A conversion factor is a numerical co-efficient used to convert the weight of one product form to that of another: most often this is to round or whole weight. In its most basic form the conversion factor for a particular product form is determined by dividing the weight of the processed fish into the weight of the fish before it was processed. The eastern Canadian Observer Programs have conducted numerous studies as a result of identification of problems stemming from inaccurate conversion factors. As a result the factor for headed and gutted silver hake was adjusted upward from 1.47 to 1.6, for skinless cod fillets from 2.81 to 3.2 and for skinless trimmed cod fillets from 2.81 to 3.7. The importance of the accuracy of the conversion factor is stressed to the observer as well as the importance of proper processing techniques, i.e. a conversion factor is not accurate if the processing technique is not on par with the technique used when determining the conversion factor.
Since the official catch and effort reporting system is based on converted product weights, (line 53 of the International Fishing Log requires the total of “round weight processed today for human consumption” and “roundweight processed for reduction”) the importance of accurate conversion factors cannot be overemphasized. Observers are instructed to monitor production methods carefully and to administer ad-hoc conversion factor analysis when they suspect a problem. If a serious misreporting situation is identified via this route then a full scale study would be launched.
On commercial offshore trawlers it is not practical to weigh catches of fish in the round form, hence, conversion factors are most useful in deriving whole weight estimates of the catch from the stored product weights or volumes. For example, the weight of blocks of various products are converted to round weight by the production foreman on factory trawlers in order to provide estimates of the original catch.
The accuracy of catch figures derived from production weights depends on the precision of the product to whole weight conversion factor used, the accuracy of the average product block weight and the accuracy in counting the number of product blocks. Usually, accurate block counts are required by the fishing companies for their own records but when this is not done, volume of the total blocks divided by the average volume of an individual block will provide a block count estimate. With this in mind a system on ongoing analyses was designed to examine the first two problems. Experimental procedures have been designed and are carried out by observers using the actual products produced at sea.
There are a number of factors which affect the amount of weight loss during processing, and a combination of these lead to high variability in conversion factors from run to run. Relatively little is known about yield levels attained in production operations at sea. Size and condition of fish, size of the catch, specific subprocess methods, differences in individual fish cutting techniques (hand processing) and different machine types are some of the factors that can lead to the observed differences in conversion factors.
* Kulka and Firth (1985)
The most important of the above mentioned effects are elaborated to provide a background to understand conversion factor experimental strategy.
Production is very complex and often in a single factory several species may each be processed into several types of products. Different machines or subprocesses may be used on a single product category. For example, fillets with skin on or off, trimmed fillets or V-fillets may be produced from the same batch of fish. Each of the subprocesses could have different conversion factors due to differential removal of body parts. Whether a fillet is produced by hand or from a machine (or specific type of machine) can also have an effect. Machines are usually more wasteful and certain models may be more efficient than the others. Size of fish in a particular catch can also have an effect. For example, smaller fish passed through a machine that was designed for larger specimens could cause a reduction in yield. These are only a few examples of how yield can be affected.
The following is a step by step description of methods used to perform a conversion factor experiment. The methods outlined pay particular attention to practical detail and it is this approach which is the key to gathering realistic data from uncontrolled conditions encountered in ships' factories.
Prior to performing any experiments, it is important to gather relevant data pertaining to the factory operation. Specifications for machines used and products produced should be noted and recorded. It is particularly important to observe and define the production subprocesses and the extent to which each is used. For example, if 30% of cod are produced as skinless boneless fillet and the rest is skinless product for a particular set, then this 70/30 ratio must be recorded.
The handling of various species can differ quite significantly between countries, fleets, or even individual vessels. To avoid confusion, each process encountered should be documented in the report and accompanied by diagrams of fish showing such details as cut angles and extent of trimming. Factors that might adversely affect product yield should also be recorded. These may range from quality of fish, biological influences such as gonad size or fish size, to quality and maintenance of equipment (poor quality leading to excessive waste) and the extent of hand trimming performed. Finally, a record must be kept of the conversion factors in current use by the vessel and their origin.
Once practical aspects such as sample site selection are worked out, consideration can be given to the actual experimental steps. In performing an experiment, the aim is to extract a normal batch of whole fish from the catch, weigh them, turn them over to the production crew for processing in the normal manner and then reweigh the product. This process must reflect as much as possible the average production sequence. Whole weight of the fish divided by product weight provides the estimate of conversion factor. At the sample selection stage, in order to simulate actual factory conditions, the sample can be selected by factory personnel. In theory, if the aim is to determine an average factor for a vessel, random size samples of a single species should be collected. A processing machine, however, is restricted in the size of fish that it can process. With a sample of all sizes of fish taken from a catch, it is possible the size of some fish may fall outside the processing efficiency limits of the machine. When this occurs, the under or oversized fish will have to be removed, leaving a preselected size range conforming to the machinery. Forcing improperly sized fish through a processor would produce uncharacteristic results. Complete randomness is not preserved by this subsampling method for testing of a particular machine but it does simulate actual conditions. However, the machinery can often accommodate a wide size range and outsizing tends to be a minor problem.
Most vessels carry several types of machines for a single process that together can handle the full size of fish being caught. In order to determine an average conversion factor, regardless of machine or combination of machines used for processing, the randomly selected sample must be partitioned into size strata conforming to the size range capabilities of each of the machines being used. Each component of the original sample can then be put through the appropriate processing machine. For this approach, the experimenter would have to select a sufficiently large sample (about 250 kg) such that there are enough fish in the smallest component. Sending an insufficient number of fish through a machine could produce unreliable results given the considerable variability in the position of flesh cuts and subsequent proportion of parts removed. The sample components, once processed, can be recombined and weighed. This multimachine approach has two drawbacks. First it does not illustrate the differences between machines. Second, two or possibly three machines would have to be monitored simultaneously. Considerable preparation, attention to detail and a team of two would be necessary to perform this type of operation, hence it would be used only in special cases. Alternatively, a series of independent size selected samples conforming to each of the various machine types can be processed individually and then post-weighed according to the actual size distribution in the catch. This is the most logical approach for the singly-deployed observer. Both of the above methods will result in an average conversion factor for the size of fish taken; the second will yield data differences between machines.
A second and more specialized type of experiment would involve choosing a range of size selected samples in order to test the effect of fish size on magnitude of conversion factor. Each sample would comprise a very narrow range of fish size. Since overall variance is quite high, relatively large numbers of samples are necessary in order to obtain a clear answer with respect to correlation between size of fish and size of conversion factor over the range of a particular machine or over the range of all sizes caught. Hence, only when there are two observers per vessel can this type of experiment be conducted.
For the sake of portability, it is necessary to use light, compact scales. Spring dial models of 100 kg capacity, similar to those used on research vessels, are adequate. Heavier but still portable triple beam balances may provide superior measures as they are affected less by vertical movement due to sea swell. Some accuracy in sample weights is sacrificed using portable models but practical matters such as sea transfers dictate that they be used. To overcome part of the problem, a second scale might be used to verify the first reading. Ships scales can often be used for this purpose since they are usually quite accurate.
Samples thus obtained (random or size selected and weighing about 250 kg) must be carefully measured and counted. The preliminary length and weight data can then be recorded with modifications on a standard length frequency sheet.
When measuring fish for a conversion factor experiment, it is not necessary to remove otoliths or sex the fish but the normal “cm” grouping should be used. After completing the length frequency, calculate mean length of the sample. This information will later be transferred to the summary sheet.
Once the appropriate morphometric data have been recorded on the length frequency sheet, the production equipment selected for the experiment should be cleared of fish and the sample can then be passed through the machinery. It is important to monitor each stage carefully making sure no fish are added or removed on purpose. Any occurrences leading to abnormal yield should be recorded. Careful monitoring procedures and attention to detail are essential through all stages of the operation.
Since loss of some product units (e.g. fillets) during normal processing is not unusual, it is important to count not only whole fish in the sample prior to processing but also the product units as they emerge from each stage. For example, if 80 fish comprise the sample, the count of fillet product must not exceed 160. Some fillets are often lost during the processing, therefore, number lost should be recorded for each experiment. This ensures that error will not be introduced into the results due to abnormal product loss or gain. Often, practical problems related to processing by the crew arise and can affect the outcome of the experiments. Any abnormalities in processing procedures should be noted. Ensure that no special measures are taken by the crew during the experiment, such as tuning of machinery or more careful trimming procedures. The aim is to obtain an average value reflecting typical product yield rather than a machine optimum factor. If there are intermediate processes such as heading and gutting before filleting, the intermediate product weights should be recorded. A separate experiment is thus created for each product substage. In this way, experimental efficiency can be maximized. The final step is to divide whole weight by product weight for each state to yield the conversion factor estimate for a particular experiment.
When the data records are complete, the raw data, sample weight, counts, and mean size of fish (unsexed) can then be summarized onto the Conversion Factor Data Summary Sheet (Appendix J). Included in the header and main body is the following information: vessel name, vessel identifier (side number), species processed, flag state of vessel (country code), sample number (consecutive sequence), set number, date, process method, whole weight of fish (kg), product weight (kg), number of fish in the sample, type of processing machinery, the area fished (NAFO area), and type of sample (i.e. random, size selected). An additional category called “problem type” may be included to identify an experiment which was detrimentally affected (problem codes are included in Appendix J). The following describes step by step how to code each category below the header.
Trip: Fill in assigned trip number.
Side #: Fill in side number of vessel.
Species Code: Code in species.
Country: Code country of vessel with appropriate code from Appendix A.
Sample No.: Number the experiments consecutively.
Set No.: Code in actual set number from which the experimental fish were extracted. This information can be obtained from the appropriate set and catch detail sheet.
Date: Code actual year, month and day for the set from which the experimental fish were taken.
Process Method: Refer to Appendix G, for a list of processing codes.
Whole Weight: Write in weight of whole fish used for the experiment, to the nearest kilogram. Be sure to subtract basket weight.
Product Weight: The weight of fish product (kg) obtained from original sample of whole fish.
Conversion Factor: This number is obtained by dividing the whole weight of the sample by the resulting product weight.
All factors are rounded to two decimal places. For example, 1.624 is coded as 1.62.
In summary, the following steps briefly summarize experimental procedures which will lead to a completed summary sheet:
| 1. | Analyse the production layout in the factory and decide exactly how to obtain the sample. Note how many products are produced in the factory operation and distribute sampling effort appropriately. | |
| 2. | Select and weigh a group of fish. Record numbers, length and weight of whole fish on the length frequency sheet. | |
| 3. | Check that the machine used for the experiment is completely clear of fish from the regular production before proceeding with the experiment. Note any lost product and insure that fish not included in the original sample do not become part of the product. The sample should be processed in the normal manner (i.e. with no special care taken or machine adjustments made to artificially improve yield). Distribute experiments over various shifts to allow for variation between factory personnel. The aim of the experiments is not to obtain values of optimum yield but rather an average figure reflecting typical factory conditions. | |
| 4. | Count and weigh the product from the sample after the fish have passed through the machine. For a serial experiment record weight at each subprocess stage. For fillets, product weight would be recorded after heading, filleting, skinning and trimming. Such serial experiments are useful in determining yield reduction attributable to each subprocessing stage. | |
| 5. | Record on the Conversion Factor Data Summary (Appendix J) the relevant data. Signify problem type and provide comments for any unusual occurences. |
