Rural Accessibility and OSM – Transcription
We’re going to get started in a minute, if you want to take your coffee into the gallery, you can continue your conversation.
Thanks, everyone. I would like to introduce Lee and Olaf, they will talk about rural accessibility with OSM. If you would like to come and share your story.
Hi, I’m thrilled to be here, it is my first time in State of the Map conference. I would like to introduce, with Dev Seed, the rural accessibility platform. We did that for indicator evaluations for our projects.
And so, this is where the story starts, this is a very rural area in China. It is a small town, it is the most poor county in China. And now, we went there, we see this amazing landscape, but there roads there are like this.
This is not hiking roads, this is – it looks like hiking roads in the United States, but it was the rural roads in China that people use every day for their daily needs. We went to the site visit, one of the roads is six miles. We drove in the car, and it took us one hour, because it is too bumpy.
So we work with local governments, and we proposed over 1,000 roads to have along with us. So we provide $200 loan to them to upgrade the existing rural roads to increase accessibility.
So, for every road – for every project that we’re then financing, we have indicator track, the results, and to evaluate how the projects impact people’s lives there.
So, how can we evaluate the impact of that accessibility if we upgrade the roads in a county? If they are financed by the World Bank, we have an indicator saying how much of the population is recovered along a 2KM zone we are supporting. We draw a buffer around the roads, and we count how many people are counting as a buffer zone.
And this means, if a road like this is covered, then the people are covered by the 2KM buffer zone, but it is not connecting them to anywhere. This is not a responsibility by us, by the way.
So we are thinking, actually, the real accessibility is that we are sponsoring – we are supporting those roads, and then the roads can connect people to real needs, like hospitals, like schools, like banks, like markets.
And also, especially we are concerning about the gender issue, like women can access the maternty center to do their routine checks. This is what we work with Dev Seed, to update the accessibility to the resources. So the indicator, converting to how many populations can access the resources, the nearest resources within a certain amount of travel time, say 30 or 60 minutes.
So, but the client that we are working with, they are resource-constrained, and this kind of analysis which requests routing the shortest path, and all the data-intensive input, they might not be able to do it with their own capacity, and their own resources, the financial resources.
So we are thinking that we can develop something for them to use, which is interface friendly, which is low-cost. So that means that we need something open source and also based on OpenStreetMap that can do the routing based on Open Data and OpenStreetMap.
So that’s the story that we come to this platform, that we work with Dev Seed, and also we trained our counterparts in China to use this platform so that, in a future, they can evaluate the impact of the projects by their own.
So the other use case for this platform is that, one, we can evaluate the impact of a network upgrading for the rural roads, what is the improvement of the network level accessibility. The second one is, we can prioritize their list of the potential road upgrade projects. Say, this road can improve the accessibility to this extent, and then we can rank them, and then we can say this is the priority in our portfolio.
And then the third one is, we have a large country suffering from a natural disaster, so we can identify the roads in this disaster zone, and if this road is impacted by the natural disaster, for three months, including all the reconstruction procedure, what is the impact of the accessibility to the people in the area.
So this is a story, then I will hand it over to Olaf to introduce this cool platform.
Olaf: So the rural accessibility map, we started working with the World Bank, building a prototype by Steven at World Bank, and really starting to scale it up. And the rural accessibility map uses OSM under the hood to do routing, and that is what really allows us to go from that – draw a 2KM buffer, to much more advanced analysis. It allows us to take road condition into account, it allows us to take surface type into account. It matters if you are driving along a goat path, or a highway. OSM relies on street map, the software, and the data, and this makes it easy for people to work with. And data acquisition, getting good road connecter data, is still tricky and, for a lot of places, OpenStreetMap is a great data source. So providing the counter parts that the World Bank works with, getting easy data and easy UI data – (speaker far from mic). And what does REM need to do the analysis? First of all, in the population data, we are doing origin destination analysis. And the origins, in this case, are so – they are population estimates. Here, you see some points on the right side, I pulled the population estimates for Boulder County, these are census population estimates. And in World Bank projects, we sometimes use villages with population counts, or world pop, and these are points. So, beyond the population, we need UI data. This can be anything that you want to measure accessibility in relation to: Hospitals, schools, markets, whatever you want to give to the platform, the platform is agnostic, as long as it is a point. And REM has easy ways to patch that automatically from OSM, or you can provide your own data. And then it needs a road network. That can be pulled from OSM for smaller areas, or you can bring your own road network data. You can point things that the road network is routeable, it is one of the most important prerequisites for the road network, we can do this analysis. And then finally, the administrative boundaries, these are the units of analysis. Later on, I will show a little bit how it actually works. But, in the end, what really matters, you want to be able to measure accessibility and be able to compare it to a county – between counties, or provinces and districts. In this case, I pulled in the example I will show you in a bit, I pulled, I think, commissionary districts here in Boulder County. And finally, the platform needs an OSM profile. In this case, I showed a little excerpt of what is in that profile, most importantly, I think in our case, are the sort of the speed profiles, and this really allows you to adapt it to the local context. In this case, it is for an OSM dataset. So you will see primary highway, secondary, etc. If you have a dataset that has a very different classification, you can adapt that and use it as well. So this is the interface, this is the upload interface that we built. And by the way, we can run REM locally on your computer, it is Dockerized, or in the cloud. And so, I uploaded, in a little bit I will show you how we are running analysis. I uploaded data for Boulder and the POIs that I used are microbreweries, so accessibility of Boulder county in relation to the microbreweries in OSM. So you can see, in the colors, each of the origins and the census block level points are visualized on the map with their travel time. There are four gray markers, those are the four microbreweries that were tagged for Boulder county. And this gives you a quick overview of accessibility in relation to the breweries. And each of these points can have multiple population estimates against it. So if you want to measure the accessibility of particular socio-economic, you know, of the poor, in relation to hospitals, or if you want to do female populations in relation to maternal health things, in relation to the male population, you can do that. And another thing is that there’s a network, a little network button there grayed out. For smaller growth networks, you can import road network data and add in the road network data to run scenarios. That is really cool, this in itself is nice, but if you want to measure the impact of a road upgrade, then you want to run scenarios and sort of see the difference. And that network – so, it will load iD editor, you will be able to drill down to the road network, and upgrade your roads. What we also do, then, is you see here the second table, the percentage of population with access to microbreweries, four, five, six, those are the commissionary districts, usually there are nicer names to them. So usually, it will say a province and it will say, 80 percent of the population with this province has access to a hospital in 10 minutes. But, again, what makes this really interesting are scenarios. So what I did in this case, I took the road network and make Boulder a living street, so living streets, I think, are in our speed profile, and they are set to maximum intent, to 15 kilometers an hour. So if in Boulder, you can only drive 10 or 15KMs an hour, what is the impact on accessibility? This the result. If you compare this to the baseline, this is the result that you will get. And here, you will see the impact, in relation to the baseline, it takes 30, 30 minutes longer to get to your POIs in some areas. So next – so next up: REM is an open-source tool, we would love to get more people involved, more people use it, more use cases. We would love to be able to bring in other datasets, for example, flood masks, bring in a flood mask and doing an an automatic intersection of the road network and see the impact of that on accessibility. More powerful analysis, we don’t take volume into analysis, but we could. And other types of analysis, critical link. The only thing that OEM returns are travel times, but we could analyze the links themselves and the criticalty of the links. And then it improves the pipeline to get data from OSM, that’s a major one on the wish list. Right now, we rely on overpass, for bigger areas, it does not work very well and it is a little bit unstable. I would love to get a better pipeline there. And this is one – this is analysis that we did for Laos that we put in, this is a nice picture. This is accessibility to education facilities near Vientiane, the capital of Laos, and this is all on OSM data. I thought this was a nice analysis. If you want more about the project, ruralaccess.info, if you want to know more, there’s a description on the website, and we will be around. Thanks. (Applause).
So I’m curious if you had any way to tell a confidence interval that is built into your analysis, because some areas have better data availability than others. I wonder if you are able to quantify that in your application as well.
We don’t have that yet within the application, but we would love to be able to do that.
Next? Any last questions? It looks like you are attributing your geometries – you are giving them each representative points, or centroids, and then you are attaching these to the network graph when you are determining accessibility. So for rural areas that don’t have a lot of road network data, if you have the geometry that is distant from a road network, how do you resolve the edge to connect that to the graph? So we will take walking speed into account, that is not perfect. We acknowledge it is not perfect. But we will find the closest road, and then we will take a walking speed into account. But that is not perfect, right, because the closest road can cross a river, or a mountain, or – yeah.
I think we can take one more.
Yeah, does your model take into account different modes of transportation, such as walking or, like, different roads will only be accessible by walking versus automobile, is that used? So, no, it doesn’t handle multi-modal. But we can set speed – you can set speed limits on roads. So in a certain way, yes. So technically, we don’t support multi-modal. If a particular road can only be, if a particular road can only be walked on, you can set the speed limit at 5KMs, or whatever it is that you want to do, to set it to. Hi, I really like the pictures that you had, but I couldn’t make out the legend. I didn’t understand the images you had, especially the one at the end. I wondered if you could narrate it for me. For this one, for example? Yeah. Okay. So what you see here, all the colored dots are the origins, in this case, are villages, in Laos. The size of the bubble is population size. So we have quintiles, if we looked on the right side. So the population estimate determines the size of the point. And then the color is the time it takes to get to the nearest POI. The POIs are on the map with those POI markers. Does that answer your question? Great. Thank you, Olaf and Lee. We will bring on the next speaker. (Applause). Live captioning by Lindsay @stoker_lindsay at White Coat Captioning @whitecoatcapx.