UNIVERSIDADE ESTADUAL PAULISTA “JÚLIO DE MESQUITA FILHO” INSTITUTO DE BIOCIÊNCIAS – RIO CLAROunesp PROGRAMA DE PÓS-GRADUAÇÃO EM ECOLOGIA, EVOLUÇÃO E BIODIVERSIDADE PREDICTING BIRD DIVERSITY USING ACOUSTIC INDICES WITHIN THE ATLANTIC FOREST BIODIVERSITY HOTSPOT LUCAS PACCIULLIO GASPAR Rio Claro – SP 2021 UNIVERSIDADE ESTADUAL PAULISTA “JÚLIO DE MESQUITA FILHO” INSTITUTO DE BIOCIÊNCIAS – RIO CLAROunesp PROGRAMA DE PÓS-GRADUAÇÃO EM ECOLOGIA, EVOLUÇÃO E BIODIVERSIDADE PREDICTING BIRD DIVERSITY USING ACOUSTIC INDICES WITHIN THE ATLANTIC FOREST BIODIVERSITY HOTSPOT LUCAS PACCIULLIO GASPAR Orientador: Prof. Dr. Milton Cezar Ribeiro Coorientador: Dr. Carlos Otávio Gussoni Dissertação apresentada ao Instituto de Biociências do Câmpus de Rio Claro, Universidade Estadual Paulista, como parte dos requisitos para obtenção do título de Meste em Ecologia, Evolução e Biodiversidade. Rio Claro – SP 2021 G249p Gaspar, Lucas Pacciullio Predicting bird diversity using acoustic indices within the Atlantic Forest biodiversity hotspot / Lucas Pacciullio Gaspar. -- Rio Claro, 2021 50 p. : tabs., fotos, mapas Dissertação (mestrado) - Universidade Estadual Paulista (Unesp), Instituto de Biociências, Rio Claro Orientador: Milton Cezar Ribeiro Coorientador: Carlos Otávio Gussoni 1. Monitoramento Acústico Passivo. 2. Ecologia Acústica. 3. Avaliação Acústica. 4. Etiquetamento Manual. 5. PELD. I. Título. Sistema de geração automática de fichas catalográficas da Unesp. Biblioteca do Instituto de Biociências, Rio Claro. Dados fornecidos pelo autor(a). Essa ficha não pode ser modificada. UNIVERSIDADE ESTADUAL PAULISTA Câmpus de Rio Claro Instituto de Biociências - Câmpus de Rio Claro - Avenida 24 A, , 1515, 13506900, Rio Claro - São Paulo https://ib.rc.unesp.br/#!/pos-graduacao/secao-tecnica-de-pos/programas/ecologia-e-biodiversidade/aprCNPJ: 48.031.918/0018-72. CERTIFICADO DE APROVAÇÃO TÍTULO DA DISSERTAÇÃO: Predicting bird diversity using acoustic indices within the Atlantic Forest biodiversity hotspot AUTOR: LUCAS GASPAR PACCIULLIO DA SILVA ORIENTADOR: MILTON CEZAR RIBEIRO COORIENTADOR: CARLOS OTÁVIO ARAUJO GUSSONI Aprovado como parte das exigências para obtenção do Título de Mestre em ECOLOGIA, EVOLUÇÃO E BIODIVERSIDADE, área: Zoologia pela Comissão Examinadora: Prof. Dr. CARLOS OTÁVIO ARAUJO GUSSONI (Participaçao Virtual) Autônomo / Rio Claro (SP) Prof. Dr. MARCO AURELIO PIZO FERREIRA (Participaçao Virtual) Departamento de Biodiversidade / UNESP - Instituto de Biociências de Rio Claro - SP Profa. Dra. ERICA HASUI (Participaçao Virtual) UNIFAL / Universidade Federal de Alfenas / MG Rio Claro, 07 de maio de 2021 AGRADECIMENTOS Agradeço primeiramente à força primária de toda criação pela existência e por experienciar as vidas na Terra, sempre no caminho da evolução. Força esta que por definição nunca terá explicação, uma vez que toda consequência possui uma causa e anterior a ela sempre há outra. Com isso, seguimos na busca infinita pelo conhecimento e questionamentos existenciais. Agradeço a toda minha família, especialmente à minha mãe, Ana Lúcia; ao meu pai Gaspar e; ao meu irmão Daniel. Por todos os anos, passados e futuros, impreterivelmente repletos de amor, carinho, respeito, esforço e superação. Minhas vivências, experiências e quem eu sou, são graças a muito suor e abnegação de cada um deles. Agradeço aos meus avós, Dona Mayre (Mayrinha) e Sr. Antônio pelo tamanho do amor transmitido. Exemplos exímios de respeito, trabalho e companheirismo. Mayrinha com sua caridade infinita e amor pelas plantas e Sr. Antônio, o coração mais puro que se pode encontrar. Obrigado pelos exemplos que levarei para toda vida guardados a sete chaves, em uma caixinha transparente para que todos possam ver, mas nunca retirá-los de mim. Aos meus tios e tias, Márcia, Luís, Marisa e Odair, que foram tão presentes na minha criação e em ensinamentos, tenho um intenso carinho. Agradeço aos meus primos e primas, Guilherme, Renato, Ana Lívia e Amanda. Sendo eu o caçula, foram meus companheiros desde os primeiros descobrimentos da vida até hoje. Possuem um lugar especial em meu coração. Agradeço a todos professores e mestres, no sentido mais amplo da palavra, que colaboraram tijolo por tijolo para construção do meu conhecimento e visão de mundo. Especialmente ao meu orientador de mestrado, Miltinho. Pela oportunidade e confiança. Por todas as conversas que por vezes são curtas, mas as reflexões permanecem por dias. Agradeço a ele também por todo conhecimento oculto passado através do seu jeito de existir e pela visão da ciência e da vida. Ao Carlos Gussoni (Pássaro) por toda a simpatia, receptividade e partilha do enorme conhecimento sobre as aves e da admiração pela natureza, além da coorientação nesta dissertação. Agradeço aos colegas, amigas e amigos do LEEC (Laboratório de Ecologia Espacial e Conservação) pelo companheirismo, conversas, ajudas, críticas, discussões e amizade. Especialmente Rafael Souza, Julia Assis, João Pena, André Regolin, Erison Monteiro, Maurício Vancine, Camila Priante, Vanesa Bejarano e Rodrigo Mataveli. Agradeço às parceiras da acústica, Marina Scarpeli, Eliziane Oliveira e Laura Honda por todas colaborações, confiança e apoio. Agradeço à Raíssa a motivação no dia-a- dia para o mestrado e forças no início da pandemia. Agradeço à turma passarinheira, Arthur Gomes, Rafaela Vitti, Rafaela Wolf e Silvia Harumi, pela colaboração essencial no etiquetamento das espécies deste trabalho, assim como a Paula Eveline, José Wagner e Samuli Laurindo, pela colaboração com os dados de anuros, que serão mais bem trabalhados em um futuro próximo. Agradeço à banca examinadora, Profa. Erica Hasui e Prof. Marco Pizo, assim como os suplentes, Prof. Carlos de Araújo e Dr. Alexandre Uezu pela honra de tê-los revisando e contribuindo com esta pesquisa. Agradeço a toda a equipe do projeto de extensão “Aves da minha Rio Claro”, pela partilha de momentos incríveis, passarinhando e incitando o amor e respeito pelo meio ambiente e pelas aves na população de Rio Claro. Agradeço todos servidores do Instituto de Biociências da Unesp de Rio Claro, ao STPG, especialmente à Ivana, pelo trabalho de excelência. O presente trabalho foi realizado com apoio da Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Código de Financiamento 001. “Princípio do Mentalismo: O todo é Mente; o universo é mental. Princípio da Correspondência: O que está em cima é como o que está embaixo. O que está dentro é como o que está fora. Princípio da Vibração: Nada está parado, tudo se move, tudo vibra. Princípio da Polariadade: Tudo é duplo, tudo tem dois pólos, tudo tem o seu oposto. O igual e o desigual são a mesma coisa. Os extremos se tocam. Todas as verdades são meias-verdades. Todos os paradoxos podem ser reconciliáveis. Princípio do Ritmo: Tudo tem fluxo e refluxo, tudo tem suas maŕes, tudo sobe e desce, o ritmo é a compensação. Princípio de Causa e Efeito: Toda causa tem seu efeito, todo o efeito tem sua causa, existem muitos planos de causalidade, mas nada escapa à Lei. Princípio de Gênero: "O Gênero está em tudo: tudo tem seus princípios Masculino e Feminino, o gênero manifesta-se em todos os planos da criação.” (O Caibalion - 1908) RESUMO A abordagem de índices acústicos em dados oriundos de monitoramento acústico passivo, torna possível sintetizar automaticamente informações de grande volume de dados. Desde o desenvolvimento destes índices, pesquisadores têm tentado entender o quão bem correlacionado é cada índice com os diferentes componentes da paisagem sonora, incluindo biofonias. Demonstramos que os índices acústicos estão mais bem correlacionados com o número de vocalizações de aves do que com a riqueza ou diversidade. Modelos que utilizam dois índices ao invés de apenas um, apresentou melhores correlações com a diversidade acústica de aves. Além disso, observamos que o tipo de ambiente tem influência nestas correlações. Portanto, defendemos o uso de índices acústicos para medir o nível e os padrões de atividade acústica das aves. No entanto, comparações entre pesquisas e inferências ecológicas realizadas a partir deste tipo de dado, precisam de cautela. Isso porque, ao variar o ambiente, também há variação na correlação entre os índices acústicos e os diferentes aspectos da diversidade de aves. Palavras chave: Monitoramento Acústico Passivo, Ecologia Acústica, Avaliação Acústica, Etiquetamento Manual, PELD. ABSTRACT Acoustic indices approach from passive acoustic monitoring data, makes it possible to automatically synthesize information from a large volume of data. Since the development of these indices, researchers have tried to understand how well correlated each index is with the different components of the soundscape, including biophonies. We demonstrated acoustic indices are better correlated with bird number of vocalizations than richness or diversity; models that use two indices instead of just one, have better correlations with bird acoustic diversity and; the environment type has influence in these correlations. So, we defend the use of acoustic indices to measure the level and patterns of birds' acoustic activity. However, the ecological inferences and comparisons made from this data type need caution, since when varying the environment, there is also variation in the correlation between acoustic indices and different aspects of bird diversity. Keywords: passive acoustic monitoring, ecoacoustics, acoustic surveys, manual labelling, long-term ecological research. Sumário 1. Introduction..........................................................................................................................................10 2. Material and Methods.........................................................................................................................13 2.1 Study area.....................................................................................................................................13 2.2 Experimental design.................................................................................................................... 14 2.3 Soundscape recordings.............................................................................................................. 15 2.3 Audio subset and species labeling............................................................................................15 2.5 Acoustic indices............................................................................................................................16 2.6 Models and Statistical analysis..................................................................................................16 3. Results..................................................................................................................................................19 3.1 Considering generalizations of environments..........................................................................19 3.2 Considering each environment separately...............................................................................20 4. Discussion............................................................................................................................................24 5. Conclusions and future directions....................................................................................................26 6. References...........................................................................................................................................28 10 1. Introduction The accelerated conversion of natural areas to anthropic land uses is causing local and regional population declines as well as species extinction. These changes affect both biodiversity and its related ecosystem services (Butchart et al., 2010; Johnson et al., 2017) ⁠ as well as important ecological functions that maintain the integrity of the ecosystems, such as seed dispersal and pollination (Duarte et al., 2018) ⁠ . In order to understand how humans affect biodiversity, ecosystem functions and services, we must expand biological monitoring both in time and space, using a wider taxonomic coverage, and yet cost efficient methods (Schmeller et al., 2017) ⁠ . Remote sensing techniques have been widely used aiming to increase sampling coverage and decrease costs. Also, it is non-invasive and allows large scale monitoring of sensitive and remote areas with difficult access. Acoustic remote sensing is usually done through Passive Acoustic Monitoring (PAM) and autonomous recording units (ARU - Shonfield and Bayne, 2017) ⁠ . ARU’s can be used in both terrestrial or marine environments, and might record data from a wide spatial range, capturing information from all directions up to 100 m away (Stowell and Sueur, 2020) ⁠ . The units may be set to record continuously and unattended for long periods of time (Stowell and Sueur, 2020), and are capable of recording soundscape information such as the biophony (sounds produced by animals - (Krause, 1987) ⁠ ; geophony (abiotic components of a landscape); and anthrophony (sounds produced by human activities and machines) (Pijanowski et al., 2011) ⁠ . Therefore, audio recordings can be used to obtain a large variety of ecological information on biodiversity, like activity patterns, composition of vocally active species and behavior (Farina et al., 2011) ⁠ , as well as on the quality and integrity of habitats (Gómez et al., 2018). ⁠ Environmental changes are known to modify bird diversity and composition and in fact birds have been used as bioindicators of environmental quality (Furness et al., 1993; Padoa- Schioppa et al., 2006) ⁠ . Besides, they occur in a great variety of environments and have their taxonomy well known, with extensive literature about behavior and responses to habitat changes. In addition, they are the group of animals which highly contribute to the biophony in soundscapes (Farina et al., 2011), as they vocalize for both inter and intra-specific communication, including 11 reproduction and territory defense (McGregor ·and Peake, 2000) ⁠ . In this way, birds are a good candidate for monitoring through PAM. PAM can collect data on a large spatial and temporal scales, making it possible to obtain species composition through manual detection (e.g. Wimmer et al., 2013) ⁠ . However, although this approach contains accurate information about the species, many other ecologically interesting questions can be answered using the entire volume of recordings obtained, such as large patterns of activity and management of the database. As a result, it is necessary to improve more automated analysis techniques in order to take full advantage of the information contained in the recordings and a macro view of standards (Sueur et al., 2014) ⁠ . Acoustic Indices (AI) are one automatic approach used to analyse environmental recordings. They consist of metrics that are calculated by using several spectral and temporal parameters of the sound recordings (Sueur et al., 2014) ⁠ . These indices are sensitive to signal characteristics such as amplitude variations and the frequency bands it occupies. Some indices were developed aiming to quantify biodiversity and soundscape patterns, based on the acoustic parameters of a soundscape recording (e.g. Villanueva-Rivera et al., 2011; Towsey et al., 2014; Scarpelli et al., 2020) ⁠ . Therefore, it is generally expected that better preserved environments will present higher levels of biophony and that those signals will be spread across different frequency bands, considering the natural selection acting in the acoustic niche (Farina et al., 2011; Araújo et al., 2020) ⁠ , despite being insensitive to species composition and functional diversity. But anyway, the acoustic variability can be used as proxies for biodiversity quantification (Sueur et al., 2014b) ⁠ . In terrestrial environments, AI are being used for rapidly assessing biodiversity (e.g. Sueur et al., 2008) ⁠ , describing habitat type and spatial heterogeneity (e.g. Bormpoudakis et al., 2013) ⁠ , characterizing soundscapes (e.g. (Towsey et al., 2014) ⁠ , quantifying anthrophony (e.g. Buxton et. al. 2017) ⁠ and monitoring protected areas (e.g. Campos et al., 2021) ⁠ . Since the development of AI, their effectiveness to predict biodiversity activity patterns from environmental recordings has been tested in areas with different vegetation structures (Buxton et al., 2018) ⁠ , climate (Eldridge et al., 2018) ⁠ , and species composition (Depraetere et 12 al., 2012) ⁠ . The level of success of translating AI on bird biodiversity estimation has varied between studies and explanatory power varies from low to high. As an example, studies that tested the correlation between bird richness and Acoustic Complexity Index (ACI - Pieretti et al., 2011) ⁠ have been presenting low positive correlation (Mammides et al., 2017; Jorge et al., 2018) ⁠ , moderate positive correlation (Eldridge et al., 2018) ⁠ , moderate negative correlation (Eldridge et al., 2018; Izaguirre et al. 2018; Shamon et al., 2021) ⁠ , and also no correlation (Dröge et al., 2021) ⁠ . In the Appendix S1 is presented as a table with previous results about other indices. Knowing that the AI summarizes characteristics of an audio, the divergent results of previous studies may be related to differences in the method of obtaining ecological indices of diversity and data treatment. In addition, different environments have different soundscape patterns, both in species composition and differences in the patterns of other signs, such as geophonies and anthropophonies. Because the patterns are not yet clear, there are still limitations and concerns on the broad use of acoustic indices as a proxy of biodiversity (Mammides et al., 2017) ⁠ . Therefore, understanding how the AI can be used as surrogate biodiversity measures at wide spatial scales is essential for its application in monitoring programs (Stowell and Sueur, 2020) ⁠ . We compared the relative contribution of AI commonly used to measure birds diversity in different types of environments. We used 7 acoustic indices (ACI, BIO, ADI, AEI, AR, Ht, and NDSI) to answer three questions: (1) What is the relative contribution of acoustic indices in explaining bird diversity?; (2) What are the explanatory power gains when combining acoustic indices (i.e bivariate models), compared to single indices (univariate models)?; (3) Does the type of environment have an influence on the strength of these correlations? We hypothesized that (1) bivariate models perform significantly better than univariate models, independent of response variables (Figure 1.A). This is due to the idea that the use of two indices in a complementary way may have a better capacity to capture different characteristics of the data. Regarding the expected differences in the relative contribution of each index in the bivariate models, we believe that some AI will have a certain redundancy between them when combined. (2) From the indices, ACI will have the higher explanatory power in univariate models due to its sensitivity in detecting amplitude variability and having resistance to constant noise, followed by BIO, ADI, AR and 13 NDSI (Figure 1.A); ADI will be negatively related with response variables, and the others will be positively related (Figure 1.B). In Figure 1.C we represent how bird diversity can behave in bivariate models with two hypothetical explanatory indices. (3) There will be a change in the relationship between AI and bird diversity among the three environments (forest, pasture and swamp). Due to structural differences and similarities, the model with greater explanatory strength for forest will be different, while for pasture and swamps the best model will be coincident. Figure 1 - Expected patterns for acoustic index (univariate or bivariate models) on explaining the bird diversity. A) explanatory power for some acoustic index alone, or combined with other indices; B) expected slopes between bird diversity and acoustic indices (the steeper the slope. the higher the relationship between bird diversity and the indices); c) expected patterns between bird diversity and bivariate acoustic models), where an AI would be a more explication contribution than others. To simplify, only some examples are presented here. 2. Material and Methods 2.1 Study area The study was conducted in northeastern São Paulo State and the southern portion of Minas Gerais State, Brazil (Figure 2). The region is the focus of a Long Term Ecological Research within the ecological Corridor of Cantareira-Mantiqueira (PELD CCM). This area is of great ecological importance as it connects two large Atlantic Forest remnants: the Cantareira State Park and the Serra da Mantiqueira mountain range (Boscolo et al., 2017) ⁠ . The Atlantic Forest is a biodiversity hotspot as it presents high levels of bird richness (620 species, Myers et. al 2000) and endemism (223 species, Vale et al., 2018) ⁠ , but also because it 14 has been highly deforested and fragmented (Laurance, 2009) ⁠ . The Atlantic Forest has had its forest cover reduced to about 28% of its original size (Rezende et al., 2018) ⁠ . The remaining patches are isolated (average distance is 1,440 m), small in size (84% has less than 50 ha), and are under severe edge effects (half of the remnants are less than 100 m from the forest edges - (Ribeiro et al., 2009) ⁠ . The main land cover types of the study area consist of forest, pasture, agriculture, forestry (mainly Eucalyptus plantation), swamps as well as urban and rural buildings (Barros et al., 2019; Boscolo et al., 2017) ⁠ . Figure 2 - Landscapes studied in the Cantareira-Mantiqueira ecological corridor, located in the Atlantic Forest biodiversity hotspot in southeastern Brazil. Zoom showing the three different environments (forest, swamp and pasture) sampled in each of the 22 landscapes varying in forest cover gradient (1% - 97%) in a 1km radius buffer. 2.2 Experimental design We collected audio data in 22 landscapes with varying forest cover (1 to 97%) within 1 km from the sampling site centroids (Figure 2). On each landscape we sampled three different 15 habitat types: forest, pasture and swamp. Habitats were sampled between October 2016 to January 2017 with the following sampling scheme: forests (October to November), swamps (November to December) and pastures (December to January), during 30 sequence days each. The forest sites were at least 50 meters away from the border to the interior of the forest; the pastures were mainly for cattle raising; and the swamps consisted of wetlands, in lower portions of the relief. The sampling period refers to the rainy season in this region, and it matches with the reproductive period of most birds within the Atlantic Forest (Develey and Peres, 2000) ⁠ . During the sampling period, the average temperature and precipitation were similar between landscapes, thus we assumed that the differences in months did not affect general biodiversity patterns. 2.3 Soundscape recordings We used Song Meter Digital Field Recorders (SM3; Wildlife Acoustics. Inc. Massachusetts) fixed on tree trunks 1.5 m above ground. The recorders were equipped with two omnidirectional microphones (frequencies 20 Hz to 20 kHz), and the sampling rate was 44.1 kHz, 16 bits and mono mode (to save storage space and battery). For the entire project and the complete sampling scheme we attained a total of 10,343 hours of recordings. The complete sampling effort of each landscape and site (i.e. habitat type) can be found in Appendix S2. Because this study focused only on diurnal birds, with the peak vocal activities in the first hours of the morning, we only used the samples from this period (05:00am until 08:25am). 2.3 Audio subset and species labeling We created a subset of the recordings by following four steps: (1) we selected five 25- minute files (05:00am, 05:45am, 06:30am, 07:15am, and 08:00am) per day (9,151 files, totalizing 228,775 minutes); 2) from each file we extracted two random minutes, totaling 18,594 minutes; 3) we grouped those minutes into nine packages of 300-random-minutes each (i.e. 2,700 minutes), being 100 minutes for forest, 100 minutes for swamps, and 100 minutes for pastures per package; 4) the packages were sent to bird experts, which labeled the occurrence of bird species on each minute (see scheme on Appendix S3). We detected 137 corrupted, blank minutes that were excluded from the analysis. So, in the end we evaluated 2,563 minutes. 16 Finally, we had 10,437 classified bird vocalizations from the labels, being 9,437 at species level, 192 at genus levels, and 808 not identified (distant signals or dubious calls). Recordings of unidentified calls were excluded from the analysis. We labeled each species once each minute, regardless of the number of times their vocalizations appeared in it, using the software Raven Pro v.1.5 (Center for Conservation Bioacoustics 2014). 2.5 Acoustic indices For each minute, we calculate eight AI that has been commonly used to measure bird diversity: (1) Acoustic Diversity Index (ADI); (2) Acoustic Evenness Index (AEI); (3) Acoustic Complexity Index (ACI); (4) Entropy Index (H); (5) Bioacoustic Index (BIO); (6) Acoustic Richness index (AR); (7) Temporal Entropy index (Ht); and (8) Normalized Difference Soundscape Index (NDSI) (description in the Table 1). As there is no consensus on which parameters are best to be followed, we used different frequency and amplitude threshold parameters, such as their combinations, to calculate the indices. We used Rstudio v.1.414 (R Development Core Team 2016) environment and the Seewave (Sueur et al., 2008) ⁠ (ACI and Ht) and Soundecology packages (Villanueva-Rivera et al., 2016) ⁠ (ADI, AEI, BIO, H, and NDSI) to perform calculations. AR was calculated using the formula (Eq. 1) due to errors in its automatic calculation from Seewave. No pre-processing techniques were applied prior to the calculation of indices as it is important to test how AI performs in different environmental conditions. (Eq.1) 2.6 Models and Statistical analysis The number of annotated minutes and number of tags varied between sampling points. Therefore, we decided to standardize the sampling effort prior to proceeding with response variable estimations. For this, we made a bootstrap procedure in which 100 sub-samples from the 17 2,563 annotated files were generated, considering the audio files of each sampling point. Each of these 100 sub-samples consisted of 1,300 random minutes totaling 6,500 groups of 20 minutes each. Then, we estimated three response variables: bird richness, number of tags, and diversity. In a sub-sample, for richness, we used the number of species noted. Tags number was considered the total of tags in a sub-sample. In order to measure diversity, we used the inverse of Simpson (Chao et al., 2010) ⁠ . The same groups of minutes (20 minutes) were used for calculating the average of each acoustic index. We used Generalized Linear Models (GLM’s) to test the relationship between bird diversity (species richness, number, and diversity of tags) and AI means as explanatory variables. The models were evaluated according to values of dAIC, wAIC, R² and correlation plots. The final models were tested for multicollinear variance inflation factors (VIF), being considered VIF < 10 (Montgomery and Peck 1992). The VIFs analysis selected the same set of acoustic indices (Table 2) to three responses variables (bird richness, number and diversity of tags). 18 Table 1 - Description of the acoustic indices used for the correlation test with acoustic diversity of diurnal birds within the region of the Long Term Ecological Research within the ecological Corridor of Cantareira-Mantiqueira (PELD CCM), Atlantic Forest, São Paulo, Brazil. Acoustic indices Description Reference Bioacoustic (BIO) A function between spectral amplitude of signals and the number of frequency bands occupied. It is an evaluation of the sound level and the amount of frequency bands used, measuring the abundance of birds. Boelman et al. (2007) ⁠ Acoustic Complexity (ACI) Since biological sounds have greater variability, whereas anthropogenic or geophonic sounds have less variation. Divides the spectrogram into multiple frequency bins and temporal subsets, and calculates the differences in the intensities of adjacent sounds, being a measure of variability on amplitude intensity. Pieretti et al. (2011b) ⁠ Acoustic Diversity (ADI) Divides the spectrogram into frequency bands and calculates the signal diversity in each of those bands similarly to a Shannon-Wiener index (Shannon and Weaver 1949). The higher the amount of species emitting sound, the more frequency bands occupied, and the higher the value of the index. Villanueva-Rivera et al. (2011) Acoustic Evenness (AEI) Divides the spectrogram into frequency bands and measures evenness using the Gini coefficient and is negatively related to ADI. Villanueva-Rivera et al. (2011) Acoustic Richness (AR) Based on Temporal Entropy (H) and the median of the overall amplitude (M). Uses the general median and Ht, developed for environments with a high signal-to-noise ratio. Depraetere et al. (2012) ⁠ Total Entropy (H) Estimated acoustic energy dispersion over the spectrum and it uses spectral and temporal amplitude patterns. The more species in a community, the more distinct signals will be produced. Sueur et al. (2008) ⁠ Temporal Entropy (Ht) Computed following the Shannon evenness index applied to the amplitude envelope of the temporal series and where the envelope points correspond to the categories. Sueur et al. (2008) ⁠ Normalized Difference Soundscape (NDSI) Estimates the level of anthropogenic disturbance on the soundscape by computing the ratio of human-generated (anthrophony) to biological (biophony). An estimate of the level of acoustic disturbance in a landscape. Kasten et al. (2012) ⁠ 19 Table 2 - Acoustic indices set selected over VIFs used in final models to test the correlation with bird richness, number and diversity of tags within the region of the Long Term Ecological Research within the ecological Corridor of Cantareira-Mantiqueira (PELD CCM), Atlantic Forest, São Paulo, Brazil. The “x” means absence of amplitude threshold. Acoustic indices Frequency range Amplitude threshold ADI 300 - 12,000 hz 75db AEI 1000 - 12,000 hz 50db ACI 1000 - 12,000 hz x BIO-1 1000 - 12,000 hz x BIO-2 1000 - 22,050 hz x NDSI-1 Antro: 300-1,000hz Bio: 1,000-12,000hz x NDSI-2 Antro: 1,000-2,000hz Bio: 2,000-22,050hz x Ht 0 - 22,050 hz x AR 0 - 22,050 hz x 3. Results We analyzed a total of 2,563 minutes. In total, we had 9,629 labels, being 9,278 at species level (199 species) distributed in 53 families and 351 at genus level (18 genus) (Appendix S4). The species with the highest number of detections were Vireo chivi (781 minutes), Zonotrichia capensis (631 minutes), Basileuterus culicivorus (533 minutes), Cyclarhis gujanensis (427 minutes), and Pitangus sulphuratus (371 minutes). 3.1 Considering generalizations of environments The acoustic indices models had great correlation with the number of tags (maximum R² = 0.48), followed by richness (maximum R² = 0.43), and less with tag diversity (maximum R² = 0.26). To all response variables, the additive model BIO1+ACI had the best performance (Figure 03; Table 3). The complete results of the models are in Appendix S5. Both in the models that considered the variable environment and in those that did not consider it, the additive models 20 with two acoustic indices, were the best correlations to three response variables. The dAICc results with big differences among the models were probably due to the use of big data as entered (e.g. 6,500 samples). So, because of this, we also used R² to obtain an idea about the explaining power of the models. Figure 3 - The best models from GLM analysis using data of all environments, within the region of the Long Term Ecological Research within the ecological Corridor of Cantareira-Mantiqueira (PELD CCM), Atlantic Forest, São Paulo, Brazil. 3.2 Considering each environment separately In general, different models had greater correlation not only for different environments, but also for different response variables. The best models and correlation power presented some variation among the data from the different environments. Swamp showed greaters results, with R² = 0.76 to richness (ADI+BIO1); R² = 0.67 to diversity (ADI+BIO1), and R² = 0.64 to number of tags (ACI+BIO1). Pasture got a number of tags R² = 0.43 (BIO1+AR), richness R² = 0.41 (BIO1+AR), and R² = 0.29 (BIO1+AR) to tag diversity. Finally, the forest environment showed results of R² = 0.50 about number of tags (BIO1+NDSI2), R² = 0.25 to richness (BIO1+ACI), and the smaller value to diversity, R² = 0.14 (NDSI1+Ht) (Table 4). The complete results are in the Appendix S6. 21 Table 3 - Four best models results for each response variable and null model. Model variables, dAICc, df (degrees of freedom), wAIC and Explanatory power (R²) of acoustic indices on explaining bird responses within the region of the Long Term Ecological Research within the ecological Corridor of Cantareira-Mantiqueira (PELD CCM), Atlantic Forest, São Paulo, Brazil. Bird responses are bird richness, number of bird tags and bird tag diversity. The acoustic indices after the symbol “~” were used in the models, which were ranked from the higher dAICc and R² values (top) to the lowest (botton) within bird responses. Detailed results are available in Appendix S6. GLMmodels [Delta]AICc df wAIC R² Richness~ACI + BIO1 + environment 0 6 1.000 0.43 Richness~BIO1 + AR + environment 474.96 6 << 0.0001 0.38 Richness~ADI + BIO1 + environment 513.07 6 << 0.0001 0.38 Richness~ACI + BIO1 605.74 4 << 0.0001 0.37 Richness~NULL 3622.95 2 0.000 0.00 Tags.Number~ACI + BIO1 + environment 0 6 1.000 0.48 Tags.Number~BIO1 + AR + environment 317.74 6 << 0.0001 0.46 Tags.Number~ADI + BIO1 + environment 466.09 6 << 0.0001 0.44 Tags.Number~AEI + BIO1 + environment 468.56 6 << 0.0001 0.44 Tags.Number~NULL 4274.76 2 0.000 0.00 Diversity~ACI + BIO1 + environment 0 6 1.000 0.26 Diversity~BIO1 + NDSI2 + environment 40.44 6 << 0.0001 0.25 Diversity~ACI + BIO1 53.97 4 << 0.0001 0.25 Diversity~BIO1 + NDSI2 67.93 4 << 0.001 0.25 Diversity~NULL 1921.78 2 0.000 0.00 To the pasture, both acoustic indices in the bivariate model had similar contributions in the explanation in the models, while in the forest data, to richness and number of tags both acoustic indices in the model were similar contributions (Figure 4). But to the diversity, Ht had much more contribution than ACI. In the swamp BIO had a greater contribution in the models than ADI to all responses variables. 22 Table 4 - Mainly models from GLM results to responses variables from each environment data. Model variables, dAICc, df (degrees of freedom), wAIC and Explanatory power (R²) of acoustic indices on explaining bird responses within the region of the Long Term Ecological Research within the ecological Corridor of Cantareira-Mantiqueira (PELD CCM), Atlantic Forest, São Paulo, Brazil. Bird responses are bird richness, number of bird tags and bird tag diversity. The acoustic indices after the symbol “~” were used in the models, which were ranked from the higher dAICc and R² values (top) to the lowest (botton) within bird responses. Detailed results are available in Appendix S6. Environment GLM models [Delta]AICc df wAIC R² Forest Richness~BIO1 + ACI 0 4 1.000 0.25 Richness~BIO1 + NDSI2 102.89 4 << 0.0001 0.21 Tags.Number~BIO1.NDSI2 0 4 1.000 0.50 Tags.Number~ACI + BIO1 365.63 4 << 0.0001 0.41 Diversity~NDSI1 + Ht 0 4 1.000 0.14 Diversity~ACI + Ht 19.39 4 << 0.0001 0.13 Pasture Richness~BIO1 + AR 0 4 1.000 0.41 Richness~ACI + AR 179.76 4 << 0.0001 0.35 Tags.Number~BIO1 + AR 0 4 1.000 0.43 Tags.Number~ACI + AR 109.97 4 << 0.0001 0.40 Diversity~BIO1 + AR 0 4 1.000 0.29 Diversity~AEI + BIO1 110.42 4 << 0.0001 0.25 Swamp Richness~ADI + BIO1 0 4 1.000 0.75 Richness~ACI + BIO1 162.05 4 << 0.0001 0.73 Tags.Number~ACI + BIO1 0 4 1.000 0.64 Tags.Number~AEI + BIO1 114.64 4 << 0.0001 0.62 Diversity~ADI + BIO1 0 4 1.000 0.67 Diversity~BIO1 + BIO2 79.91 4 << 0.0001 0.66 23 Figure 4 - The best models to each response variable (bird richness, number of tags and tags diversity) to each environment in separate within the region of the Long Term Ecological Research within the ecological Corridor of Cantareira-Mantiqueira (PELD CCM), Atlantic Forest, São Paulo, Brazil. Bird responses are bird richness, number of bird tags and bird tag diversity. All BIO correspond to the BIO1 parameters. 24 4. Discussion We investigated how seven most used different acoustic indices were related to the bird’s acoustic community in different environments of Atlantic Forest. We were able to demonstrate how the combination of two indices using models better explains diversity of vocalizations than single indices approach to all situations. Also, the environment type has influence in the correlations, being the best correlations were found in the swamps, pasture and forest, respectively. The swamp environment has found greaters results not only for richness (ADI+BIO) but also for the diversity (ADI+BIO) and number of tags (ACI+BIO). To pasture, the BIO+AR was the best model to the three responses variables. Finally, for forests, BIO+NDSI was the best model correlation with number of tags, while BIO+ACI and NDSI+Ht were similar to others, but with low correlation. The idea of a single index being unable to capture all facets of the diversity of vocalizations was discussed by (Sueur et al., 2014b) ⁠ by using some few studies (Eldridge et al., 2018; Towsey et al., 2014). Eldridge et al., 2018 discussed the combination of indices that use different strategies to measure aspects of biodiversity. Nevertheless, here we demonstrate that the two indices with the best combined performance measure the different facets of the diversity of birds tested (richness, tag abundance and diversity) were indices with similar theoretical backgrounds. (Towsey et al., 2014) ⁠ have found that bivariate models (ACI+Spectral diversity) were better than three indices models, even more than four indices models, and a low explanation was added with more indices about bird richness. Here, we used bivariate models, and found the ACI, BIO, AR and NDSI the greater indices, but the ADI, besides important, had less to explain contribution. The literature presents several inconsistencies in the relationship between AI and measures of bird diversity. This demonstrates that there is a long way to go for the widespread use of AI, mainly about what conclusions can be drawn from them. All these different patterns can be due to several reasons, such as differences in the land uses (e.g. Buxton et al., 2018) ⁠ , climate (e.g. (Eldridge et al., 2018) ⁠ , vocalization patterns (Zhao et al., 2019) ⁠ and acoustic signals diversity, like anthropogenic noises types (e.g. Ross et al., 2021). In addition, they can 25 also be caused by differences in obtaining ecological measures and data analysis. Here we saw that environments with structural differences and complexity (e.g. forest and pasture) showed different results. In addition, even environments with similar structure, that is, open vegetation (e.g. pastures and swamps) also differed between the results. Despite not being tested, this divergence may be related to the amount of anthropic activities, which tend to be larger closer to the pastures than in the swamp area, where the relief is more rugged and the soil is moist. Therefore, acoustic data obtained in different environments need to be treated separately to more accuracy, since the type of environment will have an influence on the relationship between AI and biodiversity. However, we showed the bivariate model BIO+ACI was the most appropriate model when we considered data from all environments together, however the environment variable was important in all models and not considering it, later analyzes may be superficial and mask the real standards. The differences can also be a result of ecological and biological processes. Understanding and explaining biodiversity patterns is usually the goal of the majority of studies using PAM and AI, but as there is still a lack of baseline it is hard to compare among different studies and be certain of what is causing variation. Another cause of inconsistency between studies similar to this one, may be related to differences in analyzes. Many of them vary between obtaining ecological variables, such as listening points in the field (e.g Mitchell et al., 2020; Bradfer- Lawrence et al., 2020; Mammides et al., 2017; Retamosa Izaguirre and Ramírez-Alán, 2018) ⁠ or manual detection of species in the recordings (e.g Buxton et al., 2018; Eldridge et al., 2018; Jorge et al., 2018; Machado et al., 2017; Moreno-Gómez et al., 2019; Shamon et al., 2021) ⁠ , selecting minutes and considering sample size and also, different statistical tests and presenting the results to test the correlations. In addition we believe that obtaining ecological variables from the manual detection of species in audio is important, since they will have exactly the same intentions used for calculating AI. We perform bootstrap sampling, that is, draws of sets of minutes with replacement. Thus, there is a standardization of the sampling effort of each collection point and also, different possible combinations of minutes, which decreases the temporal correlation in the data. In addition, for the selection of models we use GLM, considering mainly not only the values of dAIC but also of R². This is because, dAIC despite ordering the best models, for large samples 26 the results between the models tend to be discrepant. Since, from R² values, in addition to obtaining the best model, we can have an idea of how well related the variables are and the size of the explanation gain that is gained or lost between each model. In addition, based on R² values, it is possible to make more concrete comparisons between the different studies that seek to test this relationship in different contexts and possibilite real estimates about biodiversity numbers. As for the manual detection of species to obtain ecological variables, we use not only the most common richness, but also the amount and diversity of tags. We understand the limitation of the use of quantitative measures from acoustic data, however, it is to verify which aspects of acoustic activity certain AI are more sensitive or less. In addition, in view of the large amount of data collected and labeled species, we believe that it is necessary to move beyond just richness. Here we use the quantity of tags as a next for the quantity of vocalizations and from the richness and number of tags in a given sample we calculate the diversity of tags. In general, our results showed for each response variable there were differences not only in the correlations power, but also for the best correlated models. In the forest and pasture environment, the highest correlation of AI was with the number of tags. In a swamp environment, the greatest correlation was with richness, but also with high diversity values and number of tags. Therefore, in addition to differences in the correlations between AI and bird biodiversity in different environments, they are also correlated differently depending on the ecological variable involved. 5. Conclusions and future directions We demonstrated acoustic indices are better correlated with bird number of vocalizations than richness or diversity; models that use two indices instead of just one, have better correlations with bird acoustic diversity and; the environment type has influence in these correlations. We defend the use of acoustic indices to measure mainly the level and patterns of acoustic activity. However, the ecological inferences and comparisons made from this data type need caution, since when varying the environment, there is also variation in the correlation between acoustic indices and different aspects of bird diversity. This, mainly due to the absence of information regarding functional diversity, for example. Further studies are needed to test the acoustic indices in different conditions, considering the use of two or more indices for the analyzes. In addition, based on the correlations found, in fact perform prediction tests, relating the values of acoustic 27 indices and how much the estimated diversity of bird species in a given landscape would be. 28 6. References Araújo, C.B. de, Furtado, S.N.M., Vieira, G.H.C., Simões, C.R., 2020. O Nicho Acústico: Integrando a Física, Ecologia E Teoria Da Comunicação. Oecologia Aust. 24, 762–769. https://doi.org/10.4257/oeco.2020.2404.01 Barros, F.M., Peres, C.A., Pizo, M.A., Ribeiro, M.C., 2019. Divergent flows of avian-mediated ecosystem services across forest-matrix interfaces in human-modified landscapes. Landsc. Ecol. https://doi.org/10.1007/s10980-019-00812-z Boelman, N.T., Asner, G.P., Hart, P.J., Martin, R.E., 2007. Multi-Trophic Invasion Resistance in Hawaii: Bioacoustics, Field Surveys, and Airborne Remote Sensing, Source: Ecological Applications. Bormpoudakis, D., Sueur, J., Pantis, J.D., 2013. Spatial heterogeneity of ambient sound at the habitat type level: Ecological implications and applications. Landsc. Ecol. 28, 495–506. https://doi.org/10.1007/s10980-013-9849-1 Boscolo, D., Tokumoto, P.M., Ferreira, P.A., Ribeiro, J.W., Santos, J.S. dos, 2017. Positive responses of flower visiting bees to landscape heterogeneity depend on functional connectivity levels. Perspect. Ecol. Conserv. 15, 18–24. https://doi.org/10.1016/j.pecon.2017.03.002 Bradfer-Lawrence, T., Bunnefeld, N., Gardner, N., Willis, S.G., Dent, D.H., 2020. Rapid assessment of avian species richness and abundance using acoustic indices. Ecol. Indic. 115, 106400. https://doi.org/10.1016/j.ecolind.2020.106400 Butchart, S.H.M., Walpole, M., Collen, B., Van Strien, A., Scharlemann, J.P.W., Almond, R.E.A., Baillie, J.E.M., Bomhard, B., Brown, C., Bruno, J., Carpenter, K.E., Carr, G.M., Chanson, J., Chenery, A.M., Csirke, J., Davidson, N.C., Dentener, F., Foster, M., Galli, A., Galloway, J.N., Genovesi, P., Gregory, R.D., Hockings, M., Kapos, V., Lamarque, J.F., Leverington, F., Loh, J., McGeoch, M.A., McRae, L., Minasyan, A., Morcillo, M.H., Oldfield, T.E.E., Pauly, D., Quader, S., Revenga, C., Sauer, J.R., Skolnik, B., Spear, D., Stanwell-Smith, D., Stuart, S.N., Symes, A., Tierney, M., Tyrrell, T.D., Vié, J.C., Watson, R., 2010. Global biodiversity: Indicators of recent declines. Science (80-. ). https://doi.org/10.1126/science.1187512 Buxton, R.T., McKenna, M.F., Mennitt, D., Fristrup, K., Crooks, K., Angeloni, L. and Wittemyer, G., 2017 et al., 2017. Noise pollution is pervasive in U.S. protected areas. Science (80-. ). 356, 531–533. https://doi.org/10.1126/science.aah4783 29 Buxton, R.T., Agnihotri, S., Robin, V. V., Goel, A., Balakrishnan, R., 2018. Acoustic indices as rapid indicators of avian diversity in different land-use types in an Indian biodiversity hotspot. J. Ecoacoustics 2, GWPZVD. https://doi.org/10.22261/jea.gwpzvd Campos, I.B., Fewster, R., Truskinger, A., Towsey, M., Roe, P., Vasques Filho, D., Lee, W., Gaskett, A., 2021. Assessing the potential of acoustic indices for protected area monitoring in the Serra do Cipó National Park, Brazil. Ecol. Indic. 120, 106953. https://doi.org/10.1016/j.ecolind.2020.106953 Center for Conservation Bioacoustics. (2014). Raven Pro: Interactive Sound Analysis Software (Version 1.5) [Computer software]. Ithaca, NY: The Cornell Lab of Ornithology. Available from https://ravensoundsoftware.com Chao, A., Chiu, C.H., Jost, L., 2010. Phylogenetic diversity measures based on Hill numbers. Philos. Trans. R. Soc. B Biol. Sci. 365, 3599–3609. https://doi.org/10.1098/rstb.2010.0272 Depraetere, M., Pavoine, S., Jiguet, F., Gasc, A., Duvail, S., Sueur, J., 2012. Monitoring animal diversity using acoustic indices: Implementation in a temperate woodland. Ecol. Indic. https://doi.org/10.1016/j.ecolind.2011.05.006 Develey, P.F., Peres, C.A., 2000. Resource seasonality and the structure of mixed species bird flocks in a coastal Atlantic forest of southeastern Brazil. J. Trop. Ecol. 16, 33–53. https://doi.org/10.1017/S0266467400001255 Dröge, S., Martin, D.A., Andriafanomezantsoa, R., Burivalova, Z., Fulgence, T.R., Osen, K., Rakotomalala, E., Schwab, D., Wurz, A., Richter, T., Kreft, H., 2021. Listening to a changing landscape: Acoustic indices reflect bird species richness and plot-scale vegetation structure across different land-use types in north-eastern Madagascar. Ecol. Indic. 120, 106929. https://doi.org/10.1016/j.ecolind.2020.106929 Duarte, G.T., Santos, P.M., Cornelissen, T.G., Ribeiro, M.C., Paglia, A.P., 2018. The effects of landscape patterns on ecosystem services: meta-analyses of landscape services. Landsc. Ecol. 33, 1247–1257. https://doi.org/10.1007/s10980-018-0673-5 Eldridge, A., Guyot, P., Moscoso, P., Johnston, A., Eyre-Walker, Y., Peck, M., 2018. Sounding out ecoacoustic metrics: Avian species richness is predicted by acoustic indices in temperate but not tropical habitats. Ecol. Indic. 95, 939–952. https://doi.org/10.1016/j.ecolind.2018.06.012 Farina, A., Lattanzi, E., Malavasi, R., Pieretti, N., Piccioli, L., 2011. Avian soundscapes and cognitive landscapes : theory , application and ecological perspectives 1257–1267. https://doi.org/10.1007/s10980-011-9617-z 30 Furness, R.W., Greenwood, J.J.D., Jarvis, P.J., 1993. Can birds be used to monitor the environment? Birds as Monit. Environ. Chang. 1–41. https://doi.org/10.1007/978-94-015- 1322-7_1 Gómez, W.E., Isaza, C. V., Daza, J.M., 2018. Identifying disturbed habitats: A new method from acoustic indices. Ecol. Inform. https://doi.org/10.1016/j.ecoinf.2018.03.001 Johnson, C.N., Balmford, A., Brook, B.W., Buettel, J.C., Galetti, M., Guangchun, L., Wilmshurst, J.M., 2017. Biodiversity losses and conservation responses in the Anthropocene. Science (80-. ). 356, 270–275. https://doi.org/10.1126/science.aam9317 Jorge, F.C., Machado, C.G., da Cunha Nogueira, S.S., Nogueira-Filho, S.L.G., 2018. The effectiveness of acoustic indices for forest monitoring in Atlantic rainforest fragments. Ecol. Indic. 91, 71–76. https://doi.org/10.1016/j.ecolind.2018.04.001 Kasten, E.P., Gage, S.H., Fox, J., Joo, W., 2012. The remote environmental assessment laboratory’s acoustic library: An archive for studying soundscape ecology. Ecol. Inform. 12, 50–67. https://doi.org/10.1016/j.ecoinf.2012.08.001 Krause, B.L., 1987. Bioacoustics, Habitat Ambience in Ecological Balance," Whole Earth Review. Wild Sanctuary 57, 14–18. Laurance, W.F., 2009. Conserving the hottest of the hotspots. Biol. Conserv. 142, 1137. https://doi.org/10.1016/j.biocon.2008.10.011 Machado, R.B., Aguiar, L., Jones, G., 2017. Do acoustic indices reflect the characteristics of bird communities in the savannas of Central Brazil? Landsc. Urban Plan. https://doi.org/10.1016/j.landurbplan.2017.01.014 Mammides, C., Goodale, E., Dayananda, S.K., Kang, L., Chen, J., 2017. Do acoustic indices correlate with bird diversity? Insights from two biodiverse regions in Yunnan Province, south China. Ecol. Indic. 82, 470–477. https://doi.org/10.1016/j.ecolind.2017.07.017 Mitchell, S.L., Bicknell, J.E., Edwards, D.P., Deere, N.J., Bernard, H., Davies, Z.G., Struebig, M.J., 2020. Spatial replication and habitat context matters for assessments of tropical biodiversity using acoustic indices. Ecol. Indic. 119, 106717. https://doi.org/10.1016/j.ecolind.2020.106717 Montgomery, D. C., & Peck, E. A., 1992. Introduction to Linear Regression Analysis Wiley. New York, USA. Moreno-Gómez, F.N., Bartheld, J., Silva-Escobar, A.A., Briones, R., Márquez, R., Penna, M., 2019. Evaluating acoustic indices in the Valdivian rainforest, a biodiversity hotspot in South America. Ecol. Indic. 103, 1–8. https://doi.org/10.1016/j.ecolind.2019.03.024 31 Padoa-Schioppa, E., Baietto, M., Massa, R., Bottoni, L., 2006. Bird communities as bioindicators: The focal species concept in agricultural landscapes. Ecol. Indic. 6, 83–93. https://doi.org/10.1016/j.ecolind.2005.08.006 Peter K. McGregor · Tom M. Peake, 2000. Communication networks: social environments for receiving and signalling behaviour. Acta Ethol. 2, 71–81. https://doi.org/10.1007/s102110000015 Pieretti, N., Farina, A., Morri, D., 2011a. A new methodology to infer the singing activity of an avian community: The Acoustic Complexity Index (ACI). Ecol. Indic. 11, 868–873. https://doi.org/10.1016/j.ecolind.2010.11.005 Pieretti, N., Farina, A., Morri, D., 2011b. A new methodology to infer the singing activity of an avian community: The Acoustic Complexity Index (ACI). Ecol. Indic. 11, 868–873. https://doi.org/10.1016/j.ecolind.2010.11.005 Pijanowski, B.C., Farina, A., Gage, S.H., Dumyahn, S.L., Krause, B.L., 2011. What is soundscape ecology? An introduction and overview of an emerging new science. Landsc. Ecol. https://doi.org/10.1007/s10980-011-9600-8 Retamosa Izaguirre, M.I., Ramírez-Alán, O., 2018. Acoustic indices applied to biodiversity monitoring in a Costa Rica dry tropical forest. J. Ecoacoustics 2, 1–1. https://doi.org/10.22261/jea.tnw2np Rezende, C.L., Scarano, F.R., Assad, E.D., Joly, C.A., Metzger, J.P., Strassburg, B.B.N., Tabarelli, M., Fonseca, G.A., Mittermeier, R.A., 2018. From hotspot to hopespot: An opportunity for the Brazilian Atlantic Forest. Perspect. Ecol. Conserv. 16, 208–214. https://doi.org/10.1016/j.pecon.2018.10.002 Ribeiro, M.C., Metzger, J.P., Martensen, A.C., Ponzoni, F.J., Hirota, M.M., 2009. The Brazilian Atlantic Forest: How much is left, and how is the remaining forest distributed? Implications for conservation. Biol. Conserv. 142, 1141–1153. https://doi.org/10.1016/j.biocon.2009.02.021 Ross, S.R.P.J., Friedman, N.R., Yoshimura, M., Yoshida, T., Donohue, I., Economo, E.P., 2020. Utility of acoustic indices for ecological monitoring in complex sonic environments. Ecol. Indic. 107114. https://doi.org/10.1016/j.ecolind.2020.107114 Scarpelli, M.D.A., Ribeiro, M.C., Teixeira, F.Z., Young, R.J., Teixeira, C.P., 2020. Gaps in terrestrial soundscape research: It’s time to focus on tropical wildlife. Sci. Total Environ. https://doi.org/10.1016/j.scitotenv.2019.135403 Schmeller, D.S., Böhm, M., Arvanitidis, C., Barber-Meyer, S., Brummitt, N., Chandler, M., 32 Chatzinikolaou, E., Costello, M.J., Ding, H., García-Moreno, J., Gill, M., Haase, P., Jones, M., Juillard, R., Magnusson, W.E., Martin, C.S., McGeoch, M., Mihoub, J.B., Pettorelli, N., Proença, V., Peng, C., Regan, E., Schmiedel, U., Simaika, J.P., Weatherdon, L., Waterman, C., Xu, H., Belnap, J., 2017. Building capacity in biodiversity monitoring at the global scale. Biodivers. Conserv. 26, 2765–2790. https://doi.org/10.1007/s10531-017-1388-7 Shamon, H., Paraskevopoulou, Z., Kitzes, J., Card, E., Deichmann, J.L., Boyce, A.J., McShea, W.J., 2021. Using ecoacoustics metrices to track grassland bird richness across landscape gradients. Ecol. Indic. 120, 106928. https://doi.org/10.1016/j.ecolind.2020.106928 Shonfield, J., Bayne, E.M., 2017. Autonomous recording units in avian ecological research: current use and future applications. Avian Conserv. Ecol. https://doi.org/10.5751/ACE- 00974-120114 Stowell, D., Sueur, J., 2020. Ecoacoustics: acoustic sensing for biodiversity monitoring at scale. Remote Sens. Ecol. Conserv. 6, 217–219. https://doi.org/10.1002/rse2.174 Sueur, J., Aubin, T., Simonis, C., 2008. Sound analysis and synthesis with the package Seewave. Bioacoustics 18, 213–226. Sueur, J., Farina, A., Gasc, A., Pieretti, N., Pavoine, S., 2014a. Acoustic indices for biodiversity assessment and landscape investigation. Acta Acust. united with Acust. 100, 772–781. https://doi.org/10.3813/AAA.918757 Sueur, J., Farina, A., Gasc, A., Pieretti, N., Pavoine, S., 2014b. Acoustic indices for biodiversity assessment and landscape investigation. Acta Acust. united with Acust. 100, 772–781. https://doi.org/10.3813/AAA.918757 Sueur, Jérôme, Pavoine, S., Hamerlynck, O., Duvail, S., 2008. Rapid acoustic survey for biodiversity appraisal. PLoS One 3. https://doi.org/10.1371/journal.pone.0004065 Towsey, M., Wimmer, J., Williamson, I., Roe, P., 2014. The use of acoustic indices to determine avian species richness in audio-recordings of the environment. Ecol. Inform. 21, 110–119. https://doi.org/10.1016/j.ecoinf.2013.11.007 Vale, M.M., Tourinho, L., Lorini, M.L., Rajão, H., Figueiredo, M.S.L., 2018. Endemic birds of the Atlantic Forest: traits, conservation status, and patterns of biodiversity. J. F. Ornithol. 89, 193–206. https://doi.org/10.1111/jofo.12256 Villanueva-Rivera, L.J., Pijanowski, B.C., Doucette, J., Pekin, B., 2011. A primer of acoustic analysis for landscape ecologists. Landsc. Ecol. 26, 1233–1246. https://doi.org/10.1007/s10980-011-9636-9 Villanueva-Rivera, L.J., Pijanowski, B.C., Luis, M., Villanueva-Rivera, J., 2016. Package 33 “soundecology” Title Soundscape Ecology. Wimmer, J., Towsey, M., Roe, P., Williamson, I., 2013. Sampling environmental acoustic recordings to determine bird species richness, Ecological Applications. Zhao, Z., Xu, Z. yong, Bellisario, K., Zeng, R. wen, Li, N., Zhou, W. yang, Pijanowski, B.C., 2019. How well do acoustic indices measure biodiversity? Computational experiments to determine effect of sound unit shape, vocalization intensity, and frequency of vocalization occurrence on performance of acoustic indices. Ecol. Indic. 107. https://doi.org/10.1016/j.ecolind.2019.105588 34 Supplementary Material Appendix S1 - Programming schedule of audio recorders for data collection. 35 Appendix S2 - Scheme of the routine to preparing acoustic subdataset. 36 Appendix S3 - List of species of daytime birds identified by means of manual detection in 2,563 minutes obtained by Passive Acoustic Monitoring in fragmented landscapes of the Atlantic Forest. Names following CBRO 2015 Piacentini et. al (2015) and Conservation Status (IUCN 2020). * Endemism of Atlantic Forest biome (Vale et al. 2018). Order Family Scientific Name English Name Conservation Status Number of Tags Tinamiformes Tinamidae Tinamus solitarius Solitary Tinamou NT 3 Crypturellus obsoletus Brown Tinamou LC 17 Crypturellus parvirostris Small-billed Tinamou LC 15 Crypturellus tataupa Tataupa Tinamou LC 83 Rhynchotus rufescens Red-winged Tinamou LC 1 Anseriformes Anatidae Dendrocygna autumnalis Black-bellied Whistling-Duck LC 1 Cairina moschata Muscovy Duck LC 3 Galliformes Cracidae Penelope sp. 5 Pelecaniformes Threskiornithid ae Mesembrinibis cayennensis Green Ibis LC 8 Theristicus caudatus Buff-necked Ibis LC 3 Cathartiformes Cathartidae Coragyps atratus Black Vulture LC 2 Accipitriformes Accipitridae Leptodon cayanensis Gray-headed Kite LC 6 Rupornis magnirostris Roadside Hawk LC 26 Gruiformes Aramidae Aramus guarauna Limpkin LC 1 Rallidae Aramides cajaneus Gray-necked Wood-Rail LC 9 Aramides saracura* Slaty-breasted Wood-Rail LC 19 Laterallus melanophaius Rufous-sided Crake LC 3 Mustelirallus albicollis Ash-throated Crake LC 1 Pardirallus nigricans Blackish Rail LC 5 Charadriiformes Charadriidae Vanellus chilensis Southern Lapwing LC 118 Columbiformes Columbidae Columbina talpacoti Ruddy Ground-Dove LC 101 Patagioenas picazuro Picazuro Pigeon LC 241 Patagioenas cayennensis Pale-vented Pigeon LC 35 Patagioenas plumbea Plumbeous Pigeon LC 25 Patagioenas sp. 2 Zenaida auriculata Eared Dove LC 43 Leptotila verreauxi White-tipped Dove LC 163 Leptotila rufaxilla Gray-fronted Dove LC 62 Leptotila sp. 5 37 Order Family Scientific Name English Name Conservation Status Number of Tags Cuculiformes Cuculidae Piaya cayana Squirrel Cuckoo LC 10 Crotophaga ani Smooth-billed Ani LC 21 Guira guira Guira Cuckoo LC 36 Tapera naevia Striped Cuckoo LC 34 Dromococcyx pavoninus Pavonine Cuckoo LC 1 Caprimulgiformes Caprimulgidae Lurocalis semitorquatus Short-tailed Nighthawk LC 3 Apodiformes Apodidae Chaetura meridionalis Sick's Swift LC 1 Trochilidae Phaethornis pretrei Planalto Hermit LC 24 Phaethornis eurynome* Scale-throated Hermit LC 3 Phaethornis sp. 13 Eupetomena macroura Swallow-tailed Hummingbird LC 6 Florisuga fusca Black Jacobin LC 1 Chlorostilbon lucidus Glittering-bellied Emerald LC 9 Thalurania glaucopis* Violet-capped Woodnymph LC 2 Leucochloris albicollis White-throated Hummingbird LC 19 Amazilia lactea Sapphire-spangled Emerald LC 9 Amazilia sp. 6 Trochilidae sp. 5 Trogoniformes Trogonidae Trogon surrucura Surucua Trogon LC 3 Coraciiformes Alcedinidae Megaceryle torquata Ringed Kingfisher LC 1 Galbuliformes Galbulidae Galbula ruficauda Rufous-tailed Jacamar LC 2 Bucconidae Malacoptila striata* Crescent-chested Puffbird NT 8 Piciformes Ramphastidae Ramphastos toco Toco Toucan LC 3 Ramphastos dicolorus* Red-breasted Toucan LC 3 Picidae Picumnus sp. 38 Melanerpes candidus White Woodpecker LC 39 Veniliornis spilogaster White-spotted Woodpecker LC 5 Veniliornis sp. 1 Colaptes melanochloros Green-barred Woodpecker LC 3 Colaptes campestris Campo Flicker LC 98 Celeus flavescens Blond-crested Woodpecker LC 14 Dryocopus lineatus Lineated Woodpecker LC 3 Cariamiformes Cariamidae Cariama cristata Red-legged Seriema LC 62 38 Order Family Scientific Name English Name Conservation Status Number of Tags Falconiformes Falconidae Caracara plancus Southern Caracara LC 3 Milvago chimachima Yellow-headed Caracara LC 7 Herpetotheres cachinnans Laughing Falcon LC 5 Micrastur ruficollis Barred Forest-Falcon LC 1 Psittaciformes Psittacidae Psittacara leucophthalmus White-eyed Parakeet LC 113 Eupsittula aurea Peach-fronted Parakeet LC 1 Forpus xanthopterygius Blue-winged Parrotlet LC 8 Brotogeris tirica* Plain Parakeet LC 6 Brotogeris chiriri Yellow-chevroned Parakeet LC 62 Brotogeris sp. 11 Pionus maximiliani Scaly-headed Parrot LC 18 Amazona aestiva Turquoise-fronted Parrot LC 1 Psittacidae sp. 5 Passeriformes Thamnophilida e Dysithamnus mentalis Plain Antvireo LC 171 Herpsilochmus rufimarginatus Rufous-winged Antwren LC 61 Thamnophilus doliatus Barred Antshrike LC 1 Thamnophilus ruficapillus Rufous-capped Antshrike LC 3 Thamnophilus caerulescens Variable Antshrike LC 143 Batara cinerea Giant Antshrike LC 3 Hypoedaleus guttatus* Spot-backed Antshrike LC 60 Mackenziaena severa* Tufted Antshrike LC 2 Myrmoderus squamosus* Squamate Antbird LC 19 Pyriglena leucoptera White-shouldered Fire-eye LC 75 Drymophila ferruginea* Ferruginous Antbird LC 9 Drymophila ochropyga* Ochre-rumped Antbird NT 2 Drymophila malura* Dusky-tailed Antbird LC 1 Conopophagida e Conopophaga lineata Rufous Gnateater LC 63 Grallariidae Grallaria varia Variegated Antpitta LC 18 Rhinocryptidae Psilorhamphus guttatus* Spotted Bamboowren NT 2 Scleruridae Sclerurus scansor* Rufous-breasted Leaftosser LC 2 Dendrocolaptid ae Sittasomus griseicapillus Olivaceous Woodcreeper LC 29 Xiphorhynchus fuscus* Lesser Woodcreeper LC 3 Campylorhamphus falcularius* Black-billed Scythebill LC 1 Lepidocolaptes angustirostris Narrow-billed Woodcreeper LC 1 39 Order Family Scientific Name English Name Conservation Status Number of Tags ... Xenopidae Xenops rutilans Streaked Xenops LC 13 Furnariidae Furnarius figulus Wing-banded Hornero LC 1 Furnarius rufus Rufous Hornero LC 129 Lochmias nematura Sharp-tailed Streamcreeper LC 21 Automolus leucophthalmus* White-eyed Foliage-gleaner LC 38 Anabazenops fuscus White-collared Foliage-gleaner LC 1 Philydor rufum Buff-fronted Foliage-gleaner LC 15 Heliobletus contaminatus Sharp-billed Treehunter LC 2 Phacellodomus ferrugineigula* Orange-breasted Thornbird LC 13 Certhiaxis cinnamomeus Yellow-chinned Spinetail LC 14 Synallaxis ruficapilla* Rufous-capped Spinetail LC 32 Synallaxis frontalis Sooty-fronted Spinetail LC 4 Synallaxis spixi Spix's Spinetail LC 118 Cranioleuca pallida* Pallid Spinetail LC 22 Pipridae Chiroxiphia caudata* Swallow-tailed Manakin LC 105 Tityridae Schiffornis virescens* Greenish Schiffornis LC 12 Pachyramphus viridis Green-backed Becard LC 1 Pachyramphus validus Crested Becard LC 26 Cotingidae Pyroderus scutatus Red-ruffed Fruitcrow LC 3 Procnias nudicollis* Bare-throated Bellbird VU 55 Platyrinchidae Platyrinchus mystaceus White-throated Spadebill LC 35 Rhynchocyclida e Mionectes rufiventris* Gray-hooded Flycatcher LC 26 Leptopogon amaurocephalus Sepia-capped Flycatcher LC 22 Corythopis delalandi Southern Antpipit LC 43 Phylloscartes ventralis Mottle-cheeked Tyrannulet LC 1 Tolmomyias sulphurescens Yellow-olive Flycatcher LC 148 Todirostrum poliocephalum* Gray-headed Tody-Flycatcher LC 36 Todirostrum cinereum Common Tody-Flycatcher LC 5 Poecilotriccus plumbeiceps Ochre-faced Tody-Flycatcher LC 19 Hemitriccus diops* Drab-breasted Pygmy-Tyrant LC 1 Tyrannidae Hirundinea ferruginea Cliff Flycatcher LC 1 Tyranniscus burmeisteri Rough-legged Tyrannulet LC 1 40 Order Family Scientific Name English Name Conservation Status Number of Tags ... Camptostoma obsoletum Southern Beardless-Tyrannulet LC 91 Elaenia flavogaster Yellow-bellied Elaenia LC 78 Elaenia obscura Highland Elaenia LC 23 Myiopagis caniceps Gray Elaenia LC 2 Myiopagis viridicata Greenish Elaenia LC 3 Serpophaga subcristata White-crested Tyrannulet LC 8 Attila rufus Gray-hooded Attila LC 3 Legatus leucophaius Piratic Flycatcher LC 2 Myiarchus swainsoni Swainson's Flycatcher LC 69 Myiarchus ferox Short-crested Flycatcher LC 18 Myiarchus tyrannulus Brown-crested Flycatcher LC 26 Pitangus sulphuratus Great Kiskadee LC 371 Machetornis rixosa Cattle Tyrant LC 9 Myiodynastes maculatus Streaked Flycatcher LC 164 Megarynchus pitangua Boat-billed Flycatcher LC 31 Myiozetetes similis Social Flycatcher LC 22 Tyrannus melancholicus Tropical Kingbird LC 179 Tyrannus savana Fork-tailed Flycatcher LC 2 Empidonomus varius Variegated Flycatcher LC 26 Colonia colonus Long-tailed Tyrant LC 5 Myiophobus fasciatus Bran-colored Flycatcher LC 47 Gubernetes yetapa Streamer-tailed Tyrant LC 1 Lathrotriccus euleri Euler's Flycatcher LC 48 Vireonidae Cyclarhis gujanensis Rufous-browed Peppershrike LC 427 Hylophilus poicilotis* Rufous-crowned Greenlet LC 1 Vireo chivi Chivi Vireo LC 781 Corvidae Cyanocorax chrysops Plush-crested Jay LC 18 Hirundinidae Pygochelidon cyanoleuca Blue-and-white Swallow LC 7 Stelgidopteryx ruficollis Southern Rough-winged Swallow LC 4 Progne tapera Brown-chested Martin LC 2 Progne chalybea Gray-breasted Martin LC 1 Tachycineta sp. 1 Troglodytidae Troglodytes musculus Southern House Wren LC 274 Donacobiidae Donacobius atricapilla Black-capped Donacobius LC 1 Turdidae Turdus flavipes Yellow-legged Thrush LC 2 Turdus leucomelas Pale-breasted Thrush LC 201 41 Order Family Scientific Name English Name Conservation Status Number of Tags ... Turdus rufiventris Rufous-bellied Thrush LC 164 Turdus amaurochalinus Creamy-bellied Thrush LC 40 Turdus albicollis White-necked Thrush LC 71 Turdus sp. 84 Mimidae Mimus saturninus Chalk-browed Mockingbird LC 10 Passerellidae Ammodramus humeralis Grassland Sparrow LC 48 Parulidae Setophaga pitiayumi Tropical Parula LC 82 Basileuterus culicivorus Golden-crowned Warbler LC 533 Myiothlypis flaveola Flavescent Warbler LC 5 Myiothlypis leucoblephara White-browed Warbler LC 267 Passerellidae Arremon semitorquatus* Half-collared Sparrow LC 8 Zonotrichia capensis Rufous-collared Sparrow LC 631 Icteridae Icterus pyrrhopterus Variable Oriole LC 2 Icterus jamacaii Campo Troupial LC 1 Gnorimopsar chopi Chopi Blackbird LC 2 Chrysomus ruficapillus Chestnut-capped Blackbird LC 49 Pseudoleistes guirahuro Yellow-rumped Marshbird LC 25 Agelasticus cyanopus Unicolored Blackbird LC 3 Molothrus bonariensis Shiny Cowbird LC 2 Thraupidae Pipraeidea melanonota Fawn-breasted Tanager LC 1 Tangara desmaresti* Brassy-breasted Tanager LC 1 Tangara sayaca Sayaca Tanager LC 340 Tangara palmarum Palm Tanager LC 4 Tangara cayana Burnished-buff Tanager LC 47 Tangara sp. 9 Conirostrum speciosum Chestnut-vented Conebill LC 35 Sicalis flaveola Saffron Finch LC 89 Sicalis luteola Grassland Yellow-Finch LC 10 Volatinia jacarina Blue-black Grassquit LC 123 Coryphospingus cucullatus Red-crested Finch LC 3 Tachyphonus coronatus* Ruby-crowned Tanager LC 140 Ramphocelus carbo Silver-beaked Tanager LC 5 Tersina viridis Swallow Tanager LC 28 Dacnis cayana Blue Dacnis LC 1 42 Order Family Scientific Name English Name Conservation Status Number of Tags ... Coereba flaveola Bananaquit LC 57 Tiaris fuliginosus Sooty Grassquit LC 1 Sporophila lineola Lined Seedeater LC 55 Sporophila frontalis* Buffy-fronted Seedeater VU 1 Sporophila caerulescens Double-collared Seedeater LC 108 Sporophila leucoptera White-bellied Seedeater LC 2 Sporophila angolensis Chestnut-bellied Seed-Finch LC 1 Sporophila sp. 4 Emberizoides herbicola Wedge-tailed Grass-Finch LC 2 Saltator similis Green-winged Saltator LC 146 Saltator fuliginosus* Black-throated Grosbeak LC 1 Thlypopsis sordida Orange-headed Tanager LC 2 Thraupidae sp. 3 Cardinalidae Habia rubica Red-crowned Ant-Tanager LC 31 Cyanoloxia brissonii Ultramarine Grosbeak LC 4 Fringillidae Euphonia chlorotica Purple-throated Euphonia LC 26 Euphonia cyanocephala Golden-rumped Euphonia LC 5 Passeridae Passer domesticus House Sparrow LC 2 43 Appendix S4 -Model selection results using the General Linear Model (GLM) to data with all environments. Models with R²≤0.20 were disregarded Bird richness GLM models [Delta] AICc df weight r2 Richness~ACI_1000_12000_FFT512 + BIO_1000_12000_FFT512 + environment 0 6 1 0.43 Richness~BIO_1000_12000_FFT512 + AR + environment 474.96 6 << 0.001 0.38 Richness~ADI_300_12000_75db + BIO_1000_12000_FFT512 + environment 513.07 6 << 0.001 0.38 Richness~ACI_1000_12000_FFT512 + BIO_1000_12000_FFT512 605.74 4 << 0.001 0.37 Richness~BIO_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 + environment 658.39 6 << 0.001 0.37 Richness~AEI_1000_12000_50db + BIO_1000_12000_FFT512 + environment 774.67 6 << 0.001 0.36 Richness~BIO_1000_12000_FFT512 + Ht + environment 873.11 6 << 0.001 0.35 Richness~ADI_300_12000_75db + BIO_1000_12000_FFT512 961.08 4 << 0.001 0.34 Richness~BIO_1000_12000_FFT512 + BIO_1000_22050_FFT512 + environment 998.84 6 << 0.001 0.33 Richness~BIO_1000_12000_FFT512 + environment 1001.87 5 << 0.001 0.33 Richness~BIO_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 + environment 1003.13 6 << 0.001 0.33 Richness~ADI_300_12000_75db + ACI_1000_12000_FFT512 + environment 1065.06 6 << 0.001 0.33 Richness~ACI_1000_12000_FFT512 + BIO_1000_22050_FFT512 + environment 1177.48 6 << 0.001 0.31 Richness~ACI_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 + environment 1211.18 6 << 0.001 0.31 Richness~BIO_1000_12000_FFT512 + AR 1223.65 4 << 0.001 0.31 Richness~ACI_1000_12000_FFT512 + Ht + environment 1234.96 6 << 0.001 0.31 Richness~AEI_1000_12000_50db + BIO_1000_12000_FFT512 1267.49 4 << 0.001 0.31 Richness~ACI_1000_12000_FFT512 + AR + environment 1290.21 6 << 0.001 0.30 Richness~BIO_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 1345.58 4 << 0.001 0.30 Richness~ACI_1000_12000_FFT512 + environment 1347.33 5 << 0.001 0.30 Richness~ACI_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 + environment 1348.28 6 << 0.001 0.30 Richness~AEI_1000_12000_50db + ACI_1000_12000_FFT512 + environment 1349.22 6 << 0.001 0.30 Richness~ACI_1000_12000_FFT512 + BIO_1000_22050_FFT512 1444.54 4 << 0.001 0.28 Richness~ADI_300_12000_75db + ACI_1000_12000_FFT512 1504.32 4 0 0.28 Richness~BIO_1000_12000_FFT512 + Ht 1542.72 4 0 0.27 Richness~ACI_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 1571.49 4 0 0.27 Richness~BIO_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 1589.03 4 0 0.27 Richness~BIO_1000_12000_FFT512 + BIO_1000_22050_FFT512 1598.63 4 0 0.27 Richness~ACI_1000_12000_FFT512 + Ht 1609.02 4 0 0.27 Richness~BIO_1000_12000_FFT512 1622.76 3 0 0.26 Richness~ACI_1000_12000_FFT512 + AR 1623.80 4 0 0.26 Richness~ACI_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 1667.78 4 0 0.26 Richness~ACI_1000_12000_FFT512 1688.73 3 0 0.26 Richness~AEI_1000_12000_50db + ACI_1000_12000_FFT512 1690.31 4 0 0.26 44 Richness~ADI_300_12000_75db + BIO_1000_22050_FFT512 + environment 1923.36 6 0 0.23 Richness~ADI_300_12000_75db + BIO_1000_22050_FFT512 1941.89 4 0 0.23 Richness~N ULL 3622.95 2 0 0.00 45 Number of Tags GLMmodels [Delta] AICc df weight r2 Tags.Number~ACI_1000_12000_FFT512 + BIO_1000_12000_FFT512 + environment 0 6 1 0.48 Tags.Number~BIO_1000_12000_FFT512 + AR + environment 317.74 6 << 0.001 0.46 Tags.Number~ADI_300_12000_75db + BIO_1000_12000_FFT512 + environment 466.09 6 << 0.001 0.44 Tags.Number~AEI_1000_12000_50db + BIO_1000_12000_FFT512 + environment 468.56 6 << 0.001 0.44 Tags.Number~BIO_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 + environment 636.75 6 << 0.001 0.43 Tags.Number~BIO_1000_12000_FFT512 + Ht + environment 902.61 6 << 0.001 0.40 Tags.Number~BIO_1000_12000_FFT512 + BIO_1000_22050_FFT512 + environment 952.93 6 << 0.001 0.40 Tags.Number~BIO_1000_12000_FFT512 + environment 953.61 5 << 0.001 0.40 Tags.Number~BIO_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 + environment 955.42 6 << 0.001 0.40 Tags.Number~ADI_300_12000_75db + ACI_1000_12000_FFT512 + environment 1318.21 6 << 0.001 0.37 Tags.Number~ACI_1000_12000_FFT512 + BIO_1000_22050_FFT512 + environment 1390.49 6 << 0.001 0.36 Tags.Number~ACI_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 + environment 1468.36 6 << 0.001 0.35 Tags.Number~ACI_1000_12000_FFT512 + AR + environment 1513.94 6 0 0.35 Tags.Number~ACI_1000_12000_FFT512 + Ht + environment 1547.28 6 0 0.34 Tags.Number~AEI_1000_12000_50db + ACI_1000_12000_FFT512 + environment 1553.40 6 0 0.34 Tags.Number~ACI_1000_12000_FFT512 + BIO_1000_12000_FFT512 1572.87 4 0 0.34 Tags.Number~ACI_1000_12000_FFT512 + environment 1584.97 5 0 0.34 Tags.Number~ACI_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 + environment 1585.10 6 0 0.34 Tags.Number~ADI_300_12000_75db + BIO_1000_12000_FFT512 1864.81 4 0 0.31 Tags.Number~AEI_1000_12000_50db + BIO_1000_12000_FFT512 1918.20 4 0 0.30 Tags.Number~ADI_300_12000_75db + ACI_1000_12000_FFT512 2099.27 4 0 0.28 Tags.Number~ADI_300_12000_75db + BIO_1000_22050_FFT512 + environment 2118.81 6 0 0.28 Tags.Number~ACI_1000_12000_FFT512 + BIO_1000_22050_FFT512 2160.58 4 0 0.28 Tags.Number~BIO_1000_12000_FFT512 + AR 2264.20 4 0 0.27 Tags.Number~ACI_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 2275.08 4 0 0.26 Tags.Number~AEI_1000_12000_50db + ACI_1000_12000_FFT512 2280.09 4 0 0.26 Tags.Number~ACI_1000_12000_FFT512 + AR 2290.77 4 0 0.26 Tags.Number~ACI_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 2299.79 4 0 0.26 Tags.Number~ACI_1000_12000_FFT512 + Ht 2355.46 4 0 0.26 Tags.Number~ACI_1000_12000_FFT512 2368.62 3 0 0.25 Tags.Number~BIO_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 2430.88 4 0 0.25 Tags.Number~BIO_1000_22050_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 + environment 2483.83 6 0 0.24 Tags.Number~ADI_300_12000_75db + BIO_1000_22050_FFT512 2502.94 4 0 0.24 Tags.Number~BIO_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 2526.65 4 0 0.24 Tags.Number~BIO_1000_12000_FFT512 + BIO_1000_22050_FFT512 2594.70 4 0 0.23 Tags.Number~BIO_1000_12000_FFT512 + Ht 2619.52 4 0 0.22 46 Tags.Number~BIO_1000_12000_FFT512 2630.26 3 0 0.22 Tags.Number~AEI_1000_12000_50db + BIO_1000_22050_FFT512 + environment 2639.75 6 0 0.22 Tags.number~NULL 4274.76 2 0 0.00 47 Diversity GLM models [Delta] AICc df weight r2 Diversity~ACI_1000_12000_FFT512 + BIO_1000_12000_FFT512 + environment 0 6 1.000 0.26 Diversity~BIO_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 + environment 40.44 6 << 0.001 0.25 Diversity~ACI_1000_12000_FFT512 + BIO_1000_12000_FFT512 53.97 4 << 0.001 0.25 Diversity~BIO_1000_12000_FFT512 + NDSI_AMAX2000_AMIN1000_BMAX22050_BMIN2000_FFT512 67.93 4 << 0.001 0.25 Diversity~BIO_1000_12000_FFT512 + AR + environment 112.83 6 << 0.001 0.24 Diversity~BIO_1000_12000_FFT512 + AR 147.78 4 << 0.001 0.24 Diversity~ADI_300_12000_75db + BIO_1000_12000_FFT512 + environment 202.84 6 << 0.001 0.23 Diversity~ADI_300_12000_75db + BIO_1000_12000_FFT512 207.36 4 << 0.001 0.23 Diversity~BIO_1000_12000_FFT512 + Ht + environment 212.12 6 << 0.001 0.23 Diversity~BIO_1000_12000_FFT512 + Ht 242.16 4 << 0.001 0.23 Diversity~BIO_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 + environment 324.32 6 << 0.001 0.22 Diversity~AEI_1000_12000_50db + BIO_1000_12000_FFT512 + environment 326.04 6 << 0.001 0.22 Diversity~AEI_1000_12000_50db + BIO_1000_12000_FFT512 333.22 4 << 0.001 0.22 Diversity~BIO_1000_12000_FFT512 + BIO_1000_22050_FFT512 + environment 354.69 6 << 0.001 0.21 Diversity~BIO_1000_12000_FFT512 + NDSI_AMAX1000_AMIN300_BMAX12000_BMIN1000_FFT512 364.68 4 << 0.001 0.21 Diversity~BIO_1000_12000_FFT512 + BIO_1000_22050_FFT512 371.09 4 << 0.001 0.21 Diversity~BIO_1000_12000_FFT512 + environment 374.80 5 << 0.001 0.21 Diversity~BIO_1000_12000_FFT512 395.16 3 << 0.001 0.21 Diversity~NULL 1921.78 2 0 0.00 48 Appendix S5 - Model selection results using the General Linear Model (GLM) to data to each environment separate. Here considered only the five best models to each response variable. Forest GLM models [Delta]AICc df wAIC R² Richness~BIO1 + ACI 0 4 1.000 0.25 Richness~BIO1 + NDSI2 102.88 4 << 0.001 0.21 Richness~ACI+Ht 105.34 4 << 0.001 0.21 Richness~NDSI1+NDSI2 182.43 4 << 0.001 0.19 Richness~BIO1+Ht 208.26 4 << 0.001 0.18 Tags.Number~BIO1.NDSI2 0 4 1.000 0.50 Tags.Number~ACI + BIO1 365.63 4 << 0.001 0.41 Tags.Number~BIO+NDSI2 386.14 4 << 0.001 0.41 Tags.Number~BIO1+ADI 520.50 4 << 0.001 0.37 Tags.Number~BIO1+AR 613.98 4 << 0.001 0.34 Diversity~NDSI1 + Ht 0 4 1.000 0.14 Diversity~ACI + .Ht 19.39 4 << 0.001 0.13 Diversity~AEI+Ht 30.55 4 << 0.001 0.13 Diversity~AEI+NDSI2 51.43 4 << 0.001 0.12 Diversity~ADI+Ht 53.92 4 << 0.001 0.12 Swamp Richness~ADI + BIO1 0 4 1.000 0.75 Richness~ACI + BIO1 162.05 4 << 0.001 0.73 Richness~BIO1.BIO2 203.10 4 << 0.001 0.72 Richness~BIO1 + NDSI1 221.15 4 << 0.001 0.72 Richness~BIO1 + NDSI2 253.68 4 << 0.001 0.71 Tags.Number~ACI + BIO1 0 4 1.000 0.64 Tags.Number~AEI + BIO1 114.64 4 << 0.001 0.62 Tags.Number ~ADI + BIO1 150.65 4 << 0.001 0.62 Tags.Number~ACI + Ht 158.81 4 << 0.001 0.61 Tags.Number~BIO1 + NDSI1 204.19 4 << 0.001 0.61 Diversity~ADI + BIO1 0 4 1.000 0.67 Diversity~BIO1 + BIO2 79.91 4 << 0.0001 0.66 Diversity~ADI1 +BIO1 0 4 1.000 0.67 Diversity~BIO1 + BIO2 79.91 4 << 0.001 0.66 49 Diversity~AEI + BIO1 267.25 4 << 0.001 0.63 50 Pasture Richness~BIO1 + AR 0 4 1.000 0.41 Richness~ACI + AR 179.76 4 << 0.001 0.35 Richness~BIO1+ADI 200.27 4 << 0.001 0.35 Richness~ACI+ADI 218.22 4 << 0.001 0.34 Richness~BIO1 + ACI 261.59 4 << 0.001 0.33 Tags.Number~BIO1 + AR 0 4 1.000 0.43 Tags.Number~ACI + AR 109.97 4 << 0.001 0.40 Tags.Number~ACI+ADI 214.20 4 << 0.001 0.38 Tags.Number~BIO1 + ADI 293.02 4 << 0.001 0.35 Tags.Number~BIO1 + ACI 364.02 4 << 0.001 0.33 Diversity~BIO + AR 0 4 1.000 0.29 Diversity~AEI + BIO1 110.42 4 << 0.001 0.25 Diversity~ACI+AR 189.45 4 << 0.001 0.23 Diversity~NDSI1 + NDSI2 225.46 4 << 0.001 0.21 Diversity~BIO1 + NDSI2 262.13 4 << 0.001 0.20 Agradecimentos 1.Introduction 2.Material and Methods 2.1 Study area 2.2 Experimental design 2.3 Soundscape recordings 2.3 Audio subset and species labeling 2.5 Acoustic indices 2.6 Models and Statistical analysis 3.Results 3.1 Considering generalizations of environments 3.2 Considering each environment separately 4.Discussion 5.Conclusions and future directions 6.References