Constructing a Globally Accessible Soundbank for Brazil's Five Ecosystems.
Adao Henrique Rosa Domingos, IPBio-Biodiversity Research Institute, Atlantic Forest Research Center/ Betary Reserve, Sáo Paulo, Brazil
With the support of a bilingual volunteer we have translated section 3 and 4 of the SM4 Manual so that all our national biologists, volunteers, interns and visiting researches can understand and become familiar with the technology. Henrique Domingos, IPBio biologist, set up a test with the Song Meter SM4 recorder to get some practice with the device and software (figure 1).This test was conducted in May and June. We selected different points in terms of habitat composition and altitudinal gradient.
The first point was on the banks of a pond, at an altitude of 100 meters above sea level, in an area of forest in its initial stage of regeneration. The Song Meter SM4 was programmed to record at night to detect the diversity of amphibians in this period of the year.
The second point was situated in a forest area at an advanced regeneration stage, at an altitude of 150 meters above sea level, and the recordings were made in the morning.
The third point was a well preserved forest area, at an altitude of 356 meters above sea level. At this third point we installed the Song Meter SM4 along with a Bushnell camera trap attempting to record audio and capture images of the Sapajus nigritus.
The data is still under review, but in a short period of time, we were able to detect approximately 30 species of birds in these 3 points, among them the Carpornis melanocephala considered rare or scarce, and is categorized as "vulnerable" according to the List of Threatened Species of the Brazilian Fauna, 2014 (figure 2). Moreover, we detected two mammalian species, namely the Sapajus nigritus and Alouatta guariba, and three species of amphibians Sphaenorhynchus caramaschi, Dendropsophus werneri and Scinax sp.
After a series of tests on the Betary Reserve, we selected certain points we considered as top priority due to our past experience of encountering high levels of activity in these areas, to carry out bioacoustics monitoring. The Betary Reserve contains various stages of forest regeneration, which makes it a place with a high level of diversity. One such point M1 (Hill 1) is 356m altitude and we shall monitor it for a year. The idea is to check the variation of species according to the seasonal period. This location has high potential to record sounds of birds and some mammals. One thing about this point that caught our attention was the presence of black capuchin monkeys (Sapajus nigritus). We managed to record some sounds emitted by these animals and we also captured images of them as we are installing a Bushnell Camera Trap along with the Song Meter SM4. The next step will be to start assembling the recognizers using the scope song application.
Since October, IPBio has been monitoring the SM4 recorder on Hill M1 during peak activity hours in the morning and evening and has since set up another recorder at a second site, Pitfall 5. Where Hill M1 is a site relatively isolated from frequent human contact, Pitfall 5 is a site close to Betary Reserve headquarters with a high potential for capturing bird and amphibian vocalizations.
Every week at the Pitfall 5 site, we record the first 10 minutes of every hour for 24 hours with the aim of tracking species activity levels throughout the day. This data will prove valuable in showing us how activity levels for each species vary during the course of the day and year and will help the reserve monitor any year-to-year changes that may result from changes to the surrounding environment. Further, we shall be able to compare data collected from Hill M1 and Pitfall 5 to begin understanding the differences in biodiversity and activity levels between the sites. Significant differences between the sites may serve as a basis for future study.
Beyond monitoring species activity, IPBio will have generated a collection of sound recordings that will serve both research and educational purposes. IPBio frequently receives student groups and families interested in learning more about Brazil's Atlantic Forest, so having a diverse bank of sounds will be useful for teaching about specific animals which are difficult to see in the forest due to the region's very dense vegetation. One such recording is that of the howler monkey (Alouatta clamitans), a mammal commonly found on the Betary Reserve that can be heard clearly from 3 miles away, but is not easy to observe visually.
To date, IPBio has collected a total of approximately 1000 hours (or about 41 days) of audio from 4 sites on the Betary Reserve. The two sites previously identified as good long-term research locations will continue recording until at least February 2018 so that IPBio can study species activity levels over the course of an entire year. Both sites will continue to use camera traps nearby to complement the auditory data collected by the SM4 recorder with photos and videos of animals in the area.
Our bioacoustics project is an inventory of all species in the site locations that produce sound, and while the team at IPBio has been able to identify many of the species recorded, identifying the diverse range of birds, amphibians, mammals, and insects found on the reserve presents a unique challenge! To aid in the identification of unknown species, we are enlisting the help of experts and researchers at a few of Brazil's universities.
Over the course of 8 months, we have generated and used recognizers within the SongScope software to automatically identify species, enabling us to observe variations in activity throughout the day and from month to month. As an example, above is a graph of a single day at our Pitfall 5 site in March 2017. The variation in number of active species as well as species type throughout the day is evident, with distinct changes around sunrise (06:19) and sunset (18:24). The graph demonstrates clearly that the birds near our site are active during the day, while insects dominate the nighttime soundscape. This trend for insects dominant during the night can be seen in the below soundscape generated for the same time period, which represents an entire day's worth of audio recordings at Pitfall 5. The bright green areas in the middle of the soundscape are sounds generated by insects during the nighttime hours.
When we have finished data collection, IPBio will be able to generate graphs similar to the one above to show not only activity variation throughout the day, but also seasonal variation over the course of an entire year. We will analyse the data at both the species type and individual species level. Finally, the collection of sound bites for each species will continue to be useful as we teach visitors about the wildlife in the Atlantic Forest.
IPBio has been focused on developing a sound bank of all the species of mammals, amphibians and birds in the Atlantic Forest. Since 2016, after receiving the Scientific Product Grant, we began using the SM4 recorder to monitor a site on our reserve for over a year with the aim of understanding the daily and seasonal patterns of these animals. We began using Song Scope to analyze the recording and started developing recognizers for each species with plans of running all of the recognizers after 2 or 3 years of data collection. In May of 2018, IPBio received an additional grant from Wildlife Acoustics with their updated software named Kaleidoscope. The benefits were apparent immediately. Clustering technology was featured in the new software that removed the tedious job of manually annotating every audio, as the new software would automatically cluster similar vocalizations. This left us with a much simpler task of checking through the already automatically organized vocalizations to confirm the software was sorting efficiently. Moreover, the transition from recognizers to classifiers was massively beneficial as while recognizers were manually fed sample recording to produce a recognizer, the new classifier system essentially allowed us to consistently improve it while we went through more data.
In June of 2018, we received an intern from Antioch University who dedicated here time to graphing the vocalization of three species of frogs over a year long period of data collection to understand their seasonal variations. With the new kaleidoscope software she was able to learn how to use the software from scratch and conduct this study in approximately a month. This speed of analysis was unprecedented for IPBio and has made the process of developing recognizer, now known as classifiers, substantially more efficient.
Over the coming years IPBio will continue to develop a sound bank and study the seasonal habits of wildlife on our reserve in the Atlantic Forest and then expand to other biomes in Brazil as our organization grows.