Camera Trapping to Assess Status and Composition of Mammal Communities in a Biodiversity Hotspot in Myanmar

Occupancy of the Ethiopian endemic Moorland Francolin in pristine and degraded Afroalpine biome using a camera trap approach

Occupancy modeling is an essential tool for understanding species-habitat associations, thereby helping to plan the conservation of rare and threatened wildlife species. The conservation status and ecology of several avian species, particularly ground-dwelling birds, are poorly known in Ethiopia. We used camera trap-based occupancy modeling to investigate habitat covariate influence on occupancy (Ψ) and detection probability (ρ) estimates of Moorland Francolins Scleroptila psilolaema from spatially replicated surveys across both relatively pristine and disturbed landscapes in the Afroalpine biome of Ethiopia. Model-averaged estimate of ψ ^ across all sites was 0.76 (SD = 0.28) and ρ ^ was 0.77 (SD = 0.13) in the pristine landscape. The ψ ^ of the species in the disturbed landscape was 0.56 (SD = 0.19) and ρ ^ was 0.48 (SD = 0.06). As hypothesized, based on our model-averaged beta coefficient estimates (βmean ± SE), predators significantly negatively influenced the occupancy of Moorland Francolins in pristine habitat. We also found a significant positive association of occupancy with herb species richness. Contrary to our prediction, distance to road significantly negatively influence the occupancy of the species, suggesting that occupancy probability was highest in proximity to roadsides and trails in the pristine habitat. There was no significant influence of habitat covariates on the occupancy of the species in the disturbed habitat. The most important covariates that significantly influence the detectability of the species in pristine habitat included sampling occasion and precipitation. The greater occupancy and detectability of this endemic species in the pristine habitat could be linked with the particular conservation status and management of this biodiversity hotspot in the central highlands of Ethiopia. Our results suggest that strict legal enforcement is required to sustainably preserve Moorland Francolins and the ecological integrity of the entire Afroalpine biome. We recommend using camera traps in order to develop realistic and effective conservation and management strategies for rare, sensitive, cryptic, and ground-dwelling animals in the region.

A Camera-Trap Survey of Mammals in Thung Yai Naresuan (East) Wildlife Sanctuary in Western Thailand

The Thung Yai Naresuan (East) Wildlife Sanctuary (TYNE), in the core area of the Western Forest Complex of Thailand, harbors a diverse assemblage of wildlife, and the region has become globally significant for mammal conservation. From April 2010 to January 2012, 106 camera traps were set, and, in 1817 trap-nights, registered 1821 independent records of 32 mammal species. Of the 17 IUCN-listed (from Near Threatened to Critically Endangered) mammal species recorded, 5 species listed as endangered or critically endangered included the Asiatic elephant (Elephas maximus), tiger (Panthera tigris), Malayan tapir (Tapirus indicus), dhole (Cuon alpinus), and Sunda pangolin (Manis javanica). The northern red muntjac (Muntiacus vaginalis), large Indian civet (Viverra zibetha), Malayan porcupine (Hystrix brachyuran), and sambar deer (Cervus unicolor) were the most frequently recorded species (10-22 photos/100 trap-nights), representing 62% of all independent records, while the golden jackal (Canis aureus), clouded leopard (Neofelis nebulosa), marbled cat (Pardofelis marmorata), and Sunda pangolin were the least photographed (<0.1/100 trap-nights). Species accumulation curves indicated that the number of camera trap locations needed to record 90% of taxa recorded varied from 26 sites for herbivores to 67 sites for all mammals. TYNE holds a rich community of mammals, but some differences in photo-rates from an adjacent sanctuary and comparisons with other research on local mammals suggest that some species are rare and some are missed because of the limitations of our technique. We also conclude that the management and conservation plan, which involves the exclusion of human activities from some protected areas and strict protection efforts in the sanctuaries, is still suitable for providing key habitats for endangered wildlife populations, and that augmented and regular survey efforts will help in this endeavor.

A Lightweight Automatic Wildlife Recognition Model Design Method Mitigating Shortcut Learning

Recognizing wildlife based on camera trap images is challenging due to the complexity of the wild environment. Deep learning is an optional approach to solve this problem. However, the backgrounds of images captured from the same infrared camera trap are rather similar, and shortcut learning of recognition models occurs, resulting in reduced generality and poor recognition model performance. Therefore, this paper proposes a data augmentation strategy that integrates image synthesis (IS) and regional background suppression (RBS) to enrich the background scene and suppress the existing background information. This strategy alleviates the model's focus on the background, guiding it to focus on the wildlife in order to improve the model's generality, resulting in better recognition performance. Furthermore, to offer a lightweight recognition model for deep learning-based real-time wildlife monitoring on edge devices, we develop a model compression strategy that combines adaptive pruning and knowledge distillation. Specifically, a student model is built using a genetic algorithm-based pruning technique and adaptive batch normalization (GA-ABN). A mean square error (MSE) loss-based knowledge distillation method is then used to fine-tune the student model so as to generate a lightweight recognition model. The produced lightweight model can reduce the computational effort of wildlife recognition with only a 4.73% loss in accuracy. Extensive experiments have demonstrated the advantages of our method, which is beneficial for real-time wildlife monitoring with edge intelligence.