Abstract Estimating quantitative genetic and phenotypic (co)variances is crucial for investigating evolutionary ecological phenomena such as developmental integration, life history trade‐offs and niche specialization, as well as for describing selection and predicting multivariate evolution in the wild. While most studies assume (co)variances are fixed over short timescales, environmental heterogeneity can rapidly modify the variation of and associations among organisms' traits. Here I extend prior multilevel regression models for quantitative genetic inference (so‐called animal models) to develop a novel covariance reaction norm (CRN) model, which can be used to detect how trait (co)variances respond to continuous, multivariate, and potentially nonlinear environmental change, even in the absence of repeated individual measurements or experimental breeding designs. After introducing the CRN model, I use simulations to validate its implementation for Bayesian inference in Stan, as well as to compare its performance to standard character state and random regression approaches. Findings demonstrate superior accuracy and power for detecting environmental effects on genetic covariance with modest sample sizes. I then apply the CRN model to long‐term field data on cooperation among meerkats (Suricata suricatta). I find nonlinear effects of group size on the genetic (co)variances of cooperative behaviours, leading to increased social niche specialization among foraging and pup feeding versus babysitting tasks in larger groups. Multivariate gene‐by‐environment interactions are also observed in response to age, sex and dominance status. R code and a tutorial are provided to aid empiricists in applying CRN models to their own datasets.

Abstract This work reviews and discusses how a deeper knowledge of mosquito time‐allocation, particularly resting periods, can enhance the effectiveness and deployment of classical vector control tools. As more high‐resolution data are gathered by novel tracking devices, follow‐up questions investigating mosquito behaviour are likely to arise, requiring the development of comprehensive frameworks to properly collect, summarize and analyse them. We demonstrate how novel stereoscopic 3D mapping tools can generate unique data and discuss available metrics to adequately quantify mosquito behaviour particularly under (semi‐) field conditions. Incorporation of mosquito 3D tracking with ecological, evolutionary and quantitative genetics approaches would enable a deeper understanding of the physiological and genetic basis driving resting behaviours. The ability to better determine when, where and how mosquitoes are allocating resting periods might constitute a major input to the advancement of control strategies.

Abstract Collection of large behavioural data‐sets on wild animals in natural habitats is vital in ecology and evolution studies. Recent progress in machine learning and computer vision, combined with inexpensive microcomputers, has unlocked a new frontier of fine‐scale markerless measurements. Here, we leverage these advancements to develop a 3D Synchronized Outdoor Camera System (3D‐SOCS): an inexpensive, mobile and automated method for collecting behavioural data on wild animals using synchronized video frames from Raspberry Pi controlled cameras. Accuracy tests demonstrate 3D‐SOCS' markerless tracking can estimate postures with a 3 mm tolerance. To illustrate its research potential, we place 3D‐SOCS in the field and conduct a stimulus presentation experiment. We estimate 3D postures and trajectories for multiple individuals of different bird species, and use this data to characterize the visual field configuration of wild great tits ( Parus major ), a model species in behavioural ecology. We find their optic axes at ~±60° azimuth and −5° elevation. Furthermore, birds exhibit functional lateralization in their use of the right eye with conspecific stimulus, and show individual differences in lateralization. We also show that birds' convex hulls predicts body weight, highlighting 3D‐SOCS' potential for non‐invasive population monitoring. 3D‐SOCS is a first‐of‐its‐kind camera system for wild research, presenting exciting potential to measure fine‐scaled behaviour and morphology in wild birds.

Cover image:The cover image features a male Marenestha inconspicua, a tiny bush-cricket endemic to central Chile.Although the most common morph in this species is green, this yellow specimen seized its coloration as camouflage while feeding on Cestrum flowers, showcasing how insects like this can play important roles in pollination.This species was one of the inspirations behind Rthoptera, a Shiny-powered R package for insect bioacoustics.Although the field of bioacoustics began with studies on cricket song nearly a century ago, today researchers still lack a comprehensive tool for standardized analysis.Thus, many rely on multi-software approaches that complicate analyses and figures.To solve this gap, Rthoptera allows researchers with any level of experience in R to easily obtain accurate measurements and standard plots from focalized insect signals, especially for those produced by species in the order Orthoptera.We hope that this tool will encourage more researchers to include bioacoustic analysis when describing new species or documenting taxonomic group revisions.

Abstract Toothed whales, despite being one of the most acoustically specialized lineages of vertebrates, lack detailed vocal repertoire characterizations comparable to those of songbirds and primates. Current descriptions often lack standardization, limiting interspecies and intraspecies comparisons critical for understanding the contribution of phylogenetic, social and environmental drivers of vocal repertoire evolution. This study reviews six bioacoustics software packages (Luscinia, Beluga, ARTwarp, DeepSqueak, PAMGuard and SASLab) using machine learning for toothed whale whistle detection, extraction of whistle fundamental frequency contours and categorization. We applied these tools to acoustic recordings from four dolphin species to create a primer with guidelines and scripts for those interested in replicating the analyses. The results show that manual (Luscinia) and semi‐automated (Beluga, DeepSqueak) detection and extraction of whistles yielded a greater number of usable contours than the fully automated approaches (PAMGuard and SASLab). The integration of hard and fuzzy unsupervised machine learning analyses provided a multifaceted characterization of whistle repertoires based on identified patterns and structures. The hard approaches from ARTwarp and DeepSqueak were used to categorize signals based on contour similarity. Both packages produced similar numbers of categories. DeepSqueak was faster in categorizing contours, but ARTwarp is more flexible in how similarity thresholds are set. We used these categories to demonstrate estimating relative whistle repertoire size using three methods—Coupon Collector, Capture–Recapture and Hill Numbers. While the methods differ in their assumptions, there was consensus in the ranking of species based on repertoire size. We favour the application of Hill numbers to measure repertoire size and compositional similarity, as this method addresses uneven sampling. The fuzzy approach used the frequency and temporal features extracted from the traced contours in Luscinia to evaluate repertoire similarity in acoustic structure using the package's built‐in multivariate tools. This primer equips researchers to generate more detailed toothed whale repertoire characterizations by integrating machine learning tools with mathematical and community ecology approaches. Supplementary material is provided to support researchers with varying levels of experience in implementing these techniques and indices in their studies.

Abstract The use of drones to survey and monitor wildlife populations has increased exponentially. A common protocol used for data collection is planning flights with substantial overlap between successive photographs and lateral lines and then creating orthomosaics by merging the collected images. Because available methods for orthomosaic building assume that landscapes are static, unintended errors arise when counting moving animals. Here, we describe these sources of error and discuss potential solutions and future developments needed. Individuals can appear multiple times, be omitted or appear as faint ghosts or cut in half in the final mosaic. These errors can significantly impact abundance estimates but are rarely acknowledged. Researchers should carefully consider if using orthomosaics is really needed for surveying wildlife. Currently, there is a lack of methods to prevent these errors from arising and to explicitly accommodate them in modelling approaches. Future developments should focus on (a) creating methods to build orthomosaics that minimize these errors in the context of counting moving animals; (b) developing modelling approaches to estimate abundance while accounting for these errors; and (c) exploring alternative flight settings (e.g. amount of lateral overlap, sensor type, flight height and speed). Using an example on Giant Amazon Turtles, we illustrate potential solutions with a method for orthomosaic building that prioritizes moving animals and a modelling approach to estimate the detection errors and correct abundance estimates. The developed prototype approach for creating orthomosaics revealed many more turtle individuals than the conventional approach, although it presented more double counts as well. In the modelling approach, we found that a turtle available for detection during the survey can have a probability of 31% of being omitted or ghosted during the conventional orthomosaic building process. We also found that 12% of the turtles appearing in a conventional orthomosaic correspond to double counts.

Abstract Drones are an effective tool for animal surveys, capable of generating an abundance of high‐quality ecological data. However, the large volume of ecological data generated introduces an additional problem of the requisite human resources to process and analyse such data. Deep learning models offer a solution to this challenge, capable of autonomously processing drone footage to detect animals with higher fidelity and lower latency when compared with humans. This work aimed to develop an animal detection architecture that classifies animals in accordance to their location (terrestrial vs. arboreal). The model incorporates human pilot inspired techniques for greater performance and consistency across time. Thermal drone footage across the state of New South Wales, Australia from surveys over a 2+ year period was used to construct a diverse training and validation dataset. A high‐resolution 3D simulation was developed to workload by autonomously generating labelled data to supplement manually labelled field data. The model was evaluated on 130 hours of thermal imagery (14 million images) containing 57 unique animal species where 1637 out of 1719 (95.23%) of human pilot recorded animals were detected. The model achieved an F1 score of 0.9410, a 4.36 percentage point increase in performance over a benchmark YOLOv8 model. Simulated data improved model performance by 1.7x for low data scenarios, lowering data labelling costs due to higher quality image pre‐labels. The proposed animal detection model demonstrates strong reporting accuracy in the detection and tracking of animals. The approach enables widespread adoption of drone‐capturing technology by providing in‐field real‐time assistance, allowing novice pilots to detect animals at the level of experienced pilots, whilst also reducing the burden of report generation and data labelling costs.

Abstract Colonies of pure bacterial strains are highly dense cellular structures organised in distinct and typical arrangements. The size, shape and variability of bacterial colonies are strongly species‐dependent and also influenced by environmental conditions. However, the spatial organization of individual cells is unknown for most strains. By introducing a local biopsy technique, this study aimed to provide a means to study the local diversification of the structure of bacterial colonies. Cells were biopsied at different sites of a colony and analysed by microbial flow cytometry using cytometric fingerprints for both fixed and viable cells, which divided the biopsied samples into many heterogeneous cell states. This two‐step resolution was performed on five bacterial strains: Bacillus subtilis, Paenibacillus polymyxa, Kocuria rhizophila, Stenotrophomonas rhizophila and Pseudomonas citronellolis. The effects of biopsy tool size (27G needle and 10, 200, and 1000 μL pipette tips) and sampling site on the precision of the technique were tested by using both gate settings along Gaussian distributions of subpopulations and a grid gating tool, as well as the t‐distributed stochastic neighbour embedding (t‐SNE) method. The biopsy technique uncovered significant heterogeneity among the cells within bacterial colonies, identifying differences in cell cycle stages, the proportion of live and dead cells, and the abundance of spore types. Cells from different biopsy sites displayed distinct physiological states, revealing that colony structure is far more complex than previously understood. The technique's precision depends on the biopsy tool size, dye equilibration, and cell handling, underscoring the importance of method calibration. The biopsy method, combined with cytometric fingerprinting, provided insights within only 15–45 min and is universally applicable. The study provides a high‐resolution biopsy technique that explores the spatial distribution of cell types and their heterogeneous physiological cell states, allowing conclusions to be drawn from biopsy composition at different sites to overarching functions of the entire bacterial colony. This method also facilitates downstream analysis through further cell sorting, offering a powerful approach for future functional investigations.

Abstract Environmental DNA (eDNA) metabarcoding has revolutionized biodiversity monitoring, offering non‐invasive tools to assess ecosystem health. The complexity of eDNA metabarcoding data poses major challenges for conventional ordination methods in understanding assemblage similarities and assessing biodiversity patterns. Here, we introduce ORDNA (ORDination via Deep Neural Algorithm), a new deep learning method tailored for eDNA sample ordination. Leveraging artificial neural networks, ORDNA processes raw sequences from eDNA samples directly, bypassing potentially biased and cumbersome expert‐based bioinformatic steps. The method is trained with a contrastive self‐supervised learning approach, the triplet loss, to derive a two‐dimensional representation of eDNA samples based on their read composition. We apply ORDNA to four distinct eDNA datasets, demonstrating its robustness and superiority over traditional ordination techniques in capturing and visualizing ecological patterns. Our results underline the potential of deep learning in advancing eDNA analysis, with ORDNA serving as a promising tool for more accurate and efficient biodiversity assessments.