Elizabeth G. Mandeville

Population connectivity and genetic structure of burbot (Lota lota) populations in the Wind River Basin, Wyoming

Burbot (Lota lota) occur in the Wind River Basin in central Wyoming, USA, at the southwestern extreme of the species’ native range in North America. The most stable and successful of these populations occur in six glacially carved mountain lakes on three different tributary streams and one large main stem impoundment (Boysen Reservoir) downstream from the tributary populations. Burbot are rarely found in connecting streams and rivers, which are relatively small and high gradient, with a variety of potential barriers to upstream movement of fish. We used high-throughput genomic sequence data for 11,197 SNPs to characterize the genetic diversity, population structure, and connectivity among burbot populations on the Wind River system. Fish from Boysen Reservoir and lower basin tributary populations were genetically differentiated from those in the upper basin tributary populations. In addition, fish within the same tributary streams fell within the same genetic clusters, suggesting there is movement of fish between lakes on the same tributaries but that populations within each tributary system are isolated and genetically distinct from other populations. Observed genetic differentiation corresponded to natural and anthropogenic barriers, highlighting the importance of barriers to fish population connectivity and gene flow in human-altered linked lake-stream habitats.

Inconsistent reproductive isolation revealed by interactions between Catostomus fish species

Interactions between species are central to evolution and ecology, but we do not know enough about how outcomes of interactions between species vary across geographic locations, in heterogeneous environments, or over time. Ecological dimensions of interactions between species are known to vary, but evolutionary interactions such as the establishment and maintenance of reproductive isolation are often assumed to be consistent across instances of an interaction between species. Hybridization among Catostomus fish species occurs over a large and heterogeneous geographic area and across taxa with distinct evolutionary histories, which allows us to assess consistency in species interactions. We analyzed hybridization among six Catostomus species across the Upper Colorado River basin (US mountain west) and found extreme variation in hybridization across locations. Different hybrid crosses were present in different locations, despite similar species assemblages. Within hybrid crosses, hybridization varied from only first generation hybrids to extensive hybridization with backcrossing. Variation in hybridization outcomes might result from uneven fitness of hybrids across locations, polymorphism in genetic incompatibilities, chance, unidentified historical contingencies, or some combination thereof. Our results suggest caution in assuming that one or a few instances of hybridization represent all interactions between the focal species, as species interactions vary substantially across locations.

Machine learning to classify animal species in camera trap images: applications in ecology

Motion-activated cameras (“camera traps”) are increasingly used in ecological and management studies for remotely observing wildlife and have been regarded as among the most powerful tools for wildlife research. However, studies involving camera traps result in millions of images that need to be analyzed, typically by visually observing each image, in order to extract data that can be used in ecological analyses. We trained machine learning models using convolutional neural networks with the ResNet-18 architecture and 3,367,383 images to automatically classify wildlife species from camera trap images obtained from five states across the United States. We tested our model on an independent subset of images not seen during training from the United States and on an out-of-sample (or “out-of-distribution” in the machine learning literature) dataset of ungulate images from Canada. We also tested the ability of our model to distinguish empty images from those with animals in another out-of-sample dataset from Tanzania, containing a faunal community that was novel to the model. The trained model classified approximately 2,000 images per minute on a laptop computer with 16 gigabytes of RAM. The trained model achieved 98% accuracy at identifying species in the United States, the highest accuracy of such a model to date. Out-of-sample validation from Canada achieved 82% accuracy, and correctly identified 94% of images containing an animal in the dataset from Tanzania. We provide an R package (Machine Learning for Wildlife Image Classification; MLWIC) that allows the users to A) implement the trained model presented here and B) train their own model using classified images of wildlife from their studies. The use of machine learning to rapidly and accurately classify wildlife in camera trap images can facilitate non-invasive sampling designs in ecological studies by reducing the burden of manually analyzing images. We present an R package making these methods accessible to ecologists. We discuss the implications of this technology for ecology and considerations that should be addressed in future implementations of these methods.

artificial reference genome

This reference was made by completing a denovo assembly of a subset of the data using smng (see manuscript text) and all reads were assembled to this reference for variant calling.

Input file for entropy analysis

Genotype likelihoods (simplified from VCF) at 11,221 SNPs for 2785 individual Catostomus fish. This file was used for analyses of hybridization.

Full VCF file

All variant sites initially identified for analysis of hybridization. This VCF was produced using samtools and bcftools, and contains information for 2785 Catostomus individuals.