Search

Large areas of the tropics have been cleared of forest and converted to agriculture. The consequent human-modified landscapes (HMLs) comprise a heterogenous mix of habitats; forest fragments and riparian strips are embedded in a matrix of cattle pasture, non-native timber plantations, and urban centres. These habitat changes can have dramatic consequences for wildlife, leading to range shifts and extirpations. In turn, this can influence the integrity of ecosystem services such as frugivory and seed dispersal. Understanding how habitat conversion affects natural ecosystems is critical to inform conservation interventions, but requires long-term biodiversity monitoring and detailed knowledge of species-level responses to HMLs. The research presented in this thesis was conducted in the Emparador HML, in central Republic of Panama. In Chapter 2, we show that the regional avian community is shaped by extent of forest cover across the landscape, and to a lesser degree, extent of forest fragmentation and distance to core forest. Effects of forest cover and fragmentation were examined at local (10 ha) and landscape (500 ha) scales. Species-level responses to these factors varied widely; while abundance of many species increased with greater local-scale forest cover, greater landscape-scale forest cover was often associated with declines. Generalist species that readily persist in HMLs still responded positively to local-scale forest cover, suggesting that even smaller forest fragments in these landscapes are important for maintaining diverse avian assemblages. Critically, we found that species’ responses were not associated with particular traits such as dietary composition or forest dependence, highlighting that species may often exhibit idiosyncratic responses to landscape structure. Chapters 3 and 4 address the wider issue of long-term monitoring, and the potential for data collection over large spatiotemporal scales using remote audio recorders. Ecoacoustics, the study of environmental sound is a relatively new discipline, and as such there is still considerable uncertainty surrounding best-practice for collecting and processing recordings. One of the most straightforward means of utilising audio recordings for environmental monitoring is via acoustic indices. These are objective measures of sound based on features such as pitch and amplitude. To date, attempts to use these indices have been hindered by inconsistent or inappropriate methodologies. In Chapter 3, we determine how many recordings are required to comprehensively capture a soundscape, the acoustic energy of a location. Furthermore, we demonstrate that there are habitat-specific patterns in acoustic indices values, suggesting that these indices reflect differences in vegetation structure and wildlife. We develop this further in Chapter 4, where we show that avian species richness and abundance are clearly linked to patterns in acoustic indices values. Critically, these patterns were coherent among habitat types emphasising their potential for monitoring. Acoustic indices sensitive to the frequencies occupied by bird song have the greatest potential for monitoring an avian community. The results from these two chapters suggest that acoustic indices can be effective tools for monitoring biodiversity, with values reflecting consistent differences across habitats, and among avian assemblages. Audio recordings are a source of permanent, verifiable evidence that can be collected at much greater spatiotemporal scales than traditional biodiversity monitoring data. As the use of audio recorders grows, it is important to compare their efficacy with standard methods of data collection. In Chapter 5, we contrast data derived from audio recordings with that gathered using standard point count methods, and consider whether recorders are a feasible means of surveying antbirds (Thamnophilidae), a disturbance-sensitive avian taxon. Both approaches revealed species’ responses to landscape structure, with qualitatively similar patterns in response to forest cover and vegetation quality. We show that common species can be readily monitored using audio recorders, with greater levels of detectability compared with point counts. However, rarer species were more likely to be detected using point counts. The work presented in this thesis helps to explain the patterns seen in avian responses to Neotropical HMLs. In particular we emphasise the importance of forest cover for maintaining bird assemblages in these landscapes. We demonstrate the utility of audio recorders for data collection, and highlight their potential for future biodiversity monitoring. In the face of human population growth, and ongoing habitat disturbance and agricultural intensification, conservation efforts are essential to avoid widespread species extinctions and ecosystem collapse. Interventions must take place in HMLs, to bolster ecosystem services, provide buffer zones for protected areas, and improve connectivity in the wider landscape.

I attempt to highlight the large gap in knowledge required to pursue New Zealand’s new approach to deer management. This approach requires identifying a complex array of species and site specific factors that potentially influence population success, impacts, value, and management for one of New Zealand’s smallest deer herds, the Wakatipu white-tailed deer (Odocoileus virginianus). White-tailed deer in North America (and in introduced populations around the world) have been researched extensively, providing an unparalleled base of knowledge to review factors that may affect Wakatipu white-tailed deer population dynamics, including habitat quality, disease, hunter harvest and interspecies competition. An overlapping spatial mosaic of limiting factors that may influence white-tailed deer population success across the Wakatipu range were identified. Further research to determine the impact these factors have on the Wakatipu white-tailed deer population, individually and collectively across the landscape, is requried to support this herd’s management. Understanding how deer populations utilise local habitat is a critical first step in designing deer management plans. I used GPS telemetry technology to identify habitat use of Wakatipu white-tailed deer. Initial disturbance to deer behaviour is commonly observed following capture and handling for collar placement. We compared the behaviour of ten deer; four resident and six translocated into a novel, geographically disjunct area, over 50 days following helicopter capture and GPS collar placement. I identified the point at which behaviour disturbed by the capture deer returned to a stable nature. Habitat use by collared deer was patchy across the landscape and did not vary between resident and translocated groups. These results suggest the probability of habitat use could be mapped to predict the presence/absence of Wakatipu white-tailed deer across the wider landscape, thus informing those areas where management and monitoring of Wakatipu white-tailed deer and their impacts on the environment should be focused. A factor limiting population success unique to New Zealand was identified. Intensive ground-based searches for white-tailed deer carcasses were conducted in the Dart Valley/Routeburn catchments following the aerial application of 1080 (sodium fluoroacetate) containing cereal baits as part of a predator control program. Frequent predator control operations will likely affect the sustainable hunting of Wakatipu white-tailed deer and growth in their population. The efficacy of two deer repellents for repelling captive white-tailed deer from consumption of nontoxic cereal baits was assessed. Both repellents significantly decreased bait consumption and I observed clear displays of aversion to repellent baits by captive deer. Ten GPS collared wild white-tailed deer were then monitored following a predator control operation using one of the deer repellents; EDR. One collared deer was poisoned following the operation confirming that EDR is not 100% effective. To support management for hunting, the most effective available deer repellent should be used in future predator control operations over the Wakatipu white-tailed deer range. Deer management plan development requires an in-depth understanding of governing legislation, population management tools and monitoring techniques. I outlined the status of these pertaining to deer management in New Zealand and identify a range of opportunities and challenges within each for developing a management plan for the Wakatipu white-tailed deer herd. Many of these are related to policy development and strategic planning, funding, and implementation processes. Also, monitoring of the Wakatipu white-tailed deer population and its environment needs to be developed from scratch. Overall, I highlight the significant research required to pursue New Zealand’s new approach to deer management. A question remains, who undertakes this work and who pays for it?

Introduction: Cryptic species are often studied using trail cameras in the wild. However, their use may cause some animals to be attracted or repelled by them, skewing presence/absence studies, abundance/population estimates and general behaviour patterns. Red panda are an elusive and rare species, found in only the Himalayan mountain range. A recent study observed red panda examining trail cameras on their trails in the wild. Understanding how they respond to trail cameras may influence how cameras are used to monitor them in the future. Aims: I assessed whether trail cameras affected captive red panda behaviour. I also determined time budgets and how they may differ with different observational methods. Methods: I used three zoo study sites, Auckland Zoo, Hamilton Zoo and Currumbin Wildlife Sanctuary. Auckland Zoo had a male and female pair with their three-month-old male offspring. Hamilton Zoo had an older male, female pair with their four-year-old female offspring. Currumbin Wildlife Sanctuary housed a single older male. A Kinopta Blackeye camera was set up on enclosure fencing to record continuously for the full study period. Study periods were split into three sections, with the first being labelled as before trail cameras, this just had the Blackeye camera. The middle third had trail cameras set up inside the enclosure and was called the during period. The last third of the trial (after trail cameras removed) had only the Blackeye setup, with trail cameras removed. During the entire study period, direct personal observations were also taken, noting typical significant factors, such as weather and temperature. Statistical analysis was carried out using R studio, with a mixture of chi-square, negative binomial GLMs, emmeans, pairwise comparisons and AIC tests. Graphs were created with Excel and R studio. Results: Sleeping was the most common behaviour, followed by locomotion, resting, eating and grooming. All other behaviours were less than 1% of all behaviours. The most active periods occurred in a crepuscular pattern, as with wild panda and in some cases, followed keeper timings. This led to the significant difference between zoo activity budgets. There was a significant difference in types of behaviours recorded with the two observational methods, showing that method does affect the type of data collected. Trail cameras affected behaviour at all zoos by changing the way red panda spent their time. Captive red panda were slightly more active with trail camera presence. Temperature also had a significant impact on length of behaviours. Red panda spent more time sleeping and resting at higher temperatures. Conclusions: As trail cameras changed the way red panda spent their time (in a captive setting), care should be taken for using trail cameras in the wild. While stress responses and obvious signs of avoiding areas with trail cameras did not occur, if behaviour is being monitored, then it is likely to be skewed by trail camera presence. Red panda were more active during trail camera presence, which might suggest an inflated abundance estimate if using trail cameras in the wild.

Existing research recognises the critical role played by Unmanned Aerial Vehicles (UAVs) (also referred to as drones) to numerous civilian applications. Typical drone applications include surveillance, wireless communication, agriculture, among many others. One of the biggest challenges is to determine the deployment and navigation of the drones to benefit the most for different applications. Many research questions have been raised about this topic. For example, drone-enabled wildlife monitoring has received much attention in recent years. Unfortunately, this approach results in significant disturbance to different species of wild animals. Moreover, with the capability of rapidly moving communication supply towards demand when required, the drone equipped with a base station, i.e., drone-cell, is becoming a promising solution for providing cellular networks to victims and rescue teams in disaster-affected areas. However, few studies have investigated the optimal deployments of multiple drone-cells with limited backhaul communication distances. In addition, the use of autonomous drones as flying interactors for many real-life applications has not been sufficiently discussed. With superior maneuverability, drone-enabled autonomous aerial interacting can potentially be used on shark attack prevention and animal herding. Nevertheless, previous studies of autonomous drones have not dealt with such applications in much detail. This thesis explores the solutions to all the mentioned research questions, with a particular focus on the deployment and navigation of the drones. First, we provide one of the first investigations into reducing the negative impacts of wildlife monitoring drones by navigation control. Second, we study the optimal placement of a group of drone-cells with limited backhaul communication ranges, aims to maximise the number of served users. Third, we propose a novel method named ‘drone shark shield’, which uses communicating autonomous drones to intervene and prevent shark attacks for protecting swimmers and surfers. Lastly, we introduce one of the first autonomous drone herding systems for mustering a large number of farm animals efficiently. Simulations have been conducted to verify the effectiveness of the proposed approaches. We believe that our findings in this thesis shed new light on the fundamental benefits of autonomous civilian drones.

Estimating population abundance through time is an essential component of wildlife conservation and management. However, accurate population monitoring can be difficult and expensive for many elusive species occurring across large, dynamic landscapes. Thus, wildlife managers require methods that accurately estimate population abundance, while also minimizing field effort and cost. We estimated abundance of invasive wild pigs (Sus scrofa) on Tejon Ranch in the Tehachapi Mountains of California, using natural markings to identify individuals for mark-resight population estimation. Wild pigs in this region, like many species not traditionally identified using natural marks, are generally homogeneous in appearance with distinctive features that range dramatically in relative visibility, uniqueness and permanence. We developed a method based on standardized thresholds of image quality and animal flank distinctiveness to account for the inherent variability of natural markings between individuals. This method was tested over a fifteen-month period between March 2015 and June 2016, using an array of 48 camera traps across a 48km² survey grid. With 18.5% of wild pigs meeting our conservative standard of identifiability, we generated absolute estimates of abundance across five consecutive three-month sampling periods using the Poisson log-normal estimator under Pollocks robust design. Using left-flank photos of both naturally marked and unmarked pigs, we generated abundance estimates ranging from 352 (SE + 56) individuals in summer 2015 to 157 (SE + 43) individuals in spring 2016. These results suggest an overall decline in the wild pig population on Tejon Ranch from 2015 to 2016, which is supported by a simultaneous decline in Ranch-wide hunter harvest totals during this period. As this mark-resight method requires no trapping or tagging of any kind, it may be utilized as a cost and resource efficient alternative to traditional mark-resight techniques that rely on ear-tags or neck-bands for individual identification of species traditionally considered unidentifiable using natural marks alone.

Abstract: Camera traps are an increasingly popular tool for wildlife management. Studies that use detection rates as a simple index of relative abundance assume that movement is not density-dependent. More complex techniques such as spatially-explicit capture recapture models, occupancy models, or N-mixture models make various assumptions about animal home range size in relation to camera spacing. While some assume individual animals can visit multiple camera sites, others assume sites are independent such that no individual can be detected at more than one camera. In all these methods, variation in space use has the potential to confound population estimates and compromise current applications of camera trapping as a monitoring tool. To assess this problem, I quantified how movement rate and home range size vary both between and within populations, and investigated the implications of this variation for camera trap data analysis. To quantify the relationship between space use and population density, I conducted a meta-analysis of studies reporting movement rate, home range size, or density for at least two populations of terrestrial mammals. I found that movement rate and home range size are significantly negatively correlated with density and positively correlated with each other. Using simulations of animal movement and density, I found that density-dependent movement can obscure trends in density indexed by detection rates. True changes in density may be underestimated by up to 30%. I then investigated space use within populations using telemetry datasets for white-tailed deer, moose, and wolves. Patterns between individuals were similar to those patterns seen between populations, where individuals with larger home range sizes generally had faster movement rates. Variation between individuals within a population was an order of magnitude greater than variation between the mean movement rates and home range sizes of different populations. Carnivores in particular exhibited much greater variation between individuals than herbivores. These findings indicate that choosing camera spacing appropriate for a given model is not straightforward, as some individuals are likely violating the model assumptions regarding site independence. I recommend that practitioners treat camera trap detection rate indices of relative abundance as good indicators of directional population trends and as partially accurate indicators of the actual magnitude of density changes. Furthermore, I urge that practitioners be aware that individual variation in space use is considerable, and that assumptions regarding camera spacing relative to animal home range size are likely often violated. Animal space use is an important subject for the ongoing development of robust camera trap analyses, and I hope that this thesis will encourage a more careful consideration of its role in the design of future camera trapping studies.

Advances in data availability and methodology have greatly expanded the potential for landscape scale monitoring of biodiversity and the use of resulting insights for conservation planning. This collection of papers explores the use of modeling, monitoring, and conservation planning techniques to better facilitate landscape scale conservation in Southeast Asia.Chapter 1 focuses on the use of linear and logistic regression techniques for modeling occurrence patterns of tigers and tiger prey over a large landscape in western Thailand. I discuss the strengths and limitations of using these techniques to inform landscape scale monitoring and planning activities. Chapter 2 explores the potential for sign surveys of a landscape scale predator (tiger) to inform management decisions. I compare presence/absence and count survey frameworks and assess their value in informing landscape level management. A controlled study of tiger sign interpretation provides the much of the basis for the raw data used in this chapter and this study is presented in a detailed appendix. Chapters 3 through 5 focus on a biodiversity gap analysis conducted for Cambodia's protected area system. Chapter 3 provides a detailed account of methods used to map biodiversity surrogates for the country. Chapter 4 reviews the history and current status of Cambodia's protected area system and includes recommendations for revising IUCN category assignments for the system. Chapter 5 is a detailed account of a national level conservation planning exercise in Cambodia and how this process was translated into specific recommendations for improving Cambodia's protected area system.

Long-term, large-scale monitoring of wildlife metacommunities is needed to recognize population declines early enough to identify environmental stressors and facilitate adaptive planning. Potential outcomes include information supporting the designation of new species of conservation concern, or better yet, conservation actions that avert the need for conferring critical statuses. By surveying multiple species, declines of individual species need not be considered in isolation, but can be compared to responses of other species in the metacommunity. However, multi-species monitoring presents a variety of new challenges in terms of appropriate survey methods and analytical techniques for drawing valid ecological inferences.In this study, I investigated several related aspects of multi-species monitoring. One theme was the role of automated survey methods (e.g., audio recorders and camera stations) that leave a permanent record and easily provide for temporal replication of surveys. I applied occupancy models to repeat surveys for addressing detection probability and providing unbiased estimates of species occurrence. Lastly, I evaluated several novel quantitative methods for comparing community properties using monitoring data. In the first chapter I considered the effectiveness of automated recorders for monitoring common birds in California forests. I applied single-species occupancy models to 46 species using 5 years of monitoring data in which automated recorders were placed at 453 random sites across a 5.4-million-ha northern California study area. The devices were programmed to record sounds during up to 3 surveys each morning on 3 consecutive days during the breeding season when songbirds were singing from territories. Skilled interpreters reviewed these recordings to identify all species heard during each survey. With Monte Carlo simulation and results from occupancy models, I demonstrated 80% power for monitoring declines as small as 2.5% per year over 20 years for 32 species given a sampling effort of 100 new sites per year. I also determined an effective survey area radius of 30 m to 50 m for automated recorders, and showed that the devices provided similar occupancy estimates as traditional point counts despite lower survey-level detection probability.In the second chapter I applied multi-species occupancy models to the calculation of biodiversity indices describing metacommunity organization. I used the same automated recorder data set for birds from Chapter One. Specifically, I applied simulation and Bayesian hierarchical models to demonstrate how a failure to address detection probability heterogeneity underestimates the evenness of species occupancy distributions. In models of the bird data I found that a number of species traits (migration, foraging guild, territoriality, body size) were informative in explaining detection probability. By pooling information from common species in a multi-species model, I was able to draw stronger inferences about rarer species than by modeling these species individually. Lastly, I illustrated the ecological significance of species-traits modeling and found that warbler and woodpecker occupancies were evener than for sparrows. In the third chapter, I proposed a new quantitative method for comparing species abundance distributions. I illustrated this method using avian point count surveys from 4 research forests in California. I applied bootstrap resampling to probabilistically compare the abundances of intermediate ranks among and within species abundance distributions. I found higher abundances of intermediately-common species on 2 of the forests, and ascribed this finding to differences in forest productivity and habitat complexity leading to greater niche partitioning of resources. At the metacommunity-level, I found higher abundances of intermediately-common species for neotropical migrants compared to resident birds. . In the fourth chapter, I considered the use of baited camera stations for monitoring Pacific fisher (Pekania pennant pacifica) and other mammals. Cameras were placed at 172 randomly selected forest sites across 2.8 million ha of northwestern California. The duration of each survey was 2 to 4 weeks. I estimated regional occupancy from these data at 2 survey scales (e.g., individual sites [0.465] and pairs of sites 1.6 km apart [0.651]). I also demonstrated 80% power for monitoring declines as small as 2.0% per year over 20 years given a sampling effort of 100 new sites per year. Lastly, I calculated the median latency to first detection for 13 other species of mammals detected at > 5% of sites, showed that latency was < 6 days for 10 of these species, and argued that these results strengthen the case for expanding the use of camera traps to multi-species monitoring.