From concept to structure: The core concept of MAMBO is that new technology will bring the highest value to society through a co-design process with users and other stakeholders. Too often tools are built without thinking enough about the linkages to existing research infrastructures and products or the long-term sustainability of their use by the often highly variable end-users. The project is therefore structured across three pillars. First, we will identify existing tools, approaches and research infrastructures and the respective user needs (co-design and infrastructures). On this basis, we will build and refine novel tools based on cutting edge technology (tools and technologies). Finally, we will test and demonstrate the implementation of MAMBO’s novel monitoring tools across the EU, integrate and disseminate the outcomes, and assess their pros and cons and relevance to EU policy (Implementation and dissemination).

Taxonomic focus of MAMBO: Our taxonomic focus is intentionally broad, driven by the needs of the EU as expressed through the EU Habitats and Birds Directives, EU Red Lists and the EU list of invasive alien species. Using images and sounds as input, we specifically target rare and threatened species and knowledge gaps in European birds, bats, mammals, plants and insects, including many important pollinators (bees, butterflies, moths and hoverflies). At present there are many pilot projects and field tests for automated (AI-based) monitoring of these taxa, but major knowledge gaps exist as hardly any reliable, well standardised, and cost-efficient data pipeline is currently available for deriving quantitative (data-driven) biodiversity indicators at the European scale.

Species identification using deep learning (objective 1): In the last ten years, automated identification tools have progressed considerably thanks to deep learning techniques and the increasing availability of training data. There are now mobile applications such as Pl@ntNet, ObsIdentify, Seek, Merlin or BirdNet that allow millions of users to identify the species they encounter simply by taking pictures of them or recording the sounds they make. The data collected by this type of application, however, are subject to significant observation bias, in particular because not all species present on the observation site are listed. Data from repeated, standardised surveys, therefore, remain indispensable, if only to correct the biases of models built from opportunistic citizen data. However, the high costs of implementation of these inventories strongly limits the frequency of their realization and their replication incontrasting environments. Passive acoustic monitoring, in particular, is a valuable sampling tool for habitat assessments and the observation of environmental niches which often are endangered. However, manual processing of large collections of soundscape data remains a challenge. Similarly, the recurrent monitoring of plants carried out in the form of protocolized inventories of uniformly sized quadrats (typically 1x1m) is one of the most common approaches used by natural area managers for monitoring over time and along environmental gradients. The automation of this type of approach would transform the capacities of managers, by granting them greater autonomy as well as an unprecedented study capacity. In order to significantly reduce monitoring costs, MAMBO will develop a new generation of automated identification tools aimed at identifying species assemblages directly. In particular, we will develop new deep learning models for the detection and identification of birds in soundscapes with the objective to bridge the acoustic gap between high-quality training recordings and soundscapes with high ambient noise levels which is one of the most challenging tasks in the domain of audio event recognition. In addition, we will also work on the realization of floristic monitoring from cover pictures (quadrat photos in particular). This will require working on new deep learning approaches based on tile classification or instance segmentation utilizing the data collected and models trained within the Pl@ntNet platform.

Structured species monitoring using field sensors (objective 1): Sensors are widely used for many areas of environmental monitoring, solutions for sensor-based ecological monitoring are in their infancy. In theory, cameras have many advantages over traditional approaches. They remove observer bias and data collection and can collect standardised, high-resolution data across long time scales. Cameras can be baited in similar ways to traditional light and pheromone traps or placed over ephemeral natural resources such as flowers, fruits, dung, fungi, or carrion. Cameras can be installed in remote locations using solar, wind or battery power. “Internet of Things”- enabled hardware offers the potential of ‘edge computing’ to implement classification algorithms in the field to provide fully autonomous systems that report detection and classification data back to the user in real time. When harnessed together, camera traps and computer vision can deliver novel high-throughput systems for detection, classification, and enumeration of species and alongside functional metrics such as biomass. Machine listening/audio analysis has a similar potential for biodiversity monitoring. Originating in speech recognition technology, systems now exist for diverse general music, urban and domestic sound. Automated audio sensors are increasingly common in ecological surveys, while smartphone birdsong recordings are widely collated in projects such as Xeno Canto and iNaturalist, and can be automatically recognised (Stowell and Plumbley 2014, Stowell et al. 2018). In practice, there are still many challenges associated with the transition to sensor based monitoring. In MAMBO, we will build on newly developed sensors (by partner AU) for automated monitoring (Bjerge et al. 2021, Høye et al. 2021).

Habitat type classification and mapping (objective 2): Modern deep learning techniques have great potential for modeling species distributions (SDM) and recognizing habitats (Botella et al. 2018). In particular, experiments with convolutional neural networks (CNN) have shown that they have the ability to capture complex information about the spatial structure of the environment and landscape (Deneu et al. 2021). One limitation of CNNs, however, is that their architecture induces an intrinsic inductive bias in the sense that they process the signal only locally (e.g. in the neighbourhood of each pixel), and thus, they cannot accurately model long-distance relationships. They are, therefore, not really suitable for modelling the relationship between habitats at micro and macro-level. In contrast, transformers (Dosovitskiy and al. 2020), which are a different kind of deep learning models, go beyond local processing and exploit long-term dependencies for increased performance. However, their use in geographic information systems is still a largely unexplored but promising research field. Such models should allow segmenting habitats in a much more efficient manner than current methods without requiring expensive field surveys. In particular, MAMBO will build on Inria’s recent work on transformer models (Caron et al. 2021, Liutkus et al. 2021) to segment and classify habitats in high-resolution satellite data (typically sentinel 2 image series) coupled with environmental rasters. The models will be pre-trained in a self-supervised manner on very large volumes of data and then fine-tuned on supervised data to produce habitat type maps (EUNIS typology in particular). For this second step, we will help ourselves with recent results coupling EUNIS types with plant species composition (plant occurrence data being much more numerous than habitat type records).

Assessing habitat condition (objective 3): Being able to accurately assess a habitat’s condition is vital to cost-efficient nature management. Condition indicators vary with habitat but generally are based on metrics that range from describing the presence/abundance of plant and animal species to characterising 3-dimensional vegetation structures and temporal vegetation dynamics. Establishing habitat condition and conservation status remains a challenge. It generally involves in-situ surveying and can be semi-qualitative, and currently, there is no consistency in how condition is measured and reported across Europe. While some of these metrics (e.g. plant height, % bare ground, % woody cover, vertical woody density, individual plant abundance, disturbance feature density) are well suited for a remote sensing (RS) approach, wide-scale condition assessment has been limited so far because available relevant data is either of insufficient spatial detail (i.e. free satellite data) or of limited temporal frequency, limited coverage, costly or not easily accessible and varied across countries and regions (i.e. drone, airborne or commercial satellite data). In addition, national RS-based habitat level condition assessments would require the storage, access and processing of large data volumes. Cloud-based solutions exist, however, so far the focus is on satellite data (e.g. Google Earth Engine, Sentinel Hub Platform) or limited to drone data pre-processing using licensed software (e.g. Agisoft) and the shared open source scripts are mostly research project specific and so difficult to reuse. The lack of easily accessible computational infrastructure (with standardised analysis tools and enhanced processing workflows) for LiDAR and drone data is stifling operational use.

MAMBO will focus on developing condition metrics from airborne LiDAR and drone data. Both offer the high spatial detail required for condition assessment at scales useful to land managers and decision makers and have the potential of delivering metrics that are relevant to a wide range of habitats. These metrics will be identified through a review of field manuals and stakeholder feedback. Critically, MAMBO will evaluate in detail current barriers to airborne LiDAR and drone data use in habitat condition assessments and share its findings with stakeholder groups interested in LiDAR and drone data exploitation. Envisaging a future where LiDAR and drone data are two of many data types (e.g. satellite, in situ, citizen science) processed to deliver metrics through a user-friendly cloudbased virtual lab for biodiversity monitoring, MAMBO will convert the scripts developed by MAMBO’s RS experts into interactive computing workflows and automate workflow deployment. The idea is to produce processing pipelines that will work with locally relevant LiDAR and/or drone data but deliver metrics that are consistent across the EU, enabling both local and EU- wide reporting to the Habitat Directive.

To derive the selected condition metrics from LiDAR and drone data, MAMBO will apply current state-of-the-art methods, but will also develop new methods that can better exploit the large volume and spatially detailed information coming from these types of technology. Specifically, we will apply deep learning directly to raw LiDAR point clouds and preprocessed drone data (RGB, multi-spectral, point clouds) to avoid losing important details when using only statistical approaches. This is particularly relevant for metrics reliant on feature identification (e.g. drainage, erosion features, individual indicator plants, flower heads). We will test and validate our approaches across habitats and countries by working with available (open-access) LiDAR and newly collected drone data on 3 single large and data-rich sites (UK: Salisbury Plain; The Netherlands: Oostvaardersplassen; Denmark: Mols Bjerge), and demonstrate our computing workflows on data available for a German LTER site (Friedeburg).

Using apps to support society (objective 4): Smartphone applications that embed computer vision and machine listening democratise assessments of species and habitats. This empowers new sections of society to create their own understanding of their environment and take action in response. Smartphone applications that make use of these new technologies can also host data and information that help the user to ensure the quality of the data they collect, provide context to the observations they are making, and suggest actions in response to what is learned. In agriculture, this approach is being used by the ESurveyor app (https://assist.ceh.ac.uk/e-surveyor) to provide farmers with a tool for monitoring the wildflower habitats that they create on their land. Apps of this kind could also help to build on existing efforts (e.g. https://www.plantlife.org.uk/everyflowercounts) to encourage gardeners to improve the quality of their green spaces for biodiversity. This has the effect of raising awareness of the issues of biodiversity decline amongst the public, empowering them to make a change using data they collect about their garden, and through many small actions help to reverse declines. MAMBO will provide image classifiers through APIs that can be integrated within a range of user interfaces. This will be facilitated through community engagement, provision of exemplar demonstrations and documentation on how to access (e.g. through APIs) and deploy MAMBO tools.

Demonstration of enabling tools (objective 5): All tools developed as part of MAMBO can be applied throughout Europe. In order to test these tools, we will make use of a series of demonstration sites described below and presented in Fig. 2. Some tools, such as the image and sound recognition models, will be developed on a European scale and will be tested on larger regions by making them available to national biodiversity portals (BE, DE, DK, FR, MT, NL, UK).

• France: French Mediterranean natural reserves, which are based in the Mediterranean Basin world biodiversity hotspot. We will focus on the middle elevation habitats, and temporary swamps to highlight key habitats of the EU Habitat directive. This will be done in close collaboration with the French Mediterranean National Botanical Conservatory, which has rich data on the species and habitats of the territory for which it is responsible and which has already expressed its interest in the type of tools we are considering for plants monitoring in WP3 and WP4.
• Denmark: We will work at a 1.5 km2 nature reserve known as a biodiversity hotspot in Denmark and located within Mols Bjerge National Park. Recently, the site was divided into separate zones with trophic and passive rewilding. A monitoring program for vegetation and arthropods using traditional plot surveys, time-lapse cameras and drone surveys is combined with detailed records of the space use of introduced horses and cattle.
• UK: Salisbury Plain (Natura 2000 site code UK9011102, Special Area of Conservation, Special Protected Area) is the single largest area of chalk grassland in Western Europe and represents 45% of the remaining habitat in the UK. This military training area occupies some 38,000 ha, and of this 20,000 ha are notified as a Site of Special Scientific Interest, which includes 14,000 ha of chalk grassland. UKCEH has led major ecological surveys commencing in 1996, establishing a long-term monitoring network for vegetation within different areas of disturbance and shrub encroachment and undertaking larval web surveys for the Marsh Fritillary butterfly. A 1m 2014 airborne LiDAR survey and a time-series of 0.25m aerial photographic surveys are available.
• The Netherlands: Oostvaardersplassen Nature Reserve (part of National Park Nieuw Land in province of Flevoland, NL), an extensive, 5400 ha large restoration and rewilding site with reed marshland, dry scrubland, swamp forests, grasslands and open water. The area is of great importance to birdlife (breeding birds, migratory birds and wintering birds) and has a focus on rewilding and the effects of large grazing herbivores (horses, cattle, deer). It is a wetland of international importance according to the Ramsar Convention, part of the Natura2000 network (site code NL9802054), and a Special Protected Area (SPA) according to the EC Bird Directive. The protected area is managed by Staatsbosbeheer, the National Forest Service of the Netherlands.
• Germany: Friedeburg is a 16 km² LTER site located in the typical agricultural area of Central Germany with comparably high habitat diversity. It is also part of the Terrestrial Environmental Observatoria (TERENO) of the Helmholtz Association and provides a wealth of long-term monitoring and environmental background data (e.g. vascular plants, wild bees, butterflies, birds).
• Malta: The island of Comino and the surrounding islets are found in the Malta-Gozo Channel and are part of the Natura 2000 network (MT0000017) as a Special Area of Conservation and Special Protection Area. The site is dominated by scrub, phrygana and maquis habitats but supports other habitats of national and community importance (with a total of 10 Annex I habitats being recorded within the island and surrounding islets) and various rare and endangered endemic species. Large-scale habitat restoration of maquis, woodland and inland wetland habitats has also been carried out over the past 4 years.
• Bulgaria: LTER Petrohan is situated in the West Balkan range with a total area of 7192 ha. The region is mountainous, and steep, with deeply cut river valleys and secondary watersheds with the lowest point at an altitude of 350 m, and the highest at 1900 m. While parts of this area are forested, higher-altitude grasslands and meadows are also present. Ecosystem investigations have been carried out since the establishment of the site in 1986. Petrohan falls into the North-Bulgarian climatic zone with a moderate continental climate.

Linking data and models to promote integrated European biodiversity monitoring schemes (objective 6): The development of an integrated European biodiversity monitoring system faces several challenges related to integrating different data streams, taxonomic gaps, geographic gaps and a lack of up-scaling ability due to context dependencies. In particular, rare and other threatened species are less well covered by ‘standard’ large-scale monitoring schemes, which often impede the assessment of their status and trends in relation to changing environmental conditions based on the quantification of their ecological niches, especially at large spatial scales. MAMBO will test the usability of daily and seasonal performance patterns obtained from novel continuous monitoring tools to quantify the ecological niche of selected species and the transferability and upscaling ability of respective mechanistic ecological niche models in space and time.

A comprehensive biodiversity monitoring system also needs to integrate the wealth of citizen science data. They often underpin models that seek to understand or explain changes in species distributions across space and time at national and international levels. MAMBO will result in advances in computer vision and machine listening leading also to increases in the quality and volume of citizen science species occurrence data. As we develop and adopt these new sensing technologies, MAMBO will also develop the necessary connections between the models and the data in parallel so that new data can be effectively used to improve our understanding of the natural world, and so that model outputs can direct data collection efforts to fill geographic data gaps. However, the imminent uptake of computer vision and machine listening tools will introduce their own biases. These may originate in the sensor used, the algorithm selected, or the threshold used to accept or reject the AI classification. MAMBO will capture these metadata, and develop modelling approaches that account for these to ensure the best quality inference from the multiple types of data collected. Further to this, integrated modelling approaches will be designed to allow these new data to be used in combination with historical and future ‘traditional’ datasets to assess long-term changes.

Successfully transferring ecological models in space and time critically depends on the upscaling ability of the models, especially when local relationships between species performance and environmental conditions are dependent on a larger-scale biogeographic context. By combining high-precision local-scale data from multiple European regions, obtained by demonstrating the MAMBO tools, and respective large-scale environmental variation, MAMBO will identify the impact of environmental drivers, their interrelations and in particular their contextdependencies and test the ability and limitations of upscaling and transferring ecological models in space and time. Citizen science data often underpins models that seek to understand or explain changes in species distributions across space and time at national and international levels. We see advances in computer vision and machine listening leading to increases in the quality and volume of citizen science species occurrence data. As we develop and adopt these new sensing technologies we must develop the connections between the models and the data in parallel so that new data can be effectively used to improve our understanding of the natural world, and so that model outputs can direct data collection efforts to fill data gaps.

Efforts over the last decade have seen rapid improvements in models for detecting changes in species across time and space, accounting for the ad-hoc nature of citizen science data (Isaac et al. 2014). These models seek to account for common biases in citizen science data, such as non-random distribution in space and time, changes in detectability over time, and variation in sampling effort between site visits. The imminent uptake of computer vision and machine listening tools will introduce their own biases that models will need to account for. These may originate in the sensor used, the algorithm selected, or the threshold used to accept or reject the AI classification. Capturing these metadata, and developing modelling approaches that account for these is key to ensure the best quality inference from the data collected. Further to this, integrated modelling approaches must be designed to allow these new data to be used in combination with historical and future ‘traditional’ datasets to assess changes over long time periods. The output of models should drive data collection to ensure that new data is collected where it is needed most. Variations of this approach have long existed in citizen science projects; identifying areas on a map where data has not been collected, and targeting these for data collection in the upcoming field season. Recently the DECIDE project has taken this a step further and uses models to explicitly map uncertainty in species distributions, identifying accessible locations where new observations by citizen scientists will have the greatest chance of improving our understanding of species distributions in space. Similar efforts are being made to best inform changes in species over time (Anonymous 2021). This novel concept of ‘adaptive citizen science’ brings modelling and data collection into close alignment, each supporting the other in near real-time. These new methods will be important for directing the deployment of sensors for autonomous monitoring and demonstrating to citizen scientists the value of the observations they make.

Data available to MAMBO: RS - Sentinel-2 multispectral satellite imagery: A wide range of RS data can be acquired and are available for use to map and monitor land cover, use, dynamics and status, landscape features and objects. For habitat type classification and mapping, Mambo will use data from the Copernicus Sentinel-2 mission that since 2015 has been delivering unique 5day 10-20m global multi-spectral imagery, and jointly with the Landsat missions provides the core data source for future national, continental and global scale land cover mapping (e.g. Sentinel-2 Global Land Cover (S2GLC) http://s2glc.cbk.waw.pl) and monitoring (e.g. the Copernicus for Natura2000 http://www.cop4n2k.eu).

RS - Airborne LiDAR: Across Europe, an increasing number of countries (Denmark, UK, The Netherlands, Finland, Italy) are performing and sharing nation-wide or regional airborne surveys with light detection and ranging (LiDAR) technology, providing meter resolution terrain and habitat/vegetation structure information. Some countries (e.g. Denmark and the Netherlands) have already conducted repeated surveys, thus providing multi-temporal LiDAR data for monitoring biodiversity and habitat change. MAMBO will tap into the LiDAR repositories of the UK, The Netherlands and Denmark.

RS - Drones: Drones and drone sensor technology are evolving fast and their very high spatial resolution (mm to cm) data has the potential to transform local site monitoring. Various drone-sensor setups exist. MAMBO will focus on the most prolifically used and affordable rotary drones with red, green, blue and near infrared cameras. When these are flown to generate high overlap between image frames, structure from motion delivers 3-dimensional point clouds that are very similar to LiDAR point clouds. MAMBO will use both the mm/cm 2 and 3-dimensional drone data.

Relevant international, European and national projects and initiatives are presented in Table 2.

Interdisciplinarity and the integration of social science: MAMBO brings together natural scientists who are involved in habitat and species conservation alongside technology and software specialists to produce tools to help with monitoring, management and delivery of the habitats directives and species action plans. Social science methods and principles will be applied throughout MAMBO, most notably in the early stages of the project in WP1 (User needs and co-development) and towards the end of the project in WP6 (Synergies, gap-filling and optimisation in a European context). By directly engaging with national and international stakeholders (Task 1.4, and Task 1.6.) and exploring how actors interact with one another (Task 1.3.), the project will integrate qualitative research directly into the testing of novel technologies, identifying challenges and opportunities that would be missed by technical assessments alone. In addition, we will assess the economic benefits of the proposed technology in Task 6.4., identifying the opportunities these technologies present for:

• reducing the costs of monitoring in the short and long term,
• delivering better value for money in achieving conservation objectives and
• safeguarding economically valuable biodiversity.

The outputs of this task will include proposals to link monitoring natural capital accounting projects, including the recent Eurostat INCA system of accounts. Collectively, this will contribute to effective policy and practice at the local, national, and European levels and help maximise the uptake of the technologies through the project’s lifespan and beyond.

Gender dimension: In MAMBO the gender dimension will be considered in the co-design, stakeholder mapping, and cost-benefit analyses. Co-design: We will ensure that user needs and tool design considers all kinds of gender to make them equally attractive and intuitive. Stakeholder mapping: Questionnaires and workshop invitations will be screened for unconscious bias in the language, which can affect the outcome due to gender aspects. Cost-benefit analyses: In the context of the cost-benefit analysis, we will specifically examine gender differences in the estimates of costs and benefits associated with novel monitoring tools.

Open Science: MAMBO will address open science practices by making each stage of the developmental process open and accessible. The open access publishing platform Open Research Europe (ORE), enabling the beneficiaries of HORIZON Europe to publish their results after a thorough examination and maximally open process, will be explored for publishing the completed development stage models and documentation. The project will also make use of the highly innovative Open Access journal Research Ideas and Outcomes (RIO), where a special Open Science Collection for MAMBO is envisaged (D7.3). In addition to conventional research papers, the journal welcomes contributions documenting the entire research cycle, including data, models, methods, workflows, results, software, perspectives, and policy recommendations. Moreover, a project collection in RIO is designed to feature all project research outputs that have been published in various journals and/or platforms. The journal only displays the metadata of the article and the link to the original repository, hence fully adhering to copyright regulations. These features allow to centrally collect all project outputs and make these available beyond the project's lifetime.

In order to enforce exploitation, MAMBO will make use of EC’s Horizon Result Platform (HRP) - a tool that enables important exploitable results to be announced and disseminated to investors, stakeholders, politicians, potential business partners, etc. the project will use this platform to publish the project’s Key Exploitable Results (KERs). Beyond the HRP, MAMBO will get support from the Horizon Results Booster (HRB) to better exploit the project activities. HRB includes various services, including clustering of projects for common dissemination, enhanced exploitation plans, business plans for R&I results and commercialisation services for the more mature results. Communication to society at large using social media and easy to digest practice abstracts will be used to make the science accessible for non-scientists.

Data management: To coordinate data management within the project, MAMBO will develop a guiding Data Management Plan (DMP) (D8.3), detailing what data the project will generate, whether and how it will be exploited or made accessible for verification and re-use, and how it will be curated and preserved. The DMP will support the data management life cycle for all data that will be collected, processed or generated by the project. The DMP will be prepared in accordance with the guidelines on FAIR Data Management in Horizon Europe.

The DMP will specifically cover:

• Data types generated and their standards: MAMBO will generate a large amount of data of various types including images, data, code, software, and algorithms. These will be held by MAMBO partners but will adhere to the MAMBO policies, including archiving and preservation policy (see below). MAMBO tools and data will adhere to set data standards (e.g. DANS) and based on activities and funding to WP2 will strive to be Accessible and Interoperable.
• Datasets description: The DMP will provide the dataset metadata specification that will be used in the data registry, following an appropriate standard (e.g. http://www.dcc.ac.uk/resources/metadata-standards). In the cases where the datasets cannot be publicly shared, the reasons will be mentioned in its metadata description (e.g. ethical, rules of personal data, intellectual property, commercial, privacy-related, security-related). This will guarantee that MAMBO data will be Findable.
• Data sharing policies: These include policies for both newly generated data and data and knowledge existing among the MAMBO partners or external parties to be used in MAMBO. The DMP will outline how data can be exploited, shared, or made accessible for verification and re-use. Explanation of policies will be an integral part of MAMBO annual meetings to facilitate data sharing among MAMBO partners.
• Data archiving and preservation: Partners will be responsible for the preservation of raw data, final products will be stored in repositories maintained by the partners or open repositories (depending on data type).
• Open source policies: MAMBO will in principle favour open-source approaches. Some results, however, may be based on proprietary knowledge and components, which would not allow for open-source. In such cases, selfcontained parts of the project modules (including manuals and interfaces) will be made available through open repositories like GitHub or GitLab if possible to allow for further development by third parties.

The HORIZON-CL6-2021-BIODIV-01-02 “Data and technologies for the inventory, fast identification and monitoring of endangered wildlife and other species groups” call states four expected outcomes, and we demonstrate how MAMBO meets these in contributing to:

1) Systemic, integrated and (open-)standardised data, knowledge and models on the conservation status and ecological requirements of species and habitats, with a focus on those covered by the Birds and Habitats Directives and IUCN Red List

Impact: MAMBO will develop integrated species distribution models and habitat extent maps at high resolution, with a focus on the species covered by the Birds and Habitats Directives and IUCN Red List (WPs 4,6). These models and the resulting maps will allow computing automatically some of the IUCN criteria used to evaluate if a taxon belongs in an IUCN Red List threatened category (CRITICALLY ENDANGERED, ENDANGERED or VULNERABLE). Criterion B, in particular, focusing on the geographic range of species and habitats (B1-Area of occupation or B2-Extent of Occupancy) will be directly computable. This automated approach to conservation status determination will not replace the assessment by experts but will greatly facilitate and accelerate their work. This will allow to have a quick and recurrent estimate of the status and to identify the species for which there seems to be an urgent need of more precise assessment and/or revision by the experts. The high transferability of the models used will also allow an estimation of the status of the rarest species for which there are not enough observation data available for the standard procedure.

MAMBO will also develop a deep-learning and remote sensing based habitat type mapping approach, applicable at local, national and European scale (T4.1) which will support Europe wide data integration and standardised scaling. MAMBO will also support the development of standardised data and knowledge on habitat conservation status by developing, and demonstrating common RS derived habitat condition metrics (T4.2; T5.3) and encouraging future standardised processing of airborne LiDAR and drone data, key remote sensing data sources for condition monitoring. Reviewing the challenges of upscaling LiDAR and Drone data processing and tackling it directly by converting scripts into interactive computing workflows, and automating workflow deployment (T4.3; T5.3) will encourage future targeted interventions (investments and focussed efforts) and so speed up LiDAR and drone data use.

Impact indicators: Impact will be measured in terms of feedback from IUCN. The impact of the RS based tools is measured in the feedback from stakeholders, interest in the reviews and demonstrations (number of webinar and workshop participants T5.3), and script and computing workflows usage (access and development on GitHub).

2) Bridging of taxonomic and monitoring gaps thanks to new enabling tools, technologies, fast identification methodologies and integrated monitoring systems across Europe on wildlife species

Impact: MAMBO will result in three key developments addressing taxonomic and monitoring gaps:

• developing AI-powered identification based on sound and images,
• development of insect camera’s capable of monitoring both diversity and biomass and
• creating AI-tools allowing monitoring of vegetation quadrants.

AI based sound and image identification will be made freely available for specialist and citizen scientists throughout Europe and will include species of EU Habitats and Bird directive, EU pollinator initiative, EU Red lists and EU Invasive Alien Species. The tools developed for insect cameras make it possible to deploy these on a larger scale helping to address questions regarding insect decline, biomass decline (impact on food web) and the early detection of invasive and/or pest species. The tools for vegetation quadrants will make monitoring of vegetation more cost-efficient, less dependent on specialists and better standardized allowing for vegetation quadrant monitoring being applied on a larger scale. The tools developed in MAMBO are of importance to governmental organisations, NGO’s and citizen scientists, will be used on a large scale for both plants and animals and will help to increase the monitoring effort while at the same time reducing the cost.

Impact indicators: The impact of the three key tools developed is measured in their use. By the end of the project, we will have image and sound recognition available for all above mentioned groups with at least 30,000 European species included. The algorithms will be open access, the identification service will be available throughout Europe and will be incorporated in at least six different national biodiversity portals, and will be used annually by over 20,000 persons contributing to millions of records on biodiversity. At the end of the project at least 10 national biodiversity portals are willing to participate in a consortium aimed at further developing the identification tools. The software tools developed for the insect cameras (diversity and biomass detection of insects) will at the end of the project be used in at least 200 cameras in over 5 European countries with the aim to develop a European wide insect biomass monitoring. The tools for AI-driven analyses of vegetation quadrants will at the end of the project be freely available and will be used for spin-off projects in >5 European countries.

3) Models upscaling the results of biodiversity assessments to wider areas, based on existing datasets of environmental descriptors

Impact: MAMBO will develop novel modelling approaches that integrate multiple data streams from various monitoring tools (T6.1) and combine them with existing high-resolution environmental data and newly developed metrics of habitat condition (T6.2). These models will be upscaled to provide information on the distribution, status and trends of focal species and entire communities (T6.3), enabling the assessment of less-well monitored species or the identification of geographic monitoring gaps. Results and underlying workflows will be provided to respective citizen science-driven monitoring schemes, such as the European Butterfly Monitoring Scheme or the EU Pollinator Monitoring Scheme to allow for adaptive monitoring approaches. Results will also support local authorities inmonitoring Annex species of the EU Habitats Directives. The development of the novel integrated modeling approaches, together with the upscaling procedures will also have a considerable impact on the scientific community, by enhancing our ability to transfer ecological niche models in space and time.

Impact indicators: Improving predictive capacity via better models means publication in international peer-reviewed scientific press, where methodological advancements and respective results are independently evaluated and verified. Given the envisaged significant improvements in modelling capacity and our consolidated expertise, we expect ground-breaking scientific publications in high impact journals. In addition to these publications, we will further spread our results and engage in dialogue via targeted conference presentations. To boost this process, we will make our publications broadly available as open-source and open-data. Scientific impact will be measured by citation metrics (Web of Science) and Altmetrics (e.g. coverage in blogs, social media) and presentation at majorscientific conferences. Uptake by national and international monitoring schemes will be assessed by the number of novel monitoring sites.

4) Integrative taxonomy of inventory pollinator species (bees, butterflies, moths and hoverflies), soil fauna (mites, springtails, woodlices, millipedes and earthworms) and/or other threatened species groups

Impact: MAMBO forms the linking pin between the large number of partners and initiatives focussed on the conservation of pollinators. It will ensure that information and tools generated as part of the European Pollinator Initiative, the EU Pollinator Monitoring Scheme, SPRING (Strengthening Pollinator Recovery through INdicators and monitorinG), ORBIT and TAXOFLY will be freely accessible at a central location. This will include data on ecology, distribution, taxonomy (checklists), and identification. The latter will be partly generated as part of WP3 of MAMBO. In addition, the training data mobilised for the AI-powered image and sound recognition will be linked to the system giving taxonomist access to 100thousands of images identified by other experts and thus facilitating exchange of taxonomic know-how. A similar infrastructure will be in place for soil fauna and other groups of EU conservation concern. During the project, MAMBO will engaged in cooperation with HORIZON-CL6-2021-BIODIV-01-01 (European participation in global biodiversity genomics endeavours) ensuring that molecular data of European species of conservation concern will be readily accessible. This cooperation is especially important for soil organism as lack of DNA-barcode data is currently hampering the structured monitoring of these groups.

Impact indicators: Impact will be measured by the completeness of enclosed data, the amount of data made available and the use of the data. For pollinators the data linked will consist of at least basic data on ecology, distribution, taxonomy (checklists), the availability of DNA-data and data of relevance to the identification of butterflies, moths, hoverflies and bees of all species found in Europe (< 5,700). This will include images of use to taxonomy and identification of 80 percent of the species based both on field data and collections. For pollinators, a complete overview of available DNA barcodes will be available with a plan to fill gaps in addition. We expect at least 1000 people from all European countries accessing the information available in the portal annually, with user numbers potentially increasing over time.

The main aim of Horizon Europe is to provide the knowledge, tools and infrastructure to contribute to well-being and prosperity in Europe (and beyond). The Horizon Europe Work Programme 2021-2022 “Food, Bioeconomy, Natural Resources, Agriculture and Environment” states eight overall expected impacts, which can be summarized under four impact areas: The work programme wants to:

1. lead the development of the digital and green transformation of Europe,
2. restore Europe’s ecosystems and biodiversity and manage natural resources sustainably,
3. make Europe the first digitally-enabled circular, climate-neutral and sustainable economy, and
4. all that in a more resilient, inclusive and democratic Europe.

MAMBO will contribute to all four impact areas, but mostly to 1, 2 and 3. Related to aim (1): MAMBO tools and data will be FAIR and thus contribute to the digital literacy of Europe. MAMBO will make use of a wealth of data sources (from remote sensing to citizen science and from global and European to national and local data) and integrate them into tools and applications that can become the basis of public and private technologies to enable spatial planning (e.g. biodiversity mapping and recording), wellbeing and health (e.g. identification of disease vectors and development of green space), and green growth (e.g. tools for precision agriculture and monitoring the impact of economic activities). Related to (2): All MAMBO tools will facilitate ecological restoration as its tools make monitoring of all biodiversity (species and habitats, with a focus on EU directive targets) easier and more accurate, provide decision-support for nature conservation and guidance of nature restoration and development. Related to (3): MAMBO image and sound recognition tools will facilitate digitally-supported agriculture and economy. For example, chemical-free crop protection depends on accurate early warning of looming pests and diseases. Image-recognition tools, such as automated insect cameras, will be used to do that. Also, MAMBO tools can be used to evaluate and monitor the impact of economic activities. Not only their harmful effects (e.g. habitat degradation) but also their positive contribution (e.g. afforestation and restoration). Related to (4): Some MAMBO tools will target citizens as its end-users and will enable citizens to sense their environment directly. This will increase citizen participation level (and influence) in environmental, protection, health, and well-being areas.

Impact beyond the work programme: Biodiversity and ecosystem services in a European and global context are in crisis. One-fifth of the world’s countries are at risk of their ecosystems collapsing due to the destruction of the environment and ecosystems. Biodiversity and the ecosystem services that depend upon it face ever greater pressures due to human activity around the globe. Resolving this involves a standardised and easy-to-implement, cost-efficient adaptive monitoring. MAMBO addresses these challenges with a multi-dimensional approach, recognising the need for co-design at all levels. The project builds on an integration of sciences to deliver a set of monitoring tools to identify the root challenges in the loss of biodiversity.

MAMBO will establish a community of practice for novel monitoring technology. By networking with like-minded researchers, we will expand the scope of this field, beyond the main project partners, working towards a global ambition to develop automated biodiversity monitoring methods to generate high temporal and spatial resolution image data for monitoring species and habitats. Knowledge exchange will be invaluable in breaking down the barriers between our domains: AI, ecology, citizen science, and engineering. Effective knowledge exchange activities between the MAMBO partners and relevant stakeholder communities will take place at regular intervals to ensure user needs are considered at every stage and the uptake of tools and technologies developed during the project can be maximised. The coordinated field trials across the EU will help refine best practice guidelines and will stimulate the community of practice in challenging environments. Trials of the automated systems will be run alongside traditional sampling methods and will quantify the relative biases in the datasets from the two methods and assess how data from the two might be combined in spatiotemporal statistical analyses of species populations. We will quantify the labour and material costs of each approach and assess the efficiency of various monitoring schemes. A key MAMBO output will be a series of best-practice guides for species and habitat monitoring using automated methods.

MAMBO is expected to:

• Model the complex interdependencies between biology, and societal structures including governance and community aspiration, national and local policy and actions, financial requirements and social constraints.
• Provide opportunities for new technological innovation to be applied to the biodiversity sector that make such landscapes financially sustainable as well as ecologically stable.
• Provide targeted multi-dimensional support in the project to adopt or to co-develop innovative solutions to biodiversity challenges, and make these portfolios of monitoring solutions available for future upscaling.
• Develop and integrate innovative technologies to underpin, support and enable the upscaling of these actions by providing tools to monitor, measure and enable decision support, ensuring data is used effectively for the transformative restoration of biodiversity.
• Validate both the tools and the assessment frameworks developed in the project within the context of the demonstration sites and regions and communicate across multiple platforms the best practices we have identified.

Contribution to Sustainable Development Goals:

1. Biosphere Dimension: the project will contribute to SDG 15 (Life on Land) by addressing all targets of this SDG, with an impact on the conservation, restoration and sustainable use of terrestrial and inland freshwater ecosystems and their services, the promotion of sustainable land use planning, and the restoration of degraded soils. Importantly, MAMBO will significantly lower the costs for monitoring biodiversity.

2. Societal Dimension: MAMBO will contribute to SDG 11 (Sustainable cities). We will impact a) the capacity to ensure sustainable food production by informing the benefits of implement resilient agricultural practices for biodiversity, b) the maintenance of ecosystems, strengthening capacity for adaptation to climate change, c) the improvement of land and soil quality and the status of agrobiodiversity, genetic diversity of seeds and the crop wild relatives (SDG 2).

3. Economic Dimension: Healthy and resilient ecosystems, characterised by a complete structural and functional diversity of species, are an essential requirement for economic resilience. MAMBO contributes to SDG 8 (economic growth) through technological innovation (AI-based monitoring equipment and tools), economic productivity through entrepreneurship and job creation (e.g. promoting new start-ups using LiDAR, drone and Copernicus EO data), and through promoting sustainable tourism (in biodiverse landscapes).

Contribution to the EU Biodiversity Strategy, European Green Deal and the Habitats Directive:

1. As part of the European Green Deal, MAMBO results will support the implementation of the EU Biodiversity Strategy, for all objectives where the inventory of species is a core aspect. These are related to the following components of the BDS strategy: 2.1 A coherent network of protected areas. This includes a truly coherent Trans-European Nature Network; setting up ecological corridors to prevent genetic isolation, allow for species migration, and maintain and enhance healthy ecosystems; and investments in green and blue infrastructure. MAMBO will contribute to the aim of Member States to demonstrate significant progress in legally designating new protected areas and integrating ecological corridors. On this basis, the Commission will assess by 2024 whether the EU is on track to meet its 2030 targets or whether stronger actions, including EU legislation, are needed, for which results of MAMBO could be made available. Also for section 2.2 “An EU Nature Restoration Plan: restoring ecosystems across land and sea”, it is of importance, in particular for sub-section 2.2.2. “Bringing nature back to agricultural land”, where it is explicitly stated that farmland birds and insects, particularly pollinators, are key indicators of the health of agroecosystems and are vital for agricultural production and food security; and where an aim is to bring back at least 10% of agricultural area under high-diversity landscape features. These include, inter alia, buffer strips, rotational or non-rotational fallow land, hedges, non-productive trees, terrace walls, and ponds, where modern methods for the inventory of species as developed in MAMBO can be integrated, also for section 2.2.10. “Addressing invasive alien species” such methods should be integrated. MAMBO results can also be used for an improved EU-wide framework for the mapping and monitoring of biodiversity and ecosystem services that will be put in place with support from the EuropaBON project and the Biodiversity Partnership under Horizon Europe. It will be fully aligned with the future EU Nature Restoration Law, to be proposed in 2021.

2. For the Habitats Directive MAMBO will enlarge the toolbox for the assessment of species which are typical for the habitats listed in the Annexes of the Directive or even listed themselves (like e.g. some butterfly & moth species).

Requirements and potential barriers to expected outcomes

Stakeholder fatigue, lack of interest or conflicts of interest in participating in MAMBO activities may present a risk to the project’s impact. MAMBO will minimise this risk by following EU guidance for stakeholder engagement and co-design (e.g. http://spiral-project.eu/content/documents.html; http://www.biodiversa.org/702). We will provide training for local teams, introduce various ways for stakeholder engagement and commitment, utilise longstanding partnerships across the stakeholder spectrum and develop approaches accounting for diverse interests, values, benefits and socio-political contexts. Economic concerns driven by the COVID-19 crisis may still override the interest of local, regional and national actors in engaging with the project. However, sustainable development is central to the EU Green Deal and requires reconciling biodiversity and economic welfare. MAMBO will actively support this strategy by demonstrating the direct and indirect values and innovative implementation of its results. Possible changes in the EU and national policies can modify the framework within which MAMBO is aiming to achieve impact. We will continually horizon-scan the MS/EU policy landscape for initiatives that may affect the project (e.g. new or revised incentives or implementation targets between member states). MAMBO views this as a challenge and an opportunity to improve policy impact and decision-making by providing knowledge and methods for the EC, MS and civil society to improve designs of species and ecosystem assessments.

To maximize impacts, MAMBO will actively target its dissemination, exploitation, and communication activities to audiences and stakeholders. MAMBO will tailor various uni- and bi-directional dissemination channels to the needs of the target stakeholders and audiences, eliciting expertise, knowledge, and perceptions from stakeholders as part of the project’s co-design and engagement activities. A preliminary mapping of dissemination tools and channels, their target audiences, and relevant Key Performance Indicators (KPIs) is shown in Table 3 and will be refined in the PEDR. To effectively disseminate project results, specific dissemination tools and channels will be established by the project and existing channels used by MAMBO consortium participants will be fully exploited. MAMBO will engage participants to multiply the effects of the project’s social media channels by using their institutional social media channels. Furthermore, consortium participants will use their existing networks and develop new links, presenting the project and its results at relevant events at local, national, and international scales.

Dissemination tools and targets, measures to maximise impact and Key Performance Indicators (KPIs) are presented in Table 3.

In order to ensure effective exploitation of project results, we will explore the use of Open Research Europe (ORE), the HORIZON Europe open access publishing platform, for publishing our project outputs and documentation. The project will also make use of the highly innovative Open Access journal Research Ideas and Outcomes (RIO), where a special Open Science Collection for MAMBO is envisaged (D7.3). In addition to conventional research papers, the journal welcomes contributions documenting the entire research cycle, including data, models, methods, workflows, results, software, perspectives, and policy recommendations. Moreover, a project collection in RIO is designed to feature all project research outputs that have been published in various journals and/or platforms. The journal only displays the metadata of the article and the link to the original repository, hence avoiding any copyright issues. These features allow to central collection of all project outputs and make these available beyond the project's lifetime. Furthermore, MAMBO will make use of EC’s Horizon Result Platform (HRP) - a tool that enables important exploitable results to be announced and disseminated to investors, stakeholders, politicians, potential business partners, etc. The project will use this platform to publish the project’s Key Exploitable Results (KERs, Table 4). Beyond the HRP, MAMBO will get support from the Horizon Results Booster (HRB) to better exploit the project activities. HRB includes various services, including clustering of projects for common dissemination, enhanced exploitation plans, business plans for R&I results and commercialisation services for the more mature results.

A preliminary list of project Key Exploitable Results (KERs), their novelty, stakeholders, scale and potential outcomes and impact are present in Table 4.

In order to achieve maximum exposure and impact, we will prioritise our communication channels based on the ones that are actively used by our target audiences. With our professional audiences, our focus will be on establishing two-way communication, whereas for the general public, we will adopt the “getting our message out there” mantra (mainly through collaboration with associations and networks, as well as public administration at regional and local levels).

Internal communication: Communication and document exchange tool MS Teams will be used as part of a web-enabled communications and learning platform, to create a chat-like environment that simultaneously eases communication and streamlines information and access to documents into relevant channels to alleviate workload and stimulate fruitful and focused discussions. It will provide a place where the MAMBO project team and collaborators can communicate, share documents and work together. The platform allows for an easy exchange of messages and calls, hence avoiding the need for unnecessary email exchanges.

External communication: MAMBO's external communication strategies will be bi-directional, i.e. not only disseminating project outputs to targeted actor groups and the public at large but also eliciting expertise, knowledge and perceptions as part of the project’s engagement activities. The different Communication & dissemination tools and targets, measures to maximise impact and KPIs are outlined in Table 3.

Knowledge generated will be managed in compliance with the Consortium Agreement (CA), which will be signed at the beginning of the project. The CA will address background and foreground knowledge, ownership, protected third-party components, and protection, use, and dissemination of results and access rights. The principles are:

• Background information and knowledge contributed to the project by each participant will be listed in the CA. When included in the work plan, access to background information will be provided royalty-free to other participants for the implementation of the project’s tasks.
• Results shall be owned by the participant who generated them. Each participant will be responsible for ensuring the fulfilment of their obligations under the GA regarding results by planning with any third parties that could claim rights to them.
• Whenever results have been produced jointly by two or more participants, the ownership of the results will be shared among the participants who carried out the work. The terms of joint ownership, protection, share of ownership, and costs for possible protection will be agreed upon in writing via a joint ownership agreement.
• Each participant will be responsible for examining possibilities to protect results that may be commercially or industrially exploited. When deciding on protection, the participant must consider their own legitimate interests and the interests of the other participants. Participants will ensure that adequate steps towards protection are taken prior to DEC activities, preventing unapproved public disclosure of results, models, tools, and data.
• Access rights to results will be granted on a royalty-free basis for further research, and on fair and reasonable conditions if needed for commercial exploitation.