Global Ecology and Conservation

Managing visitor-wildlife interactions in high-use protected areas requires identifying not only where visitors concentrate but also where visitor use overlaps with wildlife space use. We estimated relative human- macaque encounter potential in Kamikochi, Chubu Sangaku National Park, Japan, where Japanese macaques (Macaca fuscata) and visitors use the same frontcountry landscape. We integrated Mobile Spatial Statistics derived from mobile phone network data, trail-camera counts, and direct-follow location data from macaques along a frontcountry-backcountry gradient. Broad-scale 500-m grid estimates of daytime visitor presence for the 2023 and 2024 open seasons were allocated to 50-m trail sections and adjusted using counts from 10 trail cameras. Automated video counts based on YOLO11 and ByteTrack showed high agreement with manual counts of 100 validation videos (MAE = 0.10 persons/video, MAPE = 1.26%, exact agreement rate = 92.0%). Visitor-use and macaque-use data were converted into kernel density surfaces, and encounter potential was calculated as their normalized product. High encounter potential was concentrated in the central frontcountry: 87.1% of areas with index values [≥] 0.9 fell within the frontcountry and 12.9% within the backcountry. Areas with high visitor use or high macaque use alone did not necessarily show high encounter potential when the two distributions were spatially offset. These results show that spatial overlap provides management information not captured by visitor counts alone. Relative encounter potential can help prioritize patrols, visitor messaging, viewing management, and other interventions in protected areas where wildlife does not consistently avoid heavily used visitor spaces.

Urban protected areas (PAs) have become important habitats for wildlife in rapidly expanding human-dominated landscapes. These PAs may provide refuge and continuously abundant resources, which may enable some adaptable carnivores to persist and attain high abundances. However, little is known about the population ecology and interactions between carnivores sharing such spatially restricted and disturbed urban habitats. We evaluated the densities and spatio-temporal interactions of leopards (Panthera pardus) and striped hyenas (Hyaena hyaena) in Jhalana-Amagarh Conservation Reserve, a protected area (PA) embedded within the metropolis of Jaipur, India. During May-July 2025, we deployed 30 paired camera-traps and used spatially explicit capture-recapture models to estimate density and individual activity centres. We further quantified their diel activity and temporal overlap using kernel density estimation. Leopard density was estimated at 7.03 {+/-} 1.31 individuals/10 km sq and striped hyena density at 4.06 {+/-} 0.96 individuals/10 km sq, with activity centres of both species occurring in close proximity and frequently towards the reserve periphery. Both species were predominantly nocturnal and showed high temporal overlap ({Delta}4 = 0.801, CI=0.73-0.86). Our results report exceptionally high leopard and striped hyena densities in an urban PA and exemplify the importance of small PAs within rapidly expanding cities, stressing the need for finer-scale studies more broadly to understand the mechanisms shaping interactions between the two species. However, such systems require careful management of food subsidies, habitat quality and connectivity along with long-term monitoring of population viability, disease dynamics and human-wildlife conflict.

Monitoring populations is fundamental for conserving threatened carnivores, but can be difficult to sustain in the long term across large landscapes. We present a collaborative framework for monitoring dholes (Cuon alpinus) in India's Western Ghats, combining fecal DNA-based individual identification, Spatially Explicit Capture-Recapture (SECR) models, state Forest Department participation, and trained citizen-volunteers. We surveyed across six sites between 2019-2023, encompassing three Protected Areas (Wayanad, Parambikulam, Periyar) and their adjacent Territorial Divisions. Fecal samples were genotyped using a panel of 96 Single Nucleotide Polymorphisms, and detections were analyzed using SECR models to estimate densities. In Wayanad Sanctuary, densities remained stable across 2019, 2022 and 2023, increasing marginally from 15.67 to 19.01 individuals per 100 sq.km. In 2023, densities varied substantially among six sites, from 9.73/100 sq.km in Parambikulam to 29.23 in Nemmara Territorial Division. Field implementation involved an average of 46 Forest Department personnel co-collecting data per site-year. Among former citizen-volunteers who were surveyed for project feedback (N=29), 83% went on to either work or pursue higher education in wildlife-related or -adjacent fields. We demonstrate that molecular individual identification coupled with SECR can provide a practical basis for dhole population assessments while enabling multi-stakeholder participation in field monitoring. Sustained engagement with wildlife managers and locally recruited volunteers can strengthen institutional capacity, facilitate knowledge exchange, and improve continuity of monitoring programs. We argue that integrating molecular ecology with collaborative governance and democratizing field-science offers a scalable pathway for landscape-level monitoring required to guide dhole conservation.

Reliable estimates of population size are fundamental to understanding wildlife population status and informing conservation. Obtaining such estimates remains challenging for unmarked species, particularly in tropical countries where logistical and financial constraints limit intensive monitoring. The endangered dhole (Cuon alpinus), a highly social canid without natural markings, is one such species for which robust population estimates remain scarce. We estimated dhole density in the Valparai plateau, a human-dominated agroforest-forest mosaic in India's Western Ghats. We compared Spatially Explicit Capture-Recapture (SECR) models based on individual identification from fecal DNA and Space-to-Event (STE) models based on camera-trap detections. We additionally applied both approaches to individually identifiable leopards (Panthera pardus) to cross-validate STE versus SECR estimates. SECR estimated dhole density at 22 individuals per 100 sq.km, with STE estimates varying substantially depending on how the camera viewshed area was defined. STE estimates based on post hoc viewshed measurements, particularly after reducing the viewshed by 10%, were comparable to SECR estimates. In contrast, STE underestimated leopard density relative to SECR (7 versus 12 per 100 sq.km), although uncertainty intervals overlapped. Ecologically, our results indicate that human-use agroforest mosaics can support dhole densities comparable to, or exceeding, those reported from several Protected Areas, highlighting their potential importance for dhole conservation. More broadly, we argue that STE models may offer a practical and relatively inexpensive approach for baseline monitoring of dholes from camera-trap surveys, but their reliability depends strongly on accurate characterization of the effective viewshed and may vary across species and landscapes.

Information on Western bongo (Tragelaphus eurycerus eurycerus) across its range remains scarce and fragmented, leaving the status of populations poorly understood, limiting effective conservation planning. We conducted three camera-trap surveys across Semuliki National Park, Uganda, between October 2019 and August 2020 (90 camera locations; 11,640 camera days), during which we detected the only currently known population of Western bongo in East Africa. We assessed the status of this population using data from the final survey (March-August 2020), when bongos were detected, applying an individual identification system adapted from Eastern bongo (Tragelaphus eurycerus isaaci) and spatially explicit capture-recapture (SECR) modelling to estimate density and abundance. To reflect known aspects of bongo ecology, we incorporated sex and habitat covariates derived from remotely sensed data as predictors of density. Bongos were detected on 17 occasions, and we identified 18 individuals (11 females and 7 males), of which seven immature individuals (5 females and 2 males). The top-ranking SECR model included the tasseled cap greenness index and sex as predictors of density. The model estimated a population size of 56.49 individuals (SE = 7.69; 95% CI = 41.43-71.56), with an estimated density of 0.58 individuals km-2 (SE = 0.171; 95% CI = 0.32-1.09) within Semuliki National Park. These results provide the first quantitative baseline for this population and suggest the presence of an established breeding population, highlighting the need for structured monitoring to support its long-term conservation.

Understanding the spatial patterns of biodiversity and their environmental determinants is fundamental to ecology and conservation, particularly in mountainous transitional zones where species assemblages can shift rapidly. Amphibians, as highly sensitive organisms, are excellent indicators of environmental change, yet quantitative assessments of their diversity in the Huangshan Mountains-a biodiversity hotspot at the junction of the Palaearctic and Oriental realms in southern Anhui, China-remain limited. We conducted systematic line-transect surveys across 200 grids (5 km x 5 km) during the spring and autumn of 2023 and 2024, recording a total of 10,342 individuals belonging to 23 species, 20 genera, 9 families, and 2 orders. Three species (Fejervarya multistriata, Microhyla fissipes, and Bufo gargarizans) were identified as dominant, and two nationally protected species (Hoplobatrachus chinensis and Andrias davidianus) were detected. Inter-annual alpha diversity did not differ significantly, but pronounced seasonal variation was observed, with spring supporting higher Shannon-Wiener and Pielou evenness than autumn in both years. Using generalized additive models and model averaging, we found that annual precipitation and the normalized difference vegetation index (NDVI) were consistently the strongest positive predictors of Shannon-Wiener diversity, Simpson dominance, and species richness, while species richness also declined significantly with increasing human footprint. Total beta diversity (Soerensen dissimilarity) was very high (0.981) and overwhelmingly driven by species turnover (97.86%) rather than nestedness. Partial Mantel tests and distance-based redundancy analysis further revealed that environmental distances-particularly annual precipitation-significantly shaped overall beta diversity and its turnover component after accounting for geographic distance. These results highlight the predominant role of climatic and vegetation gradients in structuring amphibian assemblages in this subtropical mountainous region, providing baseline data to inform local conservation strategies.

Genomic data can assist conservation planning, but implementation of genetic-based recommendations is rare. Closing this "conservation genetics gap" is crucial for restoration and translocation decisions, particularly for highly fragmented ecosystems like pine barrens. We first analyze the conservation genetics gap in pine barren management organizations using a survey. We then analyze whole-genome data to characterize population structure and diversity in the pine barren keystone plant Lupinus perennis. We find little genetic difference among L. perennis populations in the Northeast United States. While populations show signs of isolation by distance (rho = 0.785, p = 0.001), we find no correlation between population size and heterozygosity (Estimate = 0.003, p = 0.357) or inbreeding (Estimate = 0.028, p = 0.416). We also find no effect of seed source population on forecasted effective population size in restoration simulations. These findings and our recommendation that managers consider non-local seed provenancing for L. perennis restorations in the Northeast have been integrated into SWAP policy in two states. We generally recommend practitioners record seed provenance, save genetic material for future testing, and incorporate fitness metrics into restoration monitoring. To effectively close the conservation genetics gap, researchers providing conservation recommendations must consult with practitioners before designing experiments.

Urban and rural woodlands have become important habitats for wildlife, providing opportunities for ecological and ethological adaptation to changing environmental conditions. Due to limited tree resources in anthropogenic woodlands, different species are often forced to interact more closely, and one possible outcome of such interactions is hybridisation. Recent studies have shown that hybrids between the uncommon Syrian Woodpecker (Dendrocopos major) and the often common Great Spotted Woodpecker (Dendrocopos syriacus) are relatively widespread, yet rare, in sympatric populations (Bakai et al. 2026). To understand the factors determining their occurrence in rural and urban landscapes, we conducted an intensive survey across Europe. We examined the effects of environmental factors, particularly woodland characteristics, as well as the co-occurrence of hybrids with the parental species and European Starlings (Sturnus vulgaris) (hereafter Starling). Spatial analyses revealed that the parental species generally have allotopic distributions, with Syrian Woodpeckers (hereafter SW) being more common in scattered woodlands in both rural and urban landscapes, while Great Spotted Woodpeckers (hereafter GW) are more commonly associated with denser woodlands and forests. Although in some areas the two species can co-occur, which provides opportunities for interspecific hybridisation, hybrid offspring showed no clear preference for any particular woodland type. However, they tended to occur in proximity to one of the parental species, usually the species that was locally rarer. Starlings were not identified as a factor influencing hybrid distribution. These results suggest that hybrids have a relatively broad ecological tolerance, potentially facilitated by heterosis. These findings contribute to our understanding of the factors influencing interspecific hybridisation and may have implications for the management of urban and rural woodlands to preserve habitats for the SW, an uncommon and strictly protected species in the west of the European Union.

Management of marine protected areas (MPAs) require long-term monitoring and evaluation that is often limited by the cost and effort of conventional visual based surveys. We evaluated a community science-based eDNA approach to MPA monitoring over two years in the Point Dume State Marine Reserve, the Point Dume State Marine Conservation Area, and two paired unprotected reference sites. Across the sampling period, we detected 119 species of fish and marine mammals, including 19 of the 22 regional management-priority species for Southern CA MPAs. Protected and unprotected comparison sites were broadly similar in overall assemblage with 93 species shared across all sites. Multivariate analyses revealed site-level differences, although season and year explained more variation than protection status. Species presence also shifted during a La-Nina transition, highlighting the importance of long-term monitoring under different climatological contexts. Community scientists greatly expanded the temporal and spatial scope of eDNA sampling and generated high-quality data. Subsampling analysis showed that quarterly sampling, with three sample replicates and sequencing depths between 700k-990k, recovered 95% of observed biodiversity of the full time series. These results demonstrate that a community science-based eDNA program could provide a standardized, scalable, low-cost, and management-relevant approach for long-term monitoring across California's MPA network.

Trail cameras have become ubiquitous tools for ecological data collection over recent decades. Despite progress in the development of automated algorithms and artificial intelligence for image classification, our ability to process large volumes of data remain limited by the need for trained human observers to make refined judgements. We provide guidance on placement of trail cameras for observing Asian elephants (Elephas maximus) and outline a protocol for training and testing naive human observers in performing image classifications (age/sex class and group composition) that cannot yet be automated. This process can be used to develop a high-throughput workflow capable of extracting useful data from large volumes of images. Our training material consisted of 14,007 images collected from 6 trail cameras around Udawalawe National Park in Sri Lanka from 2017-2019. In the first stage, expert observers (n=3) trained a group of inexperienced participants (n=4), who engaged in an iterative process to develop a protocol document. The document was then tested on a second set of subjects (n=6) each of whom classified 350 test images in four separate sequential batches using quantitative measures of precision and accuracy. The test set was sampled from 54,435 images from an additional 25 cameras. When compared to expert observers, they achieved a fair level of precision (Fleiss' kappa = 0.247) and 82.6% accuracy. Our approach can usefully be extended to other species and contexts.

AimSpecies in mountain ecosystems often experience upslope distribution shifts in response to climate change. However, elevational range dynamics exhibit substantial complexity around this general trend, with varying magnitude and direction of shifts observed among studies. We tested whether factors other than temperature, such as habitat size and land use, are also responsible for observed elevational shifts in a cold adapted species. LocationHokkaido, Japan MethodsWe resurveyed 61 historical (1963-2007) occurrence sites covering a wide elevational range (60-2,210 m) within the distribution range of the northern pika (Ochotona hyperborea). We assessed the existence of elevational shifts using quantile regression and the relative importance of temperature, habitat size, and land use on the shift using occupancy analysis. ResultsWe detected presence of the northern pika at 41 sites, suggesting extirpations at 20 sites (32.8%) across a wide elevational range of 60-1,550 m. The concentration of extirpation sites at low to mid elevations resulted in a significant upslope shift of the distribution centroid, although the shift was nonsignificant at lower and upper portions of the range. The occupancy analysis revealed a negative effect of long-term mean of summer maximum temperature and a stronger positive effect of habitat size. ConclusionsThis highlights the susceptibility of the northern pika to heat stress and the importance of larger habitats and thus habitat heterogeneity for persistence of local populations. Given the possibility that the observed shift represents a precursor of elevational contraction, continuous monitoring of the local populations is highly needed to evaluate their long-term viability.

Over the past two decades, significant population recoveries of tigers and leopards have driven a substantial rise in human-carnivore conflict across South Asia. Sustaining the long-term conservation of both species demands effective, adaptive conflict management. However, existing research is largely confined to isolated landscapes or protected areas, hindering the identification of spatial conflict drivers across broader, heterogeneous regions. To address this gap, this study compiled 4,457 human leopard conflict (HLC) and 2,563 human tiger conflict (HTC) records spanning 2000 to 2024 across Bangladesh, Bhutan, India, and Nepal. We evaluated conflict locations against 33 landscape, anthropogenic, climatic, topographic, and hydrological predictors across three spatial scales using binomial generalized linear regression models and three alternative background sampling methods. Model performance peaked at the largest spatial scale for both species and remained consistent across sampling strategies. HTC correlated positively with forest fragmentation and proximity to human settlements, but decreased with higher overall anthropogenic disturbance. Conversely, HLC concentrated in highly productive, high-precipitation, human-modified landscapes. These contrasting spatial patterns highlight distinct ecological responses to habitat modification and carry unique conservation implications. Ultimately, this study underscores the need to integrate carnivore abundance, wild prey densities, and socioeconomic factors to achieve a holistic understanding of conflict dynamics.

Forest restoration has become a widespread response to deforestation and biodiversity loss, particularly across the tropics. Monitoring restoration is essential for tracking progress and requires accurate knowledge of baseline communities, with threatened species often used as recovery indicators. Passive Acoustic Monitoring (PAM) combined with AI classifiers is enabling this at scale. However, such tools remain immature and may underperform for rare, threatened species underrepresented in training data. Here, we demonstrate the potential of PAM assisted by the AI classifier BirdNET+ V3.0, a January 2026 beta release, to support monitoring of threatened tropical birds. By validating 995 detections from 8,940 hours of audio across a reforestation landscape in Para in the Brazilian Amazon, we found that BirdNET+ confirmed the presence of ~43% of the IUCN Red List threatened species possible in the region, five of which were the first confirmed records for the study site. We also found that this method outperformed traditional survey methods. However, BirdNET+ identified species with mixed reliability, and we could also not determine whether restoration has had any impact on them. Our results demonstrate that PAM with BirdNET+ is a viable, practical method for detecting threatened birds in tropical forest restoration landscapes, but that conventional validation approaches are not appropriate for such species, for which extensive validation effort is an essential cost. Under circumstances where this is not possible, we demonstrate validating the single highest-confidence detection per occasion as a practical alternative.

BACKGROUND: The Argentine ant (Linepithema humile) is among the world's worst invasive species. In the Donana Biological Reserve (Spain), it is associated with human structures that provide the mild, moist conditions it needs within the hot, dry scrubland. We evaluated a habitat-modification treatment, predicting that removing this suitable habitat would eradicate the invader locally. Over an 18-year period, we monitored ant communities at a treated and an untreated site, at different distances from the human structures, before and after treatment, and analysed how temperature and worker numbers affected encounters at baits between the Argentine ant and native ants. RESULTS: Distance to the human structures shaped ant communities at both sites, but only the treated site changed after treatment. There, the Argentine ant decreased close to the human structure and almost disappeared farther from it, whereas the endemic Cataglyphis tartessica and C. floricola recovered, especially close to the structure. At baits, the outcome of encounters depended mainly on how temperature changes favoured each species relative to its maximum activity temperature: Cataglyphis spp. displaced the Argentine ant in the morning, as temperature increased, and the Argentine ant recovered the baits in the late afternoon. Numerical advantage favoured only the Argentine ant. CONCLUSION: Habitat modification reduced the Argentine ant and allowed endemic ants to recover, but did not eradicate the invader, which persisted at low abundance 13 years after treatment. In protected areas, habitat management should be combined with eradication, monitoring and prevention of spread.

Extensive deforestation and disturbance have reshaped forest landscapes in West Africa, including areas surrounding Tai National Park, the largest remaining block of Upper Guinean rainforest. Although many degraded areas are currently undergoing regeneration, the capacity of these secondary forests to recover their biodiversity attributes remains insufficiently assessed within this park. This study quantifies the multidimensional restoration of biodiversity across 125 secondary forest plots and 15 primary forest plots representing 14,731 inventoried individuals in four sectors of the park. Using a Bayesian modelling framework, we estimate recovery trajectories of Shannon diversity, floristic composition, functional traits (wood density, leaf mass per area, seed mass), and conservation-relevant species. The different biodiversity attributes recovered at contrasting rates. Diversity was restored more rapidly ({lambda} = 0.06) than floristic composition ({lambda} = 0.03), and these recovery varied according to environmental variables. Among these, the presence of remnant trees showed the highest median on diversity (0.53 {+/-} 0.61) and floristic composition (0.37 {+/-} 0.17), promoting the rapid recovery of both attributes, followed by prior land use, particularly cocoa farming which also positively influenced the recovery of alpha diversity ({lambda} = 0.03) and floristic composition ({lambda} = 0.02). Regarding functional traits, they displayed contrasting dynamics: specific leaf area and seed mass recovered rapidly along successional gradients, whereas wood density followed a more gradual recovery trajectory. From a conservation perspective, although the proportion of IUCN Red List species remained stable along the successional gradient, old-growth indicator species were significantly more abundant in primary forests. To ensure the long-term conservation of biodiversity, protecting both primary and regenerating forests is essential to preserve the ecological resilience of this park.

Climate change will alter habitat suitability for plant communities in mountain ecosystems of the Western Himalaya. There will be contrasting magnitudes and spatial configurations of these changes for different plant species, and it is an important question how to translate these contrasts into effective climate-proof multi-species conservation planning. This study assessed current habitat associations and projected changes in suitable habitat for two contrasting species, Meconopsis aculeata and Rhododendron campanulatum, in Himachal Pradesh, India. Field surveys and spatially filtered occurrence records were combined with MaxEnt modelling using 10 non-collinear environmental predictors. Habitat suitability was projected to 2050 and 2070 under Shared Socioeconomic Pathways; SSP1-2.6, SSP2-4.5 and SSP5-8.5 using HadGEM3-GC31-LL. Field observations showed contrasting habitat associations, with M. aculeata occurring mainly in open alpine herbaceous communities and R. campanulatum in subalpine woody communities. Elevation was an important predictor for both species while climatic predictors contributed differently to their relative importance and response patterns. Future projections indicated reductions in suitable habitat of approximately 25% for M. aculeata and 27-28% for R. campanulatum. Habitat-change maps showed heterogeneous patterns of loss, gain and persistence rather than uniform contraction, and centroid displacement remained small (7.7-8.5 km). We performed proximity analyses to show which currently suitable areas are most close to areas of future gain, in order to support the design of climate-proof protected areas. We discuss how interspecific differences in response are relevant to multi-species conservation planning that considers the spatial relationship between current high-suitability areas and the gain, loss and redistribution of future suitable habitat.

The tidal wetlands of Louisiana comprise about 25% of those found throughout the conterminous United States yet estimates of wetland loss rates in the region between 1932 and 2016 have exceeded 60 km2 yr-1. To mitigate further degradation and wetland loss in the region, a globally unprecedented $50B, 50-year plan for coastal Louisiana is driving restoration efforts, and demand exists from multiple stakeholders for regularly updated, regional-scale, accurate land cover information. We used machine learning (random forests; RF) and cloud computing to develop a new Landsat-based, marsh vegetation community geospatial dataset. The dataset depicts wetland vegetation community types defined in a previous study at annual (1985-2025) time steps at 30-m resolution. An RF algorithm was used to integrate training samples with feature variables derived from Landsat imagery, and the resulting geospatial data product achieved an overall correct classification rate of 78%. The approach for development of the land cover dataset presented here has potential for application in other coastal wetland habitats throughout the world.

1. Assessing the conservation value of secondary forests and reforestation is important for conservation planning. Butterflies are widely used as indicators, and fruit-feeding butterflies are often used in biodiversity assessments in tropical landscapes. 2. We examined fruit-feeding butterfly diversity, abundance, and community composition across a primary forest, a secondary forest, and a plantation forest in northern Vietnam. We sampled butterflies nearly monthly during one year using baited traps. 3. Primary and secondary forest supported higher species richness than plantation forest. Moreover, while secondary forest had mainly a subset of species found in primary forest, the plantation forest had a distinct butterfly fauna. 4. Butterfly abundance and community composition varied across months, without showing a clear seasonal pattern. 5. Our results suggest that secondary and plantation forests do not fully replace primary forests, but they contribute meaningfully to butterfly conservation in human-modified tropical landscapes.

Understanding the spatial distribution of species and their patterns of endemism is necessary for establishing effective conservation priorities. Despite the vital pollination services bees provide, Californias native bee distribution patterns remain largely unexplored relative to plants and butterflies. We analyzed bee species richness and endemism across California and their concordance with plant distributions. Richness was high across areas of the California Floristic Province, including the Sierra Nevada, San Francisco Bay Area and Central Coast, South Coast Ranges, and the Transverse and Peninsular ranges. Bee endemism was more localized, concentrated in the San Joaquin Valley, eastern Sierra Nevada and adjacent Great Basin, Sierra Nevada foothills, and California deserts. Because richness and endemism appear to operate at different spatial scales and likely respond to different environmental drivers, effective conservation strategies must address both. Additionally, conservation plans that incorporate both plant and bee diversity are needed to achieve more comprehensive biodiversity protection.

Objective: This study aimed to investigate morphometric and plumage-colour variation among native chickens from Wallacea and Java, Indonesia, and to evaluate the contributions of geographic and production-system factors to phenotypic diversity. Methods: Native chickens were sampled from 259 households across eight provinces in Indonesia. Household-level information on production systems and farmer selection practices was collected through structured surveys. Morphometric traits and HSV-derived plumage-colour variables were analysed using principal component analysis (PCA), permutational multivariate analysis of variance (PERMANOVA), and hierarchical variance partitioning to assess phenotypic variation across locations, provinces, and islands. Results: Substantial phenotypic variation was observed among sampled populations, with significant province- and island-level differences detected for several morphometric traits. However, extensive overlap among locations was evident in both morphometric and plumage-colour space, indicating diffuse rather than discrete structuring. Within-location variation accounted for the largest proportion of total phenotypic variance across most traits, consistent with decentralised smallholder management and continued human-mediated exchange. Conclusion: Geographic fragmentation and agroecological heterogeneity alone appear insufficient to generate strongly differentiated phenotypic groups when connectivity among populations is maintained. Whether this phenotypic connectivity corresponds to contemporary gene flow, retained ancestral variation, or convergent environmental responses remains to be tested using genomic data.