Biology subjects

Jones, K. E.

Publications and source records attributed to Jones, K. E..

At least 19 recordsLinked to original sources

bioRxiv2026

Physiological robustness of MScanFit to simulated motor unit loss and remodelling

Introduction: MScanFit estimates motor unit number and size from compound muscle action potential (CMAP) scans, but its accuracy may depend on the physiological processes that shape motor unit loss and remodelling. We reproduced the general dynamic remodelling framework used in the original MScanFit simulation and expanded it to test MScanFit robustness across physiologically heterogeneous conditions. Methods: Generative computational models simulated CMAP scans from motor unit pools undergoing progressive loss from 160 to 5 surviving motor units. Twelve remodelling conditions combined denervation pattern (random or selective), reinnervation method (random, distributive, or size-weighted), and neuromuscular resilience (20% or 60%); two additional conditions modelled denervation without collateral reinnervation. MScanFit estimates were compared with the known motor unit numbers and mean motor unit sizes used to generate each scan. Results: MScanFit reproduced the principal behaviour reported in the original simulation and generally tracked progressive motor unit loss across heterogeneous remodelling conditions. However, motor unit number estimation (MUNE) error changed systematically with remodelling physiology. Selective denervation and distributive reinnervation increased error at several intermediate stages of motor unit loss, whereas 60% resilience produced greater error than 20% resilience at every stage after remodelling began. Motor unit size estimates also tracked the underlying increase in mean unit size, but their accuracy was more strongly affected by remodelling, particularly with 60% resilience and advanced motor unit loss. Conclusion: MScanFit motor unit number estimates were broadly robust to substantial motor unit loss and neuromuscular remodelling but were not physiologically invariant. Denervation and collateral reinnervation systematically influenced the magnitude and direction of estimation error, indicating that variation in MScanFit performance can arise from differences in the underlying motor unit population. These findings support MScanFit as a robust measure of motor unit loss across heterogeneous neuromuscular phenotypes and show that some of its estimation variability has a physiological basis.

neuroscience↗

bioRxiv2025

Accessible AI Enhances Monitoring of Coral Seeding Devices in Reef Restoration

1. Coral seeding devices (CSDs) - tools designed to deliver sexually propagated corals to target locations - offer a promising means to increase coral abundance on degraded reefs. However, evaluating CSD effectiveness for coral reef restoration across wide areas and over many years is limited by the lack of robust and efficient monitoring approaches. Large-area reef imagery offers an attractive potential solution, but manual detection of CSDs within imagery is slow and limits the scalability of CSD monitoring. 2. We investigated whether machine learning classifiers could accurately detect CSDs in reef orthoimages and tested the performance of classifiers created following minimal manual annotation effort. Using freely available software, we first evaluated classifier performance in a single-site experiment using an orthoimage containing 989 CSDs deployed in Palau. We then also evaluated performance in a multi-site experiment using orthoimages containing a different CSD design deployed across seven sites in the central Great Barrier Reef, Australia. 3. In Experiment 1, classifiers trained on just 30 CSDs annotated within 10 minutes achieved mean recall and precision of 96.7% and 97.1% respectively, reducing manual annotation time by 95.6% whilst still detecting 98.8% of the number of devices found manually. Larger training sets yielded less reliable classifiers and required more manual effort. In Experiment 2, classifiers trained on 30 CSD annotations from one site performed excellently across seven orthoimages from multiple reefs, achieving mean recall and precision of 99.2% and 93.3%. 4. We present evidence that CSD classifiers can be highly effective across both single- and multi-site CSD deployments. In using a free and user-friendly software, we also demonstrate their accessibility to reef restoration practitioners. To facilitate wider uptake of CSD monitoring, we provide a step-by-step protocol for implementing CSD classifiers. By improving access to efficient, direct assessment of intervention outcomes, this method can play a vital role in guiding the enhancement of approaches aiming to restore coral reefs.

ecology↗

bioRxiv2025

Evidence of ecosystem process recovery across a large-scale coral reef restoration programme using AI accelerated soundscape analysis

1.Coral reef restoration efforts are on the increase globally. However, reporting on the ecological outcomes of these efforts is rare and typically focuses on coral related metrics. As a result, understanding of whether restoration can recover broader aspects of reef functioning remains limited. In this study we use passive acoustic monitoring coupled with human-in-the-loop artificial intelligence to analyse >12 months of soundscape recordings from 45 sites across five biogeographically independent regions to investigate the impact of active restoration on reef functioning. We trained and rigorously evaluated machine learning models to identify 34 biological sound types within this data, generating >912,000 high-confidence detections. These detections were used to infer four key functions across healthy, degraded, early-stage (<3 months) and mid-stage (32-53 months) restored reefs. Restoration significantly enhanced: (i) biological sounds at night, key to recruiting juvenile fish; (ii) diversity of biological sounds, an indicator of fish community diversity; and (iii) snapping shrimp activity, an indicator of bioturbation. However, effects varied by region, and audible parrotfish grazing, key to algal control and bioerosion, did not differ among habitat types in four of the five regions. Our findings provide evidence that restoration can support recovery of broader ecosystem functioning when carefully implemented in the right contexts. Graphical abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=174 SRC="FIGDIR/small/678197v4_ufig1.gif" ALT="Figure 1"> View larger version (48K): org.highwire.dtl.DTLVardef@7d5462org.highwire.dtl.DTLVardef@2efa7dorg.highwire.dtl.DTLVardef@3f4b9eorg.highwire.dtl.DTLVardef@17d829c_HPS_FORMAT_FIGEXP M_FIG C_FIG

ecology↗

bioRxiv2025

Ecological inference from isolated vertebrae: Evaluating functional signal across the carnivoran spine

Understanding the ecological adaptations of extinct species is a central goal in vertebrate palaeontology, but is often limited by the incomplete nature of the fossil record. While skulls and limb bones have traditionally been emphasized in functional and ecological reconstructions, vertebrae are frequently overlooked. While isolated vertebrae are among the most commonly preserved postcranial elements, they are rarely found as complete vertebral columns, raising the question of whether isolated elements alone can yield meaningful ecological information. In this study, we assess the potential of vertebral morphology to predict two key ecological traits, running speed and hunting mode, using three-dimensional geometric morphometrics across ten presacral vertebrae from a broad sample of extant carnivorans. We evaluate the predictive power of individual vertebrae, regional groupings (cervical, thoracic, lumbar), and multi-element combinations. Our results show that certain vertebrae retain strong ecological signals on their own, especially the first thoracic and lumbar elements. However, combining multiple vertebrae often dilutes ecological signal, likely due to their differing functional roles along the axial column. This highlights the importance of treating vertebral regions independently and suggests that single, strategically informative vertebrae may outperform multi-element approaches in some contexts. We apply this framework to the extinct dire wolf (Canis dirus) and find contrasting signals along the spine, the first thoracic and lumbar vertebrae suggest adaptations for faster locomotion, while some cervical vertebrae indicate an intermediate running speed. This mosaic supports the idea that C. dirus occupied a complex ecological niche involving both active predation and scavenging. These findings underscore the power of vertebral morphology for ecological inference in fossil taxa, particularly when remains are fragmentary, and argue for a more nuanced use of isolated axial elements in reconstructing extinct carnivoran behaviour.

paleontology↗

bioRxiv2025

The terrestrial-aquatic transition impacts endocranial shape in caniform carnivorans

Mammals exhibit remarkable diversity in brain size and morphology, shaped by numerous ecological radiations throughout the Cenozoic. Although previous studies have demonstrated the influence of phylogeny, allometry, and various ecological variables on endocranial morphology in certain mammalian lineages, the extent to which locomotor behaviour influences brain shape evolution remains unclear. The suborder Caniformia, or 'dog-like' carnivorans, is an ecologically diverse clade, containing terrestrial, arboreal, fossorial, and semiaquatic species, each facing a unique set of sensory and motor challenges. Here, we evaluate the impact of phylogeny, allometry, and locomotor mode on brain shape evolution in caniforms. We examine endocranial morphology in 73 species using high-density 3D geometric morphometrics, principal component analysis and phylogenetic comparative methods. Our findings indicate that the terrestrial-to-aquatic transition significantly impacts endocranial shape across caniforms, which is associated with maximum dive duration. This effect is more pronounced in pinnipeds than in other semiaquatic fissipeds, such as mustelids and ursids, likely due to their more derived ecology and greater commitment to aquatic life. Aquatic caniforms tend to exhibit expanded neocortices alongside reduced paleocortices and olfactory bulbs. These morphological changes likely reflect selection pressures acting on both the brain and the surrounding cranial architecture in aquatic environments.

evolutionary biology↗

bioRxiv2025

Contact networks of small mammals highlight potential transmission foci of Lassa mammarenavirus.

Lassa fever (Mammarenavirus lassaense; LASV), is an endemic zoonosis in West Africa. Human infections arise from rodent-to-human transmission, mainly from the synanthropic reservoir Mastomys natalensis. In Sierra Leone, small-mammal communities vary across land use gradients, shaping LASV transmission risk. How anthropogenic environments facilitate the rodent-rodent interactions remains poorly understood. We sampled small mammals over 43,266 trap nights in Sierra Leones LASV-endemic Eastern Province, detecting 684 rodents and shrews. To assess potential transmission, we constructed space-sharing networks from co-trapping events within species-specific home range radii. These networks approximated shared space use, allowing comparison of encounter patterns across habitats. Network topology varied significantly by land use. Village networks were the most connected (highest average degree), whereas agricultural communities supported the most species (higher rarefied richness) and were the most fragmented (higher modularity). Notably, the probability of intraspecific space sharing among M. natalensis was highest in agriculture, suggesting land use modulates key intra-specific transmission pathways. LASV seroprevalence was 5.7% community-wide, with antibodies in nine species. We found no statistically significant association between overall seroprevalence and land use or aggregate network structure (mean degree). However, predictive modeling for M. natalensis indicated that higher individual degree is associated with seropositivity, suggesting complex, scale-dependent relationships. These findings show that simple ecological drivers do not fully explain LASV exposure, highlighting the importance of species-specific behaviors (i.e., M. natalensis clustering in agriculture) and the multi-host serological landscape in assessing transmission risk.

ecology↗

bioRxiv2024

Repatterning of mammalian backbone regionalization in cetaceans

The reinvasion of the aquatic realm by cetaceans is one of the most iconic ecological transitions that led to drastic modifications of the mammalian body plan, especially the axial skeleton. Relative to the vertebral column of other mammals that is subdivided into numerous anatomical regions, regional boundaries of the cetacean backbone appear obscured. Whether the traditional mammalian regions are present in cetaceans but hard to detect due to anatomical homogenization or if regions have been entirely repatterned remains unresolved. Here we combine a segmented linear regression approach with spectral clustering to quantitatively investigate the number, position, and homology of vertebral regions across species from all major cetacean clades. We propose the new "Nested Regions" hypothesis under which the cetacean backbone is composed of six homologous modules subdivided into six to nine post-cervical regions, with the degree of regionalization dependent on vertebral count and ecology. Compared to terrestrial mammals, the cetacean backbone is less regionalized in the precaudal segment but more regionalized in the caudal segment, indicating repatterning of the vertebral column associated with the transition from limb-powered to axial-driven locomotion.

evolutionary biology↗

bioRxiv2024

Unlocking the soundscape of coral reefs with artificial intelligence

Passive acoustic monitoring can offer insights into the state of coral reef ecosystems at low-costs and over extended temporal periods. Comparison of whole soundscape properties can rapidly deliver broad insights from acoustic data, in contrast to the more detailed but time-consuming analysis of individual bioacoustic signals. However, a lack of effective automated analysis for whole soundscape data has impeded progress in this field. Here, we show that machine learning (ML) can be used to unlock greater insights from reef soundscapes. We showcase this on a diverse set of tasks using three biogeographically independent datasets, each containing fish community, coral cover or depth zone classes. We show supervised learning can be used to train models that can identify ecological classes and individual sites from whole soundscapes. However, we report unsupervised clustering achieves this whilst providing a more detailed understanding of ecological and site groupings within soundscape data. We also compare three different approaches for extracting feature embeddings from soundscape recordings for input into ML algorithms: acoustic indices commonly used by soundscape ecologists, a pretrained convolutional neural network (P-CNN) trained on 5.2m hrs of YouTube audio and a CNN trained on individual datasets (T-CNN). Although the T-CNN performs marginally better across the datasets, we reveal that the P-CNN is a powerful tool for identifying marine soundscape ecologists due to its strong performance, low computational cost and significantly improved performance over acoustic indices. Our findings have implications for soundscape ecology in any habitat. Author SummaryArtificial intelligence has the potential to revolutionise bioacoustic monitoring of coral reefs. So far, a limited set of work has used machine learning to train detectors for specific sounds such as individual fish species. However, building detectors is a time-consuming process that involves manually annotating large amounts of audio followed by complicated model training, this must then be repeated all over again for any new dataset. Instead, we explore machine learning techniques for whole soundscape analysis, which compares the acoustic properties of raw recordings from the entire habitat. We identify multiple machine learning methods for whole soundscape analysis and rigorously test these using datasets from Indonesia, Australia and French Polynesia. Our key findings show use of a neural network pretrained on 5.2m hours of unrelated YouTube audio offers a powerful tool to produce compressed representations of reef audio data, conserving the datas key properties whilst being executable on a standard personal laptop. These representations can then be used to explore patterns in reef soundscapes using "unsupervised machine learning", which is effective at grouping similar recordings periods together and dissimilar periods apart. We show these groupings hold relationships with ground truth ecological data, including coral coverage, the fish community and depth.

ecology↗

bioRxiv2023

Biodiversity shows unique responses to land-use change across regional biomes

Biogeography has a critical influence on how ecological communities respond to threats and how effective conservation interventions are designed. For example, the resilience of ecological communities is linked to environmental and climatic features, and the nature of threats impacting ecosystems also varies geographically. Understanding community-level threat responses may be most accurate at fine spatial scales, however collecting detailed ecological data at such a high resolution would be prohibitively resource intensive. In this study, we aim to find the spatial scale that could best capture variation in community-level threat responses whilst keeping data collection requirements feasible. Using a database of biodiversity records with extensive global coverage, we modelled species richness and total abundance (the responses) across land-use types (reflecting threats), considering three different spatial scales: biomes, biogeographical realms, and regional biomes (the interaction between realm and biome). We then modelled data from three highly sampled biomes separately to ask how responses to threat differ between regional biomes and taxonomic group. We found strong support for regional biomes in explaining variation in species richness and total abundance compared to biomes or realms alone. Our biome case studies demonstrate that there is a high variation in magnitude and direction of threat responses across both regional biomes and taxonomic group, but all groups in tropical forest showed a consistently negative response, whilst many taxon-regional biome groups showed no clear response to threat in temperate forest and tropical grassland. Our results suggest that the taxon-regional biome unit has potential as a reasonable spatial and ecological scale for understanding how ecological communities respond to threats and designing effective conservation interventions to bend the curve on biodiversity loss.

ecology↗

bioRxiv2022

Rodent trapping studies as an overlooked information source for understanding endemic and novel zoonotic spillover.

1.Rodents a diverse, globally distributed and ecologically important order of mammals are nevertheless important reservoirs of known and novel zoonotic pathogens. Ongoing anthropogenic land use change is altering the composition of these species abundance and distribution which among zoonotic host species may alter the risk of zoonoses spillover events. These changes demand that a better understanding of the current distribution of rodent species is vital for mitigating potential zoonotic disease hazard and managing risk. However, available species distribution and host-pathogen association datasets (e.g. IUCN, GBIF, CLOVER) are often taxonomically and spatially biased. Here, we synthesise data from West Africa from 127 rodent trapping studies, published between 1964-2022, as an additional source of information to characterise the range and presence of rodent species and identify the subgroup of species that are potential or known pathogen hosts. We identify that these rodent trapping studies, although biased towards human dominated landscapes across West Africa, can usefully complement current rodent species distribution datasets and we calculate the discrepancies between these datasets. For five regionally important zoonotic pathogens (Arenaviridae spp., Borrelia spp., Lassa mammarenavirus, Leptospira spp. and Toxoplasma gondii), we identify host-pathogen associations that have not been previously reported in host-association datasets. Finally, for these five pathogen groups, we find that the proportion of a rodent hosts range that has been sampled remains small with geographic clustering. A priority should be to sample rodent hosts across a greater geographic range to better characterise current and future risk of zoonotic spillover events. In the interim, studies of spatial pathogen risk informed by rodent distributions must incorporate a measure of the current sampling biases. The current synthesis of contextually rich rodent trapping data enriches available information from IUCN, GBIF and CLOVER which can support a more complete understanding of the hazard of zoonotic spillover events. 2. Author SummaryEmerging and endemic zoonotic diseases are projected to have increasing health impacts, particularly under changing climate and land-use scenarios. Rodents, an ecologically vital order of mammals carry a disproportionate number of zoonotic pathogens and are abundant across West Africa. Prior modelling studies rely on large, consolidated data sources which do not incorporate high resolution spatial and temporal data from rodent trapping studies. Here, we synthesise these studies to quantify the bias in the sampling of rodent hosts and their pathogens across West Africa. We find that rodent trapping studies are complementary to these datasets and can provide additional, high-resolution data on the distribution of hosts and their pathogens. Further, rodent trapping studies have identified additional potential host-pathogen associations than are recorded in consolidated host-pathogen association datasets. This can help to understand the risk of zoonotic diseases based on host distributions. Finally, we quantify the current extent of known rodent presence and pathogen sampling within a species range, highlighting that current knowledge is limited across much of the region. We hope that this will support work to study rodent hosts and their pathogens in currently under sampled regions to better understand the risk of emerging and endemic zoonoses in West Africa.

ecology↗

bioRxiv2021

Characterization of cognitive-motor function in women who have experienced intimate partner violence-related brain injury

Intimate partner violence (IPV) affects at least 1 in 3 women worldwide and up to 92% report symptoms consistent with brain injury (BI). Although a handful of studies have examined different aspects of brain structure and function in this population, none has characterized potential deficits in cognitive-motor function. This knowledge gap was addressed in the current study by having participants who had experienced IPV complete the bimanual Object Hit & Avoid (OHA) task on a Kinarm End-Point Lab. BI load, as assessed by the Brain Injury Severity Assessment (BISA) tool as well as measures of comorbidities (PTSD, anxiety, depression, substance use, history of abuse) were also collected. Results demonstrated BI load accounted for a significant amount of variability in the number of targets hit and average hand speed. PTSD, anxiety, and depression also contributed significantly to the variability in these measures as well as to the number and proportion of distractor hits, and the object processing rate. Taken together, these findings suggest IPV-related BI, as well as comorbid PTSD, anxiety, and depression, disrupt the processing required to quickly and accurately hit targets while avoiding distractors. This pattern of results reflects the complex interaction between the physical injuries induced by the episodes of IPV and the resulting impacts these experiences have on mental health.

neuroscience↗

bioRxiv2021

Nodal GABA facilitates axon spike transmission in the spinal cord.

Movement and posture depend on sensory feedback that is regulated by specialized GABAergic neurons (GAD2+) that form axo-axonic contacts onto myelinated proprioceptive sensory axons and are thought to be inhibitory. However, we report here that activating GAD2+ neurons, directly with optogenetics or indirectly by cutaneous stimulation, facilitates sensory feedback to motoneurons in awake rodents and humans. GABAA receptors and GAD2+ innervation at or near nodes of Ranvier of sensory axons cause this facilitation, preventing spike propagation failure at the many axon branch points, which is otherwise common without GABA. In contrast, GABAA receptors are generally lacking from axon terminals (unlike GABAB) and do not presynaptically inhibit transmitter release onto motoneurons. GABAergic innervation near nodes and branch points allows individual branches to function autonomously, with GAD2+ neurons regulating which branches conduct, adding a computational layer to the neuronal networks generating movement and likely generalizing to other CNS axons.

neuroscience↗

bioRxiv2019

Motor axon excitability measures in the rat tail are the same: awake or anaesthetized using sodium pentobarbital.

BackgroundNerve excitability tests in sciatic motor axons are sensitive to anaesthetic choice. Results using ketamine/xylazine (KX) are different from those using sodium pentobarbital (SP). It is not clear which results are most similar to the awake condition, though results using SP appear more similar to human results.\n\nMethodsNerve excitability in tail motor axons was tested in 8 adult female rats with a chronic sacral spinal cord injury. These animals have no behavioural response to electrical stimulation of the tail and were tested awake and then anaesthetized using SP.\n\nResultsThe nerve excitability test results in the awake condition were indistinguishable from the results when the same rats were anaesthetized with sodium pentobarbital. Summary plots of the test results overlap within the boundaries of the standard error and paired t-tests on the 42 discrete measures generated by nerve excitability testing yielded no significant differences (after Bonferroni correction for multiple comparisons).\n\nConclusionsNerve excitability test results in rat motor axons are the same whether the animals are awake or anesthetized using sodium pentobarbital.

neuroscience↗

bioRxiv2019

Intentions of Canadian health professionals towards recommending exercise to people living with ALS.

BackgroundTo provide a nationwide overview of the attitudes, social pressure, perceived ability and intentions of health professionals toward exercise prescription for people living with ALS (pALS).\n\nMethodsAn online survey of physician and other health professionals (HPs) working in academic ALS clinics across Canada.\n\nResultsThe response rate was 48% (84/176) with 30% of respondents identifying as physicians, 63% as other HPs and the remainder as administrative or research personnel. Respondents were sharply divided in their intentions to provide exercise counsel: 24% unlikely and 45% likely. Respondents with low intentions were HPs that considered this activity outside their scope of practice. Measures of intention and attitude were more positive for flexibility compared to strength and aerobic exercise. Perceptions of social pressure and ability to provide exercise counsel were significantly correlated with intention across the three exercise modes in all respondents. Qualitative themes identified as barriers to exercise prescription were lack of confidence or competence (31% physicians, 32% HP), patient tolerance (30% HP), lack of evidence (22% physicians) and lack of infrastructure (22% physicians).\n\nConclusionsWhile \"lack of evidence\" for the benefit of exercise was a deterrent for physicians, the larger issue for all respondents was building competence and confidence in exercise prescription for pALS.

neuroscience↗

bioRxiv2019

Nerve Excitability Differences in Slow and Fast Motor Axons of the Rat: more than just Ih

ObjectiveThe objective was to determine if choice of anaesthetic confounded previous conclusions about the differences in nerve excitability indices between fast and slow motor axons.\n\nMethodologyNerve excitability of the rat sciatic nerve was tested while measuring responses of motor axons innervating the slow-twitch soleus (SOL) and fast-twitch tibialis anterior (TA) muscles. The experiments were conducted with sodium pentobarbital (SP) anaesthetic and compared to previous results that used ketamine-xylazine (KX).\n\nResults and ConclusionsPrevious conclusions about the differences in nerve excitability indices between TA and SOL motor axons using KX were corroborated and extended when experiments were done with SP. Nerve excitability indices sensitive to changes in hyperpolarization-activated inwardly rectifying cation current (Ih) indicated an increase in Ih in SOL axons compared to TA axons (e.g. S3 (-100 %), t=7.949 (df=10), p < 0.0001; TEh (90-100 ms), t=2.659 (df=20), p = 0.0145; hyperpolarizing I/V slope, t=4.308 (df=19), p = 0.0004). SOL axons also had a longer strength-duration time constant (t=3.35 (df=20), p = 0.0032) and a longer and larger magnitude relative refractory period (RRP (ms) t=3.53 (df=12), p = 0.0041; Refractoriness at 2 ms t=0.0055 (df=9), p = 0.0055).\n\nAnaesthetic choice affected many measures of peripheral nerve excitability with differences most apparent in tests of threshold electrotonus and recovery cycle. For example, recovery cycle with KX lacked a clear superexcitable and late subexcitable period. We conclude that KX had a confounding effect on nerve excitability results consistent with ischaemic depolarization. Results using SP revealed the full extent of differences in nerve excitability measures between putative slow and fast motor axons of the rat. These differences have important implications for the use of nerve excitability measures during processes such as ageing where it is believed that there is a selective loss of fast axons.\n\nNew & NoteworthyNerve excitability testing is a tool used to provide insight into the properties of ion channels in peripheral nerves. It is used clinically to assess pathophysiology of motor axons. Researchers customarily think of motor axons as homogeneous; however, we demonstrate there are clear differences between fast and slow axons in the rat. This is important for interpreting results with selective motor neuronopathy, like aging where fast axons are at high risk of degeneration.

neuroscience↗

bioRxiv2018

CityNet - Deep Learning Tools for Urban Ecoacoustic Assessment

O_LICities support unique and valuable ecological communities, but understanding urban wildlife is limited due to the difficulties of assessing biodiversity. Ecoacoustic surveying is a useful way of assessing habitats, where biotic sound measured from audio recordings is used as a proxy for biodiversity. However, existing algorithms for measuring biotic sound have been shown to be biased by non-biotic sounds in recordings, typical of urban environments.\nC_LIO_LIWe develop CityNet, a deep learning system using convolutional neural networks (CNNs), to measure audible biotic (CityBioNet) and anthropogenic (CityAnthroNet) acoustic activity in cities. The CNNs were trained on a large dataset of annotated audio recordings collected across Greater London, UK. Using a held-out test dataset, we compare the precision and recall of CityBioNet and CityAnthroNet separately to the best available alternative algorithms: four acoustic indices (AIs): Acoustic Complexity Index, Acoustic Diversity Index, Bioacoustic Index, and Normalised Difference Soundscape Index, and a state-of-the-art bird call detection CNN (bulbul). We also compare the effect of non-biotic sounds on the predictions of CityBioNet and bulbul. Finally we apply CityNet to describe acoustic patterns of the urban soundscape in two sites along an urbanisation gradient.\nC_LIO_LICityBioNet was the best performing algorithm for measuring biotic activity in terms of precision and recall, followed by bulbul, while the AIs performed worst. CityAnthroNet outperformed the Normalised Difference Soundscape Index, but by a smaller margin than CityBioNet achieved against the competing algorithms. The CityBioNet predictions were impacted by mechanical sounds, whereas air traffic and wind sounds influenced the bulbul predictions. Across an urbanisation gradient, we show that CityNet produced realistic daily patterns of biotic and anthropogenic acoustic activity from real-world urban audio data.\nC_LIO_LIUsing CityNet, it is possible to automatically measure biotic and anthropogenic acoustic activity in cities from audio recordings. If embedded within an autonomous sensing system, CityNet could produce environmental data for cites at large-scales and facilitate investigation of the impacts of anthropogenic activities on wildlife. The algorithms, code and pre-trained models are made freely available in combination with two expert-annotated urban audio datasets to facilitate automated environmental surveillance in cities.\nC_LI

ecology↗

bioRxiv2017

Bat Detective - Deep Learning Tools for Bat Acoustic Signal Detection

O_LIPassive acoustic sensing has emerged as a powerful tool for quantifying anthropogenic impacts on biodiversity, especially for echolocating bat species. To better assess bat population trends there is a critical need for accurate, reliable, and open source tools that allow the detection and classification of bat calls in large collections of audio recordings. The majority of existing tools are commercial or have focused on the species classification task, neglecting the important problem of first localizing echolocation calls in audio which is particularly problematic in noisy recordings.\nC_LIO_LIWe developed a convolutional neural network (CNN) based open-source pipeline for detecting ultrasonic, full-spectrum, search-phase calls produced by echolocating bats (BatDetect). Our deep learning algorithms (CNN FULL and CNN FAST) were trained on full-spectrum ultrasonic audio collected along road-transects across Romania and Bulgaria by citizen scientists as part of the iBats programme and labelled by users of www.batdetective.org. We compared the performance of our system to other algorithms and commercial systems on expert verified test datasets recorded from different sensors and countries. As an example application, we ran our detection pipeline on iBats monitoring data collected over five years from Jersey (UK), and compared results to a widely-used commercial system.\nC_LIO_LIHere, we show that both CNNFULL and CNNFAST deep learning algorithms have a higher detection performance (average precision, and recall) of search-phase echolocation calls with our test sets, when compared to other existing algorithms and commercial systems tested. Precision scores for commercial systems were reasonably good across all test datasets (>0.7), but this was at the expense of recall rates. In particular, our deep learning approaches were better at detecting calls in road-transect data, which contained more noisy recordings. Our comparison of CNNFULL and CNNFAST algorithms was favourable, although CNNFAST had a slightly poorer performance, displaying a trade-off between speed and accuracy. Our example monitoring application demonstrated that our open-source, fully automatic, BatDetect CNNFAST pipeline does as well or better compared to a commercial system with manual verification previously used to analyse monitoring data.\nC_LIO_LIWe show that it is possible to both accurately and automatically detect bat search-phase echolocation calls, particularly from noisy audio recordings. Our detection pipeline enables the automatic detection and monitoring of bat populations, and further facilitates their use as indicator species on a large scale, particularly when combined with automatic species identification. We release our system and datasets to encourage future progress and transparency.\nC_LI

ecology↗

bioRxiv2017

Role of inter-related population-level host traits in determining pathogen richness and zoonotic risk

Zoonotic diseases are an increasingly important source of human infectious diseases, and host pathogen richness of reservoir host species is a critical driver of spill-over risk. Population-level traits of hosts such as population size, host density and geographic range size have all been shown to be important determinants of host pathogen richness. However, empirically identifying the independent influences of these traits has proven difficult as many of these traits directly depend on each other. Here we develop a mechanistic, metapopulation, susceptible-infected-recovered model to identify the independent influences of these population-level traits on the ability of a newly evolved pathogen to invade and persist in host populations in the presence of an endemic pathogen. We use bats as a case study as they are highly social and an important source of zoonotic disease. We show that larger populations and group sizes had a greater influence on the chances of pathogen invasion and persistence than increased host density or the number of groups. As anthropogenic change affects these traits to different extents, this increased understanding of how traits independently determine pathogen richness will aid in predicting future zoonotic spill-over risk.

epidemiology↗