Hearing the heat: a proof-of-concept study of nonlinear sound measures for distinguishing phases of rapid organic decomposition
Abstract
Acoustic recording can follow decomposition non-destructively, but which acoustic measures carry process information is largely untested. In a proof-of-concept experiment, four nonlinear complexity measures and six conventional ecoacoustic indices were compared in eight octave bands, using 150 recordings from five buckets of one vermicompost-hay batch during one mesophilic heating-cooling cycle. Internal temperature served as a proxy for the system state. Because smooth trajectories can agree by chance, a four-parameter temperature model with a heating-cooling term was evaluated on held-out buckets and against 100000 surrogate temperature courses with the same values and smoothness but randomised timing. For the nonlinear measures, the model predicted held-out buckets with a median skill of 0.48, better than a linear time trend, a time model with the same heating-cooling break and the surrogates (p = 0.00003). The conventional indices reached 0.12 and were not distinguishable from the surrogates (p = 0.21). The group difference exceeded that of the surrogates (p = 0.0002). At equal temperature, the nonlinear measures differed between heating and cooling. Their skill was highest at 128–256Hz (median held-out skill 0.70). The temperature model nearly matched a data-driven benchmark without temperature for the nonlinear measures (0.50), but not for the conventional indices (0.18). In-sample fits overstated predictive skill (median R2 0.81 and 0.36), so held-out prediction against surrogate covariates provides a practical test for acoustic indicators in single experiments. Since the buckets replicate the measurement, not the process, replication across batches and cycles is needed.
Related
- New Scalar Measures of Geodetic Strain and Moment Rate from Kinematic Mode Decomposition: Application to the Southwestern United States
- Distinguishing leaf scorching from senescence under climate extremes
- Distinguishing run-up height from pressure distribution during avalanche impact on narrow obstacles: mechanisms and semi-empirical prediction from three-dimensional simulations
- Distinguishing Initial-Condition and Model Error Using Machine-Learning Weather Models
- Investigating the Role of Good Governance Components in Distinguishing Levels of Possible Scenarios for the Adoption of Water Management Technologies in Isfahan Province
BibTeX
@misc{eisenhauer2026hearing,
title = {{Hearing the heat: a proof-of-concept study of nonlinear sound measures for distinguishing phases of rapid organic decomposition}},
author = {Jakob Eisenhauer},
howpublished = {EarthArXiv},
year = {2026},
month = nov,
doi = {10.31223/x5123j},
url = {https://eartharxiv.org/repository/view/15995/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Hearing the heat: a proof-of-concept study of nonlinear sound measures for distinguishing phases of rapid organic decomposition
AU - Jakob Eisenhauer
PY - 2026
DA - 2026/11/07
PB - EarthArXiv
DO - 10.31223/x5123j
UR - https://eartharxiv.org/repository/view/15995/
AB - Acoustic recording can follow decomposition non-destructively, but which acoustic measures carry process information is largely untested. In a proof-of-concept experiment, four nonlinear complexity measures and six conventional ecoacoustic indices were compared in eight octave bands, using 150 recordings from five buckets of one vermicompost-hay batch during one mesophilic heating-cooling cycle. Internal temperature served as a proxy for the system state. Because smooth trajectories can agree by chance, a four-parameter temperature model with a heating-cooling term was evaluated on held-out buckets and against 100000 surrogate temperature courses with the same values and smoothness but randomised timing. For the nonlinear measures, the model predicted held-out buckets with a median skill of 0.48, better than a linear time trend, a time model with the same heating-cooling break and the surrogates (p = 0.00003). The conventional indices reached 0.12 and were not distinguishable from the surrogates (p = 0.21). The group difference exceeded that of the surrogates (p = 0.0002). At equal temperature, the nonlinear measures differed between heating and cooling. Their skill was highest at 128–256Hz (median held-out skill 0.70). The temperature model nearly matched a data-driven benchmark without temperature for the nonlinear measures (0.50), but not for the conventional indices (0.18). In-sample fits overstated predictive skill (median R2 0.81 and 0.36), so held-out prediction against surrogate covariates provides a practical test for acoustic indicators in single experiments. Since the buckets replicate the measurement, not the process, replication across batches and cycles is needed.
ER -