2 SURVEY METHODOLOGY

General

The classic method of investigating the potential of fish stocks and the effect of fishing on them, for which the models of exploitation were developed, relies on the use of statistics of catch and effort and biological sampling from the commercial fishery. This method can be inexpensive, as the basic data may already be collected for other purposes, and because it may be based on a large number of observations from many vessels, the sample variance can be small. The method depends, however, on the existence of a fishery over a number of years and of significant intensity exploiting the main component of the stock, as well as a systematic collection of statistics and biological information including age determination of the fish.

Reliable and comprehensive data of this type are in fact far from being available for all major fisheries even today and in the 1970s these conditions were seldom found in the developing regions, where often, the only long-term fishery was artisanal and small-scale, limited to exploiting the shallow littoral part of the continental shelf. Data from the often shifting effort of long-ranging foreign fleets operating in those regions seldom provided the time-series needed for assessments.

A rough appraisal of resources in unexploited areas may be made by using general information on oceanographical characteristics and basic productivity of the waters and comparing that with similar information from areas with developed fisheries. Around 1970 this formed the basis for several estimates of the potentials of unfished regions in the tropics and sub-tropics (Gulland, 1970). This type of appraisal was, however, of little tactical value for the purposes of development and investment. For sound planning of fishery development the resource information should, in addition to estimates of yield potentials, include main features of life history, distribution, behaviour, catch rates, catchability, size, quality, etc., for each target species. For most resources these basic data can best be obtained from research vessel surveys using acoustic instruments combined with sample fishing to study small pelagic fish, and fishing with bottom trawls to study the demersal species.

The method of research vessel surveys has the disadvantage of being relatively expensive and requiring advanced technology. Historically, there has been a tendency in natural renewable resource sectors such as fisheries, to expect low research costs on a total sector value basis (Sætersdal, 1975), an attitude which may have contributed to management failures and other losses. The world fisheries community in general now considers the acquisition of detailed resource data to be a basic condition for industrial fisheries, and research vessel surveys, although costly, form part of most major programmes of resources research and management. In addition to providing information on unexploited stocks, investigations with research vessels have proved indispensable also in more advanced research on those stocks which are being monitored through fishery statistics and biological sampling, for example, by providing data on recruitment and growth, independent estimates of biomass or basic biological information needed to study species interactions and ecosystems.

The main objectives of the DR. FRIDTJOF NANSEN programme were, however, especially in the first 10 years, of a basic and exploratory nature with emphasis on general descriptions of the composition, distribution and abundance of the resources and overviews of their environment. With these broad objectives the special technical problems of the methodology such as accuracy and precision of biomass estimates, were perhaps not as critical as in more advanced stock assessment studies. However, since these first surveys often described the resources at a stage of low exploitation, which is of value for later reference, there is an interest in reviewing the biomass estimates and as far as possible checking and evaluating their accuracy and reliability.

The investigation and assessment of stocks of small pelagic fish was maintained as a main task in most assignments. The description which follows of the acoustic methods used in these investigations is somewhat detailed because of the importance of this task in the programme and also because the acoustic survey techniques and instrumentation went through considerable developments during the life-span of the DR. FRIDTJOF NANSEN.

In applying the acoustic technique, fish observed in layers close to the bottom were classified and recorded as demersal or semi-demersal fish and identified by sampling with bottom trawls or pelagic trawls just above the bottom. In most of the assignments from 1978 onwards special studies of demersal resources by bottom trawl surveys were included as an additional activity. The problems encountered and experience gained in using this method are discussed below.

2.1 BOTTOM TRAWL SURVEYS

The distribution, composition and abundance of the demersal resources in a defined area of the ocean may be studied by sampling with a bottom trawl. To give valid estimates of precision the sampling should be random with respect to the distribution of the fish. Stratified random sampling meets needs of survey efficiency and is generally used. When dealing with multispecies target stocks, as was usually the case in this programme, stratification by depth ranges would seem most practical.

When combined with an acoustic survey the most convenient positioning of the trawl stations is at fixed distances along a rectangular grid pattern with adjustment, within depth ranges for the depth stratification. This results in a semi-random positioning of the trawl hauls. Although not randomly selected in the strict sense, such stations will almost certainly be randomly positioned in relation to the distribution of the fish as discussed by Saville (1977).

The precision of the abundance estimates obtained from trawl surveys varies greatly with the characteristics of distribution of the species and with sampling effort. Confidence limits from trawl surveys in the ICNAF area were quoted to range from ±25% to ±50% at the level of effort usually deployed in such programmes (Doubleday, 1981). Examples from the DR. FRIDTJOF NANSEN surveys were: ±28% from 105 fishing stations of several surveys of the Omani shelf comprising all demersal fish (Strømme, 1986), while typical confidence limits for biomass estimates of the main species in the Angola surveys ranged from ±32% to ±150% (Strømme and Sætersdal, 1991), and for Namibian hake from ±11% to ±16% within three main regions, (IMR, 1993b). The latter survey was especially planned for hake with a semi-random distribution of fishing stations.

Although the resulting estimate of variance may not satisfy all demands of statistical theory, the results may be interpreted to indicate that the effort of investigation has been at a reasonable level. For species with a highly contagious distribution such as Namibian hakes the sampling effort was increased in areas of high fish density and the data processed following post-stratification by densities.

The existence of possible systematic errors in the method may not be so important when dealing with estimates of relative abundance from a series of similar surveys. Changes in abundance in a given area may thus be observed and comparisons made between relative fish densities in different areas. For absolute abundance estimates, however, all possible sources of bias must be considered. Absolute densities in the survey area are estimated from the catch, the area swept by the trawl and the catchability coefficient, the proportion of fish in the area fished to that retained by the gear. The bottom trawl, designed as shown in Figure 2.1, was especially selected for sampling a wide range of targets: a high-opening small-meshed shrimp-cum-fish trawl with a small meshed inner lining of the cod-end. This gear was used throughout the programme. Later instrumented tows indicated a headline hight of approximately 6 m and a distance between wing tips of 18–19 m.

Figure 2.1 Design of the bottom trawl used throughout the programme

This type was chosen in the expectation that it would capture a wide size range of fish and crustaceans with little escapement through the small-meshed forward parts. The catchability of this gear for different species is, however, likely to have varied considerably especially when it was used with bobbins on the footrope as often had to be the case on unknown or rough bottoms. Indications from comparisons with commercial catch rates and comparative fishing trials were that the trawl used by the DR. FRIDTJOF NANSEN had a low efficiency for truly benthic shrimp and fish such as flatfish and monkfish (Lophius sp.). A tickler chain was at times used to improve the catch of these benthic targets.

The trawled distance was determined from the ship's log adjusted for currents as observed by navigational instruments. From 1991 on the distance was estimated directly from a GPS navigator giving a precision within 2–4% of the towed distance. The starting point of the tow was, however, based on the navigator's estimate of the time the gear made bottom contact after shooting. Later instrumented observations have shown that this estimate was reasonably accurate at depths down to 150–200 m, but that it was probably biased in hauls at the continental slope at depths of 300 m and more, resulting in overestimates of the swept-area by perhaps 15–20%.

The tow direction with respect to the current varied, but was most often with the current. The geometry of the trawl was observed with SCANMAR™ instruments under survey conditions and the direction has probably not been a source of significant bias.

The most important source of bias in trawl surveys is, however, related to the assumptions concerning catchability. Dickson (1993 a and b) found that estimates of gear efficiency varied considerably both between species and between different sizes of the same species and also by the type of ground gear used. A simplistic assumption was made that the same proportion of all fish in the path of the trawl between the wing tips, irrespective of species or size, would be caught and retained. Therefore, in most surveys the coefficient of catchability (q) was assumed to be equal to 1 for the swept-area between the wing tips, which had a mean width of 18.5 m. This does not necessarily assume that there is no avoidance of gear or vessel, but it implies that avoidance and escapement are approximately balanced by the herding of fish by the trawl bridles and sweep lines.

Trawl survey estimates may also be affected by other important sources of bias. Most demersal fish will tend to rise from the bottom at night and thus be above the headrope of the trawl. But even if fishing is limited to day-time as in most of these surveys, mid-water occurrence may still cause a bias since a number of species classified as demersal fish (snappers, hakes, various sea breams, grunts, silver smelts and others) are commonly found in mid-water and above the headrope of a bottom trawl also during the day. In some of the surveys this source of bias was overcome by making an acoustic estimate of this pelagic component of the demersal fish to be added to the stock estimated by the “swept-area method”.

In most assignments in the Indian Ocean demersal or semi-demersal fish, seen in layers or as single fish near the bottom and in the lower part of the water column were commonly recorded by the acoustic system, and identified and estimated as a special group. These observations included night-time recordings where the demersal fish were rising. The biomass estimates of this group at times exceeded those estimated from the bottom trawl surveys of the same shelf areas (Stømme and Sætersdal, 1982). The group included ponyfish and hairtails which are semi-pelagic and which may in other surveys have been classified as belonging to the pelagic community.

2.2 ACOUSTIC SURVEY METHODOLOGY

In order to be able to follow the developments in the acoustic instruments and methods used on the DR. FRIDTJOF NANSEN and to understand the adjustments that have been made in many cases to the biomass estimates based on corrections induced by later findings, it is desirable to give first a short resumé of the acoustic survey technique as applied on the DR. FRIDTJOF NANSEN. This is partly based on a note prepared by K. Olsen for an Evaluation Report (Hallenstvedt et al.,1983).

The use of vertical ranging echosounders and horizontal ranging sonars in fisheries research has a 30–40 year history. Acoustic methods were promoted through international symposia, training courses and user's manuals especially in the period 1960–80 and they now form an indispensable tool in the study of the distribution and abundance of pelagic and semi-pelagic resources. IMR has participated from the outset in the development of acoustic techniques and their application, and the NANSEN programme has benefitted from the close association with an active group of specialists at IMR.

Echosounders combined with echo integrators have proved to be particularly useful for investigation of small pelagic species such as herring, sardine, anchovy, capelin and other species with similar distributions in mid-water schools and layers. Surface schools can be observed with horizontal ranging sonars, but this method poses more problems of quantification than echosounding. Fish situated very close to the bottom may escape acoustic observation, but many demersal species - during feeding, migration and spawning - often have a major part of their biomass in mid-water and combined assessment through bottom trawl and acoustic surveys has been used for such stocks. Acoustic surveying is not suitable for fast-swimming large pelagic fish such as tunas.

The echo integration method has undergone several drastic changes since it was first introduced in the late sixties. The first system consisted of a scientific sounder SIMRAD™ EKS coupled to an analog integrator SIMRAD™ QM, which produced an output in the form of a graph measured in mm. In the next generation of instruments, the SIMRAD™ EK400 sounder coupled to the digital SIMRAD™ QD integrator, the system of using an output index expressed in mm was still maintained at first, but later it was replaced by a new type of measurement, that is independent of the performance of the instruments, the so-called fish constant. In 1991 the SIMRAD™ EK500 system was introduced, an echosounder/integrator system in one unit. All these changes in instrumentation have led to considerable improvements in the acoustic estimation procedures. Findings in later years with better instruments have often shed doubts over the validity of earlier survey results obtained with more primitive instruments.

The original echo integration method is based on the assumption that the recorded echo intensity is proportional to the fish density, when the transmission losses of the received echo are compensated for through the Time Varied Gain (TVG) applied in scientific echosounders. The received echoes are converted into voltage signals, which are squared and integrated by a separate instrument, the echo integrator. The signals are also recorded on paper, which is called an echogram. The output of the echo integrator is considered to be proportional to the density of fish along the course track, as follows:

ρ = C × M

where ρ is fish density, M is the output of the echo integrator (expressed in mm with the QM and QD integrators), while C is a conversion factor that depends on the performance and settings of the instruments and on the strength of the echo that is returned by the fish (target strength or TS).

The sonar equation forms the physical basis of the method:

EL = SL + TS - 2TL

Where EL (echo level) is the level of the reflected sound, SL (source level) is the level of the incident sound, TS is target strength, and 2 TL is two-way transmission loss due to spreading and absorption. Knowledge of the source level of the system and the target strength of the fish is thus critical for using the observed echo levels for fish abundance estimation.

The integrator output generated by a given echo level depends on the voltage response (VR) of the receiving system, gain settings and the Time Varied Gain (compensation for transmission loss and range dependent beam area).

The source level (SL) and the voltage response (VR) are monitored through acoustic calibration. Other control measurements required for proper monitoring of the system are band width, pulse length, frequency, beam width, time varied gain function (TVG) and integrator performance.

Calibration of SL, VR, TVG and pulse length should ideally take place before and after each major cruise.

From the applied value of C and available instrument records the applied target strength can be checked. The analytical relationship between C expressed in tonnes/nmi²/mm and the instrument characteristics can be written

C = 3430 antilog 0.1[-(SL + VR) + 20logR + 2αR - 10logcτ/2 - 10logψ - A + V₀ - TS/kg]

where:

A log-linear relationship has been assumed between TS/kg and fish length (L):

TS/kg = -10logL - B dB

Based on a length/weight relationship W = a L³, where a is the so-called condition factor this is equivalent to assuming the relationship between target strength and fish length for individual fish as:

TS = 20logL - A dB

With information on the condition factor (a), the transformation between the target strength of 1 kg of fish and of each of n equal-sized fishes weighing a total of 1 kg will be:

TS/kg = TS + 10log n

where TS is the target strength of one fish of that length.

The TS is defined by reference to the backscattering cross section of a target as follows:

Among the methods of studying target strength of fish mention should be made of cage experiments which in the 1970s represented an early advance. With this method the acoustic system was calibrated against a known density of caged fish. Cage calibrations could provide integrator conversion factors, C, without the need of instrument performance data, an important advantage considering the low precision of acoustic calibration at that time.

The split-beam technique, through which the position of individual fish within the beam is determined, has greatly improved facilities for in situ observations of target strength, although with hull-mounted transducers the opportunities of observing single fish targets which may be representatively sampled with fishing gear are rare, a situation confirmed by various efforts for such studies in the DR. FRIDTJOF NANSEN surveys.

The integrator output obtained over a certain sailed distance (usually 5 nautical miles) represents the total density of all species and size groups contributing to the output. The outputs are first compared with the corresponding echograms, and with corresponding catches of targets with pelagic or bottom trawl hauls made along the tracks, before being plotted on the survey track. Contour lines of equal densities are then drawn, which allows a multiplication of echo density by a certain area usually expressed in square nautical miles. Such maps can be drawn for all species combined or separately by species if there is sufficient information to do this. Subsequently, the areas occupied by the fish can be summed on a regional basis and provide estimates of the standing biomass expressed in tonnes.

In order to come up with reliable estimates based on this method it is necessary to know and check the performance of the instruments, to know as much as possible of the daily and seasonal behaviour of the target species and to take very frequent samples of the target species by fishing with pelagic and bottom trawls.

Examples of specific findings and problems encountered during the surveys with the DR. FRIDTJOF NANSEN will be presented in Section 2.4.

2.3 REVIEW OF THE ACOUSTIC INSTRUMENTATION, CONVERSION FACTORS USED AND ESTIMATES OF APPLIED LEVELS OF TARGET STRENGTH

Three generations of SIMRAD™ echo integration systems were used in succession over the programme period, each representing considerable advances made in the instrument component of the acoustic survey technique (Table 2.1). Records of the state and use of the instruments were entered in the logbooks and reports of the electronic engineers of the programme, and are still available.

Table 2.1 Echo integration systems used

The EKS instruments equipped with TVG belong to the first generation of “scientific sounders” for fishery research purposes. The performance of the EKS 38 was increased significantly with the introduction of ceramic transducers in 1979. They were used with the QM analog integrators. Imperfections of the system included: a relatively limited dynamic range of both the echosounder receiver and the QM integrator thus causing saturation in cases of dense fish recordings; an unsatisfactory bottom discrimination function resulting in the inclusion of dense fish aggregations near the bottom into the bottom echo; and a low dynamic range of the paper recorder which complicated the discrimination of recordings of fish in plankton layers.

All of these imperfections would have tended towards a bias of underestimation of fish abundance.

Table 2.2 shows the history of calibrations of the EKS 38 kHz sounder with hydrophones and later solid spheres. The first survey plan prescribed acoustic calibration once a year with dry measurement monitoring in between, but experience soon demonstrated the need for more frequent checks. The first transmitter was a radio value type, which proved unstable at high performance, necessitating a reduction of transmitting effect in mid 1975 for a more stable performance.

A more stable transistorized transmitter was installed in 1979 together with a ceramic transducer, which was expected to be 3 dB more efficient both in transmitting and receiving functions, resulting in a 6 dB increase in performance of the system.

The accuracy of measuring the performance of the systems increased greatly when the method of calibration with a solid sphere was introduced. Since the solid sphere calibrations gave consistent levels of 138 to 139 dB through 1980–83 it seems reasonable to conclude that this represented the true stable level of the system with the ceramic transducer. Assuming a 6 dB increase in efficiency after the installation of the ceramic transducer, it can be deduced that at the beginning the system had a performance of 133 dB with the nickel transducer.

Table 2.2 Record of in situ calibrations of the EKS 38 kHz sounder system, 1974–83

The original TVG function was replaced by a new improved design in April 1980. Measurements of the TVG of the EKS 38 sounder in 1979 and in 1981 respectively revealed a substantial difference, 5 to 6 dB in the resulting gain curves (Aglen et al., 1982), thought to result from a special adjustment to absorption in tropical waters of the original TVG. However, this could not be confirmed and it must be assumed that the 5 to 6 dB higher gain was the effect of drift of component performance prior to 1979. The change in gain is confirmed by differences in the results of intercalibrations with the 120 kHz system in September 1979 (EKS 120 = 0.8 * EKS 38, Nakken and Sann Aung, 1980) and in April/May 1980 (EKS 120 = 3.9 * EKS 38, Sætersdal et al., 1980). It is assumed that the TVG change took place in 1978. This performance level, 144–145 dB is also indicated by the hydrophone calibration of April 1979, a measurement undertaken during a home refit visit of the vessel.

During a period in 1980 the receiver gain was reduced from 85 to 79 dB in an attempt to avoid problems experienced at the higher setting of distinguishing between recordings of fish and of spurious sources recorded.

The factors used to convert mean echo levels to biomass, the conversion factor C, which has been reported for each survey, can be used together with the instrument performance for the respective period listed in Table 2.2 to estimate the mean value of the target strength actually applied. The data are shown by assignments in Table 2.3.

The C value used in 1975–77 (C = 10.5) was based on an intercalibration with the R/V RASTRELLIGER of the UNDP/FAO Project IND/69/593 “Survey of the Pelagic resources on the Southwest Coast of India”. RASTRELLIGER's constant was derived from calibrations of live fish in cages (Nakken, 1974). The target strength level in these calibrations was -29 dB/kg for catfish of about 17 cm.

The C value of 10.5 was adjusted to fish size by bringing along the 17 cm fish size from the RASTRELLIGER cage calibration in the form of 0.6 * L.

Table 2.3 Echosounders used for integration, estimated performance, conversion factors used and estimates of applied TS-levels by assignments (TS/kg for 17 cm fish), 1975 to April 1984

The next generation of SIMRAD™ echosounders for scientific use, the EK400 series, was brought into use in April 1984. The main improvement of this system was the digitalization of the integration process in the QD integrators, which reduced the problem of saturation at high signal levels experienced with the analog QM integrators. To facilitate in situ target strength measurements a split-beam transducer was also constructed for this system.

The EK400 sounder was used with its normal transducer from April 1984 to October 1985 (Table 2.4). The performance was repeatedly measured at 140.8 dB and with a conversion factor of 0.94 this corresponded to -33.2 dB/kg for fish of 17 cm total length.

From November 1985 onwards the ES split-beam transducer was used and the target strength which corresponded to the conversion factors used then was estimated at -33.4 dB/kg. Frequent calibrations of this system showed unexplained variations in its performance of the order of 1 to 2 dB. The conversion factors were, however, adjusted accordingly and it may be assumed that the level of target strength used was -32 to -33 dB/kg for 17 cm fish.

From October 1986 onwards the instrument constant was adjusted to give integrator readings in units of reflecting (backscattering) surface proportional to m²/nmi². This had the advantage of providing comparable acoustic indices independent of the state of the instrument. A “fish constant” only depending on the choice of target strength and the condition factor was then used for conversion to biomass estimates.

A target strength of TS = 20 log L-72 dB was used, corresponding to TS/kg (17 cm fish) = -34 dB to -35 dB depending on the condition factor used.

Table 2.4 Instruments used for integration, performance, conversion factors and applied TS-levels by assignments from May 1984 to June 1993

For multispecies estimates the condition factor applied from 1986 to 1991 showed that the TS level corresponded to -35 dB. From then on condition factors for the main species were estimated during the surveys and included in the biomass estimates. The result was that for pelagic fish TS = 20log L-72 dB was maintained, while for special surveys of demersal fish TS = 20log L-68 dB was used.

The EK500 system was brought into use in 1991. Its greatly improved dynamic range solved the problem of saturation of the previous systems at high density levels. Frequent calibrations showed very little drift, so this system gave reduced bias and improved precision. It may be appropriate to note that several of the most likely systematic errors of acoustic surveying, such as vessel avoidance by schools, and reverberation loss in dense schools, will still tend towards underestimation of the biomass.

Data logging and processing, the NAN-SIS package

The scientific observations and data were acquired, logged, processed and analysed through a set of work systems which may be classified as follows:

The origin and flow of data in these systems may be described as follows:

In the acoustic system, the observations from the echosounders were recorded as diagrams, while the backscattering from mid-water targets was at the same time quantified by the integrators. These data were then subjected to an evaluation process where the targets were classified on the basis of the characteristics of the diagrams, their contribution to the integrator output and information from the fishing system. The output was recorded in the acoustic log by groups of species or by species. Data on bottom depth and type may also have been logged.

The information flow through the fishing system provided data on catches and catch rates by species or species groups, which besides being used for abundance estimation and identification of acoustic targets also provided information on catchability and catch rates. Sampling of catches in addition provided important biological data on key species. Representative sampling of catches for species and size compositions posed special problems and the procedures adopted for this purpose in the programme are described by Strømme (1992).

The hydrographical system provided information on the ocean environment. These data were at times supplemented by observations of surface currents from the navigational system.

The flow of the information in the various systems is shown in Figure 2.2. The processed outputs were then evaluated and formed the basis for the conclusions of the survey work. All original logs were as a rule preserved and stored at IMR for permanent access.

Figure 2.2 Main data flow

The volume of survey data collected increased greatly when, after the first few years of mainly pelagic investigations, the objectives of the assignments were expanded to include studies of demersal fish by bottom trawl surveys. The processing and analysis of the data then became very time-consuming tasks and computerization became necessary. Its purpose was also to facilitate multiple access to the data. In about 1981 a comprehensive system for logging and analysis of survey data was started and this was further developed into the NANSIS software package (Strømme, 1992). It is a Survey Information System for logging, editing and analysis of scientific trawl survey data (trawl-catch data and length-frequency data). It provides summaries of user-selected sub-sets of catch and size data, defined by geographical sector, species or other groupings, depth, gear type, etc. Swept-area calculations can be made for data grouped by user-defined limits.

The multitude of species in tropical waters is handled through a mnemonic species code system. These species codes are converted into scientific or local names using project-based species name catalogues as well as a global catalogue.

NAN-SIS was originally intended only for the DR. FRIDTJOF NANSEN surveys, and was made available to the counterpart co-operating institutions. The program is, however, general and can be used for other trawl fishery resources surveys.

The published version of NAN-SIS does not include the programs for the logging, editing and analysis of acoustic data from the SIMRAD™ QD integrator and the SIMRAD™ EK500 system which were developed and used for the specific equipment of the DR. FRIDTJOF NANSEN.

Nearly all the trawl survey data from 1981 were stored in computerized files. Data collected prior to the development of NAN-SIS were recently entered in this database. The data are accessible for analysis through the NAN-SIS package at the Institute of Marine Research in Bergen, Norway.

2.4 OBSERVATIONS REGARDING ACOUSTIC SURVEY TECHNIQUES AND TACTICS AS APPLIED ON THE DR. FRIDTJOF NANSEN

Identifying fish species from target echoes cannot yet solely be done on an acoustic basis. Studies of the characteristics of echo records and related measurements of signal strength offer, on the other hand, a wide range of clues to the nature of the targets, and may often also permit estimation of their individual size. For a positive identification, however, reliance is placed on sampling by fishing with gears adjusted to the probable type and size of the target organisms, bearing in mind processes of avoidance and selectivity.

The plan for the first Arabian Sea survey stipulated that fishing for species identification and sampling should be undertaken “whenever the character of the fish recordings changed”. This proved in practice perhaps a not very useful directive and instead a simple rule of experience was developed that all targets of any substance should be sampled by fishing for species and size compositions. The incidence of fishing was found to be too low in the first surveys and therefore more time was allowed for fishing in later surveys.

Most small pelagic fish can be caught with a mid-water trawl guided by a netsonde. Gear avoidance of larger-sized fast-swimming fish (sardinellas, mackerels and horse mackerels) was, however, often experienced, especially when these were found in waters of higher temperatures. This was probably related to the limited size of the mid-water trawl and relatively low towing speed of the DR.FRIDTJOF NANSEN and was most pronounced in daytime schools. Catchability was also low for fish in the surface layer when the mid-water trawl was towed with short warps in the wake of the ship. Raising the headline by placing floats on the headline, extending warps and curved towing greatly improved the success of such fishing operations. To sample fish in mid-water in very shallow areas, the bottom trawl was often used with floats.

The bottom trawl with an approximate hight of the headline of 6 m, was used in a normal way to identify and sample fish aggregations expected to be less than 6 m from the bottom.

One survey technique, which was developed with experience included returning to an area with low catchability for sampling under different conditions, usually by taking advantage of the generally higher catchability during nighttime.

An important feature of the practice of acoustic survey methodology as developed by IMR is scrutinizing echograms, in conjunction with the outputs of the echo integrators and data from fishing. The purpose of this exercise is an identification of targets at species-group or species level and an allocation of their contribution to the integrator values. This is done on a daily basis whilst all main survey events are fresh in the minds of the engineers and biologists involved in the continuous monitoring of the instruments and in the fish sampling work.

Many aspects of acoustic surveying techniques need to be adjusted to the target stocks and must in this sense be developed through experience. In particular, account must be taken of existing patterns in the behaviour and distribution of the type of target fish, e.g., related to spawning or feeding cycles when deciding on the design and coverage of the surveys. For the DR. FRIDTJOF NANSEN surveys this basic knowledge had to be acquired for the often virgin areas and stocks covered.

Diurnal changes in behaviour pattern were often observed, such as surface schooling in daytime and the forming of scattering layers at night (except at full-moon periods). The day was then often used for mapping areas with schools while the school areas were resurveyed at night for abundance estimation using echo integration when the fish was found in dispersed layers. Sardines, sardinellas and other small pelagics at low latitudes could diurnally (or at other intervals) migrate towards and from the coast, a behaviour which at times brought parts of the fish biomass into shallow waters where the vessel could not navigate.

At times small pelagic fish in an area would disappear from the recordings of the acoustic instruments after nightfall, but could then be observed, by bioluminescence or in the lights of the vessel, swimming in small dispersed schools in the surface layer.

Sampling and mapping

The overall distribution pattern of a stock was often found to consist of a number of smaller areas of high abundance, “fish field areas”, separated by areas of zero or low fish density thus forming highly contagious distributions. Occasionally the survey data would, however, include a single or few highly dispersed observations of very high fish density in an otherwise empty area, representing an unknown incidence. Such exceptional observations were usually discarded during subsequent analyses.

For stocks with highly contagious distributions adaptive survey designs were applied. Relying on experience the basic grid net of vessel transects was designed so that it was dense enough to find the larger areas of fish abundance and a denser coverage adopted when an area with fish was encountered. This ensured an adjustment of survey intensity to fish density.

The main procedure used for analysis of the acoustic data was recording fish density indices by species or species groups by chart, with post-stratification obtained through contouring of density levels. This procedure is highly practical for comparison with environmental and fishing data, for life history studies and for generally describing distributional characteristics, but it makes estimates of the sampling variance by classic statistical methods unreliable. This approach to survey design and data processing was, however, maintained throughout because it was felt to represent the best use of costly survey effort. In consideration of the errors in acoustic assessments, the scientists applying the method seem historically to have been more concerned with the practical tasks of reducing systematic errors than with estimating precision, perhaps due to a judgement on their part of the priorities involved.

In order to obtain some information on survey precision for “tropical stocks” experiments were undertaken within the programme by repeating surveys of well-limited areas of fish aggregations within a short time period (Strømme and Sætersdal, 1987). The tests included Namibian and Moroccan sardines as well as sardinellas off Senegal. The final biomass estimates varied within 15–20%, a result similar to those of other trials of this kind (Simmonds et al., 1992).

Precision in biomass estimates is related to survey effort when the technique and method in general meet set standards. Aglen (1989) found an empirical relationship between the coefficient of the variation (CV) and the degree of coverage (d), estimated from repeated and partial surveys, where degree of coverage is defined as “distance sailed relative to the square root of the area investigated”. The larger d, the smaller the CV, although this also depended on the type of fish distribution. A common degree of coverage (value of d) for stock assessment surveys is of the order of 10 which gives coefficients of variation ranging from 0.1 to 0.4. For this degree of coverage the precision gained by a moderate increase in effort is small (Aglen, 1989). This measure of the degree of coverage will be presented for the vessel's surveys targeted on coastal small pelagic fish, in order to give an impression of the sufficiency of the survey effort.

Target strength issues

The advances in acoustic instrumentation and techniques, made during the programme period, eliminated or reduced some important sources of systematic errors. Through the sequence of generations of scientific echosounders used (SIMRAD™ EKS, EK400 and EK500), the dynamic range and stability of the systems have been greatly improved, for example solving the early problems of signal saturation. These developments have been discussed in Section 2.3.

There still exist, however, important sources of possible systematic errors: use of inaccurate target strength and the partly related avoidance reactions by fish; signal loss by attenuation in dense schools; and others. The problem of choice of target strength was discussed earlier.

In subtropical and tropical seas, backscattering of sound from non-fish targets in mid-water is at times both widespread and intensive, and creates problems for the discrimination of fish targets especially for fish found in dispersed layers. The origin of this backscattering is difficult to verify, but because of its passive appearance in echograms and as widespread distribution it is assumed to be planktonic. Observations during surveys in Sri Lanka and Myanmar showed that discrimination of fish targets was facilitated by using a higher frequency. A special study was made comparing backscattering of plankton and of fish in simultaneous operations of 38 kHz and 120 kHz systems (Sætersdal et al., 1983). Many observations from surveys in Myanmar and off West Africa confirmed a frequency-dependent backscattering with a mean of 4 dB higher volume backscattering coefficient from fish targets and from 2–7 dB lower values for “plankton” targets at 120 kHz compared with 38 kHz.

Clay and Medwin (1977) identify siphonophores as the source of “planktonic” recordings of echosounding which, with their gas-filled pneumatophores or released gas bubbles, have a profound sound scattering effect due to resonance. Estimated resonance frequencies covered 38 kHz, but did not reach 120 kHz; this may explain the observed higher volume backscattering coefficient of plankton at the lower frequency.

Experimental observations also seemed to confirm a higher target strength of fish at higher frequencies, while later in situ measurements have only shown a small difference of about 1 dB between the two frequencies (Ona, personal communication). The difference between frequencies in backscattering from the tropical type of “plankton” seems real and has been confirmed in a general way in other surveys.

For a proper assessment it is necessary to distinguish fish targets from plankton. The decrease in backscattering of fish targets and the increase of that of plankton targets at 38 kHz, reduces the possibilities of separating both types of targets. Therefore, in some areas the 120 kHz system was used, in shallow shelf areas, which could be covered with the limited range of about 100 m of this system (see Table 2.3). In other areas the 120 kHz system was mainly used as an aid in the discrimination of fish traces recorded by the 38 kHz system.

The first-generation echosounder, the SIMRAD™ EKS was less stable than the later models and its maintenance and control represented a technical challenge, especially under tropical conditions. A special problem was the recommended procedure for acoustic calibration at that time which involved the use of a test hydrophone lined up on the acoustic axis in a rig underneath the transducer. This method was found unsatisfactory by IMR and other users because it showed inexplainable variability of instrument performance. A procedure using standard targets in the form of metal spheres with known target strength was developed (Foote et al., 1987) against which the performance of the system, including the integrator could be calibrated. This method was adopted from late 1980 onwards from which calibrations are assumed to have had a high accuracy.

The general state of knowledge has, however, increased considerably since the 1970s. Accumulated data indicate a difference in backscattering cross section between physostome fish represented by clupeoids, and physoclist fish represented by gadoids and a large portion of other fish. Foote (1987) reviewing data on target strength measurements of fish found that the relationships best approximating the many estimates obtained through measurements and by theoretical computations based on swim bladder form are: TS. = 20log L - 67.4 dB for physoclist; and TS = 20log L - 71.9 dB for physostome fish. That this type of relationship between target strength and fish size seems to give the best fit to observed data is explained by the backscattering cross section being related to the square of the fish length.

The observed within-species variation in scattering cross-sections is within each group of the same order of magnitude as the reported between-species differences. The within-species variation seems to depend on such factors as behaviour, recent vertical migration and fat content. In particular, swimming behaviour, by changing the orientation of the fish relative to the acoustic axis, causes great variations in target strength.

Assuming that the within-species variation in target strength is related to fish behaviour and fish condition at the time of the survey and to the particular survey method used, this variation need not have any effect on the precision of the estimated relative abundance indices in a series of similar surveys. In order to improve the accuracy of the estimate of true abundance, however, observations of the actual in situ target strength of the fish during the survey are required.

Through the life of the programme the applied mean TS levels reflect largely views held at the time of the survey and vary accordingly. The first TS values used derived from cage calibrations and were at a level of TS/kg = -29 to -30 dB for 17 cm fish. Assuming a linear relationship between TS/kg and fish length and using a condition factor of 0.8–0.9 and proportionality between weight and length cubed, this corresponds to TS = 20 log L - 68 dB. From 1981 on, this level was thought to be too high and the conversion factors were adjusted to the relationship

W = a L³

TS/kg = -10logL - 22 dB, partly based on a similar relationship used by FAO (Aglen et al., 1982). This gives -34 dB/kg for 17 cm fish and corresponds to approximately 20 log L - 72 dB which is about 4 dB lower and results in more than a doubling of the biomass estimate. This has been the intended TS level in all subsequent surveys, but the actually applied values have at times differed somewhat from this level because of uncertainties regarding the instrument performance.

In reviewing the biomass estimates over the entire period there is a need for adjustment to a uniform TS level for time-series of surveys where the level has varied and where the level used has deviated significantly from the level intended. Since the fish in the sea are a mixture of physostome and physoclist species, use has been made for such adjustments of TS = 20 log L - 70 dB which corresponds to a TS/kg = -32 dB for fish of 17 cm total length.

It should be recalled, however, that the target strength represents only one of the likely sources of systematic error in the biomass estimates. An overestimate caused by the use of a too low target strength may have been balanced by an underestimate caused by instrument saturation, a bias likely to have affected the results from the first EKS scientific sounder period and to some extent the results of the EK400. It would thus seem most important to adjust for use of TS levels that are too high.

Acoustic distribution charts

A number of charts in this review show fish distribution by levels of density based on acoustic data: scattered, dense, very dense. Reference is made to ranges of integrator levels used for density levels. It should be noted that these levels are not directly comparable through the period of operation of the vessel as they refer to different instrument systems. The ranges chosen were also to some extent adjusted to fish abundance in the survey area and were for instance different in Northwest Africa and in the Eastern Central Pacific. The purpose of the charts is to show the main patterns of fish distribution and within regions, the depth of shading, which has been used in the charts, should demonstrate these patterns.

2.5 ENVIRONMENTAL STUDIES

The available bathymetric charts for some of the survey areas were found to have inaccuracies which at times were substantial and of significance for the survey work. Based on the greatly increased accuracy of satellite navigation simple plots of echosounder depth at intervals along the cruise tracks were in such cases used to prepare more correct bathymetric charts of the main shelf and slope areas. These were used to estimate the areas within depth contours over the shelf and slope and for purposes of survey design.

The type of bottom, with special reference to its suitability for bottom trawling - smooth, rough or steep, was observed on the basis of the echograms, and charts were prepared of these characters for all main surveys.

In addition to its significance for sampling, the type of bottom substrate has also faunistic implications, influencing the composition and distribution of demersal fish. Ordinary echo recording provides some information on bottom substrate, but usually only enables distinction between hard rocky or coral bottom and soft muddy bottom. Attempts were made in surveys off Angola and Namibia to calibrate an early version of the ROXANN™ acoustic system for bottom substrate discrimination against the bottom conditions in these regions by a grab sampling programme. These attempts met with only limited success, but there is no doubt that systematic recording of this character would represent important additional information on the environment of the demersal fishery resources.

Continuous records with a thermograph were made of the sea temperature at 5 m depth. Casts with Nansen bottles were used to record temperature and take water samples at standard depths. Salinity was analysed on board by salinometer and usually also dissolved oxygen by the Winkler method. A CTD sonde was used from 1991 onwards. Hydrographic profiles were designed to cover the shelf and slope with observations extending to 500 m depth and with some stations, usually two, occupied oceanwards beyond the 500 m depth line. In some surveys vessel drift was recorded for the description of surface currents. Logistic use was made of the vessel for mooring of current meter rigs in the Somali Current in 1976 in co-operation with the University of Miami and the Institute of Marine Science in Kiel. Co-operative programmes were conducted with the objective of comparing in situ temperature observations with data from remote sensing from satellite systems.

Besides presenting a general description of the hydrographic environment, the analysis of these data concentrated on features of the environment which were expected to influence the distribution, composition and abundance of the resources, such as surface salinity in estuarine areas, depth of the thermocline, oxygen-depleted water in the bottom layers, upwelling as an indicator of enhanced productivity, prevailing currents ocean fronts and seasonal changes of the main parameters and in the regime as a whole.

2.6 TAXONOMY AND BEHAVIOURIAL CLASSIFICATIONS

Tropical and subtropical regions are characterized by the high species diversity of their flora and fauna. The marine environment is no exception and nearly 40% of all known fish species occur in the shelf waters of tropical oceans. The Indo-Pacific region, in particular, contains by far the most diverse fauna of any marine zoogeographic region.

The DR. FRIDTJOF NANSEN programme started its operations in the Indian Ocean and in Southeast Asia and, not surprisingly, species identification constituted a major problem. In addition to the complexity of the fauna, at that time most of the taxonomic information was only available in specialized scientific journals and was thus practically inaccessible or inappropriate for field work. The two main monographs on fish taxonomy available were the J.L.B. Smith's ’Sea Fishes of Southern Africa’ (Smith, 1972) and one of the first sets of the ’FAO Species Identification Sheets for Fishery Purposes’ covering FAO fishing areas 57/71 (Fischer and Whitehead, 1974). These documents had, however, great limitations, both in the erroneous classification of many taxa and in their limited coverage of species and geographic range.

The scanty literature, coupled with the lack of experience amongst Norwegian scientists in tropical fish taxonomy, may have affected the data quality in the first years of the programme, at least as far as identification to species level was concerned. Attempts were made to overcome this problem, at least in part, with the help of international experts or experienced local counterparts. The first taxonomist was engaged for a special survey of deep-water resources off Kenya in 1980. His findings indicated that better taxonomic classification was urgently needed (Nakken, 1981; Venema, 1981).

Since a good collaboration was established with the FAO Species Identification Programme, based on mutual interests: FAO provided, whenever possible, a tropical-fish expert to help with species identification and this constituted at the same time a unique opportunity to collect valuable field information. Sets of Species Identification Sheets and field guides under development were tested in the field, specimens of ‘problem species’ collected, and distribution charts modified according to field observations. New occurrence records, both by depth and geographic range, were usually confirmed by sending specimens to relevant taxonomists. The Identification Sheets for the Eastern Central Atlantic, Fishing Areas 34/47 (in part) (Fischer et al., 1981) and those for the Western Indian Ocean, Fishing Area 51 (Fischer and Bianchi, 1984) benefitted from this collaboration.

On board the DR. FRIDTJOF NANSEN, apart from the identification work, the large database on species records was routinely updated and nomenclaturial errors corrected.

As new areas were covered by the surveys and whenever there was a clear need for compiling or updating taxonomic information on the marine resources, the programme took the initiative of starting the preparation of new field guides and provided both funds for printing and the opportunity to collect field data. This was always done in collaboration with the FAO Species Identification Programme, responsible for the implementation of this activity. The above resulted in the publication of the field guides for Angola (Bianchi, 1986), Morocco (Bianchi, 1984), Mozambique (Fischer et al., 1990), Namibia (Bianchi et al., 1993), Pakistan (Bianchi, 1985a), Sri Lanka (De Bruin et al., 1994), Tanzania (Bianchi, 1985b), Northern Coast of South America (Cervigón et al., 1993) and the Eastern Central Pacific (Fischer et al., 1995).

As years passed, the greater availability of adequate taxonomic literature and the growing experience of Norwegian participants allowed for a more detailed identification of the catches, usually including all species caught. Although identification of all species was not a priority (major food fishes were the main target of the surveys), those data constitute a unique evidence of species occurrence and diversity.

A number of scientific names used in early survey reports are now outdated and are considered synonyms. However, these denominations have been maintained in the reviews contained in this volume for reasons of consistency.

Not only taxonomic classification has been a major challenge. As already mentioned in earlier sections, the use of operational categories such as pelagic and demersal, widely used in fishery research, soon proved inadequate. In fact, there is hardly any fish species whose behaviour strictly conforms to either. From observations on board the vessel, many species traditionally assigned to either group show intermediate types of behaviour. As a consequence the allocation of any species to those categories is difficult if not impossible. For example, in waters over the continental shelf, many so-called pelagic groups are usually observed near the bottom and are caught in the bottom trawl, including several Carangidae of the genera Trachurus, Chloroscombrus, Selene, Caranx etc., barracudas, Scombridae (mainly Scomberomorus) and even Clupeidae. Also further offshore, large schools of sardinella may be found resting on the bottom during daytime.

On the other hand, it is well known that many demersal fishes rise from the bottom at night. Many species, however, occur in mid-water without a clear pattern of vertical distribution, e.g., ponyfishes, snappers, seabreams, etc. Unfavourable conditions near the bottom on a seasonal basis (for example, as with the occurrence of oxygen-depleted waters) may trigger off a pelagic behaviour. This has been observed, for example with catfishes of the genus Arius, dragonets (family Callyonimidae), threadfin breams (Nemipterus) etc. especially in the shelf areas of the Arabian Sea.

The same species may have different types of behaviour in different regions, for instance being more pelagic where upwelling is more extensive in duration or strength. Such adaptations are typical for the horse mackerel Trachurus trecae which occurs off central and southwest Africa. The changes in behaviour might be due either to temperature preferences or to changes in the availability of food.

These observations indicate that the categories ‘demersal’ and ‘pelagic’ are ecologically and operationally inadequate. Most species seem to possess a wide range of adaptations and flexibility, responding in different degrees to regular or stochastic fluctuations in ecological conditions.

A deeper analysis of various types of behaviour and of the environmental factors generating them would be desirable. Apart from increasing the understanding of the ecological adaptations to the marine environment, it would improve the application of methods used in investigations on the distribution, composition and abundance of marine resources.

In this review, in addition to the two classical categories pelagic and demersal, three other categories are used: semi-demersal, for species traditionally defined as demersal, and with a body shape more adapted to living close to the bottom, but often observed in mid-waters; and semi-pelagic indicating those species usually designated as pelagic, and with body shape adapted to mid or upper-water layers, but often found close to the bottom and caught in the bottom trawl and meso-pelagics, fish that reside during the day at depths around 300 m and that may rise to near the surface at night.