TheSecondInternationalFIT 1for1CheetahWorkshop ... ·...
Transcript of TheSecondInternationalFIT 1for1CheetahWorkshop ... ·...
The Second International FIT-‐for-‐Cheetah Workshop
N/a ‘an ku se Wildlife Sanctuary, Namibia. 15-‐16 June 2015 Sponsored by: Chester Zoo, The N/a ‘an ku se Foundation and WildTrack
Caption: From L to R. Juarez Pezzutti, Steve Swann, Luis Heyns, Jenny Noach, Sky Alibhai, Zoe Jewell, Danene van der Westhuyzen, Vincent van der Merwe, Louisa Richmond-‐Coggan, Stuart Munro, Dale Van Bommel, Linda Van Bommel, Rebecca Schoonover, Chavoux Luyt, Tendai Nekatambe
Participants
Name Email Affiliation Country Juarez Pezzuti [email protected] Universidade Federal do Para Brazil Chavoux Luyt [email protected] Stellenbosch University South Africa Linda Van Bommel [email protected] Australian National University Australia Dale Van Bommel [email protected] Software Engineering Australia Louisa Richmond-‐Coggan [email protected] Cheetah Conservation Fund Namibia Jenny Noach [email protected] AfriCat Foundation Okonjima Namibia Luis Heyns [email protected] AfriCat Foundation Okonjima Namibia Steve Swann [email protected] AfriCat North, Etosha Namibia Tendai Nekatambe [email protected] Painted Dog Conservation
Zimbabwe Zimbabwe
Danene van der Westhuyzen [email protected] Namibian Hunters association Namibia Rebecca Schoonover [email protected] Duke University USA Vincent van de Merwe [email protected] Endangered Wildlife Trust South Africa Stuart Munro [email protected] Naankuse Namibia Rudie Van Vuuren [email protected] Naankuse Namibia Marlice Van Vuuren [email protected] Naankuse Namibia Florian Weise [email protected] Manchester Metropolitan
University UK
Zoe Jewell [email protected] WildTrack/SAS/Duke USA Sky Alibhai [email protected] WildTrack/SAS/Duke USA
Background to the workshop
At present, nearly all cheetah monitoring across sub-‐Saharan Africa strongly relies on the use of invasive and expensive
research methods. These, for example, include capture-‐mark-‐release operations where animals are live trapped, fitted
with tracking units and subsequently followed in the field to directly obtain information from observations. However,
the entire process is very costly. For example GPS tracking units may cost up to USD 4,000 per individual, and they are
therefore only useful to study small sample populations. Moreover, it is impractical and unethical to capture and mark
every cheetah in a given free-‐ranging population to make inferences on the species at large. Tracking units are also
restricted to application in adults only and short operational periods of 18-‐36 months. They can cause significant
discomfort to study animals. In addition to these drawbacks, invasive handling and tagging of cheetah also incurs a
safety risks because animals need to be immobilised in remote settings.
Non-‐invasive and more cost-‐effective monitoring options, such as camera traps and FIT, have only insufficiently been
employed thus far. Considering the landscape scale at which cheetah research is carried out in Africa, indirect methods
will be required in future for reliable population assessments -‐ especially in remote areas like Namibia. In this regard,
FIT holds several crucial advantages over other existing research methods. For example, footprint identification can be
conducted opportunistically when tracks of free-‐ranging cheetah are encountered during other activities. The technique
incurs very little financial costs because it can be implemented with a standard digital camera and scale only. It does
also not require specialised training of participants. Thus, the simple FIT protocol for effective data collection can be
rolled out to larger communities of researchers as well as landowners and others, significantly increasing the sample
effort towards population assessments and capturing information from previously un-‐sampled range areas. Finally, FIT
provides multi-‐purpose data for researchers. Not only does FIT offer a reliable assessment of population structure
(including measurements of ID, gender and possibly age), it further also provides a new avenue for monitoring cheetah
movements without the necessities of capture and marking. Most importantly, in Africa’s savannah landscapes, often the
only accessible data from wild cheetah are their footprints along roads and dry riverbeds, and FIT exploits this vastly
under-‐utilised source of information effectively.
Due to its simplicity FIT has potential to be used by many different cheetah stakeholders. These include, but are not
limited to, researchers, private landowners, livestock farmers, nature guides and professional hunters. The data
collection protocol requires no specialised equipment, or training, and can easily be transferred to and applied by large
interest groups. Public science approaches, in which the general public actively participates in data collection for
scientific purposes, are more and more used around the world to yield comprehensive data sets, especially for far
ranging and widespread species. In Namibia, for example, there is considerable interest from the professional hunting
fraternity as well as livestock producers to assist researchers with large carnivore population assessments in order to
devise evidence based conservation and utilisation programmes for these species. Moreover, the public is already
actively supporting a species distribution atlas for all carnivores. The atlas simply works on an open access online link.
Through local initiatives of promotion, we anticipate that Cheetah FIT will also be rolled out at large scales and well
beyond Namibia’s boundaries.
In the first international FIT for Cheetah workshop, held at N/a ‘an ku se in 2011, the different elements of FIT were
presented and participants engaged in discussing how these could best be used for effective monitoring. WildTrack
requested participants collaborate to make further tracks available.
Subsequently additional tracks were collected by participants and added to strengthen the training database. A robust
algorithm resulted. Blind trials gave 100% accuracy in determining individual, sex and age-‐class.
The software was then streamlined within one superstructure, allowing the user to process images, apply the FIT model
analytics and map animal distributions, all within JMP data visualization software, providing a comprehensive one-‐stop
solution for conservation monitoring.
The primary aim of second workshop was provide basic instruction in the use of the new FIT v.1, and then offer the
software free-‐of-‐charge to participants who wanted to use it for conservation monitoring. The workshop was also
designed to provide a forum for discussion and establishing future collaboration between cheetah and other
conservation biologists interested in using non-‐invasive approaches.
This workshop was held at N/a’an ku sê Lodge and Wildlife Sanctuary in Namibia, a country which supports
approximately one third of the global cheetah population. The locality is ideal for this purpose due to its proximity to
Namibia’s international airport. In addition, the Wildlife Sanctuary houses orphaned and rehabilitated cheetah which
can be used for demonstration purposes. The associated wildlife reserve and surrounding farmland areas provide a
realistic opportunity for free-‐range trials and demonstrations utilising wild cheetah tracks.
Workshop Outcomes
Ø Eighteen participants from six countries attended.
Ø WildTrack demonstrated the imaging, data analytical and mapping capabilities of FIT in JMP (Figs 1-‐9 below)
Ø WildTrack gave FIT software to all participants and provided directions for a free 30 day download of JMP
software. Participants were invited to apply to Friends of JMP for a no-‐cost full copy after this expired.
Ø Stuart Munro organized and demonstrated FIT image collection in the field with captive cheetah
Ø Linda Van Bommel presented her work on the effectiveness of FIT for discrimination of individual African lions, and on livestock guardian dogs in Australia
Ø Juarez Pezzutti presented his work on wildlife conservation in the Brazilian Amazon region with specific reference to local indigenous expertise and the potential use of FIT for monitoring jaguars.
Ø Tendai Nekatambe, a Master’s student at the National University of Science and Technology in Zimbabwe, presented her work on conserving Zimbabwe’s Painted Hunting Dogs, with reference to the use of footprints in distinguishing this species from local domestic dogs.
Ø Chavoux Luyt wrote an Android/Windows app. to guide novices through the process of taking images for FIT, as part of an app. for farmers to record signs of wildlife.
Ø Rebecca Schoonover volunteered to coordinate an FIT user-‐group and investigate the best means of making incoming images and data freely available for future development.
Ø Stuart Munro and Jack Somerville from N/a’an ku sê made a video showing the sequence of protocol points for image collection for an upcoming paper on FIT for cheetah in the Journal of Visualised Experiments (JoVE).
Ø Munro/Jewell/Alibhai gave interviews for Naankuse.
Ø All participants engaged in evening discussions on the use of FIT in monitoring strategies.
Ø Danene van der Westhuyzen, representing The Namibian Hunters Association, expressed interest in trialling FIT to provide reliable density estimates to help regulate trophy hunting.
Ø A live analysis of footprints collected from Aiko revealed that his current prints separated clearly those collected 7
years ago (Fig 7. below). This revealed the need for regular updates of individuals within FIT and a new collection from many of the original animals.
Ø Florian Weise, who spearheaded the development of FIT for Cheetah, and is currently writing up his PhD, Skyped
into the meeting to provide background on the project and communicate his hopes for the future.
The next steps for Cheetah FIT
Ø To begin a structured experimental sampling of cheetah ranges across Namibia. Collaborators will include N/a’an ku sê, WildTrack, Duke University and the Polytechnic of Namibia.
Ø To publish an invited paper on FIT for Cheetah in the Journal of Visualised Experiments (JoVE), entitled: ‘Spotting the Cheetah: A non-‐invasive footprint identification technique to classify individuals’. The manuscript is in progress.
Ø To extend the existing collaboration between Duke University, WildTrack and N/a’an ku sê for the long-‐term structured application of FIT in Namibia, and for being rolled-‐out to other range countries in Africa.
Ø To place FIT as the foundation for an innovative suite of non-‐invasive and cost-‐effective tools for monitoring, anti-‐
poaching and wildlife conflict mitigation. N/a’an ku sê, WildTrack, North Carolina State University, Duke University and the Polytechnic of Namibia are already collaborating on this initiative.
The following screenshots demonstrate some of the capabilities of FITv1 as a one-‐stop tool for wildlife conservation, incorporating image optimization and feature extraction, data analytics and mapping all within JMP data visualization software.
Fig. 1. The main menu window of FIT for cheetah, where options are given to select the analytical platform. FIT software is customized for each species.
Fig 2. The Image Feature Extraction window in FIT where measurements are made of the footprint. Many of the features such as resizing and rotation of the image, substrate depth, extraction of derived points etc. are now fully automated in the latest version.
Fig.3 Robust cross-‐validated pair-‐wise analysis based on a customized model developed by Wildtrack forms the core of the analytical process which allows the identification of individuals giving both the prediction for the number of cheetahs and the classification of the trails.
Fig. 4A Fig. 4B Fig. 4. The final output from the Robust Cross-‐Validated Pair-‐wise Analysis is in the form of a cluster dendrogram providing the predicted number of cheetahs and a visual picture of the relationship between the trails. Fig 4A was generated using a 50% (19 cheetahs) holdback and FIT prediction was 19 cheetahs but, more importantly, it can be seen that the classification of the clusters was 100% accurate. In the newer version of FIT, we have introduced a measure of confidence around the prediction and Fig 4B shows the % likelihood of their being 17 cheetahs (25.5%).
Fig. 5. Sex and age-‐class classification can be performed using discriminant analysis in FIT for unknown individuals.
Fig. 6. Examples of test footprints of Jamu and Balu from N/a’ankuse for a blind trial to establish individual identification, sex and age class using FIT
Fig.7. Sex identification in the cheetah using discriminant analysis. This highlights the data distribution for the male Aiko in 2008 (black squares) and the shift for the same individual in 2015 (brown circles).
Fig 8A New home-‐range add-‐in for JMP using alpha hull constructions. This figure shows the minimum convex polygon areas for three animals.
Fig. 8B shows a more realistic home-‐range estimation for one of the animals above, achieved by adjusting the alpha value. The central area is the adjusted home-‐range, still incorporating all the points.
Fig.9. A new ‘Distance’ add-‐in tool allows the user to calculate areas and distances from utilization distribution map