Search

Context It is notoriously difficult to estimate the size of animal populations, especially for cryptic or threatened species that occur in low numbers. Recent advances with acoustic sensors make the detection of animal populations cost effective when coupled with software that can recognise species-specific calls. Aims We assess the potential for acoustic sensors to estimate koala, Phascolarctos cinereus, density, when individuals are not identified, using spatial count models. Sites were selected where previous independent estimates of density were available. Methods We established acoustic arrays at each of five sites representing different environments and densities of koalas in New South Wales. To assess reliability, we compared male koala density estimates derived from spatial count modelling to independently derived estimates for each site. Key results A total 11 312 koala bellows were verified across our five arrays. Koalas were detected at most of our sample locations (96–100% of sensors; n = 130), compared with low detection rates from rapid scat searches at trees near each sensor (scats at

This study aims to recognise frog choruses using false-colour spectrograms and machine learning algorithms with acoustic indices. This can be a useful solution for improving the efficiency of long-term acoustic monitoring. Acid frogs, our target species, are a group of endemic frogs that are particularly sensitive to habitat change and competition from other species. The Wallum Sedgefrog (Litoria olongburensis) is the most threatened acid frog species facing habitat loss and degradation across much of their distribution, in addition to further pressures associated with anecdotally-recognised competition from their sibling species, the Eastern Sedgefrogs (Litoria fallax). Monitoring the calling behaviours of these two species is essential for informing L. olongburensis management and protection, and for obtaining ecological information about the process and implications of their competition. Considering the cryptic nature of L. olongburensis and the sensitivity of their habitat to human disturbance, passive acoustic monitoring is a suitable method for monitoring this species. However, manually processing this overwhelmingly large quantities of acoustic data collected is time-consuming and not feasible in the long-term. Therefore, there is a high demand for automated acoustic recognition methods to efficiently search long-duration recordings and identify target species. In this study, we propose a two-step scheme for quickly identifying frog choruses, which is first narrowing down the search scope by inspecting long-duration false-colour spectrograms and then recognising target acoustic signals using machine learning and acoustic indices. This method is efficient, time-saving and general, which means it can easily adopted to other species. Our research also provides insights on how to choose acoustic features that efficiently recognise species from larger scale field-collected recordings. The experimental results show that these techniques are useful in identifying choruses of the two competitive frog species with an accuracy of 76.7% on identifying four acoustic patterns (whether the two species occurred).

In recent years, passive acoustic monitoring has emerged as a reliable method for monitoring soniferous fauna, with numerous studies finding estimates of species richness and community composition are comparable with estimates derived from conventional field surveys. Most of these studies compare point counts of forest bird communities with contemporaneous short‐duration acoustic recordings, but several questions remain. How do these two methods compare in more open, arid ecosystems, how does applying methods optimally influence comparisons, and how do patterns in acoustically detectable species compare with overall patterns? Here we demonstrate techniques for improving sampling of acoustic data and conduct acoustic surveys to estimate species richness from ephemeral creek‐lines in the Australian arid zone using a long‐term passively derived acoustic dataset. We examine these results in the context of long‐term observer‐based transect surveys conducted along the same creek‐lines to define species within the avian assemblage that are acoustically detectable or acoustically undetectable/cryptic. Our data suggest that some species were consistently missed by acoustic surveys, but most belonged to groups that are typically excluded from inventories prior to analysis including rare species, raptors, waterbirds, swallows and nocturnal birds. Further, the relative diversities of sites were well estimated by acoustic surveys, with variations between sites reflecting those estimated by observer‐based field surveys. This suggests that in our study system, acoustically detectable species are reliable indicators of overall species richness. Field‐based surveys will remain an important component of sampling in arid ecosystems and we highlight the value of applying acoustic and conventional field‐based surveys in a complementary manner. This approach allows large, publishable datasets to be generated by exploiting the temporal reach of acoustic sensors, while also maximizing detections of acoustically cryptic species via field‐based surveys. We argue that acoustic monitoring has the potential to facilitate greater research effort in largely research‐deficient arid ecosystems.

Cost-effective surveys of low density koala populations are challenging, but technological developments in the acoustics field offer great potential for landscape-scale surveys and monitoring. We assessed passive acoustic recording coupled with automated call identification as a survey method for koalas Phascolarctos cinereus . Surveys targeted areas of previously known koala activity based on scat surveys in southern forests of New South Wales where a low density of koalas is suspected. We set 24 Song Meters to record at night over a two week period (~3,696 hours) in the koala breeding season (October/November) in Murrah Flora Reserve. Recordings were scanned by a koala call recogniser and “matches” were manually verified. Across the 24 sites, 522 validated koala bellows were recorded at 21 sites (87.5 % detection rate). Three environmental variables had most influence on detection probability of koalas, including nightly rainfall (-ve), nightly temperature (-ve) and topographic position (lower on ridges). Calling activity peaked at midnight. Sustained site occupancy, at least in the short-term, was apparent as under optimal conditions (no rain) koalas were recorded, on average, for > 50 % of survey nights rather than for just a few nights. We suggest that only a modest survey effort (4–5 nights) in the breeding season, on nights with < 3 mm of rain, is required to achieve 90–95 % probability of koala call detection in an area of low koala density. Comparison with scat surveys at the same sites revealed that detection rates were more than three times greater with acoustic surveys. Technological advances will continue to provide improvements for wildlife survey, and perhaps most importantly, for collecting much needed long-term data to assess trends in occupancy or other population attributes over time.

Selective logging in tropical forests changes the local number of animal species (alpha diversity), but it also likely affects species turnover (beta diversity). Whilst such changes are documented in many ecosystems under different disturbances, they are poorly understood in selectively logged tropical forests. By using soundscape recordings across broad spatial scales, we measured soundscape saturation and dissimilarity of pairs of soundscapes, as a proxy of alpha and beta diversity, respectively, in selectively logged and protected tropical forest in Indonesian Borneo. Soundscapes of selectively logged forests were more homogeneous than the soundscapes of never logged forest, and that soundscape saturation of protected forest was higher during the day and lower at night in comparison with selective logging concessions. Synthesis and applications. Selectively logged forests act as an important reservoir of biodiversity. Optimizing such production forests for biodiversity conservation requires the consideration of the total continuous area that is assigned to selective logging, and spatial arrangement of annual cutting blocks, as these could affect beta diversity and its recovery.

Growing developmental activities, such as hydropower construction, farm roads, and other human activities, are affecting the critically endangered white-bellied heron (WBH). Out of a known global population of 60, 28 individuals inhabit the river basin area and freshwater lakes and ponds of Bhutan. Several constraints impede continuous monitoring of endangered species, such as the isolated and cryptic nature of the species and the remoteness of its habitat; to date, there are no long-term reference data or techniques implemented for continuous monitoring of this species. In this study, we designed acoustic detection and habitat characterisation methods using long-duration recordings from three habitat areas in Bhutan. Acoustic indices were extracted and used to implement a species-specific call detector and to generate habitat soundscape representations. Using WBH calls annotated in month-long recordings from a known site, a novel indices-based detector was implemented and tested. A total of 960 hr of continuous audio recordings from three habitats in Bhutan were analysed. We found that a species call detector implemented using a combination of acoustic indices (that includes measures of spectral and temporal entropy and different angles of spectral ridges) has a correct detection rate of 81%. Additionally, visual inspection of the species’ acoustic habitat using long-duration false-colour spectrograms enabled qualitative assessment of acoustic habitat structure and other dominant acoustic events. This study proposes a combined approach of species acoustic detection and habitat soundscape analysis for holistic acoustic monitoring of endangered species. As a direct outcome of this work, we documented acoustic reference data on the critically endangered WBH from multiple habitat areas and have analysed its temporal vocalisation patterns across sites.

There is global concern about tropical forest degradation, in part, because of the associated loss of biodiversity. Communities and indigenous people play a fundamental role in tropical forest management and are often efficient at preventing forest degradation. However, monitoring changes in biodiversity due to degradation, especially at a scale appropriate to local tropical forest management, is plagued by difficulties, including the need for expert training, inconsistencies across observers, and lack of baseline or reference data. We used a new biodiversity remote-sensing technology, the recording of soundscapes, to test whether the acoustic saturation of a tropical forest in Papua New Guinea decreases as land-use intensity by the communities that manage the forest increases. We sampled soundscapes continuously for 24 hours at 34 sites in different land-use zones of 3 communities. Land-use zones where forest cover was fully retained had significantly higher soundscape saturation during peak acoustic activity times (i.e., dawn and dusk chorus) compared with land-use types with fragmented forest cover. We conclude that, in Papua New Guinea, the relatively simple measure of soundscape saturation may provide a cheap, objective, reproducible, and effective tool for monitoring tropical forest deviation from an intact state, particularly if it is used to detect the presence of intact dawn and dusk choruses.

Species distribution models have great potential to efficiently guide management for threatened species, especially for those that are rare or cryptic. We used MaxEnt to develop a regional‐scale model for the koala Phascolarctos cinereus at a resolution (250 m) that could be used to guide management. To ensure the model was fit for purpose, we placed emphasis on validating the model using independently‐collected field data. We reduced substantial spatial clustering of records in coastal urban areas using a 2‐km spatial filter and by modeling separately two subregions separated by the 500‐m elevational contour. A bias file was prepared that accounted for variable survey effort. Frequency of wildfire, soil type, floristics and elevation had the highest relative contribution to the model, while a number of other variables made minor contributions. The model was effective in discriminating different habitat suitability classes when compared with koala records not used in modeling. We validated the MaxEnt model at 65 ground‐truth sites using independent data on koala occupancy (acoustic sampling) and habitat quality (browse tree availability). Koala bellows (n = 276) were analyzed in an occupancy modeling framework, while site habitat quality was indexed based on browse trees. Field validation demonstrated a linear increase in koala occupancy with higher modeled habitat suitability at ground‐truth sites. Similarly, a site habitat quality index at ground‐truth sites was correlated positively with modeled habitat suitability. The MaxEnt model provided a better fit to estimated koala occupancy than the site‐based habitat quality index, probably because many variables were considered simultaneously by the model rather than just browse species. The positive relationship of the model with both site occupancy and habitat quality indicates that the model is fit for application at relevant management scales. Field‐validated models of similar resolution would assist in guiding management of conservation‐dependent species.

This research explores the recognition of choruses of two co-existing sibling frog species in long-duration field recordings using false-colour spectrograms and acoustic indices. Acid frogs are a group of endemic frogs that are particularly sensitive to habitat change and competition from other species. The Wallum Sedgefrog (Litoria olongburensis) is the most threatened acid frog species facing habitat loss and degradation across much of their distribution, in addition to further pressures associated with anecdotally-recognised competition from their sibling species, the Eastern Sedgefrogs (Litoria fallax). Monitoring the calling behaviours of these two species is essential for informing L. olongburensis management and protection, and for obtaining ecological information about the process and implications of their competition. Considering the cryptic nature of L. olongburensis and the sensitivity of their habitat to human disturbance, passive acoustic monitoring is a suitable method for monitoring this species. However, manually processing the large quantities of acoustic data collected using these methods is time-consuming and not feasible in the long-term. Therefore, there is a high demand for automated acoustic recognition tools to efficiently search months of recordings and identify target species. Our research provides more insight on how to choose acoustic features that efficiently recognise species from large-scale field-collected recordings at a larger scale. The experimental results show that these techniques are useful in identifying choruses of the two competitive frog species with an accuracy of 76.7% on identifying four acoustic patterns (whether the two species occurred).

Over half of the world's tropical forests are used for timber extraction by selective logging. Even though these forests are degraded to a variable degree, they are still important for tropical forest biodiversity. It is not yet known how biodiversity is impacted during and immediately after logging, and how fast it recovers. Here, we use ecoacoustics, and specifically the recording and analysis of soundscape dawn time series, to monitor the immediate impact and early recovery of biodiversity after selective logging. We describe the results of capturing a dawn time series of soundscapes in a Bornean tropical lowland forest before, during, and after selective logging. Soundscape saturation, which is correlated with the number of different calls that make up the soundscape, dropped significantly immediately after selective logging was carried out. The reduced saturation was mostly due to the loss of bird vocalizations, whereas insects did not seem to be impacted. In a space-for-time recording from the same area, soundscape saturation during the equivalent time of day recovered after one year, however began to drop again in years 2 and 3 after logging. Our results are suggestive of a strong effect of logging and seasonality on tropical forest soundscapes. Being able to track biodiversity changes and recovery over time in tropical forests managed for timber production can ultimately improve our odds of avoiding more species extinctions, by providing evidence-based recommendations on how much time forests need to recover biodiversity.

Protected areas (PAs) monitoring is a technical bottleneck that limits the implementation of decision-making processes for natural resource and wildlife management. Recent methodological advances make passive acoustic monitoring and associated acoustic index analysis an increasingly suitable method for PAs monitoring. Acoustic indices are mathematical filters that can provide standardised comparative information about the acoustic energy, which can be applied to compare communities. In this study we test whether acoustic indices are sufficiently sensitive to detect differences in the soundscape within each of the four seasons between a PA (the Serra do Cipo National Park, Brazil) and a surrounding farmland area. Statistical analysis of results from 12 acoustic indices is used to identify which of 20 acoustic regions, defined by frequency range and time period, present the greatest differences between the two sites. The soundscapes of the two sites differed most in autumn within the acoustic region 6, representing 05:30 – 09:00am and a range of 0.988–3.609 kHz. This acoustic region exhibited significant differences for all the 12 indices tested. Visual examination of 65 long-duration false-colour (LDFC) spectrograms resulted in the selection of 865 (from 1365) sound files with acoustic events within the range of acoustic region 6. Sonotype analysis of the 865 files showed that the soundscape outside the park is strongly influenced by human activity, with domestic animals rare in the park soundscape (1% of the sound files), but very common in the surrounding farmland environment (63% of the sound files). The main goal of monitoring programmes detecting biodiversity trends across space and time, which is here achieved via passive acoustic monitoring and acoustic indices. This confirms the utility of the techniques used here for PA monitoring, especially for detecting trends in anthropogenic disturbance, which is a common threat to natural habitats in parks and reserves in the tropics.

This research explores the data selection problem in acoustic recognition of two co-existing sibling frog species from long-duration field recordings. This study explores the data selection problem in species recognition with machine learning, including instance selection and acoustic index feature selection. Our target species are two co-existing frog species. The Wallum Sedgefrog (Litoria olongburensis) is the most threatened acid frog species, facing habitat loss and degradation across much of their distribution, in addition to further pressures associated with anecdotally-recognised competition from it's sibling species, the Eastern Sedgefrog (Litoria fallax). Monitoring the calling behaviours of these two species is essential for informing L. olongburensis management and protection, and for obtaining ecological information about the process and implications of their competition. In order to recognise them from recordings, automated recognisers, instead of manual surveys, are required due to the overwhelmingly large volume of acoustic data. However, it costs much time and effort to annotate acoustic data, which results in a lack of labelled data for training recognisers. In addition, the composition of field audio recordings is complex and varies according to weather conditions, seasonal changes and animal activities. In this case, the chorusing behaviours of these two frog species can be greatly affected by weather conditions, for example, rains. Since we can only select limited amount of data from a very diverse data pool to annotate, it is important to explore how the selection of instances and features affects the performance of recognisers. In this paper, we selected two 24-h audio datasets from different weather conditions as our dataset. We first give a detailed visual comparison of them with false-colour spectrograms. Then we use datasets from individual days, their combination and a synthetic set constructed from them to evaluate the impact of different training instances on the recognition performance. We also analyses the effectiveness of acoustic index features. The experimental results show that models trained on data of a vocalisation-intense day or a normal day do not work well on each other. However, extra data from a normal-day dataset does help the recognition on a vocalisation-intense day, especially when the composition of the training set is consistent with the real-life weather condition ratio. Generally, including more data in training set is helpful, but limiting vocalisation-intense data as their ratio in real life help optimise the performance of recognisers. As for the selection of acoustic index features, a good recognition performance on different chorusing behaviours requires different sets of features.

Long duration acoustic monitoring is becoming an increasingly popular approach to extend survey effort by using autonomous sensors to passively collect data over large temporal and spatial scales. This is of particular benefit when attempting to detect a species whose temporal vocalization strategy is unknown, and whose small population size reduces detection probability. It is also of benefit in environments that are logistically difficult to access such as wetlands. We investigated the vocalization strategy of the Least Bittern (Ixobrychus exilis), a species of high conservation concern in the Western hemisphere and ‘in need of management’ in multiple states of the USA. The Least Bittern is a secretive marsh bird that is primarily detected by its vocalizations and call-playback surveys are typically used for population monitoring. To minimize disturbance to both the birds and their habitat, we deployed autonomous acoustic recording units and collected continuous 24-hour audio recordings for 30 days. The resultant accumulation of data necessitated an automated method to assist with analysis and interpretation. We successfully applied a novel soundscape technique—long-duration, false-color (LDFC) spectrograms—to visually confirm presence of Least Bittern from the ‘coo coo coo’ vocalization associated with breeding. In addition, we used a machine learning technique to automate the acoustic event detection process. Peak vocalization times were then predicted from an annotated dataset of actual calls and subsequently used to develop an optimal acoustic survey strategy. The results of this research demonstrate how machine learning methods can search large data sets for a specific species. This information can then be used to optimize existing monitoring methods, to increase detection probability and to minimize associated costs.

Monitoring the environment with acoustic sensors is now practical; sensors are sold as commercial devices, storage is cheap, and the field of ecoacoustics is recognized as an effective way to scale monitoring of the environment. However, a pressing challenge faced in many eScience projects is how to manage, analyze, and visualize very large data so that scientists can benefit, with ecoacoustic data presenting its own particular challenges. This paper presents a new zoomable interactive visualization interface for the exploration of environmental audio data. The interface is a new tool in the Acoustic Workbench, an ecoacoustics software platform built for managing environmental audio data. This Google Maps like interface for audio data, enables zooming in and out of audio data by incorporating specialized, multiresolution, visual representations of audio data into the workbench website. The ‘zooming’ visualization allows scientists to surface the structure, detail, and patterns in content that would otherwise be opaque to them, from scales of seconds through to weeks of data. The Ecosounds instance of the Acoustic Workbench contains 52 years (108 TB) of audio data, from 1016 locations, which results in a 180 million-tile, 8.3 terapixel visualization. The design and implementation of this novel big audio data visualization is presented along with some design considerations for storing visualization tiles.

Retention forestry aims to mitigate impacts of native forestry on biodiversity, but data are limited on its effectiveness for threatened species. We used acoustics to investigate the resilience of a folivorous marsupial, the koala Phascolarctos cinereus, to timber harvesting where a key mitigation practice is landscape exclusion of harvesting. We deployed acoustic recorders at 171 sites to record male bellows (~14,640 hours) for use in occupancy modelling and for comparisons of bellow rate (bellows night-1). Surveys targeted modelled medium-high quality habitat, with sites stratified by time since logging and logging intensity, including old growth as a reference. After scanning recordings with software to identify koala bellows, we found a high probability of detection (~0.45 per night), but this varied with minimum temperature and recorder type. Naïve occupancy was ~ 64% across a broad range of forests, which was at least five times more than expected based on previous surveys using alternative methods. After accounting for imperfect detection, probability of occupancy was influenced by elevation (-ve), cover of important browse trees (+ve), landscape NDVI (+ve) and extent of recent wildfire (-ve, but minor effect). Elevation was the most influential variable, though the relationship was non-linear and low occupancy was most common at tableland elevations (> 1000 m). Neither occupancy nor bellow rate were influenced by timber harvesting intensity, time since harvesting or local landscape extent of harvesting or old growth. Extrapolation of occupancy across modelled habitat indicates that the hinterland forests of north-east NSW support a widespread, though likely low density koala population that is considerably larger than previously estimated. Retention forestry has a significant role to play in mitigating harvesting impacts on biodiversity, including for forest specialists, but localised studies are needed to optimise prescriptions for koalas.

Many ecologists are using acoustic monitoring to study animals and the health of ecosystems. Technological advances mean acoustic recording of nature can now be done at a relatively low cost, with minimal disturbance, and over long periods of time. Vast amounts of data are gathered yielding environmental soundscapes which requires new forms of visualization and interpretation of the data. Recently a novel visualization technique has been designed that represents soundscapes using dense visual summaries of acoustic patterns. However, little is known about how this visualization tool can be employed to make sense of soundscapes. Understanding how the technique can be best used and developed requires collaboration between interface, algorithm designers and ecologists. We empirically investigated the practices and needs of ecologists using acoustic monitoring technologies. In particular, we investigated the use of the soundscape visualization tool by teams of ecologists researching endangered species detection, species behaviour, and monitoring of ecological areas using long duration audio recordings. Our findings highlight the opportunities and challenges that ecologists face in making sense of large acoustic datasets through patterns of acoustic events. We reveal the characteristic processes for collaboratively generating situated accounts of natural places from soundscapes using visualization. We also discuss the biases inherent in the approach. Big data from nature has different characteristics from social and informational data sources that comprise much of the World Wide Web. We conclude with design implications for visual interfaces to facilitate collaborative exploration and discovery through soundscapes.

Acoustic sensors allow scientists to scale environmental monitoring over large spatiotemporal scales. The faunal vocalisations captured by these sensors can answer ecological questions, however, identifying these vocalisations within recorded audio is difficult: automatic recognition is currently intractable and manual recognition is slow and error prone. In this paper, a semi-automated approach to call recognition is presented. An automated decision support tool is tested that assists users in the manual annotation process. The respective strengths of human and computer analysis are used to complement one another. The tool recommends the species of an unknown vocalisation and thereby minimises the need for the memorization of a large corpus of vocalisations. In the case of a folksonomic tagging system, recommending species tags also minimises the proliferation of redundant tag categories. We describe two algorithms: (1) a “naive” decision support tool (16%–64% sensitivity) with efficiency of O(n) but which becomes unscalable as more data is added and (2) a scalable alternative with 48% sensitivity and an efficiency ofO(log n). The improved algorithm was also tested in a HTML-based annotation prototype. The result of this work is a decision support tool for annotating faunal acoustic events that may be utilised by other bioacoustics projects.

Passive acoustic recording has great potential for monitoring soniferous endangered and cryptic species. However, this approach requires analysis of long duration environmental acoustic recordings that span months or years. There is a variety of approaches to analysing acoustic data. However, it is unclear which approaches are best suited for monitoring of endangered species in the wild. Specifically, this study is undertaking acoustic monitoring of the critically endangered White-bellied Heron (Ardea insignis) in Bhutan. Four different acoustic analysis methods are investigated in terms of their detection accuracy, involvement of human experts, and overall utility to ecologists for target species monitoring work. Our experimental results show that human pattern detection using a visualization technique has detection performance on par with a cluster-based recogniser, while a machine learning classifier implemented using the same acoustic features suffers from very low precision. Further, specific cases of false positives and false negatives by the different methods are investigated and discussed in terms of their overall utility for ecological monitoring. Based on our experimental results, we demonstrate how an integrated semi-automated approach of human visual pattern analysis with a recogniser is a robust system for acoustic monitoring of target species.

There is global concern about tropical forest degradation, in part, because of the associated loss of biodiversity. Communities and indigenous people play a fundamental role in tropical forest management and are often efficient at preventing forest degradation. However, monitoring changes in biodiversity due to degradation, especially at a scale appropriate to local tropical forest management, is plagued by difficulties, including the need for expert training, inconsistencies across observers, and lack of baseline or reference data. We used a new biodiversity remote-sensing technology, the recording of soundscapes, to test whether the acoustic saturation of a tropical forest in Papua New Guinea decreases as land-use intensity by the communities that manage the forest increases. We sampled soundscapes continuously for 24 hours at 34 sites in different land-use zones of 3 communities. Land-use zones where forest cover was fully retained had significantly higher soundscape saturation during peak acoustic activity times (i.e., dawn and dusk chorus) compared with land-use types with fragmented forest cover. We conclude that, in Papua New Guinea, the relatively simple measure of soundscape saturation may provide a cheap, objective, reproducible, and effective tool for monitoring tropical forest deviation from an intact state, particularly if it is used to detect the presence of intact dawn and dusk choruses.

Audio recording is a convenient and important method for large-scale terrestrial environmental monitoring. However, it is impossible to listen and make sense of all the data collected. Attempts to generalise automated analysis tasks have not been successful due to the unconstrained nature of long-term environmental recording. Our approach to this big-data challenge is to facilitate visualisation of long-term audio recording, to keep ecologists in the loop. The content of long-duration audio recordings are visualised by calculating acoustic indices. Our interface facilitates the customised visualisation and navigation of long-term audio recording by ecologists. Two case studies, one in Australia and one in Bhutan, are presented as examples.

Long-duration recordings of the natural environment have many advantages in passive monitoring of animal diversity. Technological advances now enable the collection of far more audio than can be listened to, necessitating the development of scalable approaches for distinguishing signal from noise. Computational methods, using automated species recognisers, have improved in accuracy but require considerable coding expertise. The content of environmental recordings is unconstrained, and the creation of labelled datasets required for machine learning purposes is a time-consuming, expensive enterprise. Here, we describe a visual approach to the analysis of environmental recordings using long-duration false-colour (LDFC) spectrograms, prepared from combinations of spectral indices. The technique was originally developed to visualize 24-hour “soundscapes.” A soundscape is an ecoacoustics concept that encompasses the totality of sound in an ecosystem. We describe three case studies to demonstrate how LDFC spectrograms can be used, not only to study soundscapes, but also to monitor individual species within them. In the first case, LDFC spectrograms help to solve a “needle in the haystack” problem—to locate vocalisations of the furtive Lewin’s Rail (Tasmanian), Lewinia pectoralis brachipus. We extend the technique by using a machine learning method to scan multiple days of LDFC spectrograms. In the second case study, we demonstrate that frog choruses are easily identified in LDFC spectrograms because of their extended time-scale. Although calls of individual frogs are lost in the cacophony of sound, spectral indices can distinguish different chorus characteristics. Third, we demonstrate that the method can be extended to the detection of bat echolocation calls. By converting complex acoustic data into readily interpretable images, our practical approach bridges the gap between bioacoustics and ecoacoustics, encompassing temporal scales across three orders of magnitude. Using the one methodology, it is possible to monitor entire soundscapes and individual species within those soundscapes.

Acoustic recordings play an increasingly important role in monitoring terrestrial and aquatic environments. However, rapid advances in technology make it possible to accumulate thousands of hours of recordings, more than ecologists can ever listen to. Our approach to this big-data challenge is to visualize the content of long-duration audio recordings on multiple scales, from minutes, hours, days to years. The visualization should facilitate navigation and yield ecologically meaningful information prior to listening to the audio. To construct images, we calculate acoustic indices, statistics that describe the distribution of acoustic energy and reflect content of ecological interest. We combine various indices to produce false-color spectrogram images that reveal acoustic content and facilitate navigation. The technical challenge we investigate in this work is how to navigate recordings that are days or even months in duration. We introduce a method of zooming through multiple temporal scales, analogous to Google Maps. However, the "landscape" to be navigated is not geographical and not therefore intrinsically visual, but rather a graphical representation of the underlying audio. We describe solutions to navigating spectrograms that range over three orders of magnitude of temporal scale. We make three sets of observations: 1. We determine that at least ten intermediate scale steps are required to zoom over three orders of magnitude of temporal scale, 2. We determine that three different visual representations are required to cover the range of temporal scales, 3. We present a solution to the problem of maintaining visual continuity when stepping between different visual representations. Finally, we demonstrate the utility of the approach with four case studies.

Acoustic recordings play an increasingly important role in monitoring terrestrial environments. However, due to rapid advances in technology, ecologists are accumulating more audio than they can listen to. Our approach to this big-data challenge is to visualize the content of long-duration audio-recordings by calculating acoustic indices. These are statistics which describe the temporal-spectral distribution of acoustic energy and reflect content of ecological interest. We combine spectral indices to produce false-color spectrogram images. These not only reveal acoustic content but also facilitate navigation. An additional analytic challenge is to find appropriate descriptors to summarize the content of 24-hour recordings, so that it becomes possible to monitor long-term changes in the acoustic environment at a single location and to compare the acoustic environments of different locations. We describe a 24-hour 'acoustic-fingerprint' which shows some preliminary promise.

Monitoring the environment with acoustic sensors is an effective method for understanding changes in ecosystems. Through extensive monitoring, large-scale, ecologically relevant, datasets can be produced that can inform environmental policy. The collection of acoustic sensor data is a solved problem, the current challenge is the management and analysis of raw audio data to produce useful datasets for ecologists. This paper presents the applied research we use to analyze big acoustic datasets. Its core contribution is the presentation of practical large-scale acoustic data analysis methodologies. We describe details of the data workflows we use to provide both citizen scientists and researchers practical access to large volumes of ecoacoustic data. Finally, we propose a work in progress large-scale architecture for analysis driven by a hybrid cloud-and-local production-grade website.

High quality bilingual dictionaries are rarely available for lower-density language pairs, especially for those that are closely related. Using a third language as a pivot to link two other languages is a well-known solution, and usually requires only two input bilingual dictionaries to automatically induce the new one. This approach, however, produces many incorrect translation pairs because the dictionary entries are normally are not transitive due to polysemy and the ambiguous words in the pivot language. Utilizing the complete structures of the input bilingual dictionaries positively influences the result since dropped meanings can be countered. Moreover, an additional input dictionary may provide more complete information for calculating the semantic distance between word senses which is key to suppressing wrong sense matches. This paper proposes an extended constraint optimization model to inducing new dictionaries of closely related languages from multiple input dictionaries, and its formalization based on Integer Linear Programming. Evaluations indicated that the proposal not only outperforms the baseline method, but also shows improvements in performance and scalability as more dictionaries are utilized.

Environmental monitoring is becoming critical as human activity and climate change place greater pressures on biodiversity, leading to an increasing need for data to make informed decisions. Acoustic sensors can help collect data across large areas for extended periods making them attractive in environmental monitoring. However, managing and analysing large volumes of environmental acoustic data is a great challenge and is consequently hindering the effective utilization of the big dataset collected. This paper presents an overview of our current techniques for collecting, storing and analysing large volumes of acoustic data efficiently, accurately, and cost-effectively.

Acoustic sensing is a promising approach to scaling faunal biodiversity monitoring. Scaling the analysis of audio collected by acoustic sensors is a big data problem. Standard approaches for dealing with big acoustic data include automated recognition and crowd based analysis. Automatic methods are fast at processing but hard to rigorously design, whilst manual methods are accurate but slow at processing. In particular, manual methods of acoustic data analysis are constrained by a 1:1 time relationship between the data and its analysts. This constraint is the inherent need to listen to the audio data. This paper demonstrates how the efficiency of crowd sourced sound analysis can be increased by an order of magnitude through the visual inspection of audio visualized as spectrograms. Experimental data suggests that an analysis speedup of 12× is obtainable for suitable types of acoustic analysis, given that only spectrograms are shown.