Editor’s Note: This is the first part of a two-part post about using drone technology to search for orangutans around the Sikundur research station in North Sumatra. To read the second part, click here. For all posts in the series, "Call of the Orangutan," click here.
Unmanned aerial vehicles (UAVs), or drones, are increasingly being used for environmental and conservation applications, including near real-time mapping of local land cover, monitoring of illegal forest activities and surveying large animal species.
For us orangutan researchers, the cost of doing traditional population surveys—that is cutting extensive transects, then walking along them counting nests—is painstaking and expensive, demanding both a lot of people and a lot of time. The purpose of Graham’s flights at Sikundur was to complete nest surveys from above transects to compare the results with those collected by field staff on the ground. The idea was that by running the two surveys simultaneously, we’d be able to correlate the aerial results (which will encounter a lower number of nests) with the more comprehensive (but expensive and time-consuming) traditional methods.
Do this enough times and you can just run the fast and cheap UAV survey (each drone costs between $1,000 and $2,000 depending on the set up, with another $1,000 for the equipment, software and camera) and the results can be accurately extrapolated to give the number of orangutan nests needed for population estimates.
On the first day, Graham planned to run a test flight using a Skywalker drone fitted with two GoPro HD Hero3s mounted to capture photos and video from angles directly below and in front. The purpose of this flight was two-fold: firstly to identify any large trees that needed to be avoided during the lower-altitude nest survey, and secondly, to build a real-time, geo-referenced land use and forest cover map of the site using a composite of the GoPro’s stills stitched together.
We got up early hoping for a fast start, with Graham having set up the drone and rigged a radio antenna on the roof of the building that housed our flight control operation the night before. However, the weather had other ideas, and the rain bucketed down all day, while I sat around feeling useless, watching Indonesian cooking shows with the field staff. The next morning the rain finally abated, Graham switched on all the systems, and I went out onto the grass to get loose ahead of the first of my two tasks as his assistant: throwing the drone for it to launch. Helpfully, the our field guides found me a hefty plank of wood to practice with, while (a little less helpfully), Graham offered the encouraging words that “last time out we had to give up as the guy just kept throwing it into the ground.” He was sure I’d be fine though.
Launching and landing should really be the only times when anything can go wrong. Each flight mission is programmed by defining waypoints using open-source software. The drone is thrown into the air in “fly-by-wire” mode, under which it will level out and head to the first waypoint. Then it will circle until a command from the remote control switches it to full automatic mode and it follows the programmed route.
With the route set, Graham quickly made a few final checks and the plane was ready to take off! I was feeling pretty good about my launching ability, having worked out a system whereby I’d use my feet to pick up speed so the autopilot would know it was moving, and keep my arm straight up with the drone in a horizontal position in order to gently push it into a level takeoff. However, despite my practice with the piece of wood, holding the drone was a whole different ball game, and by the time we were ready to launch I was almost too nervous to move!
Lifting the drone up high into the air, I became immediately aware of just how heavy it was. Luckily, Graham more or less immediately started the engine and the drone began to pull hard, leaving me little choice but to start running. After five steps, I pushed it into the air. Fortunately the drone stayed fairly level and cleared the trees in front of our launch sit with a couple of meters to spare.
Quickly the drone climbed to its mission altitude of 160 meters. Graham switched it to automatic and it flew off towards the west, starting a journey of around 35 minutes and 29 kilometers. We headed back inside and sat glued to the computer screen watching the drone’s progress, always making sure to keep an eye on its altitude, trim, air speed and location. Everything went perfectly through 14 turns and the drone completed its survey with plenty of time to spare. On the fifteenth and final turn, approximately 2.5 kilometers from returning to our location, the drone suddenly disappeared from the computer screen. As the “Contact Lost” warning light blinked, we scanned the sky anxiously looking for the returning drone. After five minutes we knew it wasn’t coming back and mounted a search-and-rescue mission into the jungle.
Traveling three kilometers through hilly jungle feels like a lot farther than it is, and it took over an hour and a half to get to our search location. We had rough coordinates of where the drone should have turned based on the mission parameters. However, due to the distance from our home base, the plane’s GPS was out of range of the VHF antenna, so we didn’t have an exact location.
What followed was a miserable, neck-straining afternoon spent staring up into the trees for any sign of the drone. The area we searched in was hilly and had some massive trees, making us think the unit had probably misjudged its altitude (there’s a decent amount of variation in the UAV’s sensors), and plowed into one of them. We did try to keep our spirits up by discussing more wild theories such as the drone being taken out by a raptor or flaming out spectacularly. However, our search continued fruitlessly until it started to become dark and we abandoned our efforts for the day, trudging back disconsolately to camp.