- Recognitions, 24 h
- –
- Species
- –
- Listening stations
- –
- Hydrophones live
- –
What this is
A live view of the planet's biophony: the combined sound of living things (Pijanowski et al. 2011). Thousands of free, always-on acoustic monitors report what they recognise. Drop a pin anywhere and Biophony gathers the nearest microphones: build a drum kit from what they heard, browse and download the recordings, or stitch an ambience bed for film. Behind it, each station gets a balance index and satellite context, so you can see where the soundscape is rich and where it is thin.
Listening
- BirdWeather: over 10,000 stations (BirdWeather PUCs and BirdNET-Pi) running the BirdNET model. Birds, plus bats, insects, frogs and some mammals. Each recognition comes with the original audio clip.
- Orcasound: live hydrophones in the Salish Sea. Whale calls, vessels and other sounds are flagged by listeners and by the OrcaHello model, and you can listen to the live stream.
Map layers
- Ecosystem type: MODIS land cover (IGBP classes, 500 m, yearly). Forest, shrubland, savanna, grassland, wetland, cropland, urban, snow, barren.
- Human-shaped biomes: Anthromes (Ellis & Ramankutty 2008). Wildlands, rangelands, villages, croplands, cities.
- Vegetation greenness: MODIS NDVI, 8-day.
- Canopy height: NASA GEDI lidar, 2019–2023. Tall, layered canopy means more reverberation and more vertical niches for calling animals.
- Night lights: NASA Black Marble. A rough proxy for roads, settlements and the noise that comes with them.
- Human footprint: WCS & CIESIN (Sanderson et al. 2002). Combined pressure from population, land use, roads and lights.
- Protected areas: the World Database on Protected Areas (UNEP-WCMC & IUCN).
- Terrain shading: Esri hillshade. Valleys and ridges shape how sound travels.
- EMIT (imaging spectrometer on the ISS, 285 bands, 60 m): scenes over each site, and methane plume detections.
Google's AlphaEarth satellite embeddings would be a great fit, but they are only served through Earth Engine, which needs an account, so they are left out.
The balance index
Each station gets a Soundscape Balance Index from 0 to 100, built from its last 24 hours of recognitions. The index is this project's own experimental composite; every ingredient is a published metric:
- Diversity, 50%: effective number of species, the Hill number of order 1, exp(H′) (Hill 1973; Jost 2006). Scored as a percentile against stations at a similar latitude, because diversity naturally rises toward the tropics (Hillebrand 2004).
- Evenness, 35%: Pielou's J′ = H′ / ln S (Pielou 1966). Low when one or two species dominate.
- Taxonomic breadth, 15%: whether insects, frogs, bats or mammals are heard alongside birds. Full marks at three groups.
Stations with fewer than 30 recognitions or 3 species are left out. Sample coverage (Chao & Jost 2012) and Chao1 estimated richness (Chao 1984, 1987) show how complete each sample is. The map grid and regions use the median of their stations.
Site health
Opening a station adds three more measurements and combines them with the balance index:
- Soundscape: from the station's recent audio, the Normalized Difference Soundscape Index, NDSI (Kasten et al. 2012): biophony (2–11 kHz) against anthrophony (1–2 kHz), the band where traffic and machinery sit. Also shown: Acoustic Diversity Index (Villanueva-Rivera et al. 2011) and acoustic entropy H (Sueur et al. 2008).
- Vegetation: MODIS 250 m NDVI around the station and the Vegetation Condition Index, VCI (Kogan 1995): today's greenness compared with the same time of year in past years.
- Methane: EMIT plume detections nearby, shown as a flag rather than scored.
Read with care
Detection counts are calls, not individuals. BirdNET makes mistakes and is strongest on North American and European birds. Clips are saved when a bird is recognised, which biases the soundscape indices upward. Acoustic indices track species diversity only moderately (Alcocer et al. 2022). Use the numbers to compare places and watch change over time, not as a verdict on a place.
References
- Alcocer I. et al. (2022) Acoustic indices as proxies for biodiversity: a meta-analysis. Biological Reviews 97: 2209–2236.
- Chao A. (1984) Nonparametric estimation of the number of classes in a population. Scand. J. Statistics 11: 265–270.
- Chao A. & Jost L. (2012) Coverage-based rarefaction and extrapolation. Ecology 93: 2533–2547.
- Hill M.O. (1973) Diversity and evenness: a unifying notation. Ecology 54: 427–432.
- Hillebrand H. (2004) On the generality of the latitudinal diversity gradient. Am. Naturalist 163: 192–211.
- Jost L. (2006) Entropy and diversity. Oikos 113: 363–375.
- Kahl S. et al. (2021) BirdNET: a deep learning solution for avian diversity monitoring. Ecological Informatics 61: 101236.
- Kasten E.P. et al. (2012) The remote environmental assessment laboratory's acoustic library. Ecological Informatics 12: 50–67.
- Kogan F.N. (1995) Application of vegetation index and brightness temperature for drought detection. Adv. Space Research 15: 91–100.
- Pielou E.C. (1966) The measurement of diversity in different types of biological collections. J. Theoretical Biology 13: 131–144.
- Ellis E.C. & Ramankutty N. (2008) Putting people in the map: anthropogenic biomes of the world. Frontiers in Ecology and the Environment 6: 439–447.
- Pijanowski B.C. et al. (2011) Soundscape ecology: the science of sound in the landscape. BioScience 61: 203–216.
- Sanderson E.W. et al. (2002) The human footprint and the last of the wild. BioScience 52: 891–904.
- Sueur J. et al. (2008) Rapid acoustic survey for biodiversity appraisal. PLoS ONE 3: e4065.
- Villanueva-Rivera L.J. et al. (2011) A primer of acoustic analysis for landscape ecologists. Landscape Ecology 26: 1233–1246.
Field Studio
Every recording is checked in your browser before you hear it: the signal-to-noise ratio in the band where the call is strongest (loud frames in the detection window against quiet frames outside it), plus peak and RMS level. Near-silent and very noisy clips are hidden, and the rest play levelled to the same loudness. For kits, each sound is also measured for tonality: how far its strongest spectral peak rises above the rest, at what frequency, and how much that pitch wanders. Steady whistles become the tuned voices, retuned into one key chosen to need the least correction; broadband, fast-attack sounds become drums.
Credits
Data: BirdWeather & BirdNET (K. Lisa Yang Center, Cornell Lab of Ornithology / TU Chemnitz), Orcasound, NASA EMIT via LP DAAC & CMR, NASA GIBS, MODIS via ORNL DAAC, protected areas from Protected Planet (UNEP-WCMC & IUCN). Basemap © Esri, HERE, Garmin, OpenStreetMap contributors. Made by Anish Ghosh.