Motion‐activated cameras (“camera traps”) are increasingly used in ecological and management studies for remotely observing wildlife and are amongst the most powerful tools for wildlife research. However, studies involving camera traps result in millions of images that need to be analysed, 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) that allows the users to (a) use 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 analysing images. Our r package makes these methods accessible to ecologists.

1. Advancements in wildlife telemetry techniques have made it possible to collect large data sets of highly accurate animal locations at a fine temporal resolution. These data sets have prompted the development of a number of statistical methodologies for modelling animal movement. 2. Telemetry data sets are often collected for purposes other than fine-scale movement analysis. These data sets may differ substantially from those that are collected with technologies suitable for fine-scale movement modelling and may consist of locations that are irregular in time, are temporally coarse or have large measurement error. These data sets are time-consuming and costly to collect but may still provide valuable information about movement behaviour. 3. We developed a Bayesian movement model that accounts for error from multiple data sources as well as movement behaviour at different temporal scales. The Bayesian framework allows us to calculate derived quantities that describe temporally varying movement behaviour, such as residence time, speed and persistence in direction. The model is flexible, easy to implement and computationally efficient. 4. We apply this model to data from Colorado Canada lynx (Lynx canadensis) and use derived quantities to identify changes in movement behaviour.

Summary: New methods for modeling animal movement based on telemetry data are developed regularly.With advances in telemetry capabilities, animal movement models are becoming increasingly sophisticated.Despite a need for population-level inference, animal movement models are still predominantly developed for individual-level inference. Most efforts to upscale the inference to the population-level are either post hoc or complicated enough that only the developer can implement the model. Hierarchical Bayesian models provide an ideal platform for the development of population-level animal movement models but can be challenging to fit due to computational limitations or extensive tuning required. We propose a two-stage procedure for fitting hierarchical animal movement models to telemetry data. The two-stage approach is statistically rigorous and allows one to fit individual-level movement models separately, then resample them using a secondary MCMC algorithm. The primary advantages of the two-stage approach are that the first stage is easily parallelizable and the second stage is completely unsupervised, allowing for a completely automated fitting procedure in many cases. We demonstrate the two-stage procedure with two applications of animal movement models. The first application involves a spatial point process approach to modeling telemetry data and the second involves a more complicated continuous-time discrete-space animal movement model. We fit these models to simulated data and real telemetry data arising from a population of monitored Canada lynx in Colorado, USA.

Conservation scientists and resource managers often have to design monitoring programs for species that are rare or patchily distributed across large landscapes. Such programs are frequently expensive and seldom can be conducted by one entity. It is essential that a prospective power analysis be undertaken to ensure stated monitoring goals are feasible. We developed a spatially based simulation program that accounts for natural history, habitat use, and sampling scheme to investigate the power of monitoring protocols to detect trends in population abundance over time with occupancy-based methods. We analyzed monitoring schemes with different sampling efforts for wolverine (Gulo gulo) populations in 2 areas of the U.S. Rocky Mountains. The relation between occupancy and abundance was nonlinear and depended on landscape, population size, and movement parameters. With current estimates for population size and detection probability in the northern U.S. Rockies, most sampling schemes were only able to detect large declines in abundance in the simulations (i.e., 50% decline over 10 years). For small populations reestablishing in the Southern Rockies, occupancy-based methods had enough power to detect population trends only when populations were increasing dramatically (e.g., doubling or tripling in 10 years), regardless of sampling effort. In general, increasing the number of cells sampled or the per-visit detection probability had a much greater effect on power than the number of visits conducted during a survey. Although our results are specific to wolverines, this approach could easily be adapted to other territorial species.

All rights reserved. No reuse allowed without permission.(which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. from Canada. We also tested the ability of our model to distinguish empty images from those 56 with animals in another out-of-sample dataset from Tanzania, containing a faunal community 57 that was novel to the model. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint . http://dx.doi.org/10.1101/346809 doi: bioRxiv preprint first posted online Jun. 13, 2018; 3 4. The use of machine learning to rapidly and accurately classify wildlife in camera trap images 66 can facilitate non-invasive sampling designs in ecological studies by reducing the burden of 67 manually analyzing images. We present an R package making these methods accessible to 68 ecologists. We discuss the implications of this technology for ecology and considerations that 69 should be addressed in future implementations of these methods. 70