Reconciling agricultural production with biodiversity conservation. Группа авторов
target="_blank" rel="nofollow" href="#ulink_a5bb61b4-0825-5894-959d-954cabc15f8c">Battersby, 2010). Compared to other taxa, though, it must be noted that soil biodiversity is still the less known.
Indicators of habitat diversity can be partially derived from the EMBAL monitoring system (e.g. habitat richness, habitat diversity, patch size). Moreover, the Copernicus Land Monitoring Service, part of the European Union’s Earth Observation Programme24, is providing geographical information on land cover and its changes, land use and vegetation state which can be used to derive indicators on habitat diversity and structural habitat characteristics, such as the presence of landscape elements (tree rows, hedges and groups of trees). The LUCAS module on landscape features in 2022 will collect in-situ information on smaller landscape elements in agricultural areas to complement information from Copernicus. Data (small woody features and hedgerows, grass fringes, ditches, small ponds, stone walls and terraces) will be collected on 93.000 points with the purpose to provide quantitative information at EU and Member State level. The potential of the new Sentinel imagery applied to habitat monitoring has still to be fully exploited. It is important to note that the mentioned approaches have their limitations in terms of resolution or density of sampling points, and provide results that are mostly applicable at country level or macro-regions (e.g. LUCAS grasslands, LUCAS soil, EMBAL), and satellite images are constrained by the resolution of the images (both geometric and spectral).
Despite the efforts, the first column in Fig. 2 remains substantially empty. Though projects aiming at the conservation of genetic material do exist, at global (e.g. Svalbard Global Seed Vault, https:// www.seedvault.no/), national (RIBES, the Italian network of seed banks,
http://www.reteribes.it/) and local (https://www.communityseedbanks.org/the-csb-map/) levels, an overview at EU level of the number, amount and geographical distribution of traditional breeds, cultivars, landraces, wild crop relatives, traditional and ancient varieties is not yet available. The FAO reports that ‘although crop wild relatives represent about 13 percent of the world’s gene bank holdings, about 70 percent of such species are still missing’ (FAO, 2018). To a great degree, the quantification of that part of biodiversity that is directly embedded into agriculture and constitutes its core, as well as an insurance for agricultural ecosystems to remain resilient and adaptive, still has to happen.
What analysed so far concerns initiatives in place or planned. Before reaching some conclusions, it is worth looking into the near future and understand which substantial improvements in biodiversity monitoring can be expected from the operational use of new technologies.
10New approaches and technologies
Recent and ongoing technological developments are reshaping our capacity to monitor biodiversity. Digital connectedness, cataloguing, and storage, combined with advanced analytical techniques and fast computation, provide frameworks that can gather observations across large areas. In Europe, LifeWatch ERIC was established as a European Research Infrastructure Consortium25to provide continued support to the scientific community studying biodiversity. LifeWatch ERIC is building virtual, instead of physical, e-laboratories supplied by the most advanced facilities to capture, standardize, integrate, analyse and model biodiversity (Basset and Los, 2012).
Flexibility and scalability allow designing of monitoring networks for specific tailor-made purposes. Importantly, these frameworks allow expert biologists, citizen scientists, as well as relative non-experts to make valuable contributions (e.g. Chandler et al., 2017). Indeed, these developments facilitate contributions by expert volunteers, who have contributed to biodiversity monitoring for centuries (McKinley et al., 2017). Several tools exist. Functionality varies from geo-locating ones’ own observations to sharing and mapping across the globe. The most well-known citizen-science platform in this context is probably iNaturalist (www.inaturalist.org). iNaturalist provides a generic platform to catalogue species. Another example, which started in 2010, is Pl@ntNet (https://plantnet.org).
Pl@ntNet allows a user to identify a flower, plant or tree species by photographing them with a smartphone (Joly et al., 2014). Specific examples relate to identifying weeds, cultivated and/or ornamental plants, invasive species (Botella et al., 2018) or a focus on specific geographic regions.
Using computer-vision-based algorithms relies on massive amounts of good-quality training data. Pl@ntNet for instance is able to identify a wide range of species with increasing accuracy through a novel collection and validation approach. The Pl@ntNet algorithm continuously learns and improves its accuracy. Each new picture that is submitted provides new data to train the algorithm, benefitting from user feedback stating whether the correct species was identified (or not). These developments highlight that ‘automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems’ (Bonnet et al., 2018). This will drastically reduce the time needed to generate significant biodiversity data flows.
Clearly, there are also limitations. For example, essential taxonomic details may (currently) not be visible on a picture. Resembling a Turing test, Pl@ntNet pitted its algorithms against botanical experts (Bonnet et al., 2018). One of the conclusions was the need for details to make certain taxonomic distinctions.
Ethical considerations also arise. How best to safeguard the geo-location of a protected red-listed species – say an orchid notoriously difficult to find? Surveying is thus undergoing changes. There is still a huge potential in Citizen Science and crowdsourcing. Maintaining such monitoring over longer time periods is one of the challenges. Furthermore, tapping this potential requires having procedures to set standards and for thorough quality. This starts at the collection of observations. The detection and monitoring of pollinator communities, for instance, provides guidance and training to guarantee minimum identification skills of contributing surveyors (LeBuhn et al., 2016). Besides building inventories of species occurrence, near real-time observations underpin better process understanding of why and how species move through landscapes and interact with their environment. Voluntary contributions tracking birds, for instance, have revealed that migratory tracks have been shifting due to climate change (Cooper et al., 2014). New technologies also enable fast and efficient ways to gather data. One example is the efficient collection and extraction of information from pictures with street-level cameras (e.g. d’Andrimont et al., 2018).
Green infrastructures critical for biodiversity can now be mapped at relevant temporal and spatial scales. For example, Lucas et al. (2019) used LiDAR to identify linear vegetation elements in a rural landscape. While new approaches have been developed to retrieve in situ data, the satellite remote sensing community, via the Group on Earth Observations Biodiversity Observation Network (GEO BON, https://geobon.org/ebvs) has developed the concept of Essential Biodiversity Variables (EBVs) (Pereira et al., 2013). They provide the first level of abstraction between low-level primary observations and high-level indicators of biodiversity.