Mandate

To assess the efficiency of the transportation network in the West Island, with a focus on the rail mode. 

Objectives

1. To assess potential ridership at train stations of the Montréal/Dorion-Rigaud line in the West Island.

2. To compare the potential ridership with the actual ridership to show the relative importance of each station. 

Relevance

1. To identify stations with heavy usage and propose investment plans to expand the services, such as increasing the number of parking, improving kiss-and-ride facility and increasing train frequency.

2. To identify underused stations and carry out appropriate actions, such as improving accessibility to the station, increasing the number of parking, redirecting bus routes to the station or even eliminate the station.

Data requirements

1. Street network

It is modified from a street network file available from the Statistics Canada. It is dated 1993. It contains most of the streets in the West Island and the basic attributes such as street name. The quality of the data is not perfect. A small part of the network has digitizing errors and has to be modified. On the other hand, some data has to be added in order to do the network analysis. This will be discussed in the Methodology section.

2. Train stations

It is available as a point theme from the Agence métropolitaine de transport (AMT). It is dated 1999 and contains all the stations in the West Island. It is used as a reference for the network analysis. Detailed information about each station, such as the number of parking, the schedule of trains and the number of corresponding bus routes, is available from the AMT website.

3. Actual ridership figures at different stations

Data from an on-board survey conducted by the AMT at morning peak period on September 12, 2000 are available through Mr. Paul Dorval, director of commuter train services. It can safely be assumed that the data are typical of a normal working day and that all peak hour buses were running.

4. Census data

Data are available through the Computing in the Humanities and Social Sciences (CHASS) website of University of Toronto (http://datacentre.chass.utoronto.ca/census/). Since the census data of 2001 are not available at the time of the analysis. Data from 1996 are being used. Three possible data categories can be used: total population, total population 15 years and over, total employed labour force 15 years and over. Since the data of total population come from a 100% sample, they are the most complete and accurate. It is used in this analysis. It should be noted that it is relatively easy to change the data categories. However, results are expected to be similar.

5. Geographic files of the census

It is available from Statistics Canada for the year 1996. It is essential to combine these spatial data with the census data. Enumeration area is chosen since it is the smallest scale available. It can minimize errors and uncertainties.

6. Other themes for display purposes

Railroad network (from Statistics Canada, 1993), base map and land use map (from Montréal Urban Community, 1996) of Montréal are required for display purposes and for understand the context. (See Map 1: Land Use in West Island)

Methodology

1. Data Projection

All data used in this analysis are projected to UTM Zone 18N.

2. Data preparation

Using the street network of Montréal as a start point, generalized impedance information of each street (can also be called link or arc) is added. Since actual field data are difficult to obtain, the impedance of each link is assumed according to the size and the traffic volume of that street. It is done with the help of a regular street atlas. Each segment of a residential street is assigned to a traffic speed of 30 km/h and the impedance (time) is calculated by dividing the length by the speed. 3 categories are assigned to the network: residential street (30 km/h), major arterials (50 km/h) and autoroute (100 km/h). In this analysis, the use of autoroute is not allowed. The assumption is that travelers are not likely to use an autoroute to get to a station. Turns and delays are generalized to a 30% decrease in speed. For example, a car only runs 35 km/h (50 km/h * 70%) in a major arterial. (See Map 2: West Island Street Network)

3. Analysis techniques

Once the network is setup, the analysis can be started.

1. Each station’s catchment area is found by using the Find Service Area function of the Network Analyst extension in ArcView 3.2. We assume that commuters only spend 10 minutes to go to the station including 1 minute of access time (time to get the automobile ready) and 1 minute of egress time (time to walk from the parking to the station). All segments which lie within 8 minutes of the station are selected. It is important to select a realistic time because it can greatly influence the result. Walking, cycling and bus routes are not necessary in this analysis since those modes are slower than automobile, the catchment area calculated with automobile speed will represent the maximum potential ridership, encompassing all population regardless their mode choice to go to the station. (See Map 3: Beaconsfield Train Station Service Area)
2. Since the catchment area of the stations overlaps, Thiessen polygon is used, an Avenue script available for free, to create polygons which represent areas closest to a particular station than to any other stations. This can avoid double counting of the potential ridership. The assumption is people only go to the nearest station, which is generally, but not always, true.
3. Numerous overlays, including clipping, intersecting and union with the Thiessen polygons and service areas, are done in order to get the actual catchment area for each station.
4. The census data are joined to the enumeration areas. The population of the enumeration areas range from 41 to 1729. Using another free Avenue script, their centroids are found. (See Map 4: Enumeration Area Population)
5. We then do a Select by theme to select all the centroids that lie within each catchment area. (See Map 5: Train Station Service Areas of West Island)
6. The total number of population served by each station is calculated.
7. Ratios between the actual ridership collected in the on-board survey and the potential ridership for each station are computed. We assume that the number of commuters traveling from Downtown toward West Island is negligible compared to from West Island to Downtown. So, we only use the number of commuters using the West Island stations as origin.
8. An index is derived using the average ratio as 100. Stations with an index above 100 have a high actual/potential ridership ratio while stations with an index below 100 have a low actual/potential ridership ratio. 

Results Highlights and Analysis

Total Potential Ridership: 129556

Actual Ridership: 5112

Average Ratio: 3.95%

Graph

1. Actual Ridership to Potential Ridership Ratio

1. High ratios tend to associate with high number of parking available. All train stations with above average indexes have equal to more than 115 parking available. All train stations with below average indexes have equal to less than 72 parking available.

2. High ratios tend to associate, to a less degree, with the number of bus routes reaching the stations. All train stations having no or only one corresponding bus route have below average indexes. All train stations having more than one corresponding bus routes have above average indexes, with the exception of Pine Beach.

3. Dorval’s unusually high ratio can be explained by the fact that there are travelers coming from the airport.

4. Sainte-Anne-de-Bellevue’s high ratio can be explained partly by the location of Macdonald Campus of McGill University and John Abbott College.

5. Train stations with below average indexes are sandwiched between train stations with above average indexes, suggesting ridership from those stations is stolen by adjacent stations.

6. There is no clear pattern showing train ridership varies with distance to downtown.

Limitations

1. Street network impedance does not necessarily represent the reality. However, it is almost certainly more accurate than using distance as impedance.

2. The assumptions used in the analysis are generally safe. However, if one of them proves to be seriously wrong, the results can be erroneous.

3. This is a “Competing service area” problem. There are overlaps with the catchment areas of the stations. Dividing using Thiessen Polygons might not be accurate enough since people may not go to the nearest station. Instead, they go to stations where they can reach in the least amount of time. Network Analyst extension cannot assign areas to the nearest stations in term of time. It can only assign point to the nearest stations.

4. A centroid is the centre of a polygon which includes not only residential areas, but also commercial, industrial and institutional areas. It is not population-weighted and therefore, can only provide an estimation of the potential train riders.

5. Although with many complex procedures, the analysis performed may not generate better results than using simple buffer operation around train stations.

6. Socio-economic data (from other census categories) and travel pattern (origin-destination survey) are not taken into account. The analysis assumes that all people living within the catchment area are potential train riders. However, the socio-economic profile of the residents of West Island should be relatively homogeneous since they are mostly English-speaking suburbanites.

7. Although converted to the same projection, data coming from different sources have slight inaccuracy when they are overlaid. However, influence on the results should not be significant.                                                                                    

Suggested Improvements on the Analysis

1. A turntable, which takes into account impedance at intersections, can be included in order to better model the street network. However, it is very time-consuming to set up and in this analysis, it might not change the result significantly. A sample network with turntable is setup around the station Baie d’Urfé. There are 359 arcs in total. The turntable generated has 2160 records. Each record should include an impedance value. According to Plumb, at a four-way stop intersection, the impedance for a left-turn or a right-turn is 12 seconds, going straight is 8 seconds and U-turn is not allowed. For intersection with traffic light, the impedance varies. In this analysis, I put 60 seconds for a left-turn, 30 seconds for a right turn, 45 seconds for going straight and U-turn is not allowed. There are also some prohibited turns to take into account, such as overpasses and underpasses, no left-turn intersection. All of these have to be entered. (See Map 6: Network with Turntable)

2. To perform the analysis again when the 2001 census data become available.

Conclusion

An analysis using GIS to assess the efficiency of the transportation network in West Island is done, with an emphasis on rail mode. Service areas are created using the Network Analyst extension by identifying places within a 10 minutes distance from each station. Using the centroids of the enumeration areas, population which lies within the service area are considered as potential train rider. This is then compared with the actual ridership data. The ratio ranges from 1.41% to 13.95%. This ratio is associated with the number of parking available and the number of corresponding buses. Some hypotheses are suggested to explain the results. There are several assumptions in the analysis that can influence the results. Results may be refined if more sophisticated procedures are used.

Bibliography

Agence métropolitaine de transport website. <http://www.amt.qc.ca>

Easa, Said and Yupo Chan. Ed. Urban Planning and Development Applications of GIS. Reston: American Society of Civil Engineers. 2000.

Miller, Harvey J. and Shih-Lung Shaw. Geographic Information Systems for Transportation: Principles and Applications. New York: Oxford     University Press. 2001.

Nielsen, Otto Anker. GIS-Based Method for Establishing the Data Foundation for Traffic Models. ESRI Library. <http://www.esri.com>

Plumb, Gregory. Preparing Networks for Routing Applications.  ESRI Library. <http://www.esri.com>

Appendix: Train Station Specifications

View Report in HTML
View Report in PDF format (Not available now due to limited storage space) Click here to download Adobe Acrobat Reader
View PowerPoint presentation (Not available now due to limited storage space) Click here to download PowerPoint Viewer

This web page was created by Ka Kee Alfred Chu, April 2002.
Email: Ka Kee Alfred Chu
Homepage: http://alfredchu.tripod.com