One Page Summary

To address the challenge raised by WWF Romania, we propose a solution that will enable real-time data-driven adaptive management for field conservationists to help reduce human wildlife conflict in an area inhabited by 2393 people (Armenis). The primary information needed is the location, while the current solution is costly and has a low volume and frequency of data. Satellite collars used at the moment record location every 6 hours but data is purged with as much as 2 weeks delay (WWF Ro), not helping rapid response to conflict. A Jung, Kuba study (2015) of satellite collars on free-roaming bison shows they are easily destroyable. 72.9% of collars malfunctioned before their expected deployment schedule (2 years). A Lotek Iridium collar currently used on the bison costs €3400. 4 collars were mounted since 2014 of which 1 is still active, while 2 did not function for more than 6 months (WWF Ro).
Our tools will provide detailed whole-ecosystem situational awareness based on real-time, small, low-cost wildlife trackers, while the MVP is intended for rapid scaling in other conservation initiatives. We will enable reliable and precise real-time monitoring of wildlife movements and detection of spatial patterns or anomalies in behavior and habitat use. Predictions will help conservationists take timely decisions and enable better long-term interventions to resolve HWC. The product development/testing/deployment communicated to attract feedback, peer reviews in co-lab groups such as Wildlabs where WWF is active. Stakeholders including locals, data scientists, biologists, GIS, hunting ground managers and rangers will be involved - building ITC knowledge for approximately 48 individuals. The area is a perfect “laboratory” to develop, test and refine innovative conservation technology-powered solutions strengthening the rewilding work: annual bison transports (100 bison by 2020), a field research programme for students (12 students/year), ecotourism development guides, locals offering products/services during implementation (2 guides and 6 local families). The location is one of Europe’s richest biodiversity areas, where priority species like brown bear, grey wolf, Eurasian lynx and, recently, the European bison live. Romania hosts the second-largest surface of virgin forests in Europe after Russia. This is a unique opportunity to contribute to halting the degradation of habitat at large scale. Bison, largest European herbivore, are a key instrument for naturally restoring degraded areas, maintaining biodiversity by creating corridors for the movement of other animals through its browsing capacity and big homerange (Van Der Vlasakker, 2004). The species is Vulnerable D1, while Bison bonasus Lowland-Caucasian line is Endangered C1+2a(i) (IUCN Red list). A wild bison sub population offers a chance for habitat restoration, but conservation must be sensitive to the needs of local communities. Protected areas can provide localized benefits, but they may also harm local people (Brockington, D. & Wilkie, 2015). To reduce conflict in high biodiversity areas, data-based decision making is essential to systematic conservation planning and effective proactive restoration measures like rewilding (de Vente, J., Reed, M. S., Stringer, L. C., Valente, S. & Newig, 2016). Big data analysis is needed for appropriate zoning in protected areas close to human settlements, help focus agricultural land on selected areas, while minimizing habitat fragmentation and its associated extinction risks (Tilman, Clark, Williams, Kimmel, Polasky, Packer, 2018). We design the hardware, firmware and network for real-time data capture to create data-driven predictions though machine learning (labeling, classification of specific events relevant for the species); this will enable detection of most anomalies and patterns related to habitat use, diet, ethology, food chain composition and threats. On this dataset complex neural networks will be trained on real-time GPS data to accurately predict movements and events, using Python. There is growing evidence of using machine learning and artificial intelligence used in conservation https://www.cais.usc.edu/projects/ai-for-conservation/ Automated animal identification performs at the same 96.6% accuracy level of human volunteers, saving approximately 8.2 years of human labeling effort on a 3.2- million-image data-set https://blogs.mathworks.com/headlines/2017/05/30/deep-learning-aids-wildlife-conservation-by-air-land-and-sea/. The end product for rangers and researchers to respond to pressures that favor free roaming bison from coming down to the villages is the data-crunching dashboard using WWF Globil ArcGIS platform; it will trigger live notifications including sms, based on virtual fencing, anomaly detection in behaviour or travel patterns, reactions to targeted supplementary feeding and suggest actions when some conditions are fulfilled.