I work in an organization which implements programs which impact public health at the state level. We have recently been working on analyses assessing the accessibility of services for state residents, especially in rural areas. We have been operationalizing accessibility using isochrones created using the Esri Routing Service, and while this has helped identify areas which are comparatively far away from available service locations, it can require subsequent generation of isochrone service areas to determine what the travel time might be from a service location to a faraway community. Further, if we overshoot the travel time from that community, we may have to run another isochrone analysis to get a more accurate measurement of travel time. To avoid this clumsy approach to evaluating travel times to services from across our state, I think a raster dataset with cell values representing the distance to the nearest service location would be useful.
While looking to see if something like this is already possible, I found an example of a travel time raster created by @curtvprice and shown in his July 18, 2018 reply to his postCan I use Cost Distance tools to model travel times?
In his case, though, the travel time cost surface was made using his own data, and we are not in a position to use our own data to generate a cost surface. Since the Esri Routing Service is able to generate isochrone polygons, though, I feel like it should be technically feasible to use that data as an input to generate a travel time cost surface. If possible, I believe we would significantly benefit from the ability to create a travel time cost surface and/or a derivative travel time raster using the Esri Routing Service as input data. I believe it would allow us to spend our credits and perform analyses more purposefully and enable us to generate more insights from each analysis we run.
Hello @KevinBoes2. Thanks for submitting your idea.
It sounds like the first piece of information you need is the travel time between a service location and one or more communities. If I understand correctly, you're currently running potentially multiple Service Area analyses to determine this travel time. Have you considered using the OD Cost Matrix service instead? This will tell you the travel time between a set of origins (your service locations) and a set of destinations (the communities) directly, without the need to try multiple cutoffs or overlay any geometry. The OD Cost Matrix service is also fast and quite cheap.
Service Area and your rasterization suggestion may still be a good approach if you have an extremely large number of inputs such that OD Cost Matrix is impractical, but I thought I'd suggest it and see if this leads you to any good solutions.
Hi @MelindaMorang! Thank you for this suggestion! I think what I'm hoping to get is an output which isn't dependent on prior knowledge of what locations people will want to know travel times from, so I believe a travel time raster would be the best solution.
That said, it would be good to be able to estimate people's travel times by treating their locations as their nearest municipality's centroid. The OD Cost Matrix sounds like it would be a useful tool for producing estimates from our service locations to municipalities in the meantime.
That makes sense.
Here are some additional ideas on how to approach the raster idea using existing technology:
(If you're in the US or your country has something similar): Use the centroids of census blocks (if blocks are small enough for your purposes) as the origins or destinations and calculate your OD Cost Matrix. Then, assign the block centroid travel time to any point falling within the geometry of the census block.
Do something similar but using a raster-like grid or hexagon grid.
Calculate the travel time from every street intersection or the center of every block. Any future location gets assigned to a block or intersection and uses the precalculated travel time.
I like this idea of using an OD Cost Matrix with points that will eventually become the centroids of raster cells! I can use the OD Cost Matrix tool to calculate the travel time from each centroid to each service location, and then I can summarize the resulting matrix to get a table with the minimum travel time for each centroid. Then I can join that table to the centroids point layer and use the Point to Raster tool to convert those centroids to raster cells and get a travel time raster.
Thank you for your help, @MelindaMorang!
You can also start with a raster or grid and the derive the centroids/points from it, calculate the OD, then join the results back to the original raster cells.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.