Ryan A. McCarthy
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Ryan A. McCarthy has a total of 43 co-authors. Here are the top co-authors by number of shared publications:
- Ananya Sen Gupta
Department of Electrical and Computer Engineering, The University of Iowa, Iowa City, IA, USA
6 shared publications - Adam Patrick Young
Scripps Institution of Oceanography University of California San Diego La Jolla CA USA
2 shared publications - Bernice Kubicek
Electrical and Computer Engineering, University of Iowa, Iowa City, IA 52242, USA
2 shared publications Charles Brooks
Scripps Institution of Oceanography, University of California San Diego, La Jolla, CA, USA
2 shared publicationsDerek Ung
Scripps Institution of Oceanography, University of California San Diego, La Jolla, CA, USA
2 shared publicationsEric Terrill
Marine Physical Lab., Scripps Inst. of Oceanogr., Univ. of California San Diego, San Diego, CA
2 shared publications
Latest Publications
- Open Access
- 9 September 2025
- Published by Springer Nature in npj Acoustics
- Vol. 1 (1) , 1-18
Machine Learning in Acoustics: A Review and Open-source Repository
- 9 September 2025
- Published by Springer Nature in npj Acoustics
- Vol. 1 (1) , 1-18
Open AccessMachine Learning in Acoustics: A Review and Open-source Repository
- Open Access
Machine Learning in Acoustics: A Review and Open-Source Repository
Acoustic data provide scientific and engineering insights in fields ranging from bioacoustics and communications to ocean and earth sciences. In this review, we survey recent advances and the transformative potential of machine learning (ML) in acoustics, including deep learning (DL). Using the Python high-level programming language, we demonstrate a broad collection of ML techniques to detect and find patterns for classification, regression, and generation in acoustics data automatically. We have ML examples including acoustic data classification, generative modeling for spatial audio, and physics-informed neural networks. This work includes AcousticsML, a set of practical Jupyter notebook examples on GitHub demonstrating ML benefits and encouraging researchers and practitioners to apply reproducible data-driven approaches to acoustic challenges.
Open AccessMachine Learning in Acoustics: A Review and Open-Source Repository
Acoustic data provide scientific and engineering insights in fields ranging from bioacoustics and communications to ocean and earth sciences. In this review, we survey recent advances and the transformative potential of machine learning (ML) in acoustics, including deep learning (DL). Using the Python high-level programming language, we demonstrate a broad collection of ML techniques to detect and find patterns for classification, regression, and generation in acoustics data automatically. We have ML examples including acoustic data classification, generative modeling for spatial audio, and physics-informed neural networks. This work includes AcousticsML, a set of practical Jupyter notebook examples on GitHub demonstrating ML benefits and encouraging researchers and practitioners to apply reproducible data-driven approaches to acoustic challenges.
- 11 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 50 (3) , 1668-1675
Machine Learning Transmission Loss Predictions in Acoustic Field Experiments
This study evaluates the capabilities of three distinct transmission loss (TL) models to use as an on-board decision aid to identify locations in depth and range for an autonomous underwater vehicle (AUV) and an uncrewed surface vehicle (USV) to improve acoustic communications. AUVs and USVs are computationally and resource limited vehicles that require fast in-situ decisions to reposition. This work utilizes an experiment in August 2023 off the coast of Southern California of an SV3 Wave Glider communicating with a REMUS 100 vehicle equipped with a 25 kHz acoustic modem. Observations of TL across a range dependent bathymetry are used to evaluate various decision aid models that can provide acoustical awareness for the vehicles. Models are based on: first, simple approximations of sound intensity that decreases at range; second, a ray-based physics model Bellhop; and third, a machine learning (ML) decision tree model trained off-board the vehicle using Bellhop simulations. Predictions from the ML model show good agreement with the observed field data with an overall accuracy of 85.42%. Although there are discrepancies between locations of acceptable TL and those observed by the vehicles, the ML model is able to perform as well as the Bellhop model while improving the prediction speed of TL at an individual range and depth from 764.23 to 0.57 ms.
- 11 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 50 (3) , 1668-1675
Machine Learning Transmission Loss Predictions in Acoustic Field Experiments
This study evaluates the capabilities of three distinct transmission loss (TL) models to use as an on-board decision aid to identify locations in depth and range for an autonomous underwater vehicle (AUV) and an uncrewed surface vehicle (USV) to improve acoustic communications. AUVs and USVs are computationally and resource limited vehicles that require fast in-situ decisions to reposition. This work utilizes an experiment in August 2023 off the coast of Southern California of an SV3 Wave Glider communicating with a REMUS 100 vehicle equipped with a 25 kHz acoustic modem. Observations of TL across a range dependent bathymetry are used to evaluate various decision aid models that can provide acoustical awareness for the vehicles. Models are based on: first, simple approximations of sound intensity that decreases at range; second, a ray-based physics model Bellhop; and third, a machine learning (ML) decision tree model trained off-board the vehicle using Bellhop simulations. Predictions from the ML model show good agreement with the observed field data with an overall accuracy of 85.42%. Although there are discrepancies between locations of acceptable TL and those observed by the vehicles, the ML model is able to perform as well as the Bellhop model while improving the prediction speed of TL at an individual range and depth from 764.23 to 0.57 ms.
- Open Access
Deep-ocean macrofaunal assemblages on ferromanganese and phosphorite-rich substrates in the Southern California Borderland
Mineral-rich hardgrounds, such as ferromanganese (FeMn) crusts and phosphorites, occur on seamounts and continental margins, gaining attention for their resource potential due to their enrichment in valuable metals in some regions. This study focuses on the Southern California Borderland (SCB), an area characterized by uneven and heterogeneous topography featuring FeMn crusts, phosphorites, basalt, and sedimentary rocks that occur at varying depths and are exposed to a range of oxygen concentrations. Due to its heterogeneity, this region serves as an optimal setting for investigating the relationship between mineral-rich hardgrounds and benthic fauna. This study characterizes the density, diversity, and community composition of macrofauna (>300 μm) on hardgrounds as a function of substrate type and environment (depth and oxygen ranges). Rocks and their macrofauna were sampled quantitatively using remotely operated vehicles (ROVs) during expeditions in 2020 and 2021 at depths above, within, and below the oxygen minimum zone (OMZ). A total of 3,555 macrofauna individuals were counted and 416 different morphospecies (excluding encrusting bryozoans and hydrozoans) were identified from 82 rocks at depths between 231 and 2,688 m. Average density for SCB macrofauna was 11.08 ± 0.87 ind. 200 cm−2 and mean Shannon-Wiener diversity per rock (H′[loge]) was 2.22 ± 0.07. A relationship was found between substrate type and macrofaunal communities. Phosphorite rocks had the highest H′ of the four substrates compared on a per-rock basis. However, when samples were pooled by substrate, FeMn crusts had the highest H′ and rarefaction diversity. Of all the environmental variables examined, water depth explained the largest variance in macrofaunal community composition. Macrofaunal density and diversity values were similar at sites within and outside the OMZ. This study is the first to analyze the macrofaunal communities of mineral-rich hardgrounds in the SCB, which support deep-ocean biodiversity by acting as specialized substrates for macrofaunal communities. Understanding the intricate relationships between macrofaunal assemblages and mineral-rich substrates may inform effects from environmental disruptions associated with deep-seabed mining or climate change. The findings contribute baseline information useful for effective conservation and management of the SCB and will support scientists in monitoring changes in these communities due to environmental disturbance or human impact in the future.
Open AccessDeep-ocean macrofaunal assemblages on ferromanganese and phosphorite-rich substrates in the Southern California Borderland
Mineral-rich hardgrounds, such as ferromanganese (FeMn) crusts and phosphorites, occur on seamounts and continental margins, gaining attention for their resource potential due to their enrichment in valuable metals in some regions. This study focuses on the Southern California Borderland (SCB), an area characterized by uneven and heterogeneous topography featuring FeMn crusts, phosphorites, basalt, and sedimentary rocks that occur at varying depths and are exposed to a range of oxygen concentrations. Due to its heterogeneity, this region serves as an optimal setting for investigating the relationship between mineral-rich hardgrounds and benthic fauna. This study characterizes the density, diversity, and community composition of macrofauna (>300 μm) on hardgrounds as a function of substrate type and environment (depth and oxygen ranges). Rocks and their macrofauna were sampled quantitatively using remotely operated vehicles (ROVs) during expeditions in 2020 and 2021 at depths above, within, and below the oxygen minimum zone (OMZ). A total of 3,555 macrofauna individuals were counted and 416 different morphospecies (excluding encrusting bryozoans and hydrozoans) were identified from 82 rocks at depths between 231 and 2,688 m. Average density for SCB macrofauna was 11.08 ± 0.87 ind. 200 cm−2 and mean Shannon-Wiener diversity per rock (H′[loge]) was 2.22 ± 0.07. A relationship was found between substrate type and macrofaunal communities. Phosphorite rocks had the highest H′ of the four substrates compared on a per-rock basis. However, when samples were pooled by substrate, FeMn crusts had the highest H′ and rarefaction diversity. Of all the environmental variables examined, water depth explained the largest variance in macrofaunal community composition. Macrofaunal density and diversity values were similar at sites within and outside the OMZ. This study is the first to analyze the macrofaunal communities of mineral-rich hardgrounds in the SCB, which support deep-ocean biodiversity by acting as specialized substrates for macrofaunal communities. Understanding the intricate relationships between macrofaunal assemblages and mineral-rich substrates may inform effects from environmental disruptions associated with deep-seabed mining or climate change. The findings contribute baseline information useful for effective conservation and management of the SCB and will support scientists in monitoring changes in these communities due to environmental disturbance or human impact in the future.
- Open Access
- 17 October 2024
- Published by American Geophysical Union (AGU) in Journal of Geophysical Research: Earth Surface
- Vol. 129 (10)
Subaerial Profiles at Two Beaches: Equilibrium and Machine Learning
Shoreline position (e.g., beach width) is a critical component of flooding and overtopping forecasts but difficult to predict accurately. We model beach width changes with a supervised machine learning (ML) approach informed by equilibrium principles. The time history of wave energy anomalies that force equilibrium models is used as an ML input feature. The sweeping simplifying equilibrium model assumptions relating beach width change to anomalies are replaced with data‐based ML results. Supervised learning regression methods including linear, support vector, decision trees, and ensemble regressors are tested. Observations for model training and testing includes weekly to quarterly beach elevation surveys spanning approximately 500 m alongshore and 8 years at two beaches, each supplemented with several months of ∼100 sub‐weekly surveys. These beaches, with different sediment types (sand vs. sand‐cobble mix), both widen in summer in response to the seasonal wave climate, in agreement with a generic equilibrium model. Differences in backshore erodability contribute to differing beach responses in the stormiest (El Niño) year that are reproduced by a simple extra trees regression model but not by the equilibrium model. With sufficiently extensive training data, the ML model outperforms equilibrium by providing flexibility and complexity in the response to wave forcing. The present ML and equilibrium models both fail to simulate a uniquely stunted beach recovery unlike other recoveries in the training data.
- 17 October 2024
- Published by American Geophysical Union (AGU) in Journal of Geophysical Research: Earth Surface
- Vol. 129 (10)
Open AccessSubaerial Profiles at Two Beaches: Equilibrium and Machine Learning
Shoreline position (e.g., beach width) is a critical component of flooding and overtopping forecasts but difficult to predict accurately. We model beach width changes with a supervised machine learning (ML) approach informed by equilibrium principles. The time history of wave energy anomalies that force equilibrium models is used as an ML input feature. The sweeping simplifying equilibrium model assumptions relating beach width change to anomalies are replaced with data‐based ML results. Supervised learning regression methods including linear, support vector, decision trees, and ensemble regressors are tested. Observations for model training and testing includes weekly to quarterly beach elevation surveys spanning approximately 500 m alongshore and 8 years at two beaches, each supplemented with several months of ∼100 sub‐weekly surveys. These beaches, with different sediment types (sand vs. sand‐cobble mix), both widen in summer in response to the seasonal wave climate, in agreement with a generic equilibrium model. Differences in backshore erodability contribute to differing beach responses in the stormiest (El Niño) year that are reproduced by a simple extra trees regression model but not by the equilibrium model. With sufficiently extensive training data, the ML model outperforms equilibrium by providing flexibility and complexity in the response to wave forcing. The present ML and equilibrium models both fail to simulate a uniquely stunted beach recovery unlike other recoveries in the training data.
- Open Access
- 14 November 2023
- Published by Wiley
Subaerial Profiles at Two Beaches: Equilibrium and Machine Learning
Weekly to quarterly beach elevation surveys spanning 700-800 m alongshore and 8 years at two beaches were each supplemented with several months of ∼100 sub-weekly surveys. These beaches, which have different sediment types (sand vs. sand-cobble mix), both widen in summer in response to the seasonal wave climate, in agreement with a generic equilibrium model. Results suggest differences in backshore erodability contribute to differing beach responses in the stormiest (El Niño) year. At both sites, the time dependence of the equilibrium modeled shoreline resembles the first mode of an EOF decomposition of the observations. With sufficient training, an equilibrium-informed Extra Tree Regression model, that includes features motivated by equilibrium modelling, can significantly outperform a generic equilibrium model.
- 14 November 2023
- Published by Wiley
Open AccessSubaerial Profiles at Two Beaches: Equilibrium and Machine Learning
Weekly to quarterly beach elevation surveys spanning 700-800 m alongshore and 8 years at two beaches were each supplemented with several months of ∼100 sub-weekly surveys. These beaches, which have different sediment types (sand vs. sand-cobble mix), both widen in summer in response to the seasonal wave climate, in agreement with a generic equilibrium model. Results suggest differences in backshore erodability contribute to differing beach responses in the stormiest (El Niño) year. At both sites, the time dependence of the equilibrium modeled shoreline resembles the first mode of an EOF decomposition of the observations. With sufficient training, an equilibrium-informed Extra Tree Regression model, that includes features motivated by equilibrium modelling, can significantly outperform a generic equilibrium model.
- Open Access
- 1 October 2023
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 154 (4_suppleme) , A308
Machine learning transmission loss simulations in complex undersea environments with range-dependent bathymetry
Range-dependent bathymetric variability introduces diverse scattering and multipath that can impact optimal acoustic communication ranges, and will impact autonomous underwater vehicle (AUV) operations. We are working towards building a sound-aware framework in which the AUV has knowledge of the transmission environment and can exploit or react with intended outcomes. Prior knowledge of the environment coupled with acoustic propagation modeling on-board AUVs can improve sound-awareness of a vehicle but is computationally intensive, and costly from a power-budget perspective. This work addresses the computational burden by training a machine learning (ML) model to interpret range-dependent acoustic propagation through environmental inputs to predict transmission loss (TL) from the ray tracing model, BELLHOP. A decision tree model is trained to predict locations in range and depth of acceptable TL using feature representations to reduce bathymetric information within a region. Models are trained and tested with TL field realizations for acoustic communications at 10 and 25 kHz for environments with varying bathymetry at ranges up to 3 km through BELLHOP. We present simulations and results from field experiments performed off the coast of Southern California. The communication paths are between a surface vehicle (Liquid Robotics Waveglider, Herdon, VA) and an AUV (REMUS UUV, Huntington Ingalls, Falmouth, MA). Results for model predictions and collected field data will be discussed.
- 1 October 2023
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 154 (4_suppleme) , A308
Open AccessMachine learning transmission loss simulations in complex undersea environments with range-dependent bathymetry
Range-dependent bathymetric variability introduces diverse scattering and multipath that can impact optimal acoustic communication ranges, and will impact autonomous underwater vehicle (AUV) operations. We are working towards building a sound-aware framework in which the AUV has knowledge of the transmission environment and can exploit or react with intended outcomes. Prior knowledge of the environment coupled with acoustic propagation modeling on-board AUVs can improve sound-awareness of a vehicle but is computationally intensive, and costly from a power-budget perspective. This work addresses the computational burden by training a machine learning (ML) model to interpret range-dependent acoustic propagation through environmental inputs to predict transmission loss (TL) from the ray tracing model, BELLHOP. A decision tree model is trained to predict locations in range and depth of acceptable TL using feature representations to reduce bathymetric information within a region. Models are trained and tested with TL field realizations for acoustic communications at 10 and 25 kHz for environments with varying bathymetry at ranges up to 3 km through BELLHOP. We present simulations and results from field experiments performed off the coast of Southern California. The communication paths are between a surface vehicle (Liquid Robotics Waveglider, Herdon, VA) and an AUV (REMUS UUV, Huntington Ingalls, Falmouth, MA). Results for model predictions and collected field data will be discussed.
- 15 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 48 (4) , 1149-1173
Reduced-Order Machine-Learning Model for Transmission Loss Prediction in Underwater Acoustics
Autonomous underwater vehicles (AUVs) typically rely on underwater sound to communicate with sound channel physics defining communication performance. Current acoustic propagation models incorporate channel information and physics but are computationally expensive to run onboard an AUV. This work explores reduced-order proxies for these models through a machine-learning approach that can use in-situ measurements of observed sound-speed profile (SSP) to quickly provide transmission loss (TL) propagation information for vehicles to optimize communication performance through repositioning behaviors. The vertical structure from a year-long time series of mooring observations from offshore San Diego is decomposed with empirical orthogonal functions to reduce the SSPs to three unique amplitudes that explain 95.6% of the total variance. The acoustic propagation model Bellhop is used to estimate TL for a given SSP and different thresholds of TL are used to represent communications performance. Machine-learning models are trained with binarized Bellhop outputs to indicate locations of improved underwater acoustic communications. Amplitude values, desired TL, vehicle depth, and range are used as inputs into machine-learning models for optimizing TL. The best trained model was a random forest (ensemble) network with a 93.3% dice score. The decision tree had a faster prediction speed, relative to Bellhop, with similar prediction accuracy as the ensemble. This work aims to improve sound awareness of AUVs by generating SSP-informed communication performance maps.
- 15 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Oceanic Engineering
- Vol. 48 (4) , 1149-1173
Reduced-Order Machine-Learning Model for Transmission Loss Prediction in Underwater Acoustics
Autonomous underwater vehicles (AUVs) typically rely on underwater sound to communicate with sound channel physics defining communication performance. Current acoustic propagation models incorporate channel information and physics but are computationally expensive to run onboard an AUV. This work explores reduced-order proxies for these models through a machine-learning approach that can use in-situ measurements of observed sound-speed profile (SSP) to quickly provide transmission loss (TL) propagation information for vehicles to optimize communication performance through repositioning behaviors. The vertical structure from a year-long time series of mooring observations from offshore San Diego is decomposed with empirical orthogonal functions to reduce the SSPs to three unique amplitudes that explain 95.6% of the total variance. The acoustic propagation model Bellhop is used to estimate TL for a given SSP and different thresholds of TL are used to represent communications performance. Machine-learning models are trained with binarized Bellhop outputs to indicate locations of improved underwater acoustic communications. Amplitude values, desired TL, vehicle depth, and range are used as inputs into machine-learning models for optimizing TL. The best trained model was a random forest (ensemble) network with a 93.3% dice score. The decision tree had a faster prediction speed, relative to Bellhop, with similar prediction accuracy as the ensemble. This work aims to improve sound awareness of AUVs by generating SSP-informed communication performance maps.
- Open Access
- 1 March 2023
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 153 (3_suppleme) , A175
Machine learning of acoustic propagation models for sound aware autonomous systems
Opportunities exist for autonomous underwater vehicles to optimize acoustic communications with knowledge of the sound channel, but implementation has traditionally been hindered by computationally expensive propagation models. This work addresses computational complexity by enabling machine learning to interpret environmental inputs to predict transmission loss (TL) outputs from a physics-based ray tracing model (BELLHOP). Feature representations to reduce bathymetric and sound speed profile (SSP) information through wavelet decomposition and empirical orthogonal analysis are explored, respectively. A decision tree machine learning algorithm is trained to learn the paths (depths and ranges) of acceptable transmission loss for a pair of acoustic modems within a given environment. Our test case is off the coast of Southern California, and SSPs are collected from a 3 year record sampled from an ocean mooring. Bathymetry data is based on high resolution multibeam data. TL field realizations are calculated as training and testing data for environments of varying bathymetry with frequency of 25kHz and range of 1km through the BELLHOP model. Results are shown with stationary and non-stationary emitters to demonstrate the effectiveness of this approach.
- 1 March 2023
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 153 (3_suppleme) , A175
Open AccessMachine learning of acoustic propagation models for sound aware autonomous systems
Opportunities exist for autonomous underwater vehicles to optimize acoustic communications with knowledge of the sound channel, but implementation has traditionally been hindered by computationally expensive propagation models. This work addresses computational complexity by enabling machine learning to interpret environmental inputs to predict transmission loss (TL) outputs from a physics-based ray tracing model (BELLHOP). Feature representations to reduce bathymetric and sound speed profile (SSP) information through wavelet decomposition and empirical orthogonal analysis are explored, respectively. A decision tree machine learning algorithm is trained to learn the paths (depths and ranges) of acceptable transmission loss for a pair of acoustic modems within a given environment. Our test case is off the coast of Southern California, and SSPs are collected from a 3 year record sampled from an ocean mooring. Bathymetry data is based on high resolution multibeam data. TL field realizations are calculated as training and testing data for environments of varying bathymetry with frequency of 25kHz and range of 1km through the BELLHOP model. Results are shown with stationary and non-stationary emitters to demonstrate the effectiveness of this approach.
- 1 August 2021
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 150 (2) , 906-919
Autonomous learning and interpretation of channel multipath scattering using braid manifolds in underwater acoustic communications
In this work, we explore machine learning through a model-agnostic feature representation known as braiding, that employs braid manifolds to interpret multipath ray bundles. We generate training and testing data using the well-known BELLHOP model to simulate shallow water acoustic channels across a wide range of multipath scattering activity. We examine three different machine learning techniques—k-nearest neighbors, random forest tree ensemble, and a fully connected neural network—as well as two machine learning applications. The first application applies known physical parameters and braid information to determine the number of reflections the acoustic signal may undergo through the environment. The second application applies braid path information to determine if a braid is an important representation of the channel (i.e., evolving across bands of higher amplitude activity in the channel). Testing accuracy of the best trained machine learning algorithm in the first application was 86.70% and the testing accuracy of the second application was 99.94%. This work can be potentially beneficial in examining how the reflectors in the environment changeover time while also determining relevant braids for faster channel estimation.
- 1 August 2021
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 150 (2) , 906-919
Autonomous learning and interpretation of channel multipath scattering using braid manifolds in underwater acoustic communications
In this work, we explore machine learning through a model-agnostic feature representation known as braiding, that employs braid manifolds to interpret multipath ray bundles. We generate training and testing data using the well-known BELLHOP model to simulate shallow water acoustic channels across a wide range of multipath scattering activity. We examine three different machine learning techniques—k-nearest neighbors, random forest tree ensemble, and a fully connected neural network—as well as two machine learning applications. The first application applies known physical parameters and braid information to determine the number of reflections the acoustic signal may undergo through the environment. The second application applies braid path information to determine if a braid is an important representation of the channel (i.e., evolving across bands of higher amplitude activity in the channel). Testing accuracy of the best trained machine learning algorithm in the first application was 86.70% and the testing accuracy of the second application was 99.94%. This work can be potentially beneficial in examining how the reflectors in the environment changeover time while also determining relevant braids for faster channel estimation.