OSM and small-scale crisis events – Transcription

Thank you. Can you hear me? Okay. There we go. Thank you. Next up we have Erika Kamptner. And we’re continuing the humanitarian theme. And this is OpenStreetMap and smallscale crisis events. If you haven’t been watching the news, or if you have been buried under a rock for the last two months, we’ve had a few crises around the world. So I know crisis mappers and HOT have been busy responding to the events. We’re very much looking forward to the talk. Hi, everyone, my name is Kamilah. This is research I have been doing with Penn State University. But by day I’m a New York City GIS analyst. By night, I’m working on this project. So my talk today, I’m going to be talking about really smallscale crisis events. And this is done through a study of user contributions. Comparing different events and seeing how users contribute to these events. So just a brief overview of what I’m going over today. First the objective, why I’m doing this research. There’s been a lot of related research on certain motivations and as well as how users contribute during these largescale events. I’ll talk about that. My methodology, some results I came up with. And then a summary of kind of my key findings and some limitations with this research as well. So first the objective. The big question I’m trying to answer. Do smallscale crisis events act as motivators for OSM user contribution? And to answer this question, I’ll be doing a comparison of four recent building fire incidents. And the reason I’m looking at building fires is that they’re localized in that they’re really impacting a pinpoint location on the map and it’s easy for me to kind of see how users contribute. Next, I’m doing a comparison from different areas of the world. A comparison of how users in western countries versus nonwestern, as well as areas of the map that are already really developed and less developed areas. And next, I’m going to be looking at the measure of increase of OSM contributions following that event. And then finally, who are those users that contribute? And there’s been a lot of research on kind of who are the users that contribute to humanitarian events in general? So why is this research important? Really, the majority of existing research has really focused on these largescale entitles. The earthquakes, hurricanes. But there really isn’t much that looks at are people behaving or contributing in their local communities. So I’m hoping this will kind of research a little bit or open up some of that. And second, you know, this is hopefully supporting ongoing research on contribution patterns and certain motivation factors. There’s been extensive research on what motivates us to contribute to the maps. So I’m hoping this will add to that. So I think a lot of the talks today have talked about these motivation factors. But research is really highlighted. These four key areas, I think, the most. Obviously, there’s this interest in I want to update what’s in my community. I want to update my local area. There are users that are interested in a specific area of the globe or maybe a particular feature. I want to update transit areas in the world. There’s this social aspect, which certainly a lot of us here are drawn to with, you know, mapathons. And then obviously the crisis response. So a lot of research is talked about the emergence of this crisis informatics and digital volunteers. We’re seeing this in the OSM community and Twitter and Facebook where people are openly sharing up to date information about what’s happening around them. The history of crisis informatics started with the 2010 Haiti earthquake and the emergence of the humanitarian OpenStreetMap project. This is from the hot tasking manager a couple weeks ago. Some of the really big events that have tracked in the past couple weeks with Hurricane Maria and the Mexico City earthquake. So research also looks at how are these users or who are these users? And these are, again, some of the key four areas that I’ve seen trends in a lot of research publications. There’s this division of serious versus casual mappers. Today there’s about 4.2 million registered users. But just over 1% of those users actively contribute to the project. And I’m sure a lot of them are at this conference. And then there’s this division of the local versus remote users which we have heard a lot in the talks today. How a local user contributes to the map can be very different from how a remote user contributes to the map. Country of origin is brought up a lot. The majority of OSM users I think because earlier adoption rates in Europe, you know, a lot of them are from Europe or from the U.S. So that is how kind of how users are typically classified. And finally, what are actually these users updating in the project? So the methodology for this, I’m kind of looking at this in two parts. The first part is that question, do users contribute near the location of the event? And then next, I’m classifying those users, you know, how often do they make updates to the project? What kind of users are they? Where are they located? And what kind of features they’re updating. So the four recent building fires that I’m choosing for this analysis are go ship warehouse, the Granville tower, and the prison fires. The reason I’m choosing these are that all four are similar in that they resulted in a structural change to the feature. I’ll be able to see if people updated the building that was impacted. All had international media coverage. Variable throughout the world. But, you know, this would allow me to see if remote users might be, you know, hearing about it on the news. Might be motivated to contribute to that area of the world. And then all of these are also within a oneyear time span. I really wanted to limit the analysis to a short time period since the number of users has really exponentially grown. And I didn’t want that to play a factor in the analysis. So as I mentioned before, just kind of looking at where these fall between western, nonhad with earn and developed and lessdeveloped areas of the map. So my study areas, I’m limiting also the scope of what I’m looking at with a onemile bounding box around each incident. And I’m also filtering the data to only look at the contribution activity in the 30 days before the event, and then what was the activity in the 30 days after that event. And then briefly, just the data and tools I’m using. So in the spirit of OpenStreetMap and open source, I’m using OSM convert and osmosis to process the history and chain set files. And I’m a big fan of Hintaro, it’s an ETL tool, open source, which helped me transform and make the data into a postgres database. And grass GIS and QGIS. So part one, there’s a couple methods I’m using to determine if users how users are contributing. So first looking at a temporal scale, you know, over a period of time. Is there an increase in activity? And then second, what is the spatial distribution of that contribution and what does that look like? And I’ll be looking at the number of feature edits and then the number of attributes because obviously people can contribute in different ways. And then, based on that activity, I’ll look at for part two who those users are. So part one this chart right here shows the 60day study area. Where on the left side of the graph we have 30 days before the event and 30 days after the event on the right. This is showing the contributions per day as a percent of the total edits in the 60-day period. So summarize this, the Kilinto prison fire in Ethiopia, there were no contributions after the event. The fire happened and nobody updated anything in the area. But I saw a huge increase in activity after the Grenfell fire in London, and Plasco in Tehran. Grenfell had the most drastic increase in activity. And I wanted to put a footnote there that there were a lot of outliers excluded from the Go ship fire. A user just happened to be in the area updating an area on the 29th day of the study period. So just keep that in mind. And I’ll bring that up as we go along. So next looking at how those users contribute spatially, I took all those nodes, spatially joined that to a 100foot resolution grid in my onemile bounding box. And this allowed me to get kind of really simple aggregate representation of the contribution. Where the color of the the centroid node would represent the edits and the other would represent number of attributes added to that feature. So first I wanted to get a sense of what the typical activity in those areas were before the events. These were the three events. There is activity in each of these. It is sparse throughout the area with the incident highlighted and the yellow star in the middle. Directly after those events we really see a huge increase in activity. Users are not only updating the actual building impacted, but surrounding areas. And that’s really seen a lot in the Granville incident. Users added footpaths near the event, nearby apartment buildings. Even resources like emergency relief centers that were in close proximity to that Granville tower. So next I took those users that contributed to the project and I wanted to classify them. First, looking at how often they contribute to the project. And then using three different approaches to really narrow down and see where those users might be located. Or where they’re from and looking at their editing patterns. First, the user classification. So in total there were 46 users that updated features after these events. And the vast majority were serious mappers or users that had contributed over a thousand edits to the project in their lifetime with OSM. But I wanted to take note that some of the mappers were first time mappers after the event. There might be smallscale events, something happened in the community and they want to add to the event. So and then I won’t go through the three processes I used for location analysis, but this is just one. And then this example I’m using the change set centroid. I got the centroid and then average those centroids. And there’s an assumption that most of your edits are probably going to be in your country of origin or where you’re from. Obviously, that’s not 100% accurate for all users. But it’s a good estimate of where a user might be from. And in this example, you can see and for the Granville Tour incident in London, there are a lot of local users in England that contributed. But there’s also a lot of remote users, as you can see on the image on the right. So based on those three different approaches to location, these were kind of results. So, again, user location was typically in the same location of the event. But the Granville event certainly attracted the most potentially remote users. And then lastly, I wanted to look at what types of tags these users were adding. And this can give context to whether they were remote or local to the event. I really wanted to point out with the Granville incident, footpaths, trees. These are things that maybe only someone with local knowledge of the area would be able to map. Whereas with the Plasco incident, some of the more common adds were residential highways or shops. Which is something that a remote user might be able to update. So, to summarize, you know, most of the entitles I did see an increase in user activity at the location impacted and surrounding area. And makes me believe that OSM users work in similar ways. Large or smallscale events, they’re updating content and information. The majority of the events were in the first ten-day window of the event. And serious mappers were by far contributing the most to these project areas. And while most of the users weren’t maybe local or from that exact same town, they were certainly from the country of origin, or same country as the event. And while the study only looks at four kind of case examples, I think additional research could be done to determine, you know, if there’s differences in how users contributed in western and nonwestern locations. And finally, you know, going through this research I did realize there are some limitations with this kind of research. Obviously, people are motivated by a variety of different reasons. And several outside factors could have influenced how users contributed to these specific events. The building types, you know, varied. The Granville Tower event was a residential building whereas, you know, the Plasco was mixeduse. That could play a factor in why can someone would contribute. The extent of media coverage. Certainly in the U.S. we heard a lot of the go ship incident in Oakland. Maybe that didn’t reach as far in Europe or other areas of the world. And obviously, like, the current status of the building. All these buildings had a change in them, either ruins or completely collapsed, destroyed. Maybe that played a factor in whether or not someone felt they needed to update that information in OSM. And then lastly, you know, it is kind of challenging to isolate the contributions as related to that event. In the spatial map that I showed earlier, you know, there’s certainly edits that you can really you see tied to the event. Obviously, someone updating the building is related to that event. But someone maybe updating something a quarter of a mile away, it’s kind of difficult to determine if they’re there by coincidence or because of the event. So thank you. [ Applause ] Great presentation, Erika. And it was precisely on time. That was amazing. Any questions? Tyler? Erika, thanks. Really interesting presentation. I was curious from the HOT perspective, we’re often looking at the big events. When you’re talking about big ones, we’re looking at publicity and raising awareness to bring in more mappers. We put the event up on tasking manager. I’m curious, did you look to see whether any of these events were up on the tasking manager or whether the OSM community tried to publicize them to bring in more mappers? Yeah, I looked at tasking manager when I was initially looking through this and I didn’t see any projects specific to this. I did also look I know I’m kind of new to the HOT project. But I didn’t see any trending hashtags or tags that people were using as part of updating those events. So these are really people just on their own adding. AUDIENCE: I like your presentation. It was interesting. I liked it. Now I’m very worried about political damage. People who are elected. But a lot of people will help with food, water and electricity. And it will take one month. But look at the damage. You know? Puerto Rico is damaged. Yeah, yeah. AUDIENCE: I’m very worried. Most people have no food, no water, no electricity. Okay. Thanks. AUDIENCE: Hi, amazing work. I think we need to sort of study these do these kind of event studies much more to understand how events affect contributions. But it might also be useful to take a more policy perspective to see why certain events got more contributions than others. And that point about whether this is publicized in tasking manager. The other thing you could look at it the discussion on the mailing list, whether these events are publicized. This is something that we need help with. Was that something that helped explain the variation? My question is whether you want to look at longterm implications of these events in particular when you look at the new mappers that come in. Did they keep contributing a year after this event in a completely different place. That might be an interesting followon project with this data. Yeah, definitely. Thank you. I think we have time for one more. AUDIENCE: Thanks for the analysis. It’s really great to see. I would suggest there’s a couple other factors that happened in Ethiopia. One is that Ethiopia has between 3 and 4% Internet penetration. So the population just doesn’t have access to OSM. And in September of 2016 was a pretty large civil disturbance. Which they thought was tied to this. AUDIENCE: And definitely. And the government controls the Internet. They turned the Internet off during that time. So you have the compounding factors. I guess it would be good in your research to address those points. Certainly, certainly. And, you know, the big factor I wanted to limit the, I guess, sample sites to a oneyear period. So yeah. I’ll take that into consideration. Thank you for the feedback. All right. One last round of applause for Erika. That was great. [ Applause ]