Geology Reports⌕ Search

Geology topics

Daniel P. Walsh

Publications and source records attributed to Daniel P. Walsh.

At least 37 records · Page 2Linked to original sources
USGS2024

Metagenomic sequencing sheds light on microbes putatively associated with pneumonia-related fatalities of white-tailed deer (Odocoileus virginianus)

With emerging infectious disease outbreaks in human, domestic and wild animal populations on the rise, improvements in pathogen characterization and surveillance are paramount for the protection of human and animal health, as well as the conservation of ecologically and economically important wildlife. Genomics offers a range of suitable tools to meet these goals, with metagenomic sequencing facilitating the characterization of whole microbial communities associated with emerging and endemic disease outbreaks. Here, we use metagenomic sequencing in a case-control study to identify microbes in lung tissue associated with newly observed pneumonia-related fatalities in 34 white-tailed deer ( Odocoileus virginianus ) in Wisconsin, USA. We identified 20 bacterial species that occurred in more than a single individual. Of these, only Clostridium novyi was found to substantially differ (in number of detections) between case and control sample groups; however, this difference was not statistically significant. We also detected several bacterial species associated with pneumonia and/or other diseases in ruminants ( Mycoplasma ovipneumoniae , Trueperella pyogenes , Pasteurella multocida , Anaplasma phagocytophilum , Fusobacterium necrophorum ); however, these species did not substantially differ between case and control sample groups. On average, we detected a larger number of bacterial species in case samples than controls, supporting the potential role of polymicrobial infections in this system. Importantly, we did not detect DNA of viruses or fungi, suggesting that they are not significantly associated with pneumonia in this system. Together, these results highlight the utility of metagenomic sequencing for identifying disease-associated microbes. This preliminary list of microbes will help inform future research on pneumonia-associated fatalities of white-tailed deer.

Microbial Genomics↗
USGS2023

Model-based surveillance system design under practical constraints with application to white-nose syndrome

Infectious diseases are powerful ecological forces structuring ecosystems, causing devastating economic impacts and disrupting society. Successful prevention and control of pathogens requires knowledge of the current scope and severity of disease, as well as the ability to forecast future disease dynamics. Assessment of the current situation as well as prediction of the future conditions, rely on spatially referenced information regarding the presence or absence of a pathogen, and the prevalence of the pathogen in the population. In particular, knowledge about the location of the disease front is foundational for deploying disease countermeasures to prevent further disease spread and focusing control efforts to reduce disease intensity in affected areas. In this paper, we develop a model-based approach to designing sampling strategies for wildlife disease surveillance at the disease front. Specifically, we use a mechanistic spatio-temporal model based on an underlying partial differential equation to track the disease dynamics and predict the disease prevalence in the future. We also devise an optimal surveillance system design at the disease front that takes into account the practical constraints of sampling. We evaluate the effectiveness of our proposed design via a simulation study and demonstrate the application of the proposed approach by designing a surveillance strategy for the pathogen that causes white-nose syndrome.

Environmental and Ecological Statistics↗
USGS2023

Bayesian forecasting of disease spread with little or no local data

Rapid and targeted management actions are a prerequisite to efficiently mitigate disease outbreaks. Targeted actions, however, require accurate spatial information on disease occurrence and spread. Frequently, targeted management actions are guided by non-statistical approaches that define the affected area by a pre-determined distance surrounding a small number of disease detections. As an alternative, we present a long-recognized but underutilized Bayesian technique that uses limited local data and informative priors to make statistically valid predictions and forecasts about disease occurrence and spread. As a case study, we use limited local data that were available after the detection of chronic wasting disease in Michigan, U.S. along with information rich priors obtained from a previous study in a neighboring state. Using these limited local data and informative priors, we generate statistically valid predictions of disease occurrence and spread for the Michigan study area. This Bayesian technique is conceptually and computationally simple, relies on little to no local data, and is competitive with non-statistical distance-based metrics in all performance evaluations. Bayesian modeling has added benefits because it allows practitioners to generate immediate forecasts of future disease conditions and provides a principled framework to incorporate new data as they accumulate. We contend that the Bayesian technique offers broad-scale benefits and opportunities to make statistical inference across a diversity of data-deficient systems, not limited to disease.

Scientific Reports↗
USGS2023

Assimilating ecological theory with empiricism: Using constrained generalized additive models to enhance survival analyses

1. Integrating ecological theory with empirical methods is ubiquitous in ecology using hierarchical Bayesian models. However, there has been little development focused on integration of ecological theory into models for survival analysis. Survival is a fundamental process, linking individual fitness with population dynamics, but incorporating life history strategies to inform survival estimation can be challenging because mortality processes occur at multiple scales. 2. We develop an approach to survival analysis, incorporating model constraints based on a species' life history strategy using functional analytical tools. Specifically, we structurally separate intrinsic patterns of mortality that arise from age-specific processes (e.g. increasing survival during early life stages due to growth or maturation, versus senescence) from extrinsic mortality patterns that arise over different periods of time (e.g. seasonal temporal shifts). We use shape constrained generalized additive models (CGAMs) to obtain age-specific hazard functions that incorporate theoretical information based on classical survivorship curves into the age component of the model and capture extrinsic factors in the time component. 3. We compare the performance of our modelling approach to standard survival modelling tools that do not explicitly incorporate species life history strategy in the model structure, using metrics of predictive power, accuracy, efficiency and computation time. We applied these models to two case studies that reflect different functional shapes for the underlying survivorship curves, examining age-period survival for white-tailed deer Odocoileus virginianus in Wisconsin, USA and Columbian sharp-tailed grouse T ympanuchus phasianellus columbianus in Colorado, USA.

Colorado, Wisconsin↗
USGS2023

Host vs. pathogen evolutionary arms race: Effects of exposure history on individual response to a genetically diverse pathogen

Introduction: Throughout their range, bighorn sheep ( Ovis canadensis ) populations have seen significant disease-associated declines. Unfortunately, understanding of the underlying epidemiological processes driving the disease dynamics in this species has hindered conservation efforts aimed at improving the health and long-term viability of these populations. Individual response to pathogen exposure emerges from dynamic interactions between competing evolutionary processes within the host and pathogen. The host’s adaptive immune system recognizes pathogens and mounts a defensive response. Pathogens have evolved strategies to overcome adaptive immune defenses including maintaining high genetic diversity through rapid evolution. The outcomes of this evolutionary warfare determine the success of pathogen invasion of the host and ultimately the success of conservation efforts. Methods: During an epizootic dominated by a single strain, we explore these host-pathogen dynamics by examining the variation in effects of pathogen invasion on captive bighorn sheep with differing histories of exposure to genetically diverse strains of Mycoplasma ovipneumoniae (Movi). We monitored clinical signs of disease and sampled animals and their environment to detect spread of Movi among 37 bighorn sheep separated into nine pens based on known exposure Results: We documented Movi transmission within and across pens and we detected Movi DNA in air, water, and invertebrate samples. Higher levels of antibody to Movi prior to the epizootic were associated with a lower likelihood of presenting clinical signs of pneumonia. Nonetheless, higher antibody levels in symptomatic individuals were associated with more severe progressive disease, increased probability and speed of pneumonia-induced mortality, and reduced likelihood of returning to a healthy state. Bighorn sheep with previous exposure to a strain other than the predominant epizootic strain were more likely to recover. Discussion: Our results indicate that Movi-strain variability was sufficient to overwhelm the adaptive host immunological defenses. This outcome indicates, in free-ranging herds, past exposure is likely insufficient to protect bighorn sheep from infection by new Movi strains, although it influences the progression of disease and recovery within the herd. Therefore, given Movi-strain variability and the lack of immunological protection from past exposure, focusing management efforts on minimizing the introduction of Movi into bighorn herds, through separation of domestic and bighorn sheep and avoidance of management activities that create commingling of bighorn sheep carrying differing Movi strains, will likely be the most effective approach for reducing the effects of disease and achieving bighorn sheep conservation goals.

Idaho, Oregon, South Dakota, Washington↗
USGS2023

Bayesian spatio-temporal survival analysis for all types of censoring with application to a wildlife disease study

In this article, we consider modeling arbitrarily censored survival data with spatio-temporal covariates. We demonstrate that under the piecewise constant hazard function, the likelihood for uncensored or right-censored subjects is proportional to the likelihood of multiple conditionally independent Poisson random variables. To address left- or interval-censored subjects, we propose to impute the exact event times and convert them into uncensored subjects, enabling the application of the integrated nested Laplace approximation to update model parameters using the imputed data. We introduce an iterative algorithm that alternates between imputing event times for left- and interval-censored subjects and re-estimating model parameters. The proposed method is assessed through a simulation study and applied to analyze a spatio-temporal survival dataset in a wildlife disease study investigating bovine tuberculosis in white-tailed deer in Michigan.

Environmetrics↗
USGS2023

Rapid prototyping for quantifying belief weights of competing hypotheses about emergent diseases

Emerging diseases can have devastating consequences for wildlife and require a rapid response. A critical first step towards developing appropriate management is identifying the etiology of the disease, which can be difficult to determine, particularly early in emergence. Gathering and synthesizing existing information about potential disease causes, by leveraging expert knowledge or relevant existing studies, provides a principled approach to quickly inform decision-making and management efforts. Additionally, updating the current state of knowledge as more information becomes available over time can reduce scientific uncertainty and lead to substantial improvement in the decision-making process and the application of management actions that incorporate and adapt to newly acquired scientific understanding. Here we present a rapid prototyping method for quantifying belief weights for competing hypotheses about the etiology of disease using a combination of formal expert elicitation and Bayesian hierarchical modeling. We illustrate the application of this approach for investigating the etiology of stony coral tissue loss disease (SCTLD) and discuss the opportunities and challenges of this approach for addressing emergent diseases. Lastly, we detail how our work may apply to other pressing management or conservation problems that require quick responses. We found the rapid prototyping methods to be an efficient and rapid means to narrow down the number of potential hypotheses, synthesize current understanding, and help prioritize future studies and experiments. This approach is rapid by providing a snapshot assessment of the current state of knowledge. It can also be updated periodically (e.g., annually) to assess changes in belief weights over time as scientific understanding increases. Synthesis and applications: The rapid prototyping approaches demonstrated here can be used to combine knowledge from multiple experts and/or studies to help with fast decision-making needed for urgent conservation issues including emerging diseases and other management problems that require rapid responses. These approaches can also be used to adjust belief weights over time as studies and expert knowledge accumulate and can be a helpful tool for adapting management decisions.

Journal of Environmental Management↗
USGS2023

Wildlife health surveillance: Gaps, needs and opportunities

Disease emergence represent a global threat for public health, economy, and biological conservation and most of the emerging diseases have zoonotic origin from wildlife. To prevent their spread and to support the implementation of control measures, disease surveillance and reporting systems are needed, and due to globalisation, these activities should be carried at world level. To define the main gaps affecting the performances of wildlife health surveillance and reporting systems at word level, we analysed the Wildlife Disease Surveillance Survey submitted to all Members of World Organisation for Animal Health which inquire on structure and limits of surveillance and reporting systems. The response from 103 Members, covering all world areas, showed that 54.4% of them have a wildlife disease surveillance programme and 66% implemented a strategy to manage their spread. The lack of dedicated budget affected the possibility of outbreak investigations, sampling collection and diagnostic testing. Although most Members maintain records relating to wildlife mortality or morbidity events in centralised databases, data analyses and disease risk assessment are reported as main needs. Our evaluation of surveillance capacity showed an overall low level associated to great variability between Members not restricted to specific geographical area. A better recognition of the value of wildlife disease surveillance resulting in allocation of adequate resources by governments would make wildlife disease surveillance systems more functional. Moreover, the future comprehension of the influence of socio-economic, cultural and biodiversity aspects will be essential to increase better disease surveillance under One Health approach.

Scientific and Technical Review↗
USGS2022

Diagnostic testing of chronic wasting disease in white-tailed deer (Odocoileus virginianus) by RT-QuIC using multiple tissues

Chronic wasting disease (CWD) is a fatal prion disease affecting cervids (deer, elk, moose). Current methods to monitor individual disease state include highly invasive antemortem rectal biopsy or postmortem brain biopsy. Efficient, sensitive, and selective antemortem and postmortem testing of populations would increase knowledge of the dynamics of CWD epizootics as well as provide a means to track CWD progression into previously unaffected areas. Here, we analyzed the presence of CWD prions in skin samples from two easily accessed locations (ear and belly) from 30 deceased white-tailed deer ( Odocoileus viginianus ). The skin samples were enzymatically digested and analyzed by real-time quaking-induced conversion (RT-QuIC). The diagnostic sensitivity of the ear and belly skin samples were both 95%, and the diagnostic specificity of the ear and belly skin were both 100%. Additionally, the location of the skin biopsy on the ear does not affect specificity or sensitivity. These results demonstrate the efficacy of CWD diagnosis with skin biopsies using RT-QuIC. This method could be useful for large scale antemortem population testing.

Wisconsin↗
USGS2022

Agricultural land use shapes dispersal in white-tailed deer (Odocoileus virginianus)

Background Dispersal is a fundamental process to animal population dynamics and gene flow. In white-tailed deer (WTD; Odocoileus virginianus ), dispersal also presents an increasingly relevant risk for the spread of infectious diseases. Across their wide range, WTD dispersal is believed to be driven by a suite of landscape and host behavioral factors, but these can vary by region, season, and sex. Our objectives were to (1) identify dispersal events in Wisconsin WTD and determine drivers of dispersal rates and distances, and (2) determine how landscape features (e.g., rivers, roads) structure deer dispersal paths. Methods We developed an algorithmic approach to detect dispersal events from GPS collar data for 590 juvenile, yearling, and adult WTD. We used statistical models to identify host and landscape drivers of dispersal rates and distances, including the role of agricultural land use, the traversability of the landscape, and potential interactions between deer. We then performed a step selection analysis to determine how landscape features such as agricultural land use, elevation, rivers, and roads affected deer dispersal paths. Results Dispersal predominantly occurred in juvenile males, of which 64.2% dispersed, with dispersal events uncommon in other sex and age classes. Juvenile male dispersal probability was positively associated with the proportion of the natal range that was classified as agricultural land use, but only during the spring. Dispersal distances were typically short (median 5.77 km, range: 1.3–68.3 km), especially in the fall. Further, dispersal distances were positively associated with agricultural land use in potential dispersal paths but negatively associated with the number of proximate deer in the natal range. Lastly, we found that, during dispersal, juvenile males typically avoided agricultural land use but selected for areas near rivers and streams. Conclusion Land use—particularly agricultural—was a key driver of dispersal rates, distances, and paths in Wisconsin WTD. In addition, our results support the importance of deer social environments in shaping dispersal behavior. Our findings reinforce knowledge of dispersal ecology in WTD and how landscape factors—including major rivers, roads, and land-use patterns—structure host gene flow and potential pathogen transmission.

Wisconsin↗
USGS2022

A call to action: Standardizing white-tailed deer harvest data in the Midwestern United States and implications for quantitative analysis and disease management

Recreational hunting has been the dominant game management and conservation mechanism in the United States for the past century. However, there are numerous modern-day issues that reduce the viability and efficacy of hunting-based management, such as fewer hunters, overabundant wildlife populations, limited access, and emerging infectious diseases in wildlife. Quantifying the drivers of recreational harvest by hunters could inform potential management actions to address these issues, but this is seldom comprehensively accomplished because data collection practices limit some analytical applications (e.g., differing spatial scales of harvest regulations and harvest data). Additionally, managing large-scale issues, such as infectious diseases, requires collaborations across management agencies, which is challenging or impossible if data are not standardized. Here we discuss modern issues with the prevailing wildlife management framework in the United States from an analytical point of view with a case study of white-tailed deer ( Odocoileus virginianus ) in the Midwest. We have four aims: (1) describe the interrelated processes that comprise hunting and suggest improvements to current data collections systems, (2) summarize data collection systems employed by state wildlife management agencies in the Midwestern United States and discuss potential for large-scale data standardization, (3) assess how aims 1 and 2 influence managing infectious diseases in hunted wildlife, and (4) suggest actionable steps to help guide data collection standards and management practices. To achieve these goals, Wisconsin Department of Natural Resources disseminated a questionnaire to state wildlife agencies (Illinois, Indiana, Iowa, Kentucky, Michigan, Minnesota, Missouri, Ohio, Wisconsin), and we report and compare their harvest management structures, data collection practices, and responses to chronic wasting disease. We hope our “call to action” encourages re-evaluation, coordination, and improvement of harvest and management data collection practices with the goal of improving the analytical potential of these data. A deeper understanding of the strengths and deficiencies of our current management systems in relation to harvest and management data collection methods could benefit the future development of comprehensive and collaborative management and research initiatives (e.g., adaptive management) for wildlife and their diseases.

Illinois, Indiana, Iowa, Kentucky, Michigan, Minne↗
USGS2022

Informing surveillance through the characterization of outbreak potential of chronic wasting disease in white-tailed deer

Understanding the role that an environmental prion reservoir plays in the outbreak dynamics of chronic wasting disease (CWD) in free ranging white-tailed deer ( Odocoileus virginianus ) is critical for the allocation of disease surveillance resources by state and provincial wildlife agencies. We hypothesized that demographic, ecological, and epidemiological configurations naturally attenuate epidemic risk despite the introduction of infectious prions into a susceptible population of deer, but the magnitude of infectious prions in the environmental prion reservoir complicate outbreak expectations. We developed a Susceptible-Latent-Exposed-Infective (SLEI) compartment model to represent the dynamics of CWD epidemics in free-ranging white-tailed deer, then used the basic reproductive ratio ( R 0 "> R0 ) to pinpoint counties under which pathogenic introduction (transport of infectious bodily fluids, tissues, and carcasses through natural or anthropogenic means) naturally produced (or failed to produce) an epidemic. We found that the outlook for an epidemic hinged on transmission rates, the magnitude of environmental contamination, and system type (density-, frequency-, or density/frequency-dependent). CWD can persistently infect individuals living in a contaminated environment even if direct transmission is insufficient to sustain circulation. Theoretical results show that transmission of CWD cannot be exclusively density dependent, and must behave as either a mix between frequency and density dependent, or strictly frequency dependent. While the compartment model is a simplistic representation of reality and did not contain many complicating biological considerations, it was immediately useful in hypothesis generation, motivating the collection of additional data for use in more biologically detailed models, and in the allocation of finite surveillance resources to place emphasis on data collection in areas where an introduction of infectious prions is comparatively more likely to result in an epidemic.

Ecological Modelling↗
USGS2022

Cause of death, pathology, and chronic wasting disease status of white-tailed deer (Odocoileus virginianus) mortalities in Wisconsin

White-tailed deer (WTD; Odocoileus virginianus ) are a critical species for ecosystem function and wildlife management. As such, studies of cause-specific mortality among WTD have long been used to understand population dynamics. However, detailed pathological information is rarely documented for free-ranging WTD, especially in regions with a high prevalence of chronic wasting disease (CWD). This leaves a significant gap in understanding how CWD is associated with disease processes or comorbidities which may subsequently alter broader population dynamics. In this study, we investigated unknown mortalities among collared WTD in southwestern Wisconsin, an area of high CWD prevalence. We tested for associations between CWD and other disease processes and used a network approach to test for co-occurring disease processes. Predation and infectious disease were top suspected causes of death, with high prevalence of CWD (42.4%; of 245 evaluated) and pneumonia (51.2%; of 168 evaluated) in our sample. CWD prevalence increased with age, before decreasing among older individuals, with more older females than males in our sample. Females were more likely to be CWD positive, and while this was not statistically significant when accounting for age, females were significantly more likely to die with end-stage CWD than were males and may consequently be an underrecognized source of CWD transmission. Presence of CWD was associated with emaciation, atrophy of marrow fat and hematopoietic cells, and ectoparasitism (i.e., lice, ticks). Occurrences of severe infectious disease pathologies clustered together (e.g., pneumonia, CWD), as compared to non-infectious or low severity processes (e.g., sarcocystosis). However, pneumonia cases were not fully explained by CWD status. With the prevalence of CWD increasing across North America, our results highlight the critical importance of understanding the potential role of CWD in favoring or maintaining disease processes of importance for deer population health and dynamics.

Wisconsin↗
USGS2021

Space matters: Host spatial structure and the dynamics of plague transmission

The development of models to elucidate the transmission pathways and dynamics of wildlife diseases remains challenging. Sylvatic plague, caused by the bacterium Yersinia pestis ( Yp ), is an infectious zoonotic disease that primarily affects wild rodents, including prairie dogs ( Cynomys spp.) in North America. Proposed transmission pathways for Yp include flea bites, direct contacts between hosts, and environmental reservoirs (e.g. soil, carcasses). We developed a spatially explicit, agent-based model of Yp transmission to explore the effects of alternative transmission pathways, different disease initiation mechanisms (host or fleas), parameter uncertainty, and spatial structure of hosts. A particularly novel aspect of our model was the integration of ecological models with traditional disease models. Specifically, we used estimates from spatial capture-recapture models to generate data-driven spatial distributions, densities, and contact rates to capture the spatial structure of prairie dogs. We simulated ~9 million scenarios across a wide range of parameter values and conducted sensitivity analyses to determine the most influential parameters on the number of flea-days (sum of the mean number of fleas on hosts each day of the simulation), number of newly infected hosts per day, the time to depopulation (<20 prairie dogs remaining), and the proportion of the prairie dog population remaining at the end of the simulation (after 150 days). When including spatial structure, we found the probability of transmission via environmental sources of Yp (i.e. carcasses) had the greatest influence on model results when Yp infection was initiated in prairie dog hosts, rather than in fleas. Conversely, the mechanism of transmission by fleas to prairie dogs had the greatest influence on model results when Yp infection was initiated in fleas (i.e. via introduction by carnivores, a migrant prairie dog, or other mammalian host). Uncertainty in parameter estimates, particularly those related to the transmission pathways of Yp, continue to hamper efforts to realistically model plague dynamics in wild rodents. Our results elucidate the complexity of the flea-plague-prairie dog system and reiterate the importance of research on Yp transmission mechanisms to provide a full understanding of this disease. Our results also emphasize the importance of realistic estimates of spatial structure for exploring transmission dynamics of wildlife diseases and provide a framework for generating a data-driven description of spatial structure.

Ecological Modelling↗
USGS2021

Recovering individual-level spatial inference from aggregated binary data

Binary regression models are commonly used in disciplines such as epidemiology and ecology to determine how spatial covariates influence individuals. In many studies, binary data are shared in a spatially aggregated form to protect privacy. For example, rather than reporting the location and result for each individual that was tested for a disease, researchers may report that a disease was detected or not detected within geopolitical units. Often, the spatial aggregation process obscures the values of response variables, spatial covariates, and locations of each individual, which makes recovering individual-level inference difficult. We show that applying a series of transformations, including a change of support, to a bivariate point process model allows researchers to recover individual-level inference for spatial covariates from spatially aggregated binary data. The series of transformations preserves the convenient interpretation of desirable binary regression models that are commonly applied to individual-level data. Using a simulation experiment, we compare the performance of our proposed method under varying types of spatial aggregation against the performance of standard approaches using the original individual-level data. We illustrate our method by modeling individual-level probability of infection using a data set that has been aggregated to protect an at-risk and endangered species of bats. Our simulation experiment and data illustration demonstrate the utility of the proposed method when access to original non-aggregated data is impractical or prohibited.

Northeast and Midwest United States↗
USGS2021

Efficacy of fenbendazole and ivermectin against Trichuris spp. in African green monkeys (Chlorocebus sabaeus) in Barbados West Indies

Trichuris spp. are common helminths in NHP, and benzimidazoles and avermectins have both been used to treat these intestinal parasites. The current study compared the efficacy of fenbendazole and ivermectin against natural infection of Trichuris spp. in African green monkeys ( Chlorocebus sabaeus ). Anthelmintic-naive animals ( n = 65) were randomly assigned to 4 groups: an untreated control group, and 3 groups treated with either fenbendazole, ivermectin, or both compounds. Fecal samples were collected before treatment and on days 7, 14, 28, and 60 after treatment, and fecal egg counts (FEC) were determined by using fecal flotation. The mean percentages of FEC reduction at day 60 were 100%, 86%, and 100% for treatment with fenbendazole, ivermectin, and both compounds, respectively. Analyzing the time series of FEC by using a Bayesian generalized linear model showed no significant difference in the proportional reduction in FEC among the 3 treatment groups, although all FEC from treated groups were significantly lower than the FEC of the control group. In contrast, the probability of shedding was highest in the ivermectin group and the lowest in the animals treated with both compounds. The probability of shedding differed significantly between the fenbendazole and ivermectin groups and between the ivermectin and combined-treatment groups. In conclusion, both fenbendazole and ivermectin are effective anthelmintics in treating Trichuris spp. infection in African green monkeys. All treatment groups showed significant reductions in FEC when compared with baseline counts and control animals; however, fenbendazole may be more effective than ivermectin when used solely or in combination with other anthelmintic treatments.

Journal of the American Association for Laboratory↗
USGS2020

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

Motion‐activated wildlife cameras (or “camera traps”) are frequently used to remotely and noninvasively observe animals. The vast number of images collected from camera trap projects has prompted some biologists to employ machine learning algorithms to automatically recognize species in these images, or at least filter‐out images that do not contain animals. These approaches are often limited by model transferability, as a model trained to recognize species from one location might not work as well for the same species in different locations. Furthermore, these methods often require advanced computational skills, making them inaccessible to many biologists. We used 3 million camera trap images from 18 studies in 10 states across the United States of America to train two deep neural networks, one that recognizes 58 species, the “species model,” and one that determines if an image is empty or if it contains an animal, the “empty‐animal model.” Our species model and empty‐animal model had accuracies of 96.8% and 97.3%, respectively. Furthermore, the models performed well on some out‐of‐sample datasets, as the species model had 91% accuracy on species from Canada (accuracy range 36%–91% across all out‐of‐sample datasets) and the empty‐animal model achieved an accuracy of 91%–94% on out‐of‐sample datasets from different continents. Our software addresses some of the limitations of using machine learning to classify images from camera traps. By including many species from several locations, our species model is potentially applicable to many camera trap studies in North America. We also found that our empty‐animal model can facilitate removal of images without animals globally. We provide the trained models in an R package (MLWIC2: Machine Learning for Wildlife Image Classification in R), which contains Shiny Applications that allow scientists with minimal programming experience to use trained models and train new models in six neural network architectures with varying depths.

Ecology and Evolution↗