An efficient method of evaluating multiple concurrent management actions on invasive populations.

Evaluating the efficacy of management actions to control invasive species is crucial for maintaining funding and to provide feedback for the continual improvement of management efforts. However, it is often difficult to assess the efficacy of control methods due to limited resources for monitoring. Managers may view effort on monitoring as effort taken away from performing management actions. We developed a method to estimate invasive species abundance, evaluate management effectiveness, and evaluate population growth over time from a combination of removal activities (e.g., trapping, ground shooting) using only data collected during removal efforts (method of removal, date, location, number of animals removed, and effort). This dynamic approach allows for abundance estimation at discrete time points and the estimation of population growth between removal periods. To test this approach, we simulated over 1 million conditions, including varying the length of the study, the size of the area examined, the number of removal events, the capture rates, and the area impacted by removal efforts. Our estimates were unbiased (within 10% of truth) 81% of the time and were correlated with truth 91% of the time. This method performs well overall and, in particular, at monitoring trends in abundances over time. We applied this method to removal data from Mingo National Wildlife Refuge in Missouri from December 2015 to September 2019, where the management objective is elimination. Populations of feral swine on Mingo NWR have fluctuated over time but showed marked declines in the last 3-6 months of the time series corresponding to increased removal pressure. Our approach allows for the estimation of population growth across time (from both births and immigration) and therefore, provides a target removal rate (above that of the population growth) to ensure the population will decline. In Mingo NWR, the target monthly removal rate is 18% to cause a population decline. Our method provides advancement over traditional removal modeling approaches because it can be applied to evaluate management programs that use a broad range of removal techniques concurrently and whose management effort and spatial coverage vary across time.

Accounting for animal movement improves vaccination strategies against wildlife disease in heterogeneous landscapes.

Oral baiting is used to deliver vaccines to wildlife to prevent, control, and eliminate infectious diseases. A central challenge is how to spatially distribute baits to maximize encounters by target animal populations, particularly in urban and suburban areas where wildlife such as raccoons (Procyon lotor) are abundant and baits are delivered along roads. Methods from movement ecology that quantify movement and habitat selection could help to optimize baiting strategies by more effectively targeting wildlife populations across space. We developed a spatially explicit, individual-based model of raccoon movement and oral rabies vaccine seroconversion to examine whether and when baiting strategies that match raccoon movement patterns perform better than currently used baiting strategies in an oral rabies vaccination zone in greater Burlington, Vermont, USA. Habitat selection patterns estimated from locally radio-collared raccoons were used to parameterize movement simulations. We then used our simulations to estimate raccoon population rabies seroprevalence under currently used baiting strategies (actual baiting) relative to habitat selection-based baiting strategies (habitat baiting). We conducted simulations on the Burlington landscape and artificial landscapes that varied in heterogeneity relative to Burlington in the proportion and patch size of preferred habitats. We found that the benefits of habitat baiting strongly depended on the magnitude and variability of raccoon habitat selection and the degree of landscape heterogeneity within the baiting area. Habitat baiting improved seroprevalence over actual baiting for raccoons characterized as habitat specialists but not for raccoons that displayed weak habitat selection similar to radiocollared individuals, except when baits were delivered off roads where preferred habitat coverage and complexity was more pronounced. In contrast, in artificial landscapes with either more strongly juxtaposed favored habitats and/or higher proportions of favored habitats, habitat baiting performed better than actual baiting, even when raccoons displayed weak habitat preferences and where baiting was constrained to roads. Our results suggest that habitat selection-based baiting could increase raccoon population seroprevalence in urban-suburban areas, where practical, given the heterogeneity and availability of preferred habitat types in those areas. Our novel simulation approach provides a flexible framework to test alternative baiting strategies in multiclass landscapes to optimize bait-distribution strategies.

Optimizing management of invasions in an uncertain world using dynamic spatial models.

Dispersal drives invasion dynamics of nonnative species and pathogens. Applying knowledge of dispersal to optimize the management of invasions can mean the difference between a failed and a successful control program and dramatically improve the return on investment of control efforts. A common approach to identifying optimal management solutions for invasions is to optimize dynamic spatial models that incorporate dispersal. Optimizing these spatial models can be very challenging because the interaction of time, space, and uncertainty rapidly amplifies the number of dimensions being considered. Addressing such problems requires advances in and the integration of techniques from multiple fields, including ecology, decision analysis, bioeconomics, natural resource management, and optimization. By synthesizing recent advances from these diverse fields, we provide a workflow for applying ecological theory to advance optimal management science and highlight priorities for optimizing the control of invasions. One of the striking gaps we identify is the extremely limited consideration of dispersal uncertainty in optimal management frameworks, even though dispersal estimates are highly uncertain and greatly influence invasion outcomes. In addition, optimization frameworks rarely consider multiple types of uncertainty (we describe five major types) and their interrelationships. Thus, feedbacks from management or other sources that could magnify uncertainty in dispersal are rarely considered. Incorporating uncertainty is crucial for improving transparency in decision risks and identifying optimal management strategies. We discuss gaps and solutions to the challenges of optimization using dynamic spatial models to increase the practical application of these important tools and improve the consistency and robustness of management recommendations for invasions.

Optimal spatial prioritization of control resources for elimination of invasive species under demographic uncertainty.

Populations of invasive species often spread heterogeneously across a landscape, consisting of local populations that cluster in space but are connected by dispersal. A fundamental dilemma for invasive species control is how to optimally allocate limited fiscal resources across local populations. Theoretical work based on perfect knowledge of demographic connectivity suggests that targeting local populations from which migrants originate (sources) can be optimal. However, demographic processes such as abundance and dispersal can be highly uncertain, and the relationship between local population density and damage costs (damage function) is rarely known. We used a metapopulation model to understand how budget and uncertainty in abundance, connectivity, and the damage function, together impact return on investment (ROI) for optimal control strategies. Budget, observational uncertainty, and the damage function had strong effects on the optimal resource allocation strategy. Uncertainty in dispersal probability was the least important determinant of ROI. The damage function determined which resource prioritization strategy was optimal when connectivity was symmetric but not when it was asymmetric. When connectivity was asymmetric, prioritizing source populations had a higher ROI than allocating effort equally across local populations, regardless of the damage function, but uncertainty in connectivity structure and abundance reduced ROI of the optimal prioritization strategy by 57% on average depending on the control budget. With low budgets (monthly removal rate of 6.7% of population), there was little advantage to prioritizing resources, especially when connectivity was high or symmetric, and observational uncertainty had only minor effects on ROI. Allotting funding for improved monitoring appeared to be most important when budgets were moderate (monthly removal of 13-20% of the population). Our result showed that multiple sources of observational uncertainty should be considered concurrently for optimizing ROI. Accurate estimates of connectivity direction and abundance were more important than accurate estimates of dispersal rates. Developing cost-effective surveillance methods to reduce observational uncertainties, and quantitative frameworks for determining how resources should be spatially apportioned to multiple monitoring and control activities are important and challenging future directions for optimizing ROI for invasive species control programs.

Associations between types of greenery along neighborhood roads and weight status in different climates.

Obesity is a major international health concern. Neighborhood greenery has been identified as a critical factor for promoting health in urban areas, due in part to its apparent role in facilitating healthy weight by promoting physical activity. However, studies have used diverse greenery measures and spatial analysis units to ascertain this relationship. This study examined associations between street greenery and weight status at the residential address level across 500 to 2000m buffers in two climatically distinct communities, Phoenix, AZ, and Portland, OR. Greenery was measured using one-meter landcover data. Street greenery measures were designed to quantify the pedestrian environment along a gradient of suitability for promoting physical exercise. Weight status was defined by body mass index (BMI) calculated from weight and height information on driver's license records. BMI values were dichotomized at 25 into overweight or obese vs. neither. Approximately 500,000 BMI values in Phoenix and 225,000 in Portland were modelled by community using logistic regression. Street tree cover was consistently protective for healthy weight status across all buffer sizes after adjusting for potential confounders. Herbaceous street cover showed protective associations in Phoenix but harmful associations in Portland. Every 10% increase in street tree cover within 2000m was associated with 18% lower odds of being overweight or obese (adjusted odds ratio [AOR]: 0.82, 95% CI: 0.81 - 0.84 in Phoenix; 0.82, 95% CI: 0.81 - 0.83 in Portland). When compared to residents with less than 10% street tree cover within 2000m, those with greater than 10% tree cover had at least 13% (AOR for Portland: 0.87, 95% CI: 0.81 - 0.92) lower odds of being overweight or obese. Findings support the importance of urban street trees in very different climates for facilitating healthy weight status. They can inform greenery management to prioritize vegetation type and allocation decisions in limited urban spaces.

Assessing threats of non-native species to native freshwater biodiversity: Conservation priorities for the United States.

Non-native species pose one of the greatest threats to native biodiversity, and can have severe negative impacts in freshwater ecosystems. Identifying regions of spatial overlap between high freshwater biodiversity and high invasion pressure may thus better inform the prioritization of freshwater conservation efforts. We employ geospatial analysis of species distribution data to investigate the potential threat of non-native species to aquatic animal taxa across the continental United States. We mapped non-native aquatic plant and animal species richness and cumulative invasion pressure to estimate overall negative impact associated with species introductions. These distributions were compared to distributions of native aquatic animal taxa derived from the International Union for the Conservation of Nature (IUCN) database. To identify hotspots of native biodiversity we mapped total species richness, number of threatened and endangered species, and a community index of species rarity calculated at the watershed scale. An overall priority index allowed identification of watersheds experiencing high pressure from non-native species and also exhibiting high native biodiversity conservation value. While priority regions are roughly consistent with previously reported prioritization maps for the US, we also recognize novel priority areas characterized by moderate-to-high native diversity but extremely high invasion pressure. We further compared priority areas with existing conservation protections as well as projected future threats associated with land use change. Our findings suggest that many regions of elevated freshwater biodiversity value are compromised by high invasion pressure, and are poorly safeguarded by existing conservation mechanisms and are likely to experience significant additional stresses in the future.

Inferring invasive species abundance using removal data from management actions.

Evaluation of the progress of management programs for invasive species is crucial for demonstrating impacts to stakeholders and strategic planning of resource allocation. Estimates of abundance before and after management activities can serve as a useful metric of population management programs. However, many methods of estimating population size are too labor intensive and costly to implement, posing restrictive levels of burden on operational programs. Removal models are a reliable method for estimating abundance before and after management using data from the removal activities exclusively, thus requiring no work in addition to management. We developed a Bayesian hierarchical model to estimate abundance from removal data accounting for varying levels of effort, and used simulations to assess the conditions under which reliable population estimates are obtained. We applied this model to estimate site-specific abundance of an invasive species, feral swine (Sus scrofa), using removal data from aerial gunning in 59 site/time-frame combinations (480-19,600 acres) throughout Oklahoma and Texas, USA. Simulations showed that abundance estimates were generally accurate when effective removal rates (removal rate accounting for total effort) were above 0.40. However, when abundances were small (<50) the effective removal rate needed to accurately estimates abundances was considerably higher (0.70). Based on our post-validation method, 78% of our site/time frame estimates were accurate. To use this modeling framework it is important to have multiple removals (more than three) within a time frame during which demographic changes are minimized (i.e., a closed population; ≤3 months for feral swine). Our results show that the probability of accurately estimating abundance from this model improves with increased sampling effort (8+ flight hours across the 3-month window is best) and increased removal rate. Based on the inverse relationship between inaccurate abundances and inaccurate removal rates, we suggest auxiliary information that could be collected and included in the model as covariates (e.g., habitat effects, differences between pilots) to improve accuracy of removal rates and hence abundance estimates.

Oral Rabies Vaccination of Raccoons (Procyon lotor) across a Development Intensity Gradient in Burlington, Vermont, USA, 2015-2017.

Management of the raccoon rabies virus variant in North America is conducted primarily using oral rabies vaccination (ORV). When a sufficient proportion of the population is vaccinated (∼60%), rabies transmission can be eliminated. To date, ORV programs have successfully controlled and eliminated raccoon rabies in rural areas, but there has been less success in urban areas. We studied the proportions of rabies virus neutralizing antibodies (RVNA) in a raccoon (Procyon lotor) population during a 3-yr ORV trial in developed areas of Burlington, Vermont, US. We used a modified N-mixture model to estimate raccoon abundance, RVNA seroprevalence, and capture rates jointly to examine factors that relate to ORV success to better inform management. We found that raccoon abundance was lower in less-developed areas compared to urban centers. Raccoon RVNA seroprevalence decreased as population abundance increased; it increased as the average age of the population increased. Nontarget opossum (Didelphis virginiana) captures correlated with a decrease in raccoon RVNA seroprevalence in low-development areas, suggesting that they may be competing for baits. The target bait density across the entire study area was 150 baits/km2, but a hand baiting strategy was heavily concentrated on roads, resulting in uneven bait densities within sampling sites (0-484 baits/km2). Uneven bait distribution across the study area may explain low RVNA seroprevalence in some locations. Our results suggest that increases in bait density across the study area may improve RVNA seroprevalence and support annual ORV to account for raccoon population turnover.

Evaluation of contingency actions to control the spread of raccoon rabies in Ohio and Virginia.

The raccoon (Procyon lotor) variant of the rabies virus (RRV) is enzootic in the eastern United States and oral rabies vaccination (ORV) is the primary strategy to prevent and control landscape spread. Breaches of ORV management zones occasionally occur, and emergency "contingency" actions may be implemented to enhance local control. Contingency actions are an integral part of landscape-scale wildlife rabies management but can be very costly and routinely involve enhanced rabies surveillance (ERS) around the index case. We investigated two contingency actions in Ohio (2017-2019 and 2018-2021) and one in Virginia (2017-2019) using a dynamic, multi-method occupancy approach to examine relationships between specific management actions and RRV occurrence, including whether ERS was sufficient around the index case. The RRV occupancy was assessed seasonally at 100-km2 grids and we examined relationships across three spatial scales (regional management zone, RRV free regions, and local contingency areas). The location of a grid relative to the ORV management zone was the strongest predictor of RRV occupancy at the regional scale. In RRV free regions, the neighbor effect and temporal variability were most important in influencing RRV occupancy. Parenteral (hand) vaccination of raccoons was important across all three contingency action areas, but more influential in the Ohio contingency action areas where more raccoons were hand vaccinated. In the Virginia contingency action area, ORV strategies were as important in reducing RRV occupancy as a hand vaccination strategy. The management action to trap, euthanize, and test (TET) raccoons was an important method to increase ERS, yet the impacts of TET on RRV occupancy are not clear. The probability of detecting additional cases of RRV was exceptionally high (>0.95) during the season the index case occurred. The probability of detecting RRV through ERS declined in the seasons following initial TET efforts but remained higher after the contingency action compared to the ERS detection probabilities prior to index case incidence. Local RRV cases were contained within one year and eliminated within 2-3 years of each contingency action.

Mechanistic models project bird invasions with accuracy.

Invasive species pose a major threat to biodiversity and inflict massive economic costs. Effective management of bio-invasions depends on reliable predictions of areas at risk of invasion, as they allow early invader detection and rapid responses. Yet, considerable uncertainty remains as to how to predict best potential invasive distribution ranges. Using a set of mainly (sub)tropical birds introduced to Europe, we show that the true extent of the geographical area at risk of invasion can accurately be determined by using ecophysiological mechanistic models that quantify species' fundamental thermal niches. Potential invasive ranges are primarily constrained by functional traits related to body allometry and body temperature, metabolic rates, and feather insulation. Given their capacity to identify tolerable climates outside of contemporary realized species niches, mechanistic predictions are well suited for informing effective policy and management aimed at preventing the escalating impacts of invasive species.

RABIES VIRUS SEROSURVEY OF THE SMALL INDIAN MONGOOSE (URVA AUROPUNCTATA) ACROSS MULTIPLE HABITATS IN PUERTO RICO, 2014-21.

The small Indian mongoose (Urva auropuncata) is a rabies reservoir in Puerto Rico and accounts for over 70% of reported animal rabies cases annually. The presence of rabies virus-neutralizing antibodies (RVNA) is often used as a tool to measure exposure to rabies virus in wildlife populations. We conducted a serosurvey of mongooses at 11 sites representing six habitat types across Puerto Rico. We collected a serum sample from 464 individual mongooses during 2014-21. Overall, 80/464 (17.0%; 95% confidence interval, 14.1-20.9%; 55 male, 23 female, and two sexes not recorded) of individual mongooses sampled across all habitats were RVNA positive. The geometric mean (SD) RVNA titer for 80 unique seropositive animals was 0.58 (2.92) IU/mL. Our models indicated that the probability of mongooses being RVNA seropositive mostly varied by habitat, with some influence of sex in the individual-level analyses. Population-level RVNA seroprevalence is dynamic in mongoose populations, but these data may shed light on rabies virus transmission across regions to help inform rabies management activities in Puerto Rico.

Defining County-Level Terrestrial Rabies Freedom Using the US National Rabies Surveillance System: Surveillance Data Analysis.

BACKGROUND: Rabies is a deadly zoonotic disease with nearly 100% fatality rate. In the United States, rabies virus persists in wildlife reservoirs, with occasional spillover into humans and domestic animals. The distribution of reservoir hosts in US counties plays an important role in public health decision-making, including the recommendation of lifesaving postexposure prophylaxis upon suspected rabies exposures. Furthermore, in surveillance data, it is difficult to discern whether counties have no cases reported because rabies was not present or because counties have an unreported rabies presence. These epizootics are monitored by the National Rabies Surveillance System (NRSS), to which approximately 130 state public health, agriculture, and academic laboratories report animal rabies testing statistics. Historically, the NRSS classifies US counties as free from terrestrial rabies if, over the previous 5 years, they and any adjacent counties did not report any rabies cases and they tested ≥15 reservoir animals or 30 domestic animals. OBJECTIVE: This study aimed to describe and evaluate the historical NRSS rabies-free county definition, review possibilities for improving this definition, and develop a model to achieve more precise estimates of the probability of terrestrial rabies freedom and the number of reported county-level terrestrial rabies cases. METHODS: Data submitted to the NRSS by state and territorial public health departments and the US Department of Agriculture Wildlife Services were analyzed to evaluate the historical rabies-free definition. A zero-inflated negative binomial model created county-level predictions of the probability of rabies freedom and the expected number of rabies cases reported. Data analyzed were from all animals submitted for laboratory diagnosis of rabies in the United States from 1995 to 2020 in skunk and raccoon reservoir territories, excluding bats and bat variants. RESULTS: We analyzed data from 14,642 and 30,120 county-years in the raccoon and skunk reservoir territories, respectively. Only 0.85% (9/1065) raccoon county-years and 0.79% (27/3411) skunk county-years that met the historical rabies-free criteria reported a case in the following year (99.2% negative predictive value for each), of which 2 were attributed to unreported bat variants. County-level model predictions displayed excellent discrimination for detecting zero cases and good estimates of reported cases in the following year. Counties classified as rabies free rarely (36/4476, 0.8%) detected cases in the following year. CONCLUSIONS: This study concludes that the historical rabies freedom definition is a reasonable approach for identifying counties that are truly free from terrestrial raccoon and skunk rabies virus transmission. Gradations of risk can be measured using the rabies prediction model presented in this study. However, even counties with a high probability of rabies freedom should maintain rabies testing capacity, as there are numerous examples of translocations of rabies-infected animals that can cause major changes in the epidemiology of rabies.

Data-Driven Management-A Dynamic Occupancy Approach to Enhanced Rabies Surveillance Prioritization.

Rabies lyssavirus (RABV) is enzootic in raccoons across the eastern United States. Intensive management of RABV by oral rabies vaccination (ORV) has prevented its spread westward and shown evidence of local elimination in raccoon populations of the northeastern US. The USDA, Wildlife Services, National Rabies Management Program (NRMP) collaborates with other agencies to implement broad-scale ORV and conducts extensive monitoring to measure the effectiveness of the management. Enhanced Rabies Surveillance (ERS) was initiated during 2005 and updated in 2016 to direct surveillance efforts toward higher-value specimens by assigning points to different methods of encountering specimens for collection (strange-acting, roadkill, surveillance-trapped, etc.; specimen point values ranged from 1 to 15). We used the 2016-2019 data to re-evaluate the point values using a dynamic occupancy model. Additionally, we used ERS data from 2012-2015 and 2016-2019 to examine the impact that the point system had on surveillance data. Implementation of a point system increased positivity rates among specimens by 64%, indicating a substantial increase in the efficiency of the ERS to detect wildlife rabies. Our re-evaluation found that most points accurately reflect the value of the surveillance specimens. The notable exception was that samples from animals found dead were considerably more valuable for rabies detection than originally considered (original points = 5, new points = 20). This work demonstrates how specimen prioritization strategies can be used to refine and improve ERS in support of wildlife rabies management.

Serological Responses of Raccoons and Striped Skunks to Ontario Rabies Vaccine Bait in West Virginia during 2012-2016.

Since the 1990s, oral rabies vaccination (ORV) has been used successfully to halt the westward spread of the raccoon rabies virus (RV) variant from the eastern continental USA. Elimination of raccoon RV from the eastern USA has proven challenging across targeted raccoon (Procyon lotor) and striped skunk (Mephitis mephitis) populations impacted by raccoon RV. Field trial evaluations of the Ontario Rabies Vaccine Bait (ONRAB) were initiated to expand ORV products available to meet the rabies management goal of raccoon RV elimination. This study describes the continuation of a 2011 trial in West Virginia. Our objective was to evaluate raccoon and skunk response to ORV occurring in West Virginia for an additional two years (2012-2013) at 75 baits/km2 followed by three years (2014-2016) of evaluation at 300 baits/km2. We measured the change in rabies virus-neutralizing antibody (RVNA) seroprevalence in targeted wildlife populations by comparing levels pre- and post-ORV during each year of study. The increase in bait density from 75/km2 to 300/km2 corresponded to an increase in average post-ORV seroprevalence for raccoon and skunk populations. Raccoon population RVNA levels increased from 53% (300/565, 95% CI: 50-57%) to 82.0% (596/727, 95% CI: 79-85%) during this study, and skunk population RVNA levels increased from 11% (8/72, 95% CI: 6-20%) to 39% (51/130, 95% CI: 31-48%). The RVNA seroprevalence pre-ORV demonstrated an increasing trend across study years for both bait densities and species, indicating that multiple years of ORV may be necessary to achieve and maintain RVNA seroprevalence in target wildlife populations for the control and elimination of raccoon RV in the eastern USA.

A checklist recipe: making species data open and FAIR.

Species checklists are a crucial source of information for research and policy. Unfortunately, many traditional species checklists vary wildly in their content, format, availability and maintenance. The fact that these are not open, findable, accessible, interoperable and reusable (FAIR) severely hampers fast and efficient information flow to policy and decision-making that are required to tackle the current biodiversity crisis. Here, we propose a reproducible, semi-automated workflow to transform traditional checklist data into a FAIR and open species registry. We showcase our workflow by applying it to the publication of the Manual of Alien Plants, a species checklist specifically developed for the Tracking Invasive Alien Species (TrIAS) project. Our approach combines source data management, reproducible data transformation to Darwin Core using R, version control, data documentation and publication to the Global Biodiversity Information Facility (GBIF). This checklist publication workflow is openly available for data holders and applicable to species registries varying in thematic, taxonomic or geographical scope and could serve as an important tool to open up research and strengthen environmental decision-making.

A comparison of cost and quality of three methods for estimating density for wild pig (Sus scrofa).

A critical element in effective wildlife management is monitoring the status of wildlife populations; however, resources to monitor wildlife populations are typically limited. We compared cost effectiveness of three common population estimation methods (i.e. non-invasive DNA sampling, camera sampling, and sampling from trapping) by applying them to wild pigs (Sus scrofa) across three habitats in South Carolina, U.S.A where they are invasive. We used mark-recapture analyses for fecal DNA sampling data, spatially-explicit capture-recapture analyses for camera sampling data, and a removal analysis for removal sampling from trap data. Density estimates were similar across methods. Camera sampling was the least expensive, but had large variances. Fecal DNA sampling was the most expensive, although this technique generally performed well. We examined how reductions in effort by method related to increases in relative bias or imprecision. For removal sampling, the largest cost savings while maintaining unbiased density estimates was from reducing the number of traps. For fecal DNA sampling, a reduction in effort only minimally reduced costs due to the need for increased lab replicates while maintaining high quality estimates. For camera sampling, effort could only be marginally reduced before inducing bias. We provide a decision tree for researchers to help make monitoring decisions.

Placebo Oral Rabies Vaccine Bait Uptake by Small Indian Mongooses (Herpestes auropunctatus) in Southwestern Puerto Rico.

The small Indian mongoose (Herpestes auropunctatus) is a rabies reservoir in areas of the Caribbean including Puerto Rico, but no rabies vaccination program targeting this host exists. We used two derivatives of iophenoxic acid (IPA) to evaluate placebo oral rabies vaccine bait uptake by mongooses in southwestern Puerto Rico. We hand-distributed baits at an application rate of 200 baits/km2 at three, 400 ha, sites during autumn 2016 and spring 2017. Each site contained 90-100 cage traps in a 100 ha central trapping area. We used ethyl-IPA as a biological marker during the autumn and methyl-IPA during the spring. We live captured mongooses for 10 consecutive days, beginning 1 wk following bait application. We obtained a serum sample from captured mongooses and analyzed the sera for ethyl- and methyl-IPA by liquid chromatography-mass spectrometry. During autumn 2016, 63% (55/87) mongooses sampled were positive for ethyl-IPA. In spring 2017, 69% (85/123) of mongooses were positive for methyl-IPA. Pooling seasons, accounting for recaptures between years, and disregarding marker type, 74% (133/179) unique mongooses were positive for IPA biomarker, indicating bait consumption during either the autumn, spring, or both trials. We conclude that distributing baits at an application rate of 200 baits/km2 is sufficient to reach over 60% of the target mongoose population in dry forest habitats of Puerto Rico.

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.