1. Machine (especially deep) learning algorithms are changing the way wildlife imagery is processed. They dramatically speed up the time to detect, count classify animals and their behaviours. Yet, we currently lack a systematic literature survey on its use in wildlife imagery.2. Through a literature survey (a ‘rapid’ review) and bibliometric mapping, we explored its use across: 1) species (vertebrates), 2) image types (e.g., camera traps, or drones), 3) study locations, 4) alternative machine learning algorithms, 5) outcomes (e.g., recognition, classification, or tracking), 6) reporting quality and openness, 7) author affiliation, and 8) publication journal types.3. Typically, studies have focused on single large charismatic or iconic mammalian species and used neural networks (i.e., deep learning). Additional taxa or alternative machine learning algorithms were rarely used, with limited sharing of code. There were considerable gaps, and therefore there is a great promise for deep learning to transform behavioural detection, classification, and tracking of wildlife.4. Much of the published research and focus on animals came from India, China, Australia, or the USA. There were relatively few collaborations across countries. Given the power of machine learning, we recommend increasing collaboration and sharing approaches to utilise increasing amounts of wildlife imagery more rapidly and transform and improve understanding of wildlife behaviour and conservation.5. Our survey augmented with bibliometric analyses provide valuable signposts for future studies to resolve and address shortcomings, gaps, and biases.
Abstract Almost half of Australian freshwater turtle species are formally listed as threatened, but little is known about the effects of water management and other factors on the abundance and health of freshwater turtles in arid and semi‐arid regions. This study investigated how river flows (including the controlled release of water from an upstream storage facility, or ‘environmental flow’) and water quality might affect the abundance and nutritional status of three freshwater turtle species in three dryland rivers of the Murray–Darling basin in south‐eastern Australia. No response in abundance or nutritional status of the broad‐shelled turtle, Chelodina expansa , the eastern long‐necked turtle, Chelodina longicollis , and the Macquarie turtle, Emydura macquarii , was detected in relation to river flows, possibly because of the small magnitude of the environmental flow. However, for C. expansa , the catch per unit effort (CPUE) was negatively related to increasing macrophyte cover, electrical conductivity (EC, an indicator of salinity), and turbidity. CPUE for C. longicollis was positively related to macrophyte cover and EC, and for E. macquarii it was positively related to macrophyte cover. Haematological measurements suggested that the turtles had healthy nutritional status. Body condition and blood glucose and protein were related significantly to EC, whereas haematological measurements varied significantly among species and between spring and summer. The main conclusion is that water management measures to help the conservation of these turtles should include sufficient environmental flow to produce overbank flooding, thereby creating and maintaining a wide range of habitat suitable for the differing needs of the individual species.
The classification of freshwater ecosystems is essential for effective biodiversity conser-vation and ecosystem management, particularly with increasing threats. We developed an automated approach to mapping and classifying freshwater ecosystem functional groups based on the IUCN Global Ecosystem Typology (GET), offering a scalable, dy-namic, and efficient alternative to current manual methods. Our method leveraged remote sensing data and thresholding algorithms to classify ecosystems into distinct ecosystem functional groups, accounting for challenges such as temporal and spatial variability of freshwater ecosystems, inconsistencies in manual classification, and the complexities of dynamic ecosystems. Unlike traditional approaches, which rely on manual cross-referencing to adapt existing maps and subjective biases, our system is repeatable, transparent, and adaptable to new incoming satellite data. We demonstrate the applicability of this method in the Paroo-Warrego region of Australia (~14,000,000 ha), highlighting the automated classification’s capacity to process large areas with diverse ecosystems. Although some functional groups require static datasets due to current lim-itations in satellite data resolution, the overall approach had high accuracy (84%) and potential for global scalability. This work provides a foundation for future applications to other freshwater ecosystems around the world underpinning biodiversity management, monitoring and reporting worldwide.
Abstract We compared diets of marabou storks Leptoptilos crumenifer foraging from urban landfills and natural areas in northern Botswana using stable isotope analyses and inductively coupled plasma mass spectrometry on moulted feathers. There were significant differences in the diet of marabous foraging from natural areas compared to urban waste sites, reflected by lower δ13C and less enriched δ15N concentrations in those feeding at landfills, suggesting a shift in trophic niche. Feathers from birds foraging at landfills also had significantly higher concentrations of chromium, lead, nickel, and zinc and lower levels of cadmium and potassium than feathers sampled from natural areas. We also analysed marabou regurgitant (42 kg, naturally expelled indigestible food resources) from the Kasane landfill site. More than half was plastic, with single regurgitants weighing up to 125 g. Urban waste stored in open air landfills is altering some marabou diets, affecting their natural trophic niche, resulting in the consumption (and regurgitation) of large amounts of plastic, and exposing marabou to potentially chronic levels of trace metals. Despite the marabou’s apparent resilience to this behavioural shift, it could have long-term effects on the population of the marabou stork, particularly considering Botswana has some of the few regular marabou breeding colonies in southern Africa.
Machine (especially deep) learning algorithms are changing the way wildlife imagery is processed. They dramatically speed up the time to detect, count, and classify animals and their behaviours. Yet, we currently have very few systematic literature surveys on its use in wildlife imagery. Through a literature survey (a ‘rapid’ review) and bibliometric mapping, we explored its use across: 1) species (vertebrates), 2) image types (e.g., camera traps, or drones), 3) study locations, 4) alternative machine learning algorithms, 5) outcomes (e.g., recognition, classification, or tracking), 6) reporting quality and openness, 7) author affiliation, and 8) publication journal types. We found that an increasing number of studies used convolutional neural networks (i.e., deep learning). Typically, studies have focused on large charismatic or iconic mammalian species. An increasing number of studies have been published in ecology-specific journals indicating the uptake of deep learning to transform the detection, classification and tracking of wildlife. Sharing of code was limited, with only 20% of studies providing links to analysis code. Much of the published research and focus on animals came from India, China, Australia, or the USA. There were relatively few collaborations across countries. Given the power of machine learning, we recommend increasing collaboration and sharing approaches to utilise increasing amounts of wildlife imagery more rapidly and transform and improve understanding of wildlife behaviour and conservation. Our survey, augmented with bibliometric analyses, provides valuable signposts for future studies to resolve and address shortcomings, gaps, and biases.
Using drones to count wildlife saves time and resources and allows access to difficult or dangerous areas. We collected drone imagery of breeding waterbirds at colonies in the Okavango Delta (Botswana) and Lowbidgee floodplain (Australia). We developed a semi-automated counting method, using machine learning, and compared effectiveness of freeware and payware in identifying and counting waterbird species (targets) in the Okavango Delta. We tested transferability to the Australian breeding colony. Our detection accuracy (targets), between the training and test data, was 91% for the Okavango Delta colony and 98% for the Lowbidgee floodplain colony. These estimates were within 1–5%, whether using freeware or payware for the different colonies. Our semi-automated method was 26% quicker, including development, and 500% quicker without development, than manual counting. Drone data of waterbird colonies can be collected quickly, allowing later counting with minimal disturbance. Our semi-automated methods efficiently provided accurate estimates of nesting species of waterbirds, even with complex backgrounds. This could be used to track breeding waterbird populations around the world, indicators of river and wetland health, with general applicability for monitoring other taxa.
Abstract Point 1: Portable x-ray fluorescent (pXRF) technology provides significant opportunities for rapid, non-destructive data collection in a range of fields of study. However, there are sources of variation and sample assumptions that may influence the data obtained, particularly in biological samples. Point 2: We used representative species for four taxa (fish, mammals, birds, reptiles) to test the precision of replicate scans, and the impact of sample thickness, sample state, scan location and scan time on data obtained from a pXRF. Point 3: We detected significant differences in concentration data due to sample state, scanning time and scanning location for all taxa. Infinite thickness assumptions were met for fish, reptile and mammal representatives at all body locations when samples were thawed, but not dried. Infinite thickness was not met for feathers. Scan time results found in most cases the 40, 60 and 80 second beam times were equivalent. Concentration data across replicate scans were highly correlated. Point 4: The opportunities for the use of pXRF in biological studies are wide-ranging. These findings highlight the considerations required when scanning biological samples to ensure the required data are suitably collected, while maintaining minimal radiation exposure to live animals.
'/%3e%3c/defs%3e%3cpath fill='%23012169' d='M0 0h10080v5040H0z'/%3e%3cpath stroke='%23fff' stroke-width='.6' d='m0 0 6 3m0-3L0 3' clip-path='url(%23b)' transform='scale(840)'/%3e%3cpath stroke='%23e4002b' stroke-width='.4' d='m0 0 6 3m0-3L0 3' clip-path='url(%23c)' transform='scale(840)'/%3e%3cpath stroke='%23fff' stroke-width='840' d='M2520 0v2520M0 1260h5040'/%3e%3cpath stroke='%23e4002b' stroke-width='504' d='M2520 0v2520M0 1260h5040'/%3e%3cg fill='%23fff'%3e%3cuse xlink:href='%23d' x='2520' y='3780'/%3e%3cuse xlink:href='%23a' x='7560' y='4200'/%3e%3cuse xlink:href='%23a' x='6300' y='2205'/%3e%3cuse xlink:href='%23a' x='7560' y='840'/%3e%3cuse xlink:href='%23a' x='8680' y='1869'/%3e%3cuse xlink:href='%23e' x='8064' y='2730'/%3e%3c/g%3e%3c/svg%3e)