Post provided by ALISTAIR HOBDAY (senior principal research scientist, CSIRO Australia), Tim Lynch (senior research scientist, CSIRO, Australia) and Rachael Alderman (wildlife biologist, Tasmanian Department of Primary Industry, Parks, Water and Environment, Australia).
Cameras and wildlife monitoring
Behavioural and ecological research and monitoring of wildlife populations are based on collection of field data. Demographic data, such as breeding frequency, birth rates and juvenile survival, have been critical in understanding population trends for a wide range of species.
Photography has been extensively used by field biologists and ecologists to gather these data and they have been quick to take up improvements in this technology. Many field programmes today use photography either for primary data collection or the communication of results. Advances in digital photography, image storage and transmission, image processing software and web-based dissemination of images have been extremely rapid in recent years, offering ecologists and biologists a range of powerful tools.
Digital imagery has been captured from a wide range of platforms, each of which has various advantages and limitations for biological study. The most remote images are captured from satellite-based sensors, which have been used to assess population abundance of large animals, such as elephant seals, or locate colonies of emperor penguins. Cameras mounted on aircraft can also provide large-scale perspectives but both of these platforms suffer from high cost, operational limitations due to weather, and limited temporal replication. Recent use of drones, while cheaper, still requires a person to be close to the survey location and can only be used in short bursts, typically lasting less than 20 minutes.
Land-based cameras – or those fixed onto animals – can track behaviour closely, but have low sample size as data tends to be collected at the scale of individual or small groups. To improve replication, fleets of remote cameras can be used or multiple images stitched together post hoc to form a montage. However, this increases cost, either for hardware or labour to manually construct panoramas. To date all these camera systems have had limits to their spatial and/or temporal resolution and, therefore, to the number of individuals covered. This restricts biological study at the population level.
Sampling on Mars – and also on Earth
A high-resolution camera system known as “Gigapan” was developed for deployment on NASA’s Mars Rovers by Carnegie Mellon University and Ames Intelligent Robotics Group. It was designed to allow many small images to be sent back from Mars and reassembled on Earth. In the Earth-based version, the Gigapan system allows a standard digital camera to take a tiled sequence of photographs that can then be stitched together into a single high-resolution image.
With regard to biological studies, Gigapan has already been used to study things like vegetation changes over a season and to archive insect collections. The technology is supported by a commercially available system for image export to cloud storage allowing exploration of panoramas online, including zooming into specific points of interest without substantial loss of resolution. With the use of a telephoto lens, panoramas can also be constructed at distance from the study species, minimising external influences on animal behaviour. This means that the camera systems can simulate what a researcher could achieve viewing an area from a vantage point with high powered binoculars.
Albatross and cameras
We have recently begun using Gigapan to study nesting albatross. We have found this tool is well suited for the study of aggregated wildlife populations, in our case nesting shy albatross (Thalassarche cauta) on the remote Albatross Island in Bass Strait, Tasmania.
This colony has been monitored since the early 1980s, with a small research team spending about a week collecting a range of data several times each year. These trips are planned to coincide with key phases in the breeding cycle of the shy albatross. Passage across the short stretch of Bass Strait and transfer from the boat to the island is only possible in mild conditions. Life on the island is typically cold and often wet and windy. To put it bluntly, it’s not exactly a beach holiday in Australia!
Between visits at the start and end of the breeding season, eggs hatch (or don’t), chicks die, adults abandon nests, and chicks fledge, but we have had little insight into the patterns that ultimately dictate overall breeding success each year. In seeking to understand factors influencing breeding success of this threatened species, we needed more information on the whole breeding cycle to supplement the long term, intensive but episodic field trip data. That’s where Gigapan comes in.
Sampling in remote locations
To use these systems in the field we worked with CSIRO Engineering and Technology who ruggedized the Gigapan robotic camera. In layman’s terms, they put it in a waterproof housing, built and inserted a new chip set and wrote new software to manage the power supply and ‘wake the camera up’ every day. The Gigapan robotic camera mount allows a digital camera to take multiple, zoomed images of the colony in an overlapping pattern across a wide field of view. The resulting tiled image has much higher pixel density than can be achieved from a single camera shot. These advantages mean we can now remotely monitor over 100 albatross nests with a single camera deployed for up to six months, taking two panoramic images daily!
Once the camera is retrieved and the images downloaded, we can zoom to each nest and play the entire sequence of images as a movie. This allows for ease of time-series estimates for each nest, as well as for the overall colony. With these time-series, we can get metrics such as incubation shift duration and mean hatching date, and determine the relationship between things like death of individual chicks and environmental variables such as bad weather. Incubation shift duration, for example, provides an indirect measure of foraging trip duration, which is related to climate variability and, ultimately, breeding performance.
This camera system has provided information at a fraction of the cost to deploying researchers continuously on the island for a six month period, minimises disturbance to the albatross, and the sample size (in terms of nests covered) provides suitable statistical power for our variables of interest. Gigapan systems can also be equipped to allow near real time reporting via mobile phone networks, though our mobile phone bill would be extraordinary if we did this!
Education and outreach – valuable use of images
Citizen science* in image collection and analysis is a growing area, as is the use of imagery in educational settings. The images we have collected have provided us with a range of outreach opportunities. We have partnered with an education team to develop a set of school science activities based on analysis of images. Students can use the images to follow the performance of nests over the season, just as we do in our own analysis.
The future
While remotely deployed cameras cannot replace the flexibility or precision that is possible with human observers, they can provide very useful additional data, especially for monitoring changes over time. Camera systems can also capture rare events, allow archiving of multifaceted data that can be reviewed to address additional questions, and can be widely shared and used in supplemental activities. By combining continuous in situ measurements from cameras with direct field observations, greater insights in animal behaviour and ecology are possible.
To find out more about high-resolution camera surveys, read ‘A high-resolution panorama camera system for monitoring colony-wide seabird nesting behaviour’ by Tim Lynch et al. This paper is part of our Wildlife Monitoring Virtual Issue.
*You can find out more about using volunteer data in Rob Robinson’s blog post from Wednesday
Methods Blog 