Sentinel animals : Enriching artificial intelligence with wildlife ecology to guard rhinos DOI Open Access
Jasper A.J. Eikelboom

Published: June 28, 2021

The survival of both African rhinoceros species is under threat due to large-scale poaching. pressure that poaching currently exerts on rhino populations too large solely wait for long-term conservation strategies, e.g., demand and corruption reduction campaigns, take effect. Consequently, protection efforts aimed at the short-term seem be urgently needed. Unfortunately, current fail prevent population declines as officers often localize poachers before they can kill a rhino. Therefore I develop poacher early warning system provides with more situational awareness, which therefore decrease risk shootouts between officers.For this task focused developing ''sentinel-based system'', envision nature reserves where abundant prey animals are tracked movement responses these automatically used detect presence infer location poachers. Hence term: ''sentinel'', themselves will role game wardens. benefit such it could working all times not limited Apart from obvious wildlife challenge thesis poses, also tackles major scientific challenge: able abrupt changes in an environmental variable based animal movement. In order solve challenge, myriad variables needed considered interaction single model. This premise lead me use non-traditional statistical approach ecologists: artificial intelligence.This brings together number coherent papers about conservation, ecology intelligence, investigating necessity, analytics applicability sentinel-based system. Chapter 2 critically evaluated whether actually examined by if legal international horn trade ultimate solution Through integrative review grey literature legalization, identified four main mechanisms through market would influence remaining wild populations: 1) financial viability private owners, 2) demand, 3) laundering horns, 4) behaviour consumers. Subsequently, determined plausible reasoning only increased revenue farmers potentially conservation. Conversely, global likely increase level cannot met supply. Moreover, omnipresent countries along routes, has potential negatively affect Finally, programmes reducing counteracted legalization removing stigma consuming horn. After combining insights comparing them criteria sustainable farming, concluded legalizing impact populations. To preserve suggest combine long- approaches, prioritizing within trade, increasing well-protected 'safe havens' implementing educational law enforcement targeted consumers.In 3 investigated how much tropical general impacted hunting, apart considering rhinos. did analyzing human hunters alter abundance spatial distribution tropics. systematic mixed effects meta-analysis estimated bird abundances declined average 58% (95% CI: 25-76%) mammal 83% 72-90%) hunted compared unhunted areas. Mammal densities were higher inside than outside protected areas, but hunting reduced even Furthermore, depleted 7 kilometers 40 roads settlements, function access points hunters. These results signify very large. Although effect areas less detrimental reserves, gazettement seems insufficient safeguard accompanied improved reserve management, effective on-ground efforts.In 4 studied link individual rules emergent collective properties, provide information perceived environment animals. For agent-based simulation model investigate indirect fear resources group structures. directly affected resources, self-organization became apparent size formed groups. specifically inherent variability sizes groups generated identical self-organizing processes. found coefficient variation generally lied 50 150% simulations, depending density resource scarcity/predation trade-off. Given variations already homogeneous deterministic scenarios, consider imprecise proxy conditions. Considering imprecision its time lag conditions, informative slowly-evolving conditions require recent history informative.In 5 predicted their data. different influenced multivariate entirety. linking high-resolution sensor data cows controlled various extensive feature engineering machine learning predict Using data-driven framework demonstrated possible quantify performance metrics regression algorithms. Depending chosen window engineering, scales studied. types features (e.g., individual- collective-based, or GPS- accelerometer-based) included separately combination framework. Even though aim exact contribution separate total movement, core accurately well.In 6 developed Welgevonden Game Reserve (South Africa). 138 savanna ungulates combined experimentally staged intrusions, algorithmically detected localized using three-step analytical process achieve this, namely: classification, detection, localization. first step importance interpreting deviations expectations given similar complex relationship animals' heterogeneous achieved precision 46% classify humans versus other quite achievement (given class imbalance normal response behaviour, heterogeneity study area), still leads substantial amount misclassification. However, next two steps classified collectively spatiotemporal context, allowed drastically improve upon detection localization 'poachers'. Periods present area distinguished periods without 86% accuracy balanced validation design, 500m error 54.2% intrusions. chapter thus demonstrates feasibility theme thesis, namely poachers.In automated algorithm aerial imagery intention gauge supplement replace animal-born sensors track en masse near future. deep herbivores images survey Kenya, after same With managed 90-95% layers annotation, correctly 2.8-4.0% extra missed humans. result 1.6-5.0 false positives per true positive, emphasizes manual verification automatic counts images. semi-automatic estimates. indicated detections have find can, especially when supplied taken fixed rate. aforementioned, acknowledge tracking tags. chance decreases substantially horizontal distance camera, expect cameras suitable relatively small areas.Finally, 8 synthesized my research light ecology. argued lies mainly aid during Anthropocene, concurrently reduce some negative associated 'militarized conservation' rights violations). plead collaboration conservationists short- maximize efficacy occasional trade-off success Anthropocene development society harmony nature. forecasted intelligence research, may change way understanding acquired Exciting developments related explainability causality being undertaken computer scientists, scientists do input ecologists make truly insightful applicable real world.

Language: Английский

Perspectives in machine learning for wildlife conservation DOI Creative Commons
Devis Tuia, Benjamin Kellenberger, Sara Beery

et al.

Nature Communications, Journal Year: 2022, Volume and Issue: 13(1)

Published: Feb. 9, 2022

Data acquisition in animal ecology is rapidly accelerating due to inexpensive and accessible sensors such as smartphones, drones, satellites, audio recorders bio-logging devices. These new technologies the data they generate hold great potential for large-scale environmental monitoring understanding, but are limited by current processing approaches which inefficient how ingest, digest, distill into relevant information. We argue that machine learning, especially deep learning approaches, can meet this analytic challenge enhance our capacity, conservation of wildlife species. Incorporating ecological workflows could improve inputs population behavior models eventually lead integrated hybrid modeling tools, with acting constraints latter providing data-supported insights. In essence, combining domain knowledge, ecologists capitalize on abundance generated modern sensor order reliably estimate abundances, study mitigate human/wildlife conflicts. To succeed, approach will require close collaboration cross-disciplinary education between computer science communities ensure quality train a generation scientists conservation.

Language: Английский

Citations

402

The Internet of Animals: what it is, what it could be DOI Creative Commons
Roland Kays, Martin Wikelski

Trends in Ecology & Evolution, Journal Year: 2023, Volume and Issue: 38(9), P. 859 - 869

Published: May 30, 2023

One of the biggest trends in ecology over past decade has been creation standardized databases. Recently, this included live data, formal linkages between disparate databases, and automated analytics, a synergy that we recognize as Internet Animals (IoA). Early IoA systems relate animal locations to remote-sensing data predict species distributions detect disease outbreaks, use inform management endangered species. However, meeting future potential concept will require solving challenges taxonomy, security, sharing. By linking sets, integrating automating workflows, enable discoveries predictions relevant human societies conservation animals.

Language: Английский

Citations

32

Monitoring mammalian herbivores via convolutional neural networks implemented on thermal UAV imagery DOI Creative Commons

Diego Bárbulo Barrios,

João Valente, Frank van Langevelde

et al.

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 218, P. 108713 - 108713

Published: Feb. 9, 2024

Lightweight Unmanned Aerial Vehicles (UAVs) are emerging as a remote sensing survey tool for animal monitoring in several fields, such precision livestock farming. Together with state-of-the-art computer vision techniques, UAV. technology has drastically escalated our ability to acquire and analyse visual data the field, lowering both costs complications associated collection analysis. This paper addresses mammalian herbivores using unexploited field of thermal Multi-Object Tracking Segmentation (MOTS) UAV imagery. In research, MOTS algorithm (Track R-CNN) was trained evaluated segmentation, detection tracking dairy cattle. Data carried out two farms carrying camera at various angles heights, under different light (overcast/sunny) (16.5 °C range) conditions. Our findings suggest that dataset diversity balance, especially regarding range conditions which collected, can significantly enhance efficiency specific scenarios. For training algorithm, transfer learning used knowledge migration method. The performance best model (68.5 sMOTSA, 79.6 MOTSA, 41 IDS, 100 % counting accuracy, 87.2 MOTSP), utilizes 3D convolutions an association head, demonstrates applicability optimal Track R-CNN detecting, tracking, imagery heterogenous demonstrate outperform Long-short Term Memory (LSTM) convolutions. However, LSTM also show performance, offering viable alternative. Furthermore, results highlight inability Optical Flow track motionless animals (-15 −4.1 MOTSA 2076 IDS) proficiency head differentiating static from background. research contributes growing body automated herbivore monitoring, potential applications farming wildlife conservation.

Language: Английский

Citations

6

Accelerometer-based detection of African swine fever infection in wild boar DOI Creative Commons
Кevin Мorelle, José Á. Barasona, Jaime Bosch

et al.

Proceedings of the Royal Society B Biological Sciences, Journal Year: 2023, Volume and Issue: 290(2005)

Published: Aug. 30, 2023

Infectious wildlife diseases that circulate at the interface with domestic animals pose significant threats worldwide and require early detection warning. Although animal tracking technologies are used to discern behavioural changes, they rarely monitor diseases. Common disease-induced changes include reduced activity lethargy (‘sickness behaviour’). Here, we investigated whether accelerometer sensors could detect onset of African swine fever (ASF), a viral infection induces high mortality in suids for which no vaccine is currently available. Taking advantage an experiment designed test oral ASF vaccine, equipped 12 wild boars tag quantified how affects their pattern fingerprint, using overall dynamic body acceleration. Wild showed daily reduction 10–20% from healthy viremia phase. Using change point statistics comparing individuals living semi-free free-ranging conditions, show sickness can be detected such work natural settings. Timely crucial disease surveillance control, technology on sentinel provides viable complementary tool existing management approaches.

Language: Английский

Citations

11

Real-time classification of Serengeti wildebeest behaviour with edge machine learning and a long-range IoT network DOI
Cyrus M. Kavwele, J. Grant C. Hopcraft, Juan M. Morales

et al.

Canadian Journal of Zoology, Journal Year: 2025, Volume and Issue: 103, P. 1 - 11

Published: Jan. 1, 2025

Globally, animal populations are facing increasing levels of environmental disturbance. Human activity, land-use change, and global warming altering migration routes, space use, activity budgets, the behaviour many wildlife species. Understanding impacts on at a fine scale is essential to identify locations increased disturbance, mitigate its effects, predict potential population level outcomes. In this work, we introduce low-cost tracking system that integrates open-source electronics, edge machine learning, an Internet Things network, provide real-time information location animals. The employs on-board learning algorithm distinct behaviours then transmits classification outputs along with data over long-range network. We deployed wildebeest ( Connochaetes taurinus (Burchell, 1823)), in Serengeti National Park, Tanzania, highly social migratory ungulate ecologically economically vital region. Analysis transmitted showed readings were consistent revealed biologically meaningful fluctuations daily patterns. Our introduces new dimension studying movement ecology by offering immediate insights into collared

Language: Английский

Citations

0

Automated near real-time monitoring in ecology: Status quo and ways forward DOI Creative Commons
Anna M. Davison, Koen de Koning, Franziska Taubert

et al.

Ecological Informatics, Journal Year: 2025, Volume and Issue: 89, P. 103157 - 103157

Published: April 18, 2025

Language: Английский

Citations

0

Timely poacher detection and localization using sentinel animal movement DOI Creative Commons
Henrik J. de Knegt, Jasper A.J. Eikelboom, Frank van Langevelde

et al.

Scientific Reports, Journal Year: 2021, Volume and Issue: 11(1)

Published: Feb. 25, 2021

Wildlife crime is one of the most profitable illegal industries worldwide. Current actions to reduce it are far from effective and fail prevent population declines many endangered species, pressing need for innovative anti-poaching solutions. Here, we propose test a poacher early warning system that based on movement responses non-targeted sentinel animals, which naturally respond threats by fleeing changing herd topology. We analyzed human-evasive patterns 135 mammalian savanna herbivores four different using an internet-of-things architecture with wearable sensors, wireless data transmission machine learning algorithms. show presence human intruders can be accurately detected (86.1% accuracy) localized (less than 500 m error in 54.2% experimentally staged intrusions) algorithmically identifying characteristic changes movement. These behavioral signatures include, among others, increase speed, energy expenditure, body acceleration, directional persistence coherence, decrease suitability selected habitat. The key successful identification these lies systematic deviations normal behavior under similar conditions, such as season, time day also indirect costs predation not limited vigilance, but include (1) long, high-speed flights; (2) energetically costly flight paths; (3) suboptimal habitat selection during flights. combination biologging, predictive analytics animal benefit wildlife conservation via detection, solve challenges related surveillance, safety health.

Language: Английский

Citations

23

Illegal killing associated with gamebird management accounts for up to three-quarters of annual mortality in Hen Harriers Circus cyaneus DOI Creative Commons
Steven R. Ewing, Cathleen E. Thomas,

Nigel Butcher

et al.

Biological Conservation, Journal Year: 2023, Volume and Issue: 283, P. 110072 - 110072

Published: May 11, 2023

Predators are frequently victims of wildlife crime due to conflicts with human interests. Where predators protected, killing may occur covertly and novel methods, including satellite tracking, often required assess population consequences. Wildlife persists in the British uplands, where raptors illegally killed on moorland managed for Red Grouse Lagopus lagopus scotica shooting. To understand impacts one such species, Hen Harrier Circus cyaneus, we analysed data from 148 individuals tracked across Britain between 2014 2021. Using remotely sensed land-use continuous-time survival quantified rates, contributions natural causes illegal mortality, spatial temporal associations mortality land grouse Annual was low, especially among first-year birds (males: 14 %; females: 30 %), accounting 27–43 % 75 subadult (1-2 years) harriers respectively. Illegal is likely attributable moor management because i) a 10 increase use resulted 43 risk; ii) strong overlap existed extent 20 km squares, identifying hotspots northern England northeast Scotland; iii) death showed different patterns; iv) timing peaked around shooting season during breeding territory establishment. Governments have failed reduce Harriers other our results emphasise that further legislative reform needed tackle this enduring criminality.

Language: Английский

Citations

8

Safeguarding Biodiversity: Remote Forest Poaching Camp Detection with CNN and Spatial Transformer Networks DOI

Yashu Yashu,

Vinay Kukreja,

Mukesh Kumar

et al.

Published: May 2, 2024

Language: Английский

Citations

2

Predicting groundwater contamination to protect the storm-exposed vulnerable DOI Creative Commons
Jacob Hochard, Nino Abashidze, Ranjit Bawa

et al.

Climate Risk Management, Journal Year: 2023, Volume and Issue: 40, P. 100499 - 100499

Published: Jan. 1, 2023

Domestic wells provide drinking water to 44 million people nationwide. Many of these wells, which remain federally unregulated and rarely tested for pollutants, serve rural populations clustered near surface-contaminated sites (e.g., hazardous waste sites, animal agriculture operations, coal ash ponds, etc.). The potential natural disasters deteriorate quality is well documented. Less understood whether opportunistic post-disaster sampling might underrepresent vulnerable populations. When disaster strikes, campaigns offer a glimpse into the exposed residents. We examined over 8,000 samples from 2016 2017 measure Hurricane Matthew's impact on presence indicator bacteria. Bacteria was predicted at household level following landfall. residential addresses associated with birth records as clinically estimated dates conception were used predict likelihood bacteria in sources that unsampled but likely have served pregnant women. estimate captures average contamination rates among households Our approach documents distribution risk where 2.7% sample (670 households) 75% total coliform presence. elevated small share nearby swine lagoons experienced most torrential rainfall. However, gap between sampled cannot otherwise be explained by storm event or proximity sites. Findings suggest sophisticated holistic prediction models may support targeting individual within groups are experience higher risks groundwater contamination.

Language: Английский

Citations

5