Loggingroadswood


Mikel Maron: Crowdsourcing satellite imagery to document deforestation

 

Mikel Maron has a long history in open source, open-data and mapping. He is a founding member of the OpenStreetMap Foundation, which led him to work in other collaborative community mapping projects such as Map Kibera, the Humanitarian OpenStreetMap Team (HOT) and more recently Moabi, Mapbox and Logging Roads.

The project Logging Roads maps roads built for logging in the Congo Basin in order to draw attention to illegal logging and deforestation. Since its launch in 2015, contributors have used satellite imagery to map 30,532 roads, revealing the vast interconnected system of deforestation which has developed over time. In this interview with Lisa Gutermuth, he discuss this project along with the differences between aerial imagery and other more traditional forms of mapping, verification and collaborative mapping.

 

You have been working on citizen-based mapping projects for a long time. What interests you about this form of mapping?

I was first attracted to mapping because I was interested in data visualisation. I thought putting a dot on a map was the most straightforward thing one could do. I was wrong. I do think the connection between information and what is happening out in the world is most clearly expressed through the practice of geography and mapping. Part of my motivation is to take the tools that people are using and bring them out into the world where they can have a direct impact. I think it's important to shift the framework of mapping to have citizens thinking about the world around them, and, in a structured way, think about what their agenda and priorities are and what kind of actions can take place. Mapping is not only an interesting tool in itself, but a great tool to engage in particular problems.

What has kept you involved with citizen-based projects specifically?

The more people involved in creating the map, the better the map is. The more people that have the power to contribute, the better the result, and the better picture we have of the world. I think this is demonstrated by the fact that this process is changing the way that traditional mapping is being done. A few months ago, I finished up as a Presidential Innovation Fellow, taking a different perspective on collaborative mapping, and applying it to US government activities. During my fellowship experience, I observed very traditional organisations, who work within humanitarian and government contexts, not only adopt OpenStreetMap, but also take the additional step of asking, “What's really going on here? How do we really engage with this community beyond just consuming a data-source and how does this ultimately change the practice of these institutions?” This is something that I became fascinated with as I was working with many large US government agencies and other institutions. They were people who really see this as the best way of getting, making and sharing data so to see them asking these questions was fascinating.

You can also look at it through the lens of power dynamics. There is a recognition that if someone can contribute something valuable, such as spatial data, then there are other levels that those in power will be encouraged to engage and listen to. Collaborative mapping projects brings many diverse and unlikely groups together.

How is working with aerial imagery different to traditional mapping?

It's harder to draw a community around aerial imagery simply because the data is harder to collect. Until recently we've had satellite and aircraft-acquired imagery, or maybe a kite or balloon, but the idea of a commons based aerial map didn't really become feasible until UAVs (Unmanned Aerial Vehicles, commonly known as drones) became accessible. They still are out of reach for a lot of people, and it is still a smaller group than those working in traditional mapping, but now you have the technology available for people to collect and share the imagery. The process has now gotten to the point where it's relatively straightforward if you have the right gear and process. Technically it's been smoothed out quite a bit.

Is the goal different with the collection of aerial imagery?

What you do with it, certainly in OpenAerialMap, is different, because it's not necessarily one global map. The goal is not to make a single seamless satellite image of the earth of the kind that you see in Google Earth or Mapbox Satellite, where you simply compose a single layer. When you have imagery that is made up of a composite of a huge number of different sources and times, you want to be able to retain things, like the exact time that something was collected and who collected it - so there is a somewhat different goal there. It's more of a catalogue than a single collaborative space. But the value is in creating a single catalogue where things from multiple sources are searchable and usable in the same interface.

How did Logging Roads get started?

The start of the project, or the start of my involvement in it, was to bring the collaborative processes and technology that we use in OpenStreetMap to a different domain - to the tracking and monitoring of deforestation, primarily in the Democratic Republic of Congo (DRC), but also elsewhere. The idea of Logging Roads came up in discussion between Moabi and the World Resources Institute's (WRI) Global Forest Watch programme. GFW processed large amounts of historic Landsat imagery, in order to conduct various kinds of analyses of changes in the forest, and generate alerts. A lot of this was carried out by automated algorithms, and by manually analysing the imagery to gather details. But it's a lot of work. It turns out that OpenStreetMap is a community that does a lot of this, and does it really well. You can be fairly certain that logging roads have an impact on the surrounding forest. So it's really important to look at the forest historically.

There is a tool called the OSM Tasking Manager which is used for data creation in HOT’s response to humanitarian events. This same tool was re-purposed for coordinating the mapping of forest roads. What this does is coordinate contributors and some specific data monitoring on the back-end so we could track and visualise progress on the website. The way the tasks are presented is different from typical HOT activations. It has become much more about micro-tasking, which allows you to look at logging roads one at a time, and based off multiple years of imagery, select the year when the road first appears. So that's an easier way to get started; you can just look at satellite images, click a year, and then take the next one, so it really lowers the barrier to engagement.

Screenshot from the website Logging Roads of a GIF that demonstrates the data about Logging Roads from 2000 (and any data available from before 2000) to 2014.

What are the goals and target audiences of Logging Roads?

The target audience is more on the analysis side. We are looking at the drivers of deforestation to draw conclusions of what happened to start with. For example, how did the network get set up? WRI does a lot of this kind of work, and typically this analysis is not very useful for awareness-raising, but in this case it is a way that people are getting involved in this space, and people are able to make a contribution and understand a bit more about deforestation issues.

Moabi documents their findings on a blog, and puts the information together with other resources, such as agricultural concession data which can show a more complete picture of what the deforested land is being used for, or rather, what the main drivers of deforestation in this context are. Most recently they've written about the expansion of industrial agriculture in DRC, for oil palm in particular.

Having that data is incredibly valuable to help model what might happen and to guide policy in the future. So it's not too much of a stretch to say that Logging Roads helps document deforestation, and there are a lot of reasons for which having that data explicitly drawn out helps tune analysis to exactly what is happening and what we can expect to see in the future.

Will there be a preventative function of Logging Roads?

Part of what Global Forest Watch intends to do is monitor deforestation and trigger different alerts whether they're people working more on the analysis and legal side or more on the advocacy side. I think there's a potential for this kind of data to inform models that would send alerts, but I don't know how much is yet happening in practice.

Are there standards that apply to what information gets published?

I don't think there's too much of an issue with Logging Roads. With other sorts of data that Moabi is targeting, there is that concern. One of the things they are collecting with on-the-ground monitoring is what is being removed from the forest and what is happening to forest-dwelling communities. They have a lot of protocols about how that information is collected and managed.

What are the next steps for Logging Roads. Are you looking to expand into other regions?

Yes, that's definitely the idea. Moabi has been connecting the results to the wider application in forest science. Remote-sensing and citizen participation, in particular, have so many applications for the generation and use of environmental issues and climate change, so that was a big step in getting this project started. At the same time, Moabi will embed Logging Roads in other monitoring processes such as REDD+ and other policy processes in DRC, in particular, collusion between logging companies and industrial agriculture which has lead to much of the deforestation in Indonesia.

Screenshot from the website Logging Roads comparing four different periods of time. 

What other contexts, besides forestry, could this method be applied to?

In the realm of environment and climate there are many contexts where it could be applied. Moabi is already adapting these tools to map dam development across Asia which could flood villages and block water and sedimentation flows downstream. There are similarities around identifying coastal change and changes in surface water. This is something we were talking about during the California drought - this kind of process could be very useful for identifying issues with different water sources in other regions. I think similar coastal changes like incursions into the sea could be logged - how that's driven by real-estate and development. Take

Rajwa

Read about Rajwa, a Lebanon-based activist and member of the Mashaa movement. She used tools including Google Earth to map out not only the city's seashore, but the political-corporate interests keeping the public at the gates. 


Read more
Explore Resources
Rajwa, for example, looking at the build-up of the Lebanese coast and investigating environmental implications.

Would you say that there are risks in this work either in the work-flow or verification process?

There are many aspects to that. There are the security risks to those contributing the data themselves, as well as the risk the data poses to them individually and to their community. That was definitely something we considered when starting Map Kibera. I think we've seen that it's a lot more subtle, and on the whole, more beneficial to show what's happening in these communities. A key risk for communities would be that we show that happens in a slum, and that gives the government or others, the opportunity to come in and take advantage of what's there.

With OpenStreetMap, by focusing on infrastructure and public, surveyable data, if we're not collecting personal data or personally identifiable information, it's not very high-risk for individuals. So those kind of issues haven't really come up.

There is the risk that comes with making data completely open, in that potentially bad information can be fed into OpenStreetMap, similar to Wikipedia. What we do is similar to Wikipedia in that it is completely open, but the level of vandalism is actually a lot less. We haven't had to go to the lengths that Wikipedia employs in order to protect the map. We have people that are very passionate about the maps they have made; they monitor changes to the maps that have been created in various ways using various tools. There are communication channels if something suspicious appears, or there is actually someone causing problems to the map. There is a community that is very reactive to checking anything which is suspicious or shouldn't be there.

If there's something that's beyond the scope of a community - the community can't solve it, whether it's technical, or a dispute over a name or a boundary that doesn't get resolved - and this happens very rarely - then there is a working group within the Foundation that has extra abilities to resolve the issues. That's less of a problem but it’s still a concern because OpenStreetMap is growing incredibly rapidly.

Do verification methods change with aerial imagery?

In terms of validity, I think there's an interesting question there with regard to satellite imagery, about how you ensure that this was taken at the time. If it's something that is of critical importance and you really need to have documented evidence in court, you can certainly do that and you can go back to the providers. There's a legitimate and recognised authority in that, but there are far more different types of citizen-collected data and there needs to be additional processes of verification.

How do you verify the source of satellite or UAV data? And in the event of a humanitarian crisis, how can the UAV video which might provide crucial information for the purposes of rescuing survivors be accessed?

The threshold for verifying a source of data in a time of crisis is going to be lower than in a court of law. That has allowed OpenStreetMap to emerge as a source for humanitarian response. We certainly take the same approach to imagery. We would of course take a look, and compare it against other sources, like UAV vs satellite – even if there's been a great deal of change after an earthquake or typhoon, it should be recognisable as the same place. Especially if there's local knowledge, you can look at the specifics of features. You should be able to tell something about the time of year based on what structures are there, when the imagery was possibly collected, whether it's more or less recent than something else. These are the types of things we regularly look at when we're dealing with any kind of imagery source.

 

Credits

The interviewer, Lisa Gutermuth, is a Programme Coordinator at Tactical Technology Collective. Her research focuses on the intersection of satellite imagery and agriculture. 

The header image features a wood truck for the Company Fabrique Camerounaise de paquets (FIPCAM) near the village of Ngon. District of Ebolowa, Cameroon. Photo by Ollivier Girard for CIFOR Flickr).