Corner image home
Attractants, pattern recognition, and DNA analysis paw
Zoo ResearchAttractants, pattern recognition, and DNA analysis: Combining three new tools to advance clouded leopard understanding and conservation

Dawn Tanner, PhD, Conservation Biology Program at the University of Minnesota

This project aimed to improve the success of remote-cameras sets and collection of genetic material to increase knowledge about population dynamics and help to identify critical conservation areas. Clouded leopards are almost exclusively studied in the wild by using remote cameras and collecting genetic samples through scat and hair. These studies identify presence-absence and estimate abundance of clouded leopards to improve conservation efforts. Genetic samples improve population estimates by providing insight into recent population isolation, gene flow among populations, and diversity. In her current work, Dr. Tanner, identified three areas where specific steps could be taken to improve remote-camera studies of leopards in field settings: the effectiveness of visual and scent attractants, the improvement of preservation of samples for DNA analysis, and the accuracy of pattern recognition software. Dr. Tanner and her team worked with the population of captive clouded leopards at the Smithsonian Conservation Biology Institute (SCBI), Virginia, USA, to test and refine these three objectives.

Dr. Tanner and Jessica Kordell (SCBI) worked with the population of captive clouded leopards at the Smithsonian Conservation Biology Institute (SCBI), Virginia, U.S.A., during August 2011. Scent and visual attractants are used to focus the activity of target animals (i.e. clouded leopards) and may help to overcome some of the animals’ instinctive behavior to avoid human scent and foreign objects, including remote cameras. This project tested five scent and two visual ttractants, which had been used in peer-reviewed studies on wild felids and other carnivores, against a control (no attractant). Scent attractants included: scat from a different clouded leopards, skunk gland and Hawbaker’s wild cat lure, Gusto and dried catnip, Mega Musk, and predator-survey scent disks soaked in fatty acids from USDA Animal and Plant Health Inspection Service Wildlife Services Pocatello Supply Depot. Scent attractants were applied to lamb’s wool placed in clear Pyrex containers. The control treatment contained lambs’ wool only in a Pyrex container.

Visual attractants included a suspended metal pie tin and a group of 8-12 suspended tail or primary wing feathers from wild turkeys. It is possible that feathers contained residual olfactory cues; these cues were considered small compared with the olfactory stimuli of scent attractants made with animal gland derivatives or fatty acids.

Patrick Zimmerman (School of Statistics, University of Minnesota) guided and contributed to the statistical analysis of this work. Turkey feathers and U.S. Department of Agriculture (USDA) scent disks were the best-performing visual and scent attractants (P < 0.05 vs. control). A combination of the two was not significantly different from either individual treatment (P > 0.05). That work was recently published (Tanner and Zimmerman 2012). The next step for this work will be to test optimal attractants in a field setting.

The second objective focused on improving preservation of samples for DNA analysis. Scat and hair samples collected in the field for genetic analyses often result in low rates of success in extracting target DNA. Recent advances in preservation techniques tested with wolf scat have been shown to improve extraction and amplification of target DNA. In this work, Dr. Tanner again worked with Jessica Kordell to collect scat and hair samples at SCBI. In the lab, Amy Luxbacher (Department of Ecology, Evolution, and Behavior, University of Minnesota) worked with the team to adapt a recent wolf DET preservation technique to analyze clouded leopard samples. Mitochondrial DNA was successfully amplified using this technique on both hair and scat samples; microsatellite loci successfully amplified only for hair samples. This technique is simple, inexpensive, and improves detection. Field protocols for hair and scat may be improved by including this preservation step in addition to the traditional preservation in silica gel beads. Further work is needed to determine if the technique can be used effectively for microsatellites with clouded leopard scat.

To improve the accuracy of pattern recognition software the team worked with Lex Hiby (Conservation Research Ltd, Cambridge, UK) who modified the software to best fit the characteristics of clouded leopard coats. The ExtractCompare software was originally developed for tigers and has been adapted for over 20 animals. In original tests, the leopard version of the software was able to distinguish differences among individual clouded leopards, but the match rate was low. Thanks to Lex Hiby’s work the match rate for a high-quality photograph that shows the side of a clouded leopard now correctly matches 100% of the time. This tool can now be used to rapidly identify individuals captured in remote-camera photographs in the field. Dr. Tanner and Hiby built an extensive library of photographs of differing quality for comparison. The library includes optimal photographs of a full, clear side view of a clouded leopard and 2 categories of suboptimal photographs: 1) animals photographed at irregular angles to the camera and 2) photographs that were blurred due to movement of the animals across the camera path. These kinds of photographs are commonly encountered in the field, so the team wanted to know how well the software would perform with known individuals. The software performs extremely well in correctly matching individuals in all 3 categories. To analyze a given photo, the user identifies key body points in the photograph. ExtractCompare fits a 3-dimensional model using these body points. That model is used to extract a standard area of the coat. The extracts are used to search against previous extracts stored in the library. The original image is then displayed with an image of the best matched clouded leopard from the library. The similarity score between the correctly matched images shown in the figure is 1.00. The next best-matching individual from the library had a similarity score of 0.01, suggesting that there is a high degree of confidence in the result.