| PANEL | 7 | - | MARK-RECAPTURE METHODS |
| GROUPE | METHODES DE MARQUE-RECAPTURE |
by/par
J.E. Thorpe
Freshwater Fisheries Laboratory
Pitlochry
Scotland/Ecosse
ABSTRACT
Mark-recapture methods depend on the randomness of sampling relative to the population being sampled, and to the constancy of behaviour of marked and unmarked fish. A number of statistical tests have been developed to account for differences in these factors. The use of the method is described for several areas and in general confidence limits of estimated population number at the 95 percent level range from 1–150 percent of the estimated population. The method is costly as it is additional to other catching methods but may be a useful method of calibration for these methods.
RESUME
Les méthodes de récupération des marques dépendent de la mesure dans laquelle l'échantillonnage est aléatoire par rapport à la population échantillonnée, ainsi que de la constance du comportement des poissons, marqués ou non. Un certain nombre d'outils statistiques ont été élaborés pour tenir compte des modifications de ces facteurs. L'utilisation de la méthode est décrite pour différents secteurs et, dans l'ensemble, les limites de confiance des populations ainsi évalués, de l'ordre de 95 pour cent, varient entre 1 et 150 pour cent de la population intéressée. La méthode est coûteuse étant donné qu'elle vient s'ajouter à d'autres méthodes de capture, mais elle permet d'étalonner ces autres méthodes.
CONTENTS
1. DESCRIPTION OF THE METHOD
2. USE OF METHOD
3. TYPE OF DATA OBTAINED
4. SUMMARY OF ADVANTAGES AND DISADVANTAGES
REFERENCES
The mark-recapture method for estimating the size of a population of fishes consists in releasing a known number of marked individuals, and recording recaptures in a subsequent sample. It is assumed that the fraction of the marked population recovered in the sample is equal to the fraction of the total population sampled.
According to Schaefer the principle was first applied by Laplace in 1783 to estimate the total population of France, using the register of births for the whole country and the births among the population of a restricted area. Marketing fish for individual recognition seems to be a more ancient practice. Kaiser Wilhelm II of Germany is recorded as having attached an engraved brass ring to pike at a festival on the Böcklinger See in the year 1230. The authenticity of the record is a little uncertain as the pike is said to have been recaptured in 1497, 267 years after release.
In 1889 Petersen reported a marketing experiment to determine the mortality rate of plaice, and about 20 years later Dahl marked brown trout in a small Norwegian lake. He cut off their adipose fins and redistributed the fish about the lake before seining a second time to determine what fraction of the population was marked. Since Dahl's day, the method has been elaborated and refined by many fishery workers and reviews of these developments have been published. For present purposes, in considering the advantages and disadvantages of the method as a technique for stock assessment, two general conditions for the validity of estimates must be noted. These are that:
either the marketing or the subsequent sampling should be at random with respect to the population being estimated;
marking should not alter the behaviour or the life expectancy of the fish in any way. It is unlikely that these ideals will be achieved in practice and consequently much ingenuity has been devoted to devising corrections to overcome distortions to the estimates obtained.
Condition (i): the representativeness of samples may be affected by selectivity of gear, and by time and place of sampling. By supplementary experiments allowance can be made for gear selectivity, and errors introduced due to time and place of sampling can be reduced by stratification; for example, by age, or by size.
Condition (ii): the method depends on marked and unmarked fish having the same probability of death so that the fraction of the population which is marked remains constant. Any increase in the mortality rate of marked fish relative to that of unmarked fish reduces this fraction and leads to an overestimate of the population. Such differential mortality may occur as a result of handling, in which case the effect can be corrected for by holding groups of marked and unmarked control fish captive after the marking period and comparing death rates. Also, differential mortality may be a long-term process if the marked individuals are at some slight disadvantage; in this case the ratio of marked fish in sub- sequent samples will decrease and, by extrapolation back to the release date, the true population ratio is obtained. Similarly, if tags are lost, or if the recovery of marked fishes is not reported, the effect is the same and the population is overestimated. Tag losses may be estimated by double marking experiments and incomplete reporting by periodic checks on the catch or by introducing known fish into the catch.
All these features of mark-recapture experiments have been illustrated in papers presented at this Symposium. As to the representativeness of samples, Jensen (see p.603) used gillnets to capture brown trout at two lakes in Norway. He determined the selectivity of these nets and adjusted his catch data accordingly. Raymond and Collins (see p.552) and Shearer (see p.571) with juvenile salmonids, tested representativeness in samples of down- stream migrants caught in traps by releasing groups of marked fish upstream again, at each bank and across the width of the river respectively. No differences were found in the recapture rates of fish from the different release sites. Similarly I myself compared the angling recapture rates of tagged trout released at eight different sites on Loch Leven, and again found no differences between sites. Vostradovsky (see p.651) tagged six species of fish in Lipno Reservoir, and found them all to be very restricted in their movements, so that to achieve widespread distribution of tagged fish among the populations of any of these it would be necessary to tag at many sites all over the reservoir. But in many instances restricted distribution of fish can be turned to the experimenter's advantage, since it may then be easier to obtain large enough samples for marketing or recapture to reduce statistical bias introduced into estimates. Craig (see p.502) and Kipling and Le Cren (see p.611) found wire traps useful for catching large numbers of perch but these catches were selective and, if traps were also used for the recapture of marked fish, biased and unrealistic estimates were obtained. However, Jensen (see p.600) also used wire traps to catch spawning perch and found that estimates based on a multiple mark-recapture experiment were satisfactory when compared with others obtained by retenoning later. Some workers have pointed out the need to be aware of particular activity periods of fish and of their differing vulnerability to capture at different seasons, a feature which I found important in assessing trout populations in Loch Leven. Recruitment of fish to the vulnerable stock during the experiments was allowed for by Jensen (see p.603) and by Kipling and Le Cren (see p.611) who used independent measures of growth to exclude such recruits from their analysis. Holcik and Pivnicka (see p.656) have shown the need to take large samples when the populations under study are large, in their case roach and perch in the Klicava Reservoir. Inadequate sample size (together with high vulnerability of tagged fish in gillnets) led to estimates which increased progressively with time. They overcame this problem by using the Schnabel multiple mark-recapture method in which the proportion of marked fish in the stock is continually augmented and the estimate continually refined.
As to differential mortality, that occurring in the short term was tested by several workers and shown to be absent or of negligible importance in most cases although it might be influenced by water temperature or by the use of narcotics as Koops (see p.620) found with burbot. Long-term differential mortality was expected to occur in Jensen's trout experiments where his fish were marked with Carlin tags which were expected to increase their vulnerability to capture in gillnets, as Steiner and Pechlaner (see p.672) found with Floy-tagged char, but the estimates agreed so closely with those from independent methods that this source of error was eliminated. Tag losses were checked by Kipling and Le Cren (see p.611) and by Shearer (see p.571) who fin-clipped their fish in addition to attaching tags and I myself used an injected dye-spot as the second mark. Incomplete reporting of tags was avoided by many workers who used their own staff to obtain recapture samples. Trefethen and Collins (see p.490) used an elegant automatic sorting technique to recover salmonids tagged with coded magnetic wire as these fish passed through a trap.
The problems of sample size and the precision of estimates associated with mark-recapture experiments have been reviewed by Robson and Regier (1971) and others. When using the simple Petersen ratio populations will tend to be underestimated but this bias is reduced with increasing sample size. The bias may also be reduced by several modifications of the basic estimator, the one most commonly used in the contributions to this Symposium being that of Bailey (1951):
Confidence limits for these estimates depend on the proportion of marked fish recovered and Robson and Regier have published diagrams relating sample size to given expectations of recoveries at specific levels of confidence.
In the present set of papers, confidence limits at the 95 percent level range from 1–150 percent of the estimated population number. This seems to be the range of accuracy customarily found in such experiments; in practice the limit being imposed by the amount of effort that can be expended in the tagging operation or in the recapture sampling. Multiple mark-recapture experiments increase in accuracy with successive samples but as Kozikowska noted it is often difficult to define the population that is being estimated.
The method is relatively costly as it is additional to one or more of the basic sampling techniques already described and it is restricted to the assessment of individual species.
This Symposium was conceived in the context of a need to debate the impact of environmental changes on fish stocks. We wish to be able to detect this impact simply, rapidly, and reliably, and we have considered many sampling methods. All require validation and the efficiency of the gear used needs to be measured.
Mark-recapture methods have been used explicitly to determine the efficiency of traps as instruments to measure the size of migrating salmonid populations. Lamarque (see p.173) mentioned the use of Petersen experiments to check the efficiency of electric fishing assessments. I suggest that the method can also be used to calibrate the capture instruments in still waters and then conditions can be defined for simple, rapid and cheap assessments of relative abundance thereafter. I have tried to show one way in which this may be done in my paper on Loch Leven trout (Thorpe, see p. 641). In this instance, a seining method is identified which produces the required results with less than 10 percent of the previous effort, no reward costs and much reduced laboratory analysis.
However, because of its intricacies, and the variety of subsidiary checks that must be made, it seems likely that its use will be restricted, as it is as present, to populations of particularly valuable species.
