Movement Ecology

BackgroundIn hibernating mammals, the timing of den entry and exit reflects complex interactions among environment, physiology, and energetic constraints, with consequences for fitness. Consequently, shifts in denning phenology can affect population dynamics, particularly under climate change. Reliable estimation of denning timing is therefore critical, yet current methods often rely on GPS-derived movement data, limited by coarse sampling intervals, detection issues, and the inability to distinguish true inactivity from active presence at the den site. In this study, we developed and apply a method to estimate denning phenology in a brown bear population in south-central Sweden using accelerometer-derived activity data. Our approach employs adaptive, individual-specific thresholds to account for variation in baseline activity across bears, focusing on day-to-day changes to identify the start and end of inactivity periods. This method allows flexible and reproducible detection of den entry and exit dates, overcoming limitations associated with fixed thresholds and small sample sizes. ResultsWe compared activity-based estimates with GPS-derived den occupancy and examined variation in denning behavior across demographic groups. Analyzing 388 bear-winters, the method successfully identified inactivity periods in 360 cases. The method failed to identify clear start and end dates of hibernation for 28 (7%) bear-winters, which were characterized by unusually high or low daily activity levels at the boundaries of the inactivity period. Den site occupancy ranged from September 5 to June 2, with durations of 112-260 days, whereas inactivity periods detected from activity data extended from September 6 to May 13, lasting 83-217 days. Our comparison of activity-based and GPS-based methods indicates that bears may arrive at the den site several weeks before the onset of inactivity, with timing varying among demographic groups. ConclusionWe show that activity-based analysis provides a robust framework for estimating denning phenology, distinguishing actual inactivity from site presence, and improving understanding of the timing and variability of bear denning behavior. Applying an individual-level activity-based method improves accuracy in assessing ecological mechanisms underlying hibernation in bears and other hibernators, while also enhancing interpretation of environmental drivers and providing a reliable tool to monitor phenological shifts in response to climate change.

O_LICurtailment of wind farms effectively reduces collision mortality in bats. Implementing this measure in offshore wind farms requires knowledge on the spatiotemporal occurrence and environmental predictors of migration over sea. In bats, such information can be obtained through acoustic monitoring and individual tracking. However, these techniques provide seemingly contradictory insights into migration patterns. C_LIO_LIWe used a Bayesian capture-recapture state-space model to investigate how environmental predictors influence spring departure decisions of Nathusius pipistrelle Pipistrellus nathusii migrating over the North Sea. The model was applied to both acoustic and tracking data, enabling comparable analyses across methods and incorporating uncertainty in migration dates of tracked bats. Additionally, we examined nightly offshore bat occurrence to further explore differences in movement patterns detected by the two techniques. C_LIO_LIWind conditions at 200 m above sea level were identified as key driver of Nathusius pipistrelle spring migration. In May-June, most bats migrated from the United Kingdom under westerly and northwesterly tailwinds. Tracked individuals flew in stronger supportive winds than acoustically recorded bats, which were also detected under crosswinds and headwinds. In March-April, acoustic detections occurred mainly during strong southerly winds, suggesting that early-season migrants largely consisted of individuals migrating over the European mainland and drifted northwards onto the North Sea by strong crosswinds. C_LIO_LIAcoustic detectors primarily recorded bats that landed on offshore platforms, likely because they were unable to cross the North Sea in a single flight due to less favorable wind conditions, or because they departed from more inland locations. In contrast, tracking data mainly represented bats that successfully crossed the North Sea in a non-stop flight under moderate supportive tailwinds. C_LIO_LISynthesis and applications: Combining observation techniques improves our understanding of bat migration patterns. Additionally, acoustic monitoring can capture migration from different geographic origins. Current mitigation measures for offshore wind farms at the North Sea rely solely on acoustic data, likely overlooking the part of the population that crosses over sea with optimal wind support. Acoustic and tracking data are therefore complementary rather than contradictory, and both methods should be used together when developing mitigation measures. C_LI

O_LIDemographic processes in populations are inherently heterogeneous across both space and time. Many ecological models explicitly account for temporal heterogeneity in the demographic rates that govern these processes, but assume spatial homogeneity. Ignoring spatial heterogeneity can bias inference, limit predictive performance, and obscure key spatial structure in demographic rates. Integrated population models (IPMs) offer a powerful framework to estimate spatio-temporal demographic rates by combining diverse ecological data sources collected from multiple sampling locations. However, to accomplish this, IPMs face significant statistical and computational hurdles, including misalignment between different data sources and the need to efficiently account for residual spatial autocorrelation. C_LIO_LIWe present a novel Bayesian spatially explicit integrated population model (sIPM) which integrates population count and capture-recapture data from multiple sampling locations to estimate and predict continuous spatio-temporal demographic rates, such as survival, recruitment and population growth rate, across large geographic domains. This framework employs a joint likelihood approach with change of support to flexibly accommodate spatial and spatio-temporal data misalignment, and incorporates a nearest-neighbor Gaussian process to efficiently model residual spatial autocorrelation and generate spatial predictions. C_LIO_LIWe assess the performance of our sIPM through an extensive simulation study. Results show that our approach provides unbiased and precise estimates and predictions of spatio-temporal demographic rates, even in the presence of significant data misalignment and residual spatial autocorrelation. We demonstrate the utility of our method by analyzing data on Gray Catbirds (Dumetella carolinensis) from the North American Breeding Bird Survey and the Monitoring Avian Productivity and Survivorship program across the eastern coast of the United States from 2004-2014. This analysis results in maps of apparent survival, recruitment and population growth rate, thereby revealing important spatio-temporal variations in demographic rates that would have been obscured by traditional, spatially homogeneous IPMs. C_LIO_LIOur sIPM offers a robust and computationally efficient method for studying spatio-temporal variation in demographic processes across large areas, even in the presence of data misalignment and residual spatial autocorrelation. Ultimately, this framework, applicable to many ecological monitoring programs, facilitates the development of spatially targeted strategies necessary for effective conservation and management. C_LI

Estimating the number of spawning events per female is key to understanding individual reproductive output in batch-spawning species, yet direct observation of spawning is often infeasible in the wild. Recent advances in genetic kinship inference enable the identification of maternal half siblings from young-of-the-year genomic samples, while otolith-based age determination provides reconstruction of offspring birth dates. Here we develop an offspring-based framework for estimating the number of clutches produced by individual females by integrating sibling structure inferred from genomic data with otolith-derived birth-date information. By recasting clutch identification as a richness estimation problem, we apply the Chao1 estimator to infer the total number of spawning events from incomplete offspring samples. Using simulation experiments, we evaluate how sampling effort and heterogeneity in clutch size influence clutch detection and estimation. Under uniform clutch-size distributions, modest numbers of offspring sampled per maternal family (10-20 offspring) yield accurate estimates of the total number of clutches, substantially outperforming naive counts of observed birth-date classes by recovering information from rare or unobserved spawning events. In contrast, skewed or multimodal clutch-size distributions lead to underestimation at low sample sizes, indicating that uneven reproductive output increases sampling effort required for reliable inference. Overall, our results demonstrate how offspring genomic data and otolith-derived birth dates can be jointly leveraged to reconstruct individual spawning histories under realistic sampling constraints. This perspective provides a framework for inferring within-season reproductive schedules in batch-spawning species, and highlights opportunities for integrating genomic and life-history data in fisheries monitoring and reproductive ecology.

Accurately estimating wildlife population sizes, essential for ecological theory and conservation management, yet remains challenging. Although unmanned aerial systems (UASs) combined with machine learning, have revolutionized population estimation, they face limitations in addressing the hierarchical population processes from individual behavior to colony-and population-level dynamics. To overcome this limitation, we developed a data integration framework that jointly analyzes multiple datasets, representing different scales of the same underlying process, were jointly analyzed. Using a seabird colony as a model system, we integrated UAS-based count data in the colony with bio-logging-based tracking data to estimate population size by quantifying both the number of individuals present and the proportion absent from the surveyed area. These complementary datasets were linked using state-space models allowing accurate population estimates with explicit uncertainty quantification. Furthermore, we evaluated the robustness of the estimations with respect to sample size. Sub-sampling simulations revealed that estimation uncertainty was more sensitive to sample size in bio-logging-based tracking data than in UAS-based count data. This finding highlights the importance of understanding dataset-specific properties when designing effective investigations. Overall, our resource-efficient framework is broadly applicable across species and populations and demonstrates how integrating complementary observation methods can improve population estimates and inform conservation practice.

Optimal foraging behavior is a key component of successful adaptations to natural environments. Understanding how animals decide to stay near food or to leave it for another food patch gives us insights into the underlying cognitive mechanisms that govern adaptive behaviors. 3D pose tracking was used to determine how pigeons exploit a 4 square meter arena with two separate platforms (i.e. food patches) whose absolute and relative elevations were manipulated. Detailed kinematic features of foraging and traveling behaviors were quantified using automated video tracking, without a need for manual coding. Our computational approach captured continuous, high-dimensional movement patterns and enabled precise quantification of travel costs between patches. Combined with mixed-effects survival analysis, our fine-grained behavioral tracking provided detailed insight into the moment-by-moment dynamics of patch-leaving decisions of pigeons. As expected from behavior optimization models, our results showed a preference to visit a ground food platform first, and longer latencies to leave an elevated platform. Foraging activity significantly decreased throughout the session, with shorter visits, less pecks per visit, and a decrease in inter-peck variability. However, a mixed-effects Cox regression modeled pigeons patch-leaving probability, demonstrating that current and cumulative foraging parameters between patches significantly enhanced the models predictive power beyond patch accessibility (i.e., beyond travel costs). This suggests that pigeons integrate both current environmental cues and their individual foraging history when making patch-leaving decisions. Our findings are discussed in relation to the marginal value theorem and optimal foraging theory.

Habitat selection analyses, which discern the environmental conditions individuals select, often inform conservation planning. Through a literature review, we demonstrate that recent habitat selection studies rarely include fitness and health information. With a simulation study, we show that ignoring such information could support the protection of sink habitats. Our case studies demonstrate how health and fitness proxies can modify our understanding of habitat selection: (1) incorporating mass gain of thick-billed murres shows the energetic benefit of areas deemed secondary by a naive resource selection function; (2) including number of chicks in a step selection function (SSF) exposes the complex relationships glaucous-winged gulls have with landscapes impacted by humans; and (3) including external signs of trauma in the movement kernel of SSFs demonstrate others ways in which narwhal distribution can be altered. We urge movement ecologists to collect and use health and fitness data to improve ecological inference and conservation action.

O_LIMovement paths reflect temporal shifts in behavioural states, typically driven by internal and external drivers. However, the inherently multiphasic nature of these trajectories is frequently overlooked in empirical studies, an oversight that can hinder progress in our understanding of movement ecology. While Hidden Markov Models (HMMs) can successfully identify latent states--such as foraging or travelling--they face significant challenges, particularly in determining the appropriate number of states and in interpreting their ecological relevance in the context of both movement patterns and environmental covariates. C_LIO_LIWe present a framework based on Hidden Markov Models with Step Selection Functions (HMM-SSFs) that identifies behavioural states, represented by ecologically meaningful labels linked to explicit hypotheses about animal movement, that best explain observed movement patterns. The framework imposes interpretable conditions and diagnostic criteria on the post-identified behavioural states to ensure ecological coherence. It is grounded in the evaluation of biologically motivated scenarios rather than purely data-driven partitioning. The framework proceeds in two main steps: first, movement-based states are identified using movement-derived covariates only; second, these states are refined by incorporating environmental predictors, such as habitat structure or species interactions (e.g., predator-prey dynamics). This sequential integration enables the detection of ecological responses that are conditional on behavioural context. C_LIO_LISimulations show that the framework effectively recovers behavioural states across most conditions. State decoding accuracy was notably higher when control locations were drawn from an exponential-family distribution, compared to a uniform one. The exponential-family approach improved state separation and reduced mislabelling, especially when few control locations are generated. However, low state persistence--particularly in Encamped behaviours--resulted in an overestimation of the number of states. These findings underscore the influence of transition probabilities on behavioural labelling. Finally, we applied our framework to zebra (Equus quagga) movement data by combining movement predictors with changes in direction toward the nearest preferred habitat. This enabled us to distinguish between habitat-dependent and habitat-independent travelling behaviours, as well as to identify spatially finer-scale such as encamped state. C_LIO_LIThe proposed framework balances complexity and biological interpretability by using basic movement metrics to identify the behavioural states and their sequence that best explain multiphasic movement paths, together with environmental factors directing movement in each state. Unlike traditional methods that predefine the number of states, the framework estimates both state number and labels, offering a flexible and ecologically meaningful approach for behavioural inference. C_LI

Population density can influence individual predation risk in colonial breeders through shared vigilance and predator deterrence. We investigated how predator-prey interactions are shaped by population density at two Antarctic fur seal (Arctocephalus gazella) breeding colonies at Bird Island, South Georgia, which differ four-fold in seal density. By deploying autonomous time-lapse cameras, we captured high-resolution images at one-minute intervals throughout the breeding season. Using a YOLOv8 neural network, we identified fur seal adult males, females and pups, as well as three predator-scavenger bird species: giant petrels (Macronectes spp.), brown skuas (Stercorarius antarcticus) and snowy sheathbills (Chionis alba). Abundance patterns corresponded to the known foraging and breeding behaviours of these species. Differences in seal density between the colonies were mainly driven by adult females and their pups, but not adult males. The ratios of predatory birds to pups were markedly lower at the high-density colony, while scavenger to pup ratios remained similar. Spatial analyses revealed that predators were largely excluded from areas of high seal density, whereas scavengers overlapped extensively with pups in both colonies. This study demonstrates the value of remote observation in resolving predator-prey interactions and illustrates how density can shape predation risk in a colonial breeder.

A defining feature of climate change is the increasing frequency, intensity, and severity of extreme weather events. Among them, extreme heat is recognized as a critical driver of ecological and evolutionary change. Intense heat episodes can exceed physiological limits, alter animal movement, restructure geographic ranges, and increase extinction risk more than gradual changes to mean temperatures. Yet links between extreme heat events and organismal biology remain limited, in part because definitions and metrics are not standardized, and user-friendly workflows and guides are lacking for many biologists. We present a methodological roadmap, with reproducible code, for integrating extreme heat into studies of behavior, physiology, biophysical ecology, species distribution models (SDMs), and population dynamics. First, we provide standardized computational approaches to define and quantify extreme heat. Second, we fit species distribution models for California quail (Callipepla californica) that include an extreme heat metric and showcase improved predictions of habitat suitability, particularly at range edges. Third, we compute biophysical simulations to quantify exposure to thermal stress in Sleepy lizards (Tiliqua rugosa) across distinct macro- and microclimates. Finally, accounting for temporal autocorrelation in temperature profiles in population simulation models, we show that clustered heat extremes--missed by averages--can increase the risk of population collapse. As extreme heat events become more common, incorporating their dynamics is essential for understanding ecological and evolutionary change, designing experiments across species geographic ranges, and supporting conservation in a rapidly warming world. Together, these case studies illustrate a reproducible, organism-informed roadmap to integrate extreme heat into predictions of ecological impacts and inference across levels of biological organization under ongoing climate change.

Group foraging can enhance prey detection, but depending on resource availability, it may also generate conflicts among conspecifics. To understand how animals balance these benefits and costs, foraging performance must be evaluated together with inter-individual interactions. However, under fully natural conditions, it remains challenging to quantify both simultaneously. Here, we investigated how individual foraging efficiency and pairwise interactions are shaped when more than one individuals simultaneously exploit the same foraging patch, using the Japanese large-footed bat (Myotis macrodactylus) as a model system. We monitored an entire pond functioning as a natural foraging patch using two thermal cameras and an eight-channel microphone array, and reconstructed the arrival, prey-attack, and exit times of individual bats. Using a Poisson generalized linear mixed model (GLMM), we found that prey-attack rates were approximately 25% lower during paired flights than during solitary flights. We then constructed a null model in which arrival, attack, and departure events followed independent Poisson processes parameterized from the empirical data. Compared with null-model predictions, both the total duration and the duration of individual paired flights in the empirical data were significantly shorter, indicating that bats limited the time spent co-using the same patch relative to solitary foraging. In addition, the probability that the first exiting individual was the one that arrived earlier or later did not deviate from chance levels, providing no evidence for a prior residence advantage. Together, these results demonstrate that simultaneous patch use avoidance occurs independently of arrival order and coincides with reduced prey-attack rate, suggesting that bats leave shared patches and move to alternative foraging sites to mitigate losses in prey-attack efficiency. Our findings highlight bats as an excellent model system for non-invasively linking individual behavior and foraging performance via echolocation, and for elucidating the dynamics of foraging behavior and sensory interference in the wild.

Density is a major determinant of population dynamics, with high densities exacerbating intraspecific competition and disease transmission, while low densities increase predation risk. To investigate spatiotemporal density patterns and predator-prey interactions in a breeding colony of Antarctic fur seals (Arctocephalus gazella), we deployed an autonomous camera capturing minute-by-minute, high-resolution images throughout a breeding season. Using a YOLO-based neural network, we identified adult males, females and pups, as well as three avian predator-scavengers: giant petrels (Macronectes spp.), brown skuas (Stercorarius antarcticus) and snowy sheathbills (Chionis alba). Analysis of 4.1 million automated detections from over 10,000 high-quality images revealed spatiotemporal abundance patterns corresponding with the known breeding and foraging behaviours of these species. Strong temporal associations emerged between the abundance of pups and two avian species, while fine-scale spatial analyses showed that pups grouped together with other pups and adult females but avoided avian predators and territorial males. Notably, proximity to adult fur seals of both sexes reduced pup predation risk, defined as the distance between the pup and the nearest bird, whereas proximity to other pups did not. Overall, this study provides a framework for quantifying density-dependent interactions in wild populations and emphasises the value of remote observation in ecological research.

Understanding animal movement behavior is essential for conservation and elucidating various ecological processes. In particular, assessing habitat selection is a central theme in movement ecology, traditionally evaluated by estimating travel distances per unit time across diverse environmental conditions based on tracking data. Integrated step selection analysis (iSSA : Avgar et al., 2016) has been most widely applied in conservation studies and ecosystem service quantifications due to its ease of implementation and interpretability. Despite its popularity, however, iSSA faces a critical issue since it can lead to an underestimation of the travel distance per unit time, potentially biasing estimates of step length. This is primarily due to the assumption of linear interpolation between consecutive observed points, which fails to account for the unobserved locations and the actual, non-linear trajectories taken by the animal. In this paper, we proposed a novel method to improve the estimation of travel distance in iSSA, inspired by multiple imputation, which is a statistical method for missing data. We conducted a simulation study to evaluate the extent to which our proposed method, Multiple Imputation Step Selection Analysis (MiSSA), improves the accuracy of step-length estimation (parameters of gamma distribution) compared to conventional iSSA. In simulation studies across various scenarios, MiSSA estimated the step length more accurately than iSSA. Our study demonstrates that incorporating missing data statistics into the iSSA framework improves the accuracy of travel distance estimations, which serve as the foundation for evaluating habitat selection. MiSSA maintains the core advantages of iSSA while enabling more accurate estimation of travel distances, even with low-resolution data where movement between sampling intervals is non-linear. We anticipate its broad application across various disciplines, with a primary focus on conservation.

Understanding habitat suitability and landscape connectivity is essential for conserving wide-ranging carnivores in climate-sensitive high-altitude mountain ecosystems. The first occurrence record of the Tibetan brown bear (Ursus arctos pruinosus) from the Changthang region of Ladakh, India, has raised questions about whether this individual represents an isolated dispersal event or reflects functional connectivity with source populations on the Tibetan Plateau. To evaluate potential habitat suitability and transboundary connectivity, we compiled species occurrence records and associated environmental predictors and developed an ensemble species distribution model using biomod2. We then assessed landscape connectivity using circuit theory implemented in Circuitscape to identify potential ecological corridors. Our models indicate that approximately 1,011,818 km{superscript 2} (21.34%) of the combined Ladakh (India) and the Tibetan Plateau landscape is currently suitable for the species, of which only [~]207,000 km{superscript 2} represents highly suitable habitat. Annual precipitation, precipitation of the wettest month, and precipitation of the warmest quarter were the most influential predictors of habitat suitability. Connectivity analysis identified potential corridors linking the eastern Changthang region of Ladakh with suitable habitat on the Tibetan Plateau, suggesting plausible transboundary ecological connectivity. These results indicate that the recent record from Changthang is more likely driven by landscape-scale functional connectivity than by an isolated dispersal event. Although the mapped corridors represent probable connectivity pathways rather than confirmed movement routes, this study provides the first spatially explicit assessment of habitat suitability and potential transboundary connectivity for the Tibetan brown bear across the Ladakh and Tibetan Plateau landscape.

O_LINitrogen (N) is limiting for terrestrial herbivores, particularly over winter. Caribou (Rangifer tarandus) have adapted to seasonal scarcity of N by accruing muscle mass during the growing season when N is more abundant. C_LIO_LINitrogen stored in muscle tissue is then relied upon during winter to compensate for dietary deficits. Once their diet shifts from N-rich vascular plants to N-poor lichen over winter, caribou can lose [~]30% of their muscle mass. As catabolized N is shed in urine on wintering grounds, caribou could act as elemental transport across seasons and landscapes. Furthermore, if deposited N is taken up by lichen or other winter forage, it might enrich the nitrogen-poor winter diet of caribou in the future. C_LIO_LIWe tested this potential transport via three steps. We analysed Cladonia spp. lichen and vascular plants upon which caribou forage across Fogo Island, Newfoundland, using %N content as our metric of forage quality. We then compared seasonal habitat selection responses to forage quality by caribou using integrated step selection analyses. In summer, caribou selected areas with higher vascular plant %N but did not select nor avoid Cladonia quality. In contrast, caribou selected sites with higher quality Cladonia in winter but responded neutrally to vascular plant quality. C_LIO_LIWe compared seasonal distributions of caribou to determine whether nitrogen consumed in summer and deposited in winter would occur in spatially discrete locations. Population-level kernel density estimates for summer and winter in this island herd were mostly non-overlapping, lending credence to the potential landscape effects of N transport. C_LIO_LIWhen viewed together with established seasonal changes in woodland caribou physiology, sociality, and forage preferences, the shifts in habitat selection and seasonal ranges we observe here could serve as an adaptive strategy for caribou to recycle N and mitigate winter nutrient scarcity. C_LI

Recent progress in mathematical kinship modelling has allowed one to predict the probable numbers of kin for a typical population member. In the models, kin may be structured by age and sex, both in static or time-variant demographies. Knowing the probable numbers of kin in different stages - such as parity, health status, or geographic location - however, remains an open challenge in Kinship Demography. Knowing how population structure delimits kin to distinct stages is an advance - for instance, the probability of having one sister at home and one sister away has different social implications from the probability of having two sisters. We present a novel analytical framework, grounded in branching process theory, that provides kin-number distributions jointly structured by age and stage. Using recursive compositions of probability generating functions (PGFs), we derive the joint age, stage, and age x stage kin-number distributions. All marginal distributions over either dimension naturally emerge. Simple extensions of the PGF approach additionally yield: the joint distribution of an individuals own stage and their kins stage; the probable numbers of kin deaths, both in total and by generation number; and the probabilities of being kinless and/or orphaned. We demonstrate the framework through novel results in an application using UK parity-specific fertility and mortality data. HighlightsO_LIA new method calculates probability generating functions for the number of kin structured by age and stage C_LIO_LIThe model allows predicting the probable numbers of kin organised by age and stage C_LIO_LIRecursive nesting of probability generating functions in branching processes is used C_LIO_LIAn application is presented highlighting the novel results C_LI

Migration is a well-described behavioural strategy that allows species to track variation in resources and climatic conditions by moving in response to seasonality. A common form is elevational migration, an annual short-distance movement undertaken by many mountain bird species globally. While studies show that the timing of migration may relate to food availability, the mechanisms determining which species migrate remain unclear. Our study investigated if the degree of dietary specialization explains why some high-elevation bird species in seasonal environments migrate downslope for the winter while others remain resident at high altitudes despite the apparent scarcity of their preferred food resources. We mist-netted birds along a 2300-m elevational gradient in the Eastern Himalaya and collected blood and faecal samples from 261 individual birds belonging to 18 species of high-elevation residents (ten) and elevational migrants (eight) in their breeding and wintering ranges. Using stable isotope ratios of carbon and nitrogen in whole blood and faecal DNA metabarcoding, we compared their seasonal trophic levels and dietary niches. Nitrogen isotope ratios showed that residents had a substantially lower trophic position in the winter compared to summer (-0.35 [-0.52, -0.17]), whereas migrants had a slightly higher trophic position in the winter (0.15 [-0.02, 0.32]). This trophic shift in residents was likely due to a decrease in insectivory and an increase in frugivory in the winter. The frequency of key insect orders (Lepidoptera, Hemiptera, and Coleoptera) declined by 20-35% in their winter diets alongside an increase in fruit, particularly from the family Polygonaceae (0.33 [0.18, 0.46]). Additionally, compared with residents, migrants showed greater overlap in their dietary niches between summer and winter (98% vs 80%). Because arthropod abundances in the Himalayas peak at high elevations in the summer and decline in the winter, we suggest that elevational migrants are likely dietary specialists that track resources, while high-elevation residents are dietary generalists that supplement their winter diet with fruit and nectar because of the scarcity of arthropods. These findings indicate that a species dietary specialization is linked to its migratory behaviour, providing a potential mechanistic explanation for how different species solve the challenge of seasonal resource limitation.

Habitat partitioning supports the coexistence of sympatric species and shapes their ecological roles across coastal seascapes. Understanding how sympatric species move through and use coastal habitats therefore provides fundamental ecological insight. Aerial drones provide new opportunities to monitor fine-scale movement and habitat utilisation of elasmobranchs in shallow waters. Here, we use drones to investigate fine-scale habitat partitioning and foraging behaviour among stingrays in a coastal lagoon in the central Red Sea. We conducted 30 aerial transect surveys (~17 ha each) and tracked 40 rays and 1 shark (total tracking time > 23 h). Using a double-observer protocol (manual + AI-assisted), 1,468 rays (6 species) and 4 sharks (2 species) were recorded from the transect surveys. Transect detections were dominated by bluespotted ribbontail rays (Taeniura lymma; n = 1,221) and larger-bodied whiprays (predominantly Himantura uarnak; n = 187). AI-assisted image analysis outperformed human analysts detecting 97% of these observations, compared to 76% for human analysts. We found pronounced habitat partitioning at sub-kilometre scales: bluespotted rays occupied the shallowest (< 0.4 m deep) lagoonal areas, away from open water, with foraging-related digging concentrated along the mangrove edge, identifying this zone as a key feeding ground and bioturbation hotspot. Whiprays predominated on macroalgal reef flat habitats and appeared to forage non-disruptively on epifaunal prey. Both taxa aggregated with conspecifics. Together, our results demonstrate that contrasting micro-habitat preferences and foraging strategies structure the spatial ecology of sympatric stingrays and highlight how drone-based monitoring coupled with AI can scale ecological inference in nearshore ecosystems. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=108 SRC="FIGDIR/small/710512v1_ufig1.gif" ALT="Figure 1"> View larger version (63K): org.highwire.dtl.DTLVardef@1cefdd9org.highwire.dtl.DTLVardef@7bc807org.highwire.dtl.DTLVardef@895540org.highwire.dtl.DTLVardef@3c146b_HPS_FORMAT_FIGEXP M_FIG C_FIG

Parental care can shape post-fledging behaviour through provisioning, guidance and social information, yet its absence may alter how young birds establish space use and habitat preferences. We tested the consequences of absent parental care by comparing, hand-reared juvenile herring gulls released without parents with wild, parent-reared conspecifics, focusing on the first two months after fledging. Wild juveniles frequently revisited their natal nest during the first month, whereas hand-reared birds rarely returned to the release site; revisits declined in both groups by the second month but remained more common in wild birds. Wild juveniles used smaller ranges that subsequently expanded, while hand-reared birds began with larger ranges that later contracted, leading to convergence. Contrary to expectation, wild juveniles occurred in areas with higher human population density than hand-reared birds. Habitat use also differed between groups and changed over time. Early on, wild juveniles concentrated activity in anthropogenic and marine habitats, whereas hand-reared birds used rural green habitats more. Later, both groups shifted away from marine areas towards rural green habitats, reducing but not eliminating between-group differences. Short-term survival, did not differ between hand-reared and wild juveniles, indicating that parental care primarily reshaped early space use and habitat choice rather than immediate survival.

O_LIAlthough the use of accelerometer-based behavioural classification to quantify animal activity budgets is gaining widespread traction, the interactions between key preprocessing decisions and modern classification algorithms remain poorly understood. Moreover, classification pipelines are commonly assessed using global performance metrics, despite increasing evidence that such metrics poorly reflect behaviour-specific patterns and ecological reliability. C_LIO_LIUsing a free-ranging primate (Chlorocebus pygerythrus) as a case study, we benchmarked how temporal segmentation (burst length), collar orientation correction, and model architecture jointly shape behavioural inference. We compared nine supervised algorithms spanning classical machine learning, feature-based deep learning including a tabular foundation model (TabPFN), and state of the art time-series architectures (HydraMultiROCKET). Beyond conventional metrics, performance was further evaluated using ecological validation against independent focal observations to assess model stability and biological plausibility. C_LIO_LIModel architecture exerted the strongest influence on classification outcomes. Modern deep-learning approaches substantially outperformed classical models, doubling recall for rare behaviours (e.g., grooming, self-scratching) without compromising precision. In contrast, burst length and collar orientation correction had little effect on global metrics but produced substantial, behaviour-specific trade-offs. Shorter bursts improved the detection of rare events by increasing training instances, while orientation correction suppressed dataset-specific artifacts at the cost of degrading common behaviours. Crucially, models with similar global and behaviour-level validation metrics produced divergent predictions when applied outside the annotated context. C_LIO_LIOur findings reveal that global metrics are insufficient for optimizing behavioural inference in complex wild systems. We demonstrate that modern deep-learning architectures, such as the ROCKET family, provide a robust, accessible baseline that handles class imbalance more effectively than traditional methods. We propose that reliable inference requires behaviour-aware evaluation frameworks that integrate ecological validation, and advocate for ensemble or hierarchical strategies to leverage the complementary strengths of different preprocessing and modelling configurations. C_LI