Seabed Classification in Bahía de Navidad in the Mexican Central Pacific using Single-Beam Acoustics

doi:10.21203/rs.3.rs-4717239/v1

Download PDF

Case Report

Seabed Classification in Bahía de Navidad in the Mexican Central Pacific using Single-Beam Acoustics

https://doi.org/10.21203/rs.3.rs-4717239/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

This study proposed a methodology for characterizing subtidal habitats along rocky coastlines and adjacent seabeds in the epipelagic zone using a 120 kHz single-beam echo sounder (SBES). The data collected at depths of 15 to 80 m in Bahía de Navidad on the Mexican Central Pacific coast were postprocessed and classified using QTC Impact™ software to obtain three acoustic classes.

Using the classified data and ArcMap™ software, a bathymetric map was built for the selection of 19 sampling points for ground truthing and granulometric analysis to create a matrix further subjected to two ordination analyses: principal component analysis (PCA) and nonmetric multivariate scaling analysis (MDS). Additionally, the matrix was adapted into a dummy matrix, which was subsequently used to conduct a Pearson correlation coefficient test.

The analyses revealed a strong correlation between the medium-deep-zone acoustic class (20 m to 50 m) and the medium-grain-size sand substrate, as well as a high correlation between the deep-zone acoustic class (50 to 80 m) and finer sand substrates. This study demonstrated the potential of using affordable echosounders as a reliable and feasible alternative for characterizing marine habitats, especially for research institutions in developing countries with limited financial resources.

Single beam

acoustic seabed classification

hydroacoustic

ground truthing

granulometric analysis

Mexican Central Pacific

Information on the composition of seabeds is essential for numerous fields, including marine biology, marine geology, infrastructure planning and construction both on the coast and offshore, environmental defense, fisheries management and monitoring, the establishment of marine protected areas (MPAs), and the identification of specific areas for scientific research and conservation [1–2]. The seabed plays a crucial role in ecology as well, as it influences numerous processes that frequently impact the spatial distribution of species and serves as a foundation for interactions within ecosystems [3]. To enhance policies of fruitful environmental management, it is crucial to prioritize the identification of both physical and biological connections. This approach is valuable for designing further protection schemes; thus, there is an interest in seabed classification for the creation of maps of physical properties [4].

An important first step in the establishment of ecosystem-based management (EBM) planning is to map and characterize the biophysical conditions and human uses of the ocean [5]. The involvement of scientific methodologies that enable the creation of supportive knowledge is essential, as they guide scientists and managers to comprehensively grasp the allocation of both biotic and abiotic resources to the seabed and therefore to produce baseline information about habitats, facilitating the creation of complete seabed coverage maps that establish boundaries for separating adjoining habitats [6].

However, there is still no internationally standardized methodology to guide efforts for this purpose, so proposals continue to emerge. The management of anthropogenic impacts on marine ecosystems requires the identification and mapping of natural regions on a range of hierarchically nested spatial and temporal scales since this is how their processes operate [7]. Historically, studies on soft-bottom benthic communities and their habitats have used samples of sediments and organisms for analysis through the use of corers and dredgers, leading to alterations in the structural integrity of sedimentary habitats and significant ecological information being overlooked [8]. Recently, acoustic methods for seabed characterization, classification, and imaging have been developed, offering several advantages over direct sampling using cores or dredgers [9]. Information on the type of sediment that can be gathered from the combination of echo shape change data and ground truth data from different areas is a powerful tool that helps managers and stakeholders in the decision-making process on benthic habitats [10].

Acoustic technology relies on the travel time required for an acoustic signal to reach the seabed, to cross through the water column and return to the surface to determine bathymetric measurements; moreover, this signal strength can provide useful information about the constitution of the seabed and has become the most affordable method for surveying and mapping coastal and marine resources and a customary tool for mapping and classifying the seabed worldwide [11–12]. The features of the seabed surface include the reflection and dissipation of sound waves due to various elements and the diverse composition of layers, such as large rocky outcrops, fine substrates [13], rugosity, hardness, grain size and relief shapes, which all influence the reflection of the acoustic signal; thus, the acoustic backscattering process affects the information obtained from the surface according to the shape and intensity of the returned echo [14]. The shape of the echo relies on echosounder parameters, such as the ping duration, ping shape, frequency, beam coverage angle, and output power [15]. Hydroacoustic methods offer several benefits, including rapid data acquisition, effortless automation of the entire acquisition process, efficient processing and analysis, and real-time visualization of additional data [9]. The outcomes derived from acoustic classification systems (ACSs) depend closely on seabed features, such as relief and geological composition, and on the diverse array of marine organisms supported by these systems, ranging from fish and invertebrates to benthic species [16].

The most widely used instrument for hydroacoustic surveys is the SBES due to its affordability, easy availability and calibration, data processing efficiency (lower data volumes and standardized procedures), relative operation simplicity, processing speed of backscatter data, and availability of multiple frequencies during the same survey if needed [16]. While more complex acoustic ground discrimination systems (AGDS), such as side scan sonars (SSS) and multibeam enchosounders (MBES), are capable of swath 100% seabed coverage, they require a high investment in ship time and data processing compared to SBES. One notable feature of certain single SBES systems is their capacity to convert echo signals into data matrices, which can be useful in multivariate data analysis processes that include hypothesis testing and modeling [17].

In the mid-1990s, the Canadian company Quester Tangent Corporation (QTC™) introduced a commercial system based on a phenomenological approach consisting of calculating hundreds of parameters for each bottom echo. These parameters are then subjected to techniques such as multivariate statistical analysis (including PCA) to reduce the dimensionality of the parameter space. The outcome of this process is a set of descriptors, which are subsequently used in an automatic classification system that is ultimately presented with a visually appealing graphical representation in 3D space [9].

2.1. QTC approach for seabed classification

QTC developed the QTC View™ hydroacoustic system, which includes QTC Impact™ software for acoustic data processing. According to the user manual [18], its operating principle is based on the typical geometry of the sounding of an acoustic signal toward the seabed. A pulse (or ping) of sound is transmitted by an echo sounder, and the time it takes for the echo to return provides a measure of the echosounder altitude relative to the seabed. However, the echo shape and duration often differ substantially from those of the original ping; therefore, these distortions contain information about the type of seabed, its backscattering process and other geoacoustic properties [19]. QTC Impact™ averages the backscatter signal of five pings, and each group of five echoes is normalized to a maximum amplitude of one to eliminate artifacts due to depth and water column effects [20].

According to Legendre et al. (2002) [21], the QTC system normalizes the wave by setting the peak amplitude of the waveform to unity (to effectively eliminate the energy information) and describes it by 166 shape variables. The physical or mathematical meaning of the 166 variables has been kept secret by the manufacturer; however, it is known that some variables are based on echo shape and others on spectral features (e.g., spectral moments). The variables are transformed as follows: divided by a certain maximum value since all the values of the 166 variables are within the interval [0,1] to effectively remove the physical dimensions of the variables; these variables are adapted for statistical analysis without further standardization. The method for extracting the information is carried out by means of a PCA, which serves as a classification filter, discarding noise and retaining the most relevant information based on the first three main components; otherwise, the rest of the information is dismissed. According to the user manual [18], the three main components effectively describe each echo and mostly represent approximately 90% of the total variation in a dataset. QTC Impact™ allows users to plot data points on a three-dimensional graph based on these three main components, called variables Q1, Q2, and Q3. The classification results can be plotted by clusters represented by different colors. The software can perform the classification automatically or be directed by the user. However, the user manual is ambiguous, and user-driven methods are prone to errors at any stage. Finally, the results can be plotted on a graph whose axes are the geographical X and Y coordinates of the position of every echo of the survey, and the resulting classes are represented by the colors of the previously formed clusters.

3.1. Study Area

Bahía de Navidad is located on the Mexican Central Pacific coast, south of Jalisco state (Fig. 1), between the geographic coordinates of 19°14'00''N, 104°49'00''W (northwest) and 19°7'00''N, 104°38'00'' (southeast). The climate is mostly dry, subhumid and hot, with 23°C being the annual average temperature [22]. The rainy and dry seasons are well identified, with the rainy season occurring from June to October. The predominant natural vegetation types are deciduous forests, mangrove forests and tropical scrub [23]. The hydrological system includes rivers, coastal lagoons and estuaries [24]. The oceanic part of the bay constitutes an intricate system influenced by the warm North Equatorial Countercurrent (N.E.C.C.) and the cold California Current (C.C.). A mixture of these two bodies of water produces vast marine biodiversity. Located within the bay, the littoral zone of the town of Barra de Navidad boasts a diurnal-mixed tidal pattern featuring a range of 0.5 m [25].

According to Ríos-Jara et al. (2003) [24], the continental shelf of Jalisco and Colima is very narrow. With a width of only 7–10 km, it is confined to the 200 m isobath. The irregular bottom of the continental shelf slopes sharply in the direction of the Middle America Trench (MAT), which runs through the continent, where the bottom slopes can reach an angle of 8–10° [26]. The continental shelf in this region is composed mainly of rocky substrate, expanding in an area of approximately 5,315 km² [27]. Nevertheless, the bathymetry of a significant portion of this zone is scarcely known, as highlighted by Bandy et al. (1995) [28], due to the lack of high-resolution bathymetric data available. Despite these limitations, the available information suggests that the seafloor depths are consistent with those found in undisturbed oceanic crust, which is typically observed directly south of this area. Consequently, it is inferred that the seabed of this area is predominantly undisturbed, particularly on its western side.

3.2. Hydroacoustic survey.

The hydroacoustic survey was carried out onboard the shrimp fishing type research vessel BIP-V, a 12-meter long vessel owned by Universidad de Guadalajara. An aluminum strut was installed on the port side of the vessel, on which a 120 kHz split-beam piezoelectric transducer with a coverage angle of 7° was attached and connected to a SIMRAD EY60 portable echo sounder. The echosounder is capable of recording the positioning data of each returning echo from the seabed via a GPS connected to the echosounder computer. Table 1 lists the echosounder configuration parameters used for the survey. Between June 9 and 20, 2014, the hydroacoustic survey was carried out over a mesh-like trajectory of approximately 180 km, with an average speed of 6 knots (11.11 km/hr) (see Fig. 2).

3.3. Data processing (classification of echoes)

The raw data generated by the echosounder were subjected to postprocessing using QTC Impact™ software. According to the procedure given by the user manual [18], first, the raw data are uploaded, and individual traces of the echoes are examined to identify and deactivate poor-quality echoes. The software describes the echoes with full feature vector (FFV) variables through 166 variables, including the depth at which each echo was generated (or bottom picked). The navigation data of each echo are subsequently merged, and any defective FFV record is edited for correction before useful echoes are filtered and classified. For further data analysis, a catalog of the merged FFV files was created and subjected to PCA to produce different acoustic classes (three in this case), with a classified file with the extension selected as the output. Unlike other studies that have used a cluster performance index (CPI) method to classify and determine the optimal number of classes—e.g., Freitas et al., 2005 [29]—in this study, the automatic classification given by the software was chosen based on observations made by Legendre (2002) [30], who found the CPI method to be ambiguous and prone to errors.

QTC Impact™ software concludes the acoustic classification process by generating .dat data files that can be easily exported to Excel spreadsheets and consequently integrated into ArcMap™ for map creation. After obtaining a map with the echoes of the seabed recorded, geolocated and acoustically classified, it is possible to proceed with the selection of sediment sampling points.

3.4. Groundtruthing

To determine the correlation between the acoustically classified seabed echoes and the physical properties of the sediment, a total of 19 uniformly distributed sampling points were chosen (Fig. 3), the majority of which were located at intersections between transects. On July 21, 22 and 31, 2014, 19 sediment samples (and 2 replicas) were collected using a scissor clamshell grab (Van Veen type) with a sample surface area of 0.1 m² and a volume of 20 dm³.

Table 1

The parameters of the echosounder calibration were as follows:
Parameter		Setting
Material of the calibration sphere		Tungsten
Sphere diameter		300 mm
Pulse duration (range)		512 − 128 µs
Pulses per second		2
Power		200 kW
Sphere depth at calibration		8 m
Water temperature		29°C
Salinity		34 ppt
Location	Caleta de Cuastecomates

3.5. Grain size analysis

The laboratory procedure established by Bale and Kenny (2005) [31] was followed to measure the content of organic matter (OM) using the loss on ignition (LOI) method. The dry sample was then sieved into eight meshes of sizes: 2.8 mm, 2 mm, 1.4 mm, 1 mm, 710 µm, 500 µm, 355 µm, and 150 µm. To analyze the sediment samples for their grain size distributions, the GRADISTAT version 9.1 spreadsheet [32–33] was used. The program integrated within the spreadsheet is designed to calculate the phi value (φ) based on the most commonly described grain size distributions and how they deviate from a prescribed ideal distribution. Krumbein (1934) [34], to enhance the graphical presentation and statistical manipulation of grain size frequency data, later applied a logarithmic transformation of grade scale boundaries into phi (φ) values using the expression shown in Eq. 1.

φ = -log₂ d (Eq. 1)

where d is the millimeter diameter of the grain.

The technical communication given by Blott and Pye (2001) [33] provides a detailed description of the process followed by GRADISTAT, which is briefly summarized below. Sedimentologists conventionally utilize log-normal distributions, which employ logarithmically transformed scales. For this purpose, GRADISTAT is an Excel spreadsheet that has integrated a program written in Microsoft Visual Basic that provides tabular and graphical outputs. When inputting the percentage of sediment present in various size fractions, the spreadsheet calculates a range of sediment statistics, including the mean, mode(s), range (standard deviation), skewness, kurtosis, and geometric cumulative percentile values. By employing the method of moments, the program calculates these statistics based on different distributions for metric and phi size values. These calculations follow the formulae suggested by Krumbein and Pettijohn (1938) [35] and use linear interpolation to extract specified values from a cumulative percentage curve. These extracted values are subsequently used to calculate Folk and Ward parameters, employing logarithmic and geometric scales based on log-normal distributions with phi and metric size values, as originally suggested by Folk and Ward (1957) [36]. Finally, the Udden-Wentworth grading scale was used to describe the mean grain size.

3.6. Statistical analysis

After acoustic classification, sediment sample collection, grain size analysis and calculation of the fraction of organic matter expressed as a percentage (% O.M.), a data matrix consisting of 19 sampling points (rows) and 13 environmental sampling variables (columns) was constructed. The environmental variables measured in each sample collected were as follows: the acoustic class of each sample falls into which eight average portions of grain sizes are represented as percentages; the average value of phi (φ) resulting from the granulometric analysis carried out with the GRADISTAT spreadsheet; the average fraction of O.M. expressed as a percentage (% O.M.); the depth in meters recorded by the echosounder; and the approximate value of the seabed slope (M) calculated by means of the slope equation using the depth values of the two echoes within which each sampling point is located. The average value is the result of averaging three measurements (one sample and two replicates).

The resulting environmental data matrix (Table 2) was subjected to PCA and similarity analysis (SA) among the sampling points using the normalized Euclidean distance. The normalized Euclidean distance coefficient is commonly used in environmental data analyses due to its ability to integrate variables revealed in diverse units [37]. Next, a cluster analysis was conducted to test for the existence of an a priori structure in an unstructured sample set. This analysis was performed using the SIMPROF routine, which evaluates the statistical significance of clustering patterns. The variable Acoustic Class, which comprises discrete data labels A, B, and C, was included as a factor in the analysis. Subsequently, the sedimentary data were subjected to MDS analysis. All of the above analyses were carried out with PRIMER v6 statistical software [38].

In addition to the multidimensional analyses, to determine the correlation coefficient between the clases resulting from the acoustic classification and the variables grain size, φ, % O.M., depth, and M, a dummy matrix (Table 3) was constructed and subjected to a Pearson correlation coefficient test to finally generate a correlation matrix (Table 4). As Table 3 shows, the points (listed in rows) are ordered such that the first group (in red) corresponds only to points belonging to acoustic class A (P1, P8, P14, and P16); therefore, in this case, only the column corresponding to acoustic class A is assigned a value of 1, while the columns corresponding to the two remaining acoustic classes B and C are assigned a value of 0. The same procedure was followed for the sampling points located in the two remaining acoustic classes. The Pearson correlation coefficient test allowed us to determine the correlation between the acoustic classes and the different variables, thus providing a comprehensive understanding of the relationships between these elements.

4.1. Acoustic classification mapping

As a result of the acoustic classification of the echoes performed by QTC Impact™ software and the export of the data to ArcMap™, the map shown in Fig. 3 was created. This figure also shows the 19 sampling points selected and the records of the survey echoes classified into three acoustic classes (A in red, B in green, and C in blue).

4.2. Results of the grain size analysis

The matrix described in Table 2 was created as a result of grain size analysis, organic matter content measurements and phi calculations via the GRADISTAT version 9.1 spreadsheet [32–33] and other variables described above.

4.3. Statistical analysis results

After conducting a similarity analysis on the various sampling points, a cluster analysis was performed, which produced a tree plot (Fig. 4). The plot clearly depicts a discernible grouping trend that corresponds to the acoustic class. Specifically, P8 and P14 are grouped together in class A, while P2, P5, P9, P11, and P17 form a larger group in class C.

The MDS biplot shown in Fig. 5 confirms what can be observed from the dendrogram of the plot in Fig. 4 by revealing a distinguishable group formed by the sampling points P2, P3, P5, P9, P11, P12, P13, P17, P18, and P19, where as in the dendrogram plot, only P19 belongs to a different acoustic class than class C. Figure 5 also shows a minor group in the lower left consisting of P8 and P14 (both in class C) and a group in the lower right corresponding to P4, P7, and P15, all in class B. Points P12, P13, P18 and P19 form another group with only P19 belonging to an acoustic class other than C. Finally, points P4, P6, P7 and P15 form another group, of which only P6 does not meet the acoustic class (class B).

Figure 6 shows the biplot generated after conducting the corresponding PCA, which illustrates a strong correlation between the variables Very fine sand and Silt by depicting their vectors aligned with PC1 and points toward a group of sampling points (P2, P5, P9, P11, P12, P13, P17, P18, and P19) dominated by very fine sand substrates, with only P19 falling outside class C. The variable % O.M. also points in the same direction, although to a lesser extent. The trend in the biplot was congruent with the clusters shown in the dendrogram plot (Fig. 4) and the MDS biplot (Fig. 5). Therefore, the biplot offers insight into the correlations between variables and the formation of clusters, which are consistent with the other analyses. The PCA biplot (Fig. 6) also revealed important correlations between the variables Gravel, M (%) (or slope), and coarse sand since its vectors point in the negative PC2 direction.

Table 3 shows the same matrix as Table 2 but with the inclusion of three additional columns (labeled Acoustic classes A, B and C) representing dummy variables. The matrix shown in Table 4 was generated following the application of the Pearson correlation coefficient test. Additionally, Fig. 7 depicts a zoned map that is represented by three polygons in red, green, and blue, representing the different zones derived from the acoustic classification of the study area. Finally, Table 5 shows a synoptic summary of the key features of the analyzed variables for each of the three habitats and their respective acoustic classes.

The Euclidean distance matrix (Table 4) shows the highest values of the correlation (marked in yellow) between the medium sand sediment with a grain size of 355 µm and acoustic class B, as well as between the silty sediment with a grain size less than 63 µm and acoustic class C. On the other hand, acoustic class A does not correlate significantly with any other environmental variable considered in this analysis.

Table 2

Environmental data matrix resulting from the grain size analysis and acoustic seabed classification.
	Gravel	Very coarse sand	Coarse sand	Medium sand	Medium sand	Fine sand	Very fine sand	Silt		Organic matter		Slope		Gradistat
Site	> 2 mm	1 mm	500 µm	355 µm	250 µm	150 µm	63 µm	< 63 µm	φ	% O.M.	Depth (m)	M (%)	Acoustic class	Type of sediment
P1	34.55	41.34	21.00	1.68	0.44	0.28	0.47	0.24	-0.03	3.89	76.79	23.34	A	Very coarse sand
P2	2.46	4.27	9.76	7.27	7.70	18.81	37.65	12.08	2.46	5.30	63.52	1.14	C	Fine sand
P3	1.54	2.71	4.91	8.08	19.29	40.18	21.61	1.68	2.41	3.32	50.15	1.28	C	Fine sand
P4	5.00	11.71	38.44	21.57	12.94	8.97	1.35	0.02	0.89	1.89	23.40	1.34	B	Coarse sand
P5	0.78	1.78	8.04	5.35	5.06	16.32	48.54	14.12	3.03	5.41	67.70	0.81	C	Very fine sand
P6	4.44	16.29	29.27	17.90	19.46	11.62	0.96	0.07	0.92	2.51	38.45	1.62	C	Coarse sand
P7	0.44	1.30	10.26	35.18	30.58	18.92	3.25	0.07	1.58	1.34	30.08	0.65	B	Medium sand
P8	0.04	0.29	1.01	2.09	5.91	19.91	68.79	1.96	3.22	2.23	14.37	1.63	A	Very fine sand
P9	0.07	0.16	4.84	5.72	4.89	19.54	47.33	17.45	3.22	6.33	68.20	0.65	C	Very fine sand
P10	0.15	0.72	3.20	6.21	17.13	48.22	23.63	0.75	2.51	3.61	23.40	1.63	B	Fine sand
P11	0.26	0.09	3.00	6.16	5.80	21.54	47.58	15.57	3.21	4.12	61.85	1.30	C	Very fine sand
P12	0.14	0.34	5.34	4.00	4.18	13.09	56.56	16.35	3.25	5.46	50.15	1.42	C	Very fine sand
P13	0.07	0.11	0.49	2.18	3.41	23.69	55.07	14.99	3.28	5.91	45.13	2.55	C	Very fine sand
P14	0.31	0.48	1.52	3.34	10.22	19.04	55.77	9.32	3.18	3.05	11.70	0.46	A	Very fine sand
P15	10.35	13.32	40.34	26.19	7.24	1.73	0.72	0.11	0.64	2.04	28.41	0.32	B	Coarse sand
P16	0.02	0.15	4.86	11.94	21.99	40.76	19.73	0.55	2.39	0.90	10.70	1.22	A	Fine sand
P17	0.03	0.15	8.73	6.04	5.32	9.52	45.59	24.62	3.16	6.22	66.87	2.46	C	Very fine sand
P18	0.20	0.46	9.24	5.53	4.96	12.93	50.34	16.34	3.07	6.27	50.15	4.88	C	Very fine sand
P19	1.32	1.64	3.28	3.56	6.86	14.95	62.37	6.03	3.16	3.79	35.10	4.85	B	Very fine sand

Table 3. The dummy variable matrix was subjected to a Pearson correlation coefficient test.

Table 4

Matrix resulting from the Pearson correlation coefficient test among the environmental variables
	Acoustic Class A	Acoustic Class B	Acoustic Class C		Very coarse sand (1 mm)	Coarse sand (500 µm)	Medium sand (355 µm)	Medium sand (250 µm)	Fine sand (150 µm)	Very fine sand (63 µm)
	Acoustic Class A	Acoustic Class B	Acoustic Class C	Gravel	Very coarse sand (1 mm)	Coarse sand (500 µm)	Medium sand (355 µm)	Medium sand (250 µm)	Fine sand (150 µm)	Very fine sand (63 µm)	Silt (< 63 µm)	φ	% O.M.	Depth (m)	M (%)
Acoustic Class A	1
Acoustic Class B	-0.31	1
Acoustic Class C	-0.54	-0.63	1
Gravel	0.36	0.01	-0.31	1
Very coarse sand (1 mm)	0.29	0.04	-0.27	0.96	1
Coarse sand (500 µm)	-0.17	0.41	-0.23	0.48	0.62	1
Medium sand (355 µm)	-0.27	0.61	-0.31	-0.01	0.10	0.62	1
Medium sand (250 µm)	-0.04	0.37	-0.30	-0.27	-0.16	0.08	0.68	1
Fine sand (150 µm)	0.04	-0.02	-0.02	-0.49	-0.52	-0.57	-0.16	0.50	1
Very fine sand (63 µm)	0.05	-0.41	0.32	-0.52	-0.63	-0.75	-0.72	-0.54	0.11	1
Silt (< 63 µm)	-0.33	-0.50	0.71	-0.36	-0.44	-0.44	-0.49	-0.58	-0.17	0.67	1
φ	-0.10	-0.37	0.41	-0.75	-0.84	-0.84	-0.59	-0.27	0.39	0.92	0.69	1
% O.M.	-0.41	-0.46	0.74	-0.12	-0.20	-0.37	-0.61	-0.68	-0.16	0.57	0.87	0.53	1
Depth (m)	-0.37	-0.43	0.68	0.32	0.26	-0.06	-0.33	-0.52	-0.33	0.09	0.58	0.04	0.74	1
M (%)	0.40	-0.13	-0.21	0.90	0.82	0.14	-0.29	-0.36	-0.37	-0.24	-0.17	-0.47	0.09	0.40	1

Table 5

Synoptic table with the characterization of the three acoustically detected habitats
Zone	Description
Habitat A (in red)	Shallow zone (less than 20 m), does not have a substrate of a defined grain size, although gravelly substrates and very coarse sand (greater than 1 mm) predominate slightly with little concentration of organic matter and slight slope.
Habitat B (in green)	Medium depth zone (range of 20 to 50 m). The medium sand substrate (355 µm) predominates with little presence of silt, moderate values of organic matter and very little slope.
Habitat C (in blue)	It is the deepest zone (greater than 50 m) where the silty substrate predominates (grain size < 63 µm), with the highest concentration of organic matter and very little slope.

The literature on seabed classification shows that this subject has different experimental approaches and requires the involvement of interdisciplinary teams of biologists, ecologists, mathematicians, and physicists, among others. Currently, alternative methods for studying seabeds, ranging from exhaustive point-to-point sampling to highly sophisticated techniques such as photographic spectroscopy of satellite images and lidar, continue to be considered, depending on the budget at hand. As mentioned above, echo sounders, especially SBESs, are generally available at any institution dedicated to ocean studies, especially for biomass estimation studies in the water column, making them convenient choices for seabed characterization studies.

However, one drawback of the use of hydroacoustic equipment for seabed classification may be the availability of software, which, for various reasons (costs, continuous technological development of sounders), has not been consolidated as a long-lasting tool for this field of study. For the development of this study, the availability of QTC Impact™ software was not an exception. Its main difficulties were first its high cost, which limits the duration of use for a license of only three months; second, QTC™ (the manufacturing company) ceased developing and selling products for the hydroacoustic field; however, this study has proven that acoustic classification software for acoustic signals generated with SBES and MBES are still of great use and need.

To date, studies on the classification of coastal habitats in Mexico are rare. A few studies have used a variety of remote sensing approaches, such as echo sounding, satellite imagery and seismic profiling [39–41]; therefore, there is very little evidence that Mexican managers and decision makers are using these approaches as a baseline for the management of coastal resources. This study demonstrated the value of hydroacoustic technology as a reliable and affordable approach that should be more widely used by Mexican research institutions. By establishing a baseline for the classification of seabed habitats, this approach can provide reliable cartography information for future studies and decision-making related to fisheries management.

With respect to the statistical analyses performed in this study, one noteworthy observation was the agreement among groups formed by the three different statistical analysis methods used (cluster, MDS and PCA) and the acoustic classes generated by the QTC Impact™ software. This agreement was favored by several environmental variables, most notably, the prevalence of medium sand (with a grain size of 355 µm) and silt (with a grain size < 63 µm), which significantly influenced their acoustic classification as classes B and C, respectively (as shown in Table 4). Several other logical and expected concordances between environmental variables according to the Pearson correlation coefficient test (Table 4) include the coexistence of gravely substrates with very coarse sands located both on steeper seabeds and the presence of very fine sands with silt, which was in turn correlated with organic matter content and greater depth.

To strengthen the aforementioned conclusions, the resulting PCA biplot shows that the vectors of the variables Very fine sand and Silt are very close to each other and that both point toward the positive side of the PC1 axis and to the main group formed by points P2, P5, P9, P11, P12, P13, P17, P18 and P19, which were acoustically classified in class C (in blue) only, with the exception of P19. In the same direction but at a much smaller size, the vector corresponding to the variable % O.M. (or organic matter content) was used. Like in the Pearson correlation coefficient test, the PCA biplot also showed remarkable closeness between the variables Gravel, Very coarse sand and M (%) (or slope), all three pointing toward the negative part of the PC2 axis.

The habitat classified as acoustic class B is situated at medium depths, ranging from 20 to 50 meters, and is predominantly composed of medium grain size sand. In contrast, the silty habitat, which is characterized as acoustic class C, is found in the deepest part of the area, exceeding a depth of 50 meters. These observations match previous studies conducted by Freitas and collaborators, which have shown that sediment content changes gradually with increasing depth. According to Freitas (2005) [42], the surface sediment content gradually changes with increasing depth, and in all of their studies, this gradual change was detected by acoustic classification, identifying the predominant sediment types (fine sand with low silt content and clay content, very fine silty sand and mud). Lee et al. (2015) [43] also reported a high correlation between silty and muddy sediments and depth.

Recent literature shows a growing trend in using benthic biotope mapping and characterization as baselines for managing marine reserves and fishing areas, often designated MPAs or similar protection schemes [44–45]. However, developing countries continue to lag behind in the number of such studies due to limited technological resources and financial constraints. This study employs a remote sensing mapping method that is cost effective and easily accessible, making it a standout approach, particularly in developing countries where baseline information on marine coastal management is often scarce and unreliable [46]. In addition, the management of large marine coastal areas requires continuous and detailed monitoring via field studies, which are costly [47]. Consequently, tight budgets often present a significant obstacle to the use of baseline tools and methodologies in these countries, hindering progress in marine coastal management.

Remote sensing allows the synchronized study of vast marine extents, facilitates the evaluation of spatial patterns, and provides the required constant surveying to assess temporal patterns [48]. Acoustics is widely recognized as the most powerful remote sensing tool for accurately mapping and effectively monitoring extensive seabed areas [16], and despite its limited coverage, the SBES is the most frequently used instrumentation due to its affordability, accessibility, simple operation and simplified data processing [49]. The proposed methodology uses affordable equipment at any marine research center and is feasible for periodic and repetitive use in similar studies in other coastal areas at a reasonable cost, facilitating the generation of baseline information necessary for the implementation of management programs based on Integrated Coastal Zone Management (ICZM).

This study provides detailed mapping in an area lacking such data, and through acoustic classification and ground truthing, it was possible for the classification system used (echo sounder and software) to identify two main types of habitats: one corresponding to very fine sediments (very fine sands and silt) located at the deepest side and defined as acoustic class C, and the other corresponding to medium sands located at a medium depth that were classified as class B, as is clearly shown by the PCA biplot. However, no significant correlation was found between any of the environmental variables and acoustic class A. On the other hand, the Pearson correlation coefficient test revealed an important correlation between the concentration of organic matter and very fine sediments (especially silty sediments) located in the deepest part of the study area.

Social and technological development, as well as demographic growth of human communities, has caused greater and more efficient use of natural resources; consequently, the habitats (and the ecosystems that contain them) have suffered high degradation. To combat this, there is an urgent need to harness innovation and technology to generate baseline knowledge. Such knowledge would enable the formulation of more effective management strategies that prioritize not only the restriction of natural resource use but also sustainable resource exploitation. As a first step toward this end, comprehensive mapping and characterization of the abiotic elements within the study area were carried out as a first essential product in the definition of the structural environment level proposed by Guarinello et al., 2010 [50], who established that the characterization phase is the first step in any land use planning process (terrestrial, marine or mixed).

The establishment and enactment of Integrated Coastal Zone Management (ICZM) programs in developing countries is often challenging (when efforts and methodologies are needed), mainly due to limited financial resources [46]. In this regard, the methodology proposed in this study stands out because it utilizes low-cost and easily accessible technology. Acoustics is widely recognized as the most powerful remote sensing tool for accurately mapping and effectively monitoring extensive seabed areas [16], with the SBES being the most commonly used instrumentation due to its affordability, accessibility, simple operation and simplified data processing [49].

Funding: This work was funded by Consejo Nacional de Ciencia y Tecnología (Conacyt) and Universidad de Guadalajara.

Conflicts of interest/Competing interests: The authors declare that they have no competing interests.

Data availability (data transparency): The authors confirm that the data analyzed during this study are included in this published article.

Code availability. Not applicable.

Contributions of the authors: The authors confirm their contributions to the paper as follows. Study conception and design E.G.D., A.F.H. Data collection and methodology A.F.H., E.G.D. Analysis and interpretation of results E.G.D., A.F.H. Draft manuscript preparation A.F.H., E.G.D. Funding acquisition E.G.D.

All the authors reviewed the results, commented on previous versions of the manuscript and approved the final version.

Acknowledgments: The postprocessing of the data carried out in this work was achieved with the valuable contribution of Jim Galloway, former Leader of the Sonar Systems Group of the Canadian Hydrographic Service, who provided us with a temporary license to use acoustic classification software and provided our team with indications for the best use of the echosounder.

Lee, W. S., & Lin, C. Y. (2018). Mapping of tropical marine benthic habitat: Hydroacoustic classification of coral reefs environment using single-beam (RoxAnn™) system. Continental Shelf Research, 170, 1-10. https://doi.org/10.1016/j.csr.2018.09.012
Fauziyah; Purwiyanto, A. I. S., Agustriani, F., Putri, W. A. E., Liyani, M., Aryawati, R., Ningsih, E. N., & Suteja, Y. (2020). Detection of bottom substrate type using single-beam echo sounder backscatter: a case study in the east coastal of Banyuasin. In IOP Conference Series: Earth and Environmental Science (Vol. 404, No. 1, p. 012004). IOP Publishing. https://doi.org/10.1088/1755-1315/404/1/012004
Costa, P. L., Madureira, L. A., & Pinho, M. P. D. (2013). Seabed acoustic classification in the Pelotas basin, Brazil. Brazilian Journal of Oceanography, 61, 13-22. https://doi.org/10.1590/S1679-87592013000100002
Ferretti, R., Fumagalli, E., Caccia, M., & Bruzzone, G. (2015, May). Seabed classification using a single beam echosounder. In Oceans 2015-Genova (pp. 1-5). IEEE. DOI: 10.1109/OCEANS-Genova.2015.7271709
Crowder, L., & Norse, E. (2008). Essential ecological insights for marine ecosystem-based management and marine spatial planning. Marine policy, 32(5), 772-778. https://doi.org/10.1016/j.marpol.2008.03.012
Foster-Smith, R., Connor, D., Davies, J. (2007). What is habitat mapping? In: MESH Guide to Habitat Mapping, MESH Project, 2007, JNCC, Peterborough, 66pp. Available online at http://www.searchmesh.net
Azovsky, A. I., Chertoprood, M. V., Kucheruk, N. V., Rybnikov, P. V., & Sapozhnikov, F. V. (2000). Fractal properties of spatial distribution of intertidal benthic communities. Marine Biology, 136(3), 581-590.
Guigné, J. Y., Schwinghamer, P., & Liu, Q. (1993). High resolution and broadband processing of acoustic images of the marine benthos. PROCEEDINGS-INSTITUTE OF ACOUSTICS. Vol. 15, Part 2, 15, 237-237.
Łubniewski, Z., & Stepnowski, A. (2018, September). Seafloor Characterization Using Underwater Acoustic Devices. In 2018 Joint Conference-Acoustics (pp. 1-6). IEEE.
Lied, T. T., Walday, M., Olsgard, F., Ellingsen, K. E., & Holm, S. (2004, November). SEABEC-A single beam echo sounder seabed classification system. In Oceans' 04 MTS/IEEE Techno-Ocean'04 (IEEE Cat. No. 04CH37600) (Vol. 4, pp. 2024-2028). IEEE. https://doi.org/10.1109/OCEANS.2004.1406454
Bartholomä, A. (2006). Acoustic bottom detection and seabed classification in the German Bight, southern North Sea. Geo-Marine Letters, 26(3), 177-184.
Manik H. M. (2015). Acoustic characterization of fish and seabed using underwater acoustic technology in Seribu Island Indonesia. Journal Marine Science Resource & Development 5:157-165.
Urick, R.J. (1983). Principles of underwater sound. 3rd edition. McGraw-Hill Book Company. New York. 423 pp.
Demer, D. A., Cutter, G. R., Renfree, J. S., & Butler, J. L. (2009). A statistical-spectral method for echo classification. ICES Journal of Marine Science, 66(6), 1081-1090.
Biffard, B. R., Preston, J. M., & Chapman, N. R. (2007, September). Acoustic classification with single-beam echosounders: processing methods and theory for isolating effects of the seabed on echoes. In OCEANS 2007 (pp. 1-8). IEEE. https://doi.org/10.1109/OCEANS.2007.4449250
Anderson, J. T., Van Holliday, D., Kloser, R., Reid, D. G., & Simard, Y. (2008). Acoustic seabed classification: current practice and future directions. ICES Journal of Marine Science, 65(6), 1004-1011. https://doi.org/10.1093/icesjms/fsn061
Quintino, V., Freitas, R., Mamede, R., Ricardo, F., Rodrigues, A. M., Mota, J., ... & Marcos, C. (2010). Remote sensing of underwater vegetation using single-beam acoustics. ICES Journal of Marine Science, 67(3), 594-605. https://doi.org/10.1093/icesjms/fsp251
Quester Tangent Corporation. (2004). QTC IMPACT™ User Manual (v3.40), DMN-MIMP-0001-R03, Release May 2004, Revision 3D, Quester Tangent Corporation, Marine Technology Centre, Sidney, B.C., Canada, 147 pp.
Sternlicht, D. D., & de Moustier, C. P. (2003). Time-dependent seafloor acoustic backscatter (10–100 kHz). The journal of the acoustical society of America, 114(5), 2709-2725. https://doi.org/10.1121/1.1608018
Preston, J. M., Christney, A. C., Beran, L. S., & Collins, W. T. (2004, July). Statistical seabed segmentation—from images and echoes to objective clustering. In Proceedings of the Seventh European Conference on Underwater Acoustics, ECUA (pp. 5-8).
Legendre, P., Ellingsen, K. E., Bjørnbom, E., & Casgrain, P. (2002). Rapid Communication/Communication Rapide Acoustic seabed classification: improved statistical method. Canadian Journal of Fisheries and Aquatic Sciences, 59(7), 1085-1089.
Rodríguez-Cajiga S. (1993). Macrofauna de la Laguna Barra de Navidad, Jalisco. In Biodiversidad Marina y Costera de México, S. I. Salazar-Vallejo y N. E. González (eds.). CONABIO. 499-508
Holland, T. L., Mariscal Romero, J., Davidson-Arnott, R., & Cardille, J. (2011). Landscape changes in a coastal system undergoing tourism development: implications for Barra de Navidad Lagoon, Jalisco, Mexico. Investigaciones geográficas, (74), 7-18. de http://www.scielo.org.mx/scielo.php?script=sci_arttext&pid=S0188-46112011000100002&lng=es&tlng=en.
Ríos-Jara, Eduardo, Pérez-Peña, Martín, López-Uriarte, Ernesto, & Juárez-Carrillo, Eduardo. (2003). Moluscos escafópodos de la campaña Atlas V (plataforma continental de Jalisco y Colima, México) a bordo del B/O El Puma. Ciencias marinas, 29(1), 67-76. Recuperado en 18 de julio de 2021, de http://www.scielo.org.mx/scielo.php?script=sci_arttext&pid=S0185-38802003000100007&lng=es&tlng=es.
Meyer Willerer, A. O., González, B. V., & Barragán, M. P. (2006). Ciclo anual de variables hidrológicas en el estuario Barra de Navidad, México. Los recursos pesqueros y acuícolas de Jalisco, Colima y Michoacán, Centro Regional de Investigación Pesquera de Manzanillo, Instituto Nacional de la Pesca, México, 135-148.
Filonov, A. E., Tereshchenko, I. E., Monzón, C. O., González-Ruelas, M. E., & Godínez-Domínguez, E. (2000). Seasonal variability of the temperature and salinity fields in the coastal zone of the states of Jalisco and Colima, Mexico. Ciencias Marinas, 26(2), 303-321. https://doi.org/10.7773/cm.v26i2.577
Ruíz Durá, M. F. (1985). Recursos pesqueros de las costas de México. Ed. Limusa. D.F. México.
Bandy, W., Mortera‐Gutierrez, C., Urrutia‐Fucugauchi, J., & Hilde, T. W. C. (1995). The subducted Rivera‐Cocos Plate Boundary: Where is it, what is it, and what is its relationship to the Colima Rift?. Geophysical Research Letters, 22(22), 3075-3078. https://doi.org/10.1029/95GL03055
Freitas, R., Sampaio, L., Rodrigues, A. M., & Quintino, V. (2005). Sea-bottom classification across a shallow-water bar channel and near-shore shelf, using single-beam acoustics. Estuarine, Coastal and Shelf Science, 65(4), 625-632. ISSN 0272-7714, https://doi.org/10.1016/j.ecss.2005.07.011
Legendre, P. (2002). Acoustic seabed classification methodology: a user’s statistical comparison. Departement de Sciences Biologiques, Universite de Montreal.
Bale, A. J., & Kenny, A. J. (2005). Sediment analysis and seabed characterization. Methods for the study of marine benthos, 3, 43-86.
Pye, K. (2020). Kenneth Pye associated Ltd. Scientific research, consultancy and investigations. UK. http://www.kpal.co.uk/index.html
Blott, S. J., & Pye, K. (2001). GRADISTAT: a grain size distribution and statistics package for the analysis of unconsolidated sediments. Earth surface processes and Landforms, 26(11), 1237-1248. https://doi.org/10.1002/esp.261
Krumbein, W. C. (1934). Size frequency distributions of sediments. Journal of sedimentary Research, 4(2), 65-77. https://doi.org/10.1306/D4268EB9-2B26-11D7-8648000102C1865D
Krumbein, W. C., & Pettijohn, F. J. (1938). Manual of sedimentary petrography: New York, D. Appleton-Century-Crofts Co.
Folk, R.L & Ward, W.C. (1957). Brazos River bar: a study in the significance of grain size parameters. Journal of Sedimentary Petrology. 27, 3–26. https://doi.org/10.1306/74D70646-2B21-11D7-8648000102C1865D
Ludwig, J. A., and Reynolds, J. F. (1988). Statistical ecology. A Primer on methods and computing. (Vol. 1). Chapter 9. John Wiley & Sons, USA. 337 pp.
Clarke, K.R., Gorley, R.N. (2006). PRIMER v6: User Manual/Tutorial (Plymouth Routines in Multivariate Ecological Research). Primer-E Ltd., Plymouth, pp. 192.
Riegl, B. M., Halfar, J., Purkis, S. J., & Godinez-Orta, L. (2007). Sedimentary facies of the eastern Pacific's northernmost reef-like setting (Cabo Pulmo, Mexico). Marine Geology, 236(1-2), 61-77. https://doi.org/10.1016/j.margeo.2006.09.021
Zapata-Ramírez, P. A., Blanchon, P., Olioso, A., Hernandez-Nuñez, H., & Sobrino, J. A. (2013). Accuracy of IKONOS for mapping benthic coral-reef habitats: a case study from the Puerto Morelos Reef National Park, Mexico. International Journal of Remote Sensing, 34(9-10), 3671-3687. https://doi.org/10.1080/01431161.2012.716922
Cuevas, E., de los Ángeles Liceaga-Correa, M., Rincón-Sandoval, L. A., Mexicano-Cíntora, G., Arellano-Méndez, L., Euán-Ávila, J. I., ... & Mulsow, S. (2013). Morphological and sedimentological assessment of submarine dune fields on the coast of Yucatan, Mexico. Ciencias Marinas, 39(1), 83-99. http://dx.doi.org/10.7773/cm.v39i1.2152
Freitas, R. (2005). Benthic biotopes remote sensing using single beam acoustics. Doctoral dissertation, Universidade de Aveiro, Departamento de Biologia. 169 pp.
Lee, S. T., Kelly, M., Langlois, T. J., & Costello, M. J. (2015). Baseline seabed habitat and biotope mapping for a proposed marine reserve. PeerJ, 3, e1446. https://doi.org/10.7717/peerj.1446
Lawrence E, Hayes KR, Lucieer VL, Nichol SL, Dambacher JM, Hill NA, et al. (2015) Mapping Habitats and Developing Baselines in Offshore Marine Reserves with Little Prior Knowledge: A Critical Evaluation of a New Approach. PLoS ONE 10 (10): e0141051. doi:10.1371/journal.pone.0141051
Linklater, M., Ingleton, T. C., Kinsela, M. A., Morris, B. D., Allen, K. M., Sutherland, M. D., & Hanslow, D. J. (2019). Techniques for classifying seabed morphology and composition on a subtropical-temperate continental shelf. Geosciences, 9(3), 141. https://doi.org/10.3390/geosciences9030141
Knight, D., LeDrew, E., & Holden, H. (1997). Mapping submerged corals in Fiji from remote sensing and in situ measurements: applications for integrated coastal management. Ocean & Coastal Management, 34(2), 153161-159170.
Goodman, J. A., Purkis, S. J., & Phinn, S. R. (2013). Coral reef remote sensing. A guide for mapping, monitoring and management. 436p.
Hedley, J. D., Roelfsema, C. M., Chollett, I., Harborne, A. R., Heron, S. F., Weeks, S., ... & Mumby, P. J. (2016). Remote sensing of coral reefs for monitoring and management: a review. Remote Sensing, 8(2), 118. https://doi.org/10.3390/rs8020118
Anderson, J. T., Holliday, V., Kloser, R., Reid, D., and Simard, Y. (2007). Acoustic seabed classification of marine physical and biological landscapes. ICES Cooperative Research Report, 286. 198 pp.
Guarinello, M. L., Shumchenia, E. J., & King, J. W. (2010). Marine habitat classification for ecosystem-based management: a proposed hierarchical framework. Environmental management, 45(4), 793-806. https://doi.org/10.1007/s00267-010-9430-5

No competing interests reported.

Download PDF

Editorial decision: Revision requested
05 Aug, 2024
Editor assigned by journal
31 Jul, 2024
Submission checks completed at journal
29 Jul, 2024
First submitted to journal
10 Jul, 2024

You are reading this latest preprint version

Seabed Classification in Bahía de Navidad in the Mexican Central Pacific using Single-Beam Acoustics

Status:

Version 1

Abstract

Figures

Introduction

2.1. QTC approach for seabed classification

Materials and methods

3.1. Study Area

3.2. Hydroacoustic survey.

3.3. Data processing (classification of echoes)

3.4. Groundtruthing

3.5. Grain size analysis

3.6. Statistical analysis

Results

4.1. Acoustic classification mapping

4.2. Results of the grain size analysis

4.3. Statistical analysis results

Discussion

Conclusions

Declarations

References

Additional Declarations

Status:

Version 1