Under Pl@ntNet’s polar star, GUARDEN brings citizen science and predictive machine learning together to support decision-making in biodiversity conservation.
Do plant-spotting, birdwatching, or sky-gazing make you a scientist? Partly! By documenting and recording the presence of a certain plant, bird or celestial body on citizen science platforms, your sightings enrich the databases researchers use to monitor biodiversity, track bird migration patterns, and map the universe. Paired with machine learning and AI, citizen-powered interactive maps can support local decision making in biodiversity governance.
Combining citizen geolocalisation activities with Earth observation (satellite images, time series, soil data, etc.), Deep Learning (training computers to interpret data and find patterns), and predictive hybrid modelling can be a powerful way to track biodiversity loss. The GUARDEN project develops, enhances and engages with citizen science tools, allowing for environmental observation. Ecological data collected and retrieved through these channels can inform and inspire decisions to safeguard critical ecosystems.
By October 2025, GUARDEN will deliver a forward-looking decision support application to help ecology stakeholders expand their holistic understanding of ecosystem functioning and biodiversity loss. Identifying drivers in biodiversity transformation will enhance the assessment of ecological and societal impacts of alternative policy decisions.
Led by CIRAD (French Agricultural Research Centre for International Development), this Horizon Europefunded multi-stakeholder partnership aims to transform bottom-up datasets about ecosystems into evidencebased insights to improve the monitoring and governance of biodiversity. GUARDEN aims to assist decisionmaking by using integrated data to predict future challenges to biodiversity and ecosystems from new pressures and drivers. These drivers may include human disturbance, climate change, or land management.
The GUARDEN project coordinator, Pierre Bonnet, botanist and tropical ecology expert at CIRAD, and project partner Alexis Joly, computer scientist at INRIA, both explain: “Plants don’t move! This makes them key players in characterising local ecosystems”.
Bonnet and Joly are also coordinators of Pl@ntNet, one of the world’s largest citizen science projects for automatic plant identification. They have thus taken on the GUARDEN challenge by building on existing partnerships and expertise on the topic of plant tracking.
Pl@ntNet is both a website and a mobile application with more than 20 million users worldwide which allows people to share the occurrence of plants in given areas through the automated identification of photographs. “Citizen data coming from Pl@ntNet is then merged with data from other European data infrastructures, programmers, and GEO initiatives to feed predictive AI modelling”, explains Joly.
Guided by Pl@ntNet’s experience, GUARDEN aims to refine the current scale, bringing geolocalisation to a higher resolution (up to 100 m2), and provide a set of complementary ecological indicators for local ecological monitoring and management.“
GUARDEN will extend the resolution of existing tools, adding layers and indicators to extract information,” explains Joly – “Beyond plant occurrence, also habitat features (pollinators, birds, mammals, etc.) and species interactions will be monitored to conduct deep species distribution modelling”.
The GUARDEN team is working on a mock-up: the GeoPl@ntNet prototype offers a finer resolution and the chance to access factsheets with multiple ecological indicators representative of a specific area. Scale is an issue, though. According to Bonnet, determining indicator certainty is challenging due to potential geographical bias: “Local indicators may not be shareable with the same level of precision on the European scale”. “The local angle is what makes GUARDEN original and distinctive from other macro-scale geolocalisation tools available” – confirms Bonnet – “A greater awareness of local eco-systems can also influence citizens’ perceptions of local biodiversity management and mitigate reactions to disruptive decisions”.
When it comes to ecological conservation and restoration, conflicting interests are at stake and trade-offs must be considered (e.g., between conservation policies and development agendas). GUARDEN explores these tensions by broadly involving all land management actors in assessing adaptive ecological scenarios.
Four heterogeneous local pilots (including a tropical, non-EU, setting) will address critical challenges in biodiversity governance. Case studies will consider the impacts on biodiversity and ecosystems of urban coastal management actions (in Barcelona), energy infrastructure (expansion of energy centre in Cyprus and construction of windfarms in Greece), and development of transport infrastructures (France and Madagascar).
GUARDEN recognises the complexity of managing raw data and therefore places a high priority on data transparency and trust. This helps face the scepticism due to a dual "prejudice" involving both the perceived contribution of citizen science to research and the use of deep learning models. Joly points to how “even if AIbased methods are deeply present in our everyday activities, mistrust in these tools is a big challenge”.
A responsible and transparent use of AI has the potential to enhance the value of citizen science in the realm of ecology. Safeguarding AI-mediated citizen science may support the development of increasingly precise predictive ecological models and help identify forward-looking biodiversity conservation policies. Promoting the acceptability of these approaches, also through the elaboration of computational standards on the use of hybrid biodiversity data, can boost further research and innovation in the field of biodiversity protection.
This is an article from the Horizon Future Watch Newsletter (Issue 2, July 2023), presented by Foresight on Demand