Travel Time Analysis with Uber Movement (QGIS3)

Avertissement

Uber Movement data has been discontinued [view announcement]. You can still follow the tutorial using the archive data provided in the tutorial and learn the technique of travel time analysis using QGIS.

Uber Movement partage des données de temps de trajet anonymisées et agrégées pour de nombreuses villes à travers le monde. Le produit Travel Times d’Uber Mouvement est un ensemble de données publiques mesurant les déplacements d’une zone à une autre dans une ville. Ces heures sont basées sur des trajets Uber réels et sont une représentation précise des embouteillages et des schémas de circulation dans la ville. Il s’agit d’un vaste ensemble de données librement accessible agrégé à partir de millions de trajets en taxi réels. Ce didacticiel montre les techniques permettant d’utiliser de tels ensembles de données de trafic agrégés pour effectuer une analyse du temps de trajet dans QGIS.

Aperçu du travail à faire

Nous utiliserons un ensemble de données de trafic agrégées pour la ville de Bangalore, en Inde, pour trouver les temps de trajet jusqu’à un endroit choisi dans la ville. Nous allons également créer une « carte isochrone <https://en.wikipedia.org/wiki/Isochrone_map>`_ pour un seuil de temps de trajet de 30 min.

Autres compétences que vous vous allez étudie

  • Load a GeoJSON file in QGIS

Get the data

Nous allons télécharger l’ensemble de données “Travel Times” pour la ville de Bangalore, en Inde.

  1. Go to the Uber Movement site and click on the Cities.

../../_images/data013.png

  1. Search for Banglore.

../../_images/data023.png

  1. Click on Download data.

../../_images/data032.png

  1. In the Download data popup, switch to GEO BOUNDARIES. It will contain the Banglore wards GeoJSON file. Accept the license information and click on BANGLORE_WARDS.JSON to download.

../../_images/data042.png

  1. Then switch to ALL DATA. This data is available from 2016 to 2020, and each year is divided into 4 quarters. We will download the data for 2019 Quarter 3. Select it and click the Travel Times by Hour of Day (Weekdays Only). This file will contain all anonymous information about the weekday uber travel in Bangalore from July to September 2019.

../../_images/data051.png

For your convenience, you can download the data from the links below.

bangalore_wards.json

bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate.csv

Data Source: [UBER]

Procedure

  1. Locate the bangalore_wards.json file in the Browser panel and drag it to the canvas. Next, we will load a basemap layer from OpenStreetMap. We will use the QuickMapServices plugin to access the basemap. Once installed, go to Web ‣ QuickMapServices ‣ OSM ‣ OSM Standard. A basemap tile layer from OpenStreetMap will be added to the project. Next, click the Open Data Source Manager button.

../../_images/019.png

  1. Switch to the Delimited Text tab. Browse to the bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate.csv file and select it. Since this CSV file is just tabular data, select No geometry (attribute only table) option and click Add.

../../_images/029.png

  1. The bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate layer will be added to the Layers panel. This layer contains anonymized and aggregated trip location data between different zones of the city. Each row of the table contains columns for source zone (sourceid), destination zone (dstid), hour of the day (hod) and average travel time aggregated from all trips between these zone at that hour (mean_travel_time). You can learn more about this dataset in the Movement: Travel Times Calculation Methodology (pdf). Before moving forward, let’s check how many data records are present in the layer. Right-click the bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate layer and select Show Feature Count. The total rows from the table will be displayed next to it. This is a fairly large table but we don’t need all the data rows for our analysis. We will now identify our target location and filter this table to data records for it.

../../_images/039.png

  1. We want to calculate all areas that are accessible within 30 minutes from a specific location. Using the OSM Standard basemap, you can find the location of interest. Then select the bangalore_wards layer, use the Identify tool and click on it. The results will show the attributes of the zone containing the location. For the purpose of this tutorial, let’s assume our target location is within the JP Nagar zone with the MOVEMENT_ID 193.

../../_images/049.png

  1. We can filter the travel time records to just those which have this zone as the destination. We can also restrict our analysis to the peak morning commute hour of 9am - 10am. Right-click the bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate layer and select Filter.

../../_images/059.png

  1. Enter the following filter expression and click OK.

"dstid" = 193 AND "hod" = 9
../../_images/069.png

  1. Back in the main QGIS window, you will see that the number of records in the filtered table are now down to just 197. Since there are a total of 198 zones in the city, we have records of travel times between the 1 destination ward and 197 source zones. Open the attribute table of both the layers using the Open Attribute Table button in the Attributes toolbar.

../../_images/079.png

  1. Now we have the shapes of the ward in the layer bangalore_wards and tabular information in the layer bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate. We can join the attribute information to the shapes using a common attribute. Here the MOVEMENT_ID column from the bangalore_wards layer and sourceid column from the bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate are unique ward identifiers that can be joined. This operation is called a Table Join.

../../_images/088.png

  1. Before we can join these two layers, we must ensure that the values in both columns match exactly. Thought they appear the same, they are of different type. Since GeoJSON format has no way of specifying property types, all values are assumed to be of the type String - i.e. Text. But when we import a CSV to QGIS, by default, QGIS tries to determines the types of the columns based on the values and assign appropriate field type. For the CSV file, the data type for the column sourceid was assigned as Integer. So we need to convert the column from the GeoJSON to an Integer type as well. Go to Processing ‣ Toolbox ‣ Vector Table ‣ Field Calculator algorithm. Double-click to launch it.

../../_images/098.png

  1. Choose bangalore_wards as the Input Layer. Name the Field Name as joinfield and select the Result field type as Integer. Enter MOVEMENT_ID as the Expression. Click the … button next to the Calculated and select Save to File… then enter the name of the output file as bangalore_wards_fixed.gpkg. Click Run. Close the field calculator.

../../_images/1034.png

  1. A new layer bangalore_wards_fixed will be added to the Layers panel. Now we are ready to perform the join. Go to Processing ‣ Toolbox ‣ Vector General ‣ Join attributes by field value. Double-click to launch it.

../../_images/1143.png

  1. Select bangalore_wards_fixed as the Input layer and joinfield as the Table field. Select bangalore-wards-2019-3-OnlyWeekdays-HourlyAggregate as the Input layer 2 and sourceid as the Table field 2. Name the Joined layer as uber_travel_times.gpkg and click Run.

../../_images/1237.png

  1. A new layer uber_travel_times will be added to the Layers panel. Let’s style it to visualize the result of the join. Click Open the Layer Styling Panel. Select the Graduated renderer and mean_travel_time as the Value. Select a color ramp and click Classify. You will see the map showing increasing travel times further you go from the destination.

../../_images/1335.png

  1. But we are looking to analyze and extract areas that are within 30 minutes of travel time, so we need to do some more processing. Switch the styling back to the Single symbol renderer. Right-click the uber_travel_times layer and select Filter.

../../_images/1432.png

  1. Enter the following expression to select all zones within 1800 seconds (30 minutes) of mean travel time. We also need to include our destination zone which will have 0 travel time.

"mean_travel_time" <= 1800 OR "MOVEMENT_ID" = 193
../../_images/1529.png

  1. The layer will now show the cluster of polygons that constitute the area of interest. We will merge all of them to a single polygon now. Go to Processing ‣ Toolbox ‣ Vector Geometry ‣ Dissolve. Double-click to launch it.

../../_images/1627.png

  1. Select uber_travel_times as the Input layer. Name the Dissolved layer as 30min_isochrone.gpkg. Click Run.

../../_images/1727.png

  1. A new layer 30min_isochrone will be added to the Layers panel showing the result of our analysis.

../../_images/1824.png

If you want to report any issues with this tutorial, please comment below. (requires GitHub account)