Ambisonic Earth: Recording, processing, and archiving biodiverse ecological soundscapes with higher-order ambisonics

Doctoral dissertation, UC San Diego, 2026

DOI: https://doi.org/10.5281/zenodo.21423865

Permalink: https://escholarship.org/uc/item/61n1r3tt

ProQuest ID: https://www.proquest.com/LegacyDocView/DISSNUM/32443324

Abstract:

Higher-order ambisonics (HOA) offers a powerful set of methods for immersively recording and reproducing the three-dimensional soundfields of ecological environments, yet HOA content is significantly underrepresented in public soundscape archives. Existing ambisonic collections primarily feature first-order ambisonics or urban acoustic scenes, while the practitioners whose work most compellingly demonstrates the value of immersive ecological recording have largely not made their multichannel recordings accessible. This dissertation documents Ambisonic Earth, an ongoing project to record, process, and publicly archive the soundscapes of biodiverse ecosystems in higher-order ambisonics, while serving as a practical guide for others undertaking this work. Over 500 hours of third-order ambisonic soundscape recordings have been captured from 2022 through 2025 across 35 RESOLVE ecoregions and nine terrestrial biomes. Recording sites span 27 U.S. National Wildlife Refuges, 19 Key Biodiversity Areas, four Ramsar Wetlands, two UNESCO Biosphere Reserves, and a UNESCO World Heritage Site. This dissertation presents five primary contributions. First, plans for a portable and replicable recording system centered on the ZYLIA ZM-1 array, including a Raspberry Pi-based field recorder prototype enabling autonomous long-form capture in a sealed, weatherproof enclosure. Second, a site selection framework drawing on the World Database on Protected Areas, RESOLVE Ecoregions and Biomes, Key Biodiversity Areas, and other global biodiversity and heritage datasets. Third, a noise reduction methodology, referred to as pink residual noise shaping (PRNS), that shapes the residual noise of the recording system to approximate the spectral slope of pink noise at −3 dB per octave, allowing remaining system noise to blend unobtrusively into the background of ecological recordings. Fourth, a layered sonic time-lapse (LSTL) method that generates and sums multiple time-lapses with incremented start times, processed independently across channels, to condense all sonic material from long-form ambisonic recordings into spatially coherent experiential summaries. Fifth, a multi-platform archival infrastructure, including: The Ambisonic Archive (https://zenodo.org/communities/ambisonicarchive) and Ambisonic Earth (https://zenodo.org/communities/ambisonicearth) Zenodo communities with persistent DOIs for multichannel source files and metadata, a YouTube channel with binaural audio and spectrogram video (https://www.youtube.com/@AmbisonicEarth), and an interactive sound map linking recording locations to both platforms (https://www.google.com/maps/d/viewer?mid=1-GRE1QsbZWooy2rGRFl0S8ryXhEKwxQ). Archival of the collection is proceeding on a rolling basis, with the infrastructure designed to grow incrementally as processing continues.