UAS Applications in Wildlife Ecology
Hudson Bay Project Overview
UND Biology teamed with the Hudson Bay Project (HBP) to examine utilization of an unmanned aircraft system (UAS) for surveying lesser snow geese and other aspects of the local arctic ecosystem in order to facilitate expanding non-invasive sampling methods within this long-standing research program working in what is now Wapusk National Park. This Canadian National Park is located near Churchill, Manitoba.
The team used a Trimble-UX5, fixed wing aircraft capable of pre-programmed flight (Figure 1). Images are collected by a Sony 16 megapixel camera mounted in the body of the aircraft and operated by an automatic infrared trigger. These images are then stitched together to create a larger landscape mosaic.
Figure 1 . The Trimble-UX5 mounted on its catapult launcher ready to launch from camp (Nestor
All flight operations were conducted under a Special Flight Operations Certificate (Permit Number: 5812-11-302, ATS: 14-15-00067822, RDIMS: 106100691) issued by Transport Canada on 28 April 2015. All operating protocols were approved by Parks Canada and the Wapusk National Park Board prior to the commencement of operations in Wapusk National Park. Further, the team had an approved institutional animal care and use permit (A3917-01, Protocol 1505-2) and a UND UAS Research Compliance Committee Permit (approved 10 April 2015) issued through the University of North Dakota.
Specific aims of the unmanned aircraft project in 2015 were to:
- Determine optimal flight scenarios to minimize impacts on birds but allow the resolution of imagery necessary to count birds and assess habitat characteristics.
- Evaluate the effectiveness of a UAS for counting nesting birds (snow geese and common eiders) compared to historical ground counts.
- Determine what information could be obtained from UAS imagery to evaluate plant phenology and vegetation damage caused by snow geese.
As UASs become more popular for use in wildlife research, there will be an increasing need to validate and verify the accuracy of any collected data (Jones 2006). It will also be important to determine the relationship between historical data collection to modern sampling methods to insure changes are true reflections of biological processes and not sampling bias. To do so it will be necessary to ground truth any imagery collected by a UAS (Vermeulen et al. 2013). The HBP team has conducted circular plot estimates of goose nesting density for the past 20 years that can facilitate such comparisons for snow goose nest estimates.
The imagery collected will allow for a two-way validation of techniques. As the number of nests in a circular plot should be accurately recorded by researchers on the ground, comparisons with imagery collected by aircraft should provide estimates of UAS accuracy. In order to validate the effectiveness of the UAS as a tool, we conducted a double-sampling effort by counting nests on the ground and flying the UAS over 12 circular plots in the Peter’s Rock area. Given the UAS can sample a larger area than just the 50-m radius plot, there is an opportunity to validate the circular plot method density estimates across a much larger landscape than is feasible on the ground by comparing density estimates derived from within the 50m plots with the entire area sampled using the UAS.
Preliminary imagery is proving to be adequate to identify nests due to the high contrast of nest down with the landscape. Identification of white phase geese is relatively easy, but blue phase geese are more difficult to identify (Figure 2). As the UND team continues to collate data from the summer, the validation process at both scales will be determined. Further, methods are simultaneously being developed by collaborating computer scientists to automate the counting process. Thus, comparisons between ground sampling, human counts of imagery, and computer automated methods can be compared.
Figure 2. Imagery of blue and white phase geese collected from Peter’s Rock at 75m. The left side of the photo contains 7 blue phase geese. The right side contains 4 white and 1 blue phase goose.
Behavioral Evaluation – Ground Evaluation
The growing use of UASs in wildlife research has created a need for an evaluation of potential impacts on focal species. One of the projects that coincided with the estimation of nesting density was an evaluation of behavioral impacts on nesting snow geese and common eiders as a result of a UAS surveys. Behavioral impact flights were conducted in such a way that imagery could be collected for nest counts and vegetation assessments simultaneously.
The goal of this study was to determine if flying a UAS above an active nest would result in changed behavior from either parent bird or affect nest success in any way. It has been postulated that UASs may be a source of noise disturbance that may negatively affect wildlife, so noise disturbance factors were also explored (Chabot and Craik 2015).
To answer these questions, nests were located via ground-based searches. At a subset of the nests miniature surveillance cameras were installed which would record continuously (Figure 3) day and night. Twelve lesser snow goose and 10 common eider nests were monitored with nest cameras in 2015, producing 4560 hours of footage. The cameras were miniature security cameras with 24 light emitting diodes (LEDs), camouflaged to blend into the surrounding vegetation, and mounted on a wooden dowel rod approximate 8 – 16 inches above the ground. The cameras were connected by a 25m cable to a waterproof box that housed a DVR that recorded to SD cards and the system was powered by a 12 volt, 33 amp battery. Video is being reviewed for periods at similar times during the day; on days prior to UAS flight operations, during operations, and after operations in order to determine behavioral responses to aircraft operations.
Figure 3. A cryptically positioned 24-hour surveillance camera on a COEI nest. The systems are designed to be low profile and not influence hen or predator behavior.
In addition to nest cameras capturing visual behaviors of the birds, eight nests of each species were also equipped with sound recorders to determine if the UAS could be heard at the nest as it passed overhead, and if vocal responses to the UAS were elicited during flight operations. Eight Wildlife Acoustics Sound Recorders (SM2 and SM3 models) were placed randomly amongst nests and were set to record audio continuously day and night (Figure 4). From the 16 nests monitored with song meters in 2015, we collected 2,832 hours of audio. In similar sampling fashion, audio files will be extracted and spectrograms (Figure 5) will be produced to evaluate how a UAS may impact the vocal behavior of nesting waterfowl and what level of sound the birds may have heard during flight operations.
Figure 4. A lesser snow goose nest equipped with a 24-hour surveillance camera and a Song Meter 3 acoustic recorder from Wildlife Acoustics. The large acoustic recorders are cryptically colored to not attract aerial predators.
Figure 5. A spectrogram showing differences in snow goose vocalizations collected at a monitored nest in 2015. The short peaks at 1:51:15 represent typical “grunting”, whereas the taller high frequency peaks at 1:51:18 are characteristic of honking.
Behavioral Evaluation - Flight Operations
Behavioral study flights were designed in such a way to ensure the aircraft flew transects covering specific nests at a predefined altitude. Flights were conducted at 75, 100, and 120 meters above ground level and randomized over nests (3 nests were flown at each altitude for geese). Three randomly selected nests were monitored for behaviors but were not flown over, serving as a control group. For each flight, we recorded the location of the launch and landing site so we can calculate distance between launch and landing to each nest.
Preliminary results suggest that snow geese appear to notice researcher presence during the initial set up of camera systems, or repeated visits to the nest for data collection or battery replacement. However, we observed several individuals that continued to incubate within 50 meters of the launch and landing site, or returned to the nest shortly after we arrived and remained on the nest during the flight operations from launch to landing. Video is currently under review for specific responses at nest during flight operations. It will be more difficult to quantify the impact on common eider behaviors, as a number of the cameras failed due to unfavorable weather conditions (i.e., heavy rain and wind events) or were depredated before flights occurred.
Vegetation and General Mapping
To assess vegetation, flights were conducted over historical transect lines used to characterize vegetation communities. This area has experienced substantial damage caused by the destructive feeding habits of snow geese, resulting in dramatic changes in vegetation communities (Peterson et al 2013.). A series of 75 ground-truth points for major vegetation cover classes of shrub, water, barren, and graminoids were collected to validate UAS imagery. This dataset will be compared with UAS flights conducted using both a red, green, blue (RCB) camera and a near-infrared (NIR) camera to determine if the UAS could facilitate larger scale vegetation mapping of the damage and potential recovery areas. This method has the added benefit of minimizing time researchers have to be on the ground reducing potential wildlife conflicts, specifically with the presence of polar bears in the area as well as the opportunity to characterize a larger spatial scale of vegetation than can be done on foot.
Similar UAS flights were conducted in the Thompson Point area over 3 historical transects evaluated for vegetation changes relative to snow goose use. Both the RGB and NIR cameras were flown on the UAS for these transects to provide a baseline of vegetation sampling in that area.
The inaugural season of unmanned aircraft flights with the HBP was an overwhelming success in terms of deliverables and personnel achievement. In total 87 flights were conducted, amounting to almost 55 hours of flying time, and over 80,000 images. This operation was conducted over 17 days when flight operation occurred (11 and 6 flight days in June and July, respectively) between 10 June and 16 July 2015.
Use of the UAS this summer was limited by technical failures in equipment, weather conditions unfavorable to flying, and regulations permitting only within line of sight operations. On average, the aircraft battery life was lower than was expected (rated for 50 minutes but averaged about 35 minutes), reducing the amount of ground the aircraft could cover in a single flight. In high winds this problem was exacerbated and overly windy or rainy conditions prevented field operations on several occasions.
- Dr. Robert Newman - Associate Professor & Trimble UX5 Pilot, UND Biology
- Michael Corcoran - Operations Manager and Trimble UX5 Pilot, UND EERC
- Andrew Barnas - Ph.D. Student & Trimble UX5 Pilot, UND Biology
- Christopher Felege - Instructor & Trimble UX5 Pilot, UND Biology
- Robert (Rocky) Rockwell - Professor, Hudson Bay Project Lead, American Museum of Natural History, & City University of NY
- Melissa Gibbons - Resource Conservation Manager, Wapusk National Park, Parks Canada