Mussel culture monitoring with semi-supervised machine learning on multibeam echosounder data using label spreading.

High diversity seabed habitats, such as shellfish aggregations, play a significant role in marine ecosystem sustainability but are susceptible to bottom disturbance induced by anthropogenic activities. Regular monitoring of these habitats with effective mapping methods is therefore essential. Multibeam echosounder (MBES) has been widely used in recent decades for seabed characterization due to its non-destructive manner and extensive spatial coverage compared to traditional methods like bottom sampling. Nevertheless, bottom sampling remains essential to link ground truth with acoustic seabed classification. Using seabed samples and MBES measurements, machine learning techniques are commonly employed to model their relationships and generate classification maps of an extended seabed. However, limited ground truth data, resulting from constraints in regulations, budget, or time, may impede the development of robust machine learning models. To address this challenge, we applied a semi-supervised machine learning method to classify seabed sediments of a blue mussel (Mytilus edulis) cultivation area in the Oosterschelde, the Netherlands. We utilized nine boxcore samples to generate pseudo-labels on MBES data. These pseudo-labels enlarged the training data size, facilitated the training of three comprehensive machine learning algorithms (Gradient Boosting, Random Forest, and Support Vector Machine), and helped to classify the study site into mussel and non-mussel areas. We found the geomorphological and backscatter-related features to be complementary for mussel culture detection. Our classification results were demonstrated effective through expert knowledge of this cultivation area and brought insights for future research on natural mussel habitats.

Microphones as Airspeed Sensors for Unmanned Aerial Vehicles.

This paper puts forward a novel design for an airspeed instrument aimed at small fixed-wing tail-sitter unmanned aerial vehicles. The working principle is to relate the power spectra of the wall-pressure fluctuations beneath the turbulent boundary layer present over the vehicle's body in flight to its airspeed. The instrument consists of two microphones; one flush-mounted on the vehicle's nose cone, which captures the pseudo-sound caused by the turbulent boundary layer, and a micro-controller that processes the signals and computes the airspeed. A feed-forward single-layer neural network is used to predict the airspeed based on the power spectra of the microphones' signals. The neural network is trained using data obtained from wind tunnel and flight experiments. Several neural networks were trained and validated using only flight data, with the best one achieving a mean approximation error of 0.043 m/s and having a standard deviation of 1.039 m/s. The angle of attack has a significant impact on the measurement, but if the angle of attack is known, the airspeed could still be successfully predicted for a wide range of angles of attack.

Simulating the acoustic response of cavities to improve microphone array measurements in closed test section wind tunnels.

Cavities placed along wind tunnel walls can attenuate the turbulent boundary layer (TBL) fluctuations as they propagate into the cavity. Placing microphones within the cavities can thus improve the signal-to-noise ratio of acoustic data. However, standing waves form within these cavities distorting the acoustic measurements. This work uses a finite element (FE) solver to evaluate how cavity geometry (depth, diameter, and wall angle) and wall material (hard-walled and melamine foam) affect the amplitude and eigenfrequency of standing waves when excited by an incident acoustic plane wave. Good agreement between predicted and measured acoustic transfer functions is shown. Compared to cylindrical cavities, countersunk and conical cavities improve the overall response, i.e., reducing the quality factor quantifying the resonance and damping characteristics. Stainless steel coverings also reduce the quality factor. A finding is that the shape of the external foam holder rather than the cavity shape drives the standing wave characteristics for the melamine foam cavities. The optimization problem of minimizing the acoustic response while also attenuating the TBL is thus decoupled by using the melamine foam. Consequently, these considerations can be addressed independently by optimizing the outer cavity shape for acoustics and the melamine foam insert for TBL attenuation.

Analysis of shielding of propeller noise using beamforming and predictions.

Engine noise shielding is an important measure towards low-noise aircraft configurations. Such designs are supported by prediction tools that indicate high values for shielding of engine noise. Most prediction models approximate the complex nature of engine noise to simple noise sources such as monopoles or dipoles. This work compares predictions of noise shielding with experiments using different noise sources and shielding body geometries. The experiments considered in this work concern a monopole source shielded by a flat plate and a NACA 64-008 A wing, and a propeller shielded by the same wing. Comparisons between models and measurements are made by analysis of noise levels at individual microphones and using conventional beamforming. Results show that for the monopole cases the model predictions are in agreement with the experimental data, with an average deviation of 2-3 dB. The curvature of the leading edge of the wing influences the noise shielding results. The measured values of noise shielding of propeller noise are lower than those measured for the omni-directional source. Different types of source directivity are used to approximate the propeller in the predictions: monopole, dipole and a multi-source. The dipole approximation shows the best agreement with the experiments for the case of the propeller.

Assessing the shielding of engine noise by the wings for current aircraft using model predictions and measurements.

Reducing aircraft noise is a major issue to be dealt with by the aerospace industry. In addition to lowering noise emissions from the engine and airframe, also the shielding of engine noise by the aircraft is considered as a promising means for reducing the perceived noise on the ground. In literature, noise shielding predictions indicate significant reductions in received noise levels for blended wing body configurations, but also for conventional aircraft with the engines placed above the wings. Little work has been done in assessing these potential shielding effects for full aircraft under real operational conditions. Therefore, in this work, noise shielding for current aircraft is investigated using both measurements and model predictions. The predictions are based on the Kirchhoff integral theory and the Modified Theory of Physical Optics. For the comparison between the predictions and measurements, Twenty Fokker 70 flyovers are considered. The data analysis approach for the extraction of shielding levels for aircraft under these operational conditions is presented. Directly under the flight path, the simulations predict an engine noise shielding of 6 dB overall sound pressure level. This is confirmed by some of the flyover data. On average, the measurements show somewhat lower shielding levels.

Multi-beam echo-sounder bathymetric measurements: Implications of using frequency modulated pulses.

In this contribution bathymetric uncertainties induced by the use of frequency modulated (FM) signals for multi-beam-echo-sounder (MBES) measurements are quantified and their relevance for MBES bathymetric uncertainty predictions is assessed. When switching to FM, the quality of depth measurements can get deteriorated due to the Doppler effect and baseline decorrelation. The uncertainty due to the former is divided into second-order (imperfectness of the Doppler-range correction) and first-order (effect on beamsteering) effects. The latter also holds for continuous wave (CW) signals. Here, situations of relevance for measurements in the continental shelf and ship dynamics associated to rough and calm sea-states are considered, and the vertical uncertainty induced by the above sources is quantified. The influence of the Doppler effect depends on the sea state, but is found to potentially have a significant contribution to the MBES error budget for both FM and CW [nearly 82% (rough) and 68% (calm) of the total uncertainty]. The effect of baseline decorrelation depends on the actual pulse shape. For the specifications investigated, vertical uncertainties induced by this source are predicted to be larger for FM than that of CW. This is confirmed by a comparison between the modelled and measured effect on depth uncertainties when switching to FM.

On the use of global optimization methods for acoustic source mapping.

Conventional beamforming with a microphone array is a well-established method for localizing and quantifying sound sources. It provides estimates for the source strengths on a predefined grid by determining the agreement between the pressures measured and those modeled for a source located at the grid point under consideration. As such, conventional beamforming can be seen as an exhaustive search for those locations that provide a maximum match between measured and modeled pressures. In this contribution, the authors propose to, instead of the exhaustive search, use an efficient global optimization method to search for the source locations that maximize the agreement between model and measurement. Advantages are two-fold. First, the efficient optimization allows for inclusion of more unknowns, such as the source position in three-dimensional or environmental parameters such as the speed of sound. Second, the model for the received pressure field can be readily adapted to reflect, for example, the presence of more sound sources or environmental parameters that affect the received signals. For the work considered, the global optimization method, Differential Evolution, is selected. Results with simulated and experimental data show that sources can be accurately identified, including the distance from the source to the array.

Passive probing of the sound fixing and ranging channel with hydro-acoustic observations from ridge earthquakes.

The International Monitoring System includes a hydro-acoustic part to verify the Comprehensive Nuclear-Test-Ban Treaty. Besides explosive signals, monitoring stations also detect acoustic waves from earthquakes that travel through the SOund Fixing And Ranging (SOFAR) channel. The travel times of such detections are listed in the Reviewed Event Bulletin, which is statistically evaluated for the stations located in the Pacific, Indian, and Atlantic Oceans. The celerities of ridge earthquakes are calculated to build up a homogeneous data-set, based on similar source mechanisms. The celerity is defined as the epicentral distance divided by the travel time. The global characteristics of these celerities can be well understood in terms of temperature variations in the SOFAR channel. A detailed velocity profile has been retrieved for the Atlantic Ocean where large differences (14 m/s) are found between the southern and northern parts of the basin. Propagation modeling with normal modes supports these findings, which shows that the celerity gives an estimate of the sound speed in the SOFAR channel. These results compare remarkably well with those from active experiments, showing the ability of passively probing the SOFAR channel with hydro-acoustic waves from earthquake sources.

Observations regarding coarse sediment classification based on multi-beam echo-sounder's backscatter strength and depth residuals in Dutch rivers.

This contribution investigates the behavior of two important riverbed sediment classifiers, derived from multi-beam echo-sounder (MBES)-operating at 300 kHz-data, in very coarse sediment environments. These are the backscatter strength and the depth residuals. Four MBES data sets collected at different parts of rivers in the Netherlands are employed. From previous research the backscatter strength was found to increase for increasing mean grain sizes. Depth residuals, however, are often found to have lower values for coarser sediments. Investigation of the four data sets indicates that these statements are valid only for moderately coarse sediment such as sand. For very coarse sediments (e.g., coarse gravel) the backscatter strength is found to decrease and the depth residuals increase for increasing mean grain sizes. This is observed when the sediment mean grain size becomes significantly larger than the acoustic wavelength of the MBES (5 mm). Knowledge regarding this behavior is of high importance when using backscatter strength and depth residuals for sediment classification purposes as the reverse in behavior can induce ambiguity in the classification.

Long range acoustic measurements of an undersea volcano.

A seamount 8 km southeast of Sarigan Island erupted on 29 May 2010 and was visually observed. The recordings on two sets of hydrophones, operated by International Monitoring System (IMS) of the Comprehensive Test Ban Treaty Organization (CTBTO) are analyzed. Each array is a triplet of axial single hydrophones deployed as a 2 km triangle. Measurements of acoustic intensity for the path to the southern triplet are on the order of 6 dB lower than those received on the northern triplet. Temporal cross-correlation beamforming estimation is performed and the estimated arrival angles for the two arrays, 265° and 267° were consistent with the predicted geodesic arrival of 264.6° and 267.8°, respectively. Cross-correlation between single phones on the northern and southern arrays reveals a peak at 266°, with a cross-correlation of 0.1. Nx2D parabolic equation modeling predicts complete blockage due to seamount interaction along the geodesic path. Overprediction of the seamount blockage indicates that the 2D approximation is incorrect, and three-dimensional propagation must be used to explain the observations. This is demonstrated by the computation of the Adiabatic Mode Parabolic Equation Transmission Loss, which predicts a 5-10 dB lower reception at the southern site.

An inter-comparison of sediment classification methods based on multi-beam echo-sounder backscatter and sediment natural radioactivity data.

This contribution presents sediment classification results derived from different sources of data collected at the Dordtse Kil river, the Netherlands. The first source is a multi-beam echo-sounder (MBES). The second source is measurements taken with a gamma-ray scintillation detector, i.e., the Multi-Element Detection System for Underwater Sediment Activity (Medusa), towed over the sediments and measuring sediment natural radioactivity. Two analysis methods are employed for sediment classification based on the MBES data. The first is a Bayesian estimation method that uses the average backscatter data per beam and, therefore, is independent of the quality of the MBES calibration. The second is a model-based method that matches the measured backscatter curves to theoretical curves, predicted by a physics-based model. Medusa provides estimates for the concentrations of potassium, uranium, thorium, and cesium, known to be indicative for sediment properties, viz. mean grain size, silt content, and the presence of organic matter. In addition, a hydrophone attached to the Medusa system provides information regarding the sediment roughness. This paper presents an inter-comparison between the sediment classification results using the above-mentioned methods. It is shown that although originating from completely different sources, the MBES and Medusa provide similar information, revealing the same sediment distribution.

Nitrogen deposition from aviation emissions.

Excess nitrogen deposition from anthropogenic sources of atmospheric emissions, such as agriculture and transportation, can have negative effects on natural environments. Designing effective conservation efforts requires knowledge of the contribution of individual sectors. This study utilizes a global atmospheric chemistry-transport model to quantify, for the first time, the contribution of global aviation NOx emissions to nitrogen deposition for 2005 and 2019. We find that aviation led to an additional 1.39 Tg of nitrogen deposited globally in 2019, up 72 % from 2005, with 67 % of each year's total occurring through wet deposition. In 2019, aviation was responsible for an average of 0.66 %, 1.13 %, and 1.61 % of modeled nitrogen deposition from all sources over Asia, Europe, and North America, respectively. These impacts are spatially widespread, with 56 % of deposition occurring over water. Emissions during the landing, taxi and takeoff (LTO) phases of flight are responsible for 8 % of aviation's nitrogen deposition impacts on average globally, and between 16 and 32 % over most land in regions with high aviation activity. Despite currently representing less than 1.2 % of nitrogen deposition globally, further growth of aviation emissions would result in increases in aviation's contribution to nitrogen deposition and associated critical loads.

Improving riverbed sediment classification using backscatter and depth residual features of multi-beam echo-sounder systems.

Riverbed and seafloor sediment classification using acoustic remote sensing techniques is of high interest due to their high coverage capabilities at limited cost. This contribution presents the results of riverbed sediment classification using multi-beam echo-sounder data based on an empirical method. Two data sets are considered, both taken at the Waal River, namely Sint Andries and Nijmegen. This work is a follow-up to the work carried out by Amiri-Simkooei et al. [J. Acoust. Soc. Am. 126(4), 1724-1738 (2009)]. The empirical method bases the classification on features of the backscatter strength and depth residuals. A principal component analysis is used to identify the most appropriate and informative features. Clustering is then applied to the principal components resulting from this set of features to assign a sediment class to each measurement. The results show that the backscatter strength features discriminate between different classes based on the sediment properties, whereas the depth residual features discriminate classes based on riverbed forms such as the "fixed layer" (stone having riprap structure) and riverbed ripples. Combination of these two sets of features is highly recommended because they provide complementary information on both the composition and the structure of the riverbed.

Using global optimization methods for three-dimensional localization and quantification of incoherent acoustic sources.

Complex acoustic systems typically present three-dimensional distributions of noise sources. Conventional acoustic imaging methods with planar microphone arrays are unsuitable for three-dimensional acoustic imaging, given the computational demands and the incapability to explicitly account for the presence of multiple sources. This paper proposes the use of global optimization methods to solve these shortcomings. An experiment with three incoherent speakers proved that this method can accurately determine the three-dimensional location and the respective sound level of each individual source. In addition, super-resolution is achieved beyond half the Rayleigh resolution limit.

Model-based sediment classification using single-beam echosounder signals.

Acoustic remote sensing techniques for mapping sediment properties are of interest due to their low costs and high coverage. Model-based approaches directly couple the acoustic signals to sediment properties. Despite the limited coverage of the single-beam echosounder (SBES), it is widely used. Having available model-based SBES classification tools, therefore, is important. Here, two model-based approaches of different complexity are compared to investigate their practical applicability. The first approach is based on matching the echo envelope. It maximally exploits the information available in the signal but requires complex modeling and optimization. To minimize computational costs, the efficient differential evolution method is used. The second approach reduces the information of the signal to energy only and directly relates this to the reflection coefficient to obtain quantitative information about the sediment parameters. The first approach provides information over a variety of sediment types. In addition to sediment mean grain size, it also provides estimates for the spectral strength and volume scattering parameter. The need to account for all three parameters is demonstrated, justifying computational expenses. In the second approach, the lack of information on these parameters and the limited SBES beamwidth are demonstrated to hamper the conversion of echo energy to reflection coefficient.

Riverbed sediment classification using multi-beam echo-sounder backscatter data.

A method has recently been developed that employs multi-beam echo-sounder backscatter data to both obtain the number of sediment classes and discriminate between them by applying the Bayes decision rule to multiple hypotheses [Simons and Snellen, Appl. Acoust. 70, 1258-1268 (2009)]. In deep water, the number of scatter pixels within the beam footprint is large enough to ensure Gaussian distributions for the backscatter strengths and to increase the discriminative power between acoustic classes. In very shallow water (<10 m), however, this number is too small. This paper presents an extension of this high-frequency methodology for these environments, together with a demonstration of its performance using backscatter data from the river Waal, The Netherlands. The objective of this work is threefold. (i) Increasing the discriminating power of the classification method: high-resolution bathymetry data allow precise bottom slope corrections for obtaining the true incident angle, and the high-resolution backscatter data reduce the statistical fluctuations via an averaging procedure. (ii) Performing a correlation analysis: the dependence of acoustic backscatter classification on sediment physical properties is verified by observing a significant correlation of 0.75 (and a disattenuated correlation of 0.90) between the classification results and sediment mean grain size. (iii) Enhancing the statistical description of the backscatter intensities: angular evolution of the K-distribution shape parameter indicates that the riverbed is a rough surface, in agreement with the results of the core analysis.