GUARDEN: Putting Artificial Intelligence and Citizen Science at the Service of Biodiversity
The European project GUARDEN develops digital tools to better observe biodiversity, anticipate its changes, and integrate this knowledge into land-use and spatial planning decisions. It leverages observations from Pl@ntNet and expands the platform with new services for vegetation surveys and species mapping.
Protecting biodiversity often involves balancing multiple objectives: preserving a habitat while developing infrastructure, generating renewable energy without undermining ecosystems, or enabling coastal use while protecting local flora and fauna. To make these decisions, land managers need accurate, up-to-date, and easy-to-interpret information. However, available data are often incomplete, fragmented, or difficult to translate into concrete actions.
Funded by the Horizon Europe program, the GUARDEN project (safeGUARDing biodivErsity aNd critical ecosystem services across sectors and scales) brings together an international consortium of scientists, companies, protected area managers, and public authorities. Its mission is to place biodiversity and ecosystem services at the heart of public policy and planning decisions.
From Observation to Decision-Making
GUARDEN combines naturalist observations, citizen science data, satellite imagery, climate data, and information on habitats and species interactions. Artificial intelligence and ecological modeling are then applied to address practical questions:
- Which species are present in a given area?
- Which protected or invasive species require special attention?
- What impacts could various development projects have?
- What measures could minimize these impacts?
The project also designs decision-support tools tailored for managers, environmental consulting firms, and public authorities: interactive maps, indicators, trend scenarios, and participatory tools. These solutions are co-developed with their end users, bringing together scientists, policymakers, businesses, NGOs, and citizens.
Pl@ntNet plays a dual role in the project:
- A data source, thanks to botanical observations shared by its users and made accessible to research, notably through the GBIF;
- A scientific infrastructure, hosting new services to inventory vegetation and explore potential species distributions.
Identifying Multiple Plants in a Single Image
Pl@ntNet was historically designed to identify a single plant from photographs focused on one individual. However, ecological monitoring addresses a different need: simultaneously surveying all plant species present within a small vegetation sample area, known as a quadrat or plot.
These surveys are crucial for tracking the conservation status of grasslands, dunes, or other natural habitats. When carried out manually, however, they require substantial time and deep botanical expertise. GUARDEN is therefore developing methods to analyze plot photographs containing multiple plants, in order to:
- Detect multiple species in a single image;
- Estimate plant diversity and ground cover;
- Facilitate repeated, standardized surveys over time.
This approach does not replace botanical expertise, but it can accelerate field assessments, increase the number of monitored sites, and better document shifts in vegetation.
Mapping Species Distribution
Within Pl@ntNet, GUARDEN is also developing a tool to map plant distributions. It links ground observations with climate, environmental, and satellite data to predict areas suitable for species presence.
The goal is to generate maps for approximately one thousand species across Europe, prioritizing:
- Species listed under the EU Habitats Directive;
- Invasive alien species.
These maps do not merely show locations where a plant has already been recorded. They provide an estimate of its potential distribution based on relationships between occurrences and environmental variables. They should therefore be interpreted as probabilistic predictions—with inherent uncertainties—rather than exhaustive surveys.
From Species to Habitats
Plant community composition serves as a reliable indicator of habitat type: dry grasslands, wetlands, and Mediterranean forests each host distinct species assemblages. GUARDEN is therefore developing methods to transition from predicting plant occurrences to characterizing entire habitats.
The project notably aims to:
- Jointly model the distribution of numerous species;
- Leverage Sentinel-2 satellite imagery;
- Map habitats and biodiversity indicators.
This work strengthens the scientific scope of Pl@ntNet: beyond identifying individual plants, the platform helps understand where species live, what communities they form, and how they evolve over time.
Open and Participatory Science
GUARDEN is committed to an Open Science approach. Partners develop reusable software and methodologies, adopt international standards to facilitate data interoperability, and publish resources in open repositories such as GitHub, Zenodo, and the GBIF.
The project also demonstrates that citizen participation can go far beyond data collection. Citizens can help identify local environmental challenges, take part in discussions around land-use choices, and contribute valuable local knowledge for territorial management.
A New Role for Pl@ntNet
Through GUARDEN, Pl@ntNet continues its evolution from a plant identification app to a comprehensive infrastructure for plant biodiversity observation and monitoring. The project’s breakthroughs open up several complementary applications:
- Identifying a plant from photographs;
- Surveying multiple species within a vegetation quadrat;
- Mapping potential species distributions;
- Characterizing habitats based on plant communities;
- Translating these insights into actionable conservation indicators.
By bridging citizen science, artificial intelligence, ecology, and cartography, GUARDEN helps make biodiversity data more accessible and directly actionable.
- Duration: November 2022 – October 2025
- Program: Horizon Europe
To learn more:
