The potential of soundscapes as an ecosystem monitoring tool for urban biodiversity

Abstract

As urbanization and densification often lead to significant biodiversity loss, understanding and monitoring urban biodiversity patterns is crucial. Traditional monitoring methods are often costly, time-consuming, and require specialized expertise. Passive acoustic monitoring and soundscape ecology have emerged as promising, non-invasive techniques for ecosystem monitoring. This review aims to provide an overview of methods and approaches utilized in urban soundscape ecology and discuss their limitations. We highlight exemplary studies that focus on urban soundscape and biodiversity monitoring to demonstrate that acoustic recordings can be partially used to predict biodiversity in cities, especially for avian species. To realize the potential of urban soundscape monitoring for biodiversity conservation, current challenges must be addressed. This includes data processing, data security, and missing standardized data collection methods. We call for further research that combines innovative technologies and transdisciplinary approaches for non-invasive biodiversity monitoring to develop effective conservation applications for cities.

Introduction

Conserving biodiversity in cities is crucial for improving the livability of urban environments for people and nature (Sandifer et al. 2015). The scientific interest in urban biodiversity research is strongly increasing, showing the importance of understanding biodiversity patterns and processes in cities (Rega-Brodsky et al. 2022). Urbanization and densification generally lead to a significant loss in biodiversity (Aronson et al. 2014). This is alarming because cities can also support local biodiversity, including rare species (Aronson et al. 2014; Ives et al. 2016; Spotswood et al. 2021), through habitat provision. In Germany, nearly 50% of all species found locally can also be found in cities (Sweet et al. 2022). Approaches that facilitate the systematic and long-term collection of biodiversity-related data are critical for understanding how land use change affects urban biodiversity and evaluating how conservation interventions (e.g. increased green space) may better support urban biodiversity. However, traditional methods for monitoring biodiversity are expensive, time-consuming, and require highly trained experts who may not be readily available to local government administration. Thus, new methods for monitoring urban biodiversity are necessary to understand the relationships between species abundance, richness, and diversity with urban habitat and landscape characteristics.

Passive acoustic monitoring and ecoacoustics, the ecological investigation and interpretation of environmental sound (Sueur and Farina 2015), have gained significant research interest in the last 15 years as promising and non-invasive ecosystem monitoring techniques (Alcocer et al. 2022). Soundscapes are defined as “the collection of biological, geophysical and anthropogenic sounds that emanate from a landscape and which vary over space and time reflecting important ecosystem processes and human activities” (Pijanowski et al. 2011). This includes the combination of so-called “biophonic,” “geophonic” and “anthrophonic” sounds. Soundscape ecology emphasizes the ecological characteristics of sounds (Pijanowski et al. 2011) and can thus provide valuable insights into ecosystem processes by effectively portraying the habitat and landscape structure (Fuller et al. 2015), and by helping to identify potential ecosystem stressors (Gill et al. 2017). While individual studies have shown some connection between soundscape indices and biodiversity, a meta-analysis by Alcocer et al. (2022) found only a moderate positive correlation between acoustic indices and biodiversity metrics. Yet it is essential to understand how soundscape indices function in a specific environment for them to be employed and useful. Especially for urban environments, the relationship between biodiversity and soundscape indices is largely underexplored and unclear because of the complexity of urban soundscapes (Fairbrass et al. 2017). Furthermore, the field of ecoacoustics is rapidly developing, where new techniques and technologies are being employed to better describe the acoustic environment and how it relates to biodiversity (e.g. Sethi et al. 2022). The development in acoustic monitoring is promising for applying soundscapes in biodiversity conservation (Sueur et al. 2014; Alcocer et al. 2022). However, there currently still are significant limitations around e.g. data processing and biasing effects, especially in urban environments. We argue that soundscapes have a high potential as a monitoring tool for urban biodiversity, and this can inform urban conservation science and practice, if the challenges can be overcome in the future. In this article, we: (i) provide a review of methods and approaches used in urban soundscape ecology; (ii) show best-practice examples of soundscapes as an ecosystem monitoring tool; and (iii) call for future research that combines new technologies and transdisciplinary approaches for non-invasive biodiversity monitoring to develop effective conservation applications for cities.

The state of urban soundscape monitoring

With advances in Autonomous Recording Units (ARU’s), the ability to do large-scale and long-term soundscape monitoring is easier than ever. There exists a range of affordable devices available from all the major manufacturers (Wildlife Acoustics, Frontier Labs, Open Acoustic Devices, among others). In addition to the manufacturers, there are several do-it-yourself and low-cost options (Farina et al. 2014; Bobryk et al. 2016). As such, there has been an expansion in the number of researchers using soundscapes to understand the environment and the indices for describing it.

Ecoacoustic indices such as Acoustic Complexity Index (ACI), the Acoustic Diversity Index (ADI), the Acoustic Evenness Index (AEI), the Bioacoustic Index (BI), and the Normalized Difference Soundscape Index (NSDI), among others, when used alone or in combination have in some cases been demonstrated to be successful proxies for biodiversity but also anthropogenic disturbances (i.e. Fairbrass et al. 2017; Buxton et al. 2018; Alcocer et al. 2022). The practice of using a wide array of indices to describe the acoustic environment is growing. For example, the QUT Ecoacoustics research group developed a custom software called Analysis Programs that conglomerates over 20 indices (Towsey et al. 2018). This wide array of indices was recently demonstrated to be useful for monitoring tropical forest recovery (Müller et al. 2023). However, the generalizable use of ecoacoustic indices has come into question as their usefulness for monitoring biodiversity varies across ecosystems (Sethi et al. 2023; Santos et al. 2024). Therefore, further development of ecoacoustic measurement methods is needed.

More recently, machine learning methods have been applied to understanding ecoacoustics. Methods such as Soundscape fingerprinting used a pre-trained CNN (convolutional neural network) with 128 features to quantify the soundscape (Sethi et al. 2022). With this method, the authors were able to describe the acoustic environment and predict the presence of birds and reptiles in a tropical forest. Similarly, CityNet used CNN’s to predict both biological and anthropogenic noise in urban areas and demonstrated better performance than standard acoustic indices in describing the ratio of anthropogenic to biophonic noise (Fairbrass et al. 2019). However, while the performance of these machine learning methods is promising, they come with a high initial time investment cost. Where the acoustic indices require little initial setup and can be calculated using existing software packages, i.e. seewave (Sueur et al. 2008) and soundecology (Villanueva-Rivera and Pijanowski 2016) in the R statistical environment, both CityNet and Soundscape fingerprinting require a deeper understanding of programming and machine learning. Furthermore, CityNet is trained on acoustic data from London and the authors acknowledge that it would need retraining on data from other cities to be generally applicable (Fairbrass et al. 2019), which requires large tagged training data.

While there are many studies investigating the application and function of soundscape monitoring in natural environments (e.g. Farina et al. 2011; Deichmann et al. 2017; Sethi et al. 2022), urban environments are a relatively unexplored frontier and existing work tends to focus on methods development and testing (Table 1). As urban environments are sonically complex when compared to most natural environments, they present some novel challenges, especially when it comes to the application and efficacy of soundscape indices (Santos et al. 2024). For example, acoustic indices can be significantly biased by the dominating anthropogenic sounds found in the urban environment (Fairbrass et al. 2017). Despite this, they show that under some circumstances, ACI and NDSI could be used to measure biophony or the ratio of biophony to anthrophony. Still, it is recommended to remove biasing anthropogenic noise before applying soundscape indices to the urban environment, as these can cause false readings for biophony (Fairbrass et al. 2017). Although inconsistent across studies and urban contexts (e.g. Santos et al. 2024), some studies do identify ecologically relevant patterns in acoustic recordings using soundscape monitoring in urban environments (e.g. Holgate et al. 2021).

Selected case studies on urban soundscapes, ecoacoustics and biodiversity monitoring

An increasing number of studies explore the effectiveness of ecoacoustic soundscape methodologies to monitor (urban) biodiversity (see examples above). In addition, several studies use soundscape techniques for urban sound planning (e.g. noise control) (Brown 2012; Rehan 2016; Liu et al. 2023). In one of the first studies looking at soundscapes in urban environments, Joo et al. (2011) found a negative relationship between biophony and urbanization across an urban-rural gradient in Lansing Michigan, USA. More recently Dein and Rüdisser (2020) demonstrated the application of soundscape monitoring across an urban-rural gradient in Innsbruck, Austria. To map the acoustic environment, two metrics were used: a sound exposure level metric (Merchant et al. 2015) and percent biophony as a metric that quantifies biophonic sounds which were separated from the rest of the recording to avoid a noise bias (Dein and Rüdisser 2020). They concluded that land cover indices can serve as a reasonable but scale-dependent predictor for the acoustic environment.

Recent studies also focus on using soundscapes to provide information on the habitat quality of urban green spaces. Benocci et al. (2023) developed a soundscape ranking index (SRI) to describe the environmental sound quality of urban green spaces and provide a soundscape-based measure of habitat quality. The SRI is based on a manual aural survey of sound recordings to identify biophonic activity and technophobic sounds and thus it is simple but labor intensive. Devos (2023) quantifies the biophonic component of soundscapes during the dawn chorus to assess the habitat quality of urban green spaces. The proposed Bird Dawn Chorus Strength (BDCS) is derived from year-long diurnal acoustic recordings and gives information on the songbird population and relates to the diversity of an urban green area. The BDCS is computed as the transition strength of the used ecoacoustic indices during dawn and could therefore be a valuable standardization of the used indices (Devos 2023).

Few studies explicitly aim to translate soundscape monitoring into urban biodiversity conservation; however, there are some examples. Zhao et al. (2022) test soundscape recordings as a tool to monitor biodiversity in urban parks in Beijing, China, and study their relation to habitat variables like park age, vertical heterogeneity of vegetation, percentage of bare land and human-made facilities or urban development gradient. Zhao et al. identify that a multilayered vegetation leads to higher acoustic diversity. To foster biodiversity, they suggest using ecoacoustics as a regular monitoring tool to inform urban green space management. In another example, Holgate et al. (2021) use soundscape approaches to identify hotspots of biodiversity in peri-urban landscapes in Queensland, Australia, to inform urban planning in future development sites to avoid or mitigate the negative effects of urban growth on biodiversity. The study found a positive correlation between the ACI and both bird and insect activity and promotes passive acoustic monitoring and soundscape mapping as a cost-effective method to identify spatial and temporal acoustic hotspots and peaks.

Discussion

Ecoacoustics can provide a useful approach to study, monitor, and thus understand urban biodiversity as well as potential interventions for conservation action. Soundscapes can be an indicator of species presence/absence and diversity, but also of the anthropogenic stressors of the urban environment that may negatively affect biodiversity. Thus, as we have reviewed here, ecoacoustics can partly be used for non-invasive biodiversity monitoring, and new methods in AI aim to improve soundscape quantification. However, there are still many open questions as to their efficacy and implementation in the urban environment, which can be complex to capture systematically and thoroughly. Indeed, it may be advisable to combine soundscape monitoring with traditional ecological surveys to truly capture and assess biodiversity, as some sounds may not fully represent an area’s total biodiversity (Sethi et al. 2023). Here, we summarize the opportunities, limitations, and challenges for soundscape-based monitoring in urban environments for biodiversity conservation applications.

Technological innovation

Innovative examples of intensive data collection, cheap but effective technologies, and developments in artificial intelligence (AI) show that there is a bright future for ecoacoustics in urban conservation. For example, recent advancements in AI now allow for species level identification of some taxa to a high degree of accuracy. For example, neural networks like BirdNET (Kahl et al. 2021) can identify individual species in recordings and has been shown to work well in urban environments (Fairbairn et al., in review). These technologies, coupled with the ability to identify the sources of other sounds (e.g. anthropogenic noises such as cars, planes, voices, etc.), means that a whole new set of soundscape indices could be developed. More informative indices could better describe the acoustic environment based on individual sounds as opposed to describing the environment based on characteristics of the audio files such as frequency, saturation, and amplitude. Furthermore, AI approaches now allow for screening recordings for species identification, which means that it may be possible to filter audio files for specific species, thereby identifying where in the landscape a species was found and under what habitat and acoustic habitat conditions. Useful here is the Acoustic Habitat Hypothesis (Mullet et al. 2017), which explains how the acoustic environment influences habitat selection of sound-dependent species that require specific acoustic and physical conditions for their signals to be transmitted, received and interpreted. This may inform species-specific conservation interventions for relevant species of conservation concern by considering and including the acoustic habitat not just the physical characteristics.

Nevertheless, some challenges remain, and it may be advisable to combine the ecoacoustic monitoring of the soundscape with ecological surveys of sound-producing organisms, including birds, vocal mammals and arthropods on site (Sethi et al. 2023). Furthermore, field surveys should consider the diurnal versus nocturnal differences in when sound-producing organisms are active. For example, while songbirds can contribute to the biophonic components of the diurnal soundscape to thereby strongly influence the soundscape, arthropods, anurans and vocal mammals such as bats are often active at night to influence the nocturnal soundscape (Sousa-Lima et al. 2018; Gallacher et al. 2021). While in temperate environments, birds and bird songs make up the majority of the biophony, and thus often the focus of ecoacoustic research, they do not necessarily represent the entire biodiversity of an area. A rich bird soundscape could represent a certain abundance of arthropods, as many birds need arthropods to feed their young. However, there are also biotopes where only a single bird species occurs, but which have a very high diversity of other taxa such as arthropods and plants (e.g. calcareous grasslands). And even if the soundscape of an area only indicates the presence of a few bird species, these can be of high nature conservation value. Additional monitoring or ecological assessments using remote sensing or by trained researchers may thus still be necessary to confirm or support AI-derived soundscape indices.

Call for transdisciplinary research

Transdisciplinary research in urban ecoacoustics not only holds promise for advancing conservation science but also presents unique opportunities for shaping city policies and fostering collaborative efforts with conservation organizations, city residents and local government administration. Transdisciplinarity involves more than just one branch of knowledge, and often consists of scientific researchers engaging with practitioners and government bodies on equal footing. For example, engaging citizens through innovative approaches like citizen science initiatives on acoustic monitoring can enhance data collection and community involvement in ecoacoustics, as demonstrated by Clark et al. (2023). In targeting 54 bird species that have distinct and unique vocalizations, trained citizen scientists were not only crucial to validate bird vocalizations in a reference dataset, but to also further train and improve CNN-based models (Clark et al. 2023). Other (eco)acoustic research projects that utilize citizen science include the Dawn Chorus Project. In Dawn Chorus, participants are invited to make multiple 60-second recordings of the same location over time. These recordings are analyzed using the AI algorithm BirdNET, which uses machine learning to recognize and classify bird calls to characterize bird diversity (see: https://dawn-chorus.org/science/). Citizen scientists who are experts in bird call recognition cross-check and improve AI-results. An open-source sound map further enables anyone to listen to soundscapes around the world, increasing awareness of soundscape diversity in various geographic contexts (see: https://explore.dawn-chorus.org/? lang=en). Dawn Chorus includes environmental education programs such as early morning sound walks, as well as art-science interfaces including museum exhibitions and musical albums with musicians and DJs. In addition, the platform is used by transdisciplinary research projects including the “CitySoundscape” project in Munich, Germany. Here, Dawn Chorus information will inform city biodiversity conservation strategies by expanding systematic ecoacoustic monitoring currently underway in the city to provide spatially explicit biodiversity data (Fairbairn et al. 2024). Thus, citizen science paired with ARU-derived acoustic data can be an important and useful collaborative effort to monitor bird diversity and improve conservation initiatives in the city.

Furthermore, the integration of research-practice-policy interfaces facilitates transitions from research outcomes to practical applications in urban planning (Chen et al. 2022). This transdisciplinary approach extends beyond traditional scientific domains in biodiversity management and conservation, and landscape planning, by touching on aspects of health and wellbeing. For example, the project HUSH City Lab initiated by urban architects and planners recruits city residents to record and describe their favorite quiet places in the city (see: https://opensourcesoundscapes.org/). These “quiet” places—oases from noise from traffic and construction—may also be biodiverse in their soundscapes. With this crowdsourced information as evidence, the project aims to inform and implement urban sound planning strategies and policies that conserve quiet places as well as identify characteristics of these places that can be used to create new quiet places that promote human health. This can foster synergies among biodiversity and human health if both people and nature benefit. In such initiatives, disciplines and experts in public health, environmental psychology and urban sound planning can be brought together. For example, studies in environmental psychology and psychoacoustics on the acoustic comfort of an environment can bridge biodiversity and human health research. Here, methods such as experimental soundwalks have participants evaluate the soundscape using likert scales, and describe how fascinating, interesting, and relaxing it is (Semidor 2006). This quantitative data can be related to the actual biodiversity of the place and then be evaluated to test relationships between urban biodiversity and the physical and mental well-being of city dwellers (Dearborn and Kark 2010; Ives et al. 2017). Positive impacts of health-promoting soundscapes include attention restoration, stress reduction and decreasing exposure to noise pollution (Houlden et al. 2021; Marselle et al. 2021). Recognizing the connection between urban soundscapes and human wellbeing highlights the importance of incorporating sound planning into urban development strategies that also consider biodiversity conservation. Transdisciplinary research involving social sciences, psychoacoustics, governance, planning and therapeutic action can amplify the collective expertise of scientists, policymakers, and communities to create acoustically pleasing urban environments. This is currently the goal of the above mentioned CitySoundscapes project, in which biodiversity-related indices for soundscapes are developed using field inventories, 3D laser scanning, and acoustic recordings across diverse urban green spaces. Experimental soundwalks are used to understand the impact of audible biodiversity on the subjective well-being of people who visit these places. The ecological and social research is paired with a living lab led by the Department for Climate and Environmental Protection with city stakeholders to incorporate biodiversity-based health interventions into urban planning and management.

Future directions in soundscape-based biodiversity monitoring

The future of ecoacoustics and soundscape monitoring in urban areas presents several key considerations that must be addressed to advance this field. One crucial aspect is the processing of data, wherein the challenge lies in effectively handling and analyzing large amounts of quantitative information generated by soundscape monitoring systems. Advances in machine learning and source separation models where individual sound sources can be identified in a recording could help extract more meaningful information on the effects of individual sounds on humans or biodiversity (Lin and Tsao 2020; Sethi et al. 2020). Further advances mean it is now possible, with the correct recorder deployment and little technical knowledge, to identify the direction of a sound source, potentially opening new avenues of research in urban ecoacoustics (Buchmann and Schurr 2024).

Additionally, permission issues raise questions about who has access to the collected data and under what circumstances. Data security measures also necessitate instruments in place to safely capture and store sensitive information from unauthorized access e.g. from park human users that may not realize there are monitoring devices. Here suggestions are to work closely with government officials to design monitoring schemes that meet the needs of researchers as well as fit to city or neighborhood rules and regulations. Best practice should include the anonymization of data by automatically removing human voices using voice activity detection models prior to further processing. The threat of theft and vandalism also poses a practical challenge to the maintenance and longevity of monitoring equipment in cities. New strategies are needed to protect the infrastructure and the operation of sound monitoring systems. A suggestion is to install the recorders at large heights to ensure they are out of reach of the public. For long-term monitoring schemes, recorders could feasibly be installed on buildings. Further, specific hardware could be designed that could tap into a city’s existing power and data networks, which would allow for near real-time monitoring over extended periods. In addition, the cost of equipment could be reduced to lower the barriers to ecoacoustic research and to lower the risk of lost or damaged equipment. The reduction of cost can come from using low-cost recorders, or the less expensive options from wildlife acoustics, taking into consideration the potential trade-off of reduced audio quality and ease of use.

Determining appropriate scales at which to measure acoustic environments is another challenge, as it may influence data interpretation and action. While traffic noise often emerges as a dominant component, research suggests that urban acoustic environments exhibit high variability at small scales, such as street levels, mainly depending on the local landscape characteristics and human activities. However, soundscape information on single functional urban spaces is not sufficient for urban management. Instead, larger spatiotemporal scales need to be taken into consideration for an efficient implementation of soundscapes in urban planning (Liu et al. 2013). Furthermore, if monitoring devices are placed too far or too close from one another, we may not detect differences in ecological communities in association with habitat or landscape context. Integrating contributions from multiple scales could enhance the predictive power of spatial acoustic models, as proposed by Dein and Rüdisser in 2020.

Finally, achieving comparable datasets across ecosystems and cities is necessary for meaningful cross-city and cross-ecosystem analyses. Standardizing data collection methods and ensuring consistency in monitoring practices will facilitate more accurate comparisons and a comprehensive understanding of urban soundscapes. In summary, the future directions in soundscape-based monitoring involve addressing data processing challenges, navigating permissions and security issues, mitigating theft and vandalism risks, determining appropriate measurement scales, and fostering the development of comparable datasets across diverse urban environments.

Conclusions

City landscapes and the ecosystems within them provide habitat for biodiversity. It is increasingly recognized that cities can thereby play a key role in supporting species conservation in addition to rural and natural landscapes (Grabowski et al. 2023). Consistent and systematic monitoring of biodiversity is needed in urban areas to better understand patterns and drivers of biodiversity in the city to improve management approaches specifically—but not limited to—sound-producing organisms. Passive acoustic monitoring and ecoacoustics can potentially provide a new tool to perform city-wide assessments of sound-producing organisms and understand how their presence and diversity varies with urban landscape heterogeneity. To realize this potential, it is critical to develop standardized protocols as well as reduce costs for equipment and to lower the bar for data processing and computation. Furthermore, it is important to explore transdisciplinary approaches in which natural science questions and methods are combined with social science questions around acoustic comfort and urban planning action that can improve city landscapes for biodiversity conservation and human health. Urban ecoacoustics offer a promising, non-invasive approach to understanding and conserving urban biodiversity, but further research is needed to address challenges in data processing, security, and standardization of data collection methods that are currently still limiting the applicability.

Author contributions

Sophie Arzberger (Conceptualization [equal], Funding acquisition [supporting], Writing—original draft [lead], Writing—review & editing [lead]), Andrew Fairbairn (Conceptualization [equal], Writing—original draft [equal], Writing—review & editing [equal]), Michael Hemauer (Conceptualization [supporting], Writing—original draft [equal]), Maximilian Mühlbauer (Conceptualization [equal], Writing—original draft [equal], Writing—review & editing [equal]), Julie Weissmann (Conceptualization [equal], Writing—original draft [equal], Writing—review & editing [equal]), and Monika Egerer (Conceptualization [equal], Funding acquisition [lead], Writing—original draft [equal], Writing—review & editing [equal])

Conflict of interest: The authors have no conflict of interest to report.

Funding

The project on which this report is based was funded by the German Federal Ministry of Education and Research within the Research Initiative for the Conservation of Biodiversity (FEdA) under the funding code 16LW0385. The responsibility for the content of this publication lies with the authors.

Data availability

This article does not contain empirical research and there is no data associated with it.

References