Fisheries Research 48 (2000) 157±166
Spatial structure of coastal pelagic schools descriptors in the Mediterranean Sea TaruÃb Bahri*, Pierre FreÂon I.R.D.-H.E.A., B.P. 5045, 34032 Montpellier Cedex 1, France Received 4 November 1998; received in revised form 14 September 1999; accepted 24 January 2000
Abstract Biomass estimates during acoustic surveys rely mostly on ®sh accessibility which in turn depends on the spatial distribution and structure of schools. In this paper, we ®rst investigate the spatial behaviour of schools through some of their descriptors in order to assess the interest of a study on the in¯uence of environmental factors on this behaviour. Morphological, energetic and positioning descriptors of schools were measured in the Catalan and Adriatic Seas during four acoustic surveys. The spatial structure of the descriptors was studied using geostatistical methods. The variograms were calculated on the averaged school descriptors within a sampling distance of 1 nautical mile (elementary sampling distance unit). Globally, most of the variograms are structured, depending on the type of descriptor and on the cruise. The results are discussed according to the topography of the two studied regions, the temporal variability existing in the data, the dominant species in the area and the possible biases due to the acoustic device used. # 2000 Elsevier Science B.V. All rights reserved. Keywords: Pelagic ®sh schools; Schooling behaviour; Spatial structure; Variogram; Acoustics
1. Introduction School organisation can be de®ned by the size, density and three-dimensional position and location of the school in the water column. This organisation determines the access to the schools by ®shermen, as it in¯uences detectability and catchability of ®sh. It also in¯uences biomass estimates during acoustic surveys. School characteristics (density, shape and vertical position) depend on many factors, including ®sh behaviour, physiology, biology, species and environment (FreÂon and Misund, 1999). Among these factors, the most commonly quoted are the following:
*
Corresponding author. Tel.: 33-4-67-41-94-00; fax: 33-4-67-41-94-30. E-mail address:
[email protected] (T. Bahri)
The diurnal cycle which usually induces a rise of the schools to the surface followed by a dispersion of the fish at sunset (e.g. Aoki, 1986; FreÂon et al., 1989, 1996; Ohshimo, 1996). A high variability in this behaviour due to in situ variation in light intensity and feeding behaviour has been observed; The physiological status or individual activity, such as reproduction, feeding or hunting which implies that the individuals get organised differently. Nùttestad et al. (1996) distinguished specific characteristics of the schools during the spawning season. At each stage (arrival in the area, looking for a favourable zone, spawning, feeding and leaving the spawning area), the schools display different features as far as density, shape and vertical position are concerned; Interactions between species. The presence of other species influences the behaviour, shape and position
0165-7836/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 5 - 7 8 3 6 ( 0 0 ) 0 0 1 7 6 - 4
158
T. Bahri, P. FreÂon / Fisheries Research 48 (2000) 157±166
of the schools, for trophic competition or prey/ predator relationships (Abrahams and Colgan, 1985; FreÂon et al., 1992; Parrish, 1992; Pitcher and Parrish, 1993; Soria, 1994; Masse et al., 1996); The introduction of a disruptive element, such as a the arrival of a boat causing an avoidance reaction that may modify the shape and position of the schools (e.g. Gerlotto and FreÂon, 1992); Environmental conditions are known to influence at least the vertical position of the schools, specially the presence of a pronounced thermocline (e.g. Radakov, 1973; Hara, 1985). A literature survey demonstrates a growing interest in the study of the spatial characteristics of ®sh populations. The spatial component is recognised as a major feature for understanding the interactions between schools. Several authors have discussed the strategies of the schools for occupying space in relation to the estimation of their biomass (Conan, 1988; Petitgas, 1993; Barange and Hampton, 1997; Swartzman, 1997; Gerlotto et al., 1999). In addition, spatial models can also be used to simulate the changes in abundance in relation to the spatial behaviour of schools (Pitcher, 1996). In this paper, two aspects of the spatial behaviour of the schools during acoustic surveys are investigated: the characteristics of the schools and their spatial distribution. The distribution and shape of the schools are known to be in¯uenced by many parameters. As shown above, some of these parameters have a medium scale in¯uence (i.e. environment) that should be re¯ected at the same scale in the spatial structure of the
schools descriptor, while other parameters have a too small scale in¯uence to be detectable during acoustic surveys (i.e. predator arrival). Therefore, it is not obvious whether the spatial variability of the school descriptors (morphological, position and density descriptors) is ruled by some spatial repartition laws that can be identi®ed, or if the descriptors appear randomly distributed. We examine the spatial distribution of several school descriptors to test the hypothesis of their spatial structuration (regionalisation). As far as we know, only one study (Patty, 1996) has been done on this direction. Testing this hypothesis is the necessary ®rst step prior to broader studies on the environment effect on school characteristics and to the optimisation of the design of school-based surveys. 2. Material and methods 2.1. Study areas and survey design The acoustic surveys took place in the Catalan (39± 418N/0±28E) and Adriatic Sea (438300 ±458300 N/ 128150 ±138300 E), near the estuaries of Ebra and Po rivers, respectively (Fig. 1), during the European project T-Echo (AIR1 CT92 0314). The surveys were all performed aboard the R/V Garcia Del Cid. In the Catalan Sea, two back to back coverages of the zone were performed during each of the two cruises. In the Adriatic Sea, a single coverage was performed during each of the two cruises. In total, there were four cruises, but six coverages or surveys (Table 1). The surveys were performed around the clock along
Fig. 1. Working areas during T-Echo project: Catalan Sea and Adriatic Sea.
T. Bahri, P. FreÂon / Fisheries Research 48 (2000) 157±166
159
Table 1 Date and place of the main surveys performed during T-Echo project Date
Survey
Zone
May 1994 September 1994 May 1995 September 1995
CAT1-1 and CAT1-2 ADR1 CAT2-1 and CAT2-2 ADR2
Catalan Sea Adriatic Sea Catalan Sea Adriatic Sea
parallel transects roughly perpendicular to the coastline, separated by 7 miles. Vessel speed was about 6± 7 knots. Due to the few number of schools observed during the night and beyond 100 m depth, only daily data (from sunrise to sunset) located within the 100 m depth were processed. 2.2. Material and school descriptors The schools were detected by vertical echo-sounding, using a 38 kHz Dual-Beam Biosonics model 102. It was connected to the external system, Ines-MoviesB (Weill et al., 1993), which identi®ed the schools and calculated a series of descriptors for each school. The integration was performed over the whole water column, except for two ``blind'' layers, one at the surface (®rst 5 m) and one at the bottom (1.5 m above the bottom). For each school, the software computes quantitative descriptors giving information on its geometry, position and acoustic density. Examples of these descriptors are given in Fig. 2. Qd: integrated back-scattered energy. This value is proportional to the total biomass of the school (relative value); Sv: volume reverberation index of the school, which gives an indication on the mean density of the school (dB); Length: maximum length of the school (m); MeanH: mean height of the school (m); MaxH: maximum height of the school (m); Elongation: school elongation (ratio length to maximal height); RO: roughness of the school: RO 2
ln
Perimeter=4 ln
Area
RO is close to 1 for the schools that display a very smooth outline shape and close to 2 for the very irregularly shaped schools;
Fig. 2. Example of descriptors measured by Ines-Movies-B giving the position and the geometry of the schools (MeanH and MaxH are the mean and maximum height, respectively; Min.Alt. and MinDepth are the minimum altitude and depth, respectively).
Perimeter: perimeter corrected from the beam angle error (see Weill et al., 1993 for details) (m); Area: area corrected by the beam angle effect (m2); Av.Depth: average depth of the school (m) ÿ Av:Alt: Bottom depth ÿ Min:Alt: 12 MeanH ; Min.Alt.: minimum altitude of the school, i.e. distance between the bottom and the lower limit of the school (m); Rel.Alt.: relative altitude of the school within the water column (%): Rel:Alt: 100
Min:Alt: MaxH=2 ; Depth
Min.Depth: minimum depth of the school, i.e. distance between the sea surface and the upper limit of the school (m). 2.3. Variograms The variograms measure the mean variability between two points x and xh as a function of distance h (Matheron, 1970): g
h 0:5EZ
x h ÿ Z
x2 where Z(x) is the value of the descriptor at the point
160
T. Bahri, P. FreÂon / Fisheries Research 48 (2000) 157±166
x and Z(xh) is the value of the descriptor at a distance h from x; g(h) is a function of the modulus and the orientation of the vector h and it indicates how different the values become, as the distance increases. Since our sampling is continuous, the detected schools do not correspond to speci®c sampling points. Therefore, we ®xed a sampling step (Elementary Sampling Distance Unit) of 1 nautical mile and we
counted the schools within each ESDU all along the route of the vessel. When no schools are detected, a missing value is assigned to all the descriptors. When schools are detected, we calculate the arithmetic average of the school descriptors within the sample. This method presents the advantage of homogenising the repartition of the samples. However, it does not separate ESDUs where no schools were detected from not-sampled ESDUs.
Fig. 3. Examples of experimental variograms showing the difference between the three categories adopted in our classification: (a) ``structured'' variograms; (b) ``weakly structured'' variograms; (c) ``not structured'' variograms.
T. Bahri, P. FreÂon / Fisheries Research 48 (2000) 157±166
Isotropic variograms were calculated, with a step of 1 mile. 2.4. Variograms classification The variograms are classi®ed in three categories, according to the quality of their structure: ``Structured'': the line increases quite regularly until a level from which it amply oscillates around an asymptotic value. The range can be short or long according to the survey or to the descriptor. Spherical or exponential models are fitted to the experimental variogram (Fig. 3a). ``Weakly structured'': a model can be fitted to the global aspect of the variogram, but with a poor quality and the choice of the model is not obvious. Moreover, some of these variograms present an unusual characteristic consisting in a drop of the variance below the nugget at the small distances, usually around 3 miles (Fig. 3b). ``Not structured'': variograms have a totally erratic aspect, they are purely random and the line oscillates around the level of variance from the small distances. In this category, we also classify the variograms that display atypical aspect, dome, wave or stairs shaped (Fig. 3c). In addition to the sorting of the variograms based on the ®tting of a model to the experimental variogram,
161
we performed a sorting, based on the percentage of variance due to the nugget (it varies between 0 and 85% in our data set). Variograms where the random component exceeded a pre-set value of 60% were considered to have an unstable spatial structure and were discarded from further analysis. 2.5. Species distribution During the surveys, the species were identi®ed by ®shing with a pelagic trawl, following two different sampling strategies. In the Catalan Sea, the sampling was not systematic but occurred in the zones presenting the highest acoustic densities. In the Adriatic Sea, the sampling was regular, following a grid that had been adopted in the 1970s for the collection of historical data. For each haul, we computed the species composition, the mean length and mean weight of the ®sh. 3. Results 3.1. Variograms classification Using the criteria of visual goodness of the ®t to a model and of the percentage of variance due to the nugget, 56% of the variograms have a discernible structure (Table 2). The average depth is structured
Table 2 Range (nautical miles) of the structured variograms of school descriptors according to the criteria of goodness of fit to a model and of the percentage of variance due to the nuggets (no data means no structured variogram) Descriptor Qd Sv Length MeanH MaxH Elongation RO Perimeter Area Av.Depth Min.Alt. Rel.Alt. Min.Depth Structured descriptors
ADR1
3 6 9 9 9 5 5 10 8
ADR2 8 9
CAT1-1 6 18
9 10 3 30 15 15 21 6
30 30
24 11
27
CAT1-2
14 7 3 3
CAT2-1 9 6
CAT2-2
27
1 4 4 2 3 4 2 4 2 6 4 2 5
5
43
27 10 18
15 9 16
27
12 3 3 7
6
8
5
Structured descriptors
27
162
T. Bahri, P. FreÂon / Fisheries Research 48 (2000) 157±166
for all the surveys, largely because of its link with a strongly structured feature: the bottom depth. The values of the coef®cient of correlation between the bottom depth and the average depth of the schools range from 0.77 to 0.92 according to the survey for a number of observations ranging from 120 to 2471. To a lesser extent, the same observation applies to the minimum depth (0.71
