Food biodiversity is vital for human health and the development of sustainable food systems. However, research on neglected and underutilized species is limited by funding, uneven research capacity, and the challenge of balancing ecological, cultural, and public health considerations, requiring innovative prioritization approaches. Using Brazil as a model, this study inventories 369 neglected food species across algae, aquatic fauna, wild terrestrial vertebrates, insects, mushrooms, and plants. A mixed-methods approach, combining expert knowledge and explainable AI (LightGBM and SHAP value analysis), identified key factors for prioritizing species for food composition and consumption studies. The inventory is dominated by plants (29.5%) and wild vertebrates (24.4%), with significant gaps in nutritional data, particularly for algae, insects, and wild vertebrates. Over 36,000 recipes using neglected species were identified. In both food composition (R2: 0.677) and consumption studies (R2: 0.782), recipe number and species occurrence across different states were the most influential features in predicting prioritization. These findings emphasize the role of cultural uses and local accessibility in shaping nutritional research priorities. We urge increased research on neglected species to bridge data gaps and integrate them into food systems, promoting sustainable diets in Brazil and other tropical regions.
Tag: artificial intelligence
Towards a FAIR Knowledge Graph-Based Conversational Agent for Exploring Agroforestry Research
This paper presents a framework for improving access to agroforestry research through a FAIR knowledge graph integrated with a conversational agent. The proposed system aggregates metadata and research outputs from institutional repositories of CIFOR and World Agroforestry (ICRAF), transforming scientific information into semantically enriched and machine-readable knowledge structures. The framework combines knowledge graph technologies, AGROVOC-based semantic annotation, large language models, and graph retrieval-augmented generation to support natural language querying and query-focused summarization. By linking diverse research resources and enabling intuitive access through a conversational interface, the approach aims to enhance the discoverability, accessibility, and reuse of agroforestry knowledge, particularly for researchers and stakeholders in Africa, Asia, and Latin America. The system contributes to advancing FAIR data principles and supports more effective exploration of scientific evidence in agriculture and agroforestry research.
Spatial Dimensions of Existing Conservation Areas and Other Land Uses in the Congo Basin
Remote sensing (RS) techniques and Geographic Information Systems (GIS) for earth observation have significantly improved the ability to map and measure forest composition, estimate height and biomass, detect degradation and deforestation, implement land use planning, track land use changes as well as monitoring the implementation and respect of land development policies at a finer scale. Machine learning approaches further enhance these capabilities by integrating multiple data sources to produce improved maps of forest attributes and track changes over time. Recent advances in remote sensing techniques and artificial intelligence (AI) now enable detailed mapping and modelling of conservation areas especially the demarcation of High Valued Conservation Forest. National governments follow an administrative procedure for national map design, validation, and standardization. National governments together with the national mapping institutions select particular government officials, the private sector, civil society organizations, and foreign partners to produce national maps. On the other hand, scientists use a rigorous methodology that combines cutting-edge tools and peer review approaches. Even if some of the maps mentioned have not undergone the validation process by the various governments of the sub-region, they are widely used or referred to in international negotiations and discussions. Adopting innovative policies on mapping (Indonesia’s one map policy, Rwanda National land registry) and spatial data management (Geoportal for the Region, The INSPIRE Directive) will be very helpful in planning and conserving the Congo Basin. Non the less, Gabon, the DRC and Cameroon have made exceptions by producing and using renowned National Geoportal.
The integration of traditional and Indigenous knowledge with artificial intelligence systems: A literature review for the case of fire management
This review examines concepts and literature at the intersection of traditional and Indigenous knowledge (TIK), artificial intelligence (AI), and integrated fire management (IFM) within ecosystems. The growing importance of fire management is increasingly evident as climate change intensifies ecological, economic, and public safety risks associated with wildfires. Human societies have grappled with fire since the earliest stages of development, and this history is reflected in the substantial body of traditional knowledge related to the use and management of fire within landscapes. At the same time, recent advances in AI have expanded the reach of computational systems into many domains of human–environment interaction, including wildfire monitoring, prediction, and management. The integration of TIK with AI systems for the purpose of IFM is therefore emerging as an area of growing interest that builds upon substantial bodies of work in each of these three domains. This review therefore assesses how TIK and AI may complement one another in supporting integrated fire management, while also identifying the principal challenges, ethical considerations, and knowledge gaps associated with such integration. Although the literature examined spans multiple geographic regions, particular attention is given to Canada, where extensive wildfire-prone landscapes, longstanding Indigenous fire stewardship practices, and increasing institutional interest in integrating Indigenous knowledge with technological approaches to wildfire management together provide a compelling case study.
A global biodiversity use data infrastructure acknowledging indigenous and local knowledge
Many global biodiversity datasets overlook or misrepresent the knowledge of Indigenous Peoples, Local Communities, and Afro-Descendants (IPLCAD). We propose minimum data and metadata standards for a global data infrastructure on biodiversity knowledge and use, co-designed with IPLCAD, including information on language, community attribution and consent, to ensure data traceability and ethical use. This initiative integrates ancestral and academic sciences to advance inclusive biodiversity governance, addressing historical inequities for global sustainability.
Human versus machine: Can generative AI anticipate insect biological control outcomes?
Generative artificial intelligence (AI) could transform evidence synthesis and revolutionize the global scientific enterprise, yet its agricultural applications are understudied. Here, we systematically assess the performance of three web-grounded AI engines (ChatGPT, ScholarAI and DeepSeek) in synthesizing the global literature on biological control of the fall armyworm Spodoptera frugiperda, and benchmark their outputs against a recent, near-exhaustive human review. Though all engines rapidly screened vast literature corpora, they exhibited shortcomings in factual accuracy, reporting reliability and data consistency. In machine-run syntheses, natural enemy prevalence and performance data often diverged from published records while the level of agreement in enumerating top-performing taxa was evenly low. Meanwhile, internal consistency between laboratory and field-level parasitism data for ScholarAI and DeepSeek was similar to that in human-run reviews. All models tended towards faulty data extrapolation, hallucination and data fabrication, and a sporadic exclusion of key species. While autonomous, machine-only efforts accurately capture coarse-grained patterns in natural enemy identity, abundance, and impacts, they carry limited utility for (living) evidence syntheses or rigorous decision-support. Yet, handled with prudence and due human oversight, machine power might eventually revitalize underfunded disciplines and advance nature-friendly farming.