Towards the automated detection and occupancy estimation of primates using passive acoustic monitoring

Recent advancements in technology have made possible the use of novel, cost-efficient biomonitoring techniques which facilitate monitoring animal populations at larger spatial and temporal scales. Here, we investigated using passive acoustic monitoring (PAM) for wild primate populations living in the forest of Taï National Park, Côte d’Ivoire. We assessed the potential of using a customized algorithm for the automated detection of multiple primate species to obtain reliable estimates of species occurrence from acoustic data. First, we applied the algorithm on continuous rainforest recordings collected using autonomous recording units (ARUs) to detect and classify three sound signals: chimpanzee buttress drumming, and the loud calls of the diana and king colobus monkey. Using an occupancy modelling approach we then investigated to what extent the automated, probabilistic output needs to be listened to, and thus manually cleaned, by a human expert, to approach occupancy probabilities derived from ARU data fully verified by a human. To do this we explored the robustness of occupancy probability estimates by simulating ARU datasets with various degrees of cleaning for false positives and false negative detections. We further validated the approach by comparing it to data collected by human observers on point transects located within the same study area. Our study demonstrates that occurrence estimates from ARU data, combined with automated processing methods such as our algorithm, can provide results comparable to data collected by humans and require less effort. We show that occupancy probabilities are quite robust to cleaning effort, particularly when occurrence is high, and suggest that for some species even naïve occupancy, as derived from ARU data without any cleaning, could provide a quick and reliable indicator to guide monitoring efforts. We found detection probabilities to be most influenced by time of day for chimpanzee drums while temperature and, likely, poaching pressure, affected detection of diana monkey loud calls. None of the covariates investigated appeared to have strongly affected king colobus loud call detection. Finally, we conclude that the semi-automated approach presented here could be used as an early-warning system for poaching activity and suggest additional techniques for improving its performance.     

Monitoring coyote population dynamics by genotyping faeces

Reliable population estimates are necessary for effective conservation and management, and faecal genotyping has been used successfully to estimate the population size of several elusive mammalian species. Information such as changes in population size over time and survival rates, however, are often more useful for conservation biology than single population estimates. We evaluated the use of faecal genotyping as a tool for monitoring long-term population dynamics, using coyotes (Canis latrans) in the Alaska Range as a case study. We obtained 544 genotypes from 56 coyotes over 3 years (2000-2002). Tissue samples from all 15 radio-collared coyotes in our study area had > or = 1 matching faecal genotypes. We used flexible maximum-likelihood models to study coyote population dynamics, and we tested model performance against radio telemetry data. The staple prey of coyotes, snowshoe hares (Lepus americanus), dramatically declined during this study, and the coyote population declined nearly two-fold with a 1(1/2)-year time lag. Survival rates declined the year after hares crashed but recovered the following year. We conclude that long-term monitoring of elusive species using faecal genotyping is feasible and can provide data that are useful for wildlife conservation and management. We highlight some drawbacks of standard open-population models, such as low precision and the requirement of discrete sampling intervals, and we suggest that the development of open models designed for continuously collected data would enhance the utility of faecal genotyping as a monitoring tool.     

Selecting a screening cut-off point or diagnostic criterion for comparing prevalences of disease

Comparisons of population prevalences of disease or of inadequate nutriture, across locations or over time (as in monitoring to detect a change in prevalence) are important activities in epidemiology and public health. The data collected are often counts based on a dichotomy of a continuous indicator variable, and performance of the test procedure depends on the cut-off point used. This cut-off may be chosen to optimize performance, on the assumption that the indicator has a mixed normal distribution with unknown mixing proportion corresponding to the unknown prevalence of disease. Results of James (1978, Biometrics 34, 265-275) are applied and extended to the context of monitoring. Charts are presented to facilitate determination of the optimal cut-off, and examples are given for several indicators of nutritional status.     

Participatory Ecological Monitoring of the Alaotra Wetlands in Madagascar

Participatory ecological monitoring is a realistic and effective approach in wetlands such as Alaotra, Madagascar, where important biodiversity is found in an area with high human population density. Since 2001, Durrell Wildlife Conservation Trust, government technical services, regional non-governmental organisations and local communities have collected data on key species, such as waterbirds, a locally endemic lemur and useful natural resources. The monitoring was linked with environmental quizzes and an inter-village competition, which helped raise interest in the monitoring and publicise results. The monitoring has assisted wetland management by guiding amendments to and increasing respect for the regional fishing convention, raising awareness, catalysing marsh management transfer to communities and stimulating collaboration and good governance. The sustainability of the monitoring scheme and the usefulness of the data for detecting trends and guiding local managements are discussed.     

Robust and simplified machine learning identification of pitfall trap‐collected ground beetles at the continental scale

1. Insect populations are changing rapidly, and monitoring these changes is essential for understanding the causes and consequences of such shifts. However, large‐scale insect identification projects are time‐consuming and expensive when done solely by human identifiers. Machine learning offers a possible solution to help collect insect data quickly and efficiently.
2. Here, we outline a methodology for training classification models to identify pitfall trap‐collected insects from image data and then apply the method to identify ground beetles (Carabidae). All beetles were collected by the National Ecological Observatory Network (NEON), a continental scale ecological monitoring project with sites across the United States. We describe the procedures for image collection, image data extraction, data preparation, and model training, and compare the performance of five machine learning algorithms and two classification methods (hierarchical vs. single‐level) identifying ground beetles from the species to subfamily level. All models were trained using pre‐extracted feature vectors, not raw image data. Our methodology allows for data to be extracted from multiple individuals within the same image thus enhancing time efficiency, utilizes relatively simple models that allow for direct assessment of model performance, and can be performed on relatively small datasets.
3. The best performing algorithm, linear discriminant analysis (LDA), reached an accuracy of 84.6% at the species level when naively identifying species, which was further increased to >95% when classifications were limited by known local species pools. Model performance was negatively correlated with taxonomic specificity, with the LDA model reaching an accuracy of ~99% at the subfamily level. When classifying carabid species not included in the training dataset at higher taxonomic levels species, the models performed significantly better than if classifications were made randomly. We also observed greater performance when classifications were made using the hierarchical classification method compared to the single‐level classification method at higher taxonomic levels.
4. The general methodology outlined here serves as a proof‐of‐concept for classifying pitfall trap‐collected organisms using machine learning algorithms, and the image data extraction methodology may be used for nonmachine learning uses. We propose that integration of machine learning in large‐scale identification pipelines will increase efficiency and lead to a greater flow of insect macroecological data, with the potential to be expanded for use with other noninsect taxa.

     

Estimating linear temporal trends from aggregated environmental monitoring data

Trend estimates are often used as part of environmental monitoring programs. These trends inform managers (e.g., are desired species increasing or undesired species decreasing?). Data collected from environmental monitoring programs is often aggregated (i.e., averaged), which confounds sampling and process variation. State-space models allow sampling variation and process variations to be separated. We used simulated time-series to compare linear trend estimations from three state-space models, a simple linear regression model, and an auto-regressive model. We also compared the performance of these five models to estimate trends from a long term monitoring program. We specifically estimated trends for two species of fish and four species of aquatic vegetation from the Upper Mississippi River system. We found that the simple linear regression had the best performance of all the given models because it was best able to recover parameters and had consistent numerical convergence. Conversely, the simple linear regression did the worst job estimating populations in a given year. The state-space models did not estimate trends well, but estimated population sizes best when the models converged. We found that a simple linear regression performed better than more complex autoregression and state-space models when used to analyze aggregated environmental monitoring data.     

Comparing survey methods to assess the conservation value of a community‐managed protected area in western Tanzania

The ability of low‐status protected areas under community management to achieve a conservation objective is frequently questioned, particularly in developing countries. The lack of sound, scientific‐based biodiversity monitoring frequently undermines attempts to evaluate the extent to which these areas are contributing to biodiversity conservation. Based on data collected between 2008 and 2010 in a Forest Reserve under community management in western Tanzania, our study tested fives methods: camera trapping, walking line transects, vehicle transects, opportunistic encounters and indirect signs, to find the most appropriate for future monitoring. Method comparisons confirmed a higher performance of camera trapping compared to other methods for the ability to detect species. However, our results identified the need of a better survey design to ensure a sound monitoring in the future. Besides method comparisons, our study provides the first fine‐scale data on mammal communities in such a low‐status protected area. Combined methods allow the identification of 49 species of medium and large mammals, a surprisingly high diversity for such area. These findings outline the potential conservation value of this type of protected area and call for better biodiversity monitoring throughout complexes of protected areas of different statuses and management regimes.     

The use of the alkaline comet assay with lymphocytes in human biomonitoring studies

We reviewed the data of 45 alkaline comet assay studies with lymphocytes published during the last three years with the objective of monitoring human exposure to genotoxic agents as a result of occupation, drug treatment, diseases or environmental pollution. The strengths of the studies were that: (i) a lot of data could be obtained within a relatively short period of time in a cost-effective manner, (ii) lymphocytes could be easily collected in a non-invasive way and proved to be good surrogate cells in that they picked up effects caused by agents with different cancer target organs and (iii) a remarkable concordance between comet assay and cytogenetic assay data was proved. However, our analysis revealed some shortcomings of the studies such as: (i) the inclusion of low number of study participants and bias in the number and gender of subjects between control and exposed groups, (ii) lack of qualitative and quantitative exposure data, (iii) lack of consideration of differences in physical activity and diet between control and exposed groups, (iv) difficulty in comparison of the studies due to lack of uniformity in the comet assay procedures such as duration of alkali unwinding and electrophoresis, slide scoring method and the metrics used to assess the extent of DNA damage and (v) controversy in the sensitivity of comet assay since it picked up DNA damage caused by agents such as wood dust, pesticides and hormone preparations which were found to be weak genotoxins or non-genotoxins in other tests, but gave inconsistent results with known mutagens/carcinogens such as tobacco smoke. We feel that for the alkaline comet assay to be an important tool in human biomonitoring studies, serious consideration should be given to the flaws in the design and performance of the assay.     

A single‐nucleotide polymorphism‐based approach for rapid and cost‐effective genetic wolf monitoring in Europe based on noninvasively collected samples

Noninvasive genetics based on microsatellite markers has become an indispensable tool for wildlife monitoring and conservation research over the past decades. However, microsatellites have several drawbacks, such as the lack of standardisation between laboratories and high error rates. Here, we propose an alternative single‐nucleotide polymorphism (SNP)‐based marker system for noninvasively collected samples, which promises to solve these problems. Using nanofluidic SNP genotyping technology (Fluidigm), we genotyped 158 wolf samples (tissue, scats, hairs, urine) for 192 SNP loci selected from the Affymetrix v2 Canine SNP Array. We carefully selected an optimised final set of 96 SNPs (and discarded the worse half), based on assay performance and reliability. We found rates of missing data in this SNP set of <10% and genotyping error of ~1%, which improves genotyping accuracy by nearly an order of magnitude when compared to published data for other marker types. Our approach provides a tool for rapid and cost‐effective genotyping of noninvasively collected wildlife samples. The ability to standardise genotype scoring combined with low error rates promises to constitute a major technological advancement and could establish SNPs as a standard marker for future wildlife monitoring.     

Designing better frog call recognition models

Advances in bioacoustic technology, such as the use of automatic recording devices, allow wildlife monitoring at large spatial scales. However, such technology can produce enormous amounts of audio data that must be processed and analyzed. One potential solution to this problem is the use of automated sound recognition tools, but we lack a general framework for developing and validating these tools. Recognizers are computer models of an animal sound assembled from “training data” (i.e., actual samples of vocalizations). The settings of variables used to create recognizers can impact performance, and the use of different settings can result in large differences in error rates that can be exploited for different monitoring objectives. We used Song Scope (Wildlife Acoustics Inc.) to build recognizers and vocalizations of the wood frog (Lithobates sylvaticus) to test how different settings and amounts of training data influence recognizer performance. Performance was evaluated using precision (the probability of a recognizer match being a true match) and sensitivity (the proportion of vocalizations detected) based on a receiver operating characteristic (ROC) curve‐determined score threshold. Evaluations were conducted using recordings not used to build the recognizer. Wood frog recognizer performance was sensitive to setting changes in four out of nine variables, and small improvements were achieved by using additional training data from different sites and from the same recording, but not from different recordings from the same site. Overall, the effect of changes to variable settings was much greater than the effect of increasing training data. Additionally, by testing the performance of the recognizer on vocalizations not used to build the recognizer, we discovered that Type I error rates appear idiosyncratic and do not recommend extrapolation from training to new data, whereas Type II errors showed more consistency and extrapolation can be justified. Optimizing variable settings on independent recordings led to a better match between recognizer performance and monitoring objectives. We provide general recommendations for application of this methodology with other species and make some suggestions for improvements.     

On the use of the substitution method in left-censored environmental data

In risk assessment and environmental monitoring studies, concentration measurements frequently fall below detection limits (DL) of measuring instruments, resulting in left-censored data. The principal approaches for handling censored data include the substitution-based method, maximum likelihood estimation, robust regression on order statistics, and Kaplan-Meier. In practice, censored data are substituted with an arbitrary value prior to use of traditional statistical methods. Although some studies have evaluated the substitution performance in estimating population characteristics, they have focused mainly on normally and lognormally distributed data that contain a single DL. We employ Monte Carlo simulations to assess the impact of substitution when estimating population parameters based on censored data containing multiple DLs. We also consider different distributional assumptions including lognormal, Weibull, and gamma. We show that the reliability of the estimates after substitution is highly sensitive to distributional characteristics such as mean, standard deviation, skewness, and also data characteristics such as censoring percentage. The results highlight that although the performance of the substitution-based method improves as the censoring percentage decreases, its performance still depends on the population's distributional characteristics. Practical implications that follow from our findings indicate that caution must be taken in using the substitution method when analyzing censored environmental data.     

Optimising long‐term monitoring projects for species distribution modelling: how atlas data may help

Long‐term biodiversity monitoring data are mainly used to estimate changes in species occupancy or abundance over time, but they may also be incorporated into predictive models to document species distributions in space. Although changes in occupancy or abundance may be estimated from a relatively limited number of sampling units, small sample size may lead to inaccurate spatial models and maps of predicted species distributions. We provide a methodological approach to estimate the minimum sample size needed in monitoring projects to produce accurate species distribution models and maps. The method assumes that monitoring data are not yet available when sampling strategies are to be designed and is based on external distribution data from atlas projects. Atlas data are typically collected in a large number of sampling units during a restricted timeframe and are often similar in nature to the information gathered from long‐term monitoring projects. The large number of sampling units in atlas projects makes it possible to simulate a broad gradient of sample sizes in monitoring data and to examine how the number of sampling units influences the accuracy of the models. We apply the method to several bird species using data from a regional breeding bird atlas. We explore the effect of prevalence, range size and habitat specialization of the species on the sample size needed to generate accurate models. Model accuracy is sensitive to particularly small sample sizes and levels off beyond a sufficiently large number of sampling units that varies among species depending mainly on their prevalence. The integration of spatial modelling techniques into monitoring projects is a cost‐effective approach as it offers the possibility to estimate the dynamics of species distributions in space and over time. We believe our innovative method will help in the sampling design of future monitoring projects aiming to achieve such integration.     

Evaluation of a Global Positioning System backpack transmitter for wild turkey research

Radiotelemetry is the standard method for monitoring wild turkey (Meleagris gallapavo) movements and habitat use. Spatial data collected using telemetry-based monitoring are frequently inaccurate due to triangulation error. However, new technology, such as Global Positioning Systems (GPS) has increased ecologists' ability to accurately evaluate animal movements and habitat selection. We evaluated the efficacy of micro-GPS backpack units for use on wild turkeys. We tested a micro-GPS developed specifically for avian species that incorporated a GPS antenna with a lightweight rechargeable battery and a very high frequency (VHF) transmitter. We conducted a series of static tests to evaluate performance in varying types of vegetative canopy cover and terrain. After static testing, we deployed micro-GPS on 8 adult male Rio Grande wild turkeys (M. g. intermedia) trapped in south Texas and 2 adult females trapped in the Texas panhandle. Micro-GPS units collected 26,439 locations out of 26,506 scheduled attempts (99.7% fix rate) during static testing. Mean distance error across all static tests was 15.5 m (SE = 0.1). In summer 2009, we recovered micro-GPS from 4 tagged males and both females to evaluate data collection. Units on males acquired approximately 2,500 locations over a 65-day test period (94.5% fix rate). We recovered units from the 2 females after 19 days and 53 days; those units acquired 301 and 837 locations, respectively, for a 96% fix rate. Cost analysis indicated that VHF will be cost effective when 1 location per day is required up to 181 days, but micro-GPS becomes less expensive as frequency of daily locations increases. Our results indicate that micro-GPS have the potential to provide increased reliable data on turkey movement ecology and habitat selection at a higher resolution than conventional VHF telemetric methods. © 2011 The Wildlife Society.     

Chemical process based reconstruction of exposures for an epidemiological study. III. Analysis of industrial hygiene samples

As part of an historical cohort study to investigate the mortality experience of industrial workers exposed to chloroprene (beta-CD) and other substances, all available industrial hygiene exposure monitoring data were collected and summarized. From discussions with on-site industrial hygiene personnel, it was apparent that these data were not collected for epidemiological purposes and, therefore, their use in characterization of exposures was problematic as the data mostly pertained to samples collected to investigate the performance of specific tasks. These data were, however, informative for validating the exposure modeling process used to estimate historical exposures. The data summarized below clearly indicate that exposures to beta-CD were lowered across the time period of this study. Typically, the exposures recorded were less than the occupational exposure limits of the periods in which the exposures were recorded. Additionally, exposure measurements recorded in the recent past do not represent the exposure actually experienced by the worker as a strict personal protective equipment use program has been in place for the facilities studied since the mid-1980s.     

被动声学监测技术在西黑冠长臂猿监测中的应用

近年来, 被动声学监测技术被广泛应用于陆生哺乳动物的监测, 它能以较低的价格和非侵入的方式在特定区域进行野生动物无人值守监测, 面临的主要问题是需要人工收回数据和后期数据分析整理较为困难。本研究设计了一套被动声学监测系统用于西黑冠长臂猿(Nomascus concolor)监测, 监测系统在野外由太阳能供电, 使用自研的指向性拾音器阵列采集鸣声数据, 并通过无线网桥实时传输数据至管护局办公楼的服务器进行存储, 通过后台的鸣声数据管理系统辅助研究人员识别鸣声和辨认鸣声来源方向, 简化数据采集和处理流程。该系统在哀牢山国家级自然保护区枇杷箐科研监听点对两个西黑冠长臂猿群体进行了351天的连续监测, 特点为: 系统运行长期稳定, 数据通过无线方式传输便捷高效且不受季节天气影响。指向性拾音器阵列能有效分辨长臂猿鸣声来源方向, 弥补了传统监测设备难以分辨鸣声方向的缺陷。该系统与现有人工监测方法相比在数据采集的持续性、连续性、完整度以及鸣声数据处理智能化和监测成本方面均具有一定优势, 符合西黑冠长臂猿持续长期监测需求, 未来可作为西黑冠长臂猿自动化监测的解决方案进行推广应用。     

Statistical inference for self-designing clinical trials with a one-sided hypothesis

In the process of monitoring clinical trials, it seems appealing to use the interim findings to determine whether the sample size originally planned will provide adequate power when the alternative hypothesis is true, and to adjust the sample size if necessary. In the present paper, we propose a flexible sequential monitoring method following the work of Fisher (1998), in which the maximum sample size does not have to be specified in advance. The final test statistic is constructed based on a weighted average of the sequentially collected data, where the weight function at each stage is determined by the observed data prior to that stage. Such a weight function is used to maintain the integrity of the variance of the final test statistic so that the overall type I error rate is preserved. Moreover, the weight function plays an implicit role in termination of a trial when a treatment difference exists. Finally, the design allows the trial to be stopped early when the efficacy result is sufficiently negative. Simulation studies confirm the performance of the method.     

Methods to Test Visual Attention Online

Online data collection methods have particular appeal to behavioral scientists because they offer the promise of much larger and much more representative data samples than can typically be collected on college campuses. However, before such methods can be widely adopted, a number of technological challenges must be overcome – in particular in experiments where tight control over stimulus properties is necessary. Here we present methods for collecting performance data on two tests of visual attention. Both tests require control over the visual angle of the stimuli (which in turn requires knowledge of the viewing distance, monitor size, screen resolution, etc.) and the timing of the stimuli (as the tests involve either briefly flashed stimuli or stimuli that move at specific rates). Data collected on these tests from over 1,700 online participants were consistent with data collected in laboratory-based versions of the exact same tests. These results suggest that with proper care, timing/stimulus size dependent tasks can be deployed in web-based settings.     

Can biodiversity monitoring schemes provide indicators for ecosystem services?

Recently, the science and policy agenda on biodiversity moved to include ecosystem services assessments and it is recognised that for determining the effectiveness and progress of policy frameworks monitoring is crucial.Within European monitoring schemes, data is collected following different sampling protocols for a range of biodiversity or context related aspects; from EU-wide general land cover mapping to red list species within Annex I habitats. In this paper, we analysed field instructions of seven monitoring schemes on the extent to which they can provide data on the provision of ecosystem services and what additional information may be needed for future monitoring of ecosystem services.We compared seven monitoring schemes (i.e. CORINE Land Cover, Land Use Cover Area Survey (LUCAS), European Biodiversity Observation Network (EBONE), biodiversity monitoring on organic and low-input farming systems (BioBio), National Inventory of the Landscape of Sweden (NILS) and Pan-European Common Birds Monitoring (PECBM) and UK Butterfly monitoring (UK-BM)) by scoring the quality of recorded parameters and the adequacy of sampling protocols for ecosystem service monitoring.All the examined schemes were able to provide some parameters on ecosystem services, but the quality of the parameters on average did not exceed the level of qualitative data. Additionally, the divergence between the sampling designs of the schemes and the spatial characteristics of ecosystem services reduced the potential monitoring value of all schemes. Monitoring schemes including a range of sampling methods, scales and included the recording of data on habitats, such as EBONE, BioBio and NILS, provided the best data on the provision of ecosystem services.We conclude that improvement of the monitoring of ecosystem services is hindered by several knowledge gaps: (1) a robust definition and conceptual framework of ecosystem services; (2) the linkage between biodiversity and ecosystem services; and (3) the interpretation of monitoring data.In addition to ecosystem service monitoring, biodiversity monitoring unremittingly remains very important, at least to identify trade-offs between the management for services and the resulting biodiversity status.