13.1 THE BOTTOM TRAWL
13.2 PLANNING A DEMERSAL TRAWL SURVEY
13.3 DATA RECORDING
13.4 DECK SAMPLING AND CATCH RECORDING PROCEDURES
13.5 THE SWEPT AREA
13.6 BIOMASS ESTIMATION BY THE SWEPT AREA METHOD
13.7 PRECISION OF THE ESTIMATE OF BIOMASS
13.8 ESTIMATION OF MAXIMUM SUSTAINABLE YIELD
Bottom trawl surveys are widely used for monitoring demersal stocks when only an index of abundance is required. From unfished stocks (or stocks for which no or few data on the fishery are available) biomass and annual yield estimates may also be derived by bottom trawl surveys. The estimation of total biomass from the catch per unit of effort (or unit area), however, involves several crucial assumptions, leaving such estimates rather imprecise.
The mean catch (either in weight or in numbers) per unit of effort or per unit of area is an index of the stock abundance (i.e. assumed to be proportional to the abundance). This index may be converted into an absolute measure of biomass using the so-called "swept area method" (Section 13.5). This method falls under the so-called holistic methods (cf. Section 1.8).
Reviews of the theory are given in, for example, Gulland (1975), Saville (1977), Troadec (1980), Doubleday (1980) and Grosslein and Laurec (1982). These reviews also give guidelines for conduct of surveys (planning, design, data collection, data recording, analysis and reporting), some of which are summarized in Sections 13.2 to 13.4), see also Butler et al. (1986), ICOD (1991) and Strømme (1992).
This Chapter gives first a short description of demersal trawl survey planning and a few guidelines for data recording and for field work (Sections 13.1 to 13.4). For more detailed descriptions of these subjects the reader is referred to, among others, Alverson and Pereyra (1969), Alverson (1971), Mackett (1973), FAO/UNDP (1975), Gulland (1975), Saville (1977), Flowers (1978), Doubleday (1981), Grosslein and Laurec (1982) and Fogarty (1985). The remainder of the Chapter (Sections 13.5 to 13.8) is a short account of the theory necessary to perform a stock assessment based on trawl survey data.
The bottom trawl (Fig. 13.1.1) is a conical net bag with a wide mouth fitted with weights on the ground-rope and floats on the head-rope. When the vessel is under way the net is kept open by two otter boards, wooden or iron structures which are towed by the warps attached forward of their centre so they tend to diverge. The two otter boards are connected to the net by bridles. These may be up to 200 m long and sweep the sea bed over a wide area. They frighten the fish towards the advancing net and so increase its effectiveness. The shape of net varies depending on the kinds of fish to be caught and on the types of bottom. The ground-rope may be fitted with roller gear (bobbins) so that the trawl can be used on stony bottom without being damaged. The tail end of the gear from which the captured fish are removed is called the "codend". This is where most of the size selection takes place. In most cases a relatively small mesh size is required in the codend, in order to obtain a representative sample for the entire size range of the species under investigation (see Section 13.2).
Fig. 13.1.1 Bottom trawl
Below is a list of some important items to consider before conducting a survey.
Definition of objectives
The objective(s) should be specified. Examples of objectives are:
1. Estimation of the total biomass and catch rates
2. Estimation of the biomass of selected species
3. Collection of biological data (e.g. length-frequencies) for estimation of growth and mortality parameters
4. Collection of environmental data
Information about the survey area
Information about depth and bottom conditions to point out trawlable areas and decisions on strata may be obtained from a preliminary survey with echo-sounding. Information from local fishermen may be valuable. Information on seasonal winds, currents and migration patterns of fish stocks are important as well.
Choice of gear
The design of the trawl should fit with the expected bottom conditions and the vessel used. If rough bottom is prevalent the gear should be fitted with bobbins in order to avoid damage to the gear. If semi-pelagic species are common a high-opening trawl should be used. The mesh size in the codend should be chosen so that the entire size range of the fishable part of the stock is retained by the trawl. Often 10-20 mm stretched mesh is appropriate. The mesh size used for assessment surveys is usually much smaller than the size used by the commercial fisheries, because samples of small fish are important for assessment methods based on length-frequencies.
If different trawls are used parallel or alternate hauls should be carried out in order to estimate correction factors for pooling of data.
A procedure for the selection of stations should be decided on. A fixed grid of stations ensures maximum information on the distribution throughout the area, but not necessarily the most precise estimate of biomass. For estimation of stock sizes a completely randomized design or a stratified random sampling design should be preferred. In most cases a stratified sampling design should be chosen because fish are seldom uniformly distributed and in most cases abundance is related to depth. Stratified sampling often gives a much more precise estimate for the same (or even lower) cost than simple random sampling (cf. Sections 7.2 and 13.7).
Allocation of hauls (stratification)
Strata are basically constructed in accordance with the density distribution of the fish, so that areas with high/medium/low densities are separated. To do so, some information must be available before the survey. Perhaps the first survey (or the first part of the survey) should have a completely randomized survey design, or the trawl hauls might be evenly spaced. But in the next part of the programme some information on densities and standard deviations of density estimates will be available, and that information can be used for stratification. The standard deviation of catch per unit area (CPUA) is often found to be proportional to the CPUA and will therefore be higher in areas with high abundance.
The optimum allocation of a given number of hauls between strata will be to sample each stratum in proportion to its standard deviation multiplied by the stratum size (cf. Sections 7.2 and 13.7). The distribution of hauls within strata should preferably be at random, but often practical matters dictate the sample design. For example, obstacles on the bottom may not allow a proper random distribution of stations.
Possible number of hauls
To estimate how many hauls it is possible to make in a given period the following information should be available:
Total number of days available
Time spent going to and from the fishing ground (hrs)
Duration of one haul (hrs)
Time used for shooting and hauling the trawl (hrs)
Time to cover distance between stations (average, hrs)
Number of hours available per day (depending on crew, behaviour of investigated species, navigation, etc.)
Time used for preparations: loading of ice, food, water, repair of gear and equipment (days)
Except for the first and the last day of a cruise, when T should be reduced by t1, the number of hauls per day can be calculated from:
(number of hauls per day) = T/(t2+t3+t4)
(Total number of hauls) = (N-t1-t5)*(number of hauls per day) + (hauls made first and last day)
It is important to standardize the length of a haul throughout the survey, since the catchability of species and sizes often depends on the duration of the haul. For survey purposes hauls of 0.5 hour or 1 hour are usually the most adequate (see also Section 13.6).
When setting up a plan for a trawl survey a crucial point is to decide on the data items to be recorded, the required precision and how often they should be recorded.
The data items to be recorded are determined by the models by which the objectives of the survey can be achieved, e.g. the swept area method (cf. Section 13.6), length-frequency analysis (cf. Chapter 3), mortality estimation (cf. Chapter 4). Such data items would usually include specifications of the gear used, and for each haul the time and position at start and at end of the haul, wire length, wingspread, bottom type and depth. The catch record should include total weight, species composition by weight and length-frequencies (for selected species).
The required precision depends on the subsequent use of the data. However, often the precision of a trawl survey is controlled by the number of hauls, which limits our ability to decide on the precision.
It must be known how the data are going to be processed before data recording takes place, particularly in cases where data reduction takes place before the recording stage. Consequently, before you can design appropriate log sheets you must have a rather precise idea of how data will subsequently be processed. When designing a log sheet, practical considerations are important, e.g. where the data are entered, on the bridge, in the laboratory or on the deck.
It is important that the data are well organized to facilitate processing, e.g. by computer. This is especially the case when data from more than one vessel have to be combined. Properly designed recording documents will make the correct recording much easier and more dependable. Several documents are usually required to record all the information collected during a survey. These are in the following categories:
1. A log which summarizes the whole cruise.
2. Details of individual stations (or hauls), the "fishing log", which will generally provide information on the vessel's position, time of start and end of haul, gear rigging etc. Summary information on the catch such as total weight and weight composition by species should also be recorded on the fishing log.
3. Detailed information on the catch. This may be in terms of length, weight, sex, maturity stage etc. for each specimen, or samples of length-frequency distributions.
A detailed description of a data processing system for demersal trawl surveys is given in Flowers (1978) who reproduces forms suitable for the work specified as 1 to 3 above.
It is important, before the survey begins, to make sure that the equipment and working conditions are such that the sampling can be carried out easily and without risk. Also, the crew must be instructed not to remove any part of the catch before the sampling has been completed.
The following steps pertain to methods for sorting the catch of a fishery research vessel such that the catch composition, both by weight and by number of each species (species group) can be established. The procedure outlined here is from Pauly (1980, adapted from Losse and Dwiponggo, 1977).
Step 1: Remove all sea snakes and other venomous or otherwise dangerous animals. Also remove turtles, and if alive, return these to sea. Record number and kind of animals removed.
Step 2: Remove inorganic debris and plant material. Record type of material removed.
Step 3: Remove the larger fish that are readily visible and place them in a box.
Step 4: Wash the remainder of the catch (of small fish) if necessary, and mix with shovels.
Step 5: Put the mixed catch in boxes, while continuing to remove larger fish and putting them into the box mentioned in Step 3. The boxes should be filled simultaneously, not one after the other, and it should be made certain that all boxes contain approximately the same weight of fish.
Step 6: Count the number of boxes with small fish and record.
Step 7: A rule of thumb is to take one box out of every five at random for subsampling. Record number of boxes taken for subsampling as B1, B2, B3, ... etc.
Step 8: The box(es) taken for sub-sampling is (are) then treated as follows:
- Weigh the total catch in B1 and record.
- Place fish of B1 on a sorting table and sort to species level as far as food fishes and valuable crustaceans (e.g. shrimps) are concerned, and to taxonomic groupings as well-defined as possible (e.g. genus, family, etc.) for other groups (the non-edible fish and miscellaneous crustaceans).
- Repeat procedure, if appropriate, for the other boxes, B2, B3, ... etc.
Step 9: If more than one box was sorted, compute, for each species (or higher taxonomic group) the total weight and number in all sorted boxes.
Step 10: Multiply the numbers and weight of fish and invertebrates by species (or higher taxonomic group) by the ratio of the number of unsorted to sorted boxes.
Step 11: Weigh and count the larger fish mentioned in Steps 3 and 5, by species (very large fish should be weighed individually and measured).
Step 12: Add, when there is an overlap (when the fish of a certain species occurred both in the sorted boxes of small fish and in the large fish box) the weights and numbers obtained in Step 11 to the weights and numbers in Step 10.
Step 13: Step 12 (as well as Step 11, when there is no overlap) provides estimates of total catch, both in weight and number, by species or higher taxonomic groups. Record the total, both in weight and numbers into an appropriate fishing log and convert to catch per unit if fishing time was less or more than an hour. During surveys, this step must be completed after each haul, or every evening at the latest, to preclude loss of information.
Step 14: In addition to catch sampling, identifying and recording, the work of the fishery scientist generally includes among other things:
- Collecting length-frequency data
- Collecting miscellaneous biological information on the fish caught, e.g., concerning their weight and maturity
- Collecting and preserving specimens for further studies onshore
- Collecting oceanographic data.
It is important to sort and sample the catch to the greatest extent possible. The extra cost involved in sorting and sampling all the edible species is usually only a fraction of the total costs of the survey.
From Fig. 13.5.1 it appears that the trawl sweeps a well defined path, the area of which is the length of the path times the width of the trawl, called the "swept area" or the "effective path swept". The swept area, a, can be estimated from:
a = D*hr*X2, D = V*t..........(13.5.1)
where V is the velocity of the trawl over the ground when trawling, hr is the length of the head-rope (see Fig. 13.5.1), t is the time spent trawling. X2 is that fraction of the head-rope length, hr, which is equal to the width of the path swept by the trawl, the "wing spread", hr*X2 (see Fig. 13.5.1).
For southeast Asian bottom trawls values of X2 from 0.4 (Shindo, 1973) to 0.66 (SCSP, 1978) are reported. Pauly (1980) suggests X2 = 0.5 as the best compromise. In the Caribbean a value of X2 = 0.6 was used by Klima (1976).
For the estimation of biomass we use the catch per unit area (CPUA), as will appear in the next section. It is estimated by dividing the catch by the swept area (in square nautical miles or square kilometers). This estimate thus depends on how accurately the swept area is estimated. The swept area as defined in Fig. 13.5.1 assumes that the bridles have no herding effect, which they are in fact known to exhibit. The wing spread is calculated as the fraction X2 of the head rope length. The wing spread varies with hauling speed, weather conditions, current velocity and direction and warp length and is therefore not well defined. It can be accurately measured by special devices, but an approximate value can be calculated from measurements of the distance between the warps at the gallows and at say, 1 m, towards the net.
When exact positions of the start and the end of the haul are available the distance covered can be estimated in units of nautical miles (nm), by:
Lat1 = latitude at start of haul (degrees)
Lat2 = latitude at end of haul (degrees)
Lon1 = longitude at start of haul (degrees)
Lon2 = longitude at end of haul (degrees)
Fig. 13.5.1 Swept area (cf. Fig. 188.8.131.52)
Fig. 13.5.2 Vector addition of vessel and current course and velocity
If exact positions are not available, but only the velocity of the vessel and its course together with direction and speed of the current, then the distance covered per hour can be calculated from:
VS = velocity of vessel (knots = nm/hr)
CS = velocity of current (knots)
dirV = course of vessel (degrees)
dirC = direction of current (degrees)
Eq. 13.5.3 is an application of "vector addition" (see Fig. 13.5.2).
Let Cw be the catch in weight of a haul. Then Cw/t is the catch in weight per hour, when t is the time spent hauling (in hours). Let a be the area swept (cf. Eq. 13.5.1). Then a/t is the area swept per hour, and
becomes the catch in weight per unit of area. Let X1 be the fraction of the biomass in the effective path swept by the trawl which is actually retained in the gear and let be the mean catch per unit area of all hauls. Then an estimate of the average biomass per unit area, , is:
Let A nm² be the total size of the area under investigation. Then an estimate of the total biomass, B, in this area, A, is obtained from:
It is difficult to estimate which proportion of the fish that is present in the area swept by the trawl gear is actually retained by the gear, in other words it is difficult to give a precise estimate of X1. Underwater television recordings show that the reaction of fish to trawls varies markedly between species. The value of X1 is usually chosen between 0.5 and 1.0. For trawlers in southeast Asia a value of X1 = 0.5 is commonly used in survey work (Isarankura, 1971, Saeger, Martosubroto and Pauly, 1980). Dickson (1974), on the other hand, suggests X1 = 1. The difference between these two values of X1 is difficult to resolve. Using X1 = 0.5 doubles the estimate of biomass compared to that obtained by using X1 = 1.0.
The duration of the haul is proportional to the distance covered so the duration should have no direct influence on the catch per unit of area. However, the catchability, X1, of different species may vary according to the duration of the haul because some species, when herded by the trawl get tired soon and get captured while other species are able to swim in front of the trawl for a long period and thus avoid being caught. It is therefore important to standardize the duration of the haul so that results from different hauls can be compared. To investigate the dependence of catchability on haul duration, parallel hauls of different duration (e.g. half an hour and one hour) could be made.
The total area surveyed can be estimated from a mercator projection by means of a planimeter. Another method is to copy the area on transparent paper which then is cut out and weighed. Then the weight of the paper is compared to the weight of a piece of similar paper which corresponds to a known area. This method is not very precise.
Let the biomass estimate in Eq. 13.6.3 be obtained from n hauls, and let Ca(i) be the catch (in weight) per unit of area of haul no. i, where i = 1,2,...,n. The estimate of B then becomes:
and the variance (cf. Eq. 2.1.2):
Thus, a higher precision (a smaller variance) can be obtained by increasing the number of hauls, n (cf. Section 7.1, Fig. 7.1.1). Another way of reducing the variance is to apply stratified sampling (cf. Section 7.2). Suitable stratification may reduce the variance considerably for the same number of hauls and thus improve survey efficiency for the available research vessel time. The distribution of many fish species is determined by depth and bottom type. Therefore stratification based on these factors is widely used. Fig. 13.7.1 shows an example of stratification based on depth.
The areas of the four strata in Fig. 13.7.1 are A1, A2, A3 and A4 and the total area A = A1+A2+A3+A4. Let B(i) be the estimated biomass in stratum no. i calculated by Eq. 13.7.1. Then the estimate of the total biomass of the total area, A, becomes:
B = B(1)+B(2)+B(3)+B(4)..........(13.7.3)
and the variance:
VAR(B) = VAR(B(1)) + VAR(B(2)) + VAR(B(3)) + VAR(B(4))..........(13.7.4)
where VAR(B(i)) is calculated by Eq. 13.7.2. If we work with densities, in units of biomass per square nm. (cf. Eq. 13.6.2) the parallel to Eq. 13.7.3 becomes:
b = [b(1)*A1 + b(2)*A2 + b(3)*A3 + b(4)*A4]/A..........(13.7.5)
and the parallel to Eq. 13.7.4 is
VAR(b) = VAR(b(1))*(A1/A)² +...+ VAR(b(4))*(A4/A)²..........(13.7.6)
where Ai is the size of stratum no. i and A is the total area of all strata. Eq. 13.7.6 is then seen to be analogous to Eq. 7.2.11 with A as N, Ai as N(j). Confidence limits of B and b can be calculated as described in Section 2.3.
Fig. 13.7.1 Example of stratification
Stratification may be based on the total catch of a mixture of species or on the catch of a single species. It is often desirable, however, to focus on several species, or groups of species, each with its own type of distribution. In such cases a stratification procedure must be decided on for each species, or for each group of species with similar distribution patterns within the area surveyed.
For a more detailed description of the statistical aspects of a trawl survey see Fogarty (1985) and Gavaris and Smith (1987). The latter discuss the effect of stratification by comparing the precision obtained from stratified sampling to that of simple random sampling, using the method of Sukhatme and Sukhatme (1970).
The above formulas all assume that the number caught per trawl hour is a normally distributed random variable. A better description of survey data is often obtained by assuming the so-called "delta-distribution" (Pennington, 1983). This distribution takes into account that survey data often contain a large proportion of zero observations and that the non-zero observations form a log-normal distribution rather than a normal distribution.
Fig. 13.7.2 Delta-function
A log-normal distribution is asymmetric and skewed to the right hand side. Fig. 13.7.2 shows a schematic delta-function. Note that the point zero has a positive probability. In his paper, Pennington gives the expression for the unbiased estimators of the parameters of the delta-function.
Gulland's formula for virgin stocks (Eq. 9.2.1):
MSY = 0.5*M*Bv
(M: coefficient of natural mortality; Bv: virgin stock) or Cadima's formulas for exploited stocks (Eqs. 9.3.1 and 9.3.2):
MSY = 0.5*Z*B or MSY = 0.5*(Y + M*B)
have often been applied when estimating the maximum sustainable yield (MSY) per year from the total biomass as estimated from exploratory surveys (see e.g. Saville, 1977, or Troadec, 1980). It may be better, however, to apply the modified Fox model (Garcia, Sparre and Csirke, 1987) described in Section 9.4 using Eq. 9.4.10 or Eq. 9.4.11.
Anyway, these are crude methods and should be relied upon to give only the order of magnitude (1,000 tonnes, 10,000 tonnes, etc.) of the fish stock. Yet, if nothing is known in advance, even such an estimate of the order of magnitude may be valuable.
One problem is that M is usually not known before the surveys are carried out. When the stock is virgin (unexploited) an estimate of M can be obtained by analyzing the catch curves obtained during the exploratory survey (cf. Section 4.4) because when the stock is unexploited, then F = 0 and thus Z = M. Pauly's empirical formula, Eq. 184.108.40.206 (Pauly, 1980b) can be applied when growth parameters are available.
(See Exercise(s) in Part 2.)