Invertebrate species produce taxon-specific acoustic profiles under controlled conditions

This proof-of-concept study demonstrates that distinct invertebrate species produce unique, taxon-specific acoustic signatures under controlled conditions, establishing the potential for non-destructive, scalable soil biodiversity monitoring through automated ecoacoustic classification.

The Gist

Imagine the soil beneath our feet not as just dirt, but as a bustling, invisible city. In this city, tiny creatures like worms, beetles, spiders, and snails are the workers, builders, and cleaners. They are crucial for keeping our planet healthy, growing our food, and regulating the climate. But right now, this underground city is in trouble. About 75% of the world's soil is damaged, and we need to fix it.

The problem? We don't have a good way to check if the soil is getting better. Traditional methods are like trying to count the people in a city by digging up the streets and tearing down the buildings. It's messy, destructive, and takes forever.

The New Idea: Listening to the Soil

This paper is about a new, non-invasive way to monitor soil health: listening to it.

Think of the soil as a giant, complex instrument. When creatures move across it, they make tiny vibrations and sounds, just like footsteps on a wooden floor or a cat walking on a table. The researchers wanted to know: Can we tell who is walking just by the sound of their footsteps?

The Experiment: A Soundproof "Listening Booth"

To test this, the scientists built a special "listening booth" (a soundproof box with a metal plate inside). They treated it like a recording studio for tiny animals.

They invited six different "musicians" (invertebrate species) to perform solo:

1. The House Cricket: A leggy jumper that chirps.
2. The Garden Snail: A slow, slimy glider.
3. The Red Wiggler Worm: A legless wriggler.
4. The Cockroach: A fast, six-legged runner.
5. The Huntsman Spider: An eight-legged stalker.
6. The Mealworm: A beetle larva that looks like a worm but has legs.

They recorded each animal moving around for five minutes, capturing the tiny "thumps, scrapes, and clicks" they made against the metal plate.

The Big Discovery: Each Species Has a Unique "Voiceprint"

The researchers used computers to analyze these recordings, looking for patterns in the sound waves. They found something amazing:

• Everyone sounds different: Just like you can tell your friend's voice from a stranger's on the phone, the computer could tell the difference between a cricket and a worm based on their movement sounds.
• Legs vs. No Legs: The animals naturally split into two groups. The ones with legs (crickets, spiders, cockroaches) made a "staccato" sound—lots of little taps and clicks. The ones without legs (worms, snails) made a smoother, "sliding" sound. It's like the difference between someone tapping their fingers on a table versus someone dragging a wet towel across it.
• Size doesn't matter: You might think a heavy animal would sound louder or deeper than a light one. But the study found that what the animal is (its species and body shape) mattered much more than how much it weighs. A heavy worm sounds more like a light worm than it sounds like a heavy beetle.

Why This Matters

This study is like the "proof of concept" for a new kind of soil thermometer.

• Before: To check soil health, we had to dig, kill, and count bugs. It was like checking the health of a forest by cutting down trees to count the leaves.
• Now: We can potentially drop a microphone in the ground and listen. If the "footsteps" sound like a healthy mix of spiders, worms, and beetles, the soil is doing well. If the soundscape goes silent or changes to just one type of sound, we know something is wrong.

The Future

Right now, this was done in a quiet lab with one animal at a time. The next step is to teach computers to listen to the "chaotic noise" of a real forest floor, where hundreds of animals are moving at once.

If we succeed, we could have a global network of "soil microphones" that give us a real-time report card on the health of our planet's underground city, helping us restore nature without ever having to dig a hole. It turns the soil from a silent mystery into a conversation we can finally understand.

Technical Summary

1. Problem Statement

Soil degradation is a global crisis affecting food security and biodiversity, with up to 75% of soils already degraded. Restoring these ecosystems requires effective monitoring of soil biodiversity, particularly invertebrates which drive nutrient cycling and soil aggregation. However, traditional monitoring methods (e.g., pitfall traps, Berlese funnels, DNA barcoding) are labor-intensive, destructive, and difficult to scale. While soil ecoacoustics (passive acoustic monitoring) has emerged as a non-destructive alternative, current applications rely on aggregate soundscape indices (e.g., Acoustic Complexity Index) that provide only coarse biodiversity overviews. There is a critical gap in knowledge regarding whether taxon-specific acoustic signatures can be resolved from soil invertebrates to enable species-level or functional-group identification.

2. Methodology

The study employed a proof-of-concept experimental design using a standardized, low-cost recording system to test if distinct invertebrate taxa produce unique acoustic profiles.

• Experimental Setup:
  • Environment: A custom-built, sound-attenuated chamber (54L plastic container lined with foam) to minimize ambient noise.
  • Substrate: A 30 × 45 cm lipped aluminium plate (1 mm thickness) served as the recording surface to standardize vibration transmission.
  • Hardware: A JrF C-Series Pro contact microphone (optimized for low-to-mid frequencies) was affixed beneath the plate, connected to a micBooster pre-amp and a Zoom F6 multitrack recorder (24-bit, 48 kHz sampling rate).
  • Verification: An overhead camera (Samsung Galaxy S25+) recorded video to correlate visual movement with acoustic data.
• Subjects: Six morphologically and behaviorally distinct species were selected:
  • Podous (legged): House cricket (Acheta domesticus), Garden snail (Cornu aspersum - note: snails are technically apodous but included in the study's morphological grouping for comparison, though the text clarifies snails are apodous), Red wiggler worm (Eisenia fetida), Lobster cockroach (Nauphoeta cinerea), Huntsman spider (Pediana horni), and Mealworm (Zophobas morio).
  • Correction on Morphology: The study specifically groups them into Podous (Cockroach, Spider, Cricket) and Apodous (Worm, Snail, Mealworm).
  • Sample Size: 10 individuals per species (total $N=60$), with body mass measured prior to recording.
• Data Processing & Feature Extraction:
  • Pre-processing: Audio files were converted to mono, amplitude-normalized, and trimmed (removing 10s from start/end). A 2–12 kHz band-pass filter was applied to remove low-frequency background noise.
  • Feature Extraction: Using Python (librosa, scipy), 19 audio features were extracted, including:
    • 13 Mel-frequency cepstral coefficients (MFCCs) to capture spectral envelope shape.
    • Temporal/Spatial descriptors: Peak frequency, RMS energy, spectral centroid, spectral bandwidth, zero-crossing rate, and average inter-pulse interval.
• Statistical Analysis:
  • PCA: Principal Component Analysis on scaled, uncorrelated features to visualize clustering.
  • PERMANOVA: To test for significant differences in acoustic profiles among species.
  • RDA (Redundancy Analysis): To determine if body mass explained acoustic variation after accounting for species identity.
  • Randomized ANOVAs: To test differences in individual audio parameters.

3. Key Contributions

• Proof-of-Concept for Taxonomic Resolution: Demonstrated that soil invertebrates produce distinct, quantifiable acoustic profiles that allow for statistical differentiation between species under controlled conditions.
• Morphological Drivers of Acoustics: Established that acoustic differentiation is driven by taxon identity and morphological traits (specifically the presence/absence of legs: podous vs. apodous) rather than simple body mass scaling.
• Low-Cost Standardized Protocol: Developed a reproducible, low-cost (<AUD$1500) recording and analysis pipeline suitable for scaling up soil biodiversity monitoring.
• Feature Selection: Identified that MFCCs (spectral shape descriptors) are the primary drivers of interspecific differentiation, outperforming simple energy or frequency metrics.

4. Key Results

• Distinct Clustering: The acoustic profiles of the six species differed significantly (PERMANOVA: $R^2 = 0.505$, $p < 0.001$).
• Podous vs. Apodous Separation:
  • Podous taxa (Cockroach, Spider, Cricket) showed distinct clustering, with cockroaches and spiders separated along the first principal component (PC1).
  • Apodous taxa (Worm, Snail, Mealworm) clustered closely together, suggesting that legless organisms produce more similar acoustic signatures.
  • Note: Garden worms and snails were not significantly different from each other ($p = 0.128$), but all other pairwise comparisons were significant.
• Role of Body Mass: Redundancy Analysis (RDA) revealed that body mass had no significant effect on acoustic profiles once species identity was accounted for ($p = 0.878$). This refutes the hypothesis that acoustic differences are merely a function of size.
• Feature Sensitivity: 18 out of 19 audio parameters differed significantly among species. MFCCs (specifically 3, 5, 9, 12, and 13) were the strongest drivers of separation. RMS energy was the only parameter that did not vary by species.
• Dispersion: Cockroaches exhibited higher within-species acoustic dispersion, likely due to complex behaviors (stridulation, leg tapping), whereas apodous taxa showed lower variation.

5. Significance and Future Implications

• Advancing Soil Ecoacoustics: This study moves the field beyond aggregate soundscape indices toward taxonomic resolution. It provides the empirical foundation for developing automated acoustic classifiers capable of identifying specific soil invertebrate groups.
• Non-Destructive Monitoring: The ability to detect active invertebrates acoustically offers a non-destructive alternative to pitfall traps and soil coring, crucial for long-term restoration monitoring where preserving soil structure is vital.
• Restoration Applications: Such tools could track the colonization trajectories of functional groups (e.g., decomposers) and assess restoration success more rapidly and at a larger scale than current methods.
• Limitations & Future Work: The study was conducted on a homogeneous aluminium substrate in a quiet environment. Future research must address:
  • Signal propagation through complex media (soil, leaf litter, varying moisture).
  • Discrimination in mixed-species assemblages (mock communities) rather than isolated individuals.
  • Field validation against established biodiversity assays.

In conclusion, the paper validates that substrate-borne vibrations contain sufficient taxon-specific information to distinguish between common soil invertebrates, driven by morphology rather than mass, paving the way for scalable, automated soil biodiversity monitoring.