Conference Agenda

Online wildlife trade poses increasing threats to the conservation of thousands of species globally. However, attempts to quantify online wildlife trade have often focused on a few platforms and taxonomic groups. Here, I will explain how novel methods for automated data collection and filtering can be used to investigate online wildlife trade across digital platforms and taxonomic groups. Specifically, I will focus on explaining how these methods are being used to monitor the online trade in species at high risk of extinction globally. I will also explain how these methods can be used to monitor the trade in species of conservation concern on a more regional scale. Meanwhile, I will describe how these studies can be conducted in full respect of data privacy and data protection concerns according to the European Union General Data Protection Regulation. I will conclude by highlighting what are the main challenges that we are still facing to make progress towards better investigating online wildlife trade and what are the ways forward for research on the topics.

Machine Learning methods are rapidly advancing and finding applications across diverse fields including conservation science. Here we discuss machine learning methods used to investigate the online digital environment in the context of threatened species and wildlife trade. Global biodiversity faces a significant threat from unsustainable wildlife trade, which has found a new venue in digital marketplaces and social media. With vast amount of digital content, there is a growing demand for automated techniques. First, we present an end-to-end pipeline begins from searching and downloading news articles about species listed in Appendix I of CITES and proceeds with implementing natural language processing and machine learning methods to filter and classify the data. News articles are studied with information extracted using a named entity recognition and analysed for details related to price and quantities. Next, we developed machine vision models based on Deep Neural Networks with the aim to automatically identify images of exotic pet animals for sale. We trained 24 neural-net models spanning a combination of five different architectures, three training methods and two dataset types. Further, we developed object recognition models which can help identify specific target products like elephant ivory and pangolin scales, in an image and highlight them.

Birds are one of the most charismatic groups of animals, and the bird watching community being one of the most engaged and expansive groups of amateur nature hobbyists. Birds are also relevant for the larger public, being present in a variety of cultural expressions such as songs, visual arts, films, myths, and religions. Previous studies have investigated which characteristics make birds more attractive, but were mostly focused on surveys among amateur bird watchers, which have specific interest that may not be transferable to the larger public. The emerging field of conservation culturomics (the study of human-nature interactions using digital data) provides us tools to investigate this issue at a larger scale, using data on the online behavior of a massive number of internet users all over the world. In particular, data on the use of Wikipedia, the largest online encyclopedia in the world, can shed light on the interests of people trying to learn more about bird species, in almost any language. In this study, we use Wikipedia pageviews to uncover which morphological, behavioral or ecological traits are associated with bird species that generate greater online engagement. This information can be useful for conservation marketing and educational outreach.

The great accumulation of online data together with advanced Artificial Intelligence (AI) tools, holds much promise for conservation management and policy through conservation culturomics. For example, understanding peoples’ engagements and preferences in nature and protected areas can be greatly aided analyzing social media produced while visiting such sites. Such insights can consequently guide efforts to increase sites' public appeal for visitors and improve their management for both people and nature. Here, we analyzed over 1000 sample images from Instagram that were posted at six dedicated birding sites in Israel. We aimed to identify both manually and with AI, key features of each image, their main attractors, and expressions of visitors' emotions. We analyzed images with automated image classification, object detection, image-to-text analysis, and sentiment analysis. Overall, we found that automated image classification and identification tools can be very useful to identify broad features of both images and sites. AI tools also enabled us to identify attractors and sentiments of people across and within different sites. We further highlighted unique features of manual versus automated image analysis. These results can provide managers and policymakers with efficient tools to enable grounded conservation policy and management decisions regarding nature sites visitation.

Climate change is amongst the most striking environmental challenges of modern times, producing major socio-ecological impacts with economic and conservation repercussions. More dynamic, automated, and social-oriented observatory systems are needed to tackle climate change and consider adequate mitigation and adaptation responses. Social media data has emerged as an opportunity to get insights on which climate phenomena and impacts people perceive of highest relevance and concern. Concurrently, the information shared by social media users may not always align with that from scientific facts, bringing many challenges to climate change policy and decision-making. Here we present ClimateMedia, a project that aims to: understand the extent to which climate change phenomena are reported by social media users; explore how those users perceive its climatic impacts; evaluate how divergent/congruent such reports and perceptions are to the scientific evidence. The project adopts recent advances in artificial intelligence algorithms, namely from Natural Language Processing, to explore textual content about climate change from social media data and the scientific literature. Outputs from ClimateMedia aim to help practitioners to establish appropriate political goals, enhance conservation efforts and foster biodiversity preservation. Ultimately, ClimateMedia serves as a proof-of-concept determining the feasibility of a future development of a social observatory system.

The Mediterranean Sea is a global marine biodiversity hotspot facing a biodiversity crisis. Tackling this crisis effectively and efficiently is hampered by a lack of necessary ecological and social information to guide decision-makers. To fill this gap, we developed a digital observatory with two main objectives - public attention towards and recreational fishing of Mediterranean fishes. First, understanding public attention is key to mobilise political interest, and consequently increase conservation efforts and success. Second, social and ecological aspects of recreational fishing are not well understood due to difficulty in obtaining reliable and comprehensive data. Catches of recreational fishers have a huge, underexplored potential to monitor marine ecosystems. MEDigital integrates two emerging research approaches (conservation culturomics and iEcology) to provide an unprecedented volume of data for Mediterranean fishes. First, we quantified Google search volumes (i.e., a proxy of public attention) in each Mediterranean country for 770 fishes from 2013 to 2023. Second, we assembled a machine learning workflow to automatically extract information about recreational fishing from YouTube. MEDigital will contribute to characterizing social-ecological aspects of the Mediterranean biodiversity crisis in near real-time with special focus on societal attention to species and recreational fishing.