(All times in CET)
Venue: Radisson Blu Hotel Hamburg (Congressplatz 2 D-20355 Hamburg) Room: Moskau
Online Participants:
Zoom Link
8:00-8:05 AM
Opening remarks by GeoWildLife'23 Workshop Organizers
8:05-9:00 AM
Research scientist at the Department of Evolutionary Ecology
Topic: From conflict to co-existence: Evidence-based solutions for the farmer-cheetah conflict in Namibia
Abstract: Conflicts between humans and carnivores occur worldwide, particularly when humans own livestock.
Developing sustainable solutions is a major challenge, especially for endangered carnivore species.
Namibia is home to several endangered carnivore species, mainly cheetahs. While most of these
carnivore species occur in national parks, 80% of the cheetah population lives on private farmland
where farmers keep their cattle herds. Because cheetahs occasionally kill cattle calves, a farmer-
cheetah conflict has existed for decades, with many cheetahs killed by farmers. In our long-term
project, we used GPS data to show that cheetahs are not evenly distributed across the landscape, but
maintain communication centers that are spatially dispersed. In the communication centers there is
a high level of cheetah activity, making them risk areas for cattle calves. Using also acceleration data
give us the opportunity to identify the behavior of cheetahs, including feeding site. Together with the
farmers, we developed a sustainable solution to substantially reduce cattle losses through adapted
cattle management away from the communication centers. This successful adaptation led to a
reduction in the farmer-cheetah conflict and farmers killing fewer cheetahs.
Bio: Dr. Bettina Wachter is a behavioural ecologist and evolutionary biologist who works since nearly 30
years on carnivores in eastern and southern Africa. She studied biology at the Swiss Federal Institute
of Technology (ETH) in Zurich and did her doctoral thesis on spotted hyenas in the Ngorongogo
Crater in Tanzania at the Max-Planck Institute for Behavioural Ecology in Seewiesen and the
University of Berne. Since 2001, she is a scientist at the Leibniz Institute for Zoo and Wildlife Research
and the head of the Cheetah Research Project in Namibia. She and her team work on free-ranging
cheetahs on farmland inhabited by livestock farmers. The team studies the spatial ecology and
distribution of cheetahs and develops in close collaboration with the farmers evidence based
solutions to mitigate the farmer-cheetah conflict. For this, they capture cheetahs with fully
automated box traps and collect movement data with GPS collars. The team is keen to further
develop field technologies and data analyses. The team is also interested in the reproduction and
health of cheetahs because this is crucial to predict the population dynamics of the species. For a
project overview see www.cheetah-research.org.
09:00-09:15 AM
Paper Presentation: 10-12min, QA: 3-5min
Title: A Generative Trajectory Interpolation Method for Imputing Gaps in Wildlife Movement Data
Authors: Zijian Wan and Somayeh Dodge. University of California Santa Barbara, USA.
TL;DR: Despite the surge in wildlife movement data due to improved tracking technologies, many datasets contain missing records requiring interpolation. This study introduces a generative adversarial network (GAN) combined with long short-term memory (LSTM) layers to interpolate these gaps, addressing the often-overlooked uncertainty of movement data.
09:15-09:30 AM
Paper Presentation: 10-12min, QA: 3-5min
Title: Progress toward automated migratory waterfowl census using drones and deep learning
Authors: Rowan Converse*, Christopher Lippitt*, Grant Harris+, Steven Sesnie+, Matthew Butler+ and David Stewart+. University of New Mexico*, US Fish and Wildlife Service+
TL;DR: Wildlife managers are exploring the use of automated aerial imaging via unoccupied aerial systems (UAS) and deep learning models, specifically convolutional neural networks (CNN), for more efficient waterfowl census in US wildlife refuges. Utilizing crowdsourced annotations for training, the study has identified that a limited number of representative UAS images can effectively train the CNN, with future deployments planned for New Mexico and Texas.
09:30-09:45 AM
Paper Presentation: 10-12min, QA: 3-5min
Title: Towards large-scale spatio-temporal tracking of animal behavior in the wild
Authors: Pranav C. Khandelwal*, Eric Price*, Daniel I. Rubenstein+ and Aamir Ahmad*. Institute of Flight Mechanics and Controls, University of Stuttgart, Germany* and Department of Ecology and Evolutionary Biology, Princeton University, USA+.
TL;DR: Drones are increasingly used for tracking animals, but extracting biologically relevant data from the videos is challenging and time-consuming. This research introduces an easy-to-use workflow with a graphical interface for semi-automatic tracking and classification of animal behavior from aerial footage, streamlining the process for ecologists and conservationists.
09:45-10:00 AM
Paper Presentation: 10-12min, QA: 3-5min
Title: Pipeline for open AIS Data with filtering based on vessel class
Authors: Mirjam Bayer, Tabea Fry, Soren Dethlefsen and Daniyal Kazempour. University of Kiel, Germany.
TL;DR: This study addresses the challenge of deciphering valuable information from massive data streams of vessel activities, especially focusing on the Automatic Identification System (AIS) used for maritime conservation amidst declining fish populations. We introduce a pipeline that efficiently extracts and processes AIS message streams, with a demonstration on the Danish Marine Authority's data. The system can filter and clean the voluminous data, targeting specific vessel classes such as merchandise ships, fishing vessels, or passenger ships, and can further enhance the data with features like water depth and distance to shore.
10:00-10:15 AM
Paper Presentation: 10-12min, QA: 3-5min
Title: Automated Estimation of Distance to Animals in Images: Applications for Monitoring Wildlife Abundance
Authors: Blair Mirka*, Christopher Lippitt*, Grant Harris+, Rowan Converse*, Michael Gurule*, Steven Sesnie+, Matthew Butler+, David Stewart+ and Zoe Rossman*. University of New Mexico*, US Fish and Wildlife Service+
TL;DR:This study presents a streamlined method merging camera traps and distance sampling for precise animal abundance and spatial distribution assessments. Utilizing photogrammetric techniques, the proposed system automatically gauges the target's distance from camera trap images, achieving near-accurate measurements. While currently harnessing crowd-sourced image annotations, this approach sets the stage for full AI automation in detecting and gauging wildlife from camera captures.
10:15-10:30 AM
Paper Presentation: 10-12min, QA: 3-5min
Title: Temporal and spatial pattern of wildlife attacks on human in Chitwan National Park, Nepal
Authors: Saroj Kandel, Thakur Silwal and Shailendra Kumar Yadav. Tribhuvan University, Nepal.
TL;DR: The research analyzed wildlife attacks on humans in Chitwan National Park from 2009 to 2020, revealing an increasing annual trend in such incidents. Among several species, rhinos were most frequently involved in conflicts, while elephants and tigers posed the highest fatality risks to humans. These findings, indicating a higher attack concentration in winter and near human dwellings, are vital for enhancing wildlife management and conservation strategies.
10:30-11:00 AM
Tea Break
11:00-11:40 AM
Wildlife ecologist, Natural Resource Ecology Laboratory, Colorado State University & Founding professor in the Department of Ecosystem Science and Sustainability, Colorado State University.
Topic: Evolutionary programming of animal behaviors using agent-based modeling
Abstract: Agent-based modeling is a flexible and growing method of assessing hypotheses about the traits and behaviors of animals, including their movement, habitat selection, and niche dynamics. Through the process of abduction, we hypothesize how individuals may interact with each other and their environments and assess if those interactions will grow population-level patterns of interest. Through often simple rules attributed to individuals, complex emergent population responses may be seen. Among the approaches that use interacting agents is evolutionary computation, and within that field is evolutionary programming, where the relevant phenotypes of individuals are represented. Individuals compete to maximize scores on some objective function. Poorly performing individuals are removed, and the best performers reproduce related solutions with some mutation. Over time, the solution improves. I use this approach to simulate evolutionary pathways in real-world organisms. I will provide examples simulating the annual migration of wildebeest in the Serengeti and a means of incorporating biotic relationships in niche envelope modeling of future abundances under a changing climate.
Bio: Dr. Randall Boone has been with the Natural Resource Ecology Laboratory, Colorado State University for more than two decades and is a founding professor in the Department of Ecosystem Science and Sustainability. Randall is a wildlife ecologist with training from Oregon State University and the University of Maine. He received his PH.D. in Wildlife Ecology from the University of Maine in 1996. After completing graduate work at the University of Maine, he joined Colorado State University. His experience is diverse, with research in wildlife ecology, agent-based modeling, spatial analyses and GIS, ecosystem modeling, landscape ecology, database management, biogeographical relationships of birds and plants, species/habitat relationships, wildlife and pastoral livestock mobility.
11:40-11:55 AM
Paper Presentation: 10-12min, QA: 3-5min
Title: Smart Camera Traps: Enabling Energy-Efficient Edge-AI for Remote Monitoring of Wildlife
Authors: Dhruv Tulasi, Alys Granados, Prabath Gunawardane, Abhay Kashyap, Zara McDonald and Sunil Thulasidasan. Felidae Conservation Fund, USA.
TL;DR: This study introduces an energy-efficient enhancement to camera traps using edge intelligence, aiming for extended remote wildlife monitoring. Utilizing commercially available components, the system can function for months on a minimal battery and is currently deployed for wild felid monitoring in Northern California. This innovation results from a synergy of conservation initiatives, skilled teams, and an open-source "maker culture" approach.
11:55-12:00
Best Paper Announcement and Closing remarks by GeoWildLife'23 Workshop Organizers