Periodic Reporting for period 1 - GUARDEN (safeGUARDing biodivErsity aNd critical ecosystem services across sectors and scales)
Berichtszeitraum: 2022-11-01 bis 2024-04-30
There are several legal frameworks, strategies and action plans that have been put in place by the global community and the EU to protect nature and restore habitats and species. However, a core challenge in biodiversity governance is the series of trade-offs that emerge across sectors and scales e.g. between conservation policies and development agendas or by adapting region-wide policies at the national or local level. There is therefore a need to improve the capacity of decision making, and research that informs it, to actually assess the synergies and trade-offs among the multiple policy objectives and management decisions towards a more efficient and user-oriented approach.
GUARDEN’s main mission is to safeguard biodiversity and its contributions to people by bringing them at the forefront of policy and decision-making. This will be achieved through the development of user-oriented Decision Support Applications (DSAs), and leveraging on Multi-Stakeholder Partnerships (MSPs). To do so, GUARDEN will make use of a suite of methods and tools using Deep Learning, Earth Observation to augment the amount of standardized and geo-localized biodiversity data, build-up a new generation of predictive models of biodiversity and ecosystem status indicators under multiple pressures (human and climate), and propose a set of complementary ecological indicators likely to be incorporated into local management and policy.
GUARDEN’s main mission is to safeguard biodiversity and its contributions to people by bringing them at the forefront of policy and decision-making. This will be achieved through the development of user-oriented Decision Support Applications (DSAs), and leveraging on Multi-Stakeholder Partnerships (MSPs). To do so, GUARDEN will make use of a suite of methods and tools using Deep Learning, Earth Observation to augment the amount of standardized and geo-localized biodiversity data, build-up a new generation of predictive models of biodiversity and ecosystem status indicators under multiple pressures (human and climate), and propose a set of complementary ecological indicators likely to be incorporated into local management and policy.
We established multi-stakeholder partnerships (MSPs) to capture the degree of influence and level of interest of each stakeholder. Five MSPs were established.
The overall architecture and the technical specifications of the GUARDEN system were defined, detailing functional and non-functional requirements and workflows to enhance biodiversity monitoring and stakeholder engagement through advanced decision support tools and innovative data collection platforms.
An audio-based bird species classifier has been developed, trained and evaluated. This deep learning model is focusing on 585 bird species of importance at the European scale.
A first version of the service for identifying species in plant survey images has been developed and integrated into the Pl@ntNet citizen science platform. This service available in two forms (from a web service, and a web front-end), allows to submit high-resolution images of plant survey, and to obtain in return a series of predictions characterising the community photographed.
We developed several interconnectable independent software components allowing the production of high-resolution predictive models of species compositions and habitats using in-situ observation data, remote sensing, and environmental data.
We achieve the mapping of the conservation status of the orchid family worldwide at a resolution of one kilometer. This research result has been published in Ecological informatics, early this year. Another major achievement, has been the mapping of 10K plant species at a resolution of 50 m allowing the prediction of a set of high resolution biodiversity indicators for the case studies areas.
We worked on the conceptualisation of the socio-ecological system across the four Case Study (CS) areas, in order to ultimately quantify and map future states of biodiversity and ecosystem services at the EU and case studies levels under different policy and climate scenarios.
A first version of The GUARDEN augmented reality visualisation tool has been developed.
The MINKA public participation tool was adapted for use in the case studies, in particular with the explorations of citizen data acquired for the characterisation of coastal beaches in the Barcelona metropolitan area.
The integration of the various components of the GUARDEN system was initiated using the Decision Support Application developed by eBOS.
The overall architecture and the technical specifications of the GUARDEN system were defined, detailing functional and non-functional requirements and workflows to enhance biodiversity monitoring and stakeholder engagement through advanced decision support tools and innovative data collection platforms.
An audio-based bird species classifier has been developed, trained and evaluated. This deep learning model is focusing on 585 bird species of importance at the European scale.
A first version of the service for identifying species in plant survey images has been developed and integrated into the Pl@ntNet citizen science platform. This service available in two forms (from a web service, and a web front-end), allows to submit high-resolution images of plant survey, and to obtain in return a series of predictions characterising the community photographed.
We developed several interconnectable independent software components allowing the production of high-resolution predictive models of species compositions and habitats using in-situ observation data, remote sensing, and environmental data.
We achieve the mapping of the conservation status of the orchid family worldwide at a resolution of one kilometer. This research result has been published in Ecological informatics, early this year. Another major achievement, has been the mapping of 10K plant species at a resolution of 50 m allowing the prediction of a set of high resolution biodiversity indicators for the case studies areas.
We worked on the conceptualisation of the socio-ecological system across the four Case Study (CS) areas, in order to ultimately quantify and map future states of biodiversity and ecosystem services at the EU and case studies levels under different policy and climate scenarios.
A first version of The GUARDEN augmented reality visualisation tool has been developed.
The MINKA public participation tool was adapted for use in the case studies, in particular with the explorations of citizen data acquired for the characterisation of coastal beaches in the Barcelona metropolitan area.
The integration of the various components of the GUARDEN system was initiated using the Decision Support Application developed by eBOS.
The image-based vegetation survey monitoring technology developed within Pl@ntNet is a world first. To our knowledge, no other web service or application of this type exists. Moreover, the great number of species covered (>50K worldwide) makes it an outstanding achievement in terms of generalisability and scalability.
Although the Pl@ntNet mobile application itself is not directly developed and funded by the GUARDEN project, it benefits from the new AI model and developments we conducted in the context of GUARDEN to improve the performance of the identification engine (in particular the migration to POWO checklist and the integration of SSL vision transformers). More than 20 million people around the world have used Pl@ntNet since these new developments, and identified nearly 200 million plants with the new AI model.
The MALPOLON framework is publicly available on Pl@ntNet’s GitHub account with the following current statistics: 9 Watches, 9 Stars, 4 Forks, 9 Pull requests from external developers.
The high resolution maps of species distribution and biodiversity indicators produced within this task are a world first in terms of their resolution (50m), spatial extent (continental Europe) and number of species covered (over 10K).
Minka platform has shared more than 3,500 observations with GBIF, which will potentially contribute to several new studies on local biodiversity.
Although the Pl@ntNet mobile application itself is not directly developed and funded by the GUARDEN project, it benefits from the new AI model and developments we conducted in the context of GUARDEN to improve the performance of the identification engine (in particular the migration to POWO checklist and the integration of SSL vision transformers). More than 20 million people around the world have used Pl@ntNet since these new developments, and identified nearly 200 million plants with the new AI model.
The MALPOLON framework is publicly available on Pl@ntNet’s GitHub account with the following current statistics: 9 Watches, 9 Stars, 4 Forks, 9 Pull requests from external developers.
The high resolution maps of species distribution and biodiversity indicators produced within this task are a world first in terms of their resolution (50m), spatial extent (continental Europe) and number of species covered (over 10K).
Minka platform has shared more than 3,500 observations with GBIF, which will potentially contribute to several new studies on local biodiversity.