[00:00:22] Bridget Scanlon: I'm pleased to welcome Jamon Van Den Hoek to the podcast. Jamon is an associate professor of geography at Oregon State University and he leads the Conflict Ecology Lab. Jamon uses satellite imagery and geospatial data to learn effects of violent conflict, land use and climate on people and landscapes.
Today we're going to talk about his recent work in Gaza and also Ukraine and refugee conditions in Africa with an emphasis on his work in Uganda. And we'd also discuss the various remote sensing tools that he uses and applies to quantify impacts. Thank you so much Jamon for joining me today.
[00:01:08] Jamon Van Den Hoek: Thanks, Bridget. Yeah, very happy to be here.
[00:01:11] Bridget Scanlon: So, Jamon, your most recent work I guess is on the Gaza region. And I think maybe you could provide a little bit of background to the listeners on the situation there and what you have been tracking.
[00:01:25] Jamon Van Den Hoek: Sure, yeah. Well I mean, the conflict in Gaza now, it started October, 2023. We're at about a year and a half into this large-scale armed conflict. It's a conflict that is very unique in many ways. Others have made historical comparisons to the level of damage, the number of buildings destroyed, the number of people affected, these kind of comparisons.
It's really an unprecedented war in that it's a total war. We don't often have total wars where that war dominates every facet of life. If you think about, you mentioned Ukraine, right? Ukraine of course that's a large scale armed conflict as well. But partially due to Ukraine's size, partially due to the way that the war is being fought as well, the war doesn't have a direct impact everywhere in Ukraine. It's indirect everywhere, but it's certainly not a direct impact everywhere. Similarly for Russian civilians, some Russian towns are directly affected by the war, some are indirectly affected. But in Gaza, every square kilometer of Gaza is affected by at this stage tough to find a sort of a patch of land that's not directly affected.
Gaza is also unique because this conflict is unique because, people have not been able to leave Gaza. So, there's no refugee crisis because people can't leave. So, there's about 2.3 million people estimated in Gaza prior to this conflict. There are different assessments of the number of people who have died as a result of the war, but the bulk of those people are still in the Gaza Strip.
And of course, many children have been born in this period of time. So it's a very unique war for those two characteristics. We approach this war we're not political scientists but of course our work is engaged with geopolitics and political actions as they're manifest on the landscape.
The way we come at this war as well as how we come at the war in Ukraine is through a remote, satellite based long-term assessment where we take open access data, so data that anyone, any of your podcast viewers could access the same data we do, and we use the data we do use are from the European Space Agency.
These are Sentinel One radar data sets. And we use these data to map urban damage primarily is what we've been working on, that we also have work on agriculture and water infrastructure as well. But we look at basically find signals on the landscape that of abrupt change, stability, and then abrupt change.
And we make maps of the area of damage. We make estimates of the number of buildings that have likely been damaged over time. So it's not a before/after. In the fall of 2023 when the rate of damage was so great, we were analyzing every image as soon as it was collected, as soon as it was available on the cloud, we would process these data, make a map, distribute the data to different journalist organizations, humanitarian organizations that we had been in touch with or who had reached out to ask for the data. And there were a couple of reports that we were able to generate within four to six hours. The images collected in four to six hours is out the door.
Gaza is a small region, so, the image the geo tiff that we would share was 62 kilobytes. It was a binary map of like, here's the total damage to date. So these are small, uncomplicated products by design. But the story when you pack them, when you look at them over time, one after the other, after the other, you just see this amazing, in terms of like the sheer pace of the thing, such quick damage, such extensive damage. And of course we've carried that through in a regular sense until January of this year, and we are about to do another update based on some recent attacks in the south. But it's been a really transformative sort of research effort, not just because of the close engagement and cooperation with journalists and humanitarians but also just the sort of technical capability that we needed to develop to make these data, make these maps in such a short period of time and really be persistent with it to stay with the conflict itself to learn about the conflict dynamics so that we could use that to inform how we make the assessment.
It wasn't, in no way was it a , here's a place on the earth and something's changing, let's push a button and make a map, like we were really trying to learn as much as we could about the conflict itself. Because that partially informed our analysis. It certainly informed how we thought about validating what we were seeing thinking about the data quality that we were working with.
So, it was a very transformative we had to transform how we did the work to keep up with it and to get it out. So, and it's ongoing, still. Yeah.
[00:06:15] Bridget Scanlon: Yeah, that's amazing. I mean, you described the differences between Gaza and Ukraine, and the area of Gaza is like 365 square kilometers, so it's the, like the size of Delaware in the U.S, so it's tiny.
And Ukraine, and you mentioned before that that's like a frontline war so you have a specific area where things are happening and with Gaza, it's everywhere.
But I mean, you also mentioned when we chatted before that there have been conflicts over the many years, 2014, 2021 and even before then. The Pacific Institute has a nice chronology of conflicts and describes water aspects of the different conflicts in their web database.
And so they mentioned 2021, 11 day fighting, half of water network impacted and, desalination, plant damage, it's really nice and and a great resource. So , but it seems like many of their descriptions of different events half of the water network was damaged in different times.
And so it just the whole place seems to be just a total catastrophe. I mean, seems impossible that, and, and you mentioned you use a Sentinel Two data , so that is optical data and shows you the extent. Do you also use Sentinel One to give you building elevation or anything to figure out if the building has been destroyed and the elevation has changed, or do you just use Sentinel Two?
[00:07:48] Jamon Van Den Hoek: We're actually primarily Sentinel One. The radar-
[00:07:51] Bridget Scanlon: Sentinel one! Oh, sorry. Maybe I was incorrect.
[00:07:54] Jamon Van Den Hoek: No worries. Yeah. Sentinel One is primarily what we do. We don't look at the elevation change. We have different indicators of damage. I mean, generating elevation models from radar, from open radar is challenging in its own right.
There have been some really good papers recently to talk about this but elevation change is not the only part of damage, like the sensitivity to radar, that radar offers, is actually the lateral it's the structural damage. I mean, if you can see a building, of course when buildings are surrounded by other tall buildings, you're getting blocked perspectives.
But if you can see the side of a building with radar, you should be sensitive to the damage at a very fine scale of the severity of damage cracks and foundations. There have been theoretical, not theoretical, but very fine scale detailed assessments of sensitivity of radar systems like Sentinel One, Two, structural damage.
But the Sentinel Two we have used as well that our colleague, he led a project looking at agricultural damage, greenhouse damage, and tree crop damage in Gaza. And that is an optically driven assessment. The radar doesn't help so much with that. But the optical is really the go-to for that kind of work.
We work with all of these different kinds of data sets. I mean, the validation, we can't, we don't have field access, so we have to work with available high resolution imagery both historical and those that are being generated real time. And that case we're primarily reliant on planet scope which is about four meter resolution data.
Other systems that are typically available over Gaza were not made available during the war for political reasons. So a little bit always constrained, I guess , with the data we have, but the goal is really to triangulate these different sources. And working with the journalists, in some cases, they had insights they could share with us, right? Like local reporting, people had context that checked something out in a place and that, that would feed back to us both in Ukraine and Gaza. That was extremely unique. I mean, the value of that is, it wasn't systematic, but it, there were spot checks here and there and that was very new to us and extremely helpful to have that on the ground perspective.
[00:09:58] Bridget Scanlon: Well, it's amazing that you were able to turn around the outputs in such a short time with four to six hour latency. So how frequently do they put out the Sentinel One data?
[00:10:09] Jamon Van Den Hoek: Well, in Gaza it was about every six days we would have a new shot. It's a little bit different because we have to have a consistent view angle. So, depending on how we see which direction that the satellite is coming, it might be coming from the South Pole up to the North Pole.
Effectively, that's a different orientation than the North Pole coming to the South Pole. This is just satellite orbital mechanic stuff, but we have to get those, we can't compare those directly. They just, they see things from a different perspective and radar is so sensitive to geometry. But at the end of the day, we were able to generate, it was about every week, within a week, that was our average return. There was some dates as well where the image just didn't, it wasn't sufficiently quality. It wasn't high enough quality to use because of, again, the sort of orbital misalignment. But working with Corey Scher, who was at CUNY University of New York and is now a postdoc here at Oregon State, he really did an amazing job at developing the kind of quality controls to make sure that the data we were using was consistent to get this pipeline started as quickly as we could and then to get the data out to the sort of this global network of end users.
[00:11:17] Bridget Scanlon: So, Sentinel Two, it's a 10 meter resolution,
But then when you need to compare different images and you have this orbital issues and things like that, it's more you are operating it more like 20, 30 meter resolution or what type of resolution do you think you get with it?
[00:11:36] Jamon Van Den Hoek: Well, Sentinel Two is, the true color is-
[00:11:38] Bridget Scanlon: Oh no, sorry.
You were talking, you were using Sentinel One. I keep-
[00:11:41] Jamon Van Den Hoek: Sentinel One. No, it's all right. Yeah. Sentinel One is also not, they're both 10 meter, Sentinel Two is that it's only 10 meter for the true color. Everything else is 20 meters for the, say, the near infrared and mid infrared. But Sentinel One also nominally 10 meter resolution.
But the approach that we work with is called coherent change detection. There's similarities to stereo photogrammetry where you have to have these multiple looks to develop a measurement of coherence change. I mean, coherence itself talks about similarity between two dates.
So you have to have these multiple looks. And of course, with satellite imagery, the pixels never line up one image to the next right? You get the whole, you get the same, area coverage, the general extent is the same but your pixels are shifting. So, in doing that, then you have this sort of overlap and you have to look across these multiple pixels side by side.
So, that ended up giving us a working resolution of 40 meters.
If we did some experiments to try to increase that, say, the 20 meter resolution and the noise was just too high. So, it's that 40 meter resolution we're really optimistic for using very high resolution radar data to do this going forward.
These are mainly, these are right now just commercial systems, but those would be operating closer to like one meter resolution. That is not quite the track record to do the kind of work we're doing with those kind of systems, but it's just a matter of time. Those systems will be ready to do this in the next few years and that will increase the sensitivity by an order of magnitude.
So super exciting about that. There's also another system called NISAR which is a joint NASA Indian Space Agency. Everyone's super excited about NISAR. It is going to be a game changer in lots of different ways. Multiple frequencies for radar, tons of applications that Sentinel One is helpful for but which NISAR will kind of be able to specialize in.
So, that's supposed to launch in a little under two months now in June of 2025. So, that's exciting and a great example of cooperation between two space agencies.
[00:13:46] Bridget Scanlon: Right. I know NASA also collaborates with Germany for the GRACE satellite stuff.
So, I think they, they need each other and they benefit from that collaboration. So, so that's good.
And you also mentioned planetscope, so how frequently do you get planet scope images ,
and what's the resolution like for that?
[00:14:06] Jamon Van Den Hoek: Planet Scope is, so that's from the company, Planet, San Francisco based commercial provider. Those data it actually depends on where you're at in the world. The spatial resolution for that, you get different spatial resolutions the closer you are to the equator, closer you are to the poles, but it's nominally between like four and five meters.
Sort of depends on where you're at. But that's sort of a good rule of thumb. And those data are composites. So these systems work differently, you have this big constellation of several dozen satellites orbiting at any given time. And so together they give you daily, in some cases sub daily coverage at any given location.
But the planetscope data, the data that go into that become really useful when you combine them when you aggregate them over time to weekly or monthly. Some people in agriculture mapping just do seasonal composites. You generate this whole collection of data and then you find the highest quality pixels.
And so the planet scope data they're able to give you, I mean, the weekly and monthly is a pretty typical time step for planet scope. Lots of applications for tropical deforestation harvesting agricultural dynamics. As I said, there's a lot of work in cryospheric space as well. So, for us, of course, in this Gaza application we were working with monthly composites and able to that was a reasonable time step,
Agriculture: it does literally of course it changes all the time, the morning to the evening, crops are going to have a slightly different position and condition et cetera. But for the time steps we were working in we used monthly com or for the processes we were interested in, rather, we were working with monthly composites. And capturing it did a really good job of capturing the kind of agricultural extent that was common across Gaza and also for detecting some of the changes, the damage therein. There's higher resolution available data as well, at the more of a half meter or meter level quality also from Planet, that would be their sky sat image that works differently than Planet Scope.
But it, it's more akin to the kind of data that would be on a Google Earth base map or Google maps. Like just you look at it and it's, it has high enough fidelity you can make out individual sidewalks, you could see your tiles on the roof of your house, this sorts of thing. So ,
it's a very clear photograph from space.
And so for that you can see a lot of detail that would otherwise be lost.
[00:16:20] Bridget Scanlon: What about other things like Open Street Map or Google Maps or things that. So, do you look at a lot of different things try to constrain what's happening and to kind of develop a conceptual understanding? And then it's nice to have the journalists and other people to ground reference what you're, and trying to help you interpret what you're seeing, I guess.
[00:16:41] Jamon Van Den Hoek: Like the events like we're not, we don't have a reference database that gives us like, events, right? And if we, if it is made, it's too late. Now of course there's all these conflict databases in Gaza and around the world. ACLED (Armed Conflict Location and Event Data) is probably the preeminent of these. But in this real time capacity, it doesn't help us.
Those are all collected too late. I mean, those are collected over time, and so by the time that assessment is ready, we've already missed our window to get a map out. But you mentioned OpenStreetMap. We have a really nice working collaboration with the Humanitarian Open Street Map Team, often abbreviated as, or acronomized as HOT.
And HOT is amazing, it's a global network of professionals who are employed by Open Street Map and by HOT, but also it's volunteers. There's a lot of extremely skilled volunteers around the world who are generating these data. And they are producing updates of building foot prints kinds of information on roadways or building usage or these kinds of real time assessments with whatever imagery they have available to them. So, one of the gaps we had in Gaza was just a high quality building dataset, building footprint dataset, and working with HOT, they basically had a call out to their global network of highly skilled volunteers and they refreshed the building footprint map to bring it up to , in the case of Gaza, September 2023, right before the conflict started.
So we had this very up-to-date, very detailed map that when we compared it to other data sets from Microsoft or Google, it added almost 150,000 more buildings to the map. So, we learned, right, you can't take for granted that any of these data sets are quality, but it's also really tough to show that they're not sufficient.
Right? Like how would we have shown that? Without doing exactly what HOT did, which was well, let's get an up-to-date image that you can visually interpret and then we'll go through it. So that quality of those reference data has been, that also really came through with that Gaza analysis and is still something to be considered in any sort of area where you don't have the familiarity or it's a very broad area and you need these reference data.
The quality controls what kind of count, of course, you get. If you're trying to count damaged buildings and you only have half the buildings in your dataset in the first place, you're, of course, always going to be under counting.
[00:18:58] Bridget Scanlon: Right, right. And so are you doing change detection then from one week to the next? Or are you just putting out the images as is whatever you get and to the journalists or do you do change detection or do you do both?
[00:19:12] Jamon Van Den Hoek: I mean, we perform a change to detection analysis ourselves with using these open data the Sentinel one radar data. But then we would make a map, we made two maps, one was the new damage since the previous assessment, and then one would be total damage. This was also designed to sort of support journalist reporting, which was like, what's happened lately, right?
Because at the time there's just, there was so much attention day after day after day of what was happening in Gaza. So we were trying to align the cadence of our remote sensing analysis to that as well. So we would make those changes, do that change analysis, and then get out the total and the new data.
We never, we didn't really share like the raw imagery themselves just because it was too coarse. It wasn't, there were a couple of examples of journalists publishing the, the radar derived data, the coherence maps, which was very cool to see. But by and large we just gave these derived products.
And then they have all these amazing cartographers and data visualization people and obviously journalists who flesh out these very complete narratives of change and the human cost and the political considerations and just these stories that, that we saw journalists put out were, I mean, way more compelling, way more interesting than the kind of maps we were making. But it was really, we felt very lucky to be able to be involved in that kind of collaboration where they would just add these levels of detail to this, that, of course we were, we are completely unaware of seeing this stuff from tens of thousands of kilometers away.
But also very cool when the story's lined up right? Like they have a story through interviews, and then if you look at the map, there's a synchronicity between what the map shows and what the story is that happened a lot and that ws gave us confidence that we were saying something that actually made sense.
[00:21:05] Bridget Scanlon: Right. Well, that's really cool.
Did you look at any areas specifically? I mean their desalination plant or their water treatment plant or their sewage treatment plants or energy sources or things like that, were there any regions that you focused on and that you were tracking sort of to see damage and stuff or-
[00:21:26] Jamon Van Den Hoek: Yeah, we didn't have any specific place in mind. We just did a, effectively a wall to wall Gaza wide assessment. But as you said, all of those are in there, schools, bakeries, water treatment facilities it's all there. there was a separate study that Brian Perlman, who, is one of our collaborators also led, which was specifically focused on WASH infrastructure damage towards the beginning of the war.
And that was a different kind of mode of analysis, but expressly focused on WASH, principally through a public health lens. At the time when that dataset was coming out, there were concerns of cholera. There was even, like signs of polio coming back. So, a lot of these impacts of sanitation and waste disposal is something that another collaborator of ours, whims Weinberg's, been working on,
informal waste dumps. I mean, there's so many.... We have collaborators working on damage to schools, others working on damage to hospitals. I mean this is, again, going back, this is total war. There's just name your facet of sort of infrastructure or society and there's something to talk about there, there's some impact directly of the war that is not, probably not brief and probably certainly not resolved at this point when we're at that stage of the war where it's just this grinding recurrent attacks in Gaza and a stalemate in a lot of, just about any way you could see it right between the Hamas and the Netanyahu government.
So it's very unfortunate.
[00:22:53] Bridget Scanlon: It seemed like, I mean, Gaza relied on the coastal aquifer a lot. And then, the over exploitation of the groundwater in near the coast, then leading to seawater intrusion. And then flooding of the underground systems amplifying seawater impacts on the aquifer.
But that's not anything that you were looking at or had knowledge of or anything?
[00:23:17] Jamon Van Den Hoek: No, I mean this, is a good question because this has come up many times like this sort of the sensitivity to what could be happening in the tunnel network under Gaza and, with our approach, we're only really sensitive to what's happening above ground. And there have been quite a lot of work not so much on those tunnels, but on the concern of as you said, of the aquifer intrusion.
And it's really unclear what the status is right now on the drinking water. The shortage of drinking water has been a pervasive of potable water, I should say, has been a pervasive challenge. And Gaza for a generation or more. Oh I think about a decade ago there was a UN assessment that was: Gaza would be uninhabitable by 2020.
And this was one of the main things that they identified as just the, this how to solve desalination, right? In Gaza there were certain constraints on what kinds of materials and construction equipment could come into Gaza that were imposed by the Israeli government, the blockade as it's commonly called.
And that just made it really tough to kind of keep up with these technological developments that could have addressed some of those problems. But now, things have changed so much with this war that it is unlivable for other reasons. But the question now is what is the future of Gaza?
And who's deciding what future that will be?
[00:24:35] Bridget Scanlon: Right. And when we chatted before, you mentioned some of the impacts on the ecosystems were counter to what one would expect, I mean, the coastal areas and sewage discharge and fishing and things like that. Maybe you can describe that a
little bit, Jamon.
[00:24:52] Jamon Van Den Hoek: Yeah. That's an interesting one because, so one of the challenges was just proper sewage treatment, before this war that was just the sewage treatment. There was dispersal into the Mediterranean. And so we had a little web app that we developed several years ago, which charted remote sensing proxies, indicators of pollution.
And this is, some of your podcast viewers are probably aware of the challenges of detecting pollution and water because, well, the pollution settles, and so you've gotta get certain particulate size. Certain kinds of things are detectable, but really only from a satellite perspective, they have to be like in the upper crust of the water, right?
It has to be in that first couple centimeters, otherwise, it's just the light rays won't penetrate any further, so you can't see it. But we had a little study that looked at chlorophyll as a proxy for sewage being disposed of into the Mediterranean. One of the reasons we looked at that is because one of the kind of unique counterintuitive scenario was that there was so much need for sewage treatment and water treatment more broadly in Gaza.
But as I mentioned, constraints on materials, it wasn't really until some of that sewage
the persistent sewage flowing up the coast into popular recreation tourism sites where beaches had to be shut down because of the public health concerns of going into the water with the sewage that materials and resources were allowed into Gaza to actually construct waste sewage a treatment plant or center to try to mitigate the effects of that.
So it was a very sort of, paradoxical situation where what led pre before this war, of course, but what led to that water treatment in Gaza, the sewage treatment, or rather in Gaza, was the public health impacts of people upstream. So that was interesting to sort of, retrospectively to think about how that played out.
Now with the war happening in those sewage treatment plants, yes, being destroyed, but there's also just the sewage is not being distributed into the Mediterranean anymore. So if you naively, right, look at these satellite image maps of water quality, name your name, your indicator of the Mediterranean off the Gazan coast, it will look much better now than it did then.
And similarly, there's suggestions of, well, some of the fish might be coming back. There's greater density of fish and availability and of course it's illegal for Gazan fishermen to go out to fish them. So yeah, these are some of the challenges of trying to study things by proxy, right?
If you don't know the social context, you don't know the conditions, you'll come to the wrong conclusions about what's causing it.
[00:27:25] Bridget Scanlon: Right. And guess, was that beach area Ash or something like that? Was that the name of the.
Yeah. Right. And you do a lot of work on food production and food security. And I was looking at some maps saying, deforestation, you mentioned that, so they said 2021, a third of cultivated area had forest. And then by late September, 70% of them had been damaged.
So I'm looking at images between '21 and '24 and a huge reduction in forest cover. But what about food production for people in Gaza? I think before the war they were pipelines bringing water in from Israel. I'm not sure if they discontinued those.
Do they have tankers and stuff going in? And then food, is there much food production at all in Gaza or-
[00:28:14] Jamon Van Den Hoek: Before the war, Gaza's economy was mainly agriculture actually. There's olive trees and citrus production, so, the economy as it was, as constrained as, it was primarily agricultural. I mean, tourism for example doesn't exist 'cause you can't, no one can get in and no one could get out.
But it was broadly agricultural. Under the war, during the war, then, of course all of those institutions that have been in place, all the markets that have been in place that supported the production self-sufficiency of food production, for example, all of that's collapsed.
And so it's very much a dire situation right now where the food needs, they're all met through external aid. I mean, I'm sure it's not literally a hundred percent but by far the majority is external food aid. And that is then we end up having these scenarios of humanitarian convoy trucks, aid trucks being stopped at the border or there are some cases of aid trucks being, the goods being taken off before they could get into Gaza.
it's completely dependent on external humanitarian aid at this point. Yeah.
[00:29:19] Bridget Scanlon: it's hard to imagine how they will ever be able to recover.
[00:29:23] Jamon Van Den Hoek: Yeah, it is. It's tough to see. I mean, there've been several plans put forward by different actors, different think tanks, the White House, the administration had one plan they're putting forward. RAND, which is a research group also put forward another, I mean, there's many plans out there.
Some coming from the US, some coming from Middle Eastern states. And across all of these not only the question of which one of these is actually going to be the plan to win, but who is going to be able to make that decision? Whose voice is being heard with deciding what the future of Gaza, whatever rebuilding has to happen?
It may not be in view of Gazan's needs. People have used the analogy of the modern day Marshall Plan. I mean, that could be one approach of like direct foreign investment. But the Marshall Plan was the reconstruction of certain areas for the people that were still there and who, who had a say to some extent over how those resources would be allocated.
It's difficult to, we don't know right now how that's going to happen in Gaza. It's a very I mean, just personally for me, it's an unfathomable thing to even try to think about what it is like to be consistently displaced during the course of a war. That is now about 18 months in to be displaced over and over and over and over and over and over.
And to have in certain areas of Gaza over 70% of the buildings damaged even by our estimates and others, the regents, the administrative regents, the governor of Gaza. There's not one that's less than 50% destroyed. I mean, every other building. And certainly in some neighborhoods there's just, there's nothing.
There's no building standing. That's certainly true. So reconstruction people are living amongst rebels right now. People have been living amongst rebels for a long time. The thought of reconstruction is certainly necessary but I think that's really the concern, perhaps why it's at a stalemate, it's just, it's grinded to a halt is because this next step is so important and so many people have a real investment in seeing the Gaza reconstruction going one way or the other or another way. And it's, yeah I'm not the expert on this but.
[00:31:36] Bridget Scanlon: Have you looked at the imagery at all from the previous conflicts, like 2021 or Have you seen any recovery from those? I know they weren't as,
have you looked at any of that imagery to see how much damage and how much recovery and how rapid was the recovery?
I mean, was it, just took a long time or?
[00:31:54] Jamon Van Den Hoek: Yeah, that's a really good question. A lot of the 2014 that was primarily that was like agriculture productive regions that were destroyed or damaged I should say, that did recover. I mean those areas that were damaged then it's not as if those places were just landscapes.
They were put back into production in the years following, in some cases in next season. The scale of damage now, of course is unprecedented in human history. In terms of that, you would have to find very few analogs of a single city being damaged more than all of Gaza Strip is. Like even Warsaw, after it was raised by the Germans, like that was 88, 90% damaged.
But now we're seeing that's one city as large as it is, that's still just one city. And of course Warsaw is reoccupied. People are still, people lived through that. Think about Nagasaki and Hiroshima, like those are active cities today that even two years after the nuclear bomb and Hiroshima there were tourists pamphlets saying, come to Hiroshima, right?
Like, trying to get this economy coming back. But the scale of damage in both of those places it's kind of jarring to even think about that, but like the scale of damage it's just so much more widespread in Gaza compared to that. So yeah the sort of the thought of recovery is certainly necessary and we can look to these historical analogs in Gaza and outside of it.
But even with the 2021, the border between Gaza and Israel at that time, there was basically a clearance area buffer that was created. And so the loss of land from the, this sort of, what is considered by the Israeli military as sort of security buffer around this fenced area,
that's where about a third of the land that was allocated for that security buffer was about a third of Gaza's most active agriculture at the time. So it was just, it was lost. I mean, the land is still there, it's just used differently now. And it's not for food production, so, yeah that never recovered, right?
I mean, it's like you have a farm land. This is the tale of Western expansion in the United States as well, right. All this amazing, prairie, all this amazingly productive land covered up by parking lots. We, we see that sort of development story all the time of ecological loss. In this case.
it's just a extremely acute, extremely acute case tough to ever see sort of rebounding, recovering right after such dramatic-
[00:34:15] Bridget Scanlon: Yeah. So I guess shifting to Ukraine, I know you, you were tracking the Ukraine war and its impacts and stuff like that, but not doing a ton of work on it. I guess you didn't make a career out of it. But you were tracking it. And so, were you also using Sentinel one to look at the Ukraine war?
Were you using other tools or, to track what was happening?
[00:34:41] Jamon Van Den Hoek: Very similar approach as what Corey Cher and I did in Gaza. Ukraine is much larger as we talked about earlier, so we didn't have that real time response. It was all retrospective and there we were making monthly assessments of damage across Ukraine. But it is very similar. I mean, that was sort of not the exact start of this work but one of the earlier kind of full blown trials of nationwide damage mapping.
And Ukraine also just another good collaboration with journalists in that case. Some of the initial conversations even around doing that were with the New York Times around trying to develop a visual story of the scale of damage. And that was also extremely helpful. Working with Tim Wallace and Marco Hernandez and others there, they just added so much to the reporting, to the context, to the visualizations were amazing that they produced.
And yeah, very, we just, we learned a lot through that collaboration. And it really improved our approach with the feedback and the discussions we had with them
[00:35:39] Bridget Scanlon: And so, were you looking at the destruction of the dam and the flooding and impacts on agriculture and things like that. Was that some of the stuff that you were looking, evaluating with the remote sensing?
[00:35:51] Jamon Van Den Hoek: For our work we just looked at urban damage again. Other groups have looked at agricultural damage across Ukraine and really some of the extremely difficult challenges of identifying agricultural damage in war settings in a very agriculturally productive region. The Kakhovka Dam collapse, people have assessed that from a couple of different methods as well.
But we didn't directly engage with those. We indirectly looked at those things 'cause we had to be sensitive to what those changes looked like in the signals that we were getting. But we didn't really dig into those nearly as much as other groups have. There've been several papers now on agricultural loss across Ukraine, like the risk of agricultural damage due to war.
Some really exceptional work from University of Maryland on unexploded ordinance estimations in Ukraine using very high resolution data. Just some extremely creative work by Ukrainian scholars as well who have had sort of different framings, different approaches, just asking different questions that were more regionally specific or appropriate.
So, really cool to see. And though, just on that note, I mean, that's something we're seeing increasingly over time is the perspectives that are brought from researchers and scientists and community members in these conflict affected communities, it's so much sharper and clearer than what could be generated just sort of remotely thinking about what's going on.
So, we haven't talked about this, but the Tigray Civil War in Northern Ethiopia is a another extremely dire situation, but it was all the same. Kind of a unique characteristic of that conflict was the research that was undertaken by Ethiopians, some of which were in the diaspora community around the world.
To study the agricultural impacts, the ecosystem, the biodiversity loss, as well as some of the topics that tend to be more commonly studied around war, that the sort of casualty rate, the public health impacts. All that work was ongoing, but the thinking about things you can better see from space, right?
It's landscapes, it's visible change to the surface of the earth. And it's all part of the same story, but we saw in Ethiopia most of the research that came out the last two years is Ukraine and Ethiopian conflict assessments, and a good share of the authors doing that work are Ukrainian and Ethiopian.
And then I think that's just a testament to the value of open data as well. Open science data. Sort of at the core of these missions that the European Space Agency and others that it's just putting those open data front and center, making it accessible to people around the world.
And we're very interested in trying to support those efforts for whatever scenario comes up next with the conflict affected community we want to support their engagement, their contribution, and their participation to make sure that whatever work is done is, is done in the most sensitive way that responds to local needs and context and concerns and all that.
[00:38:37] Bridget Scanlon: But it also, I guess it brings their understanding of the system and everything, and a lot of that is required, it's not a black box. It's just that, to interpret the data you need an understanding of what it might represent. So that's great that those people are engaged in doing that and bringing more intel to the process.
[00:38:59] Jamon Van Den Hoek: Yeah. Yeah, the regional expertise, that's not in a textbook, that's not on a Wikipedia page. That's just regional experts who know, who do have encyclopedic knowledge about the agricultural system or the water system or the geology, right? There's so many facets of this. When you get down to it, our look at building damage is something that, it's very apparent from certain remote sensing systems, but it's a totally different question to think about how people make decisions about agricultural damage, about migration decisions like these are things that you just, you can't see, you can't know. You don't have the information to even contextualize these decisions from space.
Sometimes it really does come from that regional experience of practice and exposure to these different kind of land use systems. So it's, it's irreplaceable.
[00:39:48] Bridget Scanlon: And so when you were looking at Ukraine in a retrospective mode, I mean, were you looking at some of the cities near the front line and farther away? And were you looking at the cities, I mean, either side of the front line or, what sort of patterns did you see, or how much damage did you see to buildings in cities that you looked at in detail or.
[00:40:10] Jamon Van Den Hoek: Yeah it was a nationwide assessment, but most of the damage in Ukraine tends to be in the east, along the frontline area. So the bulk of it, again, this was some of collaboration with Corey Scher, but lot of the work, a lot of the damage rather that was detected was very close to the frontline.
Not surprising there, but a very different pattern than what we saw in Gaza, for example. Some of the cities that had the greatest damage, like Bakhmut, I mean, those were in any sort of assessment, relative comparison you make, like those were extremely damaged cities. There were areas with very high infrastructure, rates of infrastructure, activity that were very broadly damaged.
That kind of pattern that we were seeing there is not unlike what we would see in any other, like Lebanon or Gaza, of these same kinds of techniques. But Ukraine was, it had a different analytical approach in part because there was a frontline to map. And not every, and that's mapped by other people in geospatial data from the Institute for the Study of War, that based on a whole bunch of different reports they make dynamic updateable, frontline maps.
And there, that really, when you start contextualizing the damage from satellites to the movement of the frontline, you see these very strong relationships, as the frontline moves through a city, you see this, evident, very clear change in the radar signal. This is one of the things that Corey Cher pulled out from as he was preparing this work that was really compelling.
Basically a time series of the frontline moving through and seeing the damage that resulted.
[00:41:43] Bridget Scanlon: And, you mentioned earlier, and I meant to ask you, you said that, that some people can track unexploded ordinances from satellite data. What data, what satellites are they using to do that?
[00:41:53] Jamon Van Den Hoek: Yeah. This has been worked on by Eric Duncan, specifically at University of Maryland. He was using high, very high resolution imagery, worldview. So this would be half meter, one meter resolution data. And interestingly using at the time, at least the work that was published on this initially was using a deep learning approach that came from a different kind of remote sensing and looked at individual trying to find trees.
Look for this small isolated tree in this detailed image. Well, that's not too different of a challenge from looking at a crater. From above, there's a lot of similarities between a tree crown and a crater against an agricultural or vegetated mosaic, or even bare earth. So it was very cool work where they were able to adapt that mapping approach and using very similar data to pull out thousands and thousands of craters across the country. And some of those I'm sure were associated with exploded ordinance, but the real insight there is it's a very, a dark consistently toned, roughly circular, elliptical shape in an agricultural field without obvious burn scars. Well, that's a warhead that has not exploded, but has embedded itself in the ground.
And so that tracking unexploded ordinance were there had been some precedent for that, but no one had done that at that scale. So that was another really unique part of the story that's tough to get that systematic awareness that's, it's impossible to do without satellite imagery or other kind of, you could do it with any imagery, but it has to be that, broad scale assessment,
so.
[00:43:22] Bridget Scanlon: Right. I know we're jumping around a lot. We've been to Gaza and Ukraine and you've done a lot of work in Uganda and it's really impressive at the refugee settlements. And maybe you can describe that a little.
[00:43:35] Jamon Van Den Hoek: Sure, we have a project, we're a couple years into this now , it's a NASA land cover land use change project. This is with Catherine Nakalembe at University of Maryland, and Jen Alix Garcia are the other co-investigators on this. This project tries to basically take stock of the agricultural dynamics that have resulted or been certainly followed refugee arrival, large scale refugee arrival in Northern Uganda.
Broadly these are Sudanese and South Sudanese refugees, but there's several other refugee nationalities in Uganda. And Uganda has a unique refugee policy where refugees have access to land. Land for gardening, land for building a home, or for managing a home, I should say, as well as land outside of the refugee settlement for managing land, for producing food, for selling it if you want.
There's a lot of freedom in these places. The challenge here though is really assessing the impacts, the value, the benefits of that particularly with regard to food security of refugee families. And because of the challenges of mapping this, these agricultural plots many, many settlements, very large area, but also very small plots.
It's tough to track this from space. And so there hasn't been, as well as from the ground because it's just, it's so broad, there hasn't been a comprehensive assessment of agricultural dynamics in refugee regions. So that's what our project seeks to do. Using several different remote sensing methods, some of the same kind of sensors we've talked about.
Sentinel two is really a, a go-to, but, also high resolution imagery for different kinds of applications. Tracking agricultural dynamics, looking at plot, agricultural plot changes over time. Bringing in market dynamics food market dynamics, looking at food price data, pulling all these different lines of data together to try to understand what basically the impact of refugee arrival on agricultural production.
Refugees grow food, but also non refugees grow food to then meet the demand of the newly arrived population. It's a just, it's a market of supply demand to some extent. But there's also external food aid that comes in, there's self production, there's aid, there's a lot of different things coming into that food security story.
And the, and a lot of that's tracked well, but the actual agriculture land itself has not been so that's the main contribution of the work is to try to get a better handle on that. And really take stock of some of the decision making processes that refugees have given that there are many options, self production, food aid, selling it to market, buying food from the market.
There's choices and some of that's known, but that critical component of self production is not.
[00:46:14] Bridget Scanlon: So, I mean, you mentioned that Uganda has very favorable regulations related to refugees and allowing them to buy properties and have some land to grow food and things like that. I think when we chatted before you mentioned does Kenya have similar approach to refugees.
So they allow them to buy land and grow some food and things like that. So, Kakuma refugee camp and that sort of thing. So, so that's great because I mean, Uganda ranks like fifth in terms of hosting refugees. So it has a lot of refugees,1.7 million I think I was seeing in some sites.
So, I talked to some other people about providing water to these refugee camps, BidiBidi camp and Rhino and others there. So, when you look at these data then from satellites, do you need data to train, to interpret? And do you have any ground-based data that you can , can you identify what type of crops , they're growing and do you compare with food and agricultural organization data or other data?
Then to kind of, check what you are looking at, how do you validate the data?
[00:47:25] Jamon Van Den Hoek: Yeah. We do in some areas we have crop type data. There's existing databases. Crop harvest is one that has crop type data in Uganda, some other places that we're using for this project and actually adding data on crop type presence back into that database. We, for validation, we have a field project that's spinning up this fall.
So that is going to be our main, we've basically been doing everything remotely thus far, but we'll be in the, at different refugee field sites this fall, collecting data, working with local partners to try to validate presence of agriculture, but also the type and do some work as well on seeing how well we can detect not just the presence and type, but also delineating agricultural boundaries.
Just purely through, automated remote sensing methods that we can say how close it looks, how good it looks if we look at other data that's very detailed, and then do like a hand drawn estimate. But it'll be a very different and a much more rigorous comparison when we are actually on the site and can collect those data in the field.
[00:48:26] Bridget Scanlon: That's fantastic. And so I was trying to recall, Paul Bowman is the person who works with the refugees in Uganda, helping them explore for groundwater, drill wells, and install pumps and all of that sort of thing. He's, he's just an amazing person and he's trained a lot of the people there locally to do this, and then they train others.
It's just an amazing program.
And the thing is that these refugee camps, I mean, the people can be there for decades, And the numbers of people in the different refugee camps, they grew up and down over time. And maybe you can describe that a little bit Jamon
[00:49:04] Jamon Van Den Hoek: Yeah. It's, I mean, it's a common assumption that refugees, it's a short-lived phenomenon and that people that are displaced return home, but it's actually, it's not. It's the, it depends on what year you're looking at this, but it's fair to say that the average refugee displacement is, probably over a decade.
And of course, when families are displaced, children are born into refugee settlements as well. So, a term, which is protracted refugee scenario. And there are some estimates that up to three quarters of the global refugee population live in protracted refugee scenarios, which is over 20 years. So it's a very long term challenge.
I mean, the support systems that need to be in place, the opportunities for education and healthcare and developing of livelihoods for refugee communities. Sometimes in, to be frank, like hostile, like countries that are hostile to refugee presence. Not every country is as welcoming as Uganda is or has the means to support like Uganda has.
So it's extremely dynamic, extremely diverse. There's no single refugee experience far from a monolith. It's highly variable over time. And also country by country. But to your point on the duration, some of these refugee settlements in Uganda most of them opened up in about 10 years ago now with a large Sudanese and South Sudanese refugee influx.
But there are camps going back to the sixties and eighties and or I should say refugee settlements going back to the sixties and eighties in Uganda as well. Very long. They haven't been permanently occupied that whole time, but they were opened and then they closed when the need wasn't there, and then they reopened.
And so even these sort of timelines get very complicated. Refugee camp establishment is not the same as a refugee camp opening, so you can have multiple openings and very complicated.
[00:50:52] Bridget Scanlon: One of your papers I really enjoyed, you compare different remote sensing products to evaluate the global human settlement built up, Sentinel Two high resolution, settlement layer and other products, and you work with them then to improve. Maybe you can describe that work a little bit.
And I guess also you looked at Open Street and Microsoft and Google building footprints and using AI and machine learning to harmonize a lot of different data sets.
[00:51:22] Jamon Van Den Hoek: Yeah, this initially was worked with Hannah Friedrich when she was a graduate student here, and this work really came at that question of we talked about this earlier, but data quality coverage of different footprints was central to this question. We were looking actually not so much at footprints, but actually just built up land.
Does this estimation of land from these different satellite derived products how well does it match what we think, what our best guess is of what's actually there? Refugee settlements are tough to track from space because they're so small. You have individual settlement structures,
it doesn't look like the cities that are often historically it's changing, but historically that would've been used to train up, algorithm to find a human settlement, a built up training it on Paris or Shanghai or Nairobi. It's just, it doesn't look like that in refugee settlements.
With few exceptions. So Uganda, it was a case study because we have really good data on where the refugee boundaries are, the settlement boundaries are. We know about their timing, we know about the population estimates, so we have a lot of data to sort of flesh out the picture of the settlement dynamics.
And we looked at four or five different data sets to see how well did they capture that presence of the built up environment. We're not trying to estimate population or agriculture. This was just do they even see these settlements? Do they see where people live? And at that time, they didn't do very well.
I think the best estimate was between 40 and 50%. Most of them really missed the majority. So, and that's important because I mentioned, we're not mapping population, but guess what, like, that's how you do map population. You make estimates for population from space. You figure out, well, where are these structures?
Where are the dwellings in the first place? And then you, that's one of the top-down methods that's used. So find where people live first and then kind of allocate people to those pixels, right? That's one way to do it. And so if you don't see where people live in the first place, there's not going to be a pixel there.
And so you won't have people there. And that looks like there's no one there in the maps that are used by government agencies in the UN and sustainable old development goal indicators. So it had a real potential for cascading bias. If you can't see these houses, you can't see these dwellings, then all these other things get lost too, and has all these policy implications, these blind spots of refugee presence and needs.
So that was really the motivation of this, of the paper. And what was really cool about that is the Global Human Settlement Team we were in touch with them through other means, but, I don't remember if they were aware of the work that we shared it with them, but we had really, they invited us to, to collaborate on, they were like, we're releasing a new data set.
We've really tried to improve this. And so we supported an analysis basically a refresh of what we did in that paper, but globally over I think it was about 450 refugee settlements where we had the same thing. We had the, we needed the data on the boundaries, we needed the boundaries.
And then we looked at building footprint, locations, or presence, I should say, within those boundaries. And then let's see how the new global human settlement layer does. And it was much better. It really improved, and it's still the best from this, subset of the products we looked at and,
it's not, it's far from perfect. But it was great to see the commitment of that team to obviously it wasn't a critique, it was just a raising of the issue. Like, oh, we need to do better. And they were like, we did better. That was very cool to see that. And of course, these are specialists who,
if you look at the products, you're like, this looks amazing. This looks flawless. And then you talk to the people who know about all the limitations of their work, right? Like they know how difficult it is to map structures. They, especially in informal settlements, sparsely settled areas with 10, 20 meter data.
It's really tough. It was very just refreshing, to be like, if anyone, thinks that they know how to map population from space, like, to some extent, but it's not nearly as good as people might want you to think or mapping where people live from space, it's in certain areas, it works very well.
But in a lot of areas it's really difficult and we have good guesses in a lot of places. But there's of course, tons of groups working on improving this throughout all sorts of different means. So it's just great to see the attention to these communities that are broadly underserved, under mapped.
And to have that sort of front and center was a really, a great result of that work.
[00:55:41] Bridget Scanlon: And I think when we chatted previously, you mentioned that the UNHCR does a good job of, they, they need to be registered and everything to get the populations in these refugee camps. And they report that all the time. And another thing that you mentioned is that the sustainable development goals are not tracked in refugee camps.
And so you try to evaluate a number of them with some of this mapping. Maybe you can describe that a little bit.
[00:56:10] Jamon Van Den Hoek: Yeah. Very same study areas as those Ugandans, same Ugandan settlement study areas as the previous assessment. It came from that kind of core realization that most of these refugee populations around the world, there are some exceptions, but broadly they're not included in sustainable development goal metric indicators.
Name your indicator even if you have very large refugee population, they're not in there, they're not represented. So this work in this case, in some cases you only have Open Street Map data. Open Street Map data is such a backbone of this geospatial infrastructure globally.
It has, if there's any data, in many cases it's going to be open street map data. Maybe it's out of date, maybe it's incomplete, but it's something. And in humanitarian context there's usually no data. So that's why OSM Open Street map data are so helpful. That's what we looked at in the study.
OSM is extremely dynamic. It's user, it's a, a wiki of sorts. People contribute, refine, improve over time. So it really is just a snapshot of where things are at any given time. And this study is a few years old now, so, things have surely changed. But same kind of orientation where we were saying, well, how good are these data if we want to look at indicators?
And we looked at the data that were available, we said, well, these, of the different kind of open street map data, these are relevant for these kinds of indicators, and we weren't able to track every indicator but a good number of them. And we looked at the timeliness, we looked at the coverage.
We looked at kind of, if you just worked with OSM data, what story would you see from these indicators? And part of that was helpful to see, right? The positive signs the kinds of signs you'd want to see with indicator assessments. But I think the real takeaway there was just that the challenges, again, it's not saying that you can't do this with OSM, it's all about just understand what you're working with.
It's clear communications of limitations and benefits of data, which is really just a pervasive theme. I probably used the word limitations 20 times in our talk here. It's just something we do all the time. It's just like, how, what can you even say with this? What are you confident saying? How have you proven it?
How have you aligned it and triangulated it with other data? Have you tried to show that it's not good, before you can say that it is good. And that, that's, that was one of those studies that really helped align some of our thinking around OSM. A lot of people have done very good OSM coverage checks around the world in humanitarian settings and outside looked at who contributes to OSM for what purposes, how is it used?
So it's been really cool to see as valuable as OSM has been, all these external studies that really just show how much it's used the diversity of applications around the world and the really, just the hundreds of hours that people have put into improving the database as well. It's a really special I mean data set is sort of almost diminishes what it is.
I mean, it's a whole archive, of just, of spatial data, of all different sorts.
[00:59:00] Bridget Scanlon: And did you find that certain sustainable development goals were much worse than others when you did this analysis or,
[00:59:07] Jamon Van Den Hoek: Yeah, I'd have to actually look at the paper, what we found is a few years old. But, I do recall that there are some that, especially around some of the more social justice indicators, very difficult to track, right? Like the social settings, others that had to do with kind of more simplistic indicators of just like presence, absence drinking water quality life on land.
Like a lot of those were there, right? There were hints of those, but some of these indicators were just totally absent. And in no way I would never have propose, like just do an SDG indicator assessment based on OSM. But we were saying in the absence of all other data, what could you do? The real thing that came out of it actually was the timing.
We really struggled with having a consistent timestamp on all the data. So some of these assessments were done very early into the refugee. Basically the settlement was established and then there's OSM data, but then nothing after that. And so the idea of change over time, there's no change. It was a single snapshot, like eight years before the study.
So you just don't have any confidence in it, right? It's just, it's too out of date, and that's a consideration of accuracy that's often not there. It's taken for granted with satellite imagery that you'll have a new image. But when you work with these crowdsource data, knowing the timestamp that you have to know that to, to know if it's even relevant for your question.
[01:00:28] Bridget Scanlon: Right, right. Well, thank you so much for describing all of these different programs the, the Gaza work, the Ukraine work, and all the stuff that you're doing with the refugees. And I hope you will have a great visit to Uganda, we'll really learn a lot from that. Our guest today is Jamon Van Den Hoek, and he is associate professor of geography at Oregon State University and leading the Conflict Ecology Lab.
Thanks a lot, Jamon.
[01:00:55] Jamon Van Den Hoek: Thank you so much, Bridget for the far ranging conversation, really nice to talk with you.
