Biodiversity and Artificial Intelligence

Biodiversity and Artificial Intelligence

Artificial intelligence is becoming an increasingly important tool for studying and protecting biodiversity. Advances in machine learning, deep learning, computer vision, bioacoustics, remote sensing and large language models are allowing researchers to analyze ecological information at scales that would have been extremely difficult using human labor alone.

Biodiversity research produces enormous quantities of information. Camera traps can collect millions of wildlife photographs. Acoustic recorders can operate continuously in forests, wetlands and oceans. Satellites and drones repeatedly photograph entire landscapes. Environmental DNA can reveal organisms from traces of genetic material left in water or soil. Citizen-science platforms contribute millions of additional observations. Scientific papers and conservation reports add another vast body of largely unstructured information.

Artificial intelligence can help turn these expanding data streams into usable ecological knowledge. Algorithms can identify animals in photographs, recognize species from calls, detect habitats in satellite imagery, classify environmental DNA, estimate species distributions and identify environmental changes. AI therefore has the potential to reduce one of conservation science's central problems: researchers increasingly have more environmental data than people available to analyze it.

Yet artificial intelligence is not a replacement for ecological science. The usefulness of an algorithm depends on the quality of its data, sampling methods, validation, interpretation and the conservation decisions built around it. A highly accurate model can still produce misleading ecological conclusions if it is applied outside the environments for which it was developed or if its training data poorly represent rare species, regions or ecological conditions.

Automated Species Identification

One of the most developed applications of artificial intelligence in biodiversity research is automated species identification.

Deep-learning computer-vision systems can analyze photographs and identify animals, plants, insects, fish and other organisms. Camera-trap research has been particularly important in developing this capability. Modern camera traps can operate for months in remote environments and generate hundreds of thousands or even millions of images.

Historically, researchers or volunteers had to inspect these photographs manually. Artificial intelligence can perform an initial classification much more rapidly, allowing conservation scientists to concentrate their attention on unusual observations, uncertain classifications and ecological interpretation.

Some systems can do considerably more than identify a species. AI models have been developed to:

• detect animals within photographs;
• distinguish among species;
• count individuals;
• identify behaviors;
• recognize individual animals;
• detect animals in video;
• identify tracks and footprints; and
• combine appearance with information about habitat and geographic location.

Individual animal recognition could be especially valuable for population studies. Instead of physically capturing and tagging animals, researchers may increasingly be able to recognize distinctive individuals from photographs. Such approaches have been tested with giraffes, giant pandas, primates and other species.

Automated identification nevertheless remains difficult when species resemble one another, photographs are poor, animals are partially hidden or training data contain only a small number of examples of rare species.

Camera Traps and Wildlife Monitoring

Camera traps have become one of the clearest examples of the relationship between biodiversity and artificial intelligence.

Motion-activated cameras allow researchers to observe wildlife without remaining continuously in the field. Large networks of cameras can document species presence, behavior, population changes and interactions across protected areas and other landscapes.

AI greatly increases the usefulness of these networks because automated systems can screen enormous image collections. Images containing no animals can be separated from useful observations, while photographs containing wildlife can be identified and classified.

Large-scale systems such as SpeciesNet, MegaDetector and related open-source conservation tools illustrate a broader movement toward models that can be reused across multiple conservation projects rather than requiring every research team to develop its own classifier.

However, camera-trap AI also demonstrates an important limitation of machine learning: models may inadvertently learn the background of a particular camera location rather than the animal itself. Performance can fall when the same algorithm encounters photographs from unfamiliar habitats or camera locations. This makes testing across independent sites essential.

Bioacoustics and Listening to Ecosystems

Artificial intelligence is also changing how scientists listen to biodiversity.

Passive acoustic recorders can continuously capture sounds from birds, frogs, bats, insects, whales and other vocal animals. These recordings may contain thousands of hours of sound, making complete manual analysis impractical.

Machine-learning systems can search recordings for particular calls and increasingly identify multiple species within complex soundscapes. BirdNET is one prominent example of large-scale automated bird-sound identification, while other models analyze bats, whales, frogs and terrestrial wildlife.

Bioacoustic foundation models may eventually allow knowledge learned from large sound collections to be transferred to species and environments represented by relatively little training data.

These technologies could make acoustic monitoring a powerful method for tracking biodiversity through time. Networks of inexpensive recorders may allow researchers to detect changes in species composition, migration, breeding activity and ecosystem condition.

But acoustic environments vary substantially. Weather, vegetation, equipment, background noise and regional differences in animal vocalizations can all influence model performance. Local validation therefore remains necessary.

Satellites, Drones and Remote Sensing

Artificial intelligence is extending biodiversity monitoring from individual animals to entire landscapes and seascapes.

Satellite imagery combined with machine learning can help researchers identify forests, wetlands, coral reefs, vegetation communities and other habitats. Repeated observations can reveal deforestation, wetland loss, habitat fragmentation, ecosystem restoration and other environmental changes.

AI-assisted satellite monitoring has also been explored for detecting large animals, including whales, from very-high-resolution imagery. This could supplement ship and aircraft surveys in remote ocean regions.

Drones provide another level of observation between satellites and ground surveys. High-resolution drone imagery combined with deep learning can locate animals, classify vegetation, identify threatened plants, survey wetlands and monitor habitat restoration.

Future autonomous systems may combine robotics and artificial intelligence so that drones or other vehicles can collect and process ecological information with decreasing amounts of direct human control.

These systems could be especially useful in remote environments where conventional surveys are expensive, dangerous or logistically difficult.

Marine and Freshwater Biodiversity

Aquatic ecosystems are a major area of AI development.

Researchers are applying artificial intelligence to underwater imagery, autonomous vehicles, fisheries monitoring, environmental DNA, satellite observations, acoustic recordings and coral reef surveys.

Machine-learning systems can help identify fish and other aquatic organisms, classify seafloor habitats, detect coral, measure changes in reef communities and analyze enormous collections of underwater photographs and video.

AI can also assist fisheries monitoring. Computer-vision systems installed aboard fishing vessels may identify captured species and help document bycatch. Automated analysis could improve fisheries transparency and provide information more rapidly than conventional manual review.

Marine conservation demonstrates the benefits of combining technologies. Autonomous vehicles, cameras, sonar, environmental DNA, satellites and artificial intelligence can provide different observations of the same ecosystem. Integrating these sources may eventually produce a much more comprehensive understanding of ocean biodiversity.

Environmental DNA and Genomics

Environmental DNA, commonly known as eDNA, is another rapidly developing biodiversity technology that can be strengthened by artificial intelligence.

Organisms leave traces of genetic material in water, soil and other environmental samples. Scientists can collect these traces and use DNA sequencing to determine which organisms may be present without directly observing or capturing them.

AI and machine learning can assist with the enormous computational task of interpreting these genetic datasets. Applications include classifying DNA sequences, analyzing metabarcoding data and linking genetic observations with environmental variables.

Research has begun combining eDNA with satellite imagery, geographic information and machine learning to predict biodiversity across rivers, wetlands and marine ecosystems.

AI-assisted genomics could also contribute to documenting genetic diversity within endangered species, an important dimension of biodiversity that is not visible through conventional species counts alone.

Species Distribution and Ecological Forecasting

Knowing where species currently occur is only part of conservation planning. Scientists also need to understand where species might occur in the future.

Species distribution models relate biological observations to climate, land cover, elevation, soil, water and other environmental variables. Machine learning is increasingly incorporated into these models.

AI-assisted models can help identify:

• suitable habitat;
• biodiversity hotspots;
• potential climate refuges;
• invasive-species risks;
• areas vulnerable to habitat loss;
• ecological restoration priorities; and
• possible future changes in species distributions.

Such forecasts may become increasingly important as climate and land-use change alter ecosystems.

Predictive performance, however, should not be confused with ecological understanding. Conservation decisions often require scientists to understand why a model made a prediction and how uncertain that prediction is. Explainable and interpretable AI methods are therefore becoming increasingly important.

Forests, Wetlands and Ecosystem Restoration

Artificial intelligence is increasingly being applied to ecosystem-level conservation.

Machine learning can classify forests from satellite imagery, identify causes of deforestation, estimate vegetation characteristics and detect ecological disturbance. Similar techniques can map wetlands, measure wetland loss and monitor restoration.

AI can also help locate invasive plants and predict areas where invasive species may spread. Automated wildfire-risk systems may combine environmental observations with machine-learning models to identify areas where fire prevention or management efforts should be concentrated.

In ecological restoration, AI may help identify locations where intervention could produce significant biodiversity benefits. Algorithms can potentially compare many environmental variables and management scenarios more rapidly than traditional approaches.

But choosing a restoration goal is ultimately a social and ecological judgment. An algorithm cannot independently determine what an ecosystem ought to become. Decisions about restoration involve values, historical conditions, community priorities, Indigenous and local knowledge, costs and competing uses of land.

Citizen Science and Artificial Intelligence

Citizen science and artificial intelligence can complement one another.

Members of the public already contribute enormous numbers of wildlife photographs, recordings and observations. AI can help classify these submissions and direct difficult observations toward experts.

This creates a potentially powerful feedback system. People provide observations from many locations, while machine learning makes those observations easier to organize and analyze.

AI-assisted identification tools can also make participation easier for people without formal taxonomic training. A photograph or recording can generate a preliminary species identification that encourages users to learn more about the organisms around them.

At the same time, generative AI creates a new threat to citizen-science databases. Artificially generated wildlife photographs and fabricated observations may become increasingly difficult to distinguish from genuine records. Biodiversity databases may consequently require stronger verification systems, provenance information and human review.

Large Language Models and Biodiversity Knowledge

Large language models represent another emerging application of artificial intelligence in ecology.

Much ecological knowledge remains embedded in scientific papers, environmental assessments, technical reports, historical documents and other forms of unstructured text. Language models can potentially extract species names, locations, ecological relationships and other information from these archives.

LLMs may also help researchers classify scientific literature, identify research gaps, conduct evidence synthesis and interact with structured biodiversity databases.

Multimodal models extend this idea by combining language with photographs, audio and other forms of information. Future biodiversity systems may be able to analyze an image or sound recording while simultaneously using written ecological knowledge to interpret what they observe.

However, general-purpose language models can produce incorrect information while expressing it confidently. Tests involving biodiversity information have found substantial variation in performance among tasks and taxonomic groups. Consequently, fluency should not be treated as evidence of ecological accuracy.

Artificial Intelligence and Conservation Decisions

The greatest value of AI may ultimately come not from detecting species but from improving conservation decisions.

Monitoring systems can produce information about where species occur and how habitats are changing. Decision-support systems can then help researchers and managers evaluate possible interventions.

Potential applications include:

• designing protected areas;
• prioritizing habitat restoration;
• planning wildlife corridors;
• allocating ranger patrols;
• identifying invasive-species risks;
• monitoring illegal wildlife exploitation;
• reducing fisheries bycatch;
• evaluating agricultural conservation measures; and
• identifying areas vulnerable to climate change.

Artificial intelligence can evaluate large numbers of possible scenarios, but conservation decisions remain fundamentally human decisions. Algorithms can provide evidence and predictions; governments, communities, scientists and land managers must determine objectives and acceptable trade-offs.

Risks, Bias and Uncertainty

The rapid expansion of AI creates significant scientific and ethical challenges.

One of the largest problems is biased data. Biodiversity observations are not evenly distributed across the planet. Some countries and ecosystems have extensive monitoring networks, while others remain poorly documented. Common and charismatic species generally have more photographs and observations than rare or difficult-to-detect organisms.

An algorithm trained on these unequal datasets may reproduce those inequalities.

Geographic transferability is another major issue. A wildlife classifier that performs well in one protected area may perform much worse elsewhere. Satellite models may behave differently when vegetation, climate or sensors change. Acoustic classifiers may struggle with unfamiliar soundscapes.

Rare species create an especially difficult problem because conservationists often care most about precisely those organisms for which the least training data exist.

Model uncertainty therefore needs to be explicitly considered. High classification accuracy under controlled testing does not automatically mean that ecological estimates derived from those classifications are reliable.

Human Knowledge and the Role of Conservationists

The growing use of artificial intelligence does not eliminate the need for field biology.

Ecologists are needed to determine how observations should be collected, whether datasets are representative, whether model outputs make biological sense and whether conservation interventions actually work.

Local communities and Indigenous knowledge can also provide ecological understanding that may be absent from digital datasets. Conservation systems that treat AI as a substitute for local knowledge risk reinforcing existing inequalities and overlooking important information about landscapes and species.

A productive model is therefore human-AI collaboration.

Artificial intelligence is particularly effective at processing large quantities of repetitive information. Humans remain essential for setting goals, designing studies, evaluating uncertainty, interpreting ecological meaning and making ethical and political decisions.

Governance and Responsible AI

As AI becomes integrated into conservation, governance becomes increasingly important.

Questions include who owns biodiversity data, who controls conservation algorithms, whose knowledge is incorporated into models and who benefits from new technologies.

Open-source models may make sophisticated conservation tools more accessible, but technology alone cannot eliminate differences in institutional capacity, computing resources or scientific expertise.

Responsible conservation AI therefore involves more than technical accuracy. It requires transparency, appropriate validation, explainability where important, community participation and recognition of local ecological knowledge.

Conservation organizations must also consider whether automated systems create new dependencies on private technology providers or infrastructure that cannot easily be maintained in the regions where biodiversity is greatest.

The Environmental Cost of Artificial Intelligence

Artificial intelligence can help protect nature while simultaneously creating environmental pressures of its own.

Large computing systems require electricity, data centers, water and physical infrastructure. Expansion of AI infrastructure can therefore have consequences for land, water, energy use and biodiversity.

This creates an important conservation paradox: increasingly powerful computational systems may improve environmental monitoring while contributing to environmental pressures if their infrastructure is poorly planned.

Responsible biodiversity applications should therefore consider not only what an AI system can accomplish but also the resources required to train, operate and maintain it.

The Future of AI and Biodiversity

Future biodiversity monitoring is likely to become increasingly multimodal.

Rather than analyzing one type of information at a time, conservation systems may combine satellite imagery, drones, camera traps, acoustic recordings, environmental DNA, climate observations, citizen-science records and scientific literature.

Foundation models may allow knowledge learned from large biodiversity datasets to be transferred to places and species with relatively little training information. Edge AI could allow sensors in remote areas to process information locally rather than transmitting every photograph or sound recording to distant servers.

Autonomous vehicles and sensor networks could further expand the geographic scale of monitoring.

The larger transformation may be a shift from periodically surveying ecosystems toward continuously observing them. Such systems could provide earlier warnings of habitat destruction, invasive species, wildlife declines and other ecological changes.

Whether this technological capability produces better conservation will depend on what happens after the data are collected. Monitoring biodiversity does not itself protect biodiversity. Information must ultimately lead to effective policies, resources, enforcement, restoration and changes in human behavior.

Conclusion

Artificial intelligence is rapidly expanding the scale at which humanity can observe and analyze the natural world. Machine learning can process millions of wildlife photographs, recognize animal sounds, map habitats from satellites, analyze environmental DNA, predict species distributions and extract ecological information from scientific literature.

These capabilities could help address longstanding gaps in biodiversity knowledge and allow conservation organizations to detect ecological change more rapidly.

Yet AI is not an automatic solution to biodiversity loss. Algorithms inherit limitations from their training data and can fail when applied to unfamiliar species, habitats or regions. Synthetic media can undermine biodiversity databases, opaque models can make conservation decisions difficult to evaluate, and unequal access to technology can reinforce existing inequalities.

The strongest role for artificial intelligence is therefore not to replace conservation scientists, field researchers, Indigenous knowledge holders, local communities or citizen scientists. It is to increase their ability to collect, process and interpret information.

The future of biodiversity conservation is likely to involve increasingly sophisticated partnerships between people and machines. Whether those partnerships benefit nature will depend less on the sophistication of the algorithms themselves than on the quality of the science, governance and human judgment guiding their use.

General AI in conservation, governance, ethics and decision support

1. Advanced Technologies Improve Protected Species Conservation

| NOAA Fisheries | NOAA | 2026-08-27

Describes NOAA's use of artificial intelligence, acoustics, molecular sampling, imagery and uncrewed systems to improve conservation of endangered and protected marine species. AI can help analyze enormous datasets quickly enough to influence active management decisions.

2. Deep Learning for Environmental Monitoring and Conservation: Applications, Approaches, Challenges, and Future Perspectives

| Various authors | Results in Engineering | 2026-08-24

Systematically reviews more than 100 studies applying deep learning to environmental monitoring and conservation. It examines datasets, model architectures, applications, performance metrics and challenges to practical deployment.

3. Artificial Intelligence Adoption and Corporate Biodiversity Concern: Evidence from Chinese Listed Firms

| Various authors | Sustainability | 2026-08-10

Examines whether corporate adoption of artificial intelligence is associated with greater attention to biodiversity issues. The study provides a different perspective on AI by considering its effects on corporate environmental governance rather than ecological monitoring alone.

4. Principles for navigating responsible use of AI for conservation

| Sam Reynolds et al. | Cambridge Open Engage | 2026-07-15

Proposes practical principles for responsible conservation AI. Major issues include accountability, bias, local participation, transparency, environmental cost and maintaining appropriate human control over conservation decisions.

5. Explainable AI for Biodiversity Monitoring and Ecological Image Analysis

| Brinnae Bent et al. | arXiv | 2026-06-26

Argues that explainable AI should become a routine part of biodiversity computer-vision workflows. Examples involving seals and cetaceans show how explanation tools can uncover spurious background cues, occlusion problems and other hidden sources of error.

6. Artificial intelligence and decision support in applied ecology

| Amber Cowans et al. | Journal of Applied Ecology | 2026-06-15

Examines how AI outputs should be integrated into real ecological decisions rather than treated as independent answers. Using bat monitoring as an example, the authors emphasize benchmarking, auditability, equity, data governance and clearly defined human responsibility.

7. Tools and Technology

| U.S. Geological Survey | USGS | 2026-06-12

Summarizes technologies developed by USGS for species conservation, including artificial intelligence, genetics, remote sensing and autonomous systems. These tools help researchers estimate populations, monitor distributions and anticipate ecological responses to climate change.

8. A reflexive artificial intelligence governance for transformative change in sustainability

| Various authors | Ambio | 2026-04-29

Argues that environmental AI should not reduce complex ecological decisions to technical optimization problems. Using protected-area management as an example, the paper proposes governance based on plural values, democratic judgment and transparency.

9. Can AI reveal the hidden life of a rainforest?

| Will McCarry | Conservation International | 2026-04-06

Describes an Amazon expedition testing camera traps, acoustic recorders, environmental DNA, AI-assisted drone mapping and automated insect-monitoring systems. The project explores whether multiple technologies can create a more comprehensive picture of remote rainforest biodiversity.

10. What the Tech?! Digital Innovation for Migratory Species and Protected Area Conservation

| IUCN | International Union for Conservation of Nature | 2026-03-25

Highlights AI, acoustic monitoring, IoT sensors and citizen-science platforms for migratory-species conservation. The examples show how digital technologies can be connected with protected-area management rather than operating as isolated research experiments.

11. Machine learning-driven water quality index prediction in the Dau Tieng reservoir with interpretability via SHapley additive exPlanations

| Various authors | Environmental Research | 2026-03-15

Applies machine learning and SHAP explanations to reservoir water-quality monitoring. Identifying influential factors such as dissolved oxygen and suspended solids can improve management of freshwater habitats and biodiversity.

12. Google's 2026 Environmental Report

| Google | Google | 2026

Discusses SpeciesNet and Perch among Google's environmental AI technologies. The report also acknowledges the need to consider the resource demands and environmental footprint associated with expanding artificial-intelligence infrastructure.

13. Responsible AI Use and Related Infrastructure Development

| The Nature Conservancy | The Nature Conservancy | 2026

Considers both conservation benefits and environmental costs of AI. TNC uses AI for bioacoustics, fisheries and ecosystem monitoring while warning that data centers, electricity use and water consumption can themselves affect biodiversity.

14. Artificial intelligence and conservation

| World Wildlife Fund | WWF | 2026

Describes WWF applications of AI to camera traps, acoustic monitoring, remote sensing, deforestation detection, wildlife crime and conservation information systems. The organization presents AI as an amplifier of conventional conservation rather than a replacement for it.

15. Bridging the edge-cloud gap: adaptive AI for robust image and audio wildlife monitoring

| Various authors | Frontiers in Conservation Science | 2026

Reviews adaptive AI systems that combine field-based edge computing with cloud processing for wildlife imagery and bioacoustics. Continual learning, self-supervised learning and multimodal systems may make monitoring more reliable in remote and changing environments.

16. When does management using artificial intelligence lead to unintended consequences? A case study using smart traps

| Various authors | Ecological Modelling | 2026

Models ecological consequences of AI-enabled traps designed to distinguish target pests from protected species. The study demonstrates that classification accuracy alone does not guarantee good ecological outcomes because population feedback can create unexpected effects.

17. A horizon scan of biological conservation issues for 2026

| William Sutherland et al. | Trends in Ecology & Evolution | 2026

Identifies emerging conservation issues including TinyML and low-power optical AI chips. Such technologies could allow biodiversity sensors to analyze information directly in remote environments while consuming far less power than conventional computing systems.

18. Mythology and machine: harnessing the power of traditional knowledge and artificial intelligence for biodiversity conservation in South West Nigeria

| Various authors | Journal for Nature Conservation | 2026

Explores combining Yoruba traditional ecological knowledge with artificial intelligence for biodiversity conservation. The article argues that technologically sophisticated conservation can be strengthened rather than weakened by local knowledge, cultural values and community leadership.

19. Safeguarding the role of humans in conservation science in the age of AI

| Various authors | Biological Conservation | 2026

Argues that rapid adoption of generative AI should not diminish field research, local ecological expertise, scientific judgment or community-based conservation. The article highlights concerns about AI-generated scientific material and the preservation of meaningful human contributions to conservation science.

20. PyTorch-Wildlife: A collaborative deep learning framework for conservation

| Microsoft AI for Good Lab | Microsoft / GitHub | 2024-2026

Provides an open-source framework containing wildlife detection, classification and bioacoustic models. It allows conservation practitioners to use tools such as MegaDetector within a common software environment.

21. Tech4Nature in 2025: Harnessing technology to deliver conservation impact at scale

| IUCN | International Union for Conservation of Nature | 2025-12-18

Reviews conservation projects using artificial intelligence and other digital technologies in Brazil, China, Kenya, Mexico, Spain and Türkiye. Examples include jaguar detection, mangrove conservation and marine monitoring.

22. AI for Nature: How AI can democratize and scale action on nature

| Google and World Resources Institute | Working Paper | 2025-11

Examines barriers preventing biodiversity monitoring from operating at global scale and evaluates how artificial intelligence might help. The paper also emphasizes responsible deployment, data access and conservation capacity.

23. Can AI help protect nature?

| Max Marcovitch | Conservation International | 2025-07-17

Discusses both opportunities and limitations of artificial intelligence in practical conservation. The article stresses that AI is most useful when it increases the speed and scale of ecological work without displacing field knowledge and human judgment.

24. Algorithms going wild – A review of machine learning techniques for terrestrial ecology

| Various authors | Ecological Modelling | 2025-07

Reviews more than 300 studies using machine learning in terrestrial ecology. Applications include species classification, ecosystem modeling and conservation, while major challenges include data quality, interpretability, computational demands and poor transfer between ecosystems.

25. 3 new ways we're working to protect and restore nature using AI

| Mike Werner | Google | 2025-03-03

Announces an AI-for-nature startup accelerator, funding for conservation organizations and the open release of SpeciesNet. The initiatives are intended to expand access to AI tools for ecosystem and wildlife monitoring.

26. Harnessing artificial intelligence to fill global shortfalls in biodiversity knowledge

| Laura J. Pollock et al. | Nature Reviews Biodiversity | 2025-02-20

Reviews how artificial intelligence could address major gaps in biodiversity knowledge, including species distributions, evolutionary relationships, ecological traits and interactions. The authors argue that AI's greatest future value may come from integrating images, sound, DNA, text and other complex biodiversity datasets.

27. Three ways to accelerate nature protection with AI

| Google and World Resources Institute | Google | 2025

Recommends expanding biodiversity-data collection, developing accessible AI models and increasing knowledge sharing between technology developers and conservation practitioners. The recommendations emphasize open data and local capacity.

28. 2025 Impact Report for One Conservancy Science

| The Nature Conservancy | The Nature Conservancy | 2025

Reviews conservation technologies including satellites, autonomous vehicles, camera traps, environmental DNA, acoustic sensors and artificial intelligence. The report argues that the next challenge is moving from automated sensing toward systems that better represent ecosystem processes.

29. Application of artificial intelligence in agri-tech, environmental and biodiversity conservation

| Various authors | Array | 2025

Reviews applications of machine learning, robotics, sensors and related AI technologies across agriculture and environmental conservation. The paper considers how AI can reduce environmental pressures while emphasizing technical, economic, cybersecurity and governance barriers.

30. Beyond the Hype: Navigating the Conservation Implications of Artificial Intelligence

| Chris Sandbrook | Conservation Letters | 2024-12-20

Provides a critical assessment of conservation AI, including energy and water consumption, biased datasets, job displacement, misuse and techno-optimism. The author argues for transparency and a much broader assessment of AI's indirect effects on biodiversity.

31. Harnessing Artificial Intelligence for Wildlife Conservation

| Various authors | Conservation | 2024-11-11

Reviews the Conservation AI platform and its use of computer vision to detect animals, people and possible poaching-related objects in conventional and thermal imagery. The system illustrates how automated monitoring can provide conservation information much faster than manual analysis.

32. Learning and Planning under Uncertainty for Conservation Decisions

| Lily Xu | Proceedings of the AAAI Conference on Artificial Intelligence | 2024-07-15

Explores machine learning, reinforcement learning and game theory for allocating scarce conservation resources. The work is particularly relevant to ranger patrols and conservation decisions where ecological information is incomplete, noisy and continually changing.

33. Deep learning in terrestrial conservation biology

| Zoltán Barta | Biologia Futura | 2024-01-16

Reviews deep-learning applications using camera traps, acoustic recorders, satellites and other rapidly expanding ecological data sources. The paper describes both the promise of automation and the need for careful ecological interpretation.

34. AMMonitor: Remote monitoring of biodiversity in an adaptive framework

| Laurence Clarfeld et al. | U.S. Geological Survey | 2024

Provides an open framework for integrating remote biodiversity sensors, metadata and machine-learning outputs. Automated detections can feed directly into distribution and occupancy models as additional observations arrive.

35. Hydropower, River Protection, and Artificial Intelligence

| The Nature Conservancy | The Nature Conservancy | 2024

Describes AI-assisted optimization of hydropower planning in the Amazon basin. By evaluating millions of development scenarios, computational methods can help identify energy strategies that reduce damage to free-flowing rivers and biodiversity.

36. Leveraging AI to improve evidence synthesis in conservation

| Various authors | Trends in Ecology & Evolution | 2024

Explores using artificial intelligence to accelerate systematic reviews and other evidence syntheses used in conservation decisions. The authors recommend transparent human-AI collaboration rather than treating automated systems as replacements for expert review.

37. The potential for AI to revolutionize conservation: a horizon scan

| Various authors | Trends in Ecology & Evolution | 2024

Identifies 21 promising uses of AI in conservation, including species recognition, prediction of biodiversity loss, wildlife-trade monitoring and human-wildlife conflict management. The authors also warn about unequal access, AI colonialism and erosion of conservation skills.

38. Deep learning as a tool for ecology and evolution

| Michael L. Borowiec et al. | Methods in Ecology and Evolution | 2022-05-30

Reviews more than 800 studies using deep learning across ecology and evolution. Applications range from species identification and environmental monitoring to genetics, behavior, phylogenetics and ecological modeling.

39. Perspectives in machine learning for wildlife conservation

| Devis Tuia et al. | Nature Communications | 2022

Explains how machine learning can transform data from cameras, acoustic sensors, tracking devices and other ecological technologies into usable information. The authors emphasize interdisciplinary collaboration between ecologists and computer scientists.

40. Artificial Intelligence Meets Citizen Science to Supercharge Ecological Monitoring

| Various authors | Patterns | 2020-10-09

Argues that citizen science and AI are particularly powerful when combined. Human observers provide observations, judgment and engagement while machine learning helps classify, prioritize and analyze biodiversity information at scales difficult for either approach alone.

41. Deep learning for environmental conservation

| Aakash Lamba et al. | Current Biology | 2019-10-07

Reviews early applications of deep learning to environmental imagery, spatial information and acoustic monitoring. The authors also identify technical and institutional barriers that could prevent promising models from becoming practical conservation tools.

42. Towards Ethical Deployment of AI for Conservation Systems

| Christine Kaeser-Chen et al. | Microsoft Research | 2019-08

Identifies ethical concerns that arise when large conservation datasets are processed by artificial intelligence. The paper emphasizes transparency, fairness and responsible institutional deployment.

43. Responsible AI for conservation

| Oliver R. Wearn et al. | Nature Machine Intelligence | 2019-02-11

Warns that poorly designed conservation AI can have real consequences for wildlife and people. The authors call for appropriate performance metrics, ethical safeguards and scrutiny before algorithms are deployed in conservation decisions.

44. Applications for deep learning in ecology

| Sylvain Christin et al. | Methods in Ecology and Evolution | 2019

An early broad review explaining how deep learning can process camera-trap images, audio, video and other rapidly growing ecological datasets. It also provides practical guidance to researchers considering artificial neural networks.

Large language models, foundation models, evidence synthesis and information integrity

45. Interpretable by design: Language model-derived ecological rules for species distribution modelling

| Various authors | Ecological Informatics | 2026-08

Uses language models to generate explicit ecological rules for species-distribution models. The approach attempts to achieve predictive performance while allowing conservation scientists to trace predictions to understandable environmental relationships.

46. Citizen science platforms must mitigate against the threat of generative AI

| Alexander C. Lees et al. | Nature Ecology & Evolution | 2026-07-13

Warns that synthetic wildlife images and fabricated observations generated by AI could contaminate citizen-science biodiversity databases. The authors argue that platforms need stronger verification systems to preserve the scientific value of community observations.

47. Foundation models for bioacoustics – A comparative review

| Various authors | Ecological Informatics | 2026-06

Reviews large pretrained models designed to transfer across different animal-sound recognition problems. The study compares foundation models using several biodiversity benchmarks and examines their potential to reduce the need for large labeled acoustic datasets.

48. The new burden of proving wildlife is real

| Rhett Ayers Butler | Mongabay | 2026-05-29

Discusses how increasingly convincing synthetic images force conservationists and journalists to verify whether wildlife photographs document real animals, locations and events. Generative AI therefore creates new information-integrity challenges for biodiversity communication.

49. Large language models possess some ecological knowledge, but how much?

| Various authors | Ecological Informatics | 2026-05

Benchmarks large language models on ecological tasks. The systems performed relatively well on some species-presence questions but struggled with threats and geographic ranges, illustrating why expert verification remains necessary.

50. Emerging applications of large language models in ecology and conservation science

| Christos Mammides et al. | Conservation Biology | 2026-04-13

Reviews emerging LLM applications including ecological information extraction, interaction with structured databases and large-scale literature synthesis. The authors emphasize validation, transparency and careful selection of tasks appropriate for automated assistance.

51. AI-assisted multi-target classification for research-policy alignment in conservation science

| Various authors | Ecological Informatics | 2026-03

Applies a scientific-language model to classify conservation research according to themes and policy objectives. Automated analysis could reveal where conservation evidence supports policy goals and where important research gaps remain.

52. How AI trained on birds is surfacing underwater mysteries

| Lauren Harrell | Google Research | 2026-02-09

Describes how the Perch 2.0 bioacoustic foundation model transfers knowledge learned from terrestrial animals to whale sounds. The project demonstrates how large pretrained acoustic models may generalize across very different environments.

53. AI-generated wildlife photos make conservation more difficult

| Rhett Ayers Butler | Mongabay | 2026-02-06

Examines the conservation consequences of realistic synthetic wildlife imagery. Fabricated animal encounters and behaviors can spread misinformation, distort perceptions of species and complicate the use of online photographs as ecological evidence.

54. Generative artificial intelligence and marine ecological monitoring

| Various authors | Environmental Modelling & Software | 2026-01-30

Reviews how generative AI could help integrate satellite, acoustic, sensor and robotic observations of marine ecosystems. Applications include synthetic training data, gap filling, forecasting and digital twins capable of representing changing ocean environments.

55. On the foundations of Earth foundation models

| Xiao Xiang Zhu et al. | Communications Earth & Environment | 2026-01-08

Examines the emerging development of general-purpose AI models trained on large Earth-observation datasets. Such models could eventually support biodiversity mapping by transferring learned representations across habitats, sensors and environmental tasks.

56. Towards ecologically meaningful foundation models

| Ross J. Gardiner et al. | EcoEvoRxiv / Leverhulme Centre for Nature Recovery | 2026

Proposes ecological foundation models trained on multimodal datasets spanning organisms, environments and ecological interactions. The goal is to create AI systems that learn transferable ecological representations rather than isolated single-purpose classifiers.

57. Benchmarking large language models for biodiversity assessment: Performance and biases across IUCN Red List species

| Various authors | Ecological Informatics | 2026

Tests five general-purpose LLMs against information for more than 21,000 IUCN Red List species. Performance varies sharply by task and taxonomic group, demonstrating that apparent fluency should not be confused with reliable conservation knowledge.

58. From LSA to LLM: Evolution and limitations of topic modelling methods for biodiversity conservation

| Various authors | Ecological Informatics | 2026

Traces the evolution of automated text analysis in conservation from latent semantic methods to large language models. LLMs can accelerate evidence synthesis and reveal research gaps, but limitations include weak ecological specialization, copyright concerns, energy consumption and reliability.

59. New frontiers in artificial intelligence for biodiversity research and conservation with multimodal language models

| Zhongqi Miao et al. | Methods in Ecology and Evolution | 2025-08-27

Explores multimodal models capable of interpreting images and language together. Such systems may identify species, interpret animal posture and provide natural-language explanations, potentially making advanced biodiversity AI more accessible to non-programmers.

60. Multi-modal Language models in bioacoustics with zero-shot transfer: a case study

| Zhongqi Miao et al. | Scientific Reports | 2025-02-28

Tests an audio-language model on previously unseen bird, frog, whale and other bioacoustic datasets. The model performs surprisingly well at broad categories without retraining but remains weak at fine-grained species identification.

61. Generative AI as a tool to accelerate the field of ecology

| Kasim Rafiq et al. | Nature Ecology & Evolution | 2025-01-29

Reviews ways generative AI could assist ecology, including data augmentation, analysis and scientific workflows. It also addresses risks arising from synthetic data, biases and inappropriate reliance on systems whose outputs require ecological validation.

62. From literature to biodiversity data: mining arthropod organismal traits with machine learning

| Various authors | Biodiversity Data Journal | 2025

Uses machine learning and natural-language processing to extract arthropod traits from scientific literature. Automated knowledge extraction can turn decades of inaccessible narrative information into structured biodiversity databases.

63. Large language models overcome the challenges of unstructured text data in ecology

| Various authors | Ecological Informatics | 2024

Tests LLMs for extracting biodiversity information from research papers, reports and other unstructured text. The models performed strongly on several information-extraction tasks, suggesting large archives of ecological knowledge could become more computationally accessible.

64. Foundation models in shaping the future of ecology

| Albert Morera | Ecological Informatics | 2024

Discusses foundation models trained on very large datasets as a potential new tool for integrating ecological information. Their promise comes with challenges involving interpretation, computing requirements, ethics and ecological validity.

65. Multimodal Foundation Models for Zero-shot Animal Species Recognition in Camera Trap Images

| Zalan Fabian et al. | Microsoft Research / Preprint | 2023-11

Introduces WildMatch, which generates detailed textual descriptions of camera-trap animals and compares those descriptions with external species knowledge. The method aims to recognize species without conventional species-specific training images.

66. Knowledge Augmented Instruction Tuning for Zero-shot Animal Species Recognition

| Zalan Fabian et al. | NeurIPS | 2023-11

Uses vision-language models and external species knowledge to identify wildlife that lacks conventional training examples. The approach could help address the shortage of labeled photographs for rare species.

Camera traps, wildlife computer vision, re-identification and tracking

67. Automated bird flight pattern extraction and classification using machine learning

| Various authors | Ecological Informatics | 2026-08

Shows that wingbeat patterns extracted from ordinary video can help distinguish bird species. The method offers a new biodiversity-monitoring signal beyond photographs and vocalizations.

68. An accurate, efficient, and accessible AI-powered solution for wildlife re-identification in conservation

| Shahrzad Gholami et al. | Scientific Reports | 2026-05-30

Introduces GIRAFFE, a system designed for automated identification of individual giraffes with potential expansion to additional species. Individual recognition can improve demographic analysis, capture-recapture studies and population estimates.

69. Improving wildlife track classification through human-in-the-loop method and explainable AI

| Tinao Petso et al. | Scientific Reports | 2026-04-20

Combines traditional tracking expertise with artificial intelligence to identify wildlife from footprints. Explainable-AI techniques allow human trackers to evaluate which features the algorithm uses and correct problematic predictions.

70. How our open-source AI model SpeciesNet is helping to promote wildlife conservation

| Tanya Birch and Dan Morris | Google | 2026-03-06

Describes conservation organizations adopting SpeciesNet after its open-source release. The article illustrates how general wildlife-recognition models can shorten the processing time for enormous camera-trap collections.

71. Camera traps and deep learning enable efficient large-scale density estimation of wildlife in temperate forest ecosystems

| Various authors | Remote Sensing in Ecology and Conservation | 2026

Tests the DeepFaune classifier on almost 900,000 camera-trap photographs from German protected areas. Automated classifications produced wildlife density estimates generally similar to estimates derived from manually identified images.

72. Modernizing alpine grassland biodiversity conservation: Efficient AI-powered detection of large herbivores in the Qinghai-Tibet Plateau

| Various authors | Global Ecology and Conservation | 2026

Combines lightweight object-detection networks with habitat assessment for large herbivores in alpine grasslands. The system links observations of animals directly with local habitat conditions and is designed for efficient field surveys.

73. Overcoming Fine-Grained Visual Challenges in Animal Re-Identification via Semantic Feature Alignment

| Yihao Wu et al. | IEEE/CVF WACV | 2026

Presents the CARE framework, which combines visual information with generated textual descriptions to distinguish visually similar individual animals. The authors also developed a standalone toolkit intended for real-world biodiversity monitoring.

74. From species-specific models to universal re-ID: a survey of animal re-identification

| Various authors | Information Fusion | 2026

Reviews recent research on recognizing individual animals using deep learning, transformers, multimodal systems and vision-language models. The field is moving toward models that can generalize across species rather than requiring a separate system for every population.

75. Wildlife Insights

| World Wildlife Fund | WWF | 2026

Describes a global AI-supported platform for storing and analyzing camera-trap data. Automated species identification dramatically shortens the time between collecting wildlife images and producing information that conservation managers can use.

76. Automatic re-identification of terrestrial mammals using deep learning and camera trap images

| Various authors | Global Ecology and Conservation | 2026

Develops a deep-learning framework for recognizing individual terrestrial mammals across camera-trap photographs. Individual identification could improve capture-recapture studies, population estimation and long-term monitoring without physically marking animals.

77. Using the power of AI to identify and track species

| Abby Hehmeyer | WWF | 2025-03-03

Describes SpeciesNet, an open-source AI model that rapidly identifies wildlife in camera-trap photographs. WWF examples include evaluating whether canopy bridges in Peru are successfully helping arboreal mammals cross roads.

78. DeLoCo: Decoupled location context-guided framework for wildlife species classification using camera trap images

| Various authors | Ecological Informatics | 2025-03

Investigates whether information about camera location and image backgrounds can improve wildlife identification. The framework explicitly separates animal appearance from geographic context to reduce misleading correlations.

79. Unlocking the power of artificial intelligence for pangolin protection: Revolutionizing wildlife conservation with enhanced deep learning models

| Junjie Zhong et al. | Expert Systems with Applications | 2025

Develops an enhanced YOLO-based detector for pangolins. Better automated recognition could support monitoring of one of the world's most heavily trafficked groups of mammals.

80. Addressing significant challenges for animal detection in camera trap images: a novel deep learning-based approach

| Various authors | Scientific Reports | 2025

Develops a two-stage deep-learning architecture designed for common camera-trap problems such as empty images, visually similar animals and location-specific backgrounds. Specialized expert models improve discrimination after an initial general classification stage.

81. Towards Context-Rich Automated Biodiversity Assessments: Deriving AI-Powered Insights from Camera Trap Data

| Various authors | Sensors | 2024-12-19

Explores combining wildlife-detection models with vision-language systems to extract richer ecological information from camera-trap photographs. Instead of merely assigning species labels, the approach seeks to describe environmental and behavioral context.

82. Intelligent identification system of wild animals image based on deep learning in biodiversity conservation law

| Xiaolong Liang, Derun Pan and Jiayi Yu | Journal of Computational Methods | 2024-06-01

Develops a deep-learning system for detecting and classifying wildlife in complex natural backgrounds. The authors connect automated identification with biodiversity protection and enforcement applications.

83. How artificial intelligence buys valuable time to protect wildlife

| Abby Hehmeyer | WWF | 2024-05-14

Describes AI-assisted analysis of camera-trap photographs collected after Australia's catastrophic bushfires. Rapid processing helped researchers assess wildlife recovery without waiting months for manual classification.

84. Benchmarking wild bird detection in complex forest scenes

| Various authors | Ecological Informatics | 2024-05

Compares object-detection architectures for finding small birds in complex camera-trap backgrounds. The results show that model architecture and feature backbone can substantially affect performance when animals are difficult to distinguish from vegetation.

85. Automated wildlife image classification: An active learning tool for ecological applications

| Various authors | Ecological Informatics | 2023-11

Uses active learning to prioritize which camera-trap images humans should label. Focusing scarce expert effort on the most informative examples reduces the amount of annotation needed to create useful classifiers.

86. Bag of tricks for long-tail visual recognition of animal species in camera-trap images

| Various authors | Ecological Informatics | 2023-09

Addresses the long-tail problem in wildlife datasets, where a few species have many photographs but rare species have very few. Improved training strategies can reduce systematic neglect of uncommon animals.

87. Animal Species Recognition with Deep Convolutional Neural Networks from Ecological Camera Trap Images

| Various authors | Animals | 2023

Tests several neural-network architectures for identifying snakes, lizards and toads from camera-trap imagery. Transfer learning substantially improves performance compared with a network trained from scratch.

88. An evaluation of platforms for processing camera-trap data using artificial intelligence

| Juliana Vélez et al. | Methods in Ecology and Evolution | 2023

Compares available AI platforms designed to process camera-trap photographs. The study helps conservation practitioners understand differences in accessibility, classification capability and workflow.

89. DeepWILD: Wildlife Identification, Localisation and estimation on camera trap videos using Deep learning

| Various authors | Ecological Informatics | 2023

Demonstrates a deep-learning pipeline capable of detecting, identifying and counting wildlife from camera-trap videos. Automated analysis can substantially reduce the labor required to process large monitoring datasets from protected areas.

90. Human vs. machine: Detecting wildlife in camera trap images

| Various authors | Ecological Informatics | 2022

Compares Microsoft's MegaDetector with human-reviewed camera-trap photographs from Arctic Alaska. AI performs strongly on motion-triggered images but much less reliably on some time-lapse images.

91. Class incremental learning for wildlife biodiversity monitoring in camera trap images

| Various authors | Ecological Informatics | 2022

Develops machine-learning methods that can learn newly observed wildlife species without completely forgetting species encountered during earlier training. Such continual-learning systems could make automated biodiversity monitoring more adaptable over time.

92. Application of deep learning to camera trap data for ecologists: Can captivity imagery train a model which generalises to the wild?

| Ryan Curry, Cameron Trotter and Andrew Stephen McGough | arXiv | 2021-11-24

Tests whether photographs of captive rare animals can provide training data for detecting the same species in the wild. The Scottish wildcat case illustrates both the promise and limitations of transferring artificial imagery domains into field monitoring.

93. A systematic study of the class imbalance problem: Automatically identifying empty camera trap images using convolutional neural networks

| Various authors | Ecological Informatics | 2021-09

Investigates how highly unequal numbers of empty and animal-containing photographs affect neural networks. Efficiently removing empty images is a basic but important step in automated biodiversity workflows.

94. An approach to rapid processing of camera trap images with minimal human input

| Various authors | Ecology and Evolution | 2021-08-02

Develops a workflow combining transfer learning and human review to accelerate camera-trap classification. The approach is intended for ecology projects that lack the enormous labeled datasets used by major computer-vision systems.

95. Automatic Camera-Trap Classification Using Wildlife-Specific Deep Learning in Nilgai Management

| Matthew Kutugata et al. | Journal of Fish and Wildlife Management | 2021-07-23

Trains deep-learning models to identify nilgai and other animals from camera traps. Tests in a different geographic region show substantial performance declines, illustrating the transferability problem in wildlife AI.

96. Perspectives on Individual Animal Identification from Biology and Computer Vision

| Maxime Vidal et al. | Integrative and Comparative Biology | 2021-05-29

Reviews automated computer-vision approaches for recognizing individual animals. Non-invasive photographic identification can support behavioral research and population monitoring while avoiding some drawbacks of physical tags and capture.

97. Identification of animals and recognition of their actions in wildlife videos using deep learning techniques

| Various authors | Ecological Informatics | 2021-03

Develops an end-to-end system for detecting animals and classifying their behavior in camera-trap videos. Automated behavioral information can expand the ecological value extracted from monitoring imagery.

98. Next-Generation Camera Trapping: Systematic Review of Historic Trends Suggests Keys to Expanded Research Applications in Ecology and Conservation

| Zackary J. Delisle et al. | Frontiers in Ecology and Evolution | 2021-02-26

Reviews more than 2,000 camera-trap publications and discusses emerging automation. Artificial intelligence is one of the technologies capable of expanding camera trapping from individual projects toward much larger ecological monitoring networks.

99. Robust ecological analysis of camera trap data labelled by a machine learning model

| Robin C. Whytock et al. | Methods in Ecology and Evolution | 2021-02-19

Investigates whether machine-generated camera-trap labels can be used directly in ecological analyses. The work helps distinguish classification accuracy from the reliability of downstream biological conclusions.

100. A deep active learning system for species identification and counting in camera trap images

| Mohammad Sadegh Norouzzadeh et al. | Methods in Ecology and Evolution | 2021

Combines artificial intelligence with selective human labeling so that useful wildlife classifiers can be trained with fewer manually annotated photographs. The method is particularly relevant to conservation projects lacking millions of pre-labeled images.

101. Synthetic Examples Improve Generalization for Rare Classes

| Sara Beery et al. | IEEE/CVF Winter Conference on Applications of Computer Vision | 2020

Tests synthetic wildlife imagery as a way to improve recognition of species represented by very few real photographs. Simulated examples substantially reduce classification errors for some rare classes.

102. Automatic Identification of Individual Primates with Deep Learning Techniques

| Various authors | iScience | 2020

Demonstrates an AI system trained on more than 100,000 images representing over 1,000 individual primates and additional carnivores. Automated facial identification could support non-invasive longitudinal studies of wildlife populations and behavior.

103. Identification of animal individuals using deep learning: A case study of giant panda

| Various authors | Biological Conservation | 2020

Develops a convolutional neural-network system that recognizes individual giant pandas from photographs with high accuracy. Automated individual identification could support long-term monitoring of difficult-to-observe threatened wildlife.

104. Deep Learning Methods for Multi-Species Animal Re-identification and Tracking – a Survey

| Prashanth C. Ravoor and Sudarshan T.S.B. | Computer Science Review | 2020

Reviews computer-vision methods designed to distinguish individual animals rather than merely identifying species. Reliable re-identification could permit automated tracking of individuals across multiple cameras.

105. Improving the accessibility and transferability of machine learning algorithms for identification of animals in camera trap images: MLWIC2

| Michael A. Tabak et al. | Methods in Ecology and Evolution | 2020

Presents machine-learning tools trained on approximately three million camera-trap images from multiple U.S. studies. The project focuses on improving both classification accuracy and accessibility for biologists without extensive machine-learning expertise.

106. Three critical factors affecting automated image species recognition performance for camera traps

| Various authors | Ecology and Evolution | 2020

Tests automated wildlife recognition under more realistic conditions and finds that performance falls substantially when models encounter camera locations absent from their training data. The research illustrates the importance of testing ecological AI beyond familiar backgrounds.

107. Wildlife surveillance using deep learning methods

| Ruilong Chen et al. | Ecology and Evolution | 2019-08-17

Explores deep learning for automatically detecting wildlife in large surveillance datasets. The study demonstrates how computer vision can reduce the labor required for monitoring animals and human-wildlife interactions.

108. Using machine learning to accelerate ecological research

| Stig Petersen et al. | Google DeepMind | 2019-08-08

Describes machine-learning work with Serengeti camera-trap data. Automated processing allows ecologists to examine wildlife community dynamics over spatial and temporal scales that would otherwise require enormous human labeling effort.

109. Efficient Pipeline for Camera Trap Image Review

| Sara Beery, Dan Morris and Siyu Yang | arXiv | 2019-07-15

Proposes combining a general animal detector with a relatively small number of locally labeled images. The approach helps camera-trap projects adapt AI systems to new geographic regions without creating enormous training datasets.

110. Semantic region of interest and species classification in the deep neural network feature domain

| Various authors | Ecological Informatics | 2019-07

Develops a neural-network method that first isolates likely animal regions in cluttered camera-trap photographs before classifying species. Focusing on the animal reduces interference from complex natural backgrounds.

111. Machine learning to classify animal species in camera trap images: Applications in ecology

| Michael A. Tabak et al. | Methods in Ecology and Evolution | 2019

Trains convolutional networks on more than three million wildlife photographs from multiple North American locations. The work was an important step toward reusable camera-trap classifiers rather than models tied to a single study site.

112. Insights and approaches using deep learning to classify wildlife

| Various authors | Scientific Reports | 2019

Tests convolutional neural networks on more than 100,000 images representing 20 African wildlife species. The study also examines which visual features the algorithms rely on, helping make automated species identification more understandable to ecologists.

113. Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning

| Mohammad Sadegh Norouzzadeh et al. | Proceedings of the National Academy of Sciences | 2018

Demonstrates deep neural networks on the 3.2-million-image Snapshot Serengeti dataset. The system identified species, counted animals and classified behaviors at levels approaching human accuracy, illustrating the potential for automated large-scale wildlife monitoring.

114. Towards automatic wild animal monitoring: Identification of animal species in camera-trap images using very deep convolutional neural networks

| Alejandro Gómez Villa et al. | Ecological Informatics | 2017

One of the early studies testing very deep neural networks for automated wildlife classification from camera traps. It documents important difficulties such as empty images, distant animals, class imbalance and changes in lighting.

Bioacoustics and passive acoustic monitoring

115. Twelve quick tips for applying deep learning to animal sounds

| Burooj Ghani et al. | PLOS Computational Biology | 2026-08-12

Provides practical recommendations for applying deep learning to bioacoustic datasets. Topics include training data, evaluation, transfer learning, model selection and avoiding misleading conclusions from seemingly accurate sound classifiers.

116. Automated detection of stereotyped animal sounds using data augmentation and transfer learning

| Benjamin A. Jancovich et al. | Scientific Reports | 2026-04-23

Develops automated detectors for recurring animal vocalizations from passive acoustic data. Transfer learning and data augmentation reduce the amount of species-specific training material required.

117. Promise and pitfalls: Variable performance of AI-assisted passive acoustic monitoring of birds and frogs using BirdNET and VicFrogNET

| Various authors | Ecological Informatics | 2026-03

Evaluates thousands of AI detections of birds and frogs under real field conditions. Performance varies considerably among species, demonstrating that automated acoustic classifications require validation rather than assuming a single confidence threshold works equally well for all taxa.

118. Scalable and low-power edge architecture with Wi-Fi HaLow and on-device spectrograms generation for flexible urban bioacoustics monitoring

| Various authors | Internet of Things | 2026-03

Presents a low-power edge-AI network that processes bird sounds near the point of collection rather than transmitting continuous raw audio. The approach could make large networks of biodiversity sensors more practical in cities.

119. PNW-Cnet: An evolving convolutional neural network to support broad-scale passive acoustic monitoring

| Various authors | Ecological Informatics | 2026

Presents an expandable deep-learning system capable of identifying many biological and environmental sound classes. The model is designed for operational biodiversity monitoring, endangered-species detection and soundscape assessment.

120. Integrating AI models into ecological research workflows: The case of terrestrial bioacoustics

| Justin Kitzes et al. | Methods in Ecology and Evolution | 2026

Argues that species classifiers are only one part of an ecological monitoring system. Hardware, deployment design, data management, AI detection, statistical analysis and ecological interpretation must function together to produce defensible scientific conclusions.

121. BirdNET in avian diversity monitoring: capabilities and challenges

| Various authors | Procedia Computer Science | 2026

Reviews BirdNET's real-world use and finds substantial variation in precision and recall among species and environments. The study emphasizes the need for regional validation and geographically representative training data.

122. Listening forward: emerging roles of bioacoustics in ecology, evolution, and conservation

| Various authors | Biologia Futura | 2026

Reviews rapidly expanding uses of computational bioacoustics, including neural networks capable of detecting species from large soundscape datasets. Transfer learning and automated acoustic classification could make ecosystem monitoring substantially more scalable.

123. Acoustic Monitoring Enables Multi-Taxa Conservation Assessment and Prioritisation Over Large Scales and for Rare and Cryptic Species

| Various authors | Global Ecology and Biogeography | 2025

Uses machine learning to identify bats, birds, small mammals and bush crickets from more than 34,000 hours of audio. The resulting models reveal biodiversity-priority areas and gaps in existing protected-area coverage.

124. Recent technological developments allow for passive acoustic monitoring of Orthoptera in research and conservation across broad temporal and spatial scales

| Various authors | Basic and Applied Ecology | 2025

Develops the OrthopterOSS machine-learning classifier for grasshoppers and crickets. Relatively inexpensive recorders combined with automated identification could make large-scale insect acoustic monitoring practical.

125. Wildlife Classification using Acoustic Features and Deep Learning Approach

| Various authors | Procedia Computer Science | 2025

Tests convolutional neural networks and pretrained models on a rainforest wildlife acoustic dataset. Automated recognition of animal sounds could substantially expand monitoring in dense forests where visual surveys are difficult.

126. Improving acoustic species identification using data augmentation within a deep learning framework

| Various authors | Ecological Informatics | 2024-11

Tests artificial augmentation of animal recordings to improve acoustic classifiers when real training examples are scarce. The approach is particularly useful for rare species that cannot provide thousands of labeled vocalizations.

127. EcoSonicML: Harnessing Machine Learning for Biodiversity Monitoring in South African Wetlands

| Harry Nel et al. | SN Computer Science | 2024-05-02

Applies machine learning to acoustic biodiversity monitoring in South African wetlands. Automated analysis of environmental sound could complement conventional surveys, particularly for vocal species that are difficult to observe visually.

128. Open-source machine learning BANTER acoustic classification of beaked whale echolocation pulses

| Various authors | Ecological Informatics | 2024-05

Presents an open-source acoustic classifier for identifying beaked whales from passive acoustic recordings. The framework is designed to remain useful when training datasets are relatively small, a common problem for rare marine mammals.

129. North American Bat Monitoring Program: NABat Acoustic ML

| Benjamin Gotthold et al. | U.S. Geological Survey | 2024-02-14

Provides an automated machine-learning pipeline for extracting and classifying bat echolocation calls. The system was developed to support scalable monitoring of North American bat populations.

130. Passive acoustic monitoring and convolutional neural networks facilitate high-resolution and broadscale monitoring of a threatened species

| Adam Duarte et al. | Ecological Indicators / U.S. Geological Survey | 2024

Combines passive acoustic recorders with convolutional neural networks to monitor a difficult-to-detect threatened species across large forest landscapes. Automated sound detection makes much broader surveys possible than repeated human listening.

131. Comparing detection accuracy of mountain chickadee song by two deep-learning algorithms

| Various authors | Frontiers in Bird Science | 2024

Compares a species-specific neural network with the general-purpose BirdNET classifier. The study shows that specialized models can sometimes outperform broad classifiers, reinforcing the importance of matching AI design to specific ecological objectives.

132. The bioacoustic soundscape of a pandemic: Continuous annual monitoring using a deep learning system in Agmon Hula Lake Park

| Yizhar Lavner et al. | Ecological Informatics | 2024

Uses BirdNET to analyze continuous recordings of dozens of bird species across multiple years. The system detected shifts in acoustic activity associated with a major avian-influenza outbreak, illustrating the value of continuous automated ecosystem surveillance.

133. Global birdsong embeddings enable superior transfer learning for bioacoustic classification

| Burooj Ghani et al. | Scientific Reports | 2023-12-18

Develops general representations learned from large quantities of bird audio that can be transferred to new bioacoustic tasks. Foundation-style audio features can reduce the amount of labeled data needed to build new wildlife classifiers.

134. Adapting deep learning models to new acoustic environments – A case study on the North Atlantic right whale upcall

| Various authors | Ecological Informatics | 2023-11

Uses transfer learning to adapt whale-call detectors to recordings from different acoustic environments. The study demonstrates how relatively small amounts of new data can make existing conservation AI useful in unfamiliar locations.

135. Improving deep learning acoustic classifiers with contextual information for wildlife monitoring

| Various authors | Ecological Informatics | 2023-11

Shows that incorporating location and other ecological context can improve automated identification of wildlife calls. Geographic priors substantially reduced false positives for difficult acoustic classification tasks.

136. Acoustic fish species identification using deep learning and machine learning algorithms: A systematic review

| Various authors | Fisheries Research | 2023-10

Reviews machine-learning techniques for identifying fish from sonar and echosounder data. Applications include estimating species composition, abundance and biomass during fisheries surveys.

137. PNW-Cnet v4: Automated species identification for passive acoustic monitoring

| Various authors | SoftwareX | 2023-07

Presents a deep neural network capable of detecting dozens of birds, mammals and environmental sounds in Pacific Northwest forests. The accompanying graphical software makes large-scale bioacoustic analysis more accessible.

138. Soundscapes and deep learning enable tracking biodiversity recovery in tropical forests

| Various authors | Nature Communications | 2023

Combines soundscape analysis, convolutional neural networks and metabarcoding to measure biodiversity recovery after tropical agricultural land is abandoned. Automated acoustic measurements closely reflect ecological restoration gradients.

139. NABat ML: Utilizing deep learning to enable crowdsourced development of automated, scalable solutions for documenting North American bat populations

| Various authors | U.S. Forest Service | 2023

Describes a convolutional neural network trained on hundreds of thousands of bat-call spectrograms. Incorporating species range information improves the ecological plausibility of automated acoustic predictions.

140. Method for passive acoustic monitoring of bird communities using UMAP and a deep neural network

| Various authors | Ecological Informatics | 2022-12

Combines dimensionality reduction and convolutional networks to examine a bird community through an annual soundscape. The approach detects acoustic diversity, seasonal activity and cryptic species.

141. Passive acoustic monitoring of animal populations with transfer learning

| Various authors | Ecological Informatics | 2022-09

Tests pretrained neural networks on multiple animal-sound datasets. Useful classifiers can be developed from surprisingly few verified calls, making transfer learning attractive for rare or understudied species.

142. ANIMAL-SPOT enables animal-independent signal detection and classification using deep learning

| Various authors | Scientific Reports | 2022

Introduces an open-source deep-learning framework for animal-sound detection that can be adapted to many taxa and research questions. Tests spanning multiple animal groups demonstrate the value of general tools that conservationists can retrain for new species.

143. Computational bioacoustics with deep learning: a review and roadmap

| Dan Stowell | PeerJ | 2022

Reviews deep-learning techniques for recognizing and analyzing animal sounds. The paper identifies research priorities needed to move computational bioacoustics beyond sound classification toward answering broader ecological and conservation questions.

144. Acoustic detection of regionally rare bird species through deep convolutional neural networks

| Various authors | Ecological Informatics | 2021-09

Uses deep convolutional networks and expert verification to detect two regionally rare bird species in Nepal. Data augmentation increases the amount of usable training material when recordings are scarce.

145. Bioacoustic classification of avian calls from raw sound waveforms with an open-source deep learning architecture

| Francisco J. Bravo Sanchez et al. | Scientific Reports | 2021-08-03

Tests a neural network that learns directly from raw bird-call waveforms rather than manually selected acoustic features. Open code and public data make the approach reproducible for conservation researchers.

146. A machine learning approach for classifying and quantifying acoustic diversity

| Sara C. Keen et al. | Methods in Ecology and Evolution | 2021-03-25

Develops machine-learning techniques for organizing large soundscape datasets and measuring acoustic diversity. The approach can help scientists study biological communities without manually identifying every recorded sound.

147. BirdNET: A deep learning solution for avian diversity monitoring

| Stefan Kahl et al. | Ecological Informatics | 2021

Introduces BirdNET, a deep neural network originally capable of recognizing nearly 1,000 North American and European bird species from sound. Automated acoustic identification enables large passive-monitoring datasets to be analyzed much more efficiently.

148. Automatic standardized processing and identification of tropical bat calls using deep learning approaches

| Various authors | Bioacoustics research | 2019

Applies deep-learning methods to tropical bat echolocation recordings. Standardized automated processing could make acoustic biodiversity metrics more comparable among large tropical monitoring projects.

Remote sensing, satellites, drones and habitat mapping

149. Bring on the drones: how a technology revolution is being rolled out across Africa's nature reserves

| Guardian Environment staff | The Guardian | 2026-09-03

Reports on conservation teams using thermal drones, cameras and connected technologies across African protected areas. The article emphasizes that successful conservation technology depends as much on locally trained personnel as on sophisticated hardware and AI.

150. Satellite Monitoring of Cook Inlet Beluga Whales — Geospatial Artificial Intelligence For Animals

| NOAA Fisheries | NOAA | 2026-06-25

Describes efforts to detect endangered Cook Inlet belugas in very-high-resolution satellite imagery. AI-assisted satellite monitoring could complement aircraft surveys and make wildlife observations possible across much larger and more remote areas.

151. Project SPARROW and the Future of Conservation Technology

| Juan M. Lavista Ferres et al. | Microsoft Research | 2026-06

Describes an open-source biodiversity monitoring platform combining solar energy, edge AI and satellite communications. SPARROW is designed for autonomous operation in remote areas where conventional power and internet connectivity are unavailable.

152. UAV-based deep learning for biodiversity monitoring: Advances, applications, and future directions

| Various authors | Ecological Informatics | 2026-05

Reviews the combination of drones and deep learning for biodiversity surveys. Emerging techniques include transformers, graph neural networks, self-supervised learning and automated analysis of imagery collected across difficult or inaccessible landscapes.

153. Habitat classification from ground-level imagery using deep neural networks

| Various authors | Ecological Informatics | 2026-05

Tests deep neural networks and vision transformers for classifying habitats from ordinary ground-level photographs. The best systems approached expert performance and could complement satellite-based habitat mapping at much finer scales.

154. Satellite Monitoring of North Atlantic Right Whales

| NOAA Fisheries | NOAA | 2026-04-02

Uses very-high-resolution satellite imagery and AI-assisted workflows to locate critically endangered North Atlantic right whales. Satellite observation could eventually supplement ships and aircraft in monitoring vast ocean habitats.

155. WildDrone: autonomous drone technology for monitoring wildlife populations

| U. P. S. Lundquist et al. | Frontiers in Robotics and AI | 2026-01-12

Describes interdisciplinary work toward autonomous drone systems capable of finding and monitoring wildlife populations. Integrating robotics, computer vision and ecological field methods could substantially expand large-area wildlife surveys.

156. Integrating Hyperspectral Data and Google Satellite Embedding Features for Aquatic Fish Biodiversity Prediction

| Various authors | Journal of Remote Sensing | 2026

Combines environmental DNA from 15 Chinese river basins with hyperspectral imagery and satellite foundation-model embeddings. Seven machine-learning algorithms are compared for predicting spatial patterns of fish diversity.

157. Mapping tree species diversity across the Amazon using remote sensing, diverse environmental data and machine learning

| Shoyo Nakamura et al. | International Journal of Applied Earth Observation and Geoinformation | 2026

Combines satellite remote sensing, climate, soil and vegetation data with several machine-learning methods to create kilometer-scale maps of Amazon tree diversity. Extreme gradient boosting produced the strongest results among the tested algorithms.

158. GAIA Satellite Project

| NOAA Fisheries | NOAA | 2025-07-30

Presents the Geospatial Artificial Intelligence for Animals project, which develops open-source workflows for finding marine mammals in satellite imagery. Human-in-the-loop annotation is used to build reliable datasets for future automated detection.

159. Deep learning and satellite remote sensing for biodiversity monitoring and conservation

| Nathalie Pettorelli | Remote Sensing in Ecology and Conservation | 2024-06-17

Reviews how deep learning can extract ecological information from satellite imagery at large spatial scales. Applications include habitat characterization, disturbance monitoring and biodiversity assessment, alongside concerns about training data, transferability and interpretation.

160. Remote sensing and machine learning to improve aerial wildlife population surveys

| Rebecca L. Converse et al. | Frontiers in Conservation Science | 2024-06-05

Reviews how aerial imagery, remote sensing and machine learning can improve wildlife population surveys. Automated detection may reduce processing costs while expanding the geographic area that conservation agencies can monitor.

161. Automatedly identify dryland threatened species at large scale by using deep learning

| Various authors | Science of the Total Environment | 2024-03-20

Uses high-resolution drone imagery and semantic-segmentation networks to identify rare dryland plants. Automated mapping could help conservation programs locate threatened species across landscapes that would be difficult to survey entirely on foot.

162. Merging multiple sensing platforms and deep learning empowers individual tree mapping and species detection at the city scale

| Various authors | ISPRS Journal of Photogrammetry and Remote Sensing | 2023-12

Integrates aerial and ground-level imagery with deep neural networks to map individual urban trees and identify species. The approach provides detailed biodiversity information useful for urban ecosystem management.

163. Tree species classification on images from airborne mobile mapping using ML.NET

| Maja Michałowska et al. | European Journal of Remote Sensing | 2023-11-07

Develops a machine-learning system for recognizing tree species from airborne mobile-mapping imagery. The study demonstrates how accessible software frameworks can be adapted for vegetation inventories.

164. A Biologist's Guide to the Galaxy: Leveraging Artificial Intelligence and Very High-Resolution Satellite Imagery to Monitor Marine Mammals from Space

| Christin B. Khan et al. | Journal of Marine Science and Engineering | 2023-03-11

Reviews the potential for very-high-resolution satellites and machine learning to detect whales and other marine mammals. The paper helped establish the Geospatial Artificial Intelligence for Animals approach to monitoring remote ocean areas.

165. Ensemble Machine Learning for Mapping Tree Species Alpha-Diversity Using Multi-Source Satellite Data in an Ecuadorian Seasonally Dry Forest

| Various authors | Remote Sensing | 2023-01-18

Combines several satellite datasets and machine-learning algorithms to estimate tree diversity in threatened seasonally dry tropical forest. The method provides spatial biodiversity information across areas where field sampling is sparse.

166. Tree species classification from complex laser scanning data in Mediterranean forests using deep learning

| Matthew J. Allen et al. | Methods in Ecology and Evolution | 2023

Applies deep learning to three-dimensional terrestrial laser-scanning data. Automated recognition of tree species could reduce a major processing bottleneck in detailed forest inventories.

167. Deep Learning Based Aerial Imagery Classification for Tree Species Identification

| O. C. Bayrak, F. Erdem and M. Uzar | ISPRS Archives | 2023

Evaluates YOLOv8 models on aerial imagery for tree-species classification. Automated forest mapping can support biodiversity assessments, ecosystem-health monitoring and sustainable forestry.

168. A deep learning-based mobile application for tree species mapping in RGB images

| Mário de Araújo Carvalho et al. | International Journal of Applied Earth Observation and Geoinformation | 2022-11

Tests whether tree-classification neural networks can run on mobile devices. Field-capable AI could allow rapid mapping of individual trees without requiring continuous access to powerful computers.

169. Mapping tree species proportions from satellite imagery using spectral–spatial deep learning

| Corentin Bolyn et al. | Remote Sensing of Environment | 2022-10

Uses Sentinel-2 imagery and a convolutional neural network to estimate tree-species composition across mixed forests. The resulting maps cover a much larger area than conventional individual-tree inventories.

Marine, freshwater, fisheries and coral reef applications

170. Something looks fishy! A philosophical exploration of AI for marine conservation

| Mark Ryan and Paulan Korenhof | AI and Ethics | 2026-08-31

Examines ethical and philosophical questions surrounding the use of artificial intelligence to understand fish populations and marine ecosystems. It emphasizes that conservation technologies embody assumptions about knowledge, responsibility and environmental decision-making.

171. Reproducible mapping of marine biodiversity using autonomous surface vehicles and deep learning

| Matteo Contini et al. | International Journal of Applied Earth Observation and Geoinformation | 2026-08

Uses relatively low-cost autonomous surface vehicles to collect georeferenced imagery, bathymetry and photogrammetry. Transformer-based artificial intelligence then classifies coral forms and benthic habitats to produce repeatable biodiversity maps.

172. Advancing AI-based species identification for marine bycatch monitoring: Insights from Faster R-CNN experiments

| Various authors | Ecological Informatics | 2026-07-31

Tests Faster R-CNN for recognizing high-risk marine taxa in images collected under realistic fisheries conditions. The study proposes a two-stage workflow that combines rapid screening with more detailed classification to improve scalable and auditable bycatch monitoring.

173. Combatting the data crisis: a primer on using artificial intelligence in marine biodiversity

| Jacob Badcock et al. | Frontiers in Marine Science | 2026-07-07

Provides an accessible overview of AI imaging in marine biodiversity studies, from data collection through classification. The authors discuss practical implementation, bias, training datasets and the relationship between traditional surveys and automated monitoring.

174. High-accuracy fish species identification using transfer learning on vision foundation models

| Alexandros Kofidis et al. | Frontiers in Marine Science | 2026-05-20

Develops a validated dataset of approximately 70,000 images covering 101 Mediterranean fish species and evaluates modern vision foundation models. High automated identification accuracy could improve citizen-science marine biodiversity monitoring.

175. Towards Trustworthy Artificial Intelligence for Marine Research, Fisheries and Environmental Management

| Jose A. Fernandes-Salvador et al. | Fish and Fisheries | 2026-01-21

Proposes a framework for trustworthy AI in marine science and conservation built around data governance, legal and socioeconomic viability, and technical validation. The authors stress that reliable AI requires scrutiny of the entire data-to-decision pipeline.

176. Advancing coral reef monitoring: a deep learning perspective on automated segmentation and classification

| Hafizi Malik et al. | Discover Applied Sciences | 2026-01-14

Tests deep-learning models that classify reef imagery into categories such as living coral, dead coral and sand. Automated segmentation could make monitoring more practical across extensive or difficult-to-survey reef systems.

177. A lightweight deep learning network for the precise detection and classification of variable plankton

| Various authors | Engineering Applications of Artificial Intelligence | 2026-01

Presents a computationally efficient neural network for real-time plankton detection and classification. Lightweight models are particularly important for deploying AI directly on autonomous or field-based monitoring equipment.

178. New Open-Source AI Solution Poised to Transform Global Fisheries

| The Nature Conservancy | The Nature Conservancy | 2026

Describes an open-source edge-AI system that analyzes fishing-vessel video aboard ships. Near-real-time detection and comparison with electronic logbooks could improve fisheries transparency and help combat illegal or unreported fishing.

179. Beyond classification accuracy: Uncertainty-aware deep learning for coral reef monitoring

| Various authors | Ecological Informatics | 2026

Shows that high image-classification accuracy does not automatically translate into reliable estimates of coral cover. Calibration, uncertainty analysis and appropriate sampling procedures are necessary before AI outputs are used for ecosystem reporting.

180. Technologies for marine biodiversity monitoring and mapping: A systematic review

| Gennaro Ucciero et al. | Marine Pollution Bulletin | 2026

Reviews technologies used in marine biodiversity surveys, including cameras, remotely operated vehicles, autonomous underwater vehicles and artificial intelligence. More than half of reviewed studies used AI, primarily for processing data rather than autonomous decisions.

181. GhostNetZero

| WWF and partners | GhostNetZero | 2025-2026

Uses artificial intelligence to analyze sonar data for abandoned fishing nets. Identifying the likely locations of ghost gear can help recovery teams remove debris that entangles marine wildlife and damages habitats.

182. Identification of Indigenous fish species in Lake Tana using deep learning

| Zelalem Adugna Terefe et al. | Scientific Reports | 2025-12-01

Applies several YOLO deep-learning models to a dataset of 13,000 images representing 16 indigenous fish species from Ethiopia's Lake Tana. Automated identification could aid fisheries management and conservation of freshwater biodiversity.

183. FISH-SPEC: Fast identification system for handheld spectroscopy and species classification

| Various authors | Applied Food Research | 2025-12

Combines portable spectroscopy with machine learning to distinguish fish species rapidly and non-destructively. Such methods could help verify seafood identity and strengthen enforcement of fisheries and biodiversity regulations.

184. Rapid consistent reef surveys with DeepReefMap

| Jonathan Sauder et al. | Scientific Reports | 2025-11-07

Presents an AI-assisted system for rapidly constructing standardized three-dimensional surveys of coral reefs. More repeatable reef mapping can improve comparisons among sites and measurements of ecological change.

185. Temporal comparison of coral reefs in the Galápagos using differential alignment of 3D models and Machine-Learning

| Various authors | Ecological Indicators | 2025-11

Combines three-dimensional reef reconstruction with machine learning to classify coral communities and measure ecological change. Repeated 3D surveys allow coral cover, diversity and structural complexity to be compared through time.

186. Coral reef detection using ICESat-2 and machine learning

| Various authors | Ecological Informatics | 2025-07

Combines satellite-derived bathymetric information with convolutional neural networks to detect coral reefs. The method could provide a lower-cost supplement to intensive underwater surveys.

187. Hook, line, and spectra: machine learning for fish species identification and body part classification using rapid evaporative ionization mass spectrometry

| Jesse Wood et al. | Intelligent Marine Technology and Systems | 2025-06-12

Uses machine learning to recognize fish from mass-spectrometry signatures. Rapid biochemical identification could strengthen fisheries research, traceability and enforcement.

188. Deep learning meets marine biology: Optimized fused features and LIME-driven insights for automated plankton classification

| Various authors | Computers in Biology and Medicine | 2025-06

Combines multiple neural networks with explainable-AI methods to classify plankton imagery. Automated plankton identification has applications in marine biodiversity surveys, food-web studies and water-quality monitoring.

189. Species identification for Indian seafood markets: A machine learning approach with a fish dataset

| Various authors | Data in Brief | 2025-02

Provides a large labeled image dataset covering commercially important fish taxa for machine-learning research. Better automated identification has implications for fisheries monitoring, seafood traceability and enforcement against species substitution.

190. Reviewing seas of data: Integrating image-based bio-logging and artificial intelligence to enhance marine conservation

| Marianna Chimienti et al. | NOAA Repository / Methods in Ecology and Evolution | 2025

Reviews AI analysis of image-based biologging devices carried by marine animals. Automated interpretation of underwater imagery could reveal predator behavior, prey interactions and environmental conditions that are difficult to observe directly.

191. Deep learning for automated coral reef monitoring: a novel system based on YOLOv8 detection and DeepSORT tracking

| Various authors | Ecological Informatics | 2025

Combines object detection and tracking to identify and measure corals in repeated underwater video transects. Such systems could automate comparisons of reef condition through time.

192. Artificial Intelligence in Aquatic Biodiversity Research: A PRISMA-Based Systematic Review

| Various authors | Biology | 2025

Reviews more than 300 studies involving artificial intelligence and freshwater biodiversity. Applications include automated identification, habitat assessment, ecological risk prediction and conservation planning, while geographic bias and inconsistent validation remain major problems.

193. Artificial Intelligence in Aquatic Biology: Identifying and Conserving Aquatic Species

| Various authors | Water & Ecology | 2025

Reviews machine learning and deep learning for aquatic species identification, habitat monitoring and threat detection. AI offers scalable analysis of underwater imagery and sensor data but remains dependent on representative training datasets and reliable validation.

194. Coral Reef Surveillance with Machine Learning: A Review of Datasets, Techniques, and Challenges

| Various authors | Electronics | 2024-12-20

Reviews machine learning, GIS, remote sensing and publicly available datasets used to monitor coral reefs. The paper identifies fragmented datasets and inconsistent methods as major constraints on scaling automated reef assessment.

195. A generalized machine learning model for long-term coral reef monitoring in the Red Sea

| Various authors | Heliyon | 2024-09-30

Applies a machine-learning classifier to Landsat imagery spanning nearly two decades in the Red Sea. The model detected substantial coral-cover declines and demonstrates the possibility of repeatable long-term reef surveillance from satellites.

196. Automated identification of aquatic insects: A case study using deep learning and computer vision techniques

| Various authors | Science of the Total Environment | 2024-07-20

Uses convolutional neural networks to identify 90 groups of mayflies, stoneflies and caddisflies used as freshwater ecological indicators. Automated taxonomic identification could greatly accelerate biological water-quality monitoring.

197. AquaVision: AI-Powered Marine Species Identification

| Various authors | Information | 2024

Develops image-classification models for several invasive fish species in Mediterranean waters. Smartphone-style recognition could support citizen scientists and early detection of biological invasions.

198. Deep Learning-Based Classification of High-Resolution Satellite Images for Mangrove Mapping

| Yidi Wei et al. | Applied Sciences | 2023-07-24

Uses high-resolution satellite imagery and deep learning to identify mangrove habitats, including relatively small patches. Automated mapping can help track ecosystem loss and restoration.

199. A deep learning model for measuring coral reef halos globally from multispectral satellite imagery

| Simone Franceschini et al. | Remote Sensing of Environment | 2023-07-01

Uses deep neural networks to detect and measure sandy halos surrounding coral patches in satellite imagery. Reef-halo dynamics may provide indirect information about predator-prey interactions and reef ecosystem condition.

200. Fish-TViT: A novel fish species classification method in multi water areas based on transfer learning and vision transformer

| Various authors | Heliyon | 2023-06

Uses vision transformers and transfer learning to identify both marine and freshwater fish. Grad-CAM visualizations help reveal which parts of fish images influence the model's classification.

201. Automated Freshwater Fish Species Classification using Deep CNN

| Jayashree Deka, Shakuntala Laskar and Bikramaditya Baklial | Journal of the Institution of Engineers (India): Series B | 2023-04-21

Tests convolutional neural networks on 20 indigenous freshwater fish species from northeastern India. Automated visual recognition could assist surveys where closely related species are difficult for non-specialists to distinguish.

202. Superpixel for seagrass mapping: a novel method using PlanetScope imagery and machine learning in Tauranga Harbour, New Zealand

| Nam-Thang Ha et al. | Environmental Earth Sciences | 2023-03-13

Combines very-high-frequency satellite imagery with machine learning to map seagrass. Accurate spatial information is important because seagrass habitats support marine biodiversity and significant blue-carbon storage.

203. Application of a deep learning image classifier for identification of Amazonian fishes

| Various authors | Ecology and Evolution | 2023

Develops neural networks trained on expert-verified photographs of Amazonian fish. The system aims to address the shortage of taxonomic expertise in one of the world's most species-rich freshwater ecosystems.

204. Identification of freshwater fish species based on fish feature point detection

| Various authors | Transactions of the Chinese Society of Agricultural Engineering | 2023

Develops a deep-learning approach that detects anatomically informative points on fish before classifying species. Explicit morphological measurements may make AI predictions more interpretable than purely black-box image classification.

205. Real-Time Marine Animal Detection Using YOLO-Based Deep Learning Networks in the Coral Reef Ecosystem

| J. Zhong et al. | ISPRS Archives | 2022-04-22

Applies YOLO object-detection networks to marine animals in coral-reef imagery. Real-time detection could support underwater robotic and video-based biodiversity surveys.

206. Accelerating Species Recognition and Labelling of Fish From Underwater Video With Machine-Assisted Deep Learning

| Daniel Marrable et al. | Frontiers in Marine Science | 2022

Uses deep-learning assistance to accelerate analysis of baited underwater-video surveys. Faster species identification can shorten the delay between marine field surveys and ecosystem assessments.

207. On the use of deep learning for fish species recognition and quantification on board fishing vessels

| Various authors | Marine Policy | 2022

Uses Mask R-CNN and other neural networks to identify fish and estimate their lengths aboard fishing vessels. Automated catch characterization could strengthen stock assessment and fisheries compliance.

208. Monitoring of Coral Reefs Using Artificial Intelligence: A Feasible and Cost-Effective Approach

| Manuel González-Rivero et al. | Remote Sensing | 2020

Tests convolutional neural networks against expert coral-reef image analysis and reports very high agreement. Automated processing can reduce one of the major bottlenecks in large-scale photographic reef monitoring.

209. Coral Reef Change Detection in Remote Pacific Islands Using Support Vector Machine Classifiers

| Justin J. Gapper et al. | Remote Sensing | 2019-06-27

Tests whether machine-learning classifiers trained on remote-sensing data can generalize across geographically separated Pacific coral reefs. Transferability is essential if automated reef monitoring is to operate beyond individual study sites.

210. Parsing Human and Biophysical Drivers of Coral Reef Regimes

| Various authors | NOAA Fisheries | 2019-02-13

Uses machine learning to analyze relationships among human pressure, environmental conditions, fish communities and benthic reef structure across hundreds of Hawaiian survey sites. The analysis helps identify which pressures are most strongly associated with different reef ecosystem states.

Species distribution, ecological modeling and climate forecasting

211. Risk and resilience of indicator bird species under climate and land-cover change

| Various authors | Ecological Indicators | 2026-09

Uses an ensemble of Random Forest, XGBoost, LightGBM, CatBoost and support-vector machines to project future bird habitat. Combining climate and land-cover information allows biodiversity risk hotspots to be identified for conservation planning.

212. Development of biodiversity prediction model for inland wetlands in Korea through Machine learning

| Various authors | Journal for Nature Conservation | 2026-06

Models plant, bird and fish biodiversity across more than 1,000 Korean wetlands. Explainable machine learning identifies different environmental thresholds and drivers for different taxa and wetland types.

213. Multi-scale neural networks enhance species distribution modelling across predictors and taxonomic groups

| Various authors | Ecological Informatics | 2026-06

Tests neural-network species-distribution models that analyze environmental predictors at several spatial scales. Multi-scale modeling improved performance across different taxa while explainable-AI tools identified which scales contributed most strongly to predictions.

214. Two decades of species distribution modeling: A systematic review of methods and applications

| Various authors | Ecological Modelling | 2026

Reviews two decades of species-distribution modeling and documents increasing use of machine-learning methods. Geographic bias toward North America and Europe remains an important limitation for global biodiversity forecasting.

215. Species distribution models and machine learning algorithms for medicinal and industrial plants conservation

| Various authors | Smart Agricultural Technology | 2026

Reviews 90 studies combining species-distribution models and machine learning for important plant species. AI-based habitat suitability mapping may help prioritize conservation, cultivation and climate adaptation efforts.

216. Evaluating machine learning algorithms for accuracy, stability, and among-predictors discriminability in modeling species richness across ten datasets

| Various authors | Ecological Informatics | 2025-12

Compares tree-based machine-learning methods and regression for predicting biodiversity across multiple datasets. The study shows that predictive accuracy alone is insufficient because ecological models also differ in stability and their ability to identify important environmental drivers.

217. Machine learning applied to global scale species distribution models

| Alba Fuster-Alonso et al. | Scientific Reports | 2025-10-27

Uses Bayesian machine learning to estimate global habitat suitability for marine turtles and project future distributions under climate change. The analysis illustrates how AI-enhanced species-distribution models can identify changing conservation priorities.

218. Machine learning predictions of climate change effects on nearly threatened bird species habitat in Ethiopia for conservation strategies

| Tadele Bedo Gelete et al. | Scientific Reports | 2025-10-22

Uses machine learning to project how climate change may alter habitat for the nearly threatened Salvadori's serin in Ethiopia. The resulting habitat maps identify potential refuges and areas where conservation interventions may become increasingly important.

219. Decoding drivers of multi-level ecological networks through key bird species integration: A machine learning interpretability framework for biodiversity conservation

| Various authors | Ecological Indicators | 2025-10

Uses interpretable machine learning to identify environmental drivers and thresholds within ecological networks. Bird distributions help reveal key functional areas where habitat connectivity could be preserved or restored.

220. Quantifying the non-linear response of bird diversity to landscape features in metropolitan areas: A machine learning-based analysis

| Various authors | Environmental Impact Assessment Review | 2025-08

Combines citizen-science bird data with XGBoost and explainable-AI methods to investigate urban biodiversity. The analysis identifies non-linear relationships between landscape features and bird diversity that can inform city planning.

221. Classification of animal species via deep neural networks and species distribution modeling: a systematic review

| Mateus Braga Oliveira et al. | Artificial Intelligence Review | 2025-05-03

Reviews research combining image-based deep learning with ecological species-distribution models. Integrating visual identification with information about where species are likely to occur can improve automated wildlife classification.

222. Artificial intelligence for biodiversity: Exploring the potential of recurrent neural networks in forecasting arthropod dynamics based on time series

| Various authors | Ecological Indicators | 2025-02

Compares recurrent neural networks with conventional time-series techniques for forecasting arthropod populations. AI performs well in several scenarios but also demonstrates limits when ecosystems undergo extreme invasion or extinction dynamics.

223. Modeling climate change impacts and predicting future vulnerability in the Mount Kenya forest ecosystem using remote sensing and machine learning

| Various authors | Environmental Monitoring and Assessment | 2025

Integrates Landsat vegetation indices and machine learning to identify climate-related vulnerability in the Mount Kenya forest ecosystem. The analysis supports conservation planning in a biodiversity-rich region that is also an important water source.

224. A comprehensive review of spatial distribution modeling of plant species in mountainous environments

| Various authors | Kuwait Journal of Science | 2025

Reviews remote sensing and species-distribution models for mountain plants under climate change. Machine learning is increasingly used to estimate habitat suitability, but uncertainty remains particularly important in complex mountain landscapes.

225. Evaluating the Landscape of AI in Ecology: A Systematic Review of Machine Learning for Species Distribution Models

| Renato Okabayashi Miyaji et al. | IEEE Access | 2025

Reviews more than 200 studies applying machine learning to species-distribution modeling. Random forests and related methods are widely used, but inconsistent evaluation and ecological interpretation remain barriers to reliable conservation applications.

226. Machine Learning and Its Applications in Studying the Geographical Distribution of Ants

| Shan Chen and Yuanzhao Ding | Diversity | 2022-08-26

Uses several machine-learning algorithms to estimate global ant species richness, particularly in countries with incomplete biological records. The analysis suggests AI can identify potentially overlooked biodiversity hotspots in data-poor regions.

227. Species Distribution Modeling for Machine Learning Practitioners: A Review

| Sara Beery et al. | arXiv | 2021

Introduces species-distribution modeling to machine-learning researchers and explains important ecological concepts, datasets and methodological pitfalls. The paper encourages closer collaboration between computer scientists and conservation scientists.

Taxonomy, plant/insect/species identification and biodiversity informatics

228. Herbarium specimens in the age of artificial intelligence: from herbarium image identification to Integrative Taxonomic AI

| Atsuko Takano et al. | Journal of Plant Research | 2026-08-26

Reviews how AI is expanding from simple herbarium-image classification toward integrative taxonomy involving morphology, text, metadata and genetics. The authors emphasize using AI as expert decision support rather than an autonomous taxonomic authority.

229. On class imbalance in machine learning-based taxa identification: A comparative analysis of mitigation strategies

| Various authors | Machine Learning with Applications | 2026-06

Examines a persistent biodiversity-AI problem: common species dominate image datasets while rare species have few examples. The study compares techniques for preventing automated classifiers from systematically neglecting uncommon taxa.

230. Harnessing mmWave signals and machine learning for noninvasive taxonomic classification of insects

| Various authors | PNAS Nexus | 2026-04-28

Uses millimeter-wave signals and hierarchical machine learning to classify pollinating insects without relying on photographs. The technology could function under lighting conditions that limit conventional insect-camera systems.

231. Advancing biological taxonomy in the AI era: deep learning applications, challenges, and future directions

| Suxiang Lu et al. | Science China Life Sciences | 2026-01

Reviews applications of deep learning to biological taxonomy, including images, genetics and multimodal datasets. It also considers data scarcity, explainability and how AI systems should interact with professional taxonomists.

232. Artificial Intelligence in Taxonomy: Advancing Species Identification and Classification

| Vemuganti Varaprasad Rao and Ankatwar Gajanan | Asian Journal of Research in Computer Science | 2026

Reviews AI applications to taxonomic identification using images, DNA and animal sounds. The authors discuss data imbalance, interpretability, benchmark design and the importance of keeping professional taxonomists involved in automated identification.

233. A pipeline to compile expert-verified datasets of digitised herbarium specimens for automated plant identification to accelerate taxonomy

| Various authors | Plants, People, Planet | 2026

Develops workflows for converting digitized herbarium collections into reliable machine-learning datasets with expert verification. High-quality reference data are essential if AI is to assist species identification and taxonomic discovery.

234. Herbariograph: a deep-learning tool to classify specimen images

| Fabio Andrés Ávila et al. | New Phytologist | 2026

Develops deep-learning tools for classifying digitized herbarium specimens. Automated analysis of the world's massive botanical collections could unlock information on plant diversity, historical distributions and morphological change.

235. Advances in machine learning models for plant species identification: A scoping review

| Various authors | Ecological Informatics | 2026

Reviews recent automated plant-identification research and highlights increasing adoption of deep learning and vision transformers. The authors identify opportunities for self-supervised learning, citizen-science datasets and taxonomically meaningful evaluation methods.

236. Descriptron: Artificial intelligence for automating taxonomic species descriptions with a user-friendly software package

| Alex R. Van Dam et al. | Systematic Entomology | 2025-09-29

Combines computer vision and language-model techniques to assist taxonomists in producing morphological species descriptions. Automation could accelerate documentation of undescribed biodiversity while retaining expert oversight.

237. Insect identification by combining different neural networks

| Loris Nanni et al. | Expert Systems with Applications | 2025-05-10

Develops machine-learning approaches for classifying insect species while also addressing undescribed taxa. Automated insect identification could be particularly valuable because global insect biodiversity greatly exceeds available professional taxonomic capacity.

238. Harnessing deep learning for plant species classification: A comprehensive review

| Aisha Zulfiqar et al. | Computers and Electronics in Agriculture | 2025

Reviews a decade of artificial-intelligence methods for automated plant identification, including convolutional networks, vision transformers and few-shot learning. More reliable plant recognition could help address shortages of taxonomic expertise and improve biodiversity surveys.

239. Deep learning-based detection of indicator species for monitoring biodiversity in semi-natural grasslands

| Various authors | Environmental Science and Ecotechnology | 2024-09

Trains object-detection models to recognize plant indicator species associated with high-value grasslands. A novel strategy using greenhouse-grown plants helped overcome the shortage of field training images.

240. Zero-Shot-Learning for Plant Species Classification

| Various authors | Procedia Computer Science | 2024

Tests whether vision-language models can identify plant species that were not represented in model training. Zero-shot recognition could eventually help biodiversity surveys deal with the enormous number of rare and poorly documented species.

241. Hyperparameter-tuned batch-updated stochastic gradient descent: Plant species identification by using hybrid deep learning

| Various authors | Ecological Informatics | 2023-07

Combines shallow and deep visual features to recognize plants at different stages of leaf development. Automated plant identification could reduce some of the labor associated with biodiversity inventories.

242. Individual Tree Species Identification Based on a Combination of Deep Learning and Traditional Features

| Caiyan Chen et al. | Remote Sensing | 2023-04-27

Combines learned image features with conventional remote-sensing characteristics to identify individual trees. The hybrid method seeks to capture both sophisticated AI representations and ecologically interpretable information.

243. Automated Real-Time Identification of Medicinal Plants Species in Natural Environment Using Deep Learning Models

| Various authors | Plants | 2022-07-27

Tests real-time deep-learning identification of plant species under natural field conditions in Borneo. Smartphone-compatible recognition systems could support biodiversity surveys, environmental education and documentation of useful or threatened plants.

244. Learning niche features to improve image-based species identification

| Various authors | Ecological Informatics | 2021-03

Combines ecological niche information with visual recognition. Geographic and environmental knowledge helps distinguish visually similar species and improves predictions in small or imbalanced datasets.

eDNA, genomics, metabarcoding and molecular monitoring

245. How we're helping preserve the genetic information of endangered species with AI

| Lizzie Dorfman and Andrew Carroll | Google | 2026-02-02

Describes partnerships using AI-assisted genomics to sequence endangered species. Preserving high-quality genome information can support research on population diversity, evolutionary history and long-term conservation.

246. Environmental DNA as a tool for ecosystem monitoring and conservation biology

| Various authors | Frontiers in Marine Science | 2026

Reviews environmental DNA monitoring and includes extensive discussion of machine-learning-assisted metabarcoding. AI may reveal complex relationships among species communities, environmental pressures and ecosystem condition that conventional analyses miss.

247. Machine learning, eDNA and citizen science in monitoring and assessing biodiversity and invasive alien species at sea

| Various authors | Frontiers in Marine Science | 2026

Examines the combined use of machine learning, environmental DNA, citizen science, drones and remote sensing for marine biodiversity monitoring. The authors emphasize validation and standardized biodiversity variables when integrating these emerging technologies.

248. An AI-driven deep learning pipeline for taxonomic classification and biodiversity assessment of deep-sea environmental DNA

| Various authors | Computers in Biology and Medicine | 2026

Presents a proof-of-concept deep-learning system for processing environmental DNA from deep-sea ecosystems. The approach attempts to classify known taxa while also detecting sequences that may represent previously unrecognized biodiversity.

249. The utility of combining deep learning with metabarcoding to model biodiversity dynamics at a national scale

| Various authors | Ecological Informatics | 2025-12

Combines DNA metabarcoding, geospatial information and machine-learning models to map arthropod richness across Sweden. The models successfully capture seasonal biodiversity patterns and illustrate how molecular observations can be scaled into national biodiversity maps.

250. Combining environmental DNA and remote sensing variables to model fish biodiversity in tropical river ecosystems

| Various authors | Ecological Informatics | 2025-12

Combines environmental DNA observations from tropical rivers with remotely sensed environmental variables and machine learning. The approach aims to generate spatial biodiversity information where conventional fish surveys are difficult or expensive.

251. Creating interpretable deep learning models to identify species using environmental DNA sequences

| Samuel Waggoner et al. | Scientific Reports | 2025-07-28

Develops interpretable deep-learning approaches for assigning environmental DNA sequences to species. The work attempts to make genomic AI more transparent by showing which sequence features influence identification decisions.

252. Next-Generation River Health Monitoring: Integrating AI, GIS, and eDNA for Real-Time and Biodiversity-Driven Assessment

| Various authors | Hydrobiology | 2025-07-16

Proposes integrating artificial intelligence, geographic information systems and environmental DNA into freshwater monitoring. The framework seeks to connect water-quality assessment with direct measurements of biological communities.

253. CRISPR-Based Environmental Biosurveillance Assisted via Artificial Intelligence Design of Guide-RNAs

| Benjamín Durán-Vinet et al. | Environmental DNA | 2025-05-09

Combines AI-designed CRISPR guide RNAs with environmental DNA approaches for detecting organisms in environmental samples. The technique could support rapid surveillance for invasive or otherwise high-priority species.

254. Nine (not so simple) steps: a practical guide to using machine learning in microbial ecology

| Corinne Walsh, Elías Stallard-Olivera and Noah Fierer | mSystems | 2024-01-23

Provides practical guidance for developing and evaluating machine-learning models using complex microbiome data. Microbial communities form a major component of biodiversity but present analytical challenges because of their immense dimensionality.

255. A universal tool for marine metazoan species identification: towards best practices in proteomic fingerprinting

| Various authors | Scientific Reports | 2024

Uses machine learning to classify nearly 200 marine species from protein fingerprints. The study demonstrates that molecular signals beyond DNA can support rapid biodiversity identification.

256. A combination of machine-learning and eDNA reveals the genetic signature of environmental change at the landscape levels

| Various authors | Molecular Ecology | 2023

Combines machine learning with eDNA data from freshwater monitoring sites in Switzerland. The model distinguishes reference from impacted ecological communities and demonstrates the potential for molecular AI biomonitoring.

Forests, wetlands, restoration, invasive species and wildfire

257. AI-driven forest restoration: A governance framework for people and nature

| Various authors | Environmental Science & Policy | 2026-09

Examines how artificial intelligence could scale forest restoration while warning about data gaps, opaque algorithms, unequal access and power imbalances. The authors advocate participatory knowledge integration, policy alignment and distributed governance.

258. From Ecological Monitoring to Prevention Decision Support: A Critical Review of Artificial Intelligence for Forest Fire Prevention

| Various authors | Forests | 2026-07-11

Examines how AI can translate ecological and environmental observations into forest-fire prevention decisions. Applications include ignition-risk prediction, fuel assessment, early detection and evaluation of potential fire spread.

259. Deep learning-based computer vision in forest monitoring and management: a systematic review

| Gabriel Osei Forkuo and Stelian Aleandru Borz | Biodiversity and Conservation | 2026-07-06

Reviews deep-learning computer vision applications in forests, including tree detection, forest inventory, disturbance assessment and ecosystem management. The study demonstrates increasing potential to transform large quantities of imagery into biodiversity and forest-management information.

260. DIRV: a novel deep learning-informed interpretable framework for vegetation species diversity assessment in multi-wetland scenes using multi-modal UAV images

| Bolin Fu et al. | Expert Systems with Applications | 2026-06-01

Uses multimodal drone imagery and interpretable deep learning to assess vegetation diversity across wetlands. The framework attempts to combine high predictive performance with information about the ecological features driving predictions.

261. Wetland change and its impacts on livelihood using machine learning algorithms in Chanda Beel, Gopalganj, Bangladesh

| Nazim Uddin et al. | Next Research | 2026-06

Uses Sentinel-2 imagery and machine-learning classification to examine seasonal wetland change. The study connects ecological dynamics with agriculture and community livelihoods, emphasizing the social dimension of biodiversity management.

262. A data-driven framework for forest conservation prioritisation: From citizen science data to protected area designation

| Various authors | Ecological Informatics | 2026-06

Combines citizen-science observations, species-distribution modeling and AI-based optimization to identify high-value forest areas for protection. The study demonstrates how biodiversity benefits and economic opportunity costs can be evaluated simultaneously when designing protected areas.

263. Swamp-AI: a deep learning model for monitoring wetlands change across the globe

| Charles S. Andros et al. | Scientific Reports | 2026-02-13

Develops a deep-learning system for identifying changes in wetland extent. Automated global monitoring could help identify wetland loss more rapidly than conventional mapping methods.

264. Impacts of land use land cover changes on wetland ecosystem services in Dandi Lake, Oromia, Ethiopia

| Various authors | Journal for Nature Conservation | 2026-01

Uses Google Earth Engine, satellite imagery and Random Forest classification to reconstruct three decades of wetland change. The analysis connects changing land use with ecosystem services and community observations.

265. Wetland restoration enhances soil carbon sequestration in lake ecosystems: Integrating multi-source remote sensing and optimized ensemble machine learning

| Various authors | Ecological Indicators | 2026-01

Uses remote sensing and ensemble machine learning to quantify soil carbon associated with wetland restoration. The research links biodiversity-oriented restoration with climate mitigation and long-term ecosystem recovery.

266. Solution for diagnostics of biological invasion in terrestrial ecosystems: how can deep learning help biodiversity conservation?

| Various authors | Journal for Nature Conservation | 2026-01

Uses Mask R-CNN deep learning to identify invasive Pinus elliottii trees in wetland landscapes. Automated segmentation also estimates the amount of canopy occupied by the invasive species, giving managers a scalable tool for targeting control efforts.

267. Forwarding forest restoration: Seven key socio-ecological issues for advancing forest restoration in a world in flux

| Various authors | People and Nature | 2026

Identifies artificial intelligence and related technologies as one of several major forces likely to reshape forest restoration. The authors stress that technological innovation must be considered alongside land competition, community well-being, climate adaptation and long-term financing.

268. Application of machine learning in forest monitoring: recent progress and future challenges

| Neeraj K. Maurya and Manish K. Pandey | Advanced Geospatial and Ground Based Techniques in Forest Monitoring | 2026

Reviews machine learning for forest inventory, biodiversity assessment, disturbance detection and ecosystem management. It emphasizes the growing integration of satellites, drones and ground sensors with automated analysis.

269. Cracking the black box: a technology-management-policy nexus for invasive species control

| Various authors | Frontiers in Ecology and Evolution | 2026

Reviews AI and machine-learning applications to invasive-species management. It argues that conservation tools must move beyond prediction toward explainable systems that directly support prevention, rapid response and practical control.

270. Mapping coastal wetland biodiversity and ecosystem services using hyperspectral data and machine learning

| Achraf Ben Miled et al. | Regional Studies in Marine Science | 2026

Combines hyperspectral remote sensing with support-vector machines and random forests to map coastal wetlands and associated ecosystem services. The technique demonstrates how AI can translate complex spectral data into conservation-relevant habitat information.

271. A comprehensive survey of the machine learning pipeline for wildfire risk prediction and assessment

| Various authors | Ecological Informatics | 2025-12

Reviews machine-learning approaches to wildfire prediction from data collection through operational deployment. Better forecasts could help protect ecosystems and biodiversity by supporting earlier fire detection and more targeted management.

272. Advances in machine learning for wetland classification: a comprehensive survey of methods and applications

| Derrick Effah et al. | Artificial Intelligence Review | 2025-11-25

Reviews machine-learning techniques used to classify wetlands from remote-sensing data. The study discusses algorithm selection, data integration, transferability and the growing role of deep learning in ecosystem mapping.

273. Machine learning for biosecurity: A probabilistic framework for invasive species management

| Julissa Rojas-Sandoval et al. | Journal of Applied Ecology | 2025-09-18

Develops a probabilistic machine-learning framework for estimating whether introduced plants may progress toward invasion. Testing on more than 1,000 non-native plant species from Caribbean islands shows how AI could support preventative biosecurity before damaging invasions become established.

274. Risks and benefits of artificial intelligence in locally led nature restoration

| A. Tofighi-Niaki, S. Moorthy and R. Soleymani | UK Foreign, Commonwealth & Development Office / REDAA | 2025-09-15

Evaluates AI tools for community-led restoration projects in Africa and Asia. The report considers potential benefits alongside barriers involving cost, expertise, data access, power relations and local control.

275. Human footprint with machine learning identifies risks of the invasive weed Conyza sumatrensis across land-use types under climate change

| Various authors | Global Ecology and Conservation | 2025-09

Compares multiple machine-learning algorithms to predict the global distribution of an invasive plant using climate, soil and human-pressure variables. The resulting risk maps can help identify landscapes vulnerable to future biological invasion.

276. Ecological monitoring of invasive species through deep learning-based object detection

| Various authors | Ecological Indicators | 2025-06

Develops deep-learning object detection for water hyacinth, one of the world's most damaging aquatic invasive plants. Automated identification could improve surveillance across complex waterways where manual inspection is difficult.

277. Ecological monitoring of invasive water hyacinth using deep-learning object detection

| Various authors | Ecological Indicators | 2025

Demonstrates automated detection of water hyacinth across visually complex aquatic environments. Rapid spatial assessment can help managers prioritize removal and track whether invasive-plant control is working.

278. Artificial intelligence applications in hydrological studies and ecological restoration of watersheds: A systematic review

| Various authors | Watershed Ecology and the Environment | 2025

Reviews artificial-intelligence applications in watershed assessment and ecological restoration. Remote sensing combined with machine learning can help identify degradation, evaluate restoration priorities and monitor water-related ecosystem recovery.

279. Machine and Deep Learning for Wetland Mapping and Bird-Habitat Monitoring

| Various authors | Remote Sensing | 2025

Reviews 121 studies using machine learning or deep learning to map wetlands and bird habitats. Geographic imbalance, inconsistent validation and limited ecological linkage remain important research gaps.

280. Optimising forest rehabilitation and restoration through remote sensing and machine learning: Mapping natural forests in the eThekwini Municipality

| Various authors | Remote Sensing Applications: Society and Environment | 2024-11

Uses Landsat imagery and machine learning to map natural forests and historical forest change. Such maps can identify remnants, degraded areas and potential priorities for urban forest restoration.

281. Automatic deforestation driver attribution using deep learning on satellite imagery

| Various authors | Global Environmental Change | 2024-05

Trains deep learning on expert-labeled satellite images to determine what activities caused forest loss in Indonesia. Distinguishing agriculture, plantations and other drivers can make forest and biodiversity policies more targeted.

282. Ecological restoration and artificial intelligence: whose values inform a project?

| Various authors | Restoration Ecology | 2024-03-03

Examines ethical questions surrounding AI-assisted ecological restoration. Because restoration goals inevitably reflect human values, the authors argue that algorithms cannot determine desirable ecological outcomes without transparent human judgment.

Pollinators, agriculture, poaching and field conservation operations

283. PolliCrop: A high-throughput computer vision pipeline for pollinator monitoring in agroecosystems

| Stan Chabert et al. | bioRxiv | 2026-07-13

Develops an automated image-analysis pipeline for measuring insect visitation to crop flowers. Computer vision could allow researchers to compare pollinator activity among plant varieties without manually watching thousands of individual flower visits.

284. Automating pollinator identification using artificial intelligence and participatory science

| Brian J. Spiesman | Current Opinion in Insect Science | 2026-06-18

Examines how artificial intelligence and participatory science can work together to expand pollinator monitoring. Automated identification could help address the enormous amount of observation needed to understand global pollinator declines.

285. AutoPollS: A tool for automated monitoring of pollinators using deep learning

| Matthew A.-Y. Smith et al. | Methods in Ecology and Evolution | 2026-04-30

Introduces an automated computer-vision system designed specifically for insect pollinator monitoring. Scalable detection could help scientists quantify changes in pollinator abundance across landscapes and through time.

286. Honey yield prediction and neonicotinoid risk assessment utilizing a machine learning framework in smart agriculture

| Attia Ghafoor et al. | Scientific Reports | 2026-04-10

Applies machine learning to honey production and pesticide risk while examining environmental pressures affecting honeybees. The research illustrates how agricultural AI can potentially support both production planning and pollinator protection.

287. A machine learning framework for combatting poaching challenges in wildlife

| Various authors | Human Settlements and Sustainability | 2026-03

Combines visual and acoustic machine-learning models for wildlife and potential poaching surveillance. The framework integrates animal detection, behavioral recognition and environmental sound classification.

288. Next-generation agri-environment schemes: Integrating innovations in biodiversity monitoring, farming technologies and digital tools

| Various authors | Biological Conservation | 2026

Discusses how AI, sensors and digital monitoring could improve agricultural conservation programs by targeting measures more precisely and verifying ecological outcomes. Data privacy, costs, accuracy and farmer acceptance remain important implementation issues.

289. Machine learning to detect, classify, and count blackbirds damaging agriculture using drone-based imagery

| Various authors | Ecological Informatics | 2025-12

Uses drone imagery and deep learning to detect and classify blackbirds interacting with agricultural landscapes. The research illustrates how AI could help manage human-wildlife conflict while reducing indiscriminate wildlife-control practices.

290. Machine learning for biodiversity: UAV-based flower detection as an indirect proxy for bee abundance

| Various authors | Ecological Informatics | 2025-11

Uses drone imagery and machine learning to estimate floral resources important to bees. Mapping flower availability provides an indirect, scalable indicator that can help guide pollinator habitat restoration and monitoring.

291. Image-based South Asian bee species identification: a machine learning approach

| Various authors | Journal of Insect Conservation | 2025-07-05

Tests machine learning for identifying South Asian bee species from photographs. Automated identification could increase monitoring capacity in regions where shortages of taxonomic expertise constrain pollinator conservation.

292. The rising tide of conservation technology: empowering the fight against poaching and unsustainable wildlife harvest

| A. J. Lynam et al. | Frontiers in Ecology and Evolution | 2025-05-27

Reviews conservation technologies used against poaching and unsustainable wildlife harvest, including AI, camera traps, drones, acoustic sensors and remote sensing. The paper emphasizes practical deployment, open tools and the institutional capacity needed to convert technology into conservation outcomes.

293. A deep learning pipeline for time-lapse camera monitoring of insects and their floral environments

| Various authors | Ecological Informatics | 2024-12

Uses time-lapse imagery to classify arthropods while simultaneously measuring floral resources. The approach can monitor pollinators and other insects continuously without requiring researchers to inspect every image manually.

294. Using Remote Sensing Imagery and Machine Learning to Predict Poaching in Wildlife Conservation Parks

| Rachel Guo | AAAI Conference on Artificial Intelligence | 2021

Uses public remote-sensing information to improve predictions of where poaching may occur. The approach is particularly relevant to protected areas that lack extensive historical patrol and enforcement datasets.

295. Towards Computer Vision and Deep Learning Facilitated Pollination Monitoring for Agriculture

| Malika Nisal Ratnayake et al. | CVPR Workshops | 2021

Examines computer vision as an alternative to labor-intensive manual pollinator observations. Automated monitoring could produce much larger datasets on insect visitation and pollination services.

Paleobiodiversity, evolution and deep-time applications

296. Paleo-grounded biodiversity foundation models for long-horizon species distribution forecasting

| Various authors | Frontiers in Ecology and Evolution | 2026-07-15

Proposes biodiversity foundation models that combine modern observations with fossil and paleoenvironmental information. Training models across much longer time horizons could improve understanding of how species distributions respond to climatic and ecological change.

297. Emerging uses of artificial intelligence in deep time biodiversity research

| Daniele Silvestro and Catalina Pimiento | Nature Reviews Biodiversity | 2025-08-11

Examines applications of artificial intelligence in paleontology, including automated fossil processing, morphological analysis and evolutionary modeling. AI may help researchers reconstruct biodiversity through geological time while also introducing challenges associated with incomplete and biased fossil datasets.

298. DeepDive: estimating global biodiversity patterns through time using deep learning

| Rebecca B. Cooper et al. | Nature Communications | 2024-05-17

Uses deep learning to help reconstruct historical biodiversity patterns from the incomplete fossil record. The method illustrates how AI can contribute to questions extending far beyond contemporary wildlife monitoring.