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

Accurate identification of animal species is necessary to understand biodiversity richness, monitor endangered species, and study the impact of climate change on species distribution within a specific region. Camera traps represent a passive monitoring technique that generates millions of ecological images. The vast numbers of images drive automated ecological analysis as essential, given that manual assessment of large datasets is laborious, time-consuming, and expensive. Deep learning networks have been advanced in the last few years to solve object and species identification tasks in the computer vision domain, providing state-of-the-art results. In our work, we trained and tested machine learning models to classify three animal groups (snakes, lizards, and toads) from camera trap images. We experimented with two pretrained models, VGG16 and ResNet50, and a self-trained convolutional neural network (CNN-1) with varying CNN layers and augmentation parameters. For multiclassification, CNN-1 achieved 72% accuracy, whereas VGG16 reached 87%, and ResNet50 attained 86% accuracy. These results demonstrate that the transfer learning approach outperforms the self-trained model performance. The models showed promising results in identifying species, especially those with challenging body sizes and vegetation.

Biogeographic consequences of shifting climate for the western massasauga ( Sistrurus tergeminus )

The western massasauga (Sistrurus tergeminus) is a small pit viper with an extensive geographic range, yet observations of this species are relatively rare. They persist in patchy and isolated populations, threatened by habitat destruction and fragmentation, mortality from vehicle collisions, and deliberate extermination. Changing climates may pose an additional stressor on the survival of isolated populations. Here, we evaluate historic, modern, and future geographic projections of suitable climate for S. tergeminus to outline shifts in their potential geographic distribution and inform current and future management. We used maximum entropy modeling to build multiple models of the potential geographic distribution of S. tergeminus. We evaluated the influence of five key decisions made during the modeling process on the resulting geographic projections of the potential distribution, allowing us to identify areas of model robustness and uncertainty. We evaluated models with the area under the receiver operating curve and true skill statistic. We retained 16 models to project both in the past and future multiple general circulation models. At the last glacial maximum, the potential geographic distribution associated with S. tergeminus occurrences had a stronghold in the southern part of its current range and extended further south into Mexico, but by the mid‐Holocene, its modeled potential distribution was similar to its present‐day potential distribution. Under future model projections, the potential distribution of S. tergeminus moves north, with the strongest northward trends predicted under a climate scenario increase of 8.5 W/m². Some southern populations of S. tergeminus have likely already been extirpated and will continue to be threatened by shifting availability of suitable climate, as they are already under threat from desertification of grasslands. Land use and habitat loss at the northern edge of the species range are likely to make it challenging for this species to track suitable climates northward over time.

Phylogeographic structure of the dunes sagebrush lizard, an endemic habitat specialist

Phylogeographic divergence and population genetic diversity within species reflect the impacts of habitat connectivity, demographics, and landscape level processes in both the recent and distant past. Characterizing patterns of differentiation across the geographic range of a species provides insight on the roles of organismal and environmental traits in evolutionary divergence and future population persistence. This is particularly true of habitat specialists where habitat availability and resource dependence may result in pronounced genetic structure as well as increased population vulnerability. We use DNA sequence data as well as microsatellite genotypes to estimate range-wide phylogeographic divergence, historical population connectivity, and historical demographics in an endemic habitat specialist, the dunes sagebrush lizard (Sceloporus arenicolus). This species is found exclusively in dune blowouts and patches of open sand within the shinnery oak-sand dune ecosystem of southeastern New Mexico and adjacent Texas. We find evidence of phylogeographic structure consistent with breaks and constrictions in suitable habitat at the range-wide scale. In addition, we find support for a dynamic and variable evolutionary history across the range of S. arenicolus. Populations in the Monahans Sandhills have deeply divergent lineages consistent with long-term demographic stability. In contrast, populations in the Mescalero Sands are not highly differentiated, though we do find evidence of demographic expansion in some regions and relative demographic stability in others. Phylogeographic history and population genetic differentiation in this species has been shaped by the configuration of habitat patches within a geologically complex and historically dynamic landscape. Our findings identify regions as genetically distinctive conservation units as well as underscore the genetic and demographic history of different lineages of S. arenicolus.

Searching for rare and secretive snakes: are camera-trap and box-trap methods interchangeable?

Abstract ContextAdvancements in camera-trap technology have provided wildlife researchers with a new technique to better understand their study species. This improved method may be especially useful for many conservation-reliant snake species that can be difficult to detect because of rarity and life histories with secretive behaviours. AimsHere, we report the results of a 6-month camera-trapping study using time lapse-triggered camera traps to detect snakes, in particular the federally listed Louisiana pinesnake (Pituophis ruthveni) in eastern Texas upland forests in the USA. MethodsSo as to evaluate the efficacy of this method of snake detection, we compared camera-trap data with traditional box-trapping data collected over the same time period across a similar habitat type, and with the same goal of detecting P. ruthveni. Key resultsNo differences in focal snake species richness were detected across the trap methods, although the snake-detection rate was nearly three times higher with camera traps than with the box traps. Detection rates of individual snake species varied with the trapping method for all but two species, but temporal trends in detection rates were similar across the trap methods for all but two species. Neither trap method detected P. ruthveni in the present study, but the species has been detected with both trap methods at other sites. ConclusionsThe higher snake-detection rate of the camera-trap method suggests that pairing this method with traditional box traps could increase the detection of P. ruthveni where it occurs. For future monitoring and research on P. ruthveni, and other similarly rare and secretive species of conservation concern, we believe these methods could be used interchangeably by saturating potentially occupied habitats with camera traps initially and then replacing cameras with box traps when the target species is detected. ImplicationsThere are financial and logistical limits to monitoring and researching rare and secretive species with box traps, and those limits are far less restrictive with camera traps. The ability to use camera-trap technologies interchangeably with box-trap methods to collect similar data more efficiently and effectively will have a significant impact on snake conservation.

Phylogenetic structure of Holbrookia lacerata (Cope 1880) (Squamata: Phrynosomatidae): one species or two?

Species delimitation attempts to match species-level taxonomy with actual evolutionary lineages. Such taxonomic conclusions are typically, but not always, based on patterns of congruence across multiple data sources and methods of analyses. Here, we use this pluralistic approach to species delimitation to help resolve uncertainty in species boundaries of phrynosomatid sand lizards of the genus Holbrookia. Specifically, the Spot-tailed Earless Lizard (H. lacerata) was historically divided into a northern (H. l. lacerata) and southern (H. l. subcaudalis) subspecies based on differences in morphology and allopatry, but no research has been conducted evaluating genetic differences between these taxa. In this study, patterns in sequence data derived from two genes, one nuclear and one mitochondrial, for 66 individuals sampled across 18 counties in Texas revealed three strongly supported, reciprocally monophyletic lineages, each comprised of individuals from a single geographic region. Distinct genetic variation evident across two of these regions corresponds with differences in morphology, differences in environmental niche, and lines up with the presumed geographic barrier, the Balcones Escarpment, which is the historical subspecies boundary. The combined evidence from genetics, morphology and environmental niche is sufficient to consider these subspecies as distinct species with the lizards north of the Balcones Escarpment retaining the name Holbrookia lacerata, and those south of the Balcones Escarpment being designated as Holbrookia subcaudalis.

An Improved Approach for Forecasting Ecological Impacts from Future Drilling in Unconventional Shale Oil and Gas Plays

Directional well drilling and hydraulic fracturing has enabled energy production from previously inaccessible resources, but caused vegetation conversion and landscape fragmentation, often in relatively undisturbed habitats. We improve forecasts of future ecological impacts from unconventional oil and gas play developments using a new, more spatially-explicit approach. We applied an energy production outlook model, which used geologic and economic data from thousands of wells and three oil price scenarios, to map future drilling patterns and evaluate the spatial distribution of vegetation conversion and habitat impacts. We forecast where future well pad construction may be most intense, illustrating with an example from the Eagle Ford Shale Play of Texas. We also illustrate the ecological utility of this approach using the Spot-tailed Earless Lizard (Holbrookia lacerata) as the focal species, which historically occupied much of the Eagle Ford and awaits a federal decision for possible Endangered Species Act protection. We found that ~17,000-45,500 wells would be drilled 2017‒2045 resulting in vegetation conversion of ~26,485-70,623 ha (0.73-1.96% of pre-development vegetation), depending on price scenario ($40-$80/barrel). Grasslands and row crop habitats were most affected (2.30 and 2.82% areal vegetation reduction). Our approach improves forecasts of where and to what extent future energy development in unconventional plays may change land-use and ecosystem services, enabling natural resource managers to anticipate and direct on-the-ground conservation actions to places where they will most effectively mitigate ecological impacts of well pads and associated infrastructure.

Fragmentation alters home range and movements of the Dunes Sagebrush Lizard (Sceloporus arenicolus)

Habitat fragmentation is a major driver of biodiversity loss and among reptiles has been attributed as a cause of species decline. The negative effect of habitat fragmentation has also been shown to be worse for species that are habitat specialists. The Dunes Sagebrush Lizard (Sceloporus arenicolus Degenhardt and Jones, 1972) is a species that specializes on the shinnery oak (Quercus havardii Rydb.) sand-dune landform of the Mescalero–Monahans Sandhills ecosystem in western Texas and eastern New Mexico, USA. This landform has been fragmented by roads and well pads used for the extraction of oil and gas resources. The effects of fragmentation on the home range and movements of this species can lead to the effective isolation of populations and increased risk of localized extirpations. We showed that home-range size was larger in an unfragmented area and that the mean distance of movements was greater. We also observed that roads in the fragmented areas restricted movements of S. arenicolus. We concluded that roads can be barriers to movements even though only narrow strips of habitat are altered.

Comparison of Recent Oil and Gas, Wind Energy, and Other Anthropogenic Landscape Alteration Factors in Texas Through 2014

Recent research assessed how hydrocarbon and wind energy expansion has altered the North American landscape. Less understood, however, is how this energy development compares to other anthropogenic land use changes. Texas leads U.S. hydrocarbon production and wind power generation and has a rapidly expanding population. Thus, for ~47% of Texas (~324,000 km²), we mapped the 2014 footprint of energy activities (~665,000 oil and gas wells, ~5700 wind turbines, ~237,000 km oil and gas pipelines, and ~2000 km electrical transmission lines). We compared the footprint of energy development to non-energy-related activities (agriculture, roads, urbanization) and found direct landscape alteration from all factors affects ~23% of the study area (~76,000 km²), led by agriculture (~16%; ~52,882 km²). Oil and gas activities altered <1% of the study area (2081 km²), with 838 km² from pipelines and 1242 km² from well pad construction-and that the median Eagle Ford well pad is 7.7 times larger than that in the Permian Basin (16,200 vs. 2100 m²). Wind energy occupied <0.01% (~24 km²), with ~14 km² from turbine pads and ~10 km² from power transmission lines. We found that edge effects of widely-distributed energy infrastructure caused more indirect landscape alteration than larger, more concentrated urbanization and agriculture. This study presents a novel technique to quantify and compare anthropogenic activities causing both direct and indirect landscape alteration. We illustrate this landscape-mapping framework in Texas for the Spot-tailed Earless Lizard (Holbrookia lacerata); however, the approach can be applied to a range of species in developing regions globally.

Why didn't the lizard cross the road? Dunes sagebrush lizards exhibit road-avoidance behaviour

Context Research has shown many negative effects of roads and traffic on wildlife and other biodiversity. The direct and indirect mechanisms through which roads and traffic harm animal populations vary across taxa, making mitigation of road effects a great challenge for conservation. As such, a large toolkit of species-specific management techniques may be needed to mitigate the negative effects of roads for wildlife and other biodiversity. The dunes sagebrush lizard, Sceloporus arenicolus, is a psammophilic (sand-loving) habitat specialist endemic to the Mescalero–Monahans Sandhills ecosystem of New Mexico and Texas. Within this ecosystem, roads fragment shinnery oak sand-dune landforms occupied by the species. Aims In the present study, we conducted behaviour trials in experimental enclosures to test whether the smallest roads restrict movements of the dunes sagebrush lizard. In addition, we also conducted trials to evaluate whether a sand-filled wildlife-crossing feature could facilitate road crossing. Methods We conducted behavioural trials on 24 dunes sagebrush lizards in our control enclosure and 22 lizards in our road and sand-filled wildlife-crossing enclosure. Movements were recorded for 15min. The final locations at the end of each trial were analysed using circular statistics to determine whether movements in the road or the sand-filled wildlife-crossing enclosures were different from the control. Key results Our results supported the hypotheses that dunes sagebrush lizards avoid roads and do so according to a surface-avoidance mechanism. We also found that the wildlife crossing-feature design tested here had no effect on the movements or road-crossing frequency of dunes sagebrush lizard. Conclusions Surface-avoidance behaviour indicated that roads will persistently affect the movements of dunes sagebrush lizard, even when traffic is not present. Also, more research into an effective wildlife crossing is needed to increase connectivity of fragmented populations. Implications These findings help evaluate the impact of roads in creating isolated populations that experience increased demographic stochasticity and, in some instances, localised extirpation in this species. Our study can guide conservation plans for the dunes sagebrush lizard, and contribute to our understanding of road effects on biodiversity in general.

Linking irreplaceable landforms in a self-organizing landscape to sensitivity of population vital rates for an ecological specialist

Irreplaceable, self-organizing landforms and the endemic and ecologically specialized biodiversity they support are threatened globally by anthropogenic disturbances. Although the outcome of disrupting landforms is somewhat understood, little information exists that documents population consequences of landform disturbance on endemic biodiversity. Conservation strategies for species dependent upon landforms have been difficult to devise because they require understanding complex feedbacks that create and maintain landforms and the consequences of landform configuration on demography of species. We characterized and quantified links between landform configuration and demography of an ecological specialist, the dunes sagebrush lizard (Sceloporus arenicolus), which occurs only in blowouts (i.e., wind-blown sandy depressions) of Shinnery oak (Quercus havardii) sand-dune landforms. We used matrix models to estimate vital rates from a multisite mark-recapture study of 6 populations occupying landforms with different spatial configurations. Sensitivity and elasticity analyses demonstrated demographic rates among populations varied in sensitivity to different landform configurations. Specifically, significant relationships between blowout shape complexity and vital rate elasticities suggested direct links between S. arenicolus demography and amount of edge in Shinnery oak sand-dune landforms. These landforms are irreplaceable, based on permanent transition of disturbed areas to alternative grassland ecosystem states. Additionally, complex feedbacks between wind, sand, and Shinnery oak maintain this landform, indicating restoration through land management practices is unlikely. Our findings that S. arenicolus population dynamics depended on landform configuration suggest that failure to consider processes of landform organization and their effects on species' population dynamics may lead to incorrect inferences about threats to endemic species and ineffective habitat management for threatened or endangered species. As such, successful conservation of these systems and the biodiversity they support must be informed by research linking process-oriented studies of self-organized landforms with studies of movement, behavior, and demography of species that dwell in them.

Ecological correlates of risk and incidence of West Nile virus in the United States

West Nile virus, which was recently introduced to North America, is a mosquito-borne pathogen that infects a wide range of vertebrate hosts, including humans. Several species of birds appear to be the primary reservoir hosts, whereas other bird species, as well as other vertebrate species, can be infected but are less competent reservoirs. One hypothesis regarding the transmission dynamics of West Nile virus suggests that high bird diversity reduces West Nile virus transmission because mosquito blood-meals are distributed across a wide range of bird species, many of which have low reservoir competence. One mechanism by which this hypothesis can operate is that high-diversity bird communities might have lower community-competence, defined as the sum of the product of each species' abundance and its reservoir competence index value. Additional hypotheses posit that West Nile virus transmission will be reduced when either: (1) abundance of mosquito vectors is low; or (2) human population density is low. We assessed these hypotheses at two spatial scales: a regional scale near Saint Louis, MO, and a national scale (continental USA). We found that prevalence of West Nile virus infection in mosquito vectors and in humans increased with decreasing bird diversity and with increasing reservoir competence of the bird community. Our results suggest that conservation of avian diversity might help ameliorate the current West Nile virus epidemic in the USA.

Predator-dependent species-area relationships

In addition to having a positive effect on species richness (species-area relationships [SARs]), habitat area can influence the presence of predators, which can indirectly influence prey richness. While these direct and indirect effects of area on richness occur simultaneously, no research has examined how predation might contribute to SAR variation. We extend MacArthur and Wilson's equilibrium theory of island biogeography by including predation-induced shifts in prey extinction and predict that predators will reduce slopes of prey SARs. We provide support for this with data from two insular ecosystems: orthopteran richness in Ozark glades (rocky herbaceous communities within a forested matrix) with and without insectivorous lizards and zooplankton richness in freshwater ponds with and without zooplanktivorous fishes. Our results emphasize that anthropogenic activities yield simultaneous changes in processes altering diversity and that it is critical that we understand how these components of anthropogenic change interact to impact diversity.

Genetic relationships of American alligator populations distributed across different ecological and geographic scales

Although much work has been conducted on coastal populations of the American alligator (Alligator mississippiensis), less is known about the population dynamics and genetic structure of populations of alligators confined to inland habitats. DNA microsatellite loci, derived from the American alligator, were used to investigate patterns of genetic variation within and between populations of alligators distributed at coastal and inland localities in Texas. These data were used to evaluate the genetic discreteness of different alligator stocks relative to their basic ecology at these sites. Observed mean heterozygosities across seven loci for both coastal and inland populations ranged from 0.50-0.61, with both inland and coastal populations revealing similar patterns of variation. Measures of F(st) revealed significant population differentiation among all populations; however, analyses of molecular variance (AMOVAs) failed to demonstrate any apparent geographic pattern relative to the population differentiation indicated by F(st) values. Each population contained unique alleles for at least one locus. Additionally, assignment tests based on the distribution of genotypes placed 76% of individuals to their source population. These genetic data suggest considerable subdivision among alligator populations, possibly influenced by demographic and life history differences as well as barriers to dispersal. These results have clear implications for management. Rather than managing alligators in Texas as a single panmictic population, translocation programs and harvest quotas should consider the ecological and genetic distinctiveness of local alligator populations.