Building the future of biodiversity monitoring through multimodal AI and digital ecosystem modeling at The Wilds.
We are developing The Wilds as a living laboratory for technology and nature research, supporting long-term transdisciplinary multimodal projects across diverse research areas:
Exotic Endangered Species Research
- Endangered ungulates, rhinos, and carnivores
- Behavior studies and individual identification
- Land-use pattern analysis
Conservation and Restoration Ecology
- Invasive species control and native biodiversity monitoring
- Habitat assessment and management
- Endangered prairie habitat protection
Photo Credit: The Wilds
Our work centers on developing a comprehensive digital twin of The Wilds — this 10,000-acre former strip mine in southeastern Ohio managed by the Columbus Zoo. This digital ecosystem will serve as a powerful tool for conservation planning, enabling researchers and wildlife managers to:
- Test conservation strategies in virtual environments before real-world implementation
- Optimize sensor deployment for maximum ecological insight with minimal animal disturbance
- Predict ecosystem responses to environmental changes and management decisions
- Train and validate AI models for biodiversity monitoring at scale
Our research aims to advance the field of multimodal AI for environmental monitoring by creating datasets and models that can:
- Synthesize information across sensor modalities (visual, acoustic, GPS, satellite, environmental)
- Enable real-time adaptive sampling that responds to ecological events as they unfold
- Support autonomous ecosystem monitoring with minimal human intervention
- Scale conservation efforts through AI-assisted wildlife management
This work establishes a replicable framework for digital twin development in conservation settings worldwide. Our open datasets and methodologies will enable:
- Global conservation applications across diverse ecosystems and species
- Training resources for the next generation of conservation AI models
- Collaborative research platform connecting ecology, computer science, and conservation communities
- Evidence-based conservation through data-driven decision making
This summer represents our first step toward the digital twin vision. We're conducting intensive fieldwork to collect a multimodal dataset focused on a 100-acre area used by Pere David's deer and other species at The Wilds.
Study Area: Single pasture for proof-of-concept (View on Google Earth)
Timeline:
- Field Deployment 1: June 2025
- Field Deployment 2: August 2025
Senor network deployment. Clockwise: satellite, camera trap, AudioMoth bioacoustic monitor, GPS ear-tag Moovement Platform, and LiDAR-equipped fixed-wing drone.
| Sensor Type | Data Collected | Duration | Target Species | Specs |
|---|---|---|---|---|
| GPS Ear Tags | Hourly location + ID | Continuous | Pere David's deer | Moovement Platform |
| Quadcopter Drones | Behavioral videos | 2-3 days, 4 hrs total | Zebras, giraffes, onagers, phorses, African wild dogs | Parrot Anafi, ModalAI Sentinal |
| Fixed-Wing Drones | Aerial photos + LiDAR | Single 2-hour session | Landscape + all species | LiDAR sensor |
| Camera Traps | Motion-triggered photos/video | 1 week continuous | All species (exotic + native) | Various models including GardePro |
| Bioacoustic Monitors | Continuous audio recording | 1 week continuous | Birds, insects, ungulate vocalizations | AudioMoth devices |
| Satellite Imagery | Landscape monitoring | Continuous archive | Vegetation + land use patterns | |
| Weather Station | Environmental conditions | Continuous | Context for all other data |
This summer's fieldwork will specifically address:
- Sensor modality strengths: Which sensors are most effective for tracking land use, animal interactions, and individual identification?
- Spatial-temporal overlap: How well do different sensor types capture the same ecological events?
- Species identification: Can we reliably identify species through vocalizations alone?
- Behavioral patterns: How do animals use space differently throughout the day and season?
- Environmental influences: How do weather conditions affect animal behavior and sensor performance?
We're testing the ICICLE cyber-infrastructure for autonomous monitoring, including:
- Autonomous drone navigation and adaptive flight planning
- Real-time camera trap image analysis and species detection
- Edge AI deployment for on-site data processing
- Sensor fusion techniques for multimodal data integration
We invite collaborators to help shape this research and maximize its impact:
- Access to unique multimodal training datasets with ground-truth labels
- Opportunities to develop novel sensor fusion approaches
- Testing ground for autonomous sampling algorithms
- Real-world validation of computer vision and bioacoustic models
- Biological research questions that can be explored with multimodal data
- Conservation applications of digital twin technology
- Habitat management insights from automated monitoring
- Species behavior analysis across multiple data streams
- Practical applications of digital twin technology for conservation planning
- Cost-effective monitoring strategies using autonomous sensor networks
- Evidence-based management through continuous ecosystem monitoring
- Scalable approaches for large-scale conservation efforts
- Dataset access for research and model development
- Co-authorship opportunities on resulting publications
- Collaborative research with interdisciplinary teams
- Student involvement in cutting-edge conservation technology
Ready to contribute to the future of conservation technology? We welcome:
- Research collaboration proposals and partnership ideas
- Biological research questions that could benefit from multimodal data
- Technical expertise in AI, robotics, and sensor networks
- Conservation applications and real-world deployment scenarios
Contact: Jenna Kline, [email protected]
Website: NatureLab at The Wilds
Funding: ICICLE (NSF AI Institute for Intelligent Cyberinfrastructure) and Imageomics (NSF HDR Institute: Imageomics: A New Frontier of Biological Information Powered by Knowledge-Guided Machine Learning).
Photo Credit: The Wilds