10. Assessing the effect of a new Light Rail Transit line on transport accessibility in the Tel-Aviv Metropolitan Area

This section presents the use of the Accessibility Calculator for assessing the effects of the new “Red” LRT line in the Tel Aviv Metropolitan area (TAMA). The line started functioning at the end of 2023 and became fully operational in 2024. The TAMA area is about 1500 sq. km, and it includes a dozen cities and many minor settlements. The population of TAMA in 2024 is 4.2 million and the number of buildings there is about 252,000. Figure 1 below presents the map of TAMA and the LRT Red line.

a
b

Figure 1. TAMA roads (a); Zoom to the area served by the Red LRT line (b)

The examples of this section are all computed with the Accessibility Plugin v 5_32 and QGIS v3.34. We use a Lenovo ThinkPad X1 laptop with the Intel i7 2.80GHz processor and 32MB memory and accompany each example with the estimates of computing time.

10.1. Let us arrange the data

To study TAMA transport accessibility with and without the Red LRT line, we will use TAMA layers of buildings and roads and two GTFS datasets for Israel for the years 2018 and 2024. We assume that buildings and roads did not change much between 2018 and 2024 and used the 2024 data. These datasets are provided as zip files, see here. There is also essentially a smaller dataset for the Jaffa area of Tel Aviv (~2500 buildings) that we provide for the initial fast tests see here.

The major characteristics of the exploited datasets are presented in Table 1.

Database

Type

Number of features

Size (MB)

TAMA Buildings

Shape

252,364

147

TAMA Roads

Shape

301,230

120

Israel GTFS 2018

Dataset

757

Israel GTFS 2024

Dataset

1,150

Table 1. The characteristics of the TAMA layers of roads and buildings and of two Israeli GTFS datasets. The OSM layers of roads and buildings for TAMA are already topologically cleaned (see the next section 10.2) and you don’t have to repeat cleaning for it. Topological cleaning, for roads, includes splitting road links at the points of intersection, connecting links’ ends at junctions, and deleting duplicated links. For buildings, topological cleaning includes deleting holes in the buildings’ polygons and “flatting” the layer by cutting the overlapping parts of the buildings’ polygons. We do not recommend cutting parts of GTFS. This operation demands great care, while the increase in performance that you will obtain with the smaller GTFS database will not be significant. In this tutorial, we use the full Israeli GTFS datasets in all cases, no matter what area is covered by roads and buildings. The total computation time of the data preparation stage for the TAMA dataset is about an hour and a half. For the Jaffa dataset (with yet full GTFS datasets of the public transport for 2024) it will take about 15 mins. We recommend you perform the entire data preparation procedure with the Jaffa dataset and check the changes in the road and building layers before you proceed to your data. Last but not least – there are minor differences between the GTFS datasets in different countries. We tried our best and tested many yet not all of them. Let us know if you get an error message at the stage of the transit routing database construction.

10.2. Data preprocessing

To continue with the Accessibility Calculator, we must clean the layers of buildings and roads, and construct layers for visualization. These processes are described in sections 1 and 2 of the Data Preprocessing and Constructing Databases for Fast Routing section of this tutorial. Before cleaning data and constructing databases, we recommend establishing a structure of folders to store the layers and databases. In this tutorial, the initial layers of roads and buildings are in the TAMA_tutorial/Source_layers folder. The cleaned layers or roads and buildings are stored in the TAMA_tutorial/Roads and TAMA_tutorial/Buildings and TAMA_tutorial/Visualization is used for visualization layers. Two GTFS datasets are initially in the TAMA_tutorial/gtfs2018 and TAMA_tutorial/gtfs2024 folders and we store the transit routing database in the TAMA_tutorial/gtfs2018DB and TAMA_tutorial/gtfs2024DB folders, respectively, and car routing database in the TAMA_tutorial/cars folder. Let us reproduce the steps of the data preparation procedure:

  • Click the Clean Road network menu item and choose the layer of roads (Figure 2). This layer can be chosen directly, but we recommend opening it in the current QGIS project before working with the Accessibility Calculator, to confirm that this is the layer you need.

  • Define the folder for the clean road network.

  • Road layer cleaning takes time, and for TAMA road layer with its ~300K links it took about 20 minutes.

Figure 2. Clean road network dialog

  • Click the Clean layer of buildings menu item and choose the layer of buildings (Figure 3). This layer can be chosen directly, but we recommend opening it in the current QGIS project before starting the work with the Accessibility Calculator, to confirm that this is the layer you need.

  • Define the folder for the clean layer of buildings, which can be the same as the folder for the clean layer of roads, while we suggest using a special folder for each of the datasets you use in your study.

  • Cleaning building layer cleaning is much faster than cleaning roads and cleaning ~250K TAMA buildings took about 5 minutes.

Figure 3. Clean layer of buildings dialog

  • Click the Build visualization layers menu item and choose the layer of buildings (Figure 4).

  • Define the folder for the visualization layers.

  • Building visualization layers takes 2-3 minutes.

Figure 4. Build visualization layers dialog

Summing up, for the areas of TAMA size, data preprocessing will take between half an hour and an hour. Most of the time will be spent on roads and buildings cleaning. Formally, you can work with non-clean layers of roads and buildings. However, cleaning is strongly recommended to ensure that the input data for navigation algorithms are correct.

10.3. Constructing databases

The next step is to use all four datasets – two clean layers of roads and buildings and two GTFS datasets, for constructing three databases – two for transit routing in 2018 and 2024 and one for car routing. This construction is described in sections 1 and 2 of the Data Preprocessing and Constructing Databases for Fast Routing section of this tutorial. Let us start with the transit routing database.

  • Click the Transit routing database menu item and choose the layer of roads and buildings (Figure 5). Be careful with the choice of field that represents the building ID. Establish a new folder to store the transit database.

  • Transit routing database construction takes about 15 minutes.

Figure 5. Transit routing database dialog

The log file (Figure 6) preserves all necessary data on the GTFS database construction GTFS and is stored in the database folder.

Figure 6. Log file of the Transit routing database construction

Repeat this step with the GTFS datasets of 2018 and 2024. As we already mentioned, we do not recommend cutting parts of the GTFS – it demands very careful querying of the dataset, while the gain in the database size will not be accompanied by the gain in performance. The last step in preparing data for accessibility computations is to construct a database for car routing. Before you do that, check the tables of the average car speeds by link types and congestion delay index, and edit, if necessary, the default values. If you are interested in comparing accessibility for different car speed or congestion levels during the day, build a special database for each set.

  • Click the Car routing database menu item and choose the layer of roads and buildings (Figure 7). Be careful with the choice of fields that represent the link’s speed, type, traffic direction, and building ID. Establish a new folder to store the car routing database before providing its name in the dialog box.

Figure 7. Car routing database dialog

The log file describes all data used for the TAMA car routing database construction and is stored in the database library (Figure 8).

Figure 8. Log file of the Car routing database construction

As can be seen, the time of the TAMA CAR database construction took 2 mins 11 sec. Table 2 presents characteristics of all three constructed databases. Importantly, the size of the databases is twice as small as that of the source.

Dataset

Construction time (mins)

Source files total size (MB)

Dataset size (MB)

CAR

2:11

267

194

PT2018

16:43

1,125

430

PT2024

26:21

1,417

595

Table 2. The characteristics of three constructed TAMA routing databases.

10.4. Transit accessibility, service area maps

We illustrate service area computations studying the accessibility of the Gesher (Bridge) theater in the Yafo region of Tel Aviv.

10.4.1. From/To-accessibility, fixed-time arrival/departure

Let us estimate Gesher’s transit accessibility for the visitors at 20:00 when the performance starts, and the visitors’ ability to get back home at 22:30, when it ends. In formal terms, we consider one facility and assess facility’s to-accessibility at 20:00 and facility’s from-accessibility at 22:30. Accessibility computations, in all possible regimes, demand the definition of parameters that define travelers’ behavior. Below, for the PT trips, we assume that:

  • Minimum number of transfers = 0

  • Maximum number of transfers = 1

  • Maximum walking distance from the origin building to the first PT stop = 400 m

  • Maximum distance between stops when changing lines = 200 m

  • Maximum walk distance from the last PT stop to the destination building = 400 m

  • Walking speed = 3.0 km/h

  • Maximum waiting time at the first PT stop = 10 min

  • Maximum waiting time at the transfer stop = 5 min

  • Boarding time gap = 15 sec

  • Maximum travel time = 45 min

Additional parameters for the service area computations are the arrival and departure times, and as defined, we use 20:00 for the backward, and 22:30 for the forward accessibility. The above parameters are part of the user dialog (Figure 9a). Note that we use the network and not aerial distance in all computations below (the air distance checkbox is disabled) and maximal trip duration is set to 45 minutes. As can be seen in Figure 9a, the to-accessibility is computed for the transit network of 2018. On Run, the folder of results will be created and, after the computations finish in 9 seconds (Figure 9b) this folder will contain two files: The log file log_BPTGesher.txt, and the file of results BPTGesher_45m_tot_265984731.csv. The 265984731 in the name of the result file is an OSM_ID of the Gesher Theater building. At the end of computation, the CSV file of results is joined to the visualization layer and presented as a map.

a
b

Figure 9. The dialog of the Transit accessibility map → Service area maps → To service locations – fixed time arrival (a) and the Log file of the computations (b)

The maps of accessibility to/from the Gesher theater, two before and two after the Red LRT line was introduced are presented in Figure 10. It took 2-3 seconds per scenario to compute each. You can notice that the areas accessible with up to 45 minutes’ trip after the Red LRT line was introduced are larger than the areas that were accessible during the same time before. We will compare accessibility in 2018 and 2024 and assess the Red line effect quantitatively in the Compare Accessibility section below.

a

b

c

d

Figure 10. The results of the Transit accessibility → Service area maps computations of the Gesher Theater in Yafo. To service locations – Fixed-time arrival at 20:00, in 2018 (a) and 2024 (b). From service locations – Fixed-time departure at 22:30, in 2018 (c) and 2024 (d)

10.4.2. From/To-accessibility, schedule-based arrival/departure time

Schedule-dependent accessibility considers travelers who know the transit schedule and start their trip only when the bus they have chosen arriving at the departure or arrival stop. Like a traveler who plans to go shopping between 10-10:30 in the morning or wants to get to a fish market that opens at 8:00, during the first half an hour of the market work. The schedule-based travel is optimized within the interval of the traveler’s flexibility and this flexibility is an additional parameter that defines the trip start or arrival time. Formally, in the case of from-accessibility, “Start time” is substituted by “The earliest start time,” and the “Maximum delay (in minutes)” is a parameter that defines the start time flexibility (Figure 11).

Figure 11. The part of the From service location – Schedule-based departure dialog that differs from Fixed-time departure dialog

In the case of to-accessibility, “The arrival time” is substituted by “The earliest arrival time” while the maximum lateness (in minutes) is a parameter that defines the arrival time flexibility.

Figure 12. The part of the To service location – Schedule-based arrival dialog that differs from Fixed-time departure dialog

The travel time in case of schedule-based accessibility does not include waiting at the first stop and that is why the schedule-based accessibility is always higher than the fixed-time one and is less sensitive to the chosen start or arrival time that can slide within the intervals of flexibility. As an example of the schedule-based accessibility calculations of the Gesher Theater, we consider the following outline: There is a photo exhibition in the theater foyer, and many visitors are ready to arrive at the theater any moment between 19:30 and 20:00, to see the exhibition before the performance. They also keep in mind that the theater café serves drinks and snacks long after the performance and it’s worth having a cup of tea after the performance is over and wait for the empty bus with the guaranteed sit. Four maps of schedule-based accessibility of the Gesher Theater in the year 2018 and in 2024, when the Red LRT line became fully functional, took each 2 – 3 seconds to compute. They are presented in Figure 13. As for the fixed-time route choice, the areas accessible with up to a 45-minute trip after the Red LRT line in 2024 are larger than the areas for the same maximal trip time in 2018. We will quantitatively assess the Red line effect for the users that plan their trips based on the transit schedule in the Compare Accessibility section below.

a

b

c

d

Figure 13. The results of the Transit accessibility → Service area maps computations of the Gesher Theater in Yafo. To service locations – Schedule-based arrival at 20:00, in 2018 (a) and 2024 (b). From service locations – Schedule-based departure at 22:30, in 2018 (c) and 2024 (d)

10.4.3. Single location CAR accessibility

CAR accessibility computations demand fewer parameters than transit-based calculations and the schedule-dependent accessibility is irrelevant here. However, assessment of the car travel time demands knowledge of the traffic speed along the route and this information is hardly available. The only source of systematic knowledge of the traffic speed we are aware of is Google API and we plan to relate car accessibility calculations to the Google data on traffic speed in the next version of the Accessibility Calculator. For now, to calculate CAR accessibility, we assume that the average speed on the road link is defined by the link’s type and the level of congestion in the hour of travel. The table of the characteristic speeds for the OSM classification of links is supplied with the plugin. The name of the table is Car_speed_by_link_type.csv and it can be edited by the user. See more details on this table in this section. As an example of a car service area map, we calculate car accessibility of the Gesher Theater in Yaffo: To-accessibility, to the performance that starts at 20:00, and from-accessibility at 22:30, when the performance ends (Figure 14).

a

b

Figure 14. Car accessibility → Service area maps computations dialogs. In the to-accessibility dialog (a), Gesher is a destination and TAMA buildings are origins; in the from-accessibility dialog (b), it's vice versa, Gesher is an origin and TAMA buildings are destinations

As should be expected, the maps of Gesher’s to- and from-accessibility (Figure 15) look much simpler than those of the PT accessibility. It’s worth noting, however, that the car accessibility from the theater at 22:30, when the congestion is over, is essentially higher than to-accessibility at 20:00 when the congestion is still there. Overall, car accessibility at 20:00 and 22:30 is essentially higher than PT accessibility for the same hours.

a

b

Figure 15. Car accessibility to the Gesher Theater at 20:00 (a), when the congestion is still there, and from the Theater at 22:30, without congestion (b)

10.5. Accessibility of every location in a region

Service area maps present the accessibility of one or several facilities. However, the infrastructure changes affect many locations at once. The Region part of the Accessibility Calculator assesses the effects of these changes on all locations in the region. A building remains the basic unit of the Region calculations, with accessibility calculated for every building. Calculations of regional accessibility for the region with several thousand buildings cannot end with several thousand accessibility maps computed for each building. That is why, to assess the accessibility of the region we employ aggregate measures. An example of the aggregate measure is the number of buildings that may be accessed given a maximum trip time. In the case of the from-accessibility, we compute the number of buildings that can be accessed from each building in the region. In the case of the to-accessibility, we compute the number of buildings from which each building in the region can be accessed. The aggregate accessibility measures are calculated at a user-defined time resolution, typically of 5 minutes – number of buildings accessible in maximum 5 minutes, in maximum 10 minutes, etc., up to the maximum trip time. The number of accessible buildings is a default measure of the region’s accessibility and is always calculated. The user can define more measures of this kind, like the number of accessible buildings of a certain type, the number of residents in accessible buildings, or the number of jobs there. Any characteristic that can be calculated based on the buildings’ attributes can be chosen. Let us continue the assessment of the Red LRT line’s effect on transport accessibility choosing the entire city of Tel Aviv as a region. The number of buildings in Tel Aviv is about 40K and, different from the calculation of a service area of several facilities, the computing time may be several hours. In the example below we employ the default number of accessible buildings only.

10.5.1. Region transit accessibility, fixed-time arrival/departure

Figure 16 presents the parts of Region maps dialogs that differ from the dialogs of the Service area maps. These are From every location – Fixed-time departure (Figure 16a) and Region maps → To every location – Fixed-time arrival (Figure 16b). In the case of from-accessibility, the buildings of a region are origins, and we must set the layer of destinations. In the case of the to-accessibility, the region’s buildings are destinations, and we must set the layer of origin buildings to start at. In both cases, the result will contain one record for each building – the number of destination buildings that can be reached from a current building or the number of origin buildings from which the current building can be reached with the public transport in 5, 10, etc., minutes.

a

b

Figure 16. The Transit accessibility map → Region maps dialogs for the From every location – Fixed time departure (a), and To every location – Fixed time arrival (b) cases

Computing regional accessibility demands choosing the attributes for aggregation (Figure 17). The user can aggregate any of the building’s attributes and the sum of each of these attributes over accessible buildings, by the time bins will be stored as a result. If you are interested in computing the weighted sum of some attribute, calculate this weighted attribute for each building and then sum it up with the Accessibility Calculator.

Figure 17. The choice of the number of bins and attributes to aggregate in the Transit accessibility → Region maps → From/To all locations - Fixed-time accessibility dialog

To remind, if the maximum travel time does not contain an integer number of bins, the results are also stored for the maximum travel time. Figure 18 presents maps of region accessibility in 45 minutes for Tel Aviv city in the year 2018, before the red LTR line was established, and in 2024 when this line was in full operation. To speed up computations, these maps were computed based on hexagons of 231m diameter (the diameter of the hexagons in the coverage that is obtained by constructing “200m hexagons” with the Create Grid QGIS command that is employed by the Accessibility Calculator). The number of the hexagons of this size that cover all buildings in Tel Aviv is 1750 and computations took about 30 mins, close to 1sec/origin. We will compare these two maps numerically in the next section.

a

b

Figure 18. The maps of the Tel Aviv city region accessibility in 45 minutes were computed based on the 200m hexagon in the year 2018 before the Red LTR line was established (a) and in 2024 when the Red LRT line was in full operation (b)

10.5.2. Region transit accessibility, schedule-dependent arrival/departure time

Calculation of the transit region accessibility according to the schedule-dependent view repeats the fixed-time approach in all other respects.

10.5.3. CAR accessibility

There are no any special details to mention about calculation of the car region accessibility.

10.6. Compare accessibility maps

The overall goal of our exemplary study is to assess the effects of the Red LRT line. Now, when the accessibility maps are constructed, we can compare accessibility maps before and after the line was established. To remind, the Accessibility Calculator provides three measures of difference:

  • Ratio: Result_1/Result_2: The ratio of the result of the first scenario to the results of the second scenario, for the overlapping part of the outputs.

  • Difference: Result_1 - Result_2: The difference between the result of the first scenario and the results of the second scenario, for the overlapping part of the outputs.

  • Relative difference: [Result_1 - Result_2]/Result 2: The difference between the result of the first scenario and the results of the second scenario, for the overlapping part of the outputs. The result is presented in percents.

For each of the three measures, in addition to the map of the measure, two more maps are presented. The first one presents the buildings that are accessible in Scenario 1 but not accessible in Scenario 2 (Result_1 is not NULL, while Result_2 is NULL). The second map presents the buildings that are in Scenario 2 but not accessible in Scenario 1 (Result_2 is not NULL, while Result_1 is NULL). In this tutorial, we employ the difference between two maps.

10.6.1. Comparing service area maps

Our first question is “Whether the Red Line increased the accessibility for Gesher visitors who arrive by public transport?” To answer, we compare schedule-based maps of to-accessibility to Gesher at 20:00 for the years 2024 and 2018, calculating the travel time difference:

Result_2024 – Result_2018

Figure 19 presents the difference between two to-accessibility maps, and the “only” parts. Note, that two “only” maps can be always combined into one.

a
b

Figure 19. Comparison of transit to-accessibility maps for Gesher theater visitors at 20:00 in 2018 and 2024. The travel time difference (a); Areas accessible in 2018 or 2024 only (b)

As can be seen, overall, the Red Line essentially improved transit accessibility for the visitors who get to Gesher’s performance. The green shades of the map in Figure 16a denote buildings from which the transit travel time in 2024 is lower than in 2018, and these areas cover 60% of the overlapping areas. For 20% of buildings in the overlapping areas, the travel time is almost the same, while for 20% transit travel to the Gesher Theater will take more time in 2024 than in 2018 (Figure 16a, brown shades). The Red LRT line makes the Gesher Theater accessible in not more than 45 minutes from the vast areas it was not accessible in 2018. More comparison studies will help us to understand the reasons for the differences revealed. One can compare the to-accessibility maps for the longer than 45 minutes maximum travel time or go deeper and, based on the full output of the service area computations with the full description of the trip (see Service area computations - the log file and the structure of the report), investigate how the travelers get to the Gesher Theater from each of the “only” parts in 2018 and 2024.

10.6.2. Compare fixed-time and schedule-based accessibility

To assess the effect of the schedule-based view of accessibility let us compare the schedule-based and fixed-time 2024 to-accessibility maps for the Gesher visitors, for the performance starting at 20:00. Figure 20 presents maps of the travel time difference:

Schedule_based_TO_GESHER_24 – Fixed_time_TO_GESHER_24

As can be seen, the schedule-based accessibility is always higher than the fixed-time based.

a
b

Figure 20. Comparison of the schedule-based and fixed-time accessibility maps for the Gesher Theater in 2024. To-accessibility at 20:00 (a); From-accessibility at 22:30 (b)

10.6.3. Compare Region accessibility

Comparison of the Region accessibility maps goes in the same way as the comparison of the service areas maps. Figure 21 presents this comparison of two from-accessibility maps for 2024 and 2018, computed for the fixed-time trip starting at 08:00 in the morning (Figure 18). As we have already seen in Figure 18, the Red LRT line essentially increased accessibility for most of the Tel Aviv locations – the greenish hexagons cover about 85% of the Tel Aviv area. Yet the number of buildings achievable from some locations that are far from the Red LRT line is lower in 2024 than in 2018 due to the changes in the bus network that follow the introduction of the Red LRT line. It is worth noting that in the case of service area maps when the travel times are compared, the range of difference is limited by the maximum travel time in each of the compared scenarios. In the case of the Region accessibility, the range differences can be very high. The aggregate measure of accessibility of a certain building can be very low in one scenario since there is no PT line at a walkable distance from it and all accessible buildings are reached by foot only, while in the other scenario transit lines get closer, and the number of accessible buildings becomes much higher.

Figure 21. The differences in region from-accessibility with the public transport in 2024 and 2018, for the trips starting at 08:00 in the morning