Share | 11/14/2019
We sat down with Jenny Williams, a specialist in conservation-based spatial analyses and piloting UAV remote-area surveys. She explains the research she’s been able to accomplish with the eBee fixed-wing drone technology, how drones have transformed her workflow and why she chose the eBee to help her conduct her conservation-based research.
Hi, Jenny! Thanks for chatting with us today. Let’s get started.
The Royal Botanic Gardens, Kew is one of the largest and internationally renowned botanical gardens in the world with more than 350 research scientists dedicated to documenting and understanding the globes plant and fungal diversity. At the Kew site, we have physical botanical gardens, an herbarium (dried plant material, including Darwin collections) and a genetics laboratory. We have another site, Wakehurst Place, which is home to the Millennium Seed Bank.
I work in the Biodiversity Informatics and Spatial Analysis department, so my role is to be a scientist in everything geospatial. This ranges from discovering the impact of climate change on individual species, like arabica coffee, to identifying and mapping threatened habitats for Tropical Important Plant Areas.
Most of our drone work takes places in the tropics, which is where Kew’s research interests have been historically focused. However, we spend some time at the Wakehurst Place site to check and test equipment due to its large, open areas. These flight tests help us ensure the safety and success of drone missions for when we’re in remote areas, such as Madagascar.
For our large-scale drone survey mapping, we always use the eBee fixed-wing drone. It’s the lightest and most portable that we’ve found, which allows us to fly further and longer at the resolution we want.
Before going on a drone mission, we check our kit thoroughly and make sure we have plenty of charged batteries for the drone and laptop. Once we arrive on site, we go through our pre-flight survey checks and determine whether we may need to restrict access to the public during our aerial survey flights, launches and landings. We work in pairs so there is at least one drone pilot with an observer. We always need one pair of eyes on the drone and another looking out for low-flying air traffic, birds of prey and unfavorable weather systems—anything that may threaten the drone’s flight.
We predominantly used Landsat Thematic Mapper, which has 30-meter resolution, because it’s free and has a long ranging archive of imagery. We’ve also used Sentinel data and, as projects would allow and where we had funding, we’d purchase a small area of WorldView, Quickbird or IKONOS imagery to do similar-level data collection. It’s certainly not the same resolution, but you can’t get the same resolution with anything other than a drone. Plus, a lot of the areas we work in are tropical, so trying to get satellite imagery in a cloud-free sky is almost always problematic.
A sunset view from camp after a successful day of drone missions in Ambohimahamasina Forest, Madagascar.
I had a project in Sumatra where I needed precise imagery. Despite having the rare budget for the collection of WorldView2 imagery, plus a year to work on this project, we were never able to source the imagery from the archive or task the collection because of cloud coverage. I ended up using the most current, high-resolution imagery we could find, which was a three-year-old, 10-meter SPOT satellite image. Obviously, this was not ideal or timely.
With drone imagery, we can actually count the number of trees that have fallen or been cut down to potentially discover the density of a forest or estimate future losses.
About six years ago, we looked into getting a drone so we could survey specific areas at specific times and in the specific seasons that worked for us. For some projects, we wanted to know the health of a tree during its healthiest versus most-stressed time of year. For others, we wanted to look at vegetation seasonally to determine variations in species during times of leaf-on or leaf-off. We needed to have the best chance of getting survey imagery without clouds!
Our decision to purchase a drone had a lot to do with the high spatial resolution of drone imagery. We needed the ability to identify individual tree crowns, measure and count felled trees or stumps, precisely map burn scars, and monitor small-scale vegetation regrowth. We simply cannot get timely high-resolution, cloud-free imagery any other way. When it comes to drone surveying, there is no other option for us. To get the level of detail we require, where and when we want it, we need to go and collect the imagery ourselves. Drone footage provides something that nothing else can, there’s no other way we can get that imagery.
We first flew the eBee Classic with the RGB, Near-Infra-Red (NIR) and RedEdge sensors. We now fly two eBee Plus drones and mainly use them with the S.O.D.A. camera, but often perform dual flights with the Sequoia to get multispectral resolution. The spectral information provided at different wavelengths is necessary for extracting specific vegetation features. Since vegetation is distinctively high in the NIR, this is where we are able to extract the most interesting information and indicate differences in reflectance between species, health and structures.
The main benefit of the eBee, for us, is that we can survey large areas with different cameras. There’s nothing else other than the eBee superlight fixed-wing drone that we’ve found that is able to do that.
I tried to extract felled tree trunks from drone imagery I obtained in Madagascar. There was a large shadow cast over the forest-cut area caused by intact trees. With the S.O.D.A. images, this area appeared dark grey and black because of the shadow. When I looked at the same images from the Sequoia camera’s NIR and Red-Edge bands, I could clearly identify the felled tree trunks. This is exactly why the ability to use and switch between different cameras has been so important to us.
Most of the imagery we need to collect is in remote areas. In most of the developing countries we work in, we must take the drone kit and personnel in a single 4×4 SUV to the research area. Then, we either drive or carry that kit to the base site. When we looked at potential drones, we asked questions like, “which is the lightest?”, “the most portable?”, “gives the best value for money?”, “can fly the longest?”, “can take quality pictures?” and “enables us to put different cameras in it?”.
Jenny plugs away on a laptop in eMotion flight planning software for a drone mission with the eBee Classic above the remnant forest in Ankafobe, Madagascar.
The eBee is light, flies far and is easily portable. Even the spare parts we carry are light—wings, propellers, band attachments. Most of the weight comes from the extra batteries. We bring 10 with us to enable a full day of droning. However, these are also extremely portable, despite their special LiPo fire-retardant storage bags.
First, we work out the area we want to fly. For example, when I was in Madagascar, I was given priority zones by Feedback Madagascar, the non-profit organization we worked with, who had teams on the ground. Then, I went through the eMotion software to plan potential flights and figure out the flight height, resolution and distance needed to cover the area. I had to make sure we could fly with spatial resolution sufficient enough for tree and cut detail we needed while also flying high enough to cover large areas and maximize the short amount of available flying time. Once we figured that out, I created flight plans that covered the high-risk priority zones for deforestation.
We take on far fewer extensive ground-based surveys because we can get so much information from the drone.
Next, we determined where would be best for launches and landings. With the eBee, the launch and landing zones are the most critical part of any mission. We look for big, flat areas that are open enough to get us clearance for proper, safe landings. In Madagascar, these sites were ground-checked for us in advance and only slightly altered to ensure there was camping space.
We had four teams—one of which consisted of the drone pilot and observers. The other three were “roaming teams” to specific geographic locations. The roaming teams were responsible for maintaining consistent visuals of the drone, observing weather patterns and keeping an eye out for our greatest threat: birds of prey. As the pilot, I constantly monitored the drone’s status, signal, battery as well as radio reports from the roaming teams to ensure the drone’s safety.
We then examine the drone, change its battery and initiate a new flight plan. It’s a quick turnaround process before we’re able to do it all over again.
An eBee Plus gently glides above land in Ambohimahamasina, Madagascar as it utilizes the information Jenny gathered to ensure a safe landing after a successful mission.
We use eMotion to link photos to flight logs and for the initial post-processing, and then bring it into Pix4D. We’ve used other bits and bobs, like Agisoft, which one of my colleagues uses. We’ve also been experimenting with the Pix4D bridging product, Drone2Map, in ArcGIS, but I mostly use Pix4D. They have tons of good tutorial videos on specific processing options that have helped us with more complicated photo blocks. They’re always good about answering our questions and getting back to us. Their team has always been helpful.
With drone imagery, we are looking at a completely different level of detail. It allows us to develop a new way of working so we can fully exploit the information held within these images.
For our large-scale drone survey mapping, we always use the eBee fixed-wing drone. It’s the lightest and most portable that we’ve found, which allows us to fly further and longer at the resolution we want. We use quadcopters here and there, but they can’t match the flight time. Their battery is only about half of the eBee’s. Hardly any of our Madagascar surveying would have been possible if we used a quadcopter.
The eBee Classic soars above the drone base, campsite and first-ever drop toilet (bottom-left) in the community of Ambohimahamasina, Madagascar.
We don’t really have any comparison at this level of detail. Sometimes, we could purchase high-resolution satellite imagery, but pixel sizes were always greater than one meter. It was even more than that with multispectral imagery. To extract features from the satellite imagery, we would have to trace around the features or downgrade the imagery’s resolution to something a bit more coarse so we could effectively use various automated extraction techniques, like image classification. With drone imagery, we are looking at a completely different level of detail. It allows us to develop a new way of working so we can fully explore the information held within these images.
You can’t get the same resolution with anything other than a drone.
What we’re doing with drones is completely different to our traditional satellite image interpretation. We now look for things at a much finer resolution over larger areas. For example, with drone imagery, we can actually count the number of trees that have fallen or been cut down to potentially discover the density of a forest or estimate future losses. We can identify the health of a specific tree or species, whether they are stressed or diseased. We’ve never been able to do this before we had drones. We could only extract small amounts of this information from ground-based surveys or plot work in combination with an analysis of satellite imagery.
Recent deforestation of the primary forest in Ambohimahamasina, Madagascar. The detailed imagery shows individually felled tree trunks.
A few years ago, I spent time in Sumatra doing extensive ground-based plot work. With 10-meter resolution SPOT satellite data and three teams, we completed 330 plots with a radius of 12.6 m (500 m2) in three weeks. The plots were rapid, so they weren’t full surveys, and we weren’t able to identify every species. We were mainly looking at the numbers and quantity of tree species and seedlings to determine the regeneration potential of the forest after years of selective logging. Based on the field data, I was able to classify the satellite image of the Harapan Rainforest into different forest types and estimate carbon storage as a base for future regrowth or illegal deforestation to be monitored and assessed.
Now, we take on far fewer extensive ground-based surveys because we can get so much information from the drone. Field-based teams can also ground-truth the aerial photos with far fewer surveys.
Since we mostly fly over dense forests with only sporadically cleared areas, the main challenge we have with the eBee is finding suitable launch and landing areas. Thankfully, because it can fly far, we can often position ourselves outside the forest and fly in over the portions of it we want to map.
What we’re doing with drones is completely different to our traditional satellite image interpretation. We now look for things at a much finer resolution over larger areas.
A bigger challenge is getting the country’s Civil Aviation Authority clearance to fly our drones. There are many countries with strict regulations that restrict where you can fly, if you can fly and if you can even bring a drone into the country. Some require that you get a permit to fly, which is what I had to do for Madagascar. I submitted my drone pilot certification, all the flight plan and route information, the drone details and our operations manual. All of that had to be signed off before we could fly in the country.
Jenny’s colleague Tim Wilkinson shows the village children where the safe launch and landing zones are during their mission in Ambohimahamasina, Madagascar.
When my colleagues worked in Peru, they had to provide a large down payment as collateral to provide assurance that they wouldn’t sell the drone they’d brought in. They had to do this upon arrival and go back to retrieve that money before they left. The process was very time-consuming and extremely stressful.
In Madagascar, we found the fascination of the drone to be a slight challenge, too. During a drone flight demonstration for local village elders, a large group of school children bombarded the site. There were quite possibly 100 or more children. They were released from their classes after hearing the drone takeoff—all at once. All the Malagasy staff from the Kew Madagascar Conservation Centre who were working on-site with us were suddenly on crowd-control duty to make sure the drone landed safely!
To get the level of detail we require, where and when we want it, we need to go and collect the imagery ourselves. Drone footage provides something that nothing else can, there’s no other way we can get that imagery
Even though we had the buy-in on the research side, getting the funding for it wasn’t so easy. Working for a non-profit organization, anything that’s a financial outlay is quite complicated and must be fully justified. We had to bid separately for the acquisition of this research kit as we can’t purchase it through normal funding sources, such as research grants.
I still have 38 flights of the Madagascar forest multi-spectral image data to work with – the forest’s mosaic map was just the tip of the iceberg!
I’m also applying to get my UK CAA pilots license, which is not currently a legal requirement. It now seems it’s harder in most places to fly drones due to necessary safety regulations. I think almost everyone has experienced at least one dangerous or annoying drone flight these days.
We are starting to rely on our drone images for much of the innovative research with our international partners and collaborators. There is so much scientific knowledge to be gained from all this new information collected in these unique and threatened habitats. Time to get processing!
Sam Cameron from Feedback Madagascar (center-left) and Jenny Williams (center-right) brief the village communities about the upcoming drone flights in Ambohimahamasina, Madagascar.
Thanks so much for your time, Jenny.
Good talking with you. Cheers!
Sign up to receive updates directly to your inbox.