© 2004 by ICES/CIEM International Council for the Exploration of the Sea/Conseil International pour l'Exploration de la Mer
Species discrimination of fish using frequency-dependent acoustic backscatter
National Marine Fisheries Service, Alaska Fisheries Science Center F/AKC2, PO Box 15700, 7600 Sand Point Way, Seattle, WA 98115, USA
*Correspondence to E. A. Logerwell: tel: +1 206 526 4231; fax: +1 206 526 6723. e-mail: libby.logerwell{at}noaa.gov.
The difference between mean volume-backscattering strength at 120 and 38 kHz (
MVBS) has been used to discriminate acoustically between macrozooplankton species, and between macrozooplankton and fish or small zooplankton. We examined whether MVBS could be used to discriminate between juvenile pollock (Theragra chalcogramma) and capelin (Mallotus villosus). Acoustic data at 38 and 120 kHz were collected in the Gulf of Alaska during August 2000 and 2001. We selected scattering layers of juvenile pollock and capelin that were sampled directly by midwater trawls. Although we found statistically significant differences at minimum integration thresholds ranging from –85 dB to –69 dB, the greatest difference between MVBS of juvenile pollock and capelin was observed at the highest integration threshold (–69 dB). We also found that, although there was substantial overlap between the frequency distributions of juvenile pollock and capelin MVBS at the smallest scale of analysis (0.1 nautical mile x 5 m cells), there was virtually no overlap between the MVBS distributions at the largest scale (
1 nautical mile x 20 m aggregations). We conclude that acoustic differencing at the scale of fish aggregations and at high integration thresholds can be used to distinguish between juvenile pollock and capelin.
Keywords: acoustic survey, capelin, dual frequency, Gulf of Alaska, Mallotus villosus, pollock, Theragra chalcogramma
Received 1 December 2003; accepted 6 April 2004.
 |
Introduction |
|---|
 Top Introduction Material and methodsResultsDiscussionReferences |
|---|
Acoustic differencing is a frequency-dependent technique that entails calculating the difference between mean volume-backscattering strength at two different frequencies (
MVBS). Researchers have used acoustic differencing to distinguish fish from zooplankton (Madureira et al., 1993a; Brierley and Watkins, 1996; Miyashita et al., 1997; McKelvey, 2000), and large species of zooplankton from small species of zooplankton (Everson et al., 1993; Madureira et al., 1993b; Mitson et al., 1996; Brierley et al., 1998). Acoustic differencing depends on the frequency-dependent relationship between target size and the magnitude of the echo. For acoustic differencing to be optimally effective, the frequencies used should span the transition from Rayleigh to geometric scattering (Holliday and Pieper, 1995). The optimal frequencies for fish (a few Hz to a few kHz) are less than those typically used during fisheries acoustic surveys (38 kHz and greater, MacLennan and Simmonds, 1992), so one would not expect differencing of acoustic-survey data to be effective at discriminating between different species of fish. To address this difficulty, Horne and Jech (1999) modelled the backscatter amplitude of a simulated population of threadfin shad. They found that at frequencies ranging from 38 to 420 kHz, backscatter amplitude varied, albeit unpredictably, among frequencies. Consistent with these model results, Kloser et al. (2002) were able to identify the dominant fish groups in Australian waters using frequencies typical of fisheries acoustic surveys (12, 38, and 120 kHz). In this paper, we apply acoustic differencing at 38 and 120 kHz to a species mixture that has not been previously examined, juvenile pollock and capelin in Alaskan waters. This research is part of a program at the Alaska Fisheries Science Center (AFSC) to evaluate the effect of commercial fishing activity on the prey availability to endangered Steller sea lions, Eumetopias jubatus, in the Gulf of Alaska (Wilson et al., 2003). Because the distribution and abundance of sea-lion prey are assessed with acoustic methods, one challenge is identifying the species composition of midwater scatterers, predominantly juvenile pollock (Theragra chalcogramma) and capelin (Mallotus villosus). Our goal was to test whether the multi-frequency technique of acoustic differencing could be used to aid species identification during our acoustic surveys in the Gulf of Alaska.
|
Material and methods |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
The eastern side of Kodiak Island in the Gulf of Alaska is the site of a study on the interactions between commercial fishing and the prey of Steller sea lions (Wilson et al., 2003). Replicate acoustic and trawl surveys were conducted during August 2000 and 2001 in two adjacent submarine troughs (Barnabus and Chiniak), subject to contrasting fishing impacts. The surveys comprised a series of transects spaced 3 nautical miles apart that extended across the troughs. Seventeen transects were surveyed in Chiniak Trough and 15 were surveyed in Barnabus Trough.
Fish distributions were assessed using echo-integration, trawl-survey (EIT) techniques during daylight hours aboard the NOAA ship, "Miller Freeman" (Traynor et al., 1990; Wilson et al., 2003). The acoustic data were collected with a SIMRAD EK 500 echosounder (reference to product names does not imply endorsement by the National Marine Fisheries Service, NOAA). Two SIMRAD split-beam transducers with 7° nominal beam widths, operated simultaneously, one at 38 kHz and the other at 120 kHz. The transducers were mounted 22 cm apart on the vessel's retractable centerboard, at a depth of 9 m below the water surface. The pulse lengths were 1.0 and 0.3 ms for the 38- and 120-kHz frequencies, respectively. A nominal 1-Hz ping-rate was used for both frequencies. Echo-integration data were logged with a horizontal resolution of about 5–6 m, depending on vessel speed which averaged 5–6 m s–1, and a vertical resolution of 0.1–0.5 m. Standard sphere, acoustic-system calibrations were conducted before and after the surveys to measure the performance of the echosounder at each frequency. Selection, classification, and analysis of the acoustic data were conducted with Echoview software (SonarData Pty. Ltd., Hobart, Tasmania, Australia).
A midwater, Aleutian wing trawl (AWT) was used to sample midwater acoustic-sign types (Wilson et al., 2003). The codend of the AWT was fitted with a 32-mm mesh liner in 2000, and a 9.5-mm mesh liner in 2001. The smaller mesh liner was used in 2001 in an effort to increase the ability of the AWT to sample capelin. The vertical opening of the AWT was 20 m. Trawl hauls were conducted when a significant acoustic sign was encountered to determine the species and size composition of the dominant scatterers. Trawls were deployed at the depth of the scattering, generally where little or no scattering above the layer of interest would contaminate the sample. The length of the trawl hauls varied from 1 to 2 nautical miles. A total of 40 midwater hauls was conducted in 2000 and 41 in 2001. Fish were subsampled from catches to determine the length composition. All lengths reported are "standard" length. However, capelin lengths were measured as fork length in 2000 and standard length in 2001. A regression equation was thus developed from multiple measurements on single fish to convert from fork length (FL) to standard length (SL): SL = 0.9525 x FL – 0.3142 (R2 = 0.98, n = 291).
Data from scattering layers that had been directly sampled by the trawls were selected for this analysis. The acoustic sign was classified as age-1 pollock, age-1/age-2 pollock mix, capelin, or capelin/age-0 pollock mix if that species or mix made up more than 80% of the trawl sample by weight excluding jellyfish and single large fish such as Pacific halibut. The remaining 20% (or less) of each trawl sample was made up of a variable mix of species with no particular species consistently dominating the haul catches. The acoustic data were analyzed from 15 trawls in 2000 and 16 trawls in 2001. Other trawls targeted adult pollock and were thus not deployed in portions of the water column occupied by juvenile pollock and capelin. The portion of the scattering layer sampled by the nets will be referred to as the "aggregation" for the remainder of this paper. The horizontal dimension of the aggregation was thus defined by the position where the net first reached its target depth to the position where net retrieval began. The vertical dimension was defined by the height of the net opening. Acoustic data integrated over these dimensions were used for the "aggregation-scale" analyses. The acoustic data from each aggregation were also binned into cells 0.1 nautical mile long and 5 m deep for a finer-resolution "cell-scale" analysis. A small integration cell is recommended in order to ensure that MVBS is calculated for a single species with a narrow size range (Kang et al., 2002).
Definitions of acoustic-data variables used in the analyses are (MacLennan and Simmonds, 1992; SonarData Pty. Ltd., Hobart, Tasmania, Australia):
- Sv = volume-backscattering strength (dB re 1 m–1)
- Sv min = minimum Sv value within a cell that is greater than the below threshold value
- MVBS = mean of the Sv values according to the following equation:
- Sv min = minimum Sv value within a cell that is greater than the below threshold value
|
|
|
|
|
|
- N = number of samples in area for which MVBS is to be calculated
- Svi (i = 1...N) = Sv values of the N samples
i = 0 if sample i is bad or excluded data, i = 1 otherwise
i = 0 if sample i is thresholded data, i = 1 otherwise
i = 0 if sample i is a no data sample, i = 1 otherwise
- Svi (i = 1...N) = Sv values of the N samples
Threshold analysis was carried out to determine which threshold would maximize the frequency-dependent differences among the acoustic-sign types. A range of minimum integration thresholds from –69 to –85 dB (–69, –75, –79, and –85 dB) were applied at the two scales of resolution ("aggregation" and "cell") before calculating mean volume-backscattering strength, MVBS (dB re 1 m–1) at each scale. These thresholds range from the high thresholds commonly applied to acoustic data on strong scatterers, such as gadoid fishes, to minimally thresholded data. A low threshold has the advantage of integrating weaker signals from small scatterers, whereas too low a threshold could result in the integration of unwanted scatterers and noise.
Noise is also amplified with depth as a result of the time-varied-gain (TVG) applied during echo integration. We used minimum integration thresholds to eliminate unwanted scatterers and noise. To determine whether additional efforts were needed for the removal of noise, we examined Sv min (dB re 1 m–1) at all thresholds from transect data where little biological scattering was evident. A 1-nautical mile segment of a Barnabus Trough transect, and a 2.3-nautical mile segment of a Chiniak Trough transect were selected for noise analysis. Water-depth ranges were 100–145 m and 135–155 m, respectively. In addition, data from several transects were used to quantify the noise contribution following the method reported by Korneliussen (2000). For transect segments with little biological scattering, relatively high Sv min values were evident near the surface in the Barnabus and particularly the Chiniak transect-segment data as a result of scattering layers <25 m deep (Figures 1 and 2). Moderate Sv min values were also observed at intermediate depths on the Barnabus segment (Figure 2). This can be attributed to dispersed midwater (50–150 m) scatterers evident on the echogram. A continuous increase in Sv min with depth was not observed in the 38 kHz data along either transect segment, indicating no significant noise intrusion at this range and frequency. Other researchers using the same acoustic instrumentation have reported that the relative contribution of noise is also quite small for ranges <750 m (Korneliussen, 2000). Some noise intrusion was evident in the 120 kHz data. For example, Sv min increased slightly at depths greater than approximately 120 m along the Chiniak segment at the –85 dB threshold (Figure 1). Along the Barnabus segment, Sv min increased at depths greater than approximately 110 m at the –85 dB threshold (Figure 2). Using Korneliussen's (2000) method, the results suggested that the noise contribution for the 120 kHz data at 100-m depth was approximately –79 to –85 dB and for 38 kHz was approximately –99 to –91 dB. The fish aggregations selected for acoustic differencing in our study were often quite dense, and located in the upper water column (100-m depth or less), where noise intrusions were minimal. Because of this we assumed that it was unnecessary to remove noise from the echo-integrated data beyond that removed by simply thresholding the data.
|
|
Estimates of
MVBS were calculated as MVBS at 120 kHz minus MVBS at 38 kHz for the aggregation- and cell-scale data sets. The MVBS values at the cell-scale were not normally distributed, so the non-parametric Wilcoxon test was used to test for differences between MVBS of different sign types (Sokal and Rohlf, 1995). Frequency distributions of MVBS at the aggregation-scale indicated that statistical tests at this scale were not necessary because the distributions did not overlap. Statistical analyses were conducted in S-PLUS 2000 (MathSoft, Inc., Seattle, WA, USA). The statistical tests were considered significant if p < 0.05. |
Results |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
In 2000, we observed two types of midwater fish sign: capelin/age-0 pollock mix and age-1 pollock. In 2001, we observed age-1 pollock and two additional types of sign: capelin, and age-1/age-2 pollock mix.
Fish-length distributions were similar across locations (trawl sites) and years (Figure 3). Age-1 pollock lengths averaged 224 mm in 2000 (Figure 3a), and 204 mm in 2001 (Figure 3d). The length distribution of age-1/age-2 pollock mix had one mode centered near 200 mm, indicating age-1 pollock, and a second mode near 300 mm, indicating age-2 pollock. Capelin lengths averaged 100 mm in 2000 (Figure 3b), and 98 mm in 2001 (Figure 3e).
|
Cell-scale analyses
The difference between age-1 and age-2 pollock and capelin
MVBS was greatest at the highest integration threshold, –69 dB (Figure 4). The difference in MVBS between the two types of capelin sign, with and without age-0 pollock, was greatest at the lowest integration threshold, –85 dB (Figure 4). The difference in MVBS between age-1 pollock and age-1/age-2 pollock mix was relatively unchanged across the range of integration thresholds (Figure 4).
|
Although the
MVBS values for juvenile pollock and capelin were significantly different at the –69 dB threshold, the MVBS frequency distributions of these species groups overlapped substantially (Figure 5a, b compared to Figure 5c, d). Similarly, the MVBS frequency distributions of age-1, and age-1/age-2 pollock mix MVBS showed substantial overlap, compare Figure 5a to Figure 5b, despite the fact that the difference was statistically significant. At the –85 dB threshold, MVBS of all four sign types were statistically different but, again, there was substantial overlap in the frequency distributions of the sign types (Figure 6).
|
|
Aggregation-scale analysis
The frequency distributions of juvenile pollock and capelin
MVBS at the aggregation scale showed no overlap. This is true for both the –69 dB (Figure 7) and the –85 dB (Figure 8) integration thresholds. Frequency distributions of similar sign types (age-1 pollock vs. age-1/age-2 pollock mix and capelin vs. capelin/age-0 pollock mix) did overlap (Figures 7 and 8), similar to the results of the cell-scale analysis.
|
|
|
Discussion |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
Our study is the first to apply multiple-frequency acoustics to distinguish between juvenile pollock and capelin. The
MVBS values for capelin aggregations observed during this study agree with previously published work on frequency-dependent target strength. Rose (1998) measured target strength of capelin in situ and found that target strength at 120 kHz was 1.5 dB less than target strength at lower frequencies (38 and 49 kHz). We observed MVBS for capelin ranging from –2 to –0.5 dB, depending on the integration threshold, which is consistent with Rose's (1998) findings. Miyanohana et al. (1990) measured target strengths of tethered pollock at frequencies of 25, 50, 100, and 200 kHz. The target strengths showed strong frequency dependence but the frequencies we used (38 and 120 kHz) were not tested, and furthermore there was no clear trend with frequency. Horne (2003) developed Kirchoff ray-mode (KRM) models from radiographs of walleye pollock in a tank to predict target strength (TS) at 38 and 120 kHz. The results for juvenile pollock (200–300 mm FL) indicated greater TS at 38 kHz than at 120 kHz, inconsistent with the positive MVBS values that we observed for age-1 and mixed age-1/age-2 pollock. However, the multi-modal structure of both modelled and observed target strengths makes it hard to determine with confidence the expected sign of MVBS.
At the cell-scale resolution, the greatest difference between juvenile pollock and capelin MVBS was at the highest integration threshold (–69 dB). Although statistical analysis indicated significant differences between juvenile pollock and capelin MVBS, there was considerable overlap in the frequency distributions of MVBS at this scale. This overlap diminishes the utility of acoustic differencing to discriminate between different fish species at this spatial scale. However, there was virtually no overlap in MVBS values at the aggregation scale of resolution. We conclude that acoustic differencing at the spatial resolution of aggregations and at a high integration threshold ( –69 dB) can be used to supplement net tows used for distinguishing juvenile pollock and capelin aggregations during acoustic surveys.
Korneliussen and Ona (2002) suggest that information on frequency-dependent backscattering be incorporated into post-processing of acoustic data to aid species identification. In order to examine the usefulness of acoustic differencing in improving species identification in our study area, we compared the species assignment to echosign based on MVBS with that based simply on scrutinizing. "Scrutinizing" refers to the assigning of species to acoustic data based on vertical and geographical location and morphology of the sign and on species composition of nearby net hauls (Simmonds et al., 1992; Reid et al., 1998). Scrutinizing is thus a relatively subjective process, whereas acoustic differencing is objective. Acoustic data from the survey had previously been scrutinized in 5-nautical mile segments using a threshold of –69 dB, so we applied the acoustic-differencing technique to all 5-nautical mile segments that contained juvenile pollock or capelin according to the results of scrutinizing. Our criteria for classifying sign types were based on the frequency distributions of MVBS at the aggregation scale (Figure 7). If MVBS was less than or equal 0.6, the scatterers were classified as capelin, if MVBS was greater than 0.6 the scatterers were classified as juvenile pollock. Within segments, we selected aggregations in the water column that had been delineated and classified during scrutinizing and calculated MVBS at the –69 dB threshold. During scrutinizing of the 2001 data, age-1 pollock and age-1/age-2 pollock mix sign types were not distinguished: both were classified as "juvenile pollock". The results of acoustic differencing and scrutinizing were very similar for all types of fish sign except age-1 pollock from the 2000 survey (Table 1). Acoustic differencing indicated that 6 of the 11 scrutinized age-1 pollock segments would have been classified as capelin. The acoustically determined biomass of these "misclassified" segments amounted to 54% of the total biomass originally scrutinized as age-1 pollock. This suggests that the biomass of age-1 pollock was overestimated by as much as 54%, potentially a significant error. All six misclassified segments were adjacent to scrutinized capelin (either vertically or horizontally), in such a way that it is likely that the decision to classify these as juvenile pollock was based on changes in school morphology or depth. This is an example of how acoustic differencing could be used to supplement net-tow data for species identification. A scientist deciding whether to assign a species composition to acoustic sign not directly sampled by nets could employ MVBS in addition to school morphology and depth as criteria. Simard and Lavoie (1999) applied acoustic differencing to their study of the distribution of krill aggregations in the Gulf of St. Lawrence and were thus able to classify krill- and fish-scattering layers, respectively, in areas where net sampling was not carried out.
|
In summary, our analysis of acoustic and net-trawl data collected off the east coast of Kodiak Island in the Gulf of Alaska indicates that acoustic data at two frequencies (38 and 120 kHz) can be used to discriminate between juvenile pollock and capelin aggregations. There is some indication that different age classes of pollock and different types of capelin aggregations (pure or mixed with age-0 pollock) can also be discriminated in two-frequency acoustic data. Data at additional frequencies might improve the power of this technique to distinguish among these types of aggregations composed of more similarly sized fish (Miyanohana et al., 1990; Korneliussen and Ona, 2003). Because juvenile pollock and capelin occupy similar depths in the water column, this technique will be useful as a supplement for net sampling during acoustic studies in the Gulf of Alaska.
|
Acknowledgements |
|---|
We acknowledge the invaluable assistance of the captain and crew of the NOAA Research Vessel "Miller Freeman". Thanks also to the scientific research crew: J. Burgos, C. Brothers, S. deBlois, M. Dorn, R. Foy, M. Guttormsen, D. Hanson, A. Hollowed, T. Jackson, K. Landgraf, J. O'Brien, P. Porter, M. Shima, S. Stienessen, P. Walline, and L. Watson. This paper was improved by comments from Anne Hollowed, William Karp, John Horne, Martin Dorn, Alex DeRobertis, Gary Duker, and two anonymous reviewers.
|
References |
|---|
|
TopIntroductionMaterial and methodsResultsDiscussionReferences |
|---|
-
Brierley A. and Watkins J. (1996) Acoustic targets at South Georgia and the South Orkney Islands during a season of krill scarcity. Marine Ecology Progress Series 138:51–61.[Web of Science]
Brierley A., Ward P., Watkins J., Goss C. (1998) Acoustic discrimination of Southern Ocean zooplankton. Deep-Sea Research II 45:1155–1173.[CrossRef]
Everson I., Goss C., Murray W. (1993) Comparison of krill (Euphausia superba) density estimates using 38 and 120 kHz echosounders. Marine Biology 116:269–275.[CrossRef]
Holliday D.V. and Pieper R.E. (1995) Bioacoustical oceanography at high frequencies. ICES Journal of Marine Science 52:279–296.[CrossRef][Web of Science]
Horne J.K. and Jech J.M. (1999) Multi-frequency estimates of fish abundance: constraints of rather high frequencies. ICES Journal of Marine Science 56:184–199.
Horne J.K. (2003) The influence of ontogeny, physiology, and behaviour on the target strength of walleye pollock (Theragra chalcogramma). ICES Journal of Marine Science 60:1063–1074.
Kang M., Furusawa M., Miyashita K. (2002) Effective and accurate use of difference in mean volume-backscattering strength to identify fish and plankton. ICES Journal of Marine Science 59:794–804.
Kloser R.J., Ryan T., Sakov P., Williams A., Koslow J.A. (2002) Species identification in deep water using multiple acoustic frequencies. Canadian Journal of Fisheries and Aquatic Sciences 59:1065–1077.
Korneliussen R.J. (2000) Measurement and removal of echo integration noise. ICES Journal of Marine Science 57:1204–1217.
Korneliussen R.J. and Ona E. (2002) An operational system for procession and visualizing multi-frequency acoustic data. ICES Journal of Marine Science 59:293–313.
Korneliussen R.J. and Ona E. (2003) Synthetic echograms generated from the relative frequency response. ICES Journal of Marine Science 60:636–640.
MacLennan D.N. and Simmonds E.J. (1992) Fisheries Acoustics(Chapman and Hall, London) 325 pp.
Madureira L., Everson I., Murphy E. (1993) Interpretation of acoustic data at two frequencies to discriminate between Antarctic krill (Euphausia superba Dana) and other scatterers. Journal of Plankton Research 15:787–802.
Madureira L., Ward P., Atkinson A. (1993) Differences in backscattering strength determined at 120 and 38 kHz for three species of Antarctic macroplankton. Marine Ecology Progress Series 93:17–24.[Web of Science]
McKelvey D. R. (2000) The use of two frequencies to interpret acoustic-scattering layers. Master of Science Thesis, University of Washington, Seattle, Washington.
Mitson R.B., Simard Y., Goss C. (1996) Use of a two-frequency algorithm to determine size and abundance of plankton in three widely spaced locations. ICES Journal of Marine Science 53:209–215.
Miyanohana Y., Ishii K., Furusawa M. (1990) Measurements and analyses of dorsal-aspect target strength of six species of fish at four frequencies. Rapports et Procès-Verbaux des Réunions du Conseil International pour l'Exploration de la Mer 189:317–324.
Miyashita K., Aoki I., Seno K., Taki K., Ogishima T. (1997) Acoustic identification of isada krill, Euphausia pacifica Hansen, off the Sanriku coast, north-eastern Japan. Fisheries Oceanography 6:266–271.[CrossRef][Web of Science]
Reid D. G., Fernandes P. G., Bethke E., Couperus A., Goetze E., Hakansson N., Pedersen J. (1998) On visual scrutiny of echograms for acoustic stock estimation. ICES. CM 1998/B: 1.
Rose G. (1998) Acoustic target strength of capelin in Newfoundland waters. ICES Journal of Marine Science 55:918–923.
Simard Y. and Lavoie D. (1999) The rich krill aggregation of the Saguenay–St. Lawrence Marine Park: hydroacoustic and geostatistical biomass estimates, structure, variability, and significance for whales. Canadian Journal of Fisheries and Aquatic Sciences 56:1182–1197.
Simmonds E. J., Williamson N. J., Gerlotto F., Aglen A. (1992) Acoustic-survey design and analysis procedure: a comprehensive review of current practice. ICES Cooperative Research Report, 187.
Sokal R. and Rohlf F. (1995) Biometry, the Principles and Practice of Statistics in Biological Research(W.H. Freeman and Company, New York) 887 pp.
Traynor J., Williamson N., Karp W. (1990) A consideration of the accuracy and precision of fish-abundance estimates derived from echo-integration surveys. Rapports et Procès-Verbaux des Réunions du Conseil International pour l'Exploration de la Mer 189:101–111.
Wilson C.D., Hollowed A.B., Shima M., Walline P., Stienessen S. (2003) Interactions between commercial fishing and walleye pollock. Alaska Fishery Research Bulletin 10:61–77.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. Aksland Applying an alternative method of echo-integration ICES J. Mar. Sci., January 1, 2006; 63(8): 1438 - 1452. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||