ABSTRACT
Competition, facilitation, and predation offer alternative explanations for successional patterns of migratory herbivores. However, these interactions are difficult to measure, leaving uncertainty about the mechanisms underlying body-size-dependent grazing-and even whether succession occurs at all. We used data from an 8-year camera-trap survey, GPS-collared herbivores, and fecal DNA metabarcoding to analyze the timing, arrival order, and interactions among migratory grazers in Serengeti National Park. Temporal grazing succession is characterized by a "push-pull" dynamic: Competitive grazing nudges zebra ahead of co-migrating wildebeest, whereas grass consumption by these large-bodied migrants attracts trailing, small-bodied gazelle that benefit from facilitation. "Natural experiments" involving intense wildfires and rainfall respectively disrupted and strengthened these effects. Our results highlight a balance between facilitative and competitive forces in co-regulating large-scale ungulate migrations.
Subject(s)
Animal Migration , Antelopes , Equidae , Herbivory , Parks, Recreational , Animals , Antelopes/physiology , Equidae/physiology , Poaceae , Kenya , TanzaniaABSTRACT
Classifying specimens is a critical component of ecological research, biodiversity monitoring and conservation. However, manual classification can be prohibitively time-consuming and expensive, limiting how much data a project can afford to process. Computer vision, a form of machine learning, can help overcome these problems by rapidly, automatically and accurately classifying images of specimens. Given the diversity of animal species and contexts in which images are captured, there is no universal classifier for all species and use cases. As such, ecologists often need to train their own models. While numerous software programs exist to support this process, ecologists need a fundamental understanding of how computer vision works to select appropriate model workflows based on their specific use case, data types, computing resources and desired performance capabilities. Ecologists may also face characteristic quirks of ecological datasets, such as long-tail distributions, 'unknown' species, similarity between species and polymorphism within species, which impact the efficacy of computer vision. Despite growing interest in computer vision for ecology, there are few resources available to help ecologists face the challenges they are likely to encounter. Here, we present a gentle introduction for species classification using computer vision. In this manuscript and associated GitHub repository, we demonstrate how to prepare training data, basic model training procedures, and methods for model evaluation and selection. Throughout, we explore specific considerations ecologists should make when training classification models, such as data domains, feature extractors and class imbalances. With these basics, ecologists can adjust their workflows to achieve research goals and/or account for uncertainty in downstream analysis. Our goal is to provide guidance for ecologists for getting started in or improving their use of machine learning for visual classification tasks.
Subject(s)
Computers , Neural Networks, Computer , Animals , Machine Learning , BiodiversityABSTRACT
Extreme weather events perturb ecosystems and increasingly threaten biodiversity1. Ecologists emphasize the need to forecast and mitigate the impacts of these events, which requires knowledge of how risk is distributed among species and environments. However, the scale and unpredictability of extreme events complicate risk assessment1-4-especially for large animals (megafauna), which are ecologically important and disproportionately threatened but are wide-ranging and difficult to monitor5. Traits such as body size, dispersal ability and habitat affiliation are hypothesized to determine the vulnerability of animals to natural hazards1,6,7. Yet it has rarely been possible to test these hypotheses or, more generally, to link the short-term and long-term ecological effects of weather-related disturbance8,9. Here we show how large herbivores and carnivores in Mozambique responded to Intense Tropical Cyclone Idai, the deadliest storm on record in Africa, across scales ranging from individual decisions in the hours after landfall to changes in community composition nearly 2 years later. Animals responded behaviourally to rising floodwaters by moving upslope and shifting their diets. Body size and habitat association independently predicted population-level impacts: five of the smallest and most lowland-affiliated herbivore species declined by an average of 28% in the 20 months after landfall, while four of the largest and most upland-affiliated species increased by an average of 26%. We attribute the sensitivity of small-bodied species to their limited mobility and physiological constraints, which restricted their ability to avoid the flood and endure subsequent reductions in the quantity and quality of food. Our results identify general traits that govern animal responses to severe weather, which may help to inform wildlife conservation in a volatile climate.
Subject(s)
Body Size , Cyclonic Storms , Mammals , Animals , Altitude , Biodiversity , Carnivory , Conservation of Natural Resources , Diet/veterinary , Ecosystem , Extreme Weather , Floods , Forecasting , Herbivory , Mammals/anatomy & histology , Mammals/physiology , MozambiqueABSTRACT
The landscape of fear (LOF) concept posits that prey navigate spatial heterogeneity in perceived predation risk, balancing risk mitigation against other activities necessary for survival and reproduction. These proactive behavioral responses to risk can affect individual fitness, population dynamics, species interactions, and coexistence. Yet, antipredator responses in free-ranging prey often contradict expectations, raising questions about the generality and scalability of the LOF framework and suggesting that a purely spatial, static LOF conceptualization may be inadequate. Here, we outline a 'dynamic' LOF framework that explicitly incorporates time to account for predictable spatiotemporal variation in risk-resource trade-offs. This integrated approach suggests novel predictions about predator effects on prey behaviors to refine understanding of the role predators play in ecological communities.
Subject(s)
Fear , Predatory Behavior , Animals , Food Chain , Population Dynamics , Predatory Behavior/physiologyABSTRACT
Large mammalian herbivores use a diverse array of strategies to survive predator encounters including flight, grouping, vigilance, warning signals, and fitness indicators. While anti-predator strategies appear to be driven by specific predator traits, no prior studies have rigorously evaluated whether predator hunting characteristics predict reactive anti-predator responses. We experimentally investigated behavioral decisions made by free-ranging impala, wildebeest, and zebra during encounters with model predators with different functional traits. We hypothesized that the choice of response would be driven by a predator's hunting style (i.e., ambush vs. coursing) while the intensity at which the behavior was performed would correlate with predator traits that contribute to the prey's relative risk (i.e., each predator's prey preference, prey-specific capture success, and local predator density). We found that the choice and intensity of anti-predator behaviors were both shaped by hunting style and relative risk factors. All prey species directed longer periods of vigilance towards predators with higher capture success. The decision to flee was the only behavior choice driven by predator characteristics (capture success and hunting style) while intensity of vigilance, frequency of alarm-calling, and flight latency were modulated based on predator hunting strategy and relative risk level. Impala regulated only the intensity of their behaviors, while zebra and wildebeest changed both type and intensity of response based on predator traits. Zebra and impala reacted to multiple components of predation threat, while wildebeest responded solely to capture success. Overall, our findings suggest that certain behaviors potentially facilitate survival under specific contexts and that prey responses may reflect the perceived level of predation risk, suggesting that adaptive functions to reactive anti-predator behaviors may reflect potential trade-offs to their use. The strong influence of prey species identity and social and environmental context suggest that these factors may interact with predator traits to determine the optimal response to immediate predation threat.
Subject(s)
Adaptation, Physiological , Antelopes/physiology , Behavior, Animal/physiology , Ecosystem , Herbivory/classification , Population Dynamics/statistics & numerical data , Predatory Behavior/physiology , Animals , Food ChainABSTRACT
Human activity and land use change impact every landscape on Earth, driving declines in many animal species while benefiting others. Species ecological and life history traits may predict success in human-dominated landscapes such that only species with "winning" combinations of traits will persist in disturbed environments. However, this link between species traits and successful coexistence with humans remains obscured by the complexity of anthropogenic disturbances and variability among study systems. We compiled detection data for 24 mammal species from 61 populations across North America to quantify the effects of (1) the direct presence of people and (2) the human footprint (landscape modification) on mammal occurrence and activity levels. Thirty-three percent of mammal species exhibited a net negative response (i.e., reduced occurrence or activity) to increasing human presence and/or footprint across populations, whereas 58% of species were positively associated with increasing disturbance. However, apparent benefits of human presence and footprint tended to decrease or disappear at higher disturbance levels, indicative of thresholds in mammal species' capacity to tolerate disturbance or exploit human-dominated landscapes. Species ecological and life history traits were strong predictors of their responses to human footprint, with increasing footprint favoring smaller, less carnivorous, faster-reproducing species. The positive and negative effects of human presence were distributed more randomly with respect to species trait values, with apparent winners and losers across a range of body sizes and dietary guilds. Differential responses by some species to human presence and human footprint highlight the importance of considering these two forms of human disturbance separately when estimating anthropogenic impacts on wildlife. Our approach provides insights into the complex mechanisms through which human activities shape mammal communities globally, revealing the drivers of the loss of larger predators in human-modified landscapes.
Subject(s)
Animals, Wild , Life History Traits , Animals , Ecosystem , Human Activities , Humans , Mammals , North AmericaABSTRACT
The mere threat of predation may incite behavioral changes in prey that lead to community-wide impacts on productivity, biodiversity, and nutrient cycling. The paucity of experimental manipulations, however, has contributed to controversy over the strength of this pathway in wide-ranging vertebrate systems. We investigated whether simulated gray wolf (Canis lupus) presence can induce behaviorally-mediated trophic cascades, specifically, whether the 'fear' of wolf olfactory cues alone can change deer foraging behavior in ways that affect plants and soils. Wolves were recently removed from the Cedar Creek Ecosystem Science Reserve (Minnesota, USA), such that consumptively mediated predator effects were negligible. At 32 experimental plots, we crossed two nested treatments: wolf urine application and herbivore exclosures. We deployed camera traps to quantify how white-tailed deer (Odocoileus virginianus) adjusted their spatiotemporal habitat use, foraging, and vigilance in response to wolf cues and how these behavioral changes affected plant productivity, plant communities, and soil nutrients. Weekly applications of wolf urine significantly altered deer behavior, but deer responses did not cascade to affect plant or soil properties. Deer substantially reduced crepuscular activity at wolf-simulated sites compared to control locations. As wolves in this area predominantly hunted during mornings and evenings, this response potentially allows deer to maximize landscape use by accessing dangerous areas when temporal threat is low. Our experiment suggests that prey may be sensitive to 'dynamic' predation risk that is structured across both space and time and, consequentially, prey use of risky areas during safe times may attenuate behaviorally-mediated trophic cascades at the predator-prey interface.
Subject(s)
Deer , Wolves , Animals , Ecosystem , Food Chain , Minnesota , Predatory BehaviorABSTRACT
Understanding the role of species interactions within communities is a central focus of ecology. A key challenge is to understand variation in species interactions along environmental gradients. The stress gradient hypothesis posits that positive interactions increase and competitive interactions decrease with increasing consumer pressure or environmental stress. This hypothesis has received extensive attention in plant community ecology, but only a handful of tests in animals. Furthermore, few empirical studies have examined multiple co-occurring stressors. Here we test predictions of the stress gradient hypothesis using the occurrence of mixed-species groups in six common grazing ungulate species within the Serengeti-Mara ecosystem. We use mixed-species groups as a proxy for potential positive interactions because they may enhance protection from predators or increase access to high-quality forage. Alternatively, competition for resources may limit the formation of mixed-species groups. Using more than 115,000 camera trap observations collected over 5 yr, we found that mixed-species groups were more likely to occur in risky areas (i.e., areas closer to lion vantage points and in woodland habitat where lions hunt preferentially) and during time periods when resource levels were high. These results are consistent with the interpretation that stress from high predation risk may contribute to the formation of mixed-species groups, but that competition for resources may prevent their formation when food availability is low. Our results are consistent with support for the stress gradient hypothesis in animals along a consumer pressure gradient while identifying the potential influence of a co-occurring stressor, thus providing a link between research in plant community ecology on the stress gradient hypothesis, and research in animal ecology on trade-offs between foraging and risk in landscapes of fear.
Subject(s)
Ecosystem , Lions , Animals , Ecology , Mammals , Predatory BehaviorABSTRACT
Camera trap technology has galvanized the study of predator-prey ecology in wild animal communities by expanding the scale and diversity of predator-prey interactions that can be analysed. While observational data from systematic camera arrays have informed inferences on the spatiotemporal outcomes of predator-prey interactions, the capacity for observational studies to identify mechanistic drivers of species interactions is limited. Experimental study designs that utilize camera traps uniquely allow for testing hypothesized mechanisms that drive predator and prey behaviour, incorporating environmental realism not possible in the laboratory while benefiting from the distinct capacity of camera traps to generate large datasets from multiple species with minimal observer interference. However, such pairings of camera traps with experimental methods remain underutilized. We review recent advances in the experimental application of camera traps to investigate fundamental mechanisms underlying predator-prey ecology and present a conceptual guide for designing experimental camera trap studies. Only 9% of camera trap studies on predator-prey ecology in our review use experimental methods, but the application of experimental approaches is increasing. To illustrate the utility of camera trap-based experiments using a case study, we propose a study design that integrates observational and experimental techniques to test a perennial question in predator-prey ecology: how prey balance foraging and safety, as formalized by the risk allocation hypothesis. We discuss applications of camera trap-based experiments to evaluate the diversity of anthropogenic influences on wildlife communities globally. Finally, we review challenges to conducting experimental camera trap studies. Experimental camera trap studies have already begun to play an important role in understanding the predator-prey ecology of free-living animals, and such methods will become increasingly critical to quantifying drivers of community interactions in a rapidly changing world. We recommend increased application of experimental methods in the study of predator and prey responses to humans, synanthropic and invasive species, and other anthropogenic disturbances.
Subject(s)
Introduced Species , Predatory Behavior , AnimalsABSTRACT
Developing techniques to quantify the spread and severity of diseases afflicting wildlife populations is important for disease ecology, animal ecology, and conservation. Giraffes (Giraffa camelopardalis) are in the midst of a dramatic decline, but it is not known whether disease is playing an important role in the broad-scale population reductions. A skin disorder referred to as giraffe skin disease (GSD) was recorded in 1995 in one giraffe population in Uganda. Since then, GSD has been detected in 13 populations in seven African countries, but good descriptions of the severity of this disease are not available. We photogrammetrically analyzed camera trap images from both Ruaha and Serengeti National parks in Tanzania to quantify GSD severity. Giraffe skin disease afflicts the limbs of giraffes in Tanzania, and we quantified severity by measuring the vertical length of the GSD lesion in relation to the total leg length. Applying the Jenks natural breaks algorithm to the lesion proportions that we derived, we classified individual giraffes into disease categories (none, mild, moderate, and severe). Scaling up to the population level, we predicted the proportion of the Ruaha and Serengeti giraffe populations with mild, moderate, and severe GSD. This study serves to demonstrate that camera traps presented an informative platform for examinations of skin disease ecology.
Subject(s)
Antelopes , Photogrammetry/veterinary , Skin Diseases/veterinary , Animals , Photogrammetry/methods , Skin Diseases/diagnosis , Skin Diseases/pathologyABSTRACT
Having accurate, detailed, and up-to-date information about the location and behavior of animals in the wild would improve our ability to study and conserve ecosystems. We investigate the ability to automatically, accurately, and inexpensively collect such data, which could help catalyze the transformation of many fields of ecology, wildlife biology, zoology, conservation biology, and animal behavior into "big data" sciences. Motion-sensor "camera traps" enable collecting wildlife pictures inexpensively, unobtrusively, and frequently. However, extracting information from these pictures remains an expensive, time-consuming, manual task. We demonstrate that such information can be automatically extracted by deep learning, a cutting-edge type of artificial intelligence. We train deep convolutional neural networks to identify, count, and describe the behaviors of 48 species in the 3.2 million-image Snapshot Serengeti dataset. Our deep neural networks automatically identify animals with >93.8% accuracy, and we expect that number to improve rapidly in years to come. More importantly, if our system classifies only images it is confident about, our system can automate animal identification for 99.3% of the data while still performing at the same 96.6% accuracy as that of crowdsourced teams of human volunteers, saving >8.4 y (i.e., >17,000 h at 40 h/wk) of human labeling effort on this 3.2 million-image dataset. Those efficiency gains highlight the importance of using deep neural networks to automate data extraction from camera-trap images, reducing a roadblock for this widely used technology. Our results suggest that deep learning could enable the inexpensive, unobtrusive, high-volume, and even real-time collection of a wealth of information about vast numbers of animals in the wild.
Subject(s)
Animals, Wild/physiology , Behavior, Animal/physiology , Algorithms , Animals , Artificial Intelligence , Ecology/methods , Ecosystem , Humans , Machine Learning , Neural Networks, ComputerABSTRACT
Many organisms use cues and signals beyond human sensitivity during social interactions. It is important to take into account how organisms perceive their worlds when trying to understand their behavior and ecology. Sensitivity to the ultraviolet spectrum (UV; 300 - 400 nm) is found across multiple genera of birds, fish, reptiles, amphibians, and even mammals. This protocol describes a technique for examining organisms for the presence of UV-reflecting structures and a method for testing whether these cues are used as social signals in the context of mate choice. A spectrophotometer is used to detect the presence of UV reflectance and variation in reflective intensity between individuals and sexes. An example of this technique is presented in which a dichotomous mate choice test exposes sexually receptive individuals to opposite sex individuals whose visual appearance can be manipulated by filters that either transmit full spectrum or block UV wavelengths. This system allowed for the determination that female, but not male, sailfin mollies (Poecilia latipinna) were using UV markings as part of their mating decisions. These types of studies serve to expand our knowledge of the range of organisms that utilize UV and provide insight into how UV plays a role in their lives.