"Monitoring the environment from Space" with Lisa Broekhuizen

You're listening to a stage talk titled Monitoring the Environment from Space. This week we were joined by the technical director from Space for Good as they walked us through the possible practical applications for satellite imagery when it comes to environmental monitoring, explaining how and why satellite images may be the best way to oversee large -scale rapid impacts on our climate and wildlife. This talk was hosted by my colleague Giancarlo Fiorella on Thursday the 10th of April,

2025 in the Bellingcat Discord server. Welcome everybody. My name is Giancarlo Fiorella and I'm the Director of Research and Training here at Bellingcat. I am filling in for Charlie this week. who is away at a conference. I'm really excited to introduce you to Lisa Bruckhausen. Lisa is the tech lead at Space for Good and is a remote sensing specialist. Lisa and her team use satellite imagery and remote sensing techniques to monitor changes in the environment from space.

Now, previously we'd spoken about a tool from a Bellingcat contributor named Chris Giles. And the name of the tool, you might remember, was the OSINT Forest Area Tracker. This is a tool that looks at burned areas of forest in Ukraine. And the purpose of the tool is to give users the ability to look further afield about forest fires in this part of the world. Now, that tool uses something called Sentinel -2 data. You can

find it on our GitHub. If you go to GitHub and you look for Bellingcat, this is one of the many wonderful tools you'll find there. Space for Good have just come out with their own application to help people on the ground monitor deforestation from space. And Lisa is here to talk about that application and more of her work at Space for Good. Now, if you'd like to ask a question, please do so in the chat as we talk. I'll be writing down your questions and then asking them to Lisa

towards the end of the talk. If you don't want me to read your username when I ask your question to Lisa, please make a note of that in the chat. Lisa, thank you so much for being here this afternoon in the Netherlands. I'm going to hand over the microphone to you now. Thanks very much for having me. Thanks everyone for being here, for listening in and a special thanks for Giancarlo and Charlie for setting this up and of course everyone here

for all the good work. So I'm here today to share some of the work that we are doing and share some ideas and some inspiration on hopefully open some eyes on how a lot more can be done with the eyes we have in the sky, the satellites. I'm Lisa. I'm technical director of Space for Good and today I will focus mostly on the practical use cases with satellites. So I'll talk a bit about that, like I do every day and hopefully

after this so will all of you. At Space for Good we work with satellite data to monitor the earth. We do this for projects we believe have a positive impact on the planet. I'll talk through some of the work we've done, some big developments and some smaller ones which nevertheless I think have a good impact. and hopefully share some ideas and inspirations on what's possible out there. During this session I will not get technical, but rather provide with a satellite overview.

However, you're more than welcome to ask technical questions at the end, hoping for those. And when I say monitor the planet with satellites, I don't just mean working with pictures from space. I mean uncovering the stories. Environmental stories, land use stories, sometimes very personal stories. where the start of the story starts with the data. So first of all I didn't want to talk about Forester, our product to help detect illegal

logging near real time in the field. It's a tool to give rapid alerts and really help Foresters on the ground take action rapidly rather than having long -term visuals over trends or regional actions. We rolled this out first of all in Indonesia. We're working with local partners to take direct action. Sometimes we use Sentinel -2 and need the optical data. For this, we use Sentinel -1, which is a radar data set, which means we can penetrate the clouds. Now, if we look at tropical

rainforest, then rain kind of in there. There's a lot of precipitation, a lot of cloud cover. That means that with optical data, we often have just see the clouds. With radar data, we can look through the clouds and see changes that are happening. It's a bit less intuitive to read the data, but that's what we have algorithms for. So we look for changes, rapid changes, within

areas that we before classified as forest. Within those areas, we look whether it's likely illegal logging or a legal or natural event such as a landslide. If we think it's illegal logging, we send the WhatsApp alert out to people in the fields directly. They go and they take action. So what we've seen before is that they come up with security guards to show the people that they have been spotted and to get the people

out. Sometimes they use the reports we generate to go to local authorities and get the police involved. And with that, they can remove the people before a big conversion to agriculture happens, which means that the forest can recover faster and better. We now have a system that's over 98 % accurate, so that's really nice. In

this area at least. In other areas we always need to retrain and revalidate because it's not a globally applicable model, but it's a model that really needs to be trained locally because the patterns and the nature is different in different locations. A bit about the stories. So one of the things we hear from our clients is that they can finally detect deforestation that happens away from roads. Imagine here it's a tropical rainforest condition. It's about 200 ,000 hectares

of area to monitor. Roads tend to be washed away with rainfall or any other events and suddenly the road is gone. And even if the roads are perfect, yeah, it's a rainforest. You can't look into the forest and see very far. So you can only see a few meters. So before it was almost impossible to see these events because they happen outside of the direct road network, outside of direct fuel. And one more thing that we've seen, which I really like talking about. I don't like the

story behind it. But we detect a full satellite view of the whole area, which means that we detect all deforestation events. What they noticed is with that data that around one security guard was positioned at different security posts. Whenever he was positioned, there would be a peak in illegal logging. So likely he let some family members through illegally but that's very difficult to

assess if you don't have the full data. So by having that full map it could pinpoint hey here we need to do some more research to identify what's exactly going on. Quite a nice case to see like how this information we never thought something to be pinned down to an individual but that actually was possible with the help of the data and of course merging it with all

sorts of other data sets. Now what we do here differs quite a bit from EUDR compliance because we are focused on near real -time direct action and I am assuming directly that almost everyone knows what EUDR is but some of you might not hear about it as much as I do so let me quickly expand EUDR is the European regulation for deforestation free supply chain starting next year probably you never know with EU it's intended to ensure that agricultural products entering the EU are

not causing deforestation so that you can eat your bar of chocolate guilt free. Deforestation wise I'm not talking about fair pricing or human rights or calories today, but that it doesn't cause deforestation. EODR has a single definition of forest deforestation land cover where we rather look at trees disappearing. So we don't look at is a plot of forest converted to agriculture, but we look at there were trees here, there are

none now, take action. rather than saying there were trees here, there's agriculture, two years later, so you can't import. And then we really focus on enabling people in the field to take direct action. That's a bit about forest application, but deforestation, but it's not all bad. There's also afforestation, the growth of new forest, where new forests are being planted. So, nice

story here. without naming specific names I'd like to come into a little case study we did where we got a request from a small tree planting NGO to compare how some of their different sites have been doing they used to do a lot of field work but with environmental more and more information about the flying but also safety concerns for quite a long period they couldn't go into their afforestation areas anymore so we did some analytics for them to compare how different sites were

doing, which one was healthy, which one was managed well etc etc and in some sites we found that there were no trees left so the local partners had been receiving money to take care of the trees but had failed to report the death of several full plots so we're gonna go by mistakes happen but thanks to this data the project could be changed around gain additional management and it was made clear to the people locally that there were checkups being done which is, first

is good, but validation is better. And of course it's bad that those trees didn't make it, but with satellite data, they could make early interventions and changes into the process. Now in this project, we also made some nice analytics, some quite nice algorithms we developed to compare the different sites, the missing trees. I would like to say that there was this amazing algorithm in -depth assessment. But basically it was looking at a picture and going to a colleague and, do you

see trees? I don't see trees. And having figured out that there was a major issue here. Satellite data analytics can be rocket science, but often it's not. And then having that layer of extra accountability, especially in an industry where so much greenwashing is happening, can give that additional trust. And not having that layer... and lead to more scrutiny which at least leads to us all having something to do so that's also

a win right? Looking at the current Belenka challenges there's also I also want to quickly point out some other things everything that has a location always grab a satellite image we did a project around illegal zoos small project there was no budget available so we had no commercial data but even looking at 10 meter resolution openly available satellite data It's the Sentinel -2 program that's looking at only true color images.

We could see the expansion of some of these wildlife breeding facilities as an extra source of information to help with the investigation of that party there. And animals or meat directly leads to biodiversity. We see a spike in biodiversity credits. where people are paying to support biodiversity within different areas, within different conservation sites or new nature developments. What is missing in a lot of cases is objective data. There is quite a lot we can see with satellites that relates

to biodiversity. Very rarely you look at individual species, but often you get landscape features, diversity, landscape history, all those things can be mapped. That also brings us to an issue. If we can map multiple things, we can get different outcomes. And which outcome do you share when all of the different outcomes are correct, but also saying something different? I don't always have an answer to that, but I do know that the answer can't always be the most optimistic case

or the most profitable case. Even if we don't look at the complexities of biodiversities, but look into carbon credits, which is something measurable theoretically, you can measure a tree. We see billions flowing into the carbon credit market, but are the protected areas even under threat? And that's why satellites are very important to generate a baseline to see what has been happening, because we can look back until the 70s of the last century to see evidence -based what has

been going on. We can provide support whether a project is resulting in avoided deforestation or an aforestation, and identify good reference areas, or make strong claim why a reference area chosen, so an area where we compare the prediction site with chosen, is selected wrongly. And the net satellites can give a timeline rather than snapshots to tell the whole story instead of

part of the story. We see that a lot with land cover conversion, where land is changing from forest to crop land, from pasture to urban, and we can detect whether that change is cyclic. So if you have shifting cultivation, every year part of a forest is cleared to convert to agriculture. That's not always the same as a fool on deforestation because maybe a year later a different area is cleared and it's part of the system that's being used within the region. But often it's not. Often

it's just a cascade effect. Like we see in Indonesia, where we see in certain areas a pattern from primary forest. to the graded forest, to small water fields, to commercial palm oil. And that whole pattern can happen within a few years. Also, I want to quickly talk about invasive species monitoring. We do quite a lot of work in Indonesia. We now started just out the project to map invasive grasslands to support suitable afforestation

areas. For us, that's pretty much the same as looking at making a land cover map, saying what's there. It's a bit more complex if we look at very specific species or species group. Almost everything you can see with the naked eye, especially if you can look from above and see it. You can also model and map on large scale with satellites, but also develop triggers, codes that give a trigger if something happens. That'd be quite

nice to give warnings. And then I think my current favorite project to talk about is environmental risks. Imagine you spend a lot of time and resources to map a minefield in Ukraine. Imagine a dam breach or storm causing a flood. Mines get shifted with the water. They're relocated. You don't know where. You need to remap possibly a much larger area to figure out where the mines are

in order to be cleaning them. With that we're working with the Halo Trust to map environmental risk to support in mine clearance priority setting. So where is a fire likely to happen? or a flood, like a fire you can't exactly send a fire truck into a minefield if there's a forest fire. Or where is a drought likely to happen and what

communities will that affect? That's all information that can be helped in this case for better decision making but that's all information that can also be generated globally for different purposes. So really looking into the different risks based on the patterns that we've seen and based on the patterns that we expect in the future with for example climatic differences. And if a fire happens, where, how big, the faster the information can reach the right people, the better the action

can be taken. So I can really recommend everyone to, if you haven't already, to look at the NASA Global Database firms. It provides information on patterns and trends and historic fires. So that way we can map and also make predictions. We can give alerts, we can map recovery, always globally and everywhere available. and objective. I say objective, but in this time of deep fakes and fake data, how can we be sure that satellite data can be trusted? Or how can we know when

it can't? So far, large datasets have had very few large -scale issues, and I should be careful what I say with this audience, because the satellite data can feel inherently objective. It's from space. It looks like the truth. But as soon as it hits the ground or website, it can be framed, processed, and cherry -picked to give a particular story or a particular angle. And we can't pretend like the data is always going to be correct.

So how to make sure the data is not manipulated or misrepresented and how to make sure that you don't accidentally misrepresent the data. Quite simply put, there's a lot of different satellites out there, so don't rely on a single image. Look at the time series, look at what change happened. Is it a sudden or gradual change? Is it a seasonal change? For example, if you look at the deforestation events, you can make very impressive images about

European deforestation. by comparing summer and winter images, because the trees lose their leaves and everything looks dead. It's not. Guys, the forest survives. So really look at how does this thread, how is what we're seeing normally looking over time? And how is it looking with multiple sources? So there's a lot of different satellite constellations out there, both public and commercial. Satellites are a witness. and you don't rely on one witness when you've got a dozen out there.

We have, for example, very nicely Sentinel and Landsat, where Landsat is hosted by the NASA. Sentinel is from the European Space Agency. So different sources that can look at the same thing, different sensors, different resolutions, et cetera, et cetera, but they can see the same trends. So always have a look at, hey, can I validate this with a different data set? And then if you're truly concerned about the validity of the data, heading back to the original source.

So a lot of the data you can find in different places has been processed, which is freaking amazing because I can tell you atmospheric correction is really annoying to do. And usually you don't really need to do it yourself. You can quite easily get that done automatically. But if you have reason of doubt, it's good to check the original data source and go back into that. and to check the full data source. So we're all used to look at true color images, red, green, blue,

make a very nice picture. People working a lot with satellite data tend to work a lot with the near infrared band as well, make false color composites. Looks really nice. You can see a lot more changes. But it's a lot harder to fake 23 bands of sentinel imagery, including atmospheric effects, than it is to fake a single picture. You'll look back to that single image. that original data set and see, can you validate the whole

processing chain or not? And with false position, I think that's more of a misrepresentation usually than a true hacking into the data. If people make strong claims about data, then the models, then the evidence should back that up. If they make a... If they don't share any uncertainty estimates, that's not that relevant. I now share the accuracy of our model in Eurasia and I can tell you that's tested on one side and it's amazing

there. But I can also tell you that I can make a global deforestation detection algorithm that's 99 % accurate. I would do that. I would say that in the sea, there will never be deforestation. So that's already 60 % done. I will say that outside the forest, there will never be deforestation. So then we only have a small area to check. And actually saying there's less than 1%, there's

only a couple of percent of deforestation. That means that if I say there's zero deforestation globally, I have a global model that's going to be 99 % accurate. It's also going to be absolutely useless, but it's going to be accurate. So know when people stroll around accuracy metrics, what to ask for. I can share the accuracy metrics about our model as well if you want, by the way. Those are reliable. And ask those right questions. Or make sure you or someone you can depend on

can do the actual analysis. And know that the data can be hacked. Processing pipelines can be tampered with. Cloud platforms can be compromised. Scripts can be manipulated. very easily putting in an altered timestamp, we've all seen that happening, is already giving a completely different view of what the data represents. So that's why open codes, audit trails, their reviewed methods are important. So satellite data is a powerful witness, but it does need cross -examination.

It can be mayplated, it can be wrong, but it is objective. And with the right checks, it can also be a very strong ally in the truth. So to end a bit cheesy, the eyes in the sky are watching, but they're on our side if we know where to look. Where this forest lost ghost plantation or land grabbing satellite data can be an ally. So start exploring, ask your questions now or contact me later for more information. If you can't find me based on what I've just said then you shouldn't

be in this group. So thank you all for listening. Thanks very much, Lisa, for your presentation, for sharing your knowledge and your experience with us. Folks, if you've got questions for Lisa based on what she just said, please type them in the chat. I'm watching the chat. I will copy your question and ask it to Lisa later on. Lisa, I've got a couple of questions here for you just right off the bat as people get warmed up here

with their own questions. Can you tell us a bit about what got you personally interested in satellite data? What was your journey like into this field from maybe a young person in the university or high school to where you are today? How did you get interested in this and how did you end up working at Space for Good? Yeah, thanks. It's

been a while since I thought about that. In high school I wanted to really go into nature conservation, going into biology, but then I saw that biology was going to be very much looking into cells in most universities and looking into the cell processes and that was not what drew me. I really wanted to look into ecology and the bigger processes and see how those things work. So then I started to study forest and nature conservation in Wageningen.

I did my bachelor's there. I started my master and then I got a bit into GIS remote sensing and spatial modeling. And that was for me really what I was looking for. It's working with the computer, making some models, figuring out how things actually work and can we replicate that. Basically making major conservation into a game and seeing like, how can I alter that? And that's really drew me and then... slowly but surely I started to learn more about these satellites

and everything I could see. And basically with everything I did, I was like, okay, but where can I see, what can I see that's going on? And that really got me excited to every time. But the data says, yeah, I'm kind of nerdy. Great. Thanks for that answer. And as a kind of a follow -up, if you're a person, maybe you're listening to this live here in our Discord server, you're

listening to this later. and you are interested in getting into this kind of work or you're hearing for the first time, oh, wait a minute, there's these satellite imaging platforms and one of them is called Landsat and the other one is called Sentinel. If you're just learning about this now or you're becoming interested in this now, what are some recommendations that you have for somebody like that to learn more about this field and to maybe get their hands on some satellite

imagery and start looking at them? Yeah, so the very, very first step is Google Earth, where you have the time view as well. So you can look back in the past and you can start playing around a bit and get hopefully excited. Then I would recommend to work mostly with Sentinel, Sentinel -2, which is optical data. It's a bit easy to work with and radar data, for example. And it's 10 meter resolution, so you can see quite a lot

of different things. You can download it either via one of the DS servers or Google Earth engine. Google Earth Engine is still freely available for certain use cases. So that's a nice win. Unfortunately for a lot, not anymore, which is too bad. But then Google Earth, you can, there's a lot of tutorials and a lot of different learnings out there to be looking at things. And yeah, start with seeing if you can map your own house

and different bands in different views. Look at any article you're reading like, okay, what is the latest image I can get? Get really, really frustrated by cloud cover and figure out how to deal with it. And then at one point think, okay, now I'm going to need radar data to complement this or now I'm going to need Landsat data to complement this. Have a look and play around

with it. It is coding based. You can also look at QGIS, Quantum GIS, which is a software you can install your computer and then just download the data and open it in there. But I recommend working coding based because you can all the things so much more easily. Thanks for all those tips there. By the way, I'm dropping the links to everything that Lisa's mentioning here in

our Discord server chat. So if you take a look at the chat there, I've dropped a link to Google Earth Pro, which, yes, is maybe the most beloved and used tool of the open source researcher, whether you're working primarily with satellite imagery or... You're doing geolocations for human rights investigations. The Sentinel Google Earth Engine, I've also dropped a link in the chat

there, and QGIS as well. If you're listening to this later and you don't have access to the chat, you just Google for those words that I just mentioned, and that Lisa just mentioned, and then you'll see links to these tools that you can download. We have a question from Fraser, community member Fraser. Hello Fraser, good to see you. Thanks for being here. Fraser has this question. Lisa, Fraser asks, have you had a look at the Leiden guidelines on digitally derived

evidence and satellite imagery? Yes, I had a look at it, but not recently enough to be answering this in this. I do know there's a couple of different works out there at the moment, also with CivicSpace looking to launch their own satellite constellation that is not from any... any government or any agency or any commercial entity but it's really looking at evidence -based blockchain validated

data. I've had a look at it and I definitely will have a look at it again and I recommend everyone to have a look at it and I will not go further into that. Yeah, oh great, that's an answer. Hey, that's a great answer. Thank you for that answer. Thanks for your question, Fraser. And by the way, if you have any questions, please type them in the chat here. I will ask them to Lisa. As we continue, we got about a half an hour left, so do not be shy. Lisa's an

expert in this field. You don't often have a chance to talk to somebody like Lisa, so do not be shy. Ask your questions in the chat, please. I got another question for you here, Lisa. The question is this. Towards the end there, you talked about deep fakes and whatever we want to call AI -generated images and fake data being out there. You mentioned... that there were, I think you said, very few large -scale issues

out there with imagery. I know in our own work at Bellingcat, we're often asked, can you think of examples of AI images in the wild that you've detected, that you've worked with? And certainly in our field, we've seen quite a few. So my question to you is, can you think of examples of AI -generated satellite imagery? Are they out there? Have you come across them? Are there people out there who are trying to trick others into looking at a satellite image and saying it shows something

but really it's generated by AI? Is that something that you've come across? And if not, is that something that you fear might happen soon? Yeah, I have not come across any big ones I come came across when doing some research for this to also find some an area where someone made a nice fake post about a fire in In a big park which was like obviously fake as in it looked nice, but it was also more made with Photoshop So then

there's no data behind it. I do think it's likely to be happening in the future if not now already, especially with high resolution data where there is active tasking. So there's often only one image or very few images to be altered basically. And then if you look at war zones, for example, there's very high stakes. So I can imagine it would be naive to say that it's never going to

happen. It's likely to happen. But I think with, especially in the case where you have open data, we have time series of the same data sets, it's

easy enough to disprove. oh wait that's not I think that are easy to disprove still get sometimes picked up by the crowd but it is at least easy to disprove so it is something to be conscious of but that's not yeah for the deforestation I don't see it as the direct biggest risk factor yeah thanks for that and you know it's interesting when you were going through the workflow or the suggestions on how to work with images satellite images in this time of fake images, of AI images

being out there. The steps that you outlined are the same, many of them are the same as we recommend to people who ask us about, hey, how do I know that an image from a protest is really a real image and not an image from, or how do I know that this picture from a war zone really is an actual picture from a war zone and not a fake picture? You know, you talked about looking for other images, right? So you called it the time series, like look at other pictures that

apparently show the same thing. Because I think I suspect that as with pictures from war zones or from protests or from anywhere that are generated by AI, the platforms can't, at least I'm not aware of this, can't replicate the same scene from different angles. But as you said, it's kind of naive to think it's not going to happen. I suppose for satellite imagery, something that'd be interesting to think about is satellite images

are... They show intricate patterns right like roads and houses and and and I wonder if the AI image generation technology is so good that it can actually accurately mimic those and I'm thinking specifically about the cover of the Berkeley protocol book from the Human Rights Center at Berkeley. The cover of that book is actually an AI -generated image of a satellite view. But if you take a close look at it, and an expert eye like yours could definitely tell,

hey, this is weird. I've looked at a million satellite images. None of them look like this. But as you say, it's one of those things you have to think about as being potentially possible in the next little while. There, indeed, my risk is there is especially with open satellite data, there is a lot of data available to train on. And a couple of years ago, the generation of deepfakes was very, very early on. And if you see how fast that goes, how good video material

we can get. I'm not sure why I'm here. I could have just sent an AI. How good that all gets. Then I think, yeah, the satellite data, there is a lot of data out there. is going to be feasible as well. The patterns at the moment, I don't think yet, but I'm talking to some people who are working on how to connect the roads, which is also very useful if you want to map a road network in an area. It's nice to be able to map those roads. But yeah, it can be used to generate

fakes as well. And if we talk about the wrong data, I see at the moment a bigger issue in misrepresented

data. having a data set and saying it's and that's something that we do see what people say claim it's in a different area and again you can say or a different time and you can say no wait i've got the evidence that it's not but if the information is out there coming with the evidence isn't always enough yeah thanks for that lisa i have another question here since we're talking about you know we're talking about what may be coming in the future in terms of ai image generation you've

just mentioned this ability to make realistic roads, for example. Are there any advances in remote sensing technology that you think are coming soon or in the next couple of years that you're really excited about? A lot, a lot. So what's happening right now is Google has just announced a better version of a system to be talking to maps, basically. So rather than doing all the investigation yourself, being able to just ask, hey, for this area of interest, how

much deforestation has there been? I think there's a big risk there, but I also think there's a lot of opportunities there to take that first step into how can we actually see what's going on on our planet? I think the roads, that's for me to make the system of automatically connecting different features. based on a lot of data, but also based on quite complex models. Having a detailed road network can mean so, so much for, for example, detecting illegal logging, but also

predicting it. We see that illegal logging almost always happens near roads and almost always first a new road is made and then logging happens and it expands. If you can see that earlier by having

just a few points. then you can really map out a lot so that's really interesting to me and yeah biodiversity making some sort of integrated solution how can we map biodiversity to well it's not really clear what biodiversity there's not one definition of biodiversity but to be able to say hey this is going well and this is not going well Thanks, Lisa. We have a question from Chris. And if you have a question, folks,

please type it in the chat. We've got about 20 minutes left here with Lisa from Space for Good. Chris is asking this question here, Lisa. Now, there's some jargon in here that I don't know. So I'm going to read it out as it is written. And maybe this is the jargon that you know. And the question is this. I'm curious to hear more about how you are ground -truthing satellite data to help detect distortions. So we ourselves are not earlier, but let me first like phrases

a bit for the rest. So ground rooting that's looking at what's on the ground and perceiving that as the truth and you can use that to retrain

models. I think here the distortion, the distortions, but Chris, chat if it's wrong, is really looking at the distortions of the images, which is why you're looking at, for example, something that we've seen in the past quite a lot, the earlier generations of planet data had a lot of misalignment of different bands now i'm getting a bit technical here so that's when you have a picture of the red green and blue and they're all not exactly aligned so you could have a situation where the

red value of a certain pixel was 10 meters below blue value so you would have a shift there with ground rooting you can basically align pictures so they're all exactly at the same spot. And then you can look for example, changes and then you know where something is exactly. Usually that is done by the satellite data providers. And I think right now, most of those have good enough on -train models that we can focus on our own strengths, which is mostly the data analytics

and going into depth on the AI models. Thank you for that question. And thanks for that answer, Lisa. I have another question. What is your kind of biggest frustration about the limitations of the current technology? What's the thing that you run across the most that makes you go, oh my gosh, I wish it wasn't like this. I wish we could fix this issue with the technology. What's something that's missing in the technology or a capability that's missing at the moment? Yeah,

it really depends on use case to use case. What is, for me, I think my biggest frustration is not as much in the technology, but in the technology communication. So I'm spending a lot of time with different users and different potential users talking about, hey, what can be done and what cannot be done. And then sometimes you have people who say, who think we're working with satellites and AI. So I want to see this individual

tree and how well it's doing. And sometimes you have people who still think like satellites and AI, no, that's not working in forest, forest is special. It's somewhere in the middle and that's always really annoying. If we go back to really technical, I think a lot is possible.

For me, the challenges with working on a large scale with high resolution data, especially with time series, are still really annoying because I know that there's a lot of data out there, but most of it is commercial and then it tends to be nuts. available for a lot of the applications that I feel need it most. You can do a lot if you want to go into oil and gas, but if you want to make sure there's a world left, then it's

a constant uphill battle that way. Again, maybe not the most technical issue, so I think, yeah. I kind of like the technical challenges. Yeah, clouds are annoying, but it's part of the fun, right? Gives us a puzzle to solve. Yeah, whenever we work with, we have a really wonderful network of volunteers and sometimes we get satellite imagery from a satellite image provider that they request. They'll be working on a case and they'll go, hey, I need a picture from here.

And if the volunteer, if the individual isn't familiar with satellite imagery, we have lots of folks who've never worked with it in the past and are now working with it for the first time at BellyCat. you know, 95 % of the time we get the image, we send it to them and we go, it's cloudy. Like we could, you know, here's the result, or you can't see anything. And they go, oh, of course, like there's clouds. Yeah, there's lots of clouds. And depending on the place, there's

clouds like every day. So, yeah, definitely sympathize with that. We did a project in Ivory Coast and I think in a year's time we could get one image. Yeah, that sounds about right. Yeah. We have another question here from another community member named BQN. And BQN asks, Lisa, what is the most interesting or innovative or unusual or creative application of satellite image data that you have come across? So I kind of want

to talk about this project we've done. It wasn't technically with satellite image because here we could get through a partner, really like the project as well, some aerial data. But basically for us, the processing is pretty much the same. It's a bit less atmospheric correction needed, but data is data, right? So I was working with aerial data and that was to map out latrines. Where MREF held Africa, this project, this wash intervention to support, to give out micro credits

for people to build latrines. And they wanted to know how effective the project was. Now they needed to do a comparison to a different site. to see how many latrines people would normally build, at least would build us outhouses. And it's kind of invasive to be knocking on people's

door and ask, hey, can I see your toilets? So they were looking for a different solution and then we developed a model to see, hey, where can we see outhouses and how can we distinguish latrines from chicken coops, which is very challenging. And I really like talking about that, like looking at the different angle, like, yeah, sure, sure, we can map latrines, why not? Yeah, I don't think I've ever heard of anything like that. That's fascinating. Thanks for sharing that. We have

a question here from Aikan. Hello, Aikan. Thanks for your question. And the question is, do you have any tips for using platforms like Google Earth or Sentinel for identifying and confirming smaller scale environmental sites like artisanal mines, for example? So artisanal minds, we've also done some work in. I believe there might even be some tutorials using it in Google Earth Engine. Otherwise, what I would really look at is look for bare soil index. So really look at

bare soil and mapping changes into that. So looking at the time series analysis, you can look into,

for example, BFAST, B -F -A -S -T. that's normally done on the NDVI but you can also do it on for example bare soil index and using that to map where you see for example bare soils popping up in certain area sizes so you want to look at then quite small area sizes where there's quite a rapid change ideally you make some some smart selections about your area of interest where you expect it to happen so for example close to rivers and then Second point of advice

is don't be afraid to get dirty and just click through images and look at things. So use these algorithms. If you don't have a lot of training data, don't spend too much time on building complex models. If you don't have the data to put in there and the model is not going to be great anyway and click through things, look at the different images and use your own eyes as well because our own eyes are very good at change

detection and pattern recognition. Use the computer to take the first step, but don't try to get everything done perfectly Rather just look and click and do the dirty work yourself as well We have a question now from wit Fox and you've talked You know, we just talked a little bit about this. Let's let's see if we can get more On this Lisa. You've talked a bit about deforestation biodiversity fires mines. We talked about latrines a moment ago What are some things that you think

satellites are being underutilized for? Or what are some other perhaps non -obvious use cases for satellite imagery data, either optical or multispectral? Thanks. Take your moment. Because I know satellite data is being used for loads

of different things. even for mapping penguin colonies by looking at their shit you can do with satellite so what they're not being used for is kind of difficult to look at yeah I think a lot of places a lot more can be done I think for mapping these environmental claims for big organizations I think they can be used a lot more because I feel like for a lot of these cases companies will make a claim, for example, in

the offsetting. They will share some sort of location with it or they will share some sort of report with it. But the data tends to be based on field work, which is nice, but which must be validated. We've seen it a lot where even companies we work with who do field work professionally and who do it for the project we asked them, give us data where there's just errors in there. That makes me wonder how much validation is being done to support these claims for, for example,

the regulated carbon market. I know it's being done there, and I know it's kind of an open and short case, but I think more can be done there. Thanks, Lisa. We have about 10 minutes here left with Lisa Brughausen from Space for Good. If you have any questions, please type them in the chat. You've got 10 minutes to ask her. A question, we've got one here from SirSkippy. And SirSkippy asks, do you make the data that you collect available on OpenStreetMap for ECUs by others? No, we do

not. And the data we collect, we don't have our own satellites. So the data we collect is the satellite data itself, the raw satellite data. And the data, the results, the information we generate We don't make available and I think so we also work for, for example, commercial entities to take their direct action. I think in most of those cases, it would not be beneficial to them to publish it. And then we also need to be cautious with that and rather focus on

the direct action and on the mapping. Plus in OpenStreetMap, if we do, for example, deforestation

detection. So like I said, what we do is we look at deforestation as there was a forest it's gone now there's no more forest if you look at most land cover maps they look at what is the land cover we the official land cover that's here the land use that's here so if you cut a forest and a new forest grows it is still the land cover class of forest and then there's no rapid change for change that we detect and then in the other land cover mapping we tend to have a bit more

classes than are relevant for This road networks could make sense. But even then I think that working with the GPS tracker so people have locally are probably going to be more, uh, more directly applicable also because it's in private areas. So they don't want the date. They don't want people on the roads. So no, we do not. The short answer. Thanks, Lisa. We got another one here from, um, uh, this one's from Charlie. Um, are you Charlie from, from Bellingham? And Charlie

asks, I'm going to paraphrase here. Imagine you're listening to this and you go, wow, Space for Good sounds incredible. I think I have some skills that might make me a candidate. I'm going to look for postings on the Space for Good website. Maybe if they're hiring somebody, I'm going to apply. And the question to you, Lisa, is what kinds of skill sets do you look for in team members? Just in case somebody in the audience here wants to maybe give a shot at working with Space for

Good. Yeah, thanks. So what we're looking for is really in -depth remote sensing and modeling skills. So if we look for a data analyst or data scientist and we look for people who really have experience in the data, but also have experience in one or two fields of interest. So we have, I myself have a mixed background with forestry. We have someone who has a background in doing carbon projects in the fields, getting that access with someone who's done a lot of work around

fires in the past, agriculture. So knowing, and that's where I think most value lays in, combining the data sites with direct action in the field with something, someone who knows the background. If you don't know how a tree grows, then monitoring from above is going to be a lot harder and monitoring it when it's gone is going to be very difficult to assess. So that's ready for the data. They're the analytics, they're the scientist position.

Then we have, if we work with, we have, for example, at the moment, one full stack developer that was looking for someone with cloud experience, scalability, et cetera, et cetera. But for us, the soft side is also very important where you've got a very cool team, if I may say so, and really working together as a team to build something

that's better. And then, of course, around commercial, well, we've got a very small team around that, but that's also always good to have people who can try to get it to the final users and clients. Thanks very much. We have about eight minutes left here with Lisa. Get your questions in, folks. We have a question from Dusty Wieners. By the way, Wieners, if you folks don't know. is the skin on your elbow is the weenus. I'm not sure if you are familiar with that term, but Dusty

Weenus is asking the following question. They say, Dusty Weenus says, is there any way to verify direct carbon emissions from industrial sites from space? For example, A company claims a certain amount of scope one emissions, but the satellite data actually indicates a larger amount. Is that possible? For example, methane mapping. Thanks, Dusty or Venus, I'm not sure how to refer to you. For example, for methane, there's some satellites out there that can really map at different resolutions

the direct methane emissions. It's always important to look at then the full scale of events. If they report CIRTOSCOPE -1 emissions, you need to have the correct overpass times, the correct models behind it. And there's always the issue that you look with the satellite at the full column. So at the entire atmosphere, basically.

So there is a potential for a mix -up. However, these satellites have been used, for example, to identify when there have been methane leaks or large -scale methane emissions where they shouldn't have been due to any technical issues. So yes, it can be done for larger scale operations. The farmer next door with 20 cows doesn't exist anymore. But anyway, you're not going to be able to find that. But for the large factories, yes. And for the pipelines where there might be maybe

slight issues with the maintenance as well. I'm being positive here. Thanks. Thanks Lisa. I think this is going to be the last question here. Somebody notices we're talking about, you know, work in the field and maybe working space for good. Somebody noticed that there's an astropreneur position on the website. Are you familiar with that one? I'm asking it because Chris is asking about this, but also I've never heard of that term. Do you know about the position, the vacancy? Can you

talk about it? It might be a word we made up. More on the data side than that side. So let me quickly, where is that? It's in the, yeah, spaceforgo .com forward slash vacancies. Marketing and sales, market research and analysis, finance and investment. Oh yeah, talented and engaged entrepreneurs. So that's basically an open, always open vacancy where people who are interested

can. let us know and we can see whether there is there is anything out there uh that always depends a bit on projects on current needs on skills whether we can fit something in or not um lately usually not i have to admit because we're with quite quite a strong team at the moment uh but quite frequently we can find figure out something so um yeah entrepreneur combination of uh astro and entrepreneur so also looking at people who want to We do work as well with

a shell of freelancers who take specific projects or who we support in giving specific projects when there are specific skill sets we don't have in the team, but we need in certain projects or that just want to be somehow connected because we are a lot of fun. Great. Thank you so much for that. And we've got this last question here, folks, again, from SIRS -KP, has the change in US government funding priorities affected Space for good at all. We get that question about again

all the time. We're passing it on to you Lisa space for good. Yeah, you don't have to ask the book Affected us. I mean emotionally financially directly or indirectly. I think it's affecting affecting the entire world At the moment, it has not affected any of the current of the projects that we do. And that's also where we try to have a mix between commercial projects, projects with

NGOs and governmental projects. We do know that a lot of our partners have been affected quite strongly and we do see quite some risk here because the budgets are also affecting the data sets. And then we're looking at the climatological data sets, which we use if we make predictive models. It's important to have those available globally rather than only in certain regions because the climate is kind of a global thingy.

So there we do expect it to affect us on the data set quite a bit and on the projects not on the active projects we have but on the yeah future ones unfortunately I think that's yeah that's great well On that depressing note, we're going to call it. Thank you so much, Lisa. Brooke Hellsend from Space for Good. Thanks for your wonderful presentation. I learned a lot. I know folks here learned a lot as well. Thank you very much, Lisa, for spending this hour with us. Thank

you. And have a good evening and day, everyone. Thank you for listening to the Stage Talk. If you'd like to catch a Stage Talk live where you can ask the guest questions, join the Bell and Cat Discord server by visiting www. The music you've heard is titled Dawn by Newer Self and is courtesy of Artlist.