Rainforest Connection | Publications

Multispecies Bioacoustic Classification using transfer learning of deep convolutional
neural networks with pseudo-labeling

April 9, 2020

In this study, we evaluated deep convolutional neural networks for classifying the calls of 24 birds and amphibian species detected in ambient field recordings. In classifying a test set of manually validated positive and negative template based detections, our proposed model achieves 97.7% sensitivity (true positive rate), 96.4% specificity (true negative rate) and 99.5% Area Under a Curve (AUC). This multi-label multi-species classification methodology and its framework can be easily adopted by other acoustic classification problems. Please email us at contact@rfcx.org for a copy of this article.

How does FSC forest certification affect the acoustically active fauna in Peru?

May 22, 2019

Despite several efforts to quantify the effectiveness of forest certification in developing sustainable use of forest resources, there is little evidence that certified forests are more effective in conserving fauna than non-certified managed forest. Our findings correspond with the conclusions of other studies that certified forests can maintain levels of fauna biodiversity similar to those of undisturbed primary forest in the Amazon region.

Using soundscapes to assess biodiversity in Neotropical oil palm landscapes

April 6, 2019

Expanding oil palm plantations have caused widespread deforestation and biodiversity loss in Southeast Asia, stigmatizing the industry around the world regardless of regional context. In Latin America, oil palm plantations are primarily replacing other agroindustrial land uses with uncertain implications for local biodiversity. Our aim was to create empirical baseline data to help guide development of future plantations into areas where biodiversity impacts are minimized. We used soundscapes to assess fauna in oil palm landscapes of Colombia, the world’s 4th largest palm oil producer.

It’s time to listen: there is much to be learned from the sounds of tropical ecosystems

June 8, 2018

Soundscape recordings provide a permanent record of a site at a given time and contain a wealth of invaluable and irreplaceable information. In this commentary, we (1) argue for the need to increase acoustic monitoring in tropical systems; (2) describe the types of research questions and conservation issues that can be addressed with passive acoustic monitoring (PAM) using both shortand long-term data in terrestrial and freshwater habitats; and (3) present an initial plan for establishing a global repository of tropical recordings.

Changes in the acoustic structure and composition along a tropical elevational gradient

December 6, 2017

To improve our understanding of how environmental gradients influence patterns of animal communities and to test the relationship between soundscapes and animal composition we investigated how variation in bird and anuran composition affect the acoustic structure and composition of the soundscapes along an elevation gradient. This study shows how different animal taxa respond to environmental gradients and provide strong evidence for the use of soundscapes as a tool to describe and compare species distribution and composition across large spatial scales.

Species Richness (of Insects) Drives the Use of Acoustic Space in the Tropics

October 27, 2017

In this manuscript, we evaluate recordings from eight tropical forest sites that vary in species richness, from a relatively low diversity Caribbean forest to a megadiverse Amazonian forest, with the goal of understanding the relationship between acoustic space use (ASU) and species diversity across different taxonomic groups. We show a strong positive relationship between ASU and regional and acoustic morphospecies richness. Premontane forest sites had the highest ASU and the highest species richness, while dry forest and montane sites had lower ASU and lower species richness. Furthermore, we show that insect richness was the best predictor of variation in total ASU, and that insect richness was proportionally greater at high-diversity sites.

Have bird distributions shifted along an elevational gradient on a tropical mountain?

September 28, 2017

An upward shift in elevation is one of the most conspicuous species responses to climate change. Nevertheless, downward shifts and, apparently, the absences of response have also been recently reported. Given the growing evidence of multiple responses of species distributions due to climate change and the paucity of studies in the tropics, we evaluated the response of a montane bird community to climate change, without the confounding effects of land-use change.

Species-specific audio detection: a comparison of 3 template based detection algorithms using random forests

April 24, 2017

We developed a web-based cloud-hosted system that allow users to archive, listen, visualize, and annotate recordings. The system also provides tools to convert these annotations into datasets that can be used to train a computer to detect the presence or absence of a species. The algorithm used by the system was selected after comparing the accuracy and efficiency of three variants of a template-based detection.

Soundscape analysis and acoustic monitoring document impacts of natural gas exploration on biodiversity in a tropical forest

November 18, 2016

We used passive acoustic monitoring in a pre-montane forest in Peru to investigate how soundscape composition and richness of acoustic frequencies varied with distance from a natural gas exploratory well. Results demonstrate that acoustic monitoring and soundscape analyses are useful tools for evaluating the impact of development activity on the vocalizing community, and should be implemented as a best practice in monitoring biodiversity and for guiding specific mitigation strategies.

Audio segmentation using Flattened Local Trimmed Range for ecological acoustic space analysis

June 27, 2016

In this paper, we describe a new spectrogram-based approach for extracting individual audio events. Our goal is to develop an algorithm that is not sensitive to noise, does not need any prior training data and works with any type of audio event. To do this, we propose: (1) a spectrogram filtering method, the Flattened Local Trimmed Range (FLTR) method, which models the spectrogram as a mixture of stationary and non-stationary energy processes and mitigates the effect of the stationary processes, and (2) an unsupervised algorithm that uses the filter to detect audio events.