© 2003 by ICES/CIEM International Council for the Exploration of the Sea/Conseil International pour l'Exploration de la Mer
Acoustic semi-tomography in studies of the structure and function of the marine ecosystem
Sea Fisheries Institute ul. Ko
taja 1, 81-332 Gdynia, Poland
*tel: +48 58 620 1728, x 215; fax: +48 58 620 2831. e-mail: orlov{at}miryb.mir.gdynia.pl.
This paper describes a new, enhanced version of the macrosounding method. It was first introduced over 10 years ago by the author. The purpose was to integrate acoustic data with environmental and biological parameters in a time and space scale, for use in marine-ecosystem analyses. Previous visualizations were replaced in 2002 by a vari-coloured topographical space- or time-based matrix of fish volume-backscattering strength. The matrix is calculated from acoustic measurements, collected during the surveys for each elementary distance unit (ESDU) in standardized slices (depth intervals) of insonified volume (semi-tomography). Each sv visualization is accompanied by a cross-section of the same space unit, which provides a distribution of the selected environmental parameter measured during the same survey. Charts of such parameters are produced with the same procedures as for sv. This presentation of multidisciplinary data greatly improves recognition of time and space gradients. It exposes ecological inter-correlations that are difficult or impossible to estimate by numerical analysis of local diversity of the observed processes. The application of the method is illustrated by examples showing its practical use for data from 1994 to 2000 in the southern Baltic.
Keywords: acoustic tomography, Baltic fish, environmental variability, fish behaviour, fish distribution, structure of marine ecosystem, visualization
Received 6 December 2002; accepted 16 April 2003.
 |
Introduction |
|---|
 Top Introduction Materials and methodsResults and discussionReferences |
|---|
The increasing global climate changes that primarily influence aquatic ecosystems together with the accumulated contamination of those systems by human activity (Barnes and Mann, 1991) increase the need to develop adequate methods of monitoring the hydrosphere. Large-scale satellite observations of the biosphere are limited to the surface of the oceans because of the limited range of water penetration by electromagnetic waves. The total volume of the oceans which contain most of the accumulated thermal energy is sampled in a very limited way by direct or remote methods.
Systematic acoustic surveys are one of the most effective ways of producing vertical cross-sections of fish and plankton aggregations in marine ecosystems. The development of visualization methods for acoustic data is considered in a wide range of papers (Orlowski, 1990, 1998, 1999, 2000; Socha et al., 1996; Jech and Luo, 2000; Reid et al., 2000; Mayer et al., 2001; Roman et al., 2001). Most of these deal with monitoring relationships among biotic and abiotic factors of the marine ecosystem. Others are intended to verify numerical models of the aquatic habitat (Horne et al., 1996; Orlowski, 1998, 1999).
This paper describes the most recent version of the macrosounding method, introduced by the author in 1989 (Orlowski, 1990) and subsequently modified as a result of increasing needs, experience, and technical advances (Orlowski, 1998, 1999, 2000). A previous version was limited to the conversion of acoustic density of the main layer of fish echoes to space density of black dots within the depth limits of fish occurrence. The picture was supplemented by adding all available data on environmental factors as isolines. That solution is good for the quick assessment of basic features of data collection. A new version, called "T-macrosounding" (T – tomography), fully applies a slice structure (as in tomography) of acoustic and environmental data, enabling very precise observations of the details of spatial and temporal gradients within a full range of insonified depths. The observations can be performed over a wider range of database filtering possibilities, i.e. the area of analysis can be defined by selecting limits of the parameters recorded (geographic position, bottom depth, fish depth, time of day, etc.). Additional graphical transformation by commercial software of a calculated cross-section pattern is applied to give maximum readability of the final visualization.
|
Materials and methods |
|---|
|
TopIntroductionMaterials and methodsResults and discussionReferences |
|---|
Systematic acoustic surveys of the southern Baltic area have been conducted by the Polish Sea Fisheries Institute since 1981. The recording of samples 24 h a day for each nautical mile (nmi) distance unit, the elementary distance unit (ESDU), in a slice-structured database started aboard RV "Baltica" in 1994. An EK400 echosounder and a QD echo-integrating system and bespoke software were used. In 1998 an EY500 scientific system was introduced to meet international standards of acoustic measurements and allow the research to continue. Both systems were using a frequency of 38 kHz and the same hull-mounted transducer of 7.2°x8.0°. Calibration took place with a standard target in Swedish fjords in 1994–1997 and in Norwegian fjords in the period 1998–2000. The cruises were carried out in October and lasted for 2–3 weeks so that samples were collected over a distance of between 1000 and 1500 nmi. The survey tracks of all cruises followed the same grid to give high comparability of measurements. A schematic chart of the area and survey tracks over the period 1994–2000 is shown in Figure 1.
|
Biological samples were collected over the period from 1994 to 2000 by the same pelagic trawl every 37 nmi of the transect on average. The fish observed during all surveys were mostly pelagic, viz. herring and sprat (Clupeidae). Hydrographic measurements (temperature, T; salinity, S; and oxygen level, O2) were made by a Neil-Brown CTD system. These were sampled at haul positions in the main, so the sampling density is similar to that of the biological samples. At each hydrographic station values of the measured parameters were recorded at 2-m depth intervals (slices).
The results of the echo integration for each ESDU and for each "slice" (hydrographic station) were converted into values of normalized area-backscattering coefficient (SA), following Knudsen's formula (Knudsen, 1990):
|
| (1) |
is the conversion constant [m2 nmi–2 sr], sv(y) is the volume-backscattering strength [m–1 sr–1], y, yi, yi+1 are the depth and i-slice layer limits [m]. Due to the draught of the vessel, hull reverberations, and the aeration zone, the first layer of integration had to start at 15-m depth. Each ESDU unit was associated with a series of (sA)i values (slices), its geographic position, date, time of day, and bottom depth. The process of converting acoustic and environmental data into T-macrosounding visualization was carried out in two stages. In the first stage software prepared in the Polish Sea Fisheries Institute is applied to prepare and calculate a three-dimensional (3D) (x, y, z coordinates) matrix of a single cross-section of the marine ecosystem for the analysed factors (sv, T, S, and O2). This T-matrix is calculated with standardized space or temporal resolution, within stated limits of the selected parameters. Interpolation can be involved in the calculation process. A graphical chart in which each rectangle gets the colour corresponding to the mean value of analysed parameter is presented as the output of this stage of the process. The range of colours is based on a cumulative percentage of an empirical probability distribution of each parameter value. Such a method of forming the scale steps was presented by the author in 1999. One basic colour (see Figures 2 and 3) corresponds to approximately a 10% range of the cumulative distribution of the parameter empirical values. Black corresponds to the rows and columns of T-matrix for which data are not available (i.e. below the bottom depth, or where sampling did not occur). Two basic types of cross-section, geographic- and time-related, are taken into account.
|
|
In the first case, X expresses a distance along a parallel of latitude (Figure 1), or meridian. Y corresponds to a depth and the contour level to sv, estimated by transformation of SA values. The width of the transect is limited by the range of the longitude or latitude values of data collected. Bottom depth is taken into consideration in this process also. "Visualizations" have to be produced separately for daytime and night-time as the fish-distribution pattern differs significantly between those periods (Orlowski, 1999). Daytime is characterized by the presence of shoal-like concentrations within a wide depth range. During the night most pelagic fish, in the form of scattering layers, inhabit a reduced depth range in the warmer, near-surface waters.
The second type of cross-section, called TDS (TDS, time–depth–sv), describes the 24-h variability of sv against depth, illustrating basic characteristics of diel behaviour of fish, i.e. the dynamics of the migration pattern in an environmental background.
The data used in the calculations can be filtered and up to seven independent factors within defined ranges are taken into consideration. As a consequence, fish-distribution patterns can be observed in different geographical cross-sections or within the characteristic sub-areas of the ecosystem, taking into consideration all the limitations of the remaining factors (i.e. bottom depth, etc.).
A similar process is used in the calculation of the T-matrix of environmental factors. Due to a lower spatial density of the hydrographic samples, an adequate T-matrix is calculated with lower geographical resolution. Day and night options are not needed in this case.
In the second stage a visualization produced by our software is transformed by commercial graphic software using procedures selected after a long series of trials for the optimal results. The Jasc-Software PSP 7.04 (Anon., 2001 and personal communication with Jasc-Software) was chosen as a useful and efficient tool for this purpose. A procedure called "topography effect" was applied to convert a primary version of visualization into an enhanced multi-colour chart in which single rectangles of colour matrix elements are transformed into topographical patterns.
The effect works by applying three different operations. In the first Gaussian blur with the parameter corresponding to the blur radius is applied. The "blur radius" can be interpreted as the data-correlation radius. In the second operation a colour spectrum is generated using the density parameter. The parameter indicates the number of transition colour steps. Patterns in the form of interpolated terraces (in order to show x, y, and contour levels) are calculated in the final operation by the terrace-relief function to produce a pseudo-3D distribution of the parameter associated with the Oz axis by the shading effect. This is determined by the angle and the colour of the highlight. An example of the full transformation is given in Figure 2. All five measured parameters are estimated and entered into the software transformation via the T-matrix grid calculated at the first stage. Parameters should be stabilized for similar visualization series.
In the visualization presented in Figures 2 and 3 the parameters were as following: blur radius – 47 pixels, density – 18 steps per colour, highlight colour – white, and light direction – 45°. The topographical effect enhances the dynamic range of observation of the contour values of analysed factor very efficiently. It gives a quick estimation of the absolute values and the gradients of the factor in relation to the selected geographical 3D area.
|
Results and discussion |
|---|
|
TopIntroductionMaterials and methodsResults and discussionReferences |
|---|
Figures 2 and 3 show some examples of the practical application of the T-macrosounding method. In the first case the visualization gives a condensed overview of the diel mean migration pattern of fish estimated for the Polish EEZ in the warm years of 1994, 1995, and 1999. The relative density of the fish expressed by sv(x,y) is given with a resolution of 0.5 h, as a function of time of day (x) and a resolution of 4 m, as a function of sea depth (y). Important gradients of fish density and the diel transformation of its pattern are easily compared to the values and gradients of environmental factors. In this instance, the latter could be given as traditional depth-dependent diagrams (T(y), S(y), and O2(y)).
During the night-time fish are concentrated uniformly in the top 55 m in the main. This layer is strongly dependent on hydrographic factors, and in this period, in particular, plays an important role in the metabolism process (Helfman et al., 1997). During the day fish are active and occur over a wide range of depths, feeding or looking for nourishment, and this, as a consequence, gives a more dispersed sv pattern and lower (Sv < –67.6 dB) volume densities, though a single fish school can exceed many times the highest sv values. Such analyses are very important in research on fish behaviour in separate sub-areas of the ecosystem, characterized by different static and dynamic environmental characteristics, and frequently influenced by anthropogenic factors. The T-macrosounding method gives great possibilities of easily standardizing and comparing the different options with such observations.
In Figure 3 the T-macrosounding procedure has been applied to the distribution of fish along the most significant cross-section of the Polish EEZ. This area plays an important role in the final fish-biomass distribution in the southern Baltic and the transect represents ICES rectangles having the dominant role in that part of the Baltic (Orlowski, 2003). The profile takes in the areas along the 19°E meridian between Gdansk Bay and South Gotland Deep (see Figure 1). The cross-section has the form of a strip 96-nmi long and 1°-longitude wide (approximately 34 nmi). The "strip" was divided into 48 elements (each 2-nmi long) and the depth was divided into 25, 4-m slices for sv visualization T-matrix (1200 cells) purposes. A smaller spatial density of hydrographic stations means that the visualizations of environmental factors were calculated for 4-nmi units. Given the higher resolution of the depth slices (2-m intervals) the environmental visualization T-matrix had the same number of cells (1200) (24x50).
An analysis of the results for fish-stock assessment from 1989 to 2001 has shown that during warmer years, when temperature in the southern part of the profile was significantly higher, a significant increase of the fish stock was recorded. The opposite phenomenon was observed during the cold years (Barnes and Mann, 1991; Helfman et al., 1997; Orlowski, 2000). A detailed analysis of the influence of thermal conditions on fish-distribution structure in the southern Baltic is described in Orlowski (2003).
The application of the T-macrosounding method enables a detailed analysis of fish behaviour and correlated environmental factors to be provided in the critical zone. Visualizations in the upper section of Figure 3 show the geographical structure of the fish distribution, expressed by sv values along the profile. The phenomenon was analysed separately for the cold years of 1996, 1998, and 2000 and warm years, 1994, 1995, and 1999 data series. The analysis was performed for night-time when fish are more stabilized by hydrographic factors.
The temperature structure is shown over the sv pattern by isolines, not by the T-macrosounding chart, which was possible due to its simpler form. T-macrosounding visualization has to be used to produce a more complicated oxygen-level chart and to show details of the method in full. No differences in the salinity pattern between warm and cold years were observed and so the factor may be ignored.
The patterns of fish distribution for cold years showed a significant limitation of the zone of high sv values (sv > –67.6 dB) to areas of higher water temperature (T>7°C) in the top 45-m depth. The depth of a more abundant layer was decreasing down to 20 m towards the north, while it was at about 40-m depth in the southern part of the profile.
However, during the warm years, higher values of sv were found over a significantly wider range of depths from 75 m in the south to 65 m in the north. The areas mentioned were very well correlated with a water mass with a temperature that exceeded 7°C (see temperature isolines in Figure 3).
When we consider the charts of oxygen distribution (lower section of Figure 3) the nature of the phenomenon becomes more complicated. In this case we can see clearly how T-macrosounding cross-sections can expose two parallel phenomena: (1) during cold years the fish-distribution layer depends on water temperatures (lower depth limit) and (2) the fish layer is perfectly correlated with the oxygen-level range (O2 > 7 ml l–1).
During the warm years, layers of fish concentrations (sv > –67.6 dB) were correlated in the main with warm water masses with an exception in the northern part of the area (north to 55°30'N). At the slope of the South Gotland Deep, the concentrations were strongly correlated to water characterized by a high oxygen level (O2 > 7 ml l–1). The area of main fish concentration (Gdansk Bay and Deep) was characterized by lower (O2 < 6.8 ml l–1) oxygen. This means that the temperature can be considered as a factor of the first rank in the environmental influence on fish distribution. It must also be added that the biological structure of fish is different: in the south there is a high percentage of young fish while in the north of the profile there are high year classes of adult fish.
Recapitulating on both examples we can conclude that the application of a series of T-macrosounding visualizations facilitates the identification and characterization of important local differences in environmentally conditioned fish behaviour over a wide range of absolute values and gradients of the factors observed in the same 3D geographic area. This information is very necessary for the better understanding of particular marine-ecosystem elements at all time scales. It provides a new approach to the treatment of inter-disciplinary data collected during research cruises, with respect to both the small- and large-scale analyses of the ecosystem in the context of global variability.
This method complements the range of visualization methods already reported. Horne et al. (1996), for example, applied a similar visualization technique to produce patterns of numerical density of fish and temperature distribution along single transects of Lake Erie in verification of a bio-energetic model of the aquatic habitat. Very small cells (40 m x 0.5 m) correlating with model assumptions were used. A similar visualization technique for single transects was used by Roman et al. (2001) to present research on zooplankton and salinity distribution in Chesapeake Bay (near Baltimore) carried out in 1996. Reid et al. (2000) described a way of extracting acoustic and environmental data to characterize and visualize individual fish schools. This is important in some regions where fish aggregation patterns play a more significant role in fish-stock assessment and exploitation. Finally, Mayer et al. (2001) described initial studies related to the transition from single- to multi-beam applications that aimed to introduce geomatics and 3D-visualization software to the assessment of fish stocks and enhance knowledge of pelagic fish schools.
All these references, however, were dealing with a single series of measurements. The technique described in this paper is intended for long-term and large-scale research on fish behaviour in relation to environmental characteristics of ecosystem using measurements taken over several years and a grid of stations. The type of visualizations shown here could be useful for direct ecological studies, based on acoustic semi-tomography, or to define sub-areas where numerical solutions can be looked for or verified. The method could quickly define areas of important correlations of the values and the gradients of the pairs of factors observed i.e. fish-volume density and temperature or oxygen distributions along a selected 3D geographic area as in Figure 3.
|
References |
|---|
|
TopIntroductionMaterials and methodsResults and discussionReferences |
|---|
-
Anon. (2001) Paint Shop Pro 7 and Animation Shop 3 – Reference Guide(Jasc-Software, Inc., Eden Prairie) pp. 1–505.
Barnes R.S.K. and Mann K.H. (1991) Fundamentals of Aquatic Ecology(Blackwell Science, Oxford) pp. 1–270.
Helfman G.S., Colette B.B., Facey D.E. (1997) The Diversity of Fishes(Blackwell Science, Oxford) pp. 1–528.
Horne J. K., Jech J. M., Brandt S. B. (1996) Spatial modelling of aquatic habitat from fish's perspective. Proceedings, Third International Conference/Workshop on Integrating GIS and Environmental Modeling(CD, Santa Fe, NM, Santa Barbara, CA).
Jech J.M. and Luo J. (2000) Digital echo visualization and information system (DEVIS) for processing spatially-explicit, fisheries acoustic data. Fisheries Research 47:115–124.[CrossRef][Web of Science]
Knudsen H.P. (1990) The Bergen echo integrator: an introduction. ICES Journal of Marine Science 47:167–174.
Mayer L., Yanchao L., Melvin G. (2001) 3D visualization for pelagic fisheries research and assessment. ICES Journal of Marine Science 10:1–10.
Orlowski A. (1990) Macrosounding as a new concept in the study of marine ecosystems. Colloque de Physique 51:269–72.
Orlowski A. (1998) Acoustic methods applied to fish environmental studies in the Baltic Sea. Fisheries Research 34:3227–237.[CrossRef][Web of Science]
Orlowski A. (1999) Acoustic studies of spatial gradients in the Baltic: implication for fish distribution. ICES Journal of Marine Science 56:561–570.
Orlowski A. (2000) Diel dynamic of acoustic measurements of Baltic fish. ICES Journal of Marine Science 57:1196–1203.
Orlowski A. (2003) Influence of thermal conditions on biomass of fish in the Polish EEZ. Fisheries Research 63:3367–377.[CrossRef][Web of Science]
Reid D., Scalabrin C., Petitgas P., Masse J., Aukland R., Carrera P., Georgakarakos S. (2000) Standard protocols for analysis of school-based data from echosounder surveys. Fisheries Research 47:125–136.[CrossRef][Web of Science]
Roman M.R., Holliday D.V., Sanford L.P. (2001) Temporal and spatial patterns of zooplankton in the Cheaspeake Bay turbidity maximum. Marine Ecological Progress Series 213:215–227.
Socha D.G., Watkins J.L., Brierley A.S. (1996) A visualization-based, post-processing system for the analysis of acoustic data. ICES Journal of Marine Science 53:2335–338.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Orlowski Experimental verification of the acoustic characteristics of the clupeoid diel cycle in the Baltic ICES J. Mar. Sci., January 1, 2005; 62(6): 1180 - 1190. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||