ABSTRACT
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.
ABSTRACT
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.
