Soundscape Analytics: A New Frontier of Knowledge Discovery in Soundscape Data

Here, we describe a new evolving research area of focus for soundscape ecology called soundscape analytics. Soundscape analytics follows traditional, as well as state-of-the-art, data and visual analytics that chain together tools and approaches to explore and analyze massive data. The theoretical underpinning of soundscape analytics is anchored in recent advances in machine learning that have been very successful in other applications, such as business, medicine, and psychology. We present and summarize four main components – data processing, data mining, integration and interpretation - of soundscape analytics pipelines that are being used today by soundscape ecologists. In the last five years, the number of tools advancing our ability to analyze big acoustic data for soundscape ecology research has increased considerably, especially those leveraging generic deep learning methods. A considerable portion of this work has focused on using soundscape recordings to assess biodiversity trends across space and time. Many of these implementations are based on R and Python routines designed to be executed on supercomputers with specialized data storage arrays as well as cloud-based user interface software. Quick porting of interactive data visualizations to the web enables scientists around the world to collaborate and share information for management and with the public. Big data in ecology has arrived, illustrated by the massive amounts of acoustic data being collected by soundscape ecologists. The challenge of soundscape analytics is to make the most of each available computational resource so many application problems can be solved from similar data.