Índices ecoacústicos como ferramentas para investigar a variação temporal de uma paisagem sonora na Baía da Ilha Grande

Abstract

Soundscapes can diversify over time and space, with the analysis of temporal sound variations being an important mechanism for understanding the ecological dynamics of an environment. Ecoacoustic indices are quantitative metrics used to characterize sound environments and have proven to be a crucial tool in soundscape ecology for estimating biodiversity and ecosystem processes. Passive acoustic monitoring has proven to be an effective solution for accessing soundscapes in remote locations; however, this technique often generates large volumes of data, making manual analysis impractical. Machine learning algorithms stand out as an efficient approach to analyzing this data and sound recognition. Marine soundscapes are composed of multiple sounds. Many marine organisms produce sounds and have a significant connection with their soundscape. Anthropogenic noise is increasingly present in marine environments, often causing changes in the sound space, which may affect ecosystem dynamics and the life history of populations. During the global COVID-19 pandemic, significant transformations occurred in the soundscapes of various environments, including the marine ecosystem, primarily due to the reduction of anthropogenic noise, mainly caused by the reduction in boat and ship traffic. Including several environmental protection areas, Ilha Grande Bay presents distinct ecological attributes and is considered a hotspot for Brazilian marine biodiversity. Despite this, the negative impacts of noise pollution pose a significant threat to the region. The objective of this study is to analyze the temporal variations in the underwater soundscape of Baía da Ilha Grande, employing a global acoustic method through the application of six ecoacoustic indices (ACI, ADI, AEI, BI, NDSI, and NP), based on sound data collected during two distinct periods through a passive acoustic monitoring system. The first data collection took place between December 2020 and January 2021, and the second occurred between December 2022 and January 2023. It was assessed whether the soundscape showed daily variations in both collection periods. It was also investigated whether the COVID-19 pandemic, due to the reduction of commercial and tourist activities in the region, might have influenced the seasonal variations in the soundscape. The results of the ecoacoustic indices point to the existence of daily variation, with the early hours of the day, at sunrise, and the hours near sunset and moonrise showing higher activity and sound diversity compared to other daily periods. It was also possible to observe that, during the first data collection period, the soundscape exhibited higher levels of diversification and sound complexity, suggesting a seasonal change in the acoustic composition between the two collection periods. Furthermore, a review of the state of the art on machine learning techniques was conducted, describing how neural networks have been used to assist in the automatic recognition of sounds in marine soundscapes.