Building a scalable WILDlife monitoring system by integrating remote camera sampling and artificial INTELligence with Essential Biodiversity Variables

野生动物监测对于理解、应对和制止当前的生物多样性危机至关重要。生物多样性遥感的最新进展，如相机捕捉、图像分类技术、公民科学平台和机器学习，提供了具有成本效益的野生动物监测；然而，仍然存在一些瓶颈，例如人工图像审查的高成本和缺乏自动化工作流程。这些限制阻碍了我们创新和协调收集和管理生物多样性监测数据的方法和工具以及及时采取养护和管理行动的能力。我们将开发一个可扩展的监测框架，该框架建立在整个数据周期内制定协调和可重复的程序的基础上，从捕获图像到基本生物多样性变量的大数据处理、注释、共享和下游估计。我们将在代表不同生态系统和欧洲生物地理区域的四个研究区域测试该系统。我们将结合相机捕捉、公民科学、人工智能和分层建模来获得对物种种群和群落动态的公正估计，这将使以自动化方式开发空间全面的基本生物多样性变量数据产品和生物多样性变化指标成为可能。我们的项目将生成任何利益相关者都可以重复使用的产品。除其他外，这些产品包括处理图像并将其转换为基本生物多样性变量和其他与生物多样性有关的统计数据所需的信息基础设施，以及为四个研究系统量身定做的四个人工智能系统，这些系统易于使用，可以针对新的物种或系统进行再培训。我们将通过项目专用网站和应用程序实时传播我们的结果，并将通过全球生物多样性信息基金动员数据。我们将提供支持工具和能力建设，以促进应用到其他领域和升级，帮助移动和优化现有数据，并培育相机陷阱项目。我们由五个研究小组组成的经验丰富的联盟包括来自欧洲和其他地区的生态和保护生物学、数学、计算科学和大数据管理的专家。这将确保一个高度跨学科和先进的科学环境，以实现项目目标。我们将联盟的技能与与利益攸关方和政策制定者的坚实合作网络结合在一起，这对于实施和扩大已开发的监测框架至关重要。该项目将通过为科学家、管理者和政策制定者提供生物多样性研究和监测工具，以产生关键知识，支持准确的评估和预测，并实施基于证据的及时管理战略，为欧洲的科学卓越和能力建设做出贡献。

Wildlife monitoring is critical for comprehending, reacting to and halting the current biodiversity crisis. Recent advances in biodiversity sensing, such as camera trapping, image classification technologies, citizen-science platforms, and machine learning, provide cost- effective wildlife monitoring; however, there are still some bottlenecks, such as the high costs of manual image reviews and the lack of automated workflows. These constraints have hampered our ability to innovate and harmonise methods and tools for collecting and managing biodiversity monitoring data and take timely conservation and management actions. We will develop a scalable monitoring framework that builds on elaborating harmonised and reproducible procedures across the data cycle, from capturing images to big-data processing, annotation, sharing, and downstream estimation of Essential Biodiversity Variables. We will test this system in four study areas representative of different ecosystems and European biogeographical regions. We will combine camera trapping, citizen science, artificial intelligence and hierarchical modelling to obtain unbiased estimates of species populations and community dynamics that will enable the development of spatially comprehensive Essential Biodiversity Variables data products and indicators of biodiversity changes in an automatized manner. Our project will generate products that any stakeholder can reuse. These products include, among other things, the informatic infrastructure required to process the images and translate them to Essential Biodiversity Variables and other biodiversity-related statistics, as well as four artificial intelligence systems tailored to the four study systems easy to use and that can be re-trained for new sets of species or systems. We will disseminate our results in real-time through a project-dedicated website and application and will mobilise the data through the Global Biodiversity Information Facility. We will provide supporting tools and capacity building to facilitate the application to other areas and upscaling, helping mobilise and optimise existing data and nurture camera trap projects. Our experienced Consortium of five research groups includes experts from ecology and conservation biology, mathematics, computational sciences, and big data management, from across Europe and beyond. This will ensure a highly interdisciplinary and advanced scientific environment to achieve the project goals. We combine the Consortium's skills with a solid collaboration network with stakeholders and policymakers, which is critical for implementing and scaling up the developed monitoring framework. This project will contribute to European scientific excellence and capacity building by providing scientists, managers, and policy makers with a biodiversity research and monitoring tool to generate critical knowledge, support accurate assessments and predictions, and implement evidence- based and timely management strategies.