Remote sensing of underwater vegetation using single-beam acoustics

Abstract

Quintino, V., Freitas, R., Mamede, R., Ricardo, F., Rodrigues, A. M., Mota, J., Pérez-Ruzafa, Á., and Marcos, C. 2010. Remote sensing of underwater vegetation using single-beam acoustics. – ICES Journal of Marine Science, 67: 594–605.

A single-beam, acoustic, ground-discrimination system (QTC VIEW, Series V) was used to study the distribution of underwater macrophytes in a shallow-water coastal system, employing frequencies of 50 and 200 kHz. The study was conducted in Mar Menor, SE Spain, where the expansion of Caulerpa prolifera has contributed to the silting up of the superficial sediments. A direct relationship was identified between algal biomass and sediment-fines content. Acoustic information on sediment grain size and data on algal biomass were obtained in muddy and sandy sediments, including vegetated and non-vegetated seabed. Non-vegetated muddy areas were created by diving and handpicking the algae. The multivariate acoustic data were analysed under the null hypotheses that there were no acoustic differences between bare seabeds with contrasting superficial sediment types or among low, medium, and high algal-biomass areas, having in mind that grain size can act as a confounding factor. Both null hypotheses were rejected, and the results showed that 200 kHz was better than 50 kHz in distinguishing cover levels of algal biomass. The relationship between the 200-kHz acoustic data and algal biomass suggests utility in modelling the latter using the former.

acoustic methods, algal cover, remote sensing, sediment types

Introduction

Benthic macrophytes include both attached macroalgae (seaweeds) and seagrasses (rooted marine plants), and they can form dense canopies in shallow coastal areas. They are widely recognized as an important coastal benthic structuring element, providing a number of valuable ecosystem services, including habitat for a variety of organisms (Whitfield, 1989; Irlandi et al., 1999; Boström et al., 2006). Many studies have revealed the importance of macrophytes to other communities, such as invertebrate macrofauna (Bowden et al., 2001; Sfriso et al., 2001; Hovel and Lipcius, 2002; Rueda and Salas, 2003; Eklöf et al., 2005; Unsworth et al., 2007). The effects of macrophytes on other communities could be beneficial, by increasing ecological niches, providing shelter for juveniles, and aiding the retention of larvae (López De La Rosa et al., 2006), or detrimental, when associated with eutrophication. Sánchez-Moyano et al. (2001) reported the impoverishment of bivalve populations as a result of the development of anoxic conditions near the seabed caused by the development of dense algal cover, as well as an overall decrease in species diversity associated with subsequent organic enrichment. Moreover, the spreading of certain macrophyte species, as in Mar Menor, Southeast Spain, where Caulerpa prolifera has taken much of the seabed area previously occupied by Cymodocea nodosa, can alter the species composition of the fish and cause the stocks of some species to decrease (Pérez-Ruzafa et al., 1991, 2006).

Considering their worldwide distribution, vital ecosystem services, and their ability to structure the benthic invertebrate and fish communities, there is value in developing cost-effective methods to study the distribution of macrophytes at large spatial scales. Studies investigating the distribution of underwater vegetation usually involve the direct collection of samples using grabs or by diving to classify the species and to assess biomass (Pérez-Ruzafa et al., 1989, 2006, 2008). Although such methodology provides the most accurate representation, it is time-consuming and hence more suitable for small areas. That is why remote-sensing methods for the study of aquatic vegetation are becoming more common, mainly for mapping purposes. Depending on the apparatus used, these methods can be subdivided into optical and acoustic, and both require validation through ground-truthing.

Optical remote-sensing methods, such as aerial photography and satellite imagery, are generally used to map the distribution of macrophyte beds. Aerial photography can allow for the examination of extensive seagrass landscapes at a fine spatial resolution, but it can be expensive and have limited ability to discriminate species. Remote sensing using the latest generation of multi- or hyperspectral, fine spatial resolution (<5 m²) satellite or airborne platforms (e.g. CASI, IKONOS, or Quickbird), are cost-effective at mapping shallow-water marine biotopes including those dominated by seagrass (Dekker et al., 2005; Phinn et al., 2008). Fornes et al. (2006) showed that vast meadows of the seagrass Posidonia oceanica situated within shallow coastal areas of the Mediterranean could be distinguished successfully from non-vegetated seabed using IKONOS imagery. Other authors have considered underwater videography to be more time-efficient than satellite imagery, particularly in deep water out of the range of areal platforms, but that method is less practical for covering very large areas and turbid waters (Norris et al., 1997). In fact, all optical-based methods for habitat surveillance will be ineffective or substantially less useful in waters of greater depth with high optical-attenuation coefficients. This can be the case in many coastal areas inhabited by benthic macrophytes, so such methods are then at a disadvantage.

Acoustic remote-sensing methods are less affected by water turbidity and water depth, which makes them potentially useful in differentiating bare from vegetated seabed and mapping these important benthic biotopes. Sidescan sonar has been used for scientific purposes in the marine environment (Fish and Carr, 1990; Brown et al., 2002; Kenny et al., 2003), including distinguishing between bare sediment and vegetated seabed (Siljeström et al., 1996; Pasqualini et al., 1998; Piazzi et al., 2000; Leriche et al., 2006; Lefebvre et al., 2009). Further, multibeam echosounders have been applied in quantifying the spatial distribution of seagrass beds (Komatsu et al., 2003a, b) with the advantage, compared with other acoustic methods such as sidescan sonar, of mapping large continuous strips of the seabed as well as acquiring data on the topography. There are few examples in the literature of the use of acoustic ground-discrimination systems based on single-beam echosounders (AGDS) for studying underwater vegetation, and there is a tendency for wider use of swath systems. AGDS, such as that used here (QTC VIEW), have been used successfully in discriminating benthic habitats characterized by different superficial sediment types (Freitas et al., 2003a), although their utility in studying the distribution of coral biotopes (Moyer, et al., 2005; Riegl and Purkis, 2005) and aquatic vegetation (Riegl et al., 2005; Preston et al., 2006) has also been reported. Work conducted by Preston et al. (2006) in the Seto Inland Sea, Japan, revealed the ability of the acoustic system to distinguish bare seabed from the areas covered by two species of seaweed (Sargassum fulvellum and Ecklonia kurome). Riegl et al. (2005), conducting field studies and work in a controlled environment in Florida (USA), reported an acoustic distinction between seagrass, sparse algae, and dense algae. Those authors suggested that the sensitivity of the acoustic system to sediment type may hinder the classification of macrophyte cover, leading them to question whether acoustic methods can diagnose the presence of macrophytes. If a large proportion of the acoustic echo originates from the sediment, then the information left to diagnose the macrophytes may be of little use. In fact, no acoustic study of underwater vegetation has been able to demonstrate specifically whether the areas covered by vegetation and those with bare seabed have the same type of sediment. If that is not the case, then the superficial sediment may act as a confounding factor in the distinction of the acoustic signals between the vegetated and bare seabed. This issue is addressed herein. The field experiment was carried out in Mar Menor, a shallow-water coastal area in which the alga C. prolifera has increased its distribution in recent times and today covers most of the seabed (Pérez-Ruzafa et al., 1989, 1991, 2006). The work was done in areas of contrasting superficial sediment types and used two different echosounder frequencies (50 and 200 kHz). Two null hypotheses were tested: H₀1, no significant acoustic differences exist between bare sandy and muddy areas; H₀2, no significant acoustic differences exist among areas of low, medium, and high algal biomass.

Material and methods

Mar Menor is a hypersaline coastal lagoon located on the Mediterranean coast of SE Spain (37°42′00″N 00°47′00″W; Figure 1). It has a total area of ∼135 km² and mean and maximum water depths of ∼3.6 and 6.0 m, respectively. Spatial and temporal variations in oceanographic conditions, nutrient concentration, and macrophyte assemblages were studied by Pérez-Ruzafa et al. (2005, 2008). Water salinity ranges from 38 to 51 and temperature from 11°C to 30°C, depending on location and season. Muddy substrata dominate, mainly in the central area of the lagoon, but also on some protected shores. Muddy areas are covered by a dense meadow of the macroalga C. prolifera, the distribution of which expanded significantly in the early 1970s, following the enlargement and deepening of the channel El Estacio, the main lagoon inlet currently (Figure 1; Pérez-Ruzafa et al., 1991). Sandy areas are located in the shallow areas on the margins and in the small bays surrounding the islands, where sparse patches of the phanerogame C. nodosa grow.

Figure 1.

Mar Menor, showing the location of the study areas, Molino de La Calcetera (MC), and La Puntica (Lo Pagán; LP). As an example, the 200 kHz acoustic-survey lines are shown for LP and MC, area B. The squares and rectangles identify the site replicates for the acoustic-sampling area, the sediment grain size, and the algae biomass. Dashed squares and rectangles correspond to bare seabed.

The study was conducted on sandy and muddy substrata, located, respectively, at Molino de la Calcetera (MC) and along the shore close to La Puntica (Lo Pagán, LP; Figure 1), between 1.5 and 2.5 m deep. The acoustic survey conducted in MC included sites with and without C. prolifera. The superficial sediment in the MC area is mainly sandy, and the Caulerpa meadows are less dense than those located on the muddy seabed. Sites with and without algae were replicated in two areas, A and B. In the study area close to LP, the seabed is muddy, covered everywhere by the algae, which form dense meadows in some locations. It was intended to replicate the experimental work in both muddy and sandy areas, but this was not possible because no bare muddy areas could be found. Therefore, an area of non-vegetated seabed at LP had to be created artificially by handpicking the macroalgae. This was performed by divers operating in an area of ∼100 m² (5 × 20 m; Figure 1). All fieldwork, i.e. handpicking the algae, acoustic surveying with both frequencies, and ground-truth sampling for sediments and algal biomass, was carried out over a period of 10 d.

Acoustic data

The acoustic survey was run with the seabed-classification system QTC VIEW Series V connected to a dual-frequency echosounder, Hondex 7300II (50 and 200 kHz). The base settings of the echosounder were transmit power 600 W, pulse duration 265 µs, and 7 pings s⁻¹ for both frequencies. Beam width for 50 kHz was 28°, and for 200 kHz it was 10°. For both frequencies, the QTC VIEW used automatic gain control. Positional information was logged continuously via a Differential Global Position System with the acoustic data allowing post-processing in a Geographic Information System (GIS). Each survey frequency was recorded separately, in consecutive surveys. The whole acoustic survey spanned 4 d. Because of the small size of the survey vessel, the transducer was mounted at the side of the boat and as far as possible from the turbulence created by the engine. The survey speed did not exceed 2 knots, and each area was surveyed repeatedly (Figure 1), given the small size of the survey areas and to acquire as many echoes as possible while avoiding ping collision. The minimum water depth that could be sampled safely to avoid it was ∼1 m. Nevertheless, all sampling areas were located between 1.5 and 2.5 m deep.

Superficial sediments and algae

The areas covered by the acoustic survey were sampled for superficial-sediment grain size and C. prolifera biomass. At each area in MC (MCA and MCB), four sampling sites were established, two on non-vegetated seabed (MCA1, MCA2, MCB1, MCB2) and two on seabed covered by macroalgae (MCA3, MCA4, MCB3, MCB4), a total of eight sampling sites. At each site, four replicates (a–d; Figure 1) were taken for grain size and for algal biomass. At LP, three sampling sites were established (LP1, LP2, LP3), LP3 located between the other two and selected for removal of macroalgae. At each site, four replicate samples were taken for both descriptors, except at LP3, where sediment samples were taken before and immediately after handpicking the algae to test whether sediment disturbance would significantly alter grain size.

All samples of sediment and macroalgae were obtained by scuba divers immediately after the final acoustic survey with both frequencies. The algae (leafs and root systems) were handpicked from an area of 20 × 20 cm and placed in bags of 1-mm mesh. The sediment samples were collected with a corer or a shovel, depending on sediment compactness, and retained in plastic bags.

Analysis of sediment grain size and algal biomass

Sediment grain-size analysis was performed by wet- and dry-sieving, according to the methodology described by Quintino et al. (1989): (i) chemical destruction of organic matter with H₂O₂; (ii) measurement of total sediment dry weight (DW), followed by chemical dispersion with tetra-sodium pyrophosphate (30 g l⁻¹) and wet-sieving through a screen of 63 µm mesh; (iii) measurement of the second DW of the material left on the 63 µm screen; (iv) dry-sieving of the sand fraction, i.e. particles with diameter from 63 µm to 2 mm, and the gravel fraction, particles with diameter >2 mm, through a battery of sieves spaced at size intervals of 1 ϕ, where ϕ = −log₂ particle diameter, expressed in millimetres. The silt and clay fraction, fine particles, with diameter <63 µm, was expressed as a percentage of the total sediment DW.

For each sample, the quantity of sediment in each grain-size class was expressed as a percentage of the whole sediment DW. The results were used to calculate the median value, corresponding to the diameter that has half the grains finer and the other half coarser. Given that no detailed grain-size analysis was performed for the fines fraction, i.e. particles with diameter <63 µm, the median could not be calculated for the samples with >50% fines content. These sediment samples were classified as mud. Sands, sediments with <50% fines, were classified using the median, expressed in units of phi (ϕ), according to the Wentworth scale (Doeglas, 1968): very fine sand (median between 3–4 ϕ); fine sand (2–3 ϕ); medium sand (1–2 ϕ); coarse sand (0–1 ϕ). The final classification adopted the description “clean”, “silty”, or “very silty”, with fines ranging from 0 to 5%, from 5 to 25%, and from 25 to 50%, respectively, of total sediment DW (Quintino et al., 1989). For each sample, the sediment data matrix includes six grain-size classes (>1, 0.5–1.0, 0.25–0.50, 0.125–0.250, 0.063–0.125, and <0.063 mm).

Benthic macrophytes were separated from the sediment, cleaned, and dried in air. All the organisms attached to the macrophytes were also removed. Caulerpa prolifera was the only true benthic macroalga, that is, permanently attached to the sediment, encountered in abundance in the study area. After species-level identification, the standing-crop biomass of C. prolifera was measured as DW and ash-free DW (AFDW). DW determination was made by drying to constant weight (48 h) in an oven at 80°C, and AFDW was determined subsequently by burning in a furnace at 500°C for 5 h. Weight was calculated with a precision of ±0.01 g. All data analysis was performed using AFDW.

Analysis of acoustic data

QTC VIEW Series V records the pressure waves that arrive at the transducer (the echo) and generates data in full waveform (fwf file) and/or based on the amplitude of the full waveform, called the envelope (env file). The post-processing software QTC IMPACT™ loads both files. For our purpose, only the fwf files were used because they represent the whole echo, although no real difference should be expected from the analysis of either file.

The 50 and 200 kHz acoustic datasets were processed separately by QTC IMPACT which, using a series of algorithms, describes each echo as 166 variables (the Full Feature Vectors, FFV file). This file was quality-controlled inside IMPACT, and inspected for the descriptors depth and positioning of each record. Suspicious FFVs were excluded from further analysis. Analysis using QTC IMPACT generally proceeds by inputting the FFV file to principal component analysis (PCA), the output file of which only includes, for each echo, the scores on the three first axes. This output file is then submitted to classification analysis to obtain acoustic classes. The novelty introduced by this work in the acoustic-data analysis is the direct use of the FFV file. To access this file, it was saved in ASCII format and written out using a Microsoft^® Windows^® net-software application converter (v1.0.8.45), created using a Microsoft^® Visual^® c# 2005 developing tool. Because of the very large number of records in the FFV ASCII file, special attention was given to the processing algorithm, to minimize CPU processing time and memory usage. This application converts the file to a CSV format that will open in Microsoft^® Office Excel^®. The excel file can be saved as a dbf (database file) and exported into a GIS environment, or other data-analysis software. The application converter developed for this study is available on request from the authors.

Within the GIS environment, the position of the acoustic-survey lines, the grain size, and the biomass samples were visualized, and the acoustic data were assigned to ground-truth samples, by selecting an area around them, named the acoustic-sampling area (Figure 1). The data matrix representing each of these included between 40 and 50 records on average. For each ground-truth sample, the final acoustic-data matrix corresponded to the mean value of those records for each of the individual 166 variables. These data were used in the analysis described below.

Experimental design

The grain size and the acoustic-data matrices (50 and 200 kHz) were submitted to hypothesis testing using PERMANOVA (permutational multivariate analysis of variance; Anderson et al., 2008) from the software PRIMER v6 (Clarke and Gorley, 2006), following the calculation of Euclidean distance matrices among samples. The pseudo-F values in the PERMANOVA main tests and the t-statistic in the pairwise comparisons were evaluated in terms of significance among different groups for the factors tested. A factor is defined as a categorical variable that identifies the groups we wish to compare. Values of p ≤ 0.05 revealed that groups differed significantly. Analysis of the acoustic data also included sediment grain size and algal biomass as covariates. For grain size, the site scores of the first axis from a PCA of the grain-size data were selected as the covariate.

The data were also submitted to non-metric, multidimensional scaling (NMDS), which was used as an unconstrained ordination method to visualize multivariate patterns from each dataset. NMDS diagrams are accompanied by a stress value that quantifies the mismatch between the distances among datapoints in the Euclidean distance matrix and in the ordination diagram. Values <0.10 are considered to represent the original distance matrix accurately (Clarke and Warwick, 2001).

Two null hypotheses were tested:

• H₀1: No significant acoustic differences exist between bare sandy- and muddy-seabed areas. Bare sandy-seabed areas were located in MC, and bare muddy-seabed areas were artificially created by diving and handpicking the algae in LP. At the non-vegetated site created, LP3, four sediment replicates were collected before and four immediately after removing the algae to test the null hypothesis that no significant differences in sediment grain size resulted from the activity. To test H₀1 with the grain-size data and the 200 kHz data, a three-factor nested model was used, with sediment type as the fixed factor with two levels, such as mud and sand, areas as a random factor and nested in sediment (LP for mud, MCA and MCB for sand), and sites, also as a random factor and nested in areas and in sediment (LP3 for LP and mud; MCA1 and MCA2 for MCA and sand; MCB1 and MCB2 for MCB and sand). The model is asymmetric because neither replicate sites nor areas could be obtained for the mud-sediment type. Each site was represented by four replicates. For the 50-kHz data, a simplified version of the model was used, with two factors, because no data could be obtained successfully in area MCA. The acoustic-data analysis was completed using sediment grain size as covariate.

• H₀2: No significant acoustic differences exist among low, medium, and high algal-biomass areas. With the 200 kHz data, this null hypothesis was tested using a two-factor nested model, using algal biomass/sediment-fines content as a fixed factor, and sites as a random factor and nested in the three levels, with two sites per level and four replicates per site: sites MCB3 and MCB4 for the level low; sites MCA3 and MCA4 for medium; sites LP1 and LP2 for the level high biomass/fines content. For the 50 kHz data, a simplification of this model was used, given that no data with that survey frequency could be available from the medium biomass/fines content level. The acoustic data were also analysed with covariates: algal biomass, sediment grain size, and the site scores on axis 1 of a PCA of the algal biomass/sediment-fines content data matrix.

Results

Sediment grain size and algal biomass

Sediment grain-size descriptors and algal-biomass mean values for each sampling site are listed in Table 1. All sites studied at LP had muddy sediment with >80% fines content. At MC, the superficial sediment was sandy with variable fines content, and the sites with C. prolifera always had finer sediments with higher silt and clay content than the sites without vegetation, corresponding to clean medium sand, in both areas (A and B; Table 1). Figure 2 shows the sediment-fine-particles content and the biomass of the algae. The relationship is direct, and although the dispersion is greater with increasing values of the descriptors, there was a positive significant Spearman correlation (ρ = 0.72; p < 0.01). The samples obtained in the three areas with C. prolifera, LP, MCA, and MCB, established a gradient corresponding, respectively, to high, medium, and low algal biomass/sediment-fines content. These levels were considered subsequently to test whether the respective acoustic signatures would differ significantly.

Experimental design

• H₀1: No significant acoustic differences exist between bare sandy- and muddy-seabed areas

Figure 2.

The relationship between sediment-fines content (percentage of total sediment, DW) and algal biomass (AFDW).

Table 1.

Sediment grain size, expressed as a percentage of total sediment DW, median value, in units of phi (ϕ), algal biomass, expressed in grammes of AFDW (AFDW, g), and sediment classification.

	Grain-size class (mm)
Sites	1.000	0.500	0.250	0.125	0.063	<0.063	Median (ϕ)	Biomass (AFDW, g)	Sediment classification
MCA1 bare	2.08	14.58	48.47	33.79	0.89	0.29	1.70	0.00	Clean medium sand
MCA2 bare	4.26	13.03	45.17	34.68	1.84	1.02	1.76	0.00	Clean medium sand
MCA3 vegetated	0.87	0.63	1.45	13.98	38.53	44.32	3.86	8.435	Very silty, very fine sand
MCA4 vegetated	0.67	0.73	15.15	19.41	21.13	42.82	3.56	8.755	Very silty, very fine sand
MCB1 bare	1.21	13.92	38.20	45.53	0.96	0.22	1.89	0.00	Clean medium sand
MCB2 bare	0.87	8.45	43.84	45.73	0.76	0.47	1.93	0.00	Clean medium sand
MCB3 vegetated	2.47	4.55	21.86	63.35	3.69	4.08	2.33	6.355	Clean fine sand
MCB4 vegetated	2.62	3.96	20.20	62.63	6.07	4.53	2.37	6.238	Clean fine sand
LP1 vegetated	2.54	3.27	4.32	5.63	3.00	80.71	>4.00	21.588	Mud
LP2 vegetated	0.93	1.05	1.61	2.74	2.20	91.23	>4.00	14.178	Mud
LP3 vegetated	1.56	1.98	3.47	3.27	1.63	86.87	>4.00	–	Mud
LP3a bare	1.33	1.95	3.72	4.37	1.92	86.00	>4.00	0.00	Mud

The values per site correspond to the mean of four replicates.

At the LP study area where the non-vegetated site was created, sediment samples were compared before and after removal of the algae. The results indicated that the associated sediment disturbance did not alter grain size significantly (pseudo-F = 0.134; p = 0.90).

The main test results for H₀1 are listed in Table 2, and they allow similar conclusions for the grain-size data and for both two survey frequencies. There were no significant differences between sites nested in areas and in sediment types, or between areas nested in sediment types, but significant differences between sediment types. Using the sediment grain size or the sediment-fines content as covariates, the difference between sediment types was no longer significant for both survey frequencies. These results indicate that both 50 and 200 kHz frequencies were sensitive to the structural properties of the sediment and clearly distinguished muddy from sandy sediment. An ordination analysis of the sediment and acoustic data is presented in Figure 3. There is clear separation between sandy and muddy seabeds in terms of grain size and acoustics, as shown by the PERMANOVA results (Table 2).

• H₀2: No significant acoustic differences exist among low, medium, and high algal-biomass areas

Figure 3.

Ordination diagrams (NMDS) of the sedimentary and acoustic data (50 kHz and 200 kHz), obtained in the non-vegetated seabed areas in LP and MC, area A (MCA) and area B (MCB).

Table 2.

PERMANOVA table of results for the test of the null hypothesis that no acoustic (50 and 200 kHz) significant differences exist between non-vegetated seabed characterized by different sediment types.

Source	d.f.	SS	MS	Pseudo-F	p-value
Main test for sediment grain size
Se	1	32 969	32 969.00	72.69	0.005
Ar(Se)	1	701.08	701.08	8.51	0.059 (n.s.)
Si[Ar(Se)]	2	164.68	82.34	0.11	0.964 (n.s.)
Res	15	11 505	766.99
Total	19	45 340
Main test for sediment-fines content
Se	1	23 390	23 390	50 119	0.0001
Ar(Se)	1	0.39	0.39	0.658	0.509 (n.s.)
Si[Ar(Se)]	2	1.17	0.59	0.0398	0.963 (n.s.)
Res	15	221.28	14.75
Total	19	23 613
Main test for 200 kHz
Se	1	2 259.60	2 259.60	69.17	0.002
Ar(Se)	1	46.39	46.39	3.84	0.097 (n.s.)
Si[Ar(Se)]	2	24.16	12.08	0.25	0.918 (n.s.)
Res	15	728.86	48.59
Total	19	3 059
Main test for 50 kHz
Se	1	1 485.50	1 485.50	23.72	0.020
Si(Se)	1	62.62	62.62	2.30	0.087 (n.s.)
Res	9	244.90	27.21
Total	11	1 793
200 kHz with sediment grain size as covariate
Cov	1	2 219.70	2 219.70	63.14	0.0004
Se	1	65.08	65.08	1.33	0.250 (n.s.)
Ar(Se)	1	49.22	49.22	3.80	0.085 (n.s.)
Si[Ar(Se)]	2	23.84	11.92	0.24	0.928 (n.s.)
Res	14	701.18	50.08
Total	19	3 059
200 kHz with sediment-fines content as covariate
Cov	1	2 230.8	2 230.8	67.955	0.0001
Se	1	58.10	58.10	1.171	0.290 (n.s.)
Ar(Se)	1	46.695	46.695	3.981	0.093 (n.s.)
Si[Ar(Se)]	2	23.46	11.73	0.235	0.932 (n.s.)
Res	14	699.98	49.998
Total	19	3 059
50 kHz with sediment grain size as covariate
Cov	1	1 464.60	1 464.60	23.94	0.001
Se	1	28.77	28.77	0.94	0.441 (n.s.)
Si(Se)	1	61.32	61.32	2.06	0.116 (n.s.)
Res	8	238.30	29.79
Total	11	1 793
50 kHz with sediment-fines content as covariate
Cov	1	1 471.20	1 471.20	23.593	0.002
Se	1	22.20	22.20	0.740	0.558 (n.s.)
Si(Se)	1	62.71	62.71	2.117	0.111 (n.s.)
Res	8	236.92	29.62
Total	11	1 793

d.f., degrees of freedom; SS, sum of squares; MS, mean square; Se, sediment type; Ar(Se), areas nested in sediment type; Si(Se), sites nested in sediment type; Si[Ar(Se)], sites nested in areas and sediment types; Res, residuals; Cov, covariate; n.s., non-significant.

These three algal-biomass levels are represented, respectively, by the sites sampled in the areas MCB, MCA, and LP. The main test results and pairwise comparisons between levels are listed in Table 3. For all datasets tested, the main factor was strongly significant, and the pairwise comparisons showed that all three levels were significantly different from each other. Given the rejection of the null hypothesis H₀1, these sediment differences may act as a confounding factor in the test of null hypothesis H₀2. The results for the 200 kHz survey, including all three levels of the main factor, are shown in Table 4, and they demonstrate that with this acoustic frequency, the main factor is significant in rejecting the null hypothesis H₀2. Table 4 also shows its significance when several potentially confounding factors are introduced in the PERMANOVA model as covariates. Neither grain size nor fines content were able to eliminate the significance of the main factor. On the contrary, algal biomass alone and algal biomass together with the sediment-fines content, represented by the site scores on axis 1 of a PCA analysis for these two variables, eliminated the significance of the main factor (Table 4). Compared with the results obtained in the analysis of H₀1, the 200-kHz survey frequency is clearly sensitive to the presence of the macroalgae.

Table 3.

PERMANOVA table of results for the test of the null hypothesis that no significant differences exist among low, medium, and high algal biomass/sediment-fines content areas: main tests and pairwise comparisons between levels.

Main test						Pairwise comparisons
Source	d.f.	SS	MS	Pseudo-F	p-value	Source	t	p-value
Algal biomass and sediment-fines content
Ar	2	37.23	18.62	18.44	0.004	High vs. medium	3.12	0.044
Si(Ar)	3	3.03	1.01	3.17	0.032	High vs. low	4.85	0.015
Res	18	5.74	0.32			Medium vs. low	33.36	0.0001
Total	23	46.00
Sediment grain size
Ar	2	47 134	23 567	53.04	0.0002	High vs. medium	4.07	0.016
Si(Ar)	3	1 333	444.33	1.99	0.086 (n.s.)	High vs. low	17.04	0.001
Res	18	4 009.1	222.73			Medium vs. low	5.81	0.004
Total	23	52 476
Sediment-fines content
Ar	2	26 686	13 343	176.88	0.0007	High vs. medium	7.98	0.014
Si(Ar)	3	226.3	75.44	0.96	0.437 (n.s.)	High vs. low	15.51	0.003
Res	18	1 421.7	78.98			Medium vs. low	50.23	0.0004
Total	23	28 334

d.f., degrees of freedom; SS, sum of squares; MS, mean square; Ar, areas with low, medium, and high algal biomass/sediment-fines content; Si(Ar), sites nested in areas; Res, residuals; n.s., non-significant.

Table 4.

PERMANOVA table of results for the test of the null hypothesis that no 200 kHz acoustic differences exist among low, medium, and high algal biomass/sediment-fines content areas.

Source	d.f.	SS	MS	Pseudo-F	p-value
Main test
Ar	2	2 740.30	1 370.20	8.61	0.014
Si(Ar)	3	477.72	159.24	5.64	0.0001
Res	18	507.95	28.22
Total	23	3 726
Sediment grain size as covariate
Cov	1	1 676.80	1 676.80	13.52	0.0001
Ar	2	1 206.90	603.45	6.89	0.0124
Si(Ar)	3	364.46	121.49	4.32	0.0006
Res	17	477.80	28.11
Total	23	3 726
Sediment-fines content as covariate
Cov	1	1 844.9	1 844.9	15.66	0.0001
Ar	2	1 047.9	523.97	6.22	0.0115
Si(Ar)	3	354.63	118.21	4.20	0.0010
Res	17	478.54	28.15
Total	23	3 726
Algal biomass as covariate
Cov	1	2 478.70	2 478.70	27.69	0.0001
Ar	2	484.91	242.45	2.73	0.0708 (n.s.)
Si(Ar)	3	285.10	95.04	3.39	0.0032
Res	17	477.30	28.08
Total	23	3 726
Algal biomass and sediment-fines content as covariate
Cov	1	2 485	2 485	18.78	0.0001
Ar	2	365.16	182.58	1.82	0.186 (n.s.)
Si(Ar)	3	394.78	131.59	4.65	0.0012
Res	17	481.09	28.30
Total	23	3 726

d.f., degrees of freedom; SS, sum of squares; MS, mean square; Ar, areas with low, medium, and high algal biomass/sediment-fines content; Si(Ar), sites nested in areas; Res, residuals; Cov, covariate; n.s., non-significant.

Table 5 lists the PERMANOVA results obtained when using just the low and the high levels of the main factor for both acoustic frequencies. Sediment descriptors, i.e. grain size or fines content, were insufficient to eliminate the significance of the main factor, although with a borderline p-value of 0.05 for the 50-kHz data (Table 5). This suggests that 50 kHz is more sensitive to the structural properties of the sediment than 200 kHz. Algal biomasses eliminated the significance of the main factor for the 200-kHz survey, but not the 50-kHz survey (Table 5). The results indicate that 200 kHz is more sensitive to the presence and biomass of the macrophyte than 50 kHz. This is confirmed when introducing as covariate the site scores on axis 1 of a PCA of algal biomass together with fines content: the main factor is no longer significant for any of the survey frequencies, but the p-value associated with the pseudo-F statistic is much higher for 200 kHz (Table 5). An ordination analysis of the sedimentary and acoustic data (50 and 200 kHz) representing low, medium, and high levels of algal biomass/sediment-fines content is displayed in Figure 4. There is clear separation between the three levels, in terms of grain size and acoustics, illustrating the PERMANOVA results just presented (Tables 3–5).

Figure 4.

Ordination diagrams (NMDS) of the sedimentary and acoustic data (200 and 50 kHz) representing the low, medium, and high algal biomass/sediment-fines content levels for the bottom areas vegetated with C. prolifera. The three levels correspond to study areas: low, MC area B; medium, MC area A; high, LP.

Table 5.

PERMANOVA table of results for the test of the null hypothesis that no acoustic (50 and 200 kHz) differences exist between low and high algal biomass/sediment-fines content areas.

	50 kHz					200 kHz
Source	d.f.	SS	MS	Pseudo-F	p-value	d.f.	SS	MS	Pseudo-F	p-value
Main test
Ar	1	1 836.30	1 836.30	15.22	0.008	1	1 697.40	1 697.40	9.09	0.037
Si(Ar)	2	241.37	120.68	3.94	0.002	2	373.47	186.74	6.24	0.001
Res	12	367.33	30.61			12	359.15	29.93
Total	15	2 445				15	2 430
Sediment grain size as covariate
Cov	1	1 752.20	1 752.20	16.16	0.0002	1	1 525	1 525	16.31	0.0004
Ar	1	170.22	170.22	3.09	0.046	1	457.79	457.79	9.66	0.005
Si(Ar)	2	184.64	92.32	3.01	0.007	2	159.17	79.59	3.04	0.016
Res	11	337.94	30.72			11	288.02	26.18
Total	15	2 445				15	2 430
Algal biomass as covariate
Cov	1	1 490.90	1 490.90	10.77	0.0004	1	1 754.10	1 754.10	15.92	0.0003
Ar	1	418.25	418.25	3.86	0.053	1	166.43	166.43	1.90	0.1977 (n.s.)
Si(Ar)	2	244.98	122.49	4.63	0.001	2	195.02	97.51	3.41	0.0119
Res	11	290.84	26.44			11	314.46	28.59
Total	15	2 445				15	2 430
Algal biomass and sediment-fines content as covariate
Cov	1	1 783.30	1 783.30	12.11	0.0003	1	1 785.30	1 785.30	10.75	0.002
Ar	1	113.50	113.50	1.59	0.2192 (n.s.)	1	40.23	40.23	0.50	0.697 (n.s.)
Si(Ar)	2	259.25	129.63	4.94	0.0007	2	292.40	146.20	5.15	0.004
Res	11	288.90	26.26			11	312.03	28.37
Total	15	2 445				15	2 430

d.f., degrees of freedom; SS, sum of squares; MS, mean square; Ar, areas with low and high algal biomass/sediment-fines content; Si(Ar), sites nested in areas; Res, residuals; Cov, covariate; n.s., non-significant.

In Figure 5, the relationship is shown between the covariates used in the model and the site scores on axis 1 of a PCA of the acoustic data: 50 kHz with two levels of the algal biomass/sediment-fines content factor, and 200 kHz with two and three levels. All graphs are accompanied by their appropriate correlation coefficients. For both survey frequencies, the strongest relationship is with the site scores on axis 1 of a PCA of the algal biomass and sediment-fines content, demonstrating that both frequencies are sensitive and related to a variable combining the two descriptors. The relationship between each acoustic frequency and the individual variables is stronger with algal biomass for 200 kHz and with sediment-fines content for 50 kHz. This indicates that the 200 kHz frequency is potentially more appropriate for use in a survey dedicated to studying macrophytes, whereas 50 kHz would be more appropriate for a survey dedicated to studying superficial sediments. The strong correlation between the 200 kHz PCA1 scores and algal biomass provides good perspectives for future modelling aimed at predicting the biomass of C. prolifera from acoustic data.

Figure 5.

Relationship between the site scores on axis 1 of a PCA of acoustic data (50 and 200 kHz) and the covariates (sediment grain size, sediment-fines content, algal biomass, and algal biomass/fines content). r is the correlation coefficient.

Discussion

The work was conducted in Mar Menor, SE Spain, a shallow-water system with most of the seabed covered by the macroalgae C. prolifera (Pérez-Ruzafa et al., 1991, 2008). Sampling took place at a seabed depth of 1.5–2.5 m and included areas characterized by contrasting superficial sediments (mud and clean sand), with and without C. prolifera, and a range of algal biomass. Caulerpa prolifera has been described as able to alter the superficial sediment by trapping fine particles in the root system (Pérez-Ruzafa et al., 1991). Our data and diving observations confirm this notion. In the MC area only, where the algae are patchy, a thin, superficial layer of finer sediment was observed beneath the algal patches. This was in contrast to the bare seabed found where no such accumulation of finer sediment was observed on the surface. As the distribution of the algae became less patchy, the seabed was covered homogenously by a muddy layer beneath the algal cover. This layer was quite thick in the LP sites, where algal biomass per sampling area was greater. In fact, a direct relationship between algal biomass and sediment-fines content was demonstrated. It is therefore possible that the spatial spread of the algae could cause a progressive silting of the superficial sediment in Mar Menor, as suggested by Pérez-Ruzafa et al. (1991). Because the present work has also shown that the single-beam AGDS used in this work was effective at distinguishing sandy from muddy seabed, the grain-size alteration resulting from the silting of the superficial sediment could then be successfully used as a surrogate descriptor for the study of the spatial extent of C. prolifera. This is especially the case if a wider survey confirms that no muddy areas exist in Mar Menor without C. prolifera. Therefore, the seabed-coverage limitations of single-beam acoustic systems compared with swath systems can be compensated for by habitat predictive modelling of the data collected by them. Single-beam AGDS can be an effective alternative approach to remote-sensing acoustic surveys using multibeam systems, especially if the survey area is sufficiently homogeneous to preclude the need to cover a greater proportion of the bottom. In areas of high bottom heterogeneity, swath systems would be more cost-effective because single-beam AGDS is not designed to produce extensive spatial coverage of the seafloor.

Several studies have shown the ability of the single-beam AGDS QTC VIEW Series V used in this work to distinguish different sediment types (Hamilton et al., 1999; Freitas et al., 2003b; Hewit et al., 2004; Wienberg and Bartholomä, 2005). In each of these cases, the QTC VIEW's own data-analysis workflow was used. The work reported here proposes and uses a novel approach to analysing the data recorded by the QTC VIEW system, by working directly with the acoustic data matrix in which the echoes are described by 166 variables, the FFV file. The results obtained here concerning the ability of the system to distinguish sediments characterized by different grain size are similar to those of previous studies that used more conventional data analysis involving the identification and interpretation of acoustic classes. Moreover, although the QTC VIEW Series V can sample adequately in water as shallow as 1 m, studies from shallow-water systems are relatively uncommon (Hutin et al., 2005; Moyer et al., 2005; Riegl and Purkis, 2005). Our work has shown the ability of both survey frequencies, 50 and 200 kHz, to distinguish sediment types in very shallow depth, as indicated by the significant difference in the acoustic data acquired from bare mud and medium sand. This conclusion agrees partially with the results of previous studies conducted with the same acoustic system. Freitas et al. (2008) showed for the inner Bay of Cádiz, also a shallow-water system, a close relationship between the 50-kHz acoustic diversity and the superficial-sediment-type pattern, whereas the acoustic classes based in the 200 kHz survey could not be related to sediment types. Riegl and Purkis (2005), also working with both frequencies in a shallow coral-reef area in the Arabian Gulf, demonstrated that the 50-kHz acoustic-seafloor classification was able to distinguish between unconsolidated sand and hard seabed, whereas the 200-kHz survey only determined high rugosity (corals and ripples) and low rugosity (flat areas).

The experimental set-up here to test the ability of the two survey frequencies to distinguish a range of algal biomass cover levels can shed some light on this issue. Part of the relevance of the present study is that there are very few examples in the literature discussing the use of single-beam AGDS, and particularly QTC VIEW series V, for the study of underwater vegetation (Riegl et al., 2005; Preston et al., 2006). The results we obtained indicated that both frequencies were able to distinguish areas characterized by different algal biomass/sediment-fines content, but that the 200-kHz frequency was more sensitive to the presence and biomass of the macrophyte than 50 kHz. Freitas et al. (2008) also showed that the 200 kHz acoustic diversity obtained in the inner Bay of Cádiz could reflect not only the influence of the sediment but also that of biological features including underwater vegetation. Moreover, Riegl et al. (2005), in a study on the drift macroalgae in the Indian River Lagoon, FL, USA, found that 50 kHz acoustic diversity was in better agreement with the sedimentary diversity than 200 kHz. Our results suggest that this is possibly a result of the higher acoustic energy passing through the vegetation layer to interact primarily with the substratum, so that most of the 50 kHz acoustic signals may be more strongly influenced by the underlying sediment than the overlying vegetation, which can then be sampled better by a frequency of 200 kHz. Preston et al. (2006) also revealed the ability of the 200-kHz survey frequency to distinguish bare from vegetated seabed.

Overall, we conclude that the AGDS QTC VIEW Series V is potentially a valuable tool for the remote assessment of underwater vegetation, whatever the sedimentary characteristics of the seabed. Although the amount of data was limited, our results also indicated that it is possible to model the biomass distribution of C. prolifera based on an acoustic survey at 200 kHz. This modelling approach will pose major challenges in areas with lower and higher biomass values per unit seabed surface. Under a patchy distribution of the macrophytes, or a low biomass, it is in fact possible that the system will not be able to detect the vegetation and show a resolution below the actual disappearance of the macrophytes. At the other end of the scale, it would be interesting to test the capacity of the system to detect changes in a range of very high biomass levels. In either case, surveys devoted to the study of underwater vegetation should include routinely the study of the superficial-sediment grain-size characteristics, because these will significantly affect the ability of the acoustic system to sample the vegetation cover adequately.

Acknowledgements

The work was supported by the Portuguese FCT (Fundação para a Ciência e a Tecnologia) under the project “ACOSHELF: Coastal-shelf ecosystem studies using acoustics” (POCI/MAR/56441/2004 and PPCDT/MAR/56441/04), by the programme “Acções Integradas Luso-Espanholas”, under the project “ACOFAN: shallow-water acoustic mapping of seagrass meadows and other benthic biotopes” (E-97/08) and by CESAM research funds as well as those of the research group “Ecología y Ordenación de Ecosistemas Marinos Costeros”, University of Murcia. Our colleague Rui Marques helped in preparing the acoustic system and in data collection. We also thank the Puerto and Club Náutico de Lo Pagán for harbour facilities. Finally, we acknowledge the valued comments from two anonymous reviewers.