Improving Animal Monitoring Using Small Unmanned Aircraft Systems (sUAS) and Deep Learning Networks

In recent years, small unmanned aircraft systems (sUAS) have been used widely to monitor animals because of their customizability, ease of operating, ability to access difficult to navigate places, and potential to minimize disturbance to animals. Automatic identification and classification of animals through images acquired using a sUAS may solve critical problems such as monitoring large areas with high vehicle traffic for animals to prevent collisions, such as animal-aircraft collisions on airports. In this research we demonstrate automated identification of four animal species using deep learning animal classification models trained on sUAS collected images. We used a sUAS mounted with visible spectrum cameras to capture 1288 images of four different animal species: cattle (Bos taurus), horses (Equus caballus), Canada Geese (Branta canadensis), and white-tailed deer (Odocoileus virginianus). We chose these animals because they were readily accessible and white-tailed deer and Canada Geese are considered aviation hazards, as well as being easily identifiable within aerial imagery. A four-class classification problem involving these species was developed from the acquired data using deep learning neural networks. We studied the performance of two deep neural network models, convolutional neural networks (CNN) and deep residual networks (ResNet). Results indicate that the ResNet model with 18 layers, ResNet 18, may be an effective algorithm at classifying between animals while using a relatively small number of training samples. The best ResNet architecture produced a 99.18% overall accuracy (OA) in animal identification and a Kappa statistic of 0.98. The highest OA and Kappa produced by CNN were 84.55% and 0.79 respectively. These findings suggest that ResNet is effective at distinguishing among the four species tested and shows promise for classifying larger datasets of more diverse animals.

A Test for the Underlying State-Structure of Hidden Markov Models: Partially Observed Capture-Recapture Data

Hidden Markov models (HMMs) are being widely used in the field of ecological modeling, however determining the number of underlying states in an HMM remains a challenge. Here we examine a special case of capture-recapture models for open populations, where some animals are observed but it is not possible to ascertain their state (partial observations), whilst the other animals' states are assigned without error (complete observations). We propose a mixture test of the underlying state structure generating the partial observations, which assesses whether they are compatible with the set of states observed in the complete observations. We demonstrate the good performance of the test using simulation and through application to a data set of Canada Geese.

Effects of liberalizing harvest regulations on Canada goose (Branta canadensis) demography in Nebraska.

Harvest regulations are used to manage populations of game species. Across their range, Canada goose Branta canadensis populations have recovered from near extirpation and are now perceived as overabundant and even a nuisance or a threat to human safety in many regions. Like many states, Nebraska has liberalized harvest regulations to increase recreation opportunities for consumptive users and to control increasing numbers of Canada geese. However, the efficacy of harvest regulations to control populations of geese is unclear. We used a live capture-recapture and dead recovery data set of more than 19,000 Canada geese banded in Nebraska 2006-2017 to determine the effect of liberalized harvest regulations on goose survival and overall growth rate. Our goals were to 1) estimate demographic parameters for Canada geese in five different regions in Nebraska 2) estimate the effect of increasing daily bag limits during the early September season and regular season on survival of hatch-year, juvenile, and adult Canada geese and 3) relate the effect of estimated changes in survival to population growth rate. We found survival (0.54-0.87), fidelity (0.14-0.99), and productivity (number of young per adult, 0.17-2.08) varied substantially among regions within Nebraska. We found increasing early season bag limits, but not regular season bag limits, reduced survival in Canada geese. However, this effect was most pronounced when comparing years without an early season to years with the highest daily bag limits used in Nebraska (eight). Survival of juvenile geese (2-3 years post-hatch) were unaffected by changes in daily bag limits during any season, though the probability of reporting was greatest for this age-class. The observed reductions in survival probability of hatch-year and adult geese due to increased daily bag limits during the early season (<10%) had only weak effects on regional growth rates. Regional growth rate estimates appeared more responsive to changes in adult survival, but only decreased ~5% between years with the most liberal early-season daily bag limits to years without an early season. Our results suggest increased bag limits during the early season may reduce Canada goose survival, but has a weak impact on population growth in Nebraska.