 I'm very happy to be able to introduce Serge Wich, who's a professor at Liverpool John Moore's University. He's an ecologist who studies both the behavior of primates and the ecology of forests, primarily in Indonesia but also elsewhere in the world. He's also one of the co-founders of Conservation Drones, which is a pioneering group which does exactly what its name sounds like. Serge, I think, has both done tremendous work himself and has a very good command of the work that other scientists, ecologists, environmental scientists have been doing around the world with drones, where there's really been an explosion over the last two, three, four years where there's already a pretty vast scientific literature, which Serge talks about, sort of, a wonderful read in a chapter in this book that we're putting out today. So, without further ado, I'll let Serge speak to us a little bit about his work and conservation and drones more generally. Thanks, Serge. Well, thank you very much for inviting me and for a wonderful introduction. So, I'm an ecologist and I spend most of my life doing pretty low-tech work. I camp out in the forest and with pen and paper and more recently with a GPS, we collect data to find out where animals are and to do ground-true things for satellite images. And that's all very good and fun, but it's very slow, it's very expensive because sometimes it takes us two weeks to get to one data point, which is not very efficient. Mostly, my work has been focused in Indonesia on orangutans and as many of you know, the landscape where these animals occur and many other animals around the world is changing very rapidly, so it becomes more difficult to map that as these changes go very fast, so we need to find out where these animals are and where their habitat is and how it changes into areas that are not particularly suitable for them, but in which they still might occur. So a couple of years ago, I met up with Professor Leon Pintko and we thought it would be wonderful if we can just fly over these forests and collect data with drones, so we started with the approach that Gregory and others have mentioned as well, do it yourself systems because we wanted it to be affordable, fixable in the field and available for a reasonable price for local conservation NGOs and local communities that we were planning to train. Over the past few years, we've been training people in about, I don't know, 25 countries and a lot of that work focuses on basically three things that we're trying to do with conservation. One of the aims is to count wildlife. The other thing is to map the areas where these animals occur and how that landscape is changing and the last effort is focused on anti-poaching, which is an increasingly large problem in Africa, but also in Southeast Asia. So when people try to count wildlife, one of the main aims that we try to accomplish was to count orangutan nests, which orangutans like any great ape make a nest every night or a sleeping platform and we try to find those on aerial images, which works really quite well. So in addition to that, we do a lot of ground proofing to see whether the aerial data that we get is actually comparable to ground data, so our densities and distributions are similar. So can we actually really use drone-based data as an alternative to ground-based data? And in this image, you can just see that we collected a lot of nest locations in the field, overlaid those with one of these high-resolution automosaics and then put the aerial nests on there as well, the yellow stars, and there's a very good correlation in the number of nests you get per transect from the ground and from aerial-based platforms. So this looks as a very promising technique to really be used as an alternative to ground-based analyses and obtain data that is a much higher resolution than land-set data as well in terms of mapping. Of course, many people are interested in just observing animals and finding out where they are, so we've been taking pretty pictures of animals in many different areas, for many different projects, and of course, as you start collecting these pictures, you after a while realize that you're collecting lots of pictures and you're supposedly trying to be more efficient, but then what do you do with all those pictures? In the beginning, we went through them manually, and that was fun for the first hundred pictures, but then you're like, hmm, this is not really the future. So we started to talk to colleagues and are now interacting with people from computer vision labs to do automatic object recognition for things like orangutan nests, counting birds automatically in images, counting cows, and my most recent collaboration is with my neighbor actually, who's an astronomer at the same university, and we're using the algorithms they use to find stars to automatically detect animals in thermal imaging, videos and photos. So it's really exciting that this field brings together lots of different disciplines, which for scientists like myself is great fun and a good intellectual challenge because it sort of broadens your horizon. So there's a lot more to be done there. There's also a lot more to be done for land cover mapping. I totally agree that it's fairly easy to fly over these areas now, get high resolution maps, get photos that illustrate differences between a diverse forest and a more homogeneous patch of land near a lake, and you can try to identify trees on these images, but this is also a lot of manual work still, if you want to try to look at change detection or do land cover classification, that still requires quite a bit of work, and one of the things that we're trying to do is we're posting all our maps on Google Maps Engine, so we can link it to the Google Earth Engine where we can do classification, so we're hoping to get at some stage to sort of an easier workflow where you can get these ortho mosaics and then have trained algorithms that can automatically classify areas that are primary forest, log forest, oil pump plantations, et cetera, so that we don't really need to do all that manually and in different software packages, but getting to one workflow for that is something that I think is very important for the future. We're also very much aware that drones stay in the air for a limited amount of time and that therefore they can't map the same amount of areas that satellites can map, so we're trying to use drone data as training data to improve the land cover classification from satellite images, and that's working quite well, I won't go into details, but you can see that the accuracy of land cover classification improves as you start to use drone data as training data for these classifications, and I think that's an exciting field because there's not so much work going on on that, and I think that will help a lot so that we actually see these technologies not as alternatives but as complementary to each other, and I think that's something that could use quite a bit more effort. We're of course also using the structure for motion things to make these three dimensional maps of forest, which is great for us because then all of a sudden we can map the size of trees, we can map the size of gaps, we get a view that we never had before of the areas where we work, in a way it's poor man's lidar, but it still is quite good because it allows us to, for instance, this is a study area that I work in in Sumatra, to derive tree heights from these structure for motion maps, and that allows us to, for instance, look at where orangutans travel and where potential corridors are for wildlife, and that's also data that we couldn't really collect earlier. So we do a lot of these maps, we do use RGB cameras, a lot of people are using near-infrared cameras to help land use classification and to produce these, what they call NDVI maps, which gives you information about the photosynthetic activity of plants. So the greener areas in these images are areas where there's more trees than the blue areas and where there's more photosynthetic activity, and that's also very useful data for conservation people and of course also for agronomists. So the last part that we focus on is poaching. We're trying to use these systems and train local poaching, anti-poaching teams, in this case in Nepal, where there's tigers, elephants and rhinos, and we're trying to, we've trained them there to use these systems and operate them to be part of their management and we started with that in 2012. Now it's 2015, we're in the third year that there's no poaching in that area. Of course that's probably not causally linked to using drones, but it's part of a toolkit that they now have that can work to reduce poaching and it might potentially only be sort of a deterrence in some cases where it's publicized widely in the media that these systems are being used now and so that people are maybe not going into these national parks anymore because the risk is just a little bit higher. Of course with poachers it's very difficult to find them particularly if they're under the rainforest, even with thermal imaging cameras, it will not penetrate through because there's a lot of cloud coverage in between. So one of the things we're trying to do is to use these systems to detect smoke plumes from drying racks that poachers are using to dry bushmeat. This is particularly an issue in Africa where poachers go in for a week, they dry bushmeat, then they come out with dried bushmeat, but these smoke plumes give their position away, so this might potentially be really good. This is the one slide that never wants to, okay. So we fly over these areas and then you see these smoke plumes and you can basically see them from a few kilometers away, so this could potentially be very effective, particularly if you would use small quadcopters to just fly up, oh, let's see if this runs, no. Well particularly if you have smoke plumes or small quadcopters, you can just send them up, do a 360 above the canopy and you can detect where those smoke plumes are. You know the compass or the degrees where these smoke plumes are and then patrol teams can go to those areas and that's much more efficient than what they're doing at the moment, which is more or less sort of a random walk through the forest in some cases. There's many people working on anti-proaching efforts around the world here in Washington. Thomas Snitch is working on that in Africa. So there's lots of exciting things that are going on in the conservation world and there's lots of things that I think will need to happen to help this effort more. And one of them is increased flight duration, particularly for multi-rotors which people already have said they're easier to fly. You need less landing space and their flight duration is increasing. So we now have systems that can fly for about 40 minutes, fly 10 kilometers. So that's quite a lot better than one or two kilometers. So that will change the way that we can use these systems in terms of the ease it is to use these systems but also where you can apply them. I think we'll see a lot of rapid developments in all sorts of sensors, regular cameras, thermal imaging, multi-spectro hyperspectral lidar, and that will be exciting because we can just fit those, more of those on small drones. We'll see an increase in user-friendliness all the time, even with the open source systems. They're becoming very user-friendly. Even a field biologist like myself can put one together and fly them. And that means it has to be quite easy. There's a need for improved data analysis. Now it's still a cumbersome workflow of detecting animals on an image, get the GPS points from that, putting that into some software to create a distribution map or allow for density estimates. So that whole workflow needs to be improved and I hope that will happen over the coming years. I'm quite positive that it will happen. I think another thing that we would like to see is use automatic object detection on drones so that we can use the limited bandwidth that we have to send rangers only the data that they need, not thousands of images of savanna trees but only an image of a rhino of a poacher so that they know where to go. So I think there's quite a bit of work need to be done there as well. I think in the conservation world nobody is using swarms at the moment. I think that will change rapidly as well and that will mean we can cover larger areas and we can cover areas for a longer duration which will be very valuable as well. And the last thing I think is that I hope there will be a lot of integration between all these sensor platforms in a way, drones, cameras on the ground, microphones on the ground, other sensors that are detecting activities on the ground so that GPS tags on animals, VHF transmitters, GPS loggers, et cetera. Now those things are not integrated yet but I think we'll see a lot of integration in the coming years and that will be extremely helpful for overall management of conservation areas and data gathering. And that's where I want to end with a picture of the core team of conservation drones. There's a lot of more information on our website but I hope this gave you sort of a whirlwind overview of what people are using these systems for in conservation over the past few years. Thank you very much. Sure. Yeah. If there's questions please fire away. Walter? Conservation efforts tremendously. They might just help the poacher exponentially more than the protectors of the wildlife. What kind of ideas do you have to prevent the technology actually accelerating co-poaching skills and putting the making the conservation effort even harder than it is now already? Well it's a very good point and it's concerned that many people have. I think it really depends. There's most of the poachers are not well funded and will not for the coming years be able to use these technologies. There's of course very well funded poaching teams out there that I think will be using these technologies and I think there are even cases where they are already using them. So it will really depend on the local context whether we will see poachers using these technologies and if they will that will become a very interesting situation for sure that we'll have to deal with in some way or another. Yeah. All right. Thank you very much.