Acoustic bird surveys

Using acoustic methods alongside standard surveys can provide the best of both worlds. It will enable the collection of both spatial information and data on non-vocalising species from the walkover surveys, combined with the longer-term data from a number of days and nights of acoustic recording. The benefits of using both methods together has been clearly demonstrated in numerous scientific studies.

The standard approach within the Bird Survey Guidelines is for six transect visits to be undertaken at a site between late March and early July. In some circumstances, depending on the needs of the survey, its habitat context, and scoping of particular species or groups, the addition of automated acoustic surveys may allow for a reduction in transect surveyor effort – particularly where it is considered that adding acoustic recording would provide a better overall dataset. However, as stated previously in these Guidelines, it is the responsibility of the consultant to justify and evidence such methodological decisions, based upon factors such as the species of interest, the habitats on site, the scale and nature of the proposed development, and the available desk study information. Surveyors should also take a precautionary and proportionate approach where there are uncertainties about species or habitats present on site, or where evidence cannot be presented to justify a reduction of survey effort.

When integrating acoustic and standard survey methods, it should also be ensured that both methods span the appropriate spatial and temporal requirements for the survey.  For example, it would not normally be appropriate to conduct transect surveys only at the start of the breeding season in March-April, and replace these with only acoustic methods in May and June. Similarly the use of acoustic methods should provide a representative sample of all relevant key habitat areas at a site, in the same way that a walked survey route would pass through appropriate locations at least once during a survey visit.

In conclusion, integration of acoustic and standard survey methods should always seek to expand the range and quality of bird data recorded at a site, and not simply be used to reduce surveyor effort or costs.

This section of the guidelines makes evidence-based recommendations (where possible) for aspects of acoustic study design, including survey effort considerations, how to deploy recording equipment, and how to deal with the recorded audio data. It is based on a draft protocol developed by Abrahams (2018), and also takes into account guidance set out in Metcalf (2023).  Example case studies from the scientific literature are given in support of each recommendation below – but these are not exhaustive reviews of each aspect of the guidance provided.

Recommendations:

As for any method, the seasonal and daily timing of acoustic surveys should be appropriate to the species/groups of interest. For generic breeding bird surveys, recorder deployments should include a minimum of two periods: firstly in April to mid-May, and then mid-May to end of June. For non-breeding species or assemblages, surveyor knowledge and desk study findings should inform an appropriate timing schedule.

A minimum of five days recording should be undertaken per deployment.

For breeding bird surveys, recordings should cover both dawn and dusk periods. This should cover a minimum of one hour before sunrise until two hours after, and an hour either side of sunset – for a total of five hours recording per day.

A time-sampling approach can be used to reduce data volume and analytical requirements, without significantly affecting the results. This should employ frequent short samples through the recording period, e.g. a one minute sample taken every five minutes.

Rationale

The recording and data volume requirements of any survey will vary depending on the project characteristics and the species concerned (Bayne et al. 2017). The seasonal programme and daily timing of recording both need to be considered, to maximise the long-term data capture benefits of automated recorders, whilst avoiding an overwhelming data mountain (Klingbeil & Willig 2015).

Discontinuous time-sampling recordings have been repeatedly shown to significantly reduce data volume with only minimal reductions in capture of species richness. For example, studies have demonstrated that a stratified ‘on-off’ time sampling programme (e.g. recording 1 minute in every 10), can capture comparable data to continuous recording, with consequent benefits in terms of battery life, data storage and processing time (La & Nudds 2016; Bayne et al. 2017).

Bird detection probability normally varies with time of the day, so recording times distributed throughout the day will sample the entire community most effectively (La & Nudds 2016). At present, most studies have focused recordings on the main dawn and evening chorus times. With prices reducing and availability of data storage increasing, continuous recording, that can be sub- sampled later in the processing stage, is a potentially realistic option for fieldwork.

Supporting evidence

Shaw et al. (2022) investigated the effort required to estimate bird species richness and composition in European forests. They compared sampling intensity for 1 minute files, in intervals from 1-in-3 (n = 20 per hour) to 1-in-60 minutes (n = 1 per hour). The highest species richness was with recordings at the highest intensity of one every 3 minutes. They also compared recording in a dawn period (1 hour before sunrise), a morning period (1 hour beginning 3 hours after sunrise), and a combined period including both day phases. Species richness was significantly higher when including both day phases compared to dawn alone, and was slightly higher in the morning compared to dawn (yielding 80% of recorded species). However, certain nocturnal/crepuscular species could only be observed in the dawn period.

Franklin et al. (2021) studied bird assemblages in eucalypt forests, and recommended an optimal survey method of five 20-minute sampling periods immediately following dawn for 2 days.

Metcalf et al. (2022) found that if a time-sampling approach is being used, then for the same total recording effort, shorter file lengths (i.e. one minute or less) are far better than longer recordings. For example, instead of 4 x 15 minute samples across a three-hour period, then 240 x 15 second samples produce faster species accumulation and greater species richness.

Cifuentes et al (2021) suggest that short recordings sampled throughout the survey period accurately represent acoustic patterns, with an optimal schedule of ten 1 minute samples per hour.

de Araújo et al (2021) found it was necessary to sample at least 55% of a 12 hour total recording time to obtain the full composition of bird species, while sampling only 20% focussed on the morning peak allowed detection of 81% of the species assemblage.

Wood et al (2021) showed from simulations and case studies that a 50% reduction in overall recording duration may only result in a 5%–17% decrease in the detected species richness of the assemblage present. In terms of outcome per unit of effort, there was an increase in performance for 3 days/continuous (12 hr total) and 3 days/every-other 5 minutes (6 hour) scenarios compared to 7 days/every other 5 minutes (14 hours) and 7 days/15 minutes per hour (7 hour) scenarios.

Franklin et al (2020) studied the completeness of forest bird assemblages detected in acoustic surveys, and found that the highest levels required 300 minutes of recordings.

Cook & Hartley (2018) compared a continuous five-minute audio recording with an intermittent sample, using 30 x ten-second samples from a 30 minute period to create a five-minute composite recording. A significantly greater number of species were detected with the latter method.

Watson (2017) reviewed 194 avian studies conducted between 2004 and 2016, and found that ~78% sampled for a total of 240 minutes or less per site survey.

Bayne et al. (2017) state that, for singing birds, deployment over several days results in higher detection and occupancy rates than using a single day. However, there are diminishing returns – with fewer benefits from month-long deployments in comparison to covering more locations.

Thompson et al. (2017) recommend that having more recordings of shorter duration (1-4 minutes) is more efficient than fewer recordings of 10 minutes duration.

La and Nudds (2016) found that morning-only acoustic recordings underestimated species richness, and that the greatest number of species per unit of sampling effort was detected with on-the-hour samples between 07:00 and 12:00, and at 21:00.

Both Furnas and Callas (2015) and Thompson et al. (2017) demonstrated how detection probability varies for different species with time of the day, such that recording throughout the day will sample the entire bird community more effectively.

Wimmer et al (2013) found that 120 one-minute recordings, randomly selected from the three hours following sunrise, was the most efficient way to detect the highest number of species from a five-day recording period.

Recommendations:

Recorder placement considerations will vary depending upon whether the overall breeding bird assemblage across a site is being assessed, or particular habitat areas are being targeted for key priority species (see the Guidelines section on species-specific surveys).

For a generic survey of breeding birds, recorders should be spatially deployed to adequately sample the range of key bird habitats present within the site, e.g. using a selected or stratified random sample. If only a limited number of recorders are available, then locations can be rotated across a site (increasing the number of deployments proportionally) to allow greater area coverage.

To avoid ‘double-counting’ from recording the same birds on more than one recorder, a minimum distance between sampling locations of 250 m should be applied.

Recorders should be located 1-2 m from the ground, on tripods, narrow poles, or trees <0.2 m diameter, avoiding branches/leaves around the unit as far as possible.

Rationale

For coverage of a site, the aim should be to sample across the range of the habitats and species of interest, with recorders placed to limit overlap of detection radii so that counts are independent (O’Donnell & Williams 2015). The effective radius of most recorders for passerines is in the region of 50 m, so a minimum separation distance of at least 100 m should be used (Yip et al. 2017).

As guidance, a 250 m spacing between recorder locations would provide 16 sampling locations/km², while 500 m spacing provides 4 sampling locations/km². These spacings are likely to be dense enough to provide a good level of survey data in varied vs homogenous sites, and are also likely to be relevant to the territory sizes of bird species of interest within ecological assessments. However, alternative separation distances between recorders could be used, depending on survey requirements – and the integration with standard survey methods.

To accommodate a large number of sampling locations with a limited number of available recorders, units can be rotated across the site, e.g. four recorders could each be deployed twice to cover eight sampling locations.  If this approach is taken, then the number of deployment periods needs to be increased to maintain the appropriate level of temporal coverage within the survey programme.

When placing recorders in the field, omnidirectional microphones should be used, located horizontally 1-2 m from the ground (or higher if security is an issue), and in a mounting position that does not block the field of sound or increase the levels of background noise from wind and water (Klingbeil & Willig 2015, La & Nudds, 2016).

Supporting evidence

Pérez-Granados et al (2019) simulated the requirements for recorder layout with Dupont’s Lark, and found that within a 1km² survey area and at a density of 10 males km^-2 , four recorders (= 500m spacing) were sufficient to detect species presence with 90% confidence.

Symes et al (2022) performed a simulation to test trade-offs between recording at more sites, or for longer durations, when total sampling time was the same. Adding locations resulted in more species per unit of analysis effort than adding more days, e.g. species detection saturated at 30 species when sampling in one location and at 41 species when the same sampling duration was spread was across ten sites.

Furnas & Bowie (2020) argue that effective survey design for passerine bird point counts, usually entails independent sampling locations at least 250 m apart. This guidance, intended to minimise the potential for double counting and spatial autocorrelation, is considered to apply well to the use of sound recorders.

Recommendations:

Recordings should be made as non-compressed .WAV files, with a sample rate of 48 kHz and 16-bit depth or greater.

If storage capacity and battery life are critical issues, e.g. for long-term deployments, then a lower sample rate, such as 24 kHz, may be used. This will still record the frequency range required to capture bird vocalisations, but will have a reduced sound quality.

Before deployment, ensure that hardware and software settings are recorded and standardised across all units.

Rationale

For good quality audio data, a non-compressed digital file format (i.e. .WAV rather than MP3) should be used. If possible, recordings should use a sample rate of 48 kHz and 16-bit depth (or greater). These settings will cover the entire audible range, producing detailed data on frequency and amplitude to produce clear spectrograms and analysis information. If, however, the study is focussed on particular target species with lower frequency calls, or if file size is critical, then a lower sample rate (e.g. 24 kHz) can be used to save on storage and battery life.

Supporting evidence

Darras et al. (2022) describes the Global Soundscapes Project database, where 56% of the 325 projects listed used a 48 kHz sample rate.

Alcocer et al. (2022) reviewed 35 acoustic index studies, and found that the most common (37%) sampling rate used was 44.1 kHz, and that 94% of the studies used the .WAV file format.

Darras et al. (2018) recommend recording all audible sound (i.e. with 44.1 or 48 kHz sample rate) in uncompressed audio file formats such as .WAV or FLAC.

Kahl et al (2021) developed the BirdNET analyzer algorithm for detecting and classifying bird vocalizations.  This uses a 48kHz sample rate for inputs to the software.

Recommendations:

At the start of each deployment, record the date/time, surveyor name, sampling location, recorder/microphone identifiers and settings. Photographs of location and set-up should be taken. Weather conditions during the survey period should also be recorded.

Many bioacoustic recorders embed metadata within the header of the .WAV sound files, and within each filename, potentially including parameters such as date, time, location coordinates, recorder identification code, and audio settings. This information should be retained alongside the audio data.

Project metadata should align with recognized guidance, such as Cherrill 2020.

Rationale

With each survey deployment, appropriate metadata including location, dates/times, habitat and equipment identifiers should be recorded. This can be done using paper/tablet, or by speaking into the microphone while the unit is recording, so the metadata becomes part of the recorded data itself. This background data is clearly needed to accurately organise and archive recordings, and can be used for any detailed analysis of how environmental characteristics determine the bird acoustic assemblage. It is also important to make acoustic data as comparable as possible across different surveys, allowing use in larger-scale monitoring projects and contributions to databases.

Supporting evidence

Oswald et al. (2022) recognise that the types of metadata required for a project may vary with the aims and scope of each study, but that it will commonly include information such as the time of the recordings, location, equipment and settings used, and weather conditions.

Cherrill (2020) sets out general principles for metadata management in reporting, sharing and archiving of ecological data. It does not specifically cover acoustic recordings, but includes many relevant recommendations for ecological metadata.

Roch et al (2016) shows how the effective collection, storage and analysis of acoustic data is dependent on consistent, organised and transparent metadata, which should include, at the minimum, details on the equipment deployment and environmental conditions.

Recommendations:

The minimum standard for a bird acoustic survey will commonly be a verified species list for the study site, with associated levels of vocal activity. This may be broken down into data for different recorder locations, and across sampling dates.

There are two simple approaches that can be taken to analyse and present the data from surveys – either: (i) report the number of detections identified for each species, or (ii) identify the presence/absence of each species in one minute audio samples and calculate the proportion of samples in which each species is recorded. Both of these outputs can be used to provide a summary of species observations per day or deployment period at each recorder location.

If using any automated recogniser or clustering process to identify species vocalizations, then a degree of manual verification is required, so that error rates can be checked and the quality of the recogniser can be properly assessed. This quality assurance process need not review all detections, but can use a suitable sub-sample of the complete dataset, or may focus on rare/unusual species in the context of the site or habitat being surveyed. This verification process, and its results, should be fully set out in the survey report.

Rationale

The analysis of data gained from acoustic recorders is perhaps the most difficult area in which to make standardised recommendations. A range of software is available to manipulate, view and analyse acoustic recordings, e.g. Kaleidoscope Pro, Raven, Audacity, BirdNET, BTO Acoustic Pipeline and packages in R (with use subject to licence as necessary). Some of software options allow the clustering or automated recognition of bird calls. However, much scientific research has also relied upon ornithologists listening to audio files and viewing spectrograms. At present, a human-supervised semi- automated process probably offers the best balance between accuracy of call classification and time required for analysis. Whichever method is used, the data analysis protocol should be fully described, and identification error rates calculated, providing metrics such as precision and recall if a recogniser has been used (Knight et al. 2017).

Data collected through acoustic monitoring techniques should be checked by experienced ornithologists, familiar with bird vocalisations and species distribution/behaviour.

Following survey completion, audio recordings, particular those of rare and priority species, should be stored to allow review at a later date. They should also be archived with county recorders/research organisations, or with online, open-source bird call repositories such as Xeno-Canto – unless the client explicitly requires that this is not done.

Supporting evidence

Pérez-Granados (2023) reviewed a number of published studies that used BirdNET to detect and classify bird sounds.  Amongst these, average precision (% detections correctly identified) usually ranged around 72–85%, and recall rate (% target species vocalizations detected) ranged around 33–84%. The review showed that the use of confidence score thresholds for classifications increased the percentage of detections correctly identified to species, but lowered the proportion of calls and bird species detected. 

Symes et al. (2022) recognise that bird acoustic data are currently often analyzed by humans who review  spectrograms on screen, however this is not efficient, and automated machine-learning approaches are being developed to replace this process for large datasets.

Cole et al. (2022) assessed the performance of BirdNET by comparing automated and manual classifications of 13 breeding bird species. They found that BirdNET correctly identified most bird species detected during manual bird identification, and concluded that BirdNET is suitable for annotating multispecies recordings for extended recording periods.

Toenies & Rich (2021) assessed the ability of BirdNET to correctly identify species, when employed with a data subsetting process for quality control. This process achieved a high rate of true positive species identifications and a misidentification rate of less than 4%, which compares well to misidentification rates of 6-22% recorded in studies of human analysts.

Shaw et al. (2022) compared species richness values in forest habitats derived from point counts and bioacoustic monitoring methods, with manual analysis of recordings being employed. Species richness was significantly higher from acoustic recordings than point counts, although only when the time spent analysing acoustic data exceeded the time spent conducting point counts. However, the bioacoustic method enabled additional metrics of bird activity, enabling insights not provided by point count data. For example, increased acoustic activity was related significantly with habitat structural complexity, indicating that automated recording is more effective in identifying high quality habitat patches than point count methods.

Perez-Granados & Traba (2021) reviewed studies that used acoustic recorders for estimating bird densities or bird abundance. The most common approach was to estimate the relationship between the number of vocalizations per recording time with bird density or bird abundance estimated in the field, with the result that 79% of studies showed agreement between estimates obtained by human surveyors and those by acoustic methods.

Abrahams, C. (2018). Bird Bioacoustic Surveys – developing a standard protocol. In Practice, 102(102), 20–23.

Alcocer, I., Lima, H., Sugai, L.S.M. and Llusia, D. (2022), Acoustic indices as proxies for biodiversity: a meta-analysis. Biol Rev, 97: 2209-2236. https://doi.org/10.1111/brv.12890

Alquezar, R. D., & Machado, R. B. (2015). Comparisons Between Autonomous Acoustic Recordings and Avian Point Counts in Open Woodland Savanna. The Wilson Journal of Ornithology, 127(4), 712–723. doi:10.1676/14-104.1

Bayne, E., Knaggs, M., & Solymos, P. (2017). How to Most Effectively Use Autonomous Recording Units When Data are Processed by Human Listeners. Bioacoustic Unit, University of Alberta and Alberta Biodiversity Monitoring Institute (p. 72). The Bioacoustic Unit.

Budka M, Jobda M, Szałański P, Piórkowski H (2022) Acoustic approach as an alternative to human-based survey in bird biodiversity monitoring in agricultural meadows. PLOS ONE 17(4): e0266557. https://doi.org/10.1371/journal.pone.0266557

Cherrill, A. (2020) Metadata Standards. Professional Guidance Series PGS9. Chartered Institute of Ecology & Environmental Management. Available at: https://cieem.net/wp-content/uploads/2021/10/Metadata_Standards_2020.pdf

Cifuentes, E., Vélez Gómez, J., & Butler, S. J. (2021). Relación entre índices acústicos, duración de las grabaciones y tiempo de procesamiento: una prueba metodológica. Biota Colombiana, 22(1), 26–35. https://doi.org/10.21068/c2021.v22n01a02

Cole, J.S., Michel, N.L., Emerson, S.A., Siegel, R.B. (2022) Automated bird sound classifications of long-duration recordings produce occupancy model outputs similar to manually annotated data, Ornithological Applications, 124(2), duac003, https://doi.org/10.1093/ornithapp/duac003

Cook, A., & Hartley, S. (2018). Efficient sampling of avian acoustic recordings: Intermittent subsamples improve estimates of single species prevalence and total species richness. Avian Conservation and Ecology, 13(1), art21. doi:10.5751/ACE-01221-130121

Darras, K., Furnas, B., Fitriawan, I., Mulyani, Y., & Tscharntke, T. (2018). Estimating bird detection distances in sound recordings for standardizing detection ranges and distance sampling. Methods in Ecology and Evolution, 2018, 1–11. doi:10.1111/2041-210X.13031

Darras, K., Batáry, P., Furnas, B. J., Grass, I., Mulyani, Y. A., & Tscharntke, T. (2019). Autonomous sound recording outperforms human observation for sampling birds: A systematic map and user guide. Ecological Applications, 29(6). doi:10.1002/eap.1954

de Araújo, C.B., Jardim, M., Saturnino, N.S.F. et al. The optimal listening period for an effective assessment of bird richness and composition: a case study of Neotropical forest. J Ornithol 162, 303–306 (2021). https://doi.org/10.1007/s10336-020-01812-6

Franklin, M., Major, R., Bedward, M. & Bradstock, R. (2020). Establishing the adequacy of recorded acoustic surveys of forest bird assemblages. Avian Conservation and Ecology, 15 (1), 1-16.

Franklin, M. J. M., Major, R. E., Bradstock, R. A., Franklin, M. J. M., Major, R. E., & Bradstock, R. A. (2021). How much survey effort is required to assess bird assemblages in fire-prone eucalypt forests using acoustic recorders? Wildlife Research. doi:10.1071/WR20099

Furnas, B.J. & Bowie, R.C.K. (2020). Design and analytical considerations for improving effectiveness of bird surveys that use autonomous sound recorders, Ostrich, 91:3, 271-273, DOI: 10.2989/00306525.2020.1788829

Furnas, B. J., & Callas, R. L. (2015). Using automated recorders and occupancy models to monitor common forest birds across a large geographic region. Journal of Wildlife Management, 79(2), 325–337. doi:10.1002/jwmg.821

Holmes, S. B., McIlwrick, K. A., & Venier, L. A. (2014). Using automated sound recording and analysis to detect bird species-at-risk in southwestern Ontario woodlands. Wildlife Society Bulletin, 38(3), 591–598. doi:10.1002/wsb.421

Kahl, S., Wood, C.M., Eibl, M. & Klinck, H. (2021). BirdNET: A deep learning solution for avian diversity monitoring. Ecological Informatics 61: 101236

Kevin F.A. Darras, Rodney Rountree, Steven Van Wilgenburg, Amandine Gasc, 松海 李, 黎君 董, Yuhang Song, Youfang Chen, & Thomas Cherico Wanger. (2022). Worldwide Soundscapes project meta-data (1.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7415473

Klingbeil, B. T., & Willig, M. R. (2015). Bird biodiversity assessments in temperate forest: The value of point count versus acoustic monitoring protocols. PeerJ, 3, e973. doi:10.7717/peerj.973

La, V. T., and Nudds, T. D. (2016). Estimation of avian species richness: biases in morning surveys and efficient sampling from acoustic recordings. Ecosphere 7(4): e01294.10.1002/ecs2.1294

Lankau, H.E. (2019). Bioacoustic Unit SongMeter (SM4) Maintenance Protocol. Bioacoustic Unit, University of Alberta and Alberta Biodiversity Monitoring Institute. Edmonton, Alberta. Available at: https://www.wildtrax.ca/home/resources/methods-and-protocols.html

Leach, E. C., Burwell, C. J., Ashton, L. A., Jones, D. N., & Kitching, R. L. (2016). Comparison of point counts and automated acoustic monitoring: Detecting birds in a rainforest biodiversity survey. Emu, 116(3), 305–309. doi:10.1071/MU15097

Metcalf, O.C., Barlow, J., Marsden, S., Gomes de Moura, N., Berenguer, E., Ferreira, J. and Lees, A.C. (2022). Optimizing tropical forest bird surveys using passive acoustic monitoring and high temporal resolution sampling. Remote Sens Ecol Conserv, 8: 45-56. https://doi.org/10.1002/rse2.227

Metcalf, O. et al. (2022). Good practice guidelines for long-term ecoacoustic monitoring in the UK. UK Acoustics Network.

Oswald, J.N.; Van Cise, A.M.; Dassow, A.; Elliott, T.; Johnson, M.T.; Ravignani, A.; Podos, J. (2022) A Collection of Best Practices for the Collection and Analysis of Bioacoustic Data. Appl. Sci. 12, 12046. https://doi.org/10.3390/ app122312046

Pérez-Granados, C. (2023). BirdNET: applications, performance, pitfalls and future opportunities. Ibis. https://doi.org/10.1111/ibi.13193

Pérez-Granados, C., & Traba, J. (2021). Estimating bird density using passive acoustic monitoring: A review of methods and suggestions for further research. Ibis, 163(3), 765–783. doi:10.1111/ibi.12944

Pérez-Granados, C., Gómez-Catasús, J., Bustillo-de la Rosa, D., Barrero, A., Reverter, M., & Traba, J. (2019). Effort needed to accurately estimate Vocal Activity Rate index using acoustic monitoring: A case study with a dawn-time singing passerine. Ecological Indicators, 107. doi:10.1016/j.ecolind.2019.105608

Roch, M. A., Batchelor, H., Baumann-Pickering, S., Berchok, C. L., Cholewiak, D., Fujioka, E., … Soldevilla, M. S. (2016). Management of acoustic metadata for bioacoustics. Ecological Informatics, 31, 122–136. doi:10.1016/j.ecoinf.2015.12.002

Shaw, T., Schönamsgruber, S.-R., Cordeiro Pereira, J. M., & Mikusiński, G. (2022). Refining manual annotation effort of acoustic data to estimate bird species richness and composition: The role of duration, intensity, and time. Ecology and Evolution, 12, e9491. https://doi.org/10.1002/ece3.9491

Shonfield, J., & Bayne, E. M. (2017). Autonomous recording units in avian ecological research: Current use and future applications. Avian Conservation and Ecology, 12(1)(1), 14. doi:10.5751/ACE-00974-120114

Symes, L. B., Kittelberger, K. D., Stone, S. M., Holmes, R. T., Jones, J. S., Castaneda Ruvalcaba, I. P., Webster, M. S., & Ayres, M. P. (2022). Analytical approaches for evaluating passive acoustic monitoring data: A case study of avian vocalizations. Ecology and Evolution, 12, e8797. https://doi.org/10.1002/ece3.8797

Tegeler, A. K., Morrison, M. L., & Szewczak, J. M. (2012). Using extended-duration audio recordings to survey avian species. Wildlife Society Bulletin. doi:10.1002/wsb.112

Thompson, S. J., Handel, C. M., & McNew, L. B. (2017). Autonomous acoustic recorders reveal complex patterns in avian detection probability. Journal of Wildlife Management, 81, 1228– 1241. https://doi.org/10.1002/jwmg.21285

Toenies, M. and Rich, L.N. (2021). Advancing bird survey efforts through novel recorder 471 technology and automated species identification. California Fish and Game. 107(2):56-70.

Turgeon, P. J., Van Wilgenburg, S. L., & Drake, K. L. (2017). Microphone variability and degradation: Implications for monitoring programs employing autonomous recording units. Avian Conservation and Ecology, 12(1), art9. doi:10.5751/ACE-00958-120109

Watson, D. M. (2017). Sampling effort determination in bird surveys: do current norms meet best-practice recommendations? Wildlife Research 44:183-193. https://doi.org/10.1071/WR16226

Wimmer, J., Towsey, M., Roe, P., & Williamson, I. (2013). Sampling environmental acoustic recordings to determine bird species richness. Ecological Applications, 23(6), 1419–1428. doi:10.1890/12-2088.1

Wood, CM, Kahl, S, Chaon, P, Peery, MZ, Klinck, H. Survey coverage, recording duration and community composition affect observed species richness in passive acoustic surveys. Methods Ecol Evol. 2021; 12: 885– 896. https://doi.org/10.1111/2041-210X.13571

Zwart, M. C., Baker, A., McGowan, P. J. K., & Whittingham, M. J. (2014). The use of automated bioacoustic recorders to replace human wildlife surveys: An example using nightjars. PLoS ONE, 9(7). doi:10.1371/journal.pone.0102770