4
Seafloor characterization using time- dependent acoustic backscatter: study at Arabian Sea Bishwajit Chakrabortyl and Chanchal De2 'National Institute of Oceanography, Dona Paula, Goa- 403 004 2 Naval Physical and Oceanographic Laboratory, Thrikkakara, Kochi- 682021 Abstract- This paper presents application of "inversion modeling" technique to echo shape data acquired shallow water regions off Goa. The estimated sediment and seafloor roughness parameter for two different frequencies (33 and 210 kHz) provide comparable results. The mean grain size of the sediment ( MO ) estimated through the inversion of echo envelope data deviates for mixed sediments of clayey-silt type of five locations for both the operating frequencies, though the roughness characteristics (w2) and volume characteristics (U,) show a good agreement for relatively mixed and coarse-grained sediments. The extent of deviation in estimated grain size parameters can be related well the estimated sediment volume roughness parameters. Location wise higher sediment volume inhomogeneities in turn show difference in the estimated mean grain size (interface roughness) parameter for five mixed sediment type (clayey-silt). I. INTRODUCTION Seafloor characterization using high frequency acoustic methods are considered as useful technique to understand fine scale processes. Among these, backscatter models provide a means for remote seafloor classification and characterization using measured acoustic energies from the interface of water and marine sediments. However, recent developments show that the shape of the echo backscatter contains information regarding seafloor roughness (interface and sediment volume) and mean sediment grain size. In general, there are two approaches exist for characterization of the seafloor sediments. The first is to measure the seafloor parameters (roughness and sediment grain size etc.) of the limited seafloor spot locations by simultaneous use of geological sampling and associated instrumentations. It is impossible to acquire such information continuously especially during real time application. However, there exists numerical approach called "inversion modeling". This approach is conditional and the physical parameters obtained from inversion modeling are dependent on the accuracies of forward model developed. In general, the measured seafloor physical parameters are precise for the areas close to the spot locations where ground-truth measurements are made. In this work, use of inversion modeling is carried out to echo shape data acquired from the central part of the western continental shelf of India i.e., at Arabian Sea off Betul, southern end of the Goa area for estimation of seafloor parameters. Pouliquen and Lurton [1] proposed a method to identify the sea bottom type by comparing the signal time envelopes from standard echo sounders with a set of reference signals computed for different sea bottom types. Sternlicht and de Moustier [2, 3] have presented a time-dependent seafloor acoustic backscatter model using the physical concepts of acoustic backscattering from seafloor based on the references therein. In this paper, the seabed roughness parameters are estimated using inversion technique presented by Sternlicht and de Moustier [4] and the inverted parameters are then compared with the ground truth data to validate the applicability of the time dependent acoustic backscatter model in the shallow water region in Indian Ocean. II. MODEL AND DATA The total time dependent echo intensity of the bottom acoustic backscattering signal is the summation of acoustic backscattering energy at the water-sediment interface (Ii) and the intensity backscattered from the sediment volume (I,). The backscattered signal by the water-sediment interface is derived analytically using the Kirchhoff approximation of the pressure field and the reflection coefficients. The total roughness I(t) is expressed by (1). I(t) = Ii (t) + IV(t)(1 Ii depends on water-sediment interface characteristics and the transducer characteristic and Iv depends on the sediment volume scattering coefficient and the transducer characteristics. The computations of I(t) incorporates the characteristics and geometry of the echo sounder, the environmental factors and the characteristics of the transmitted sound pulse. The modeled shape and intensity of the average bottom echo envelope are compared with the measured data by echo sounder to infer bottom roughness characteristics. The normal incidence dual frequency single beam echo sounder (RESON NS420) was used to collect the backscattered data off west coast India using 33 kHz frequencies with pulse length 0.97 ms and 210 kHz frequencies with pulse length 0.61 ms. Present data were acquired at six different locations having three different seabed sediment type. The two sediment types are sand and 978-1-4244-2126-8/08/$25.00 ©2008 IEEE

[IEEE OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Kobe, Japan (2008.04.8-2008.04.11)] OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Seafloor characterization using time-dependent acoustic

Embed Size (px)

Citation preview

Page 1: [IEEE OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Kobe, Japan (2008.04.8-2008.04.11)] OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Seafloor characterization using time-dependent acoustic

Seafloor characterization using time-dependentacoustic backscatter: study at Arabian Sea

Bishwajit Chakrabortyl and Chanchal De2'National Institute of Oceanography, Dona Paula, Goa- 403 004

2 Naval Physical and Oceanographic Laboratory, Thrikkakara, Kochi- 682021

Abstract- This paper presents application of "inversion modeling" technique to echo shape data acquired shallow water regions offGoa. The estimated sediment and seafloor roughness parameter for two different frequencies (33 and 210 kHz) provide comparableresults. The mean grain size of the sediment (MO ) estimated through the inversion of echo envelope data deviates for mixed sedimentsof clayey-silt type of five locations for both the operating frequencies, though the roughness characteristics (w2) and volumecharacteristics (U,) show a good agreement for relatively mixed and coarse-grained sediments. The extent of deviation in estimatedgrain size parameters can be related well the estimated sediment volume roughness parameters. Location wise higher sediment volumeinhomogeneities in turn show difference in the estimated mean grain size (interface roughness) parameter for five mixed sediment type(clayey-silt).

I. INTRODUCTION

Seafloor characterization using high frequency acoustic methods are considered as useful technique to understand fine scaleprocesses. Among these, backscatter models provide a means for remote seafloor classification and characterization usingmeasured acoustic energies from the interface of water and marine sediments. However, recent developments show that the shapeof the echo backscatter contains information regarding seafloor roughness (interface and sediment volume) and mean sedimentgrain size. In general, there are two approaches exist for characterization of the seafloor sediments. The first is to measure theseafloor parameters (roughness and sediment grain size etc.) of the limited seafloor spot locations by simultaneous use ofgeological sampling and associated instrumentations. It is impossible to acquire such information continuously especially duringreal time application. However, there exists numerical approach called "inversion modeling". This approach is conditional and thephysical parameters obtained from inversion modeling are dependent on the accuracies of forward model developed. In general,the measured seafloor physical parameters are precise for the areas close to the spot locations where ground-truth measurementsare made. In this work, use of inversion modeling is carried out to echo shape data acquired from the central part of the westerncontinental shelf of India i.e., at Arabian Sea off Betul, southern end of the Goa area for estimation of seafloor parameters.

Pouliquen and Lurton [1] proposed a method to identify the sea bottom type by comparing the signal time envelopes fromstandard echo sounders with a set of reference signals computed for different sea bottom types. Sternlicht and de Moustier [2, 3]have presented a time-dependent seafloor acoustic backscatter model using the physical concepts of acoustic backscattering fromseafloor based on the references therein. In this paper, the seabed roughness parameters are estimated using inversion techniquepresented by Sternlicht and de Moustier [4] and the inverted parameters are then compared with the ground truth data to validatethe applicability of the time dependent acoustic backscatter model in the shallow water region in Indian Ocean.

II. MODEL AND DATA

The total time dependent echo intensity of the bottom acoustic backscattering signal is the summation of acousticbackscattering energy at the water-sediment interface (Ii) and the intensity backscattered from the sediment volume (I,). Thebackscattered signal by the water-sediment interface is derived analytically using the Kirchhoff approximation of the pressurefield and the reflection coefficients. The total roughness I(t) is expressed by (1).

I(t) = Ii(t) + IV(t)(1

Ii depends on water-sediment interface characteristics and the transducer characteristic and Iv depends on the sediment volumescattering coefficient and the transducer characteristics. The computations of I(t) incorporates the characteristics and geometry ofthe echo sounder, the environmental factors and the characteristics of the transmitted sound pulse. The modeled shape andintensity of the average bottom echo envelope are compared with the measured data by echo sounder to infer bottom roughnesscharacteristics.The normal incidence dual frequency single beam echo sounder (RESON NS420) was used to collect the backscattered data off

west coast India using 33 kHz frequencies with pulse length 0.97 ms and 210 kHz frequencies with pulse length 0.61 ms. Presentdata were acquired at six different locations having three different seabed sediment type. The two sediment types are sand and

978-1-4244-2126-8/08/$25.00 ©2008 IEEE

Page 2: [IEEE OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Kobe, Japan (2008.04.8-2008.04.11)] OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Seafloor characterization using time-dependent acoustic

clayey-silt. The study area is shown in Fig. 1. The raw analogue output on the receiver circuit board was tapped and connected toa PCL 1712L 12-bit A/D converter with a sampling frequency 1 MHz. The transducer beam width for 33 kHz and 210 kHzfrequencies were 20 degrees and 9 degrees respectively. Hilbert transform was used to obtain the echo envelope within the range0-5 volt. Ground-truth samples were collected by grab at all locations where the echo data acquisitions were made. The surfacesediment samples were analyzed in the laboratory using standard methods.

7300

I16 O0O

I 5 0

74 "00'

Fig. 1 Study Area off Betul

In general, echo shapes depend on various factors such as noise, natural variability of the study area, and echo-sounderinstability. To obtain acoustic signal stability ten pings are usually averaged. The averages with 20 echoes and 100 echoes arealso used [2, 4]. The average with higher number of echoes may reduce the amplitude level of the echo from their true value,which may lead to wrong interpretation. Here in the study area we have used the averaged echo with 10, 20, 30, 40 and 50 echoesand found that the average with 10 echoes are slightly noisy and produce a degraded results when compared with the ground truth.It is observed that the averages with 20, 30, 40 echoes produce comparatively better results and slightly less noisy. Hence, in thisstudy, for comparison with the temporal model 20 echo envelopes are averaged. The echo envelopes, which are affected byvessel's roll and pitch, are removed from the stack and only good ones are selected by visual checking for further analysis. Theminimum threshold alignment method is used for taking the average of the echo envelope. To take into account the variability ofthe bottom, the echoes are averaged with 20 echoes successively with 5000 overlap, i.e., the echoes are averaged with 1-20sequences, 11-30 sequences, 21-40 sequences and so on till the end ofthe number ofgood echo traces.

III. DATA ANALYSIS METHOD

The various model parameters used are the mean grain size MO, defined as MO =-log2 Ug, where Ug is the sedimentsmean grain diameter, spectral strength (w2) and spectral exponent (7y) and sediment volume scattering coefficient (aj) The

978-1-4244-2126-8/08/$25.00 ©2008 IEEE

.6......

.I

Page 3: [IEEE OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Kobe, Japan (2008.04.8-2008.04.11)] OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Seafloor characterization using time-dependent acoustic

interface roughness is modeled by a power-law spectrum W(k) = w2k ;", where k is the 2D wave number vector for bottomrelief. These inverted parameters can be used to infer the sediment bottom types [5]. Sternlicht and de Moustier [4] usedsimulated annealing with downhill simplex method for inversion of these parameters by echo envelope matching. The averageecho is matched with the modeled temporal echo envelope with the specified mean altitude and sediment geo-acoustic parameters.The objective function for measuring the mismatch between the averaged data (Pa) and the modeled echo energy (pm) is expressedby (2)

(Pa Pm)2 (2)

,PaThe goal is to minimize the functionf(p) using simulated annealing with down-hill simplex algorithms. Global optimization is

essential for this type of problem as in the multi-dimensional search space; it is highly possible that there exist a number of localminima. The model-data matching procedure initially comprises of ID search to find out the general sediment type (MO ) and thespectral exponent ( 7). After establishing the general sediment type, three-dimensional global search is carried out for theroughness spectral strength (w2), the sediment volume scattering coefficient (UcT) and the mean grain size (MO ). The ID searchspace generally has one extremum with respect toMO, when other variables are dependent on MO . The 'best' solution is foundby iteratively bracketing the minimum using golden section search algorithm coupled with inverse parabolic interpolation. Theoutput of this ID search process provides the starting MO value for the next 3D global search method. In the second stage thevalue of the spectral exponent is chosen based on the value of MO . For finer sediments it is assigned as 3.3 and for sandysediments it is assigned as 3.0. The geo-acoustic parameters contained in the temporal model have numerous local maxima orminima. To achieve a better result, it is necessary to constrain the solution space using a priori information. This typical two-stagesearch process increases the possibilities that a true global maximum will be found. The ID search method provides the a prioriknowledge on the mean grain size (MO ). The MO value determines the type of sediments and the roughness spectral exponent isfound out based onMO . The MO value is allowed to vary betweenMO + 1 5, and the 3D search process over the limited searchrange of MO, UV and w2 is carried out to fine-tune the match between model and data. If the volume scattering energy componentincreases compared to the surface scattering energy, an empirical penalty is added to the objective function. The penalty dependson the severity of the violation of the fact that the surface scattering coefficient is comparatively low compared to volumescattering coefficient [4]. The penalty using the maximum values of the volume (Ivmax) and interface (Jimax) scattering intensities isgiven by (3).

f(p) = f(p).4.{1 + 5 vmax _0.634, for the case vmax > 0.63 (3)L Krnimax 'imax

IV. DISCUSSIONS

The envelope data collected over two types of sediments (e.g., five clayey-silt locations and one sandy location) are analyzed.The averaged echo enveloped then matched with modeled echo envelope using the two-stage envelope matching technique.Simulated annealing with downhill simplex algorithm is used for finding the global minimum of the objective function. Theestimated variability of the bottom surface in terms of MO, UV and w2 values for 33 kHz and 210 kHz (from inversion) for the sixregions are presented in Table I. The estimated sediment and seafloor roughness parameter for two different frequencies providecomparable results. The mean grain size of the sediment (MO ) estimated through the inversion of echo envelope data deviatesfor mixed sediments of clayey-silt type for both the operating frequencies, though the roughness characteristics (w2) and (UV)shows a good agreement for relatively mixed sediments. However, deviations of estimated values are minimum for singlelocation of sandy sediment area. Fig. 2 shows one typical comparison of the modeled data with the observed data for 210 kHzfrequencies in clayey silt area with MO value 6.67 (ground truth). This deviation ofmean grain size of the sediment (MO ) fromground truth data of mixed sediment type is correlated well with the estimated sediment volume roughness ( av ) parameters. Theexistence of the seafloor benthic animal from off Goa area is reported earlier [6]. Location wise sediment volume inhomogenityshow difference in the estimated interface roughness parameters [5] for five mixed sediment type (clayey-silt).

ACKNOWLEDGMENTS

The Department of Information Technology, New Delhi, supported the echo data acquisition activity of this work. We arethankful to Dr. S. R. Shetye, Director, NIO for encouragement and also thankful to R. G. Prabhudesai for his suggestions and G.S.Navelkar and William Fernandes during data acquisition activities. The second author is also grateful to Shri S.Ananthanarayanan, Director, NPOL for his encouragement.

978-1-4244-2126-8/08/$25.00 ©2008 IEEE

Page 4: [IEEE OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Kobe, Japan (2008.04.8-2008.04.11)] OCEANS 2008 - MTS/IEEE Kobe Techno-Ocean - Seafloor characterization using time-dependent acoustic

413.5 44 44.5 41Tiine (iiis)

46 46.5

Fig. 2 Comparison of the model output and the observed envelope (210 kHz) for mixed seafloor sediment

TABLE IRESULTS FOR 33 /210 KHZ AFTER EMPLOYING INVERSION TECHNIQUE

Laboratory Measured MO (Estimated) W2 avMO (Sediment Type)

Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.6.71 (clayey-silt) 4.76 /4.24 0.67 /0.20 8.6E-4 / 1.lE-3 5.5E-4 / 1.5E-4 9.1E-2 / 4.9E-1 5.2E-2 / 9.3E-26.68 (clayey-silt) 6.93 /5.12 0.83 /0.56 6.5E-4 / 5.6E-4 1.3E-4 / 2.4E-4 1.OE-2 / 2.1E-1 4.4E-3 / liE-I6.67 (clayey-silt) 5.20 /5.17 0.80/ 1.09 1.4E-3 / 4.3E-3 5.6E-4 / 1.4E-2 2.6E-2 / 1.2E-2 2.5E-2 / 1.0E-26.32 (clayey-silt)) 4.96 /4.67 0.74 /0.39 8.6E-4 / 1.8E-3 7.2E-4 / 5.3E-4 3.5E-2 / 1.7E-1 2.2E-2 / 7.3E-25.82 (clayey-silt) 5.18 /4.29 0.33 /0.22 6.7E-4 / 1.lE-3 3.6E-4 / 1.4E-4 1.3E-2 / 1.8E-1 1.lE-2 / 2.5E-21.69 (sand) 1.58 / 1.31 0.56 /0.30 6.7E-3 / 1.5E-3 1.9E-3 / 3.2E-4 6.5E-2 / 7.4E-1 3.1E-3 / 2.6E-1

REFERENCES

[1] E. Pouliquen and X. Lurton, "Sea-bed identification using echo-sounder signals," in European conference on underwater acoustics Ed. By. M. Weydert,Elsevier Applied Science (London) ,1992, pp. 535-538.

[2] D.D. Sternlicht and C.P. de Moustier, "Temporal modeling of high frequency (30-100 kHz) acoustic seafloor backscatter: shallow water results", in Highfrequency acoustics in shallow water, CP-45, Lerici, Italy,1997, pp. 509-516.

[3] Daniel D. Sternlicht and Christian P. de Moustier, "Time dependent seafloor acoustic backscatter (10-100 kHz)," J Acoust. Soc. Amer., vol. 114, pp. 2709-2725, November 2003.

[4] Daniel D. Sternlicht and Christian P. de Moustier, "Remote sensing of sediment characteristics by optimized echo-envelope matching," J Acoust. Soc.Amer., vol. 114, pp. 2727-2743, November 2003.

[5] Darrell R. Jackson, D.P. Winebrenner and Akira Ishimaru, "Application of the composite roughness model to high-frequency bottom backscattering," JAcoust. Soc. Amer., vol. 79, pp. 1410-1422, May 1986.

[6] Bishwajit Chakraborty, V. Mahale, G.S. Navelkar, B.R. Rao, R.G. Prabhudesai, B. Ingole and G. Janakiraman, "Acoustic Characterization of seafloorhabitats on the western continental shelf of India", ICES Jour. Mar. Sci., pp. 551-558, April 2007.

978-1-4244-2126-8/08/$25.00 ©2008 IEEE

0.7

0.6

0.5

0.4ol 0.3

0.0.1

0.-

013