O21- Miniature recorders embedded in radio transmitters help to elucidate behaviours and inform monitoring practices in a New Zealand endemic flightless bird

Alberto De Rosa^1x, Daryl Olsen², Isabel Castro^1x, Stephen Marsland^4x

^x AviaNZ – Making Sure New Zealand Birds Are Heard – http://www.avianz.net
¹ Massey University, Palmerston North, Aotearoa New Zealand
² Kiwitrack Ltd, Havelock North, Aotearoa New Zealand
³ Victoria University of Wellington, Wellington, Aotearoa New Zealand

We describe the development of a miniature acoustic recorder that can
be fitted into the radio transmitter of the North Island Brown Kiwi
(NIBK | Apteryx mantelli, Bartlett), a New Zealand native nocturnal,
ground dwelling bird. Our aim is to obtain sufficiently accurate data
to determine whether this species produces recognisable individual
calls. As well as providing useful information about behaviour, this
will also assist in the establishment of a link between between the
number of vocalisations detected and the number of calling birds
present.
We have equipped birds from two high density populations with these
‘microrecorders’ and deployed traditional ARUs in both their resting
and foraging areas. By comparing the vocal activity of individuals with
the community level we can inform acoustic based population density
estimates, providing a
method to establish both the number of silent birds and the
distribution of vocalisations over the vocal individuals. Employing
these devices on a larger scale, including multiple kiwi populations of
known different densities could eventually lead to (passive) acoustic
population density estimates, which would be useful for conservation
decisions.

O22-  Sound event detection in long-term recordings using changepoints

• Julius Juodakis, Stephen Marsland, and Nirosha Priyadarshani
• School of Mathematics and Statistics, Victoria University of Wellington,Wellington, NZ

Passive acoustic recordings have emerged as an efficient means to monitor populations of vocalizing species, and also provide unique data for studying animal behaviour in natural environments. However, the processing of such data is currently limited by the ability to efficiently and reliably detect target sounds, which are often sparse in long-term recordings and overshadowed by environmental noise. We present a new sound event detector based on changepoint theory.
A wavelet pre-filter is used to extract a set of frequency bands from the recordings. The filtered data is then analysed to localise any changes in power. Transient signal changes are separated from longer shifts in the background noise, using a maximum length bound. Training requires only a small amount of data to set this bound and other parameters. In contrast to existing methods, this framework allows us to establish theoretical guarantees of the detector’s accuracy and computational complexity. New statistical and algorithmic developments to support these claims are presented.
We applied the method to acoustic surveys of two bird species in New Zealand: Australasian bittern and little spotted kiwi. The detections by the proposed and reference methods were reviewed, and used in a population inference method (spatial capture-recapture). Compared to commonly-used detectors, the proposed method consistently produced fewer false alarms, leading to two-fold higher precision of the inferred population size. This shows that survey data can be analysed more efficiently if the workflow design is driven by the ecological goals. Potential application examples for other types of sounds are
demonstrated as well.

O23- Individual recognition in rifleman, Acanthisitta chloris: Machine learning classification helps identify individual bird calls from audio recordings

Moran, I.G., Biological Sciences, University of Auckland, Auckland, New Zealand, Young, N., Centre for eResearch, University of Auckland, Auckland, New Zealand, Loo, Y.Y., Biological Sciences, University of Auckland, Auckland, New Zealand, Cain, K.E., Biological Sciences, University of Auckland, Auckland, New Zealand.

Machine learning classification is increasingly used to identify individual bird vocalizations from audio recordings. Machine learning classifiers capable of recognizing individual vocalizations are essential tools to answer questions about animal vocal behaviour such as seasonality of vocalizations and vocal interactions between individuals. In our study, we created machine learning classifiers to recognize individual birds in hundreds of hours of recordings. Our aim is to understand how rifleman recognize each other’s vocalizations. We hypothesized that if individual vocal recognition is accurately performed by machine learning algorithms, then similarly, birds should be using similar vocal parameters to differentiate between one another. We first developed and trained a machine learning classifier using zebra finches vocalizations. We then collected audio recordings from wild rifleman, a basal passerine species endemic to New Zealand whose vocal behaviour at nests is not well known. We found that our trained classifier successfully discriminated between individual zebra finches. We found that random forest models based on Mel-frequency Cepstral Coefficients (MFCC) performed better than Keras Models which are call-type independent models working regardless of what call type zebra finches were making. We predicted that our classifier will perform equally on discriminating between individual rifleman’s nest calls. In order to gain new insight into animal vocal recognition and vocal behaviour, machine learning classifiers are a tool that can help identify individual bird calls.

O24-  Diversity favours the old: metrics of avian diversity increase in aging regrowth Acacia woodlands of semi-arid eastern Australia.

• Doohan, B. School of Biology & Environmental Science, Queensland University of Technology, Brisbane, Australia.
• Fuller, S. School of Biology & Environmental Science, Queensland University of Technology, Brisbane, Australia.

Understanding how native fauna use regrowth vegetation is critical because of increased land clearing rates and biodiversity loss, yet it remains poorly studied in Australia’s semi-arid regions. This study used acoustic sensors to monitor avian diversity in three different age classes (new regrowth <15 years, intermediate regrowth 15-30 years, and old growth >30 years) of Acacia dominated, predominately mulga (Acacia aneura) woodlands in south-west Queensland. Manual analysis of 300 random minutes from 15 days of dawn recordings per site made during 2017 revealed that species richness, functional diversity and phylogenetic diversity increased with time since last clearance, with statistically significant differences between new regrowth and old growth. Each age class had unique species, yet intermediate regrowth and old growth shared a large number of species suggesting a convergence in species composition. Overall, results of this study show that while old growth vegetation sustains the highest level of biodiversity, intermediate and new regrowth still support a range of bird species. Therefore, regrowth mulga vegetation should be protected as it represents important habitat for avian biodiversity in semi-arid Australia.

In a follow-up study in the three age classes, recordings collected during the same time period over three years (2017-2019) were analysed using 18 acoustic indices. GLMMs revealed that three indices (mid frequency cover, high frequency cover and acoustic complexity) showed some relationship with avian species richness. However, modelling revealed that random effects (site and time) explained more variation than species richness. Remote sensing greenness data (soil adjusted vegetation index) varied over the three years reflecting local climatic changes and ephemeral conditions. Further studies are required to determine whether ecological factors, such as shifts in avian community composition and functional groups, insect activity, and phenology, are drivers of the patterns observed.

O25- The soundscape patterns of urban spaces with varying degrees of connectivity in metropolitan Seoul

Soyeon Chae
Interdisciplinary Program of EcoCreative, Ewha Womans University, Seoul 03760, R Korea

Urban green spaces differ in size as well as in connectivity. Theories predict that species richness may increase with size or connectivity of habitats, but the effects of size and connectivity on soundscape are not well studied. Using GIS and Graphab, we selected 4 sites with four classes of area (1000, 10000, 100000, over 1000000 m2) and four classes of connectivity, as well as habitat type, in metropolitan Seoul. To record soundscapes of these sites, we used Audiomoth 24h twice a month unitl may to august in 2019. We measured 10 α indices (spectral entropy, acoustic complexity index, acoustic evenness index, acoustic richness index, normalized difference soundscape index, acoustic entropy index, bioacoustics index, acoustic diversity index, amplitude index, temporal entropy) for each 24h recording. The acoustic complexity was low in sites with low connectivity. Size was a significant factor for α indices only between the biggest sites and the other lower classes sites. Thus, patterns of α indices during daytime were similar among sites with three lower classes of size. Our findings show that connectivity is more critical for determining patterns of soundscape in relatively small urban green spa

O26- Biotremology as a new tool of ecoacoustics

Imane Akassou¹,² Marco Ciolli², Valerio Mazzoni¹

¹Research and Innovation Centre, Fondazione Edmund Mach, San Michele all’Adige (TN), Italy
²Department of Civil, Environmental and Mechanical Engineering, University of Trento, Trento, Italy

Biotremology is the study of the production, perception, and transmission of mechanical vibrations through a substrate. Research in the field of biotremology has contributed to improve the understanding of insect behavior, ecology, and evolution. However, due to the technical challenges encountered in field trials, less research is devoted to investigating the interactions between insect vibrational signaling and the environment. We developed and tested an approach that could be implemented as a monitoring tool of insect vibrational signals occurring in an agroecosystem and evaluated the effect of environmental factors on the vibrational signaling of the insect community composing its vibroscape. The approach consisted of recording vibrational signals throughout the day and in different parts of an organic vineyard. Results show that the signaling activity was highly influenced by environmental factors. High temperature and wind velocity represented unsuitable conditions for insects which therefore significantly reduced their signaling activity. The approach allowed us to determine the daily signaling pattern of the two vineyard pests Scaphoideus titanus and Halyomorpha halys and the spatial occurrence of their signals. Our conclusion is that biotremology techniques could be profitably used, as a new tool of ecoacoustics, to monitor not only quantitative information but also biodiversity associated to insect vibrational signaling in ecosystems. In particular, in agroecosystems, this method could be employed to compare the environmental quality of cultivated areas at different management systems.

O27- Sonosfera – An ecoacoustic theatre for science-based listening experiences

Monacchi, D.
Department of New Technologies and Music Languages, State Conservatory G. Rossini, Pesaro, Italy

Several field recording campaigns have been conducted, over the past 19 years, in undisturbed primary equatorial rainforests in Amazon, Africa and Borneo, within the scope of the project “Fragments of Extinction – the sonic heritage of ecosystems”, generating complex three-dimensional data sets for archival, ecoacoustic analysis and public sharing. Due to the extreme hi-definition and space-preservative characteristics of the multichannel data collected, these soundscape recordings of animal communities required an increasingly sophisticated technology to be then properly rendered to audiences. The Eco-acoustic Theatre has been developed as a concept and instrument to fill this gap and enable science-based listening experiences of entire ecosystems.

From the first project in 2006, a series of different realizations of permanent and mobile spaces (S.P.A.C.E. Soundscape Projection Ambisonics Control Engine at Conservatorio G. Rossini, IT; Mobile Space at ECSITE-MUSE, IT; Soundscape Theatre at Naturama Natural History Museum, DK) have constituted the engineering basis for the construction of the Sonosfera (Sonosphere), a functional theatre for ecoacoustics which optimizes all passive-acoustics and electroacoustic characteristics, employing the most advanced custom-built audio technologies. This mobile 16 tons theatre is a 10-meter diam. amphitheatre which hosts 60 people at the centre of a 45 loudspeakers array and an ultra hi-definition 360° screen for scientific visualizations.

The Sonosfera® opened in December 2019 to the public in Pesaro UNESCO Creative City of Music – Italy; it represents a unique spherical ‘machine’ to enable both sensorial perception of complex sonic habitats and visualization of spatial and spectral characteristics of ecosystems.

O28- Polycultures show higher acoustic diversity than monocultures in a Panamanian tree diversity experiment.

• Sandra Müller, Geobotany, Faculty of Biology, University of Freiburg, Freiburg, Germany
• Linda Oschwald, Geobotany, Faculty of Biology, University of Freiburg, Freiburg, Germany
• Michael Scherer-Lorenzen, Geobotany, Faculty of Biology, University of Freiburg, Freiburg, Germany
• Catherine Potvin, Department of Biology, McGill University, Montreal, QC, Canada & Smithsonian Tropical Research Institute, Panama City, Panama

The analysis of soundscapes offers easy, rapid and sustainable methods for assessing biodiversity. Recently the quantification of regional or global acoustic variability in sounds and the analysis of soundscapes has evolved into an important tool for biodiversity conservation, especially since case studies confirm relationships between land-use management, forest structure and acoustic diversity. Tree diversity experiments like in Sardinilla, Panama examine the influence of tree species richness on various facets of diversity and ecosystem functioning. Diurnal sound recordings made in dry and wet seasons, combined with the experimental design of different tree species mixtures, offer possibilities to also investigate acoustic diversity as a function of diurnal and temporal patterns as well as vegetation structure. In the frame of this study, different acoustic indices were calculated to investigate acoustic diversity within (α-diversity) and between different tree species mixtures (β-diversity). Acoustic α-diversity was highest at night, dusk and dawn, which confirmed former studies stating dusk and dawn as being acoustically richest, and could be related to vocalizing insect species being most active at night and dominating the vocalizing community in Sardinilla. The dry season was more acoustically diverse than the wet season, which might be due to the dominance of insects rather than amphibians. During all seasons and day phases, except for day-time recordings during the dry season, monocultures showed a lower acoustic diversity than polycultures (2-4 tree-species-mixtures). This aligns with findings that monocultures had lower forest structural diversity than polycultures. Beta-diversity was highest between monocultures and polycultures. In conclusion, this study confirms that biodiversity, as measured by acoustic diversity, of tropical plantations can be substantially increased by planting two or more tree-species instead of monocultures.

O29- Assessing the presence and species richness of owls and woodpeckers through bioacoustics in two differently managed Alpine forests

Alessia Portaccio(1), Andrea Favaretto(2), Thomas Campagnaro(1), Giovanni Trentanovi(1), Tommaso Sitzia(1)

1) Department of Land, Environment, Agriculture and Forestry, Università degli Studi di Padova, V.le dell’Università 16, 35020 Legnaro, PD (Italy)

2) Via De Leva 2, 35128, Padova, Italy

Birds present large differences in species-specific responses to forest features such as structure and composition, and therefore they act as an indicator of forest ecosystem biodiversity. Woodpeckers and owls, specifically, are considered important forest naturalness indicators and their presence reflects those of numerous other forest-dwelling species. Because of their habits and behaviour, they are more difficult to detect than the passerine species, and therefore the implementation of the bioacoustics discipline is suited to the survey of such taxa. In the present study we collected data about the presence and species richness of owls and woodpeckers in relation to the main forest structure and composition variables (basal area, dominant height, tree species diversity, living trees’ vegetational conditions, canopy closure percentage, total deadwood volume, decomposition stage) in two forest sites which share similar ecological characteristics, but differ in terms of management: Cajada (non-intensively managed) and Tovanella (abandoned). Both Cajada and Tovanella forests report the presence of bird species belonging to the taxa of woodpeckers and owls, which are indicators of the forest at late-successional stages. In Cajada we contacted woodpeckers and owls a significantly higher number of times than in Tovanella, and the decomposition stage of deadwood and the level of canopy closure seem to the main explanatory factors of such results. Since management practices in Tovanella have been only recently abandoned we think that our results might change in the future. Therefore, further research is needed, also to better assess how sustainable forest management might conserve key forest features which are crucial for the thriving of most demanding owls and woodpeckers.

O30- Does noise matter ? Passive Acoustic Monitoring reveals the co-occurring presence of two threatened sympatric vocal species (Sciaena umbra and Umbrina cirrosa; Sciaenidae) in highly anthropized Venice inlets

Marta Picciulin1*, Chiara Facca1, Marta Bolgan2, Matteo Zucchetta3, Riccardo Fiorin4, Federico Riccato4 and Stefano Malavasi1

¹Department of Environmental Sciences, Informatics and Statistics, Ca’ Foscari; University of Venice, Venice, Italy
²Laboratoire de Morphologie Fonctionnelle et Evolutive, Institut de Chimie, Université de Liège, Belgium
³Institute of Polar Sciences, ISP-CNR, Venice-Mestre, Italy;
⁴Laguna Project S.N.C., Venice, Italy

Human impacts on marine ecosystems are accelerating, and the number of fish species listed in the Red List is growing. In the Mediterranean Sea, seven of the 10 bony fishes defined as Threatened by the IUCN are known to be vocal, including the shi drum (Umbrina cirrosa) and the brown meagre (Sciaena umbra). As a result, non-invasive Passive Acoustic Monitoring (PAM) can be used to pinpoint their distribution at sea. This is of particular interest since these species are closely related, elusive, sympatric, vocal sciaenids.

During summer 2019, three PAM surveys were conducted on a total of 40 listening points along the three tidal inlets of the Venice lagoon (Italy), which connect the open sea to one of the principal ports of the Northern Adriatic Sea. Here, the calls of both species have been recognized according to their temporal features: shi drum sounds were made up of a lower number of longer pulses with a different envelope, repeated at a lower rate than those of the brown magre. Call discrimination highlighted a partially overlapping distribution of the two species, inhabiting these highly anthropized inlets. Furthermore, S. umbra was found to emit longer sounds, with a higher number of faster repeated pulses during the chorus; these sound features are related to spawning activities in captive Sciaenids and were therefore used as proxy of spawning events in the study area. A cluster analysis based on S. umbra vocalizations separated the listening points in three areas; the areas in which vocal activity was highest were also characterized by the highest noise levels and number of vessel passages. This indicates that S. umbra spawning grounds are located in the noisier areas of the inlets, despite vessel noise is known to affect the efficiency of fitness-related behaviors. Results are discussed in a conservation perspective.

O31- Data mining soundscapes for Orthoptera monitoring: challenges and opportunities

Riede, K., Zoological Research Museum Alexander Koenig, Bonn, Germany

The majority of acoustically signalling species are insects. Orthoptera often dominate tropical soundscapes : most o f the 7,720 known katydids (Tettigonioidea ) and 5,780 crickets (Grylloidea) stridulate [1]. Adding about 4,000 short horned grasshoppers singing mainly in grasslands, there is an estimated total of 17,500 sound producing Orthoptera, plus hitherto undescribed species. Ecoacoustic data sets are a most valuable data source for Orthoptera bioacoustics: once properly archived, they could be (re)used for detection of threatened species or data mining for hitherto unidentified Orthoptera song patterns. Here I present some general features of Orthoptera songs in complex soundscapes from Borneo and South America, including ultrasound recordings made with programmable audiomoth recorders. Up to 20 Orthoptera signals were detected in a single sound track, from overall sets of about 100 insufficiently identified Orthoptera species at the respective localities. In contrast, soundscapes from Greece allow species determination thanks to reliable, well curated song libraries and a well-known limited set of species. It is evident that these data are highly relevant for Redlist Threat Assessment, monitoring of conservation measures and species discovery.
Taking into account the high number of ecoacoustic projects worldwide, there is a huge potential for datamining Orthoptera songs, if ecoacoustic datasets would be freely accessible. While bioacoustic sound archives can be federated via Global Biodiversity Information Facility ( www.gbif.org ) protocols, this is much more difficult for ecoacoustic datasets, due to their size and complexity. I describe requirements for data management such as stable URLs and metadata annotation within snippets extracted from soundscape recordings, which could serve as a first step towards an acoustic file interchange protocol, resulting in improved data sharing and efficient re use.
[1] Higher classification and species numbers from Cigliano, M.M., H. Braun, D.C. Eades & D. Otte. Orthoptera Species File . Version 5.0/5. 0 9/02/2020 ] http://Orthoptera.SpeciesFile.org

O32-  Using acoustic monitoring to measure the effect of forest management intensity on bird and bat communities

• Taylor Shaw, Geobotany, Faculty of Biology, University of Freiburg, Freiburg, Germany
• Raluca Hedes, Chair of Wildlife Ecology and Management, Faculty of Environment and Natural Resources, University of Freiburg, Freiburg, Germany
• Grzegorz Mikusiński, School for Forest Management, Swedish University of Agricultural Sciences SLU, Skinnskatteberg, Sweden

Resident birds in boreal forests can serve as indicators of habitat quality and are often species of conservation interest, particularly in multifunctional forests also used for timber production. It would therefore be useful to establish reliable methods for monitoring their presence and activity during the critical winter season and to evaluate the degree to which structural features in forest patches provide habitats useful for their winter survival. In establishing these methods, we will have efficient ways to identify patches of high conservation value for birds. We employed three monitoring approaches in 19 sites in a Swedish boreal forest landscape to identify the optimal method for monitoring resident birds in winter. We conducted a vegetation survey, traditional point count surveys and collected automated acoustic recordings from December – February, 2019. First we directly compared species richness values derived from point count and bioacoustic monitoring methods. Bioacoustic species identification yielded additional metrics of bird activity that point counts cannot: the number of cumulative bird identifications per site (No. Visits), number of observed flocking events per site (No. Flocks), and the number of recordings containing multiple species (non-flocks) simultaneously (No. Multiple Birds). We tested the response of all point count and bioacoustic metrics to variables of structural heterogeneity and complexity. Lastly, using an ecoacoustic approach, we calculated six of the most common acoustic indices and tested if any could effectively reflect the relationships between bird activity and vegetation structure described above. This is the first winter acoustic study to monitor bird assemblages in detail; it employed both bioacoustic and multi-index ecoacoustic approaches, and the results we will present provide strong evidence that automated acoustic recording can be an effective and superior method for monitoring resident forest birds in winter, providing a high-resolution ability to identify links between bird diversity and different components of structural complexity.

O33- Trade-off between song complexity and colorfulness in parid birds

• Tietze, D.T., Center of Natural History, Universität Hamburg, Hamburg, Germany
• Hahn, A., Center of Natural History, Universität Hamburg, Hamburg, Germany
• Johansson, U.S., Department of Zoology, Swedish Museum of Natural History, Stockholm, Sweden

Complex singing as well as plumage coloration of male birds are honest signals for potential partners and provide information about the males’ quality. To function as honest signals, both traits must be costly for the males. Due to limited resources, we expect a trade-off between the expression of both traits. This study researches the relation between song complexity and plumage coloration in tits (Paridae). These belong to the songbirds (Oscines) and show great variability in song and plumage coloration across species. For statistical analysis we implemented a phylogenetically generalized model of least squares containing potential explanatory variables. In the best model, body size besides colorfulness had a negative impact on song complexity: Large colorful tits sing less complex and vice versa. This result supports the hypothesis of a trade-off between costly traits and their likely intense signal function. This study contributes to the better understanding of how sexual selection influences diversification of traits.

O34-  The application of Direction Of Arrival methods for animal localisation

• David Warren Wallis
• Aarhus University, Denmark

Multilateration with Time Difference Of Arrival (TDOA) is the standard method for locating animals from their calls. An alternative method is to use compact microphone arrays, measuring only a few millimetres in size, to obtain Direction Of Arrival (DOA) measurements. Locations are found by triangulating angle measurements from two or more arrays. A ground location can be found from two arrays positioned on one edge of the monitored area. This is useful if the site has restricted access, for example one edge of a wetland area or the bank of a lake. There is particular benefit for locating flying animals (bats and birds). We have developed a two-dimensional array that measures azimuth and angle of elevation. A point in space can be found by triangulating the unit direction vectors recorded by each array. We will show models of the error estimates for two- and three-dimensional localisation methods. We will also introduce ‘Flightpath’, a web application that is being developed to process DOA recordings for animal localisation.