The GAIA i³ approach: Creating an AI-powered early-warning system for ecosystem disruptions

Vultures are a highly intelligent group of birds that provide important ecosystem services. Yet, populations of old-world vultures decreased dramatically in the last decades owing to anthropogenic factors. Efficient conservation strategies that address critical threats such as indiscriminate poisoning or depleted food sources need to be developed. At the same time, their behaviour including social interactions is still poorly understood. Building on high-tech tracking equipment and AI-based analytical tools, GAIA aims at better understanding how vultures communicate, interact and cooperate, forage, breed and rear their young. Additionally, the GAIA scientists research the social foraging strategies of white-backed vultures and the information transfer within carnivore-scavenger-communities. In the animal kingdom it is common across taxa that the search for food is undertaken not only as individuals but in a group. Animals forage together or rely on knowledge from other individuals to find food. This so-called social foraging presumably yields benefits, for example concerning the amount of food that is found, the size of prey that can be hunted or the time required to access food. GAIA investigates species-specific mechanisms in behaviour and communication as well as the incentives, benefits and possible disadvantages for individuals.

By further understanding these intra- and inter-speficic connections and interactions, GAIA also contributes to better understanding roots of human-wildlife conflicts (which are often connected to carnivore behaviour) and to species management. In Namibia for example, research into the lion communities helps understanding their spatial behaviour and mitigating contacts with the local population (e.g. cattle farmers) to manage human-wildlife conflict (GBF target 4). This knowledge is also utilized to observe and manage local lion populations sustainably to benefit people (GBF target 9), balancing conflict mitigation and tourism.

Satellites and aircrafts play a crucial role in gathering environmental data from the distance, helping us to better understand our climate and ecosystems. Remote sensing, often conducted from aircraft, balloons, or satellites, allows us to monitor large areas and remote regions over extended periods. These “eyes in the sky” are invaluable complements to land-based observations, helping us understand ocean and air currents, land cover changes, and climate change. However, animals also possess extraordinary senses and a unique ability to detect changes in their habitats. By combining animal capabilities with remote sensing technologies, GAIA aims at enhancing our ability to monitor and understand our planet. Animals have superior sensory abilities and behavioural strategies that enable them to sense subtle and dramatic changes in their ecosystems, as well as to detect critical incidents. Vultures, for example, act as “sentinel species” and can elevate the concept of remote sensing to new heights. They regularly patrol vast areas in search of food, operating without emissions, additional resources, or repairs. Furthermore, their patrols are guided by their exceptional vision and the mission to find carcasses. The way they patrol, what they search for, and the incidents they lead us to may be linked to specific environmental changes and ecological events.

To fully exploit the potential of vulture-borne remote sensing, GAIA focuses on two essential aspects. Firstly, powerful tracking devices are attached to vultures to monitor their movements and behaviour on detailed temporal and spatial scales. Secondly, new technological solutions are being developed to better understand what the animals observe and do. This includes a newly developed camera tag featuring an integrated camera, artificial intelligence algorithms for behaviour detection and image recognition, and satellite uplink for real-time coverage in remote regions. With these tools, animals can capture imagery and provide data of their surroundings faster, with higher resolution and specificity than satellite imagery. This innovative approach allows us to see nature through the eyes of animals.

GAIA has adopted a minimum waste strategy: Only technical equipment that is absolutely essential is used and developed. Collars and tags either remain for long periods of time (e.g. vultures) or are collected routinely (e.g. lions) to extract data. No transmitters remain in the landscape: If a transmitter drops of or the animal carrying the tag dies, it is located and removed from the landscape. This way, the GAIA system is a “leave no trace” system with significant benefits for the ecosystems.

For ecological research as well as for GAIA use cases, it is necessary to reliably and accurately recognise the behaviour of different animal species over a long period of time in remote wilderness regions. To do this, GAIA scientists have developed and trained an artificial intelligence (AI) that can perform behavioural classification from GPS and acceleration data and tell us exactly what, for example, white-backed vultures fitted with animal tags are doing at any given time and place. This AI will eventually run directly on the GAIA animal tags and generate behavioural information from sensor data. In a second step, the scientists combined the behaviour thus classified with the GPS data from the tags. Using algorithms for spatial clustering, they identified locations where certain behaviours occurred more frequently. In this way, they obtained spatially and temporally finely resolved locations where vultures fed. Last but not least, GAIA is developing an AI for image recognition that will analyse photos taken by the integrated camera of the new tag system. All those algorithms will run directly on the tag and can perform efficient embedded data processing. This also places very special demands on image recognition AI, which must operate particularly sparingly and with small amounts of data. To this end, GAIA teams are developing appropriate strategies and models for sparse AI.

This novel carcass detection pipeline is a key asset in halting species extinction and managing human-wildlife conflicts and therefore aligns with GBF target 4. The pipeline allows for the swift detection of either vultures' death or the death of the animal that the vultures are feeding on. Both scenarious are relevant to halting species extinction: Poisoning at carcasses contributes significantly to the decline in populations of many vulture species. As vultures use social strategies in their search for food, one poisoned carcass can kill hundreds of birds. Scientists from the GAIA Initiative have shown that tagging vultures allows for an early detection of deaths and the carcass to be removed. Tagging vultures and using the AI pipelines described here can substantially reduce further mortalities. Secondly, early detection of poaching incidents of threatened species can put a local full stop to poaching and contribute significantly to combating extinction.

To meet the goal of the GAIA Initiative to develop and put into practice a high-tech early-warning system for environmental changes, a new generation of animal tags is a key component. GAIA teams are working on the hardware and software development of miniaturized animal tags with lowest-power sensor technology with camera and image processing. The tags will be energy-autonomous, optimally adapted to the anatomy of vultures and are the basis for further technological features under development such as on-board artificial intelligences for behaviour detection and image recognition as well as a satellite-based IoT communication system.

Additionally, GAIA is developing concepts of distributed artificial intelligence and networks of micro-processors – animal tags that act just like a swarm. Analogous to natural swarm intelligence, the GAIA initiative is mapping digital swarm intelligence in an ad hoc network of microprocessors. These spontaneously forming networks are the foundation for distributed and sensor-based analysis of large amounts of data. Following this path will make it possible for vulture tags, for example, that are present at the same location during feeding events, to link and share tasks such as artificial intelligence analyses and data transmission.

Relevant ecological processes and incidents that are of interest in environmental change research typically occur in remote areas beyond the reach of terrestrial communication infrastructures. Data generated in the field using animal tags in these regions can often only be transmitted with a delay of days or even weeks. To overcome this delay and ensure no delay in the early-warning system, GAIA develops a satellite communication module for the tags as well as a nanosatellite operating in low earth orbit (LEO): In order to be able to transmit collected data and information directly from the transmitting node to the LEO satellite (Low Earth Orbit), a high-performance satellite IoT radio module will be integrated into the new tags. This guarantees immediate, secure and energy-efficient transmission of the extracted data. The communication system is based on the terrestrial mioty® technology and will be adapted to satellite-typical frequency bands such as L- and S-band for the project. Typical communication protocols, which are sometimes used in the IoT sector, are usually designed for small packet sizes. Further development of the mioty® system will therefore also aim to increase the data rate and message size to enable application scenarios such as image transmissions.

The satellite IoT system will be key for a no-delay communication and thus for an early-warning system. It greatly contributes to the GAIA system in achieving GBF target 4 "Halt Extinction, Protect Genetic Diversity and Manage Human-Wildlife Conflicts". 

Modern Zoological gardens and aquariums worldwide provide unique opportunities by contributing expertise in animal care, species conservation, and public education, forming a strong foundation for modern conservation and scientific research. By working closely with these institutions and utilizing the data and insights they generate, the GAIA Initiative aims at bridging the gap between in-situ and ex-situ conservation efforts. Animals under human care can serve as valuable models for understanding species’ biology, behaviour, and responses to environmental changes. Furthermore, the controlled conditions of zoological gardens allow for the development and testing of advanced technologies, such as animal-borne sensors and AI systems, under more predictable and accessible settings before deployment in the wild.

Key focus areas of this building block include:

• Generating reference and training data for the development of the AI pipeline for the sensor data. By deploying the tags on vultures in captivity in a large aviary and recoding their behaviour simultaneously, we were able to create a paired dataset for the training of the AI.  With the trained AI there is no more need to observe the animals to detect relevant behaviour, e.g. feeding; the AI can very reliably predict behaviour from the sensor data giving us insights in the behaviour of the target animals throughout their life.
• Education and public engagement: Zoo Berlin integrates GAIA’s findings into its educational programs and collaborates in media relations and public outreach, fostering public awareness and participation in biodiversity conservation and technological innovations. Visitors are introduced to cutting-edge tools and their impact on wildlife conservation.

Having minimal and only strictly necessary impact on individual animals is a key goal of the GAIA Initiative. For both lions and vultures, there were extensive testing procedures conducted (within the German system of Animal Testing and Animal Welfare) in Berlin Zoo and Berlin Tierpark. Techniques were developed and tested by veterinary experts for both zoo animals and wildlife and are considered safe and compatible with strict animal welfare considerations. Additionally, both within GAIA and by other research groups there is long-term experience and data on the effects of tagging and collaring of the respective species. It has been proven, for example, that tagging vultures does not have any detrimental effects on the birds’ well-being, health or reproduction. Vulture were found to live many years with tags, to have similar movement and foraging behavior, and to have offspring.

The partnership of GAIA with the Zoo Berlin also emphasises the communication and knowledge transfer objectives of the Initiative in the sense of the GBF target 21 "Ensure that Knowledge is Available and Accessible to Guide Biodiversity Action". This field of activitey not only targets the wider public for raising awareness for biodiversity conservation and technological innovations, but also aims at political decision makers on national and international levels. GAIA has been very active in consulting with political stakeholders in Germany and Namibia for example as well as participating in the IUCN Regional Conservation Forum 2024 in Bruges, Belgium.

The GAIA Initiative conducts important capacity building measures as the developed early-warning system is put into practice together with local parks and authorities in many African countries such as Namibia, Mozambique and Uganda. Park staff, officers in relevant authorities and in ministries are trained while implementing the system. This includes empowering local communities to conduct collaring, tagging, and tracking with the GAIA system as well as implementing the early-warning pipeline using the designated frontend. 

Additionally, GAIA staff is actively educating students in various disciplines and research fields to support novel technologies for conservation and life scienes. In the past 6 years more than 250 students successfully partipated in courses conducted by GAIA staff at the University of Namibia in veterinary science and wildlife biology with special focus on, for example, human-wildlife conflict, animal tracking as well as vulture, lion and hyena behaviour. 

Both professional capacity building and student training directly targets local communities to enable them to run the GAIA early-warning system largely with local knowledge and resources alone. This building block puts the GBF target 20 "Strengthen Capacity Building, Technology Transfer, and Scientific and Technical Cooperation for Biodiversity" at the core of the GAIA Initiative as this block is not an addendum to the research and development part of the Inititiative, but a key field of action from the very beginning.