Having established what the basic data requirement is, with some indication of priorities, we now turn to the problems of collection. In designing collection methods the advice of a competent and experienced statistician is absolutely essential, as the procedure to follow is by no means simple. The statistician is concerned with the method and design of measurement and will try to establish:
1. What is to be measured?
2. What questions should be asked?
3. What precision is needed?
4. How can the survey beet be carried, out to provide the information desired with the desired precision and no more?
5. What will the survey or sampling system coat?
6. What do the results mean?
7. How can objective measures of the sampling errors and biases be obtained, so that the reliability and meaning of the results can be assessed and methods improved?
Table 5: Disposition channels and commodity types (FAO, 1966)
Once again we must return to consider the overall objectives and the objectives of collecting data in order to specify what reliability is to be aimed at and what cost we are prepared to pay in order to obtain it. These are questions which a statistician cannot answer, but which he should help us to pose clearly. He may, for instance, design alternative surveys or sampling systems to show the relationship between cost and precision, taking account of the personnel and facilities available and the restrictions likely to be en countered. Finally he should, appraise the precision actually attained and evaluate the results generally, with a view to improving the design. The three steps of specifying, designing and appraising are clearly interrelated, so that a change in specification may alter the design and the result of an appraisal may be a change in specification.
The precision of a sampling procedure can only be estimated if the sample is drawn - randomly from the population being sampled. Precision can be represented as the scatter of points about the mean of a large number of such samples (see for example Bazigos, 1974). The accuracy of this mean is measured by the difference between it and the true mean value in the population. In practice this "true mean value" is not known and often may not even be established by complete census, but some idea of accuracy may be gained by the use of different estimation procedures for the same statistic and by a careful appraisal of possible sources of bias. Apart from clearly defined objectives and a defined level of precision or cost to work to, the statistician requires some information on the structure of the population being studied and on the way in which the sampling may be organized.
In this chapter the problems of collecting from industrialized fisheries based at a few large ports are considered separately from the problems of collecting from scattered, small scale or artisanal fisheries, because, although the objectives are very similar in both, the methods which can be used are rather different. As will no doubt become obvious from later comments, we are using the terms "small scale" and "large scale or industrialized." in a rather general way to indicate difference in size of vessels, landing sites and markets; degree of technological advance and general level of development. There are of course intermediate fisheries of many kinds, snob as lobster fisheries - scattered fisheries using small boats with advanced technical aide and with international markets.
It is worth mentioning again that many of the methods of collection can be carried out only when the personnel and institutional framework of the collecting system have bad tints to develop their capacities. In the initial stages the emphasis should be on training of staff, pilot and feasibility studies and continuing reappraisal of the objectives and methods employed to attain them. Although the statistical collection system finally adopted should be capable of adapting to changing circumstances and objectives, institutional inertia often prevents such change. In any case the value of most statistical series lies largely in their continuity and it may not be advisable to make minor improvements which disrupt this continuity. It is important to get the thing right from the start.
Against this must often be set the need for quick answers to a particular problem. There is little that a general discussion can do to help resolve such conflicts. Speed in producing a set of data does not preclude a subsequent reappraisal of the reliability of the data. Unwillingness to re-examine old methods and fear of admitting that data may be unreliable should play no part in the collection of statistics,
The costs of collecting different kinds of routine statistics varies greatly and depends very much on the existing facilities within a country. Where the industry is sufficiently organized and advanced to have centralised marketing facilities or an auction it will often be possible to obtain some records of total catch and value without the need for sampling arid recording directly, except for checking occasionally for greater detail. It is usually worth spending quite a lot on checking and improving existing records if the only alter native is to set up an independent collecting system. Where no records exist the collecting system must start with a survey to obtain the most basic information about ports, landing sites, numbers and types of vessels, species landed and so on. This is known as a frame survey and part of the cost is non-recurring. The further cost of collecting regular statistics of catch, effort etc., can only be assessed once the basic information from the frame survey is available. Some of the techniques for this type of survey will be introduced in Section 3.3.
The collection of data on fishing effort is usually more expensive than catch data because it is seldom available from existing records in a form which can be used, On the other hand we have seen that detailed statistics on effort may only be required for a part of the fleet. Once arrangements have been made to collect, some statistics in detail, the cost of collecting others in detail is often no greater. For example, if the skipper or mate is being interviewed or asked to complete log books to find out the exact fishing grounds used, it takes very little extra time to record details of the gear, time spent fishing etc., on the form. Even if not required for the immediate objective, the possibility of its future use justifies the negligible cost of collection. This factor should always be borne in mind when planning a questionnaire or collection system.
The cost of collecting data on the size and, where the biological characteristics of the fish allow, age structure of various specie is high compared with the cost of collecting catch and effort data because it involves special sampling of the landings, usually in two stages, and the training of skilled, age readers. The statistics are however essential for the more powerful modelling techniques used in assessment and it is possible to justify the cost in terms of improved management advice in both the short and long term. In the short term detailed size and age statistics make possible yearly forecasts of catch rates and these will help fishing companies to plan their fleet deployment and processing plants to adjust their capacity. An example of the costs and potential benefits of a yearly forecasting system for sockeye salmon runs is given by Mattheus (1971). In the long term the size and age compositions of the population are needed for detailed studies of growth, migration and reproductive patterns. Where the age composition cannot be obtained directly by determining the age of individual fish, indirect techniques may be used.
Although the category "industrialized fishery" includes a wide variety of different kinds of fishing operations and landing patterns, which give rise to differences in sampling methods, there are a number of common factors. There is little difficulty in identifying the number and types of vessels and their ports. Because there are relatively few ports it is possible to set up port based sampling systems. The skippers will generally be literate and able to fill in log books or questionnaires. There will usually be records of catch weights, sales and crew payments.
Because of all these factors it is possible to obtain statistical records of catch weights, sales, vessel characteristics, numbers of tripe etc. simply by arranging to get copies of sales sheets and crew settlement forms. Well informed, collecting staff in the ports are still needed to check on the completeness and reliability of the statistics provided and to collect other data which cannot be found in existing records. Of these data the moat important is fishing effort (fishing input). Whether the aim is to record the fishing effort of every boat landing or to sample part of the fleet, the skipper or mate should be interviewed as soon as possible after the boat returns, in order to obtain inform- at ion on exact grounds fished, number of hauls and hours fishing.
In some cases a great deal of work can be saved by introducing a log book which is kept by the skipper and returned at the end of each trip, or perhaps once a week if the trips are short. With goodwill and cooperation on the part of the skipper this can give very detailed. and accurate information, but if it is simply another chore to be fitted into an already exhausting job then the quality of the data will be poor, Paying to have log books kept or making them a condition of licensing may improve the situation, but if there is some way of giving back the information, which does not destroy the necessary degree of confidentiality but is of interest to the fishermen, this will be a stronger incentive.
Methods of collecting catch and effort data from a demersal fishery landing at a major port and from a purse seine fishery are described by Holden and Raitt (1974). The actual methods of collection will have to be adapted to the particular port and type of fishery, but it is important that the type of form used and the method of completing it should be standardised as far as possible at a national or regional level. This can be done by running central training and refresher courses for port staff, during which they can compare notes with colleagues in other parts of the country, and by producing manuals on how the various forms should be filled in (e.g., Shultz, in preparation). The fisheries scientists or statisticians who run training courses for port staff will also learn the details of the fishing industry which are lost in statistical aeries.
The design of good sampling forms of all kinds, which are straightforward to fill in and easy to check and process is worth spending a great deal of care on. The aim should always be to reduce the number of transcriptions of -the data to a minimum since these are costly and introduce errors. If ADP (automatic data processing) is being used then the aim should be to punch the data straight from the original record.
We have assumed that in industrialised ports the aim will be to record the catch from every landing and to record effort for all or part of the fleet. The aim in collecting length and age data is to estimate, for the more important species at least, the actual numbers landed in each age or length group with the greatest level of precision, given the limits to available funds and manpower. In order to do this a series of small samples is taken and then multiplied up to give the estimate of numbers of that species in the total catch. Establishing the best way in which to carry out this sampling can be very difficult, but since the cost of collecting the length and age data is usually rather high it is worth spending the time and money -to get it right. At present far too little is spent on ensuring that such sampling systems are efficient and the wastage is enormous.
If the fleet landing at a port consists of similar boats using one type of gear and fishing for the same species in a uniform area then the problem of sampling will be fairly simple. If the fish have been sorted into size or quality categories before sampling then randomly chosen samples within each category can be taken and raised to the total weight for the category. If the boats are fishing different areas with different gears then a number of divisions or "strata" will have to be taken into account and the problem is to strike a balance between the degree of subdivision and the number and size of samples which it is possible to take (e.g., Gulland, 1955; Deming, 1950.)
The location at which sampling for size and age are carried out and the time at which the samples are taken may be very important. In order to reduce the problems of transportation and to be sure that a sample does indeed come from a particular vessel it is preferable to sample on the quayside during landing. If this is very inconvenient, uncomfortable or impossible then it may be necessary to buy sample through a merchant for sampling later or to sample the fish at a later stage, e.g., processing. In these cases very great care should be taken to ensure that samples are not biased (e.g., the merchant may always buy from the same boats).
Many of the comments made about standardisation of forms and design of sampling scheme apply to small scale as well as industrialised fisheries but the major problems of collection are quite different. It is simply not possible to attempt to obtain records of the catch from each landing because there are usually no existing records at any level and the number of landings is so large and scattered that total coverage cannot be undertaken. Carefully planned, executed and monitored survey techniques can provide the necessary information by sampling methods and the cost need not be very great. Bazigos (1974) deals with the design of such surveys and refers to several practical applications. Recent examples of the application of these techniques to marine fisheries come from West Africa (Banerji, 1974).
Like a census of human population, a fishery survey will have a number of aims each of which may beat, be achieved by slightly different means, but an estimate of the total catch taken by all vessels will generally be the first concern. The vessels or fishing units, which may be of different sizes, use different gears etc., are grouped according to landing site and there may be further regional groupings to take account of different types of fishery or simply different administrative regions. The first requirement will be to produce a list of all these landing sites with enough information about numbers of vessels and type of fishery to enable um to set up at least a preliminary "sampling frame" for use in estimating total catch. The criteria for deciding what to use as our sampling unit and how to group these units are exactly the same as those briefly discussed when considering how to choose a sample for measurement and age determination in the previous section. A balance must be struck between the degree of subdivision and the number and size of samples which it is possible to take.
For most purposes the landing site is chosen as the sampling unit and -they may be grouped according to their size and location in order to take account of known differences in the kind of area which their fleets fish. Other factors may be shown to be important by a preliminary survey. For example in Sierra Leone fishing vessels were classified into five types, but a preliminary survey showed that a major source of variation in catch arose from landing sites operating the largest class of vessel since these land a far greater amount than the other four types. As a result the landing centres were divided into those with the largest class of vessel and those without.
Once the sampling units (landing sites) have been selected and divided into categories or strata (e.g., by area, number of vessels, type of vessel) and a complete list of sites exists, one can begin to decide how many sampling units within each stratum to sample and which sampling unite to select. In the absence of any information about the variability of catch among the sampling units within each stratum it is best to keep the fraction of units sampled constant for each stratum (e.g., if it is proposed to sample at ten landing sites within a country and there are three strata with 20, 30, and 50 landing sites respectively, then one would allocate 2, 3 and 5 samples). The process of selecting which landing sites to sample must be carried out randomly in order to be able to use statistical techniques in analysing the results. The method used for randomisation can have a great effect on the level of precision attained and it needs careful investigation. One simple method is to use patterned or systematic sampling in which one sample from a stratum is selected by a random start and the others are taken at a constant interval. For the example given previously in which 5 landing sites are to be sampled from a stratum containing 50 such sites, the 50 sites could be listed in geographical order along the coastline; one site is chosen by random number and the others are taken at intervals of 10.
So far we have talked only about estimating the total catch since this is usually the first objective. Many other statistics such as fishing effort, costs and earnings etc., can be collected in the course of frame surveys and catch assessment surveys either by integrating them into the survey questionnaire or by selecting a smaller subsample, within the existing sample design for more intensive investigation.
Two factors of the utmost importance in the subsequent use of information obtained by means of survey techniques are:
(a) the completeness of frames, validity of sampling design and methods, reliability of the collectors, accuracy of copying and processing must be thoroughly checked at all times;
(b) the fullest possible records must be kept of methods of work, techniques for randomisation, suspected or known biases in collection and breakdowns in the sampling scheme.
These factors are obviously also important for statistics collected by routine census-type systems, but are additionally important for surveys in order to compensate for any lack of continuity in personnel and difficulties in subsequent checking. The results should be presented clearly, with indications of the level of precision achieved and with no attempts to conceal possible shortcomings. It is essential that those using them in future should have confidence in them and should be able to learn from any mistakes made.
Because most of the major marine fish stocks are the subject of exploitation by more than one country, their management must be carried out internationally by regional fisheries commissions or through multilateral agreements. Just as it is difficult or impossible to manage a fishery unless all the participants cooperate, so it is difficult to assess a fishery and give advice on management unless all the participants collect data and provide it in time. The assessment will be as good as the poorest data used in producing it.
With very few exceptions (e.g., Inter-American Tropical Tuna Commission) the international regulatory bodies rely on national agencies to provide the statistics and their function is one of compiling rather than collecting. For example, the annual Bulletin Statistique of ICES provides records of the total catch of each species, by each country in each of the ICES areas since 1906. More detailed stock information, giving length and age compositions and the
catch and effort by statistical square (1° latitude by 1/2° latitude) is published, in the Statistical Newsletter. A number of standardised reporting forms have been developed for international compilations such as these and examples are given in Fig 10a and 10b.
The regional commissions do not as a rule collect and publish information on disposition of catches, production of fishery commodities and employment. Very few economic analyses are carried out for the commissions although they can usually be handled by the same machinery as is used for biological studies. Detailed economic studies are generally conducted at the national or lower level. International trade and commodity data are compiled annually from data, provided by national statistical agencies and published by FAO in the Yearbook of Fishery Statistics.