme (Scenario 3) reduced total system costs only marginally and the share of road transport. while higher, did not increase to a level which would indicate a much larger role for road transport in the future. Rail sector scenarios In the rail sector, the scenario analyses concentrated on the potential for closure of specific segments oJ’tlle rail network (Scenarios 4 and S), and removal of the critical bottleneck for rail operations at the Jamuna rail ferry crossing (Scenario 6. Figure 2). Roth scenarios for closure of parts of the rail network (non-core network and north-west MG network) resulted in a reduction in system costs - primarily in passenger operations. From the overall transport svstem viewpoint such closures are not likely to be ol significance, even though in the short run they ma> lead to some dislocations. The most important result for the rail sector is the one relating to rcducca’ rt,agon ,fhrry delays at the Jarnuria rail croxsing (Scenario 6). Currently, freight traffic on this transJamuna route suffers a delay of about IS days in each direction, and ferry operations in themselves are unlikely to cause more than a #-hour delay il operations arc carried out efficiently. This scenario indicated that the railway can substantially improve its share of the freight market (almost double) simply by reducing the wagon ferry delay. Whereas the reduction in overall system costs is not large. the effect on BR’s performance is likely to be very substantial. The investments required for the fcrr) and in view of their larger potential are minor, benefits should receive a high priority.
Table I. Results of scenario
simulations Passengers
Freight Mode share (O/b)
Scenario description
7 Cow progrzimme 1WT under FFYP
output (Bil Tkm)
System cost (Bil Tk)
Operating cost (TkiTkm)
I% 1
5x II 31 100 60 II) 2s
I Xl6
IO0 70 23 7
1.797
I00
7i)trll
100
Roxl R,uI W,itcr
61 IO 3
rl)ttrl
100
Road RUI Water To tcrl Road Rail Water Totctl
hl s 31 100 hl 7 32 1110
Water Tutu1 Road Rail Water Totrtl Road Rail Water Torcrl Road Rail
77 IO0 51 IO 3’) 100 51 Y iY IO0 5X II
output (Bil P km)
System cost (Bil Tk)
Total Operating system cost cost (TkIP Km) (Bil Tk)
h7
Ro,~d Rul \\‘atcr To tttl Road Rail Water
3 s
7.7 20 h
I.792
100
I. m5
74 I7 x 100 72 IY X
I x9
I(10
I 72.3
I. 75.5
67 ‘5 Y 100
I 59.3
I 8/h Reduced trelght taritf\ IO
Mode share
70 21 Y I 00
rail
I 704
1 895 13. Combined 2. 3. 7. x, 0. IO & I I I 4.3
tl4 38 x IO0 73 IX Y 100
Trattlc costs and \olumc\ r&r to Inter-dstrlct transport They do not include Intra-district Svtcm co\t Includes vchlclc operntlng and value of tlmc co\ts dpcrating co5t per unit IS total operating cost/output. BII = Blll~on. Thm = Ton kllometre: Pkm = Passenger kllometrc. Tk = Taka.
Water transport
scenarios
The most significant finding in the water transport sector resulted from rationalized allocation of freight betweerl the two seaports (Scenario 8), Chittagong and Mongla. Presently, Chittagong handles 80% of
and urban
transport.
the country’s foreign trade. The scenario analysed the effect of improving Mongla Port (primarily through dredging and maintenance of approach channels) to the point where the choice between the two ports becomes strictly dependent on the cost of transport to the hinterland. It appears that were
Mongla Port to be so improved, a majority of cargo now using Chittagong would switch to Mongla. This is primarily due to the availability of low-cost sheltered inland waterway routes connecting Mongla with Dhaka. In addition, traffic originating in or destined for the western part of the country (particuarly the north-west) would shift from Chittagong. This scenario also resulted in the largest system cost reduction for freight transport (from Tk 74.4 billion to Tk 20.5 billion).
Three types of policy measures were considered: enhnnced passeqer tariffs (Scenario 9), reduced rail freglzt tarqfs (Scenario IO). and an irzcrrasc itz file1 prices to .5W7 (Scenario 11). In Scenario 9, passenger tariffs were increased to cost recovery levels for all three modes. The major impact of this was a reduction in the rail share (from 25% to ?I%), with road transport picking up most of this reduction. Interestingly, system costs reduced substantially showing the efficiency gains obtainable through this policy change. The reduced rail freight tariff scenario is based on the high tariff now charged by BR for (mostly Government account) freight. The reduced freight tariff increased rail share of freight substantially from 11% to lS%, and resulted in reduced system costs. This indicates that the high rail freight tariffs now in place are leading to inefficient allocations of traffic among transport modes. The most interesting result with regard to policy measures was the effect of an increase in fuel prices which translates to a proportionately higher increase in road transport costs. As a result. two effects occurred: (a) there was a small reduction in transport demand (measured in ton-kilometres): and (b) road transport lost a significant portion to rail and water transport. Consequently, overall system costs increased by less than 3%. This indicates that fuel taxes may be a relatively efficient means of mobilizing resources for the transport sector.
Conclusions Comprehensive transportation models like the BTMS are often criticized for being good describers of the past but poor predictors of the future (Lewis et
110
(11, 1990). Such criticism misses the point. These models should be measured not only against the future they predict, but also against the predictions of competing methodologies. In that sense. comprehensive systems like the BTMS have an important role to play. They give us a consistent methodology on which to base analyses. Development agencies spend a good deal of time and money on partial annl~scs of individual projects which may not be consistent with changes elsewhere in the system. Using comprehensive systems built around a GIS lessens the likelihood of various interventions in the transport sector being mutually inconsistent or incompatible. The GIS aspects of systems such as the BTMS should also be mentioned. The ability systematically to collect. update and display network information is important in and of itself. Further. maps which show how regions interact when linked with a neu connector reinforce the idea that economic planning is more than just national accounts and sector shares. It is about interactions between people and regions. Marrying GIS and transportation analysis packages can only reinforce this often overlooked aspect of development planning.
Magnanti. T (19X4)‘Models and algorithm\ tar prcdlctlng urban traffic cquillbria’. in Florian. M. (cd. ) Trurwporttrtwrr Pltrnrtttlg Models. New York North Holland, p, 153-l X5 Owen, W. ( I YX7) Trumporttrtwrt und World I)~,~~~~If~I)t~l~‘~lt. Baltimore Johns Hopkins Univcrwty Press Rendel Palmer and Trltton (IYXY) ‘The Jamuna Bridge ProJcct, fcaaihllitv report to the Government ot Phase 11 Study’. Bangladesh. the United Natlow Development Programmc. and the World Bank Van Vursn. T. and Janscn. G. R. M. (14Xx) ‘Recent dcvclop;L rcbicw‘. Trrrrlcpr)rttrtron mcntv in path findlng algorithm\ Plunrm~ md Tuholog~, I?. pp S7-71 Wdson. A.G. ( lY70) Etrtropy u! Urhutr crtttl ReC~wrd hl~dcllrq. London: Pion Press