AI-assisted recognition for passive acoustic monitoring of birds in urban wetland parks

doi:10.17520/biods.2024188

Abstract

Abstract:

Aims: This study aims to explore the effectiveness of AI recognition-based passive acoustic monitoring of bird species in urban wetland parks, as well as to compare its results with traditional transect survey.

Methods: A three-month concurrent monitoring has been carried out from March to May in 2023 at Wanzuitou Wetland Park in Guangzhou City, China. The transect method involved a twice-monthly survey, while acoustic monitoring method utilized two acoustic monitoring devices operating in triggered recording mode throughout the day. Audio files were transmitted via a 4G mobile network to a server, and bird species were identified using an AI model based on the Pearl River Delta bird list, filtered by confidence scores and manually reviewed.

Results: The transect method recorded 2,200 individuals, whereas the acoustic monitoring collected 96,848 audio files, and obtained 34,117 valid records after screening and validation. Two methods identified a total of 70 bird species: 48 species by the transect survey and 49 species by the acoustic monitoring, with 27 species common to both survey methods.

Conclusions: The proportion of species overlapping between the two survey methods was less than half of the total species, suggesting that neither method can fully replace the other in this type of wetland park habitat. Transect survey is more accurate and makes it easier to estimate the population density, but it requires a higher level of bird species identification skills and involve more workloads. The acoustic monitoring can be automated and unmanned, making it easy to expand the scale of monitoring. However, processing the data from audio files is more challenging, and AI species identification results still need manual correction. The combination of traditional transect survey methods and AI recognition-based passive acoustic monitoring will provide higher accuracy and broader application prospects in the future.

Key words: bird diversity, transect survey method, passive acoustic monitoring, AI recognition model

Haotian Bai, Shang Yu, Xinyuan Pan, Jiale Ling, Juan Wu, Kaiqi Xie, Yang Liu, Xueye Chen. AI-assisted recognition for passive acoustic monitoring of birds in urban wetland parks[J]. Biodiv Sci, 2024, 32(8): 24188.

Add to citation manager EndNote|Ris|BibTeX

URL: https://www.biodiversity-science.net/EN/10.17520/biods.2024188

https://www.biodiversity-science.net/EN/Y2024/V32/I8/24188

Figures/Tables 8

References 46

[1]	Azeem M, Ali G, Amin RU, Babar ZUD (2022) Bird calls identification in soundscape recordings using deep convolutional neural network. In: Big Data Analytics and Computational Intelligence for Cybersecurity (eds Ouaissa M, Boulouard Z, Ouaissa M, Khan IU, Kaosar M), pp. 325-335. Springer, Cham.
[2]	Bian Q, Wang C, Cheng H, Han D, Zhao YL, Yin LQ (2023) Exploring the application of acoustic indices in the assessment of bird diversity in urban forests. Biodiversity Science, 31, 22080. (in Chinese with English abstract) DOI
	[边琦, 王成, 程贺, 韩丹, 赵伊琳, 殷鲁秦 (2023) 声学指数在城市森林鸟类多样性评估中的应用. 生物多样性, 31, 22080.] DOI
[3]	Bicudo T, Llusia D, Anciães M, Gil D (2023) Poor performance of acoustic indices as proxies for bird diversity in a fragmented Amazonian landscape. Ecological Informatics, 77, 102241.
[4]	Bino G, Steinfeld C, Kingsford RT (2014) Maximizing colonial waterbirds’ breeding events using identified ecological thresholds and environmental flow management. Ecological Applications, 24, 142-157.
[5]	Blumstein DT, Mennill DJ, Clemins P, Girod L, Yao K, Patricelli G, Deppe JL, Krakauer AH, Clark C, Cortopassi KA, Hanser SF, McCowan B, Ali AM, Kirschel ANG (2011) Acoustic monitoring in terrestrial environments using microphone arrays: Applications, technological considerations and prospectus. Journal of Applied Ecology, 48, 758-767.
[6]	Boelman NT, Asner GP, Hart PJ, Martin RE (2007) Multi‐trophic invasion resistance in Hawaii: Bioacoustics, field surveys, and airborne remote sensing. Ecological Applications, 17, 2137-2144.
[7]	Browning E, Gibb R, Glover-Kapfer P, Jones KE (2017) Passive acoustic monitoring in ecology and conservation. https://www.wwf.org.uk/sites/default/files/2019-04/Acousticmonitoring-WWF-guidelines.pdf. (accessed on 2024-05-16)
[8]	Budka M, Sokołowska E, Muszyńska A, Staniewicz A (2023) Acoustic indices estimate breeding bird species richness with daily and seasonally variable effectiveness in lowland temperate Białowieża forest. Ecological Indicators, 148, 110027.
[9]	Catchpole CK, Slater PJB (2003) Bird Song:Biological Themes and Variations. Cambridge University Press, Cambridge.
[10]	Chung NC, Miasojedow B, Startek M, Gambin A (2019) Jaccard/Tanimoto similarity test and estimation methods for biological presence-absence data. BMC Bioinformatics, 20 (Suppl. 15), 644.
[11]	Engen S, Grøtan V, Sæther BE (2011) Estimating similarity of communities: A parametric approach to spatio‐temporal analysis of species diversity. Ecography, 34, 220-231.
[12]	Farina A, Gage SH (2017) Ecoacoustics: The Ecological Role of Sounds. John Wiley & Sons, Hoboken.
[13]	Farina A, James P, Bobryk C, Pieretti N, Lattanzi E, McWilliam J (2014) Low cost (audio) recording (LCR) for advancing soundscape ecology towards the conservation of sonic complexity and biodiversity in natural and urban landscapes. Urban Ecosystems, 17, 923-944.
[14]	Gilbert NA, Clare JD, Stenglein JL, Zuckerberg B (2021) Abundance estimation of unmarked animals based on camera‐trap data. Conservation Biology, 35, 88-100.
[15]	Goyette JL, Howe RW, Wolf AT, Robinson WD (2011) Detecting tropical nocturnal birds using automated audio recordings. Journal of Field Ornithology, 82, 279-287.
[16]	Gupta G, Kshirsagar M, Zhong M, Gholami S, Ferres JL (2021) Comparing recurrent convolutional neural networks for large scale bird species classification. Scientific Reports, 11, 17085. DOI PMID
[17]	Hao ZZ, Zhang CY, Li L, Gao BT, Wu RC, Pei NC, Liu Y (2024) Anthropogenic noise and habitat structure shaping dominant frequency of bird sounds along urban gradients. iScience, 27, 109056.
[18]	He XL, Deng Y, Dong AR, Lin LX (2022) The relationship between acoustic indices, vegetation, and topographic characteristics is spatially dependent in a tropical forest in southwestern China. Ecological Indicators, 142, 109229.
[19]	Kahl S, Wood CM, Eibl M, Klinck H (2021) BirdNET: A deep learning solution for avian diversity monitoring. Ecological Informatics, 61, 101236.
[20]	Kasten EP, Gage SH, Fox J, Joo W (2012) The remote environmental assessment laboratory’s acoustic library: An archive for studying soundscape ecology. Ecological Informatics, 12, 50-67.
[21]	Kati VI, Sekercioglu CH (2006) Diversity, ecological structure, and conservation of the landbird community of Dadia Reserve, Greece. Diversity and Distributions, 12, 620-629.
[22]	Klingbeil BT, Willig MR (2015) Bird biodiversity assessments in temperate forest: The value of point count versus acoustic monitoring protocols. PeerJ, 3, e973.
[23]	Kumar Y, Gupta S, Singh W (2022) A novel deep transfer learning models for recognition of birds sounds in different environment. Soft Computing, 26, 1003-1023.
[24]	Leach EC, Burwell CJ, Ashton LA, Jones DN, Kitching RL (2016) Comparison of point counts and automated acoustic monitoring: Detecting birds in a rainforest biodiversity survey. Emu, 116, 305-309.
[25]	Lei FM (1999) A preliminary study of the role of bird songs in the systematic study of birds. Zoological Systematics, 24, 461-466. (in Chinese with English abstract)
	[雷富民 (1999) 鸟类鸣声在鸟类系统学研究中的作用初探. 动物分类学报, 24, 461-466.]
[26]	Lin TY, Goyal P, Girshick R, He K, Dollar P (2017) Focal loss for dense object detection. In:Proceedings of the IEEE International Conference on Computer Vision, pp. 2980-2988. IEEE, Honolulu.
[27]	Linke S, Gifford T, Desjonquères C, Tonolla D, Aubin T, Barclay L, Karaconstantis C, Kennard MJ, Rybak F, Sueur J (2018) Freshwater ecoacoustics as a tool for continuous ecosystem monitoring. Frontiers in Ecology and the Environment, 16, 231-238.
[28]	Meng L, Xu J, Tan X, Wang J, Qin T, Xu B (2021) Mixspeech:Data augmentation for low-resource automatic speech recognition. In: International Conference on Acoustics, Speech and Signal Processing (ICASSP 2021), pp. 7008-7012. ICASSP, Toronto.
[29]	Neate-Clegg MH, Horns JJ, Adler FR, Kemahlı Aytekin MÇ, Şekercioğlu ÇH (2020) Monitoring the world’s bird populations with community science data. Biological Conservation, 248, 108653.
[30]	Obrist MK, Pavan G, Sueur J, Riede K, Llusia D, Márquez R (2010) Bioacoustics approaches in biodiversity inventories. ABC Taxa, 8, 68-99.
[31]	Pérez-Granados C, Traba J (2021) Estimating bird density using passive acoustic monitoring: A review of methods and suggestions for further research. Ibis, 163, 765-783.
[32]	Ralph CJ, Droege S, Sauer JR (1995) Managing and monitoring birds using point counts: Standards and applications. United States Department of Agriculture (USDA) Forest Service General Technical Report, 49, 161-169.
[33]	Shamon H, Paraskevopoulou Z, Kitzes J, Card E, Deichmann JL, Boyce AJ, McShea WJ (2021) Using ecoacoustics metrices to track grassland bird richness across landscape gradients. Ecological Indicators, 120, 106928.
[34]	Slabbekoorn H, Smith TB (2002) Bird song, ecology and speciation. Philosophical Transactions of the Royal Society B: Biological Sciences, 357, 493-503.
[35]	Stephenson PJ (2018) A global effort to improve species monitoring for conservation. Oryx, 52, 412-413.
[36]	Stowell D, Wood MD, Pamuła H, Stylianou Y, Glotin H (2019) Automatic acoustic detection of birds through deep learning: The first bird audio detection challenge. Methods in Ecology and Evolution, 10, 368-380. DOI
[37]	Sueur J, Pavoine S, Hamerlynck O, Duvail S (2008) Rapid acoustic survey for biodiversity appraisal. PLoS ONE, 3, e4065.
[38]	Sugai LSM, Silva TSF, Ribeiro JW, Llusia D (2019) Terrestrial passive acoustic monitoring: Review and perspectives. BioScience, 69, 15-25. DOI
[39]	Tan M, Le Q (2021) EfficientNetV2: Smaller Models and Faster Training. https://proceedings.mlr.press/v139/tan21a.html. (accessed on 2024-09-25)
[40]	Thompson SJ, Handel CM, Mcnew LB (2017) Autonomous acoustic recorders reveal complex patterns in avian detection probability. The Journal of Wildlife Management, 81, 1228-1241.
[41]	Venier LA, Mazerolle MJ, Rodgers A, McIlwrick KA, Holmes S, Thompson D (2017) Comparison of semiautomated bird song recognition with manual detection of recorded bird song samples. Avian Conservation and Ecology, 12, 120202.
[42]	Villanueva-Rivera LJ, Pijanowski BC, Doucette J, Pekin B (2011) A primer of acoustic analysis for landscape ecologists. Landscape Ecology, 26, 1233-1246.
[43]	Wrege PH, Rowland ED, Keen S, Shiu Y (2017) Acoustic monitoring for conservation in tropical forests: Examples from forest elephants. Methods in Ecology and Evolution, 8, 1292-1301.
[44]	Zhang H, Cisse M, Dauphin YN, Lopez-Paz D (2017) mixup: Beyond empirical risk minimization. arXiv preprint, arXiv, 1710.09412.
[45]	Zheng GM (2012) Ornithology, 2nd edn. Beijing Normal University Publishing Group, Beijing. (in Chinese)
	[郑光美 (2012) 鸟类学(第2版). 北京师范大学出版社, 北京.]
[46]	Zuberogoitia I, Martínez JE, González-Oreja JA, de Buitrago CG, Belamendia G, Zabala J, Laso M, Pagaldai N, Jiménez-Franco MV (2020) Maximizing detection probability for effective large-scale nocturnal bird monitoring. Diversity and Distributions, 26, 1034-1050.

声景指数 Acoustic indices	计算公式 Formula	说明 Explanation	参考文献 Reference
声学多样性指数 Acoustic diversity index (ADI)	$ADI=\underset{i=1}{\overset{S}{\mathop \sum }}\,{{P}_{i}}\ln {{P}_{i}}$	P_i为第i个频段内信号强度高于阈值-50 dB的信号比例, S为频段数量 P_i is the proportion of signals in the i-th band with signal strength above the threshold -50 dB, S is the number of bands	Villanueva-Rivera et al, 2011
声熵指数 Acoustic entropy index (H)	H=H_f× H_t	H_f为频谱熵, H_t为时间熵 H_f is spectral entropy, H_t is temporal entropy	Sueur et al, 2008
归一化声景指数 Normalized difference soundscape index (NDSI)	$NDSI=\frac{{{B}_{s}}-{{A}_{s}}}{{{B}_{s}}+{{A}_{s}}}$	B_s为生物声, 一般集中在2-11 kHz,; A_s为人工声; 一般集中在1-2 kHz B_s is biophony, generally concentrated in 2-11 kHz; A_s is anthropophony; generally concentrated in 1-2 kHz	Kasten et al, 2012
生物声学指数 Bioacoustic index (BI)	$BI=\underset{i=1}{\overset{t}{\mathop \sum }}\,\left( {{S}_{i}}-{{S}_{\min }} \right)\times \Delta f$	S_i为子带内信号强度, ∆f为子带宽度, S_min为子带内最小信号强度 S_i is the signal intensity within the sub-band, ∆f is the sub-band width, and S_min is the minimum signal intensity within the sub-band	Boelman et al, 2007
Jaccard相似性指数Jaccard similarity index (J)	$J=\frac{\left\| A\bigcap B \right\|}{\left\| A\bigcup B \right\|}$	A为样线调查物种集合, B为声学调查物种集合 A is the transect survey species set, B is the acoustic monitoring species set	Chung et al, 2019
Sørensen相似性指数Sørensen similarity index (D)	$D=\frac{2\cdot \left\| A\bigcap B \right\|}{\left\| A \right\|+\left\| B \right\|}$	A为样线调查物种集合, B为声学调查物种集合 A is the transect survey species set, B is the acoustic monitoring species set	Engen et al, 2011

声景指数 Acoustic indices	计算公式 Formula	说明 Explanation	参考文献 Reference
声学多样性指数 Acoustic diversity index (ADI)	$ADI=\underset{i=1}{\overset{S}{\mathop \sum }}\,{{P}_{i}}\ln {{P}_{i}}$	P_i为第i个频段内信号强度高于阈值-50 dB的信号比例, S为频段数量 P_i is the proportion of signals in the i-th band with signal strength above the threshold -50 dB, S is the number of bands	Villanueva-Rivera et al, 2011
声熵指数 Acoustic entropy index (H)	H=H_f× H_t	H_f为频谱熵, H_t为时间熵 H_f is spectral entropy, H_t is temporal entropy	Sueur et al, 2008
归一化声景指数 Normalized difference soundscape index (NDSI)	$NDSI=\frac{{{B}_{s}}-{{A}_{s}}}{{{B}_{s}}+{{A}_{s}}}$	B_s为生物声, 一般集中在2-11 kHz,; A_s为人工声; 一般集中在1-2 kHz B_s is biophony, generally concentrated in 2-11 kHz; A_s is anthropophony; generally concentrated in 1-2 kHz	Kasten et al, 2012
生物声学指数 Bioacoustic index (BI)	$BI=\underset{i=1}{\overset{t}{\mathop \sum }}\,\left( {{S}_{i}}-{{S}_{\min }} \right)\times \Delta f$	S_i为子带内信号强度, ∆f为子带宽度, S_min为子带内最小信号强度 S_i is the signal intensity within the sub-band, ∆f is the sub-band width, and S_min is the minimum signal intensity within the sub-band	Boelman et al, 2007
Jaccard相似性指数Jaccard similarity index (J)	$J=\frac{\left\| A\bigcap B \right\|}{\left\| A\bigcup B \right\|}$	A为样线调查物种集合, B为声学调查物种集合 A is the transect survey species set, B is the acoustic monitoring species set	Chung et al, 2019
Sørensen相似性指数Sørensen similarity index (D)	$D=\frac{2\cdot \left\| A\bigcap B \right\|}{\left\| A \right\|+\left\| B \right\|}$	A为样线调查物种集合, B为声学调查物种集合 A is the transect survey species set, B is the acoustic monitoring species set	Engen et al, 2011

	3月March	4月 April	5月 May	总计 Total
声学监测种数 Species number of acoustic monitoring	43	37	29	49
样线调查种数 Species number of transect survey	35	35	31	48
样线调查只次数 Individuals of transect survey	540	831	829	2,200
鸣声数 Number of sounds	14,479	13,760	5,795	34,034
独立事件数 Number of sound-independent events	3,922	3,606	2,058	9,586
总计种数 Total number of species	57	53	41	70
共同物种数 Number of shared species	21	19	19	27
声学监测独有物种数 Number of species unique to acoustic monitoring	22	18	10	22
样线独有物种数 Number of species unique to transect survey	14	16	12	21
Jaccard相似性指数 Jaccard similarity coefficient	0.37	0.36	0.46	0.39
Sørensen相似性指数 Sørensen similarity coefficient	0.54	0.53	0.63	0.56

	3月March	4月 April	5月 May	总计 Total
声学监测种数 Species number of acoustic monitoring	43	37	29	49
样线调查种数 Species number of transect survey	35	35	31	48
样线调查只次数 Individuals of transect survey	540	831	829	2,200
鸣声数 Number of sounds	14,479	13,760	5,795	34,034
独立事件数 Number of sound-independent events	3,922	3,606	2,058	9,586
总计种数 Total number of species	57	53	41	70
共同物种数 Number of shared species	21	19	19	27
声学监测独有物种数 Number of species unique to acoustic monitoring	22	18	10	22
样线独有物种数 Number of species unique to transect survey	14	16	12	21
Jaccard相似性指数 Jaccard similarity coefficient	0.37	0.36	0.46	0.39
Sørensen相似性指数 Sørensen similarity coefficient	0.54	0.53	0.63	0.56