UrbanAccess: GTFS and OSM network accessibility analysis

[blanchard]

This thread is to introduce a new library that will soon be released in the Urban Data Science Toolkit, called UrbanAccess. UrbanAccess is tool for creating multi-modal graph networks for use in multi-scale (e.g. address level to the metropolitan level) transit accessibility analyses with the network analysis tool Pandana. UrbanAccess uses open data from General Transit Feed Specification (GTFS) data to represent disparate operational schedule transit networks and pedestrian OpenStreetMap (OSM) data to represent the pedestrian network. UrbanAccess provides a generalized, computationally efficient, and unified accessibility calculation framework by linking tools for: 1) network data acquisition, validation, and processing; 2) computing an integrated pedestrian and transit weighted network graph; and 3) network analysis using Pandana.

UrbanAccess offers the following tools:

• GTFS and OSM network data acquisition via APIs
• Network data validation and regional network aggregation
• Compute network impedance:
  • by transit schedule day of the week and time of day
  • by transit mode
  • by including average passenger headways to approximate passenger transit stop wait time
• Integrate pedestrian and transit networks to approximate pedestrian scale accessibility
• Compute cumulative accessibility metrics using the open source network analysis tool Pandana

UrbanAccess was developed by Sam Blanchard while completing his graduate degree in City and Regional Planning at UC Berkeley. UrbanAccess is currently being prepared for a public release expected by late February 2017. UrbanAccess will be integrated within the UrbanSim Cloud Platform at a later date. We will provide updates here, so stay tuned. If you have questions about this project, this is the right place to ask them.

[stefancoe]

Thanks for putting this together Sam!

When doing network queries, there will be paths that only include walk links, correct? This seems fine for short distances, but if I do a query for the number of jobs within 45 minutes, will the results include jobs that are accessible by a 45 minute walk on the network (paths that never traverse a transit edge)? Thanks!

-Stefan

[blanchard]

Hi Stefan!

Thank you and I am glad you are finding helpful!

To answer your question:

Yes that is correct the full integrated network will include all transit edges, all transit to pedestrian (and visa versa) edges and all pedestrian edges within your bounding box meaning that yes in your aggregation query you will get results that include travel on pedestrian only edges that may not be within a walkable distance to transit. Further processing would need to be done to remove walking edges that are not within a certain distance - this could be done with buffers around transit nodes. A function to do this inside of UrbanAccess would be a nice addition. If this is something of interest, any ideas or contributions are welcome.

[stefancoe]

I think buffers would take care of most of the issue. There might be some cases where parallel corridors have overlapping buffers that would create walk paths connecting nodes that are not accessible by transit links (because of transfers, you would have to go beyond the max time/distance on the transit network to get to a place that is accessible on the walk network, yet beyond the buffer distance). But I think I am largely talking about edge cases at this point.

As for built in solutions. I agree that eliminating walk edges that are beyond a certain distance/time of a transit node would be a nice feature. Another potential, yet quite involved, solution would be to add pandana functionality that would allow the user to specify distance/time limits on certain types of edges. But this seems like a lot of development time so its probably not reasonable at this point.

-Stefan

[blanchard]

Agreed, these are good ideas.

Your suggested feature for Pandana would be a good idea for a future version. Feel free to describe it in an issue on the Pandana GitHub repo.

[proma12]

Hello,
I'm a beginner using Urban Access, I'm trying to follow https://udst.github.io/urbanaccess/data_loaders.html instructions but it gives me different errors.

import urbanaccess
urbanaccess.osm.load.ua_network_from_bbox(lat_min=52.2312, lng_min=12.5807, lat_max=52.8019, lng_max=14.1957, bbox=14.1957,52.2312,12.5807,52.8019, network_type='public_transport', timeout=180, memory=None, max_query_area_size=2500000000L, remove_lcn=True)

I used that, but it says that there is a "SyntaxError: non-keyword arg after keyword arg". Could you help me?

Thank you so much,

Paula

[proma12]

I have another issue, while trying to compute the headways, it says that routes, stops and stop_times are empty, but it's not true, what I should do?

Thanks,

Paula

[proma12]

Hi again, I have a second problem, I just left the headways and I tried to Loading Processed GTFS data, it gives me that error: ValueError: Unable to find key: stop_times_int. Keys found: ['/calendar', '/calendar_dates', '/routes', '/stop_times', '/stops', '/trips']

This time I checked and this file doesn't exist, so how can I downloaded?

Thank you,

Paula

[blanchard]

Hi Paula,

We have just created a demo for UrbanAccess that shows you an entire workflow in a jupyter notebook that can be accessed here. Before running the notebook make sure you are running the latest version of urbanaccess as changes have been made as of today, use git pull from the cloned directory or re-clone the library. If you go through the demo you should have a better understanding of the API and the workflow.

Once you complete the demo (restart the notebook to clear memory), you can substitute the code specific to the demo in the notebook for your own code and try your process again. Let us know if after you have done this you have any issues.

In direct response to your messages though:

From the messages and code you provided it looks like the parameters you are passing to ua_network_from_bbox are incorrect. The function docstrings and documentation should provide the necessary information to run the function correctly. For example, you should only specify the four lat and long parameters or the bbox parameter, not both. If using the bbox it needs to be a tuple eg: (14.1957,52.2312,12.5807,52.8019) and network_type must be either walk or drive. For the headway function to work you need to have loaded GTFS data with the gtfsfeed_to_df function. To use the gtfsfeed_to_df function you must specify the location of your GTFS feeds on disk in the gtfsfeed_path parameter or use the global variable for this.

If you run through the demo though this should clear up the confusion.

[proma12]

Hi blanchard,

thank you for your help.

I'm following the demo and I have this problem: AttributeError: 'DataFrame' object has no attribute 'sort'
when I'm doing that part:

Add average headways to network travel time
Calculate route stop level headways
The network we have generated so far only contains pure travel times. UrbanAccess allows for the calculation of and addition of route stop level average headways to the network. This is used as a proxy for passenger wait times at stops and stations. The route stop level average headway are added to the pedestrian to transit connector edges.
Let's calculate headways for the same AM Peak time period. Statistics on route stop level headways will be added to your GTFS transit data object inside of headways
In [ ]:

ua.gtfs.headways.headways(gtfsfeeds_df=loaded_feeds,
headway_timerange=['07:00:00','10:00:00'])
In [ ]:
loaded_feeds.headways.head()

Until now I didn't have any problem.

Thanks!

[blanchard]

Hi Paula,

It looks like you are not running the latest version of UrbanAccess. That issue was due to the library using a depreciated Pandas sort function that is now sort_values. This has been fixed in the latest version. If you used git to clone the repo, go to your cloned repo directory and in a terminal do git log -1 and note the date and number of the last commit you have it should be this:

commit 9a0c287651e903f7c004997fc09b68da92239f71
Date: Mon Jun 19 22:53:12 2017 -0700

added plots of nearest poi examples

If not then use git pull to get the latest version. I am assuming you installed using develop so you should be set to go. If you didnt, first uninstall your current version then reinstall with the new version.

[proma12]

Thank you so much Blanchard, problem solved it.

[elai]

I am trying to get the latest version of Urban Access via conda, but seems like the current version is 0.3.0. Is it possible to get 0.4.0 via conda install?

Thanks!

[blanchard]

Hi elai, UrbanAccess is not yet available on conda or pip but you can install it via Git by following these steps until it is made available on conda.

Your comment on 0.3.0 and 0.4.0 looks like you are referring to Pandana and not UrbanAccess. Currently Pandana is available on conda for 0.3.0 for OSX, Windows, and Linux but only 0.4.0 is available for OSX right now, Windows and Linux are in progress.