A REVIEW OF THE BIOLOGY AND FISHERIES FOR SKIPJACK TUNA, KATSUWONUS PELAMIS, IN THE PACIFIC OCEAN

Alex Wild
Inter-American Tropical Tuna Commission
La Jolla, California

and

John Hampton
South Pacific Commission
Noumea, New Caledonia

1. INTRODUCTION

This paper is intended as a review of the biology, resource, and fisheries associated with skipjack tuna, Katsuwonus pelamis, in the Pacific Ocean. Because of the distribution of the fisheries, it has been convenient in several sections to discuss skipjack in the eastern Pacific Ocean (EPO) and western and central Pacific Ocean (WPO) separately. This division may, for the purpose of these discussions, be taken to occur at approximately 150°W. Certain sections contain information from the Indian and Atlantic Oceans in order to offer a wider perspective. The task of compiling the material was simplified considerably through the extensive use of three major sources of information: the synopses of biological data on skipjack by Forsbergh (1980) and Matsumoto et al. (1984), and a study on the environmental factors affecting their apparent abundance (Forsbergh, 1989). In Section 13, the material that refers to the EPO is intended to be both a general and an interaction-specific review for the purse-seine and baitboat fisheries.

2. CLASSIFICATION

The family Scombridae is composed of 15 genera comprising 49 species of epipelagic marine fishes, which include the mackerels, Spanish mackerels, bonitos and tunas (Collette and Nauen, 1983). The family is divisible into two subfamily categories, the Gasterochismatinae, which contains the single species Gasterochisma melampus, and the Scombrinae. On the basis of internal osteological characters the Scombrinae in turn have been divided into two groups of tribes: the primitive mackerels and Spanish mackerels, and the more advanced group consisting of the bonitos (Sardini) and tunas (Thunnini). Among other characteristics, the Sardini differ from the Thunnini in that the latter possess a subcutaneous, counter-current vascular system that permits members of the tribe to be warmer than the surrounding water (Collette et al., 1984). In this respect the four genera of the Thunnini tribe are unique relative to other bony fishes. The three more primitive members of the tribe (Auxis, Euthynnus, and Katsuwonus), together with the yellowfin group (Thunnus), have central as well as lateral heat exchangers. The diagram below indicates the position of skipjack tuna within the family Scombridae after following Lindberg's (1971) scheme to the family level:

Phylum Chordata
    Subphylum Vertebrata
        Superclass Gnathostomata
            Class Teleostomi
                Subclass Actinopterygii
                    Order Perciformes
                        Suborder Scombroidei
                            Family Scombridae
                                Sub-family Scombrinae
                                    Tribe Thunnini
                                        Genus Katsuwonus
                                            Species pelamis

The classification of skipjack tuna in particular, and in relation to other tunas, is discussed by Matsumoto et al. (1984) and Collette et al. (1984), respectively.

3. EARLY LIFE HISTORY

Ripe skipjack ovarian eggs are spherical, smooth, transparent, and usually contain a single oil droplet (Brock, 1954; Yabe, 1954; Yoshida, 1966). The eggs are 0.80 to 1.17 mm in diameter with variable-sized oil droplets ranging from 0.22 to 0.45 mm. Because the size and appearance of the eggs of artificially-reared skipjack are similar to those of other artificially-reared scombrids (Harada et al. 1971; Ueyanagi et al., 1973), it is not possible to identify those of skipjack in field collections of tuna eggs.

Net tows by Japanese research vessels recorded the distribution (Figure 1) of skipjack larvae in the world oceans (Ueyanagi, 1969; Nishikawa et al., 1985). In particular, in the Pacific they are found near 35°N off Japan and as far south as 37°S off the southeastern part of Australia. This wide distribution is maintained eastwardly until slightly past the Hawaiian Islands in the northern hemisphere and the Society Islands to the south. Matsumoto et al. (1984) concluded that, “… the distribution then narrows abruptly toward the equator at about 145°W and remains within 10°–15° of latitude on either side of the equator to the Central and South American coasts.” Klawe (1963) found few larvae in the eastern Pacific, while Ueyanagi (1969) and Ueyanagi et al. (1969) showed that they appeared to be concentrated in the equatorial region with an increase in abundance from east to west. After adjusting for differences between net tows, Matsumoto (1975) found that the greatest concentration occurred between 160°E and 140°W. Recent work by Leis et al. (1991) identified particularly high concentrations of skipjack and Thunnus spp. larvae in the vicinity of coral reefs of the islands of Moorea, Rangiora, and Takapoto in French Polynesia.

Skipjack larval distribution is strongly influenced by temperature. Forsbergh (1989) showed that the concentration of larvae in the Pacific approximately doubles with each 1°C increase in the sea-surface temperature from 24°–29°C, and decreases by a variable amount after 30°C. However, it is not known whether the relationship is brought about by the larvae, the spawners, or both. The larvae are therefore concentrated in the tropics, although the warm, poleward-flowing Kuroshio and East Australian Currents facilitate a seasonal (summer) expansion of the larval distribution into subtropical waters in the far WPO.

Figure 1. Distribution of skipjack tuna larvae, 1952–1975. (After Matsumoto, et al., 1984).

In terms of depth, Klawe (1963) did not find any larvae below the mixed layer in the eastern Pacific, but there is a suggestion in Strasburg's (1960) closing-net sampling data from 11°N–5°S, 140°W that this may not always be the case. Upon examining Japanese data, Forsbergh (1980) concluded that 22 percent of the larvae were captured in surface tows and 55 percent in subsurface tows. The larvae are scarce at the surface during the day, and although they are more common there at night, there are still fewer than at deeper levels (Ueyanagi, 1969). Richards and Simmons (1971) found that, in the northwestern Gulf of Guinea and off Sierra Leone, the larvae of yellowfin (T. albacares) and bigeye (T. obesus) tunas migrate to the surface during the day while skipjack larvae migrate to the surface at night. In contrast to yellowfin, the swimbladder in skipjack larvae degenerates before the fish reaches a size of two cm (Richards and Dove, 1971).

Most juvenile skipjack have been collected from the stomachs of larger tunas and billfishes, from experimental mid-water trawls and by dipnetting at night under a light (Forsbergh, 1980). Higgins (1967) examined the records of 518 juveniles (1.2–30 cm) caught in the Pacific in the period from 1916 to 1966. Of the 372 caught in the WPO west of 180°, the majority were found in tuna stomachs; 135 were found in the central Pacific (180°-140°W) in either midwater trawls or stomachs, and 11 were caught by dipnetting in the EPO east of 140°W. Mori (1972) defined juvenile skipjack as those less than 15 cm long, and young skipjack as those between 15 and 35 cm. He examined the data from 3,778 juveniles and young recovered from the stomachs of billfishes caught by longlines in the Pacific and concluded, together with Yoshida (1971), that the distribution of juveniles was similar to that of the larvae. The apparent abundance of the juveniles was greatest west of 180° between 10°N and 10°S, and least east of 120°W. They were also more prevalent in areas with surface waters warmer than 24°C. On the other hand, skipjack young were more widely distributed and more common in areas with surface temperatures less than 24°C. Based on the distribution of baitboat-caught skipjack with juvenile skipjack in their stomachs, Argue et al. (1983) concluded that, in the vast area of the central Pacific, the abundance of juveniles was greatest during October–March between the equator and 25°S in two broad geographical areas, one including eastern Polynesia (130°–150°W) and the other adjacent to Papua New Guinea, the Solomon Islands and Vanuatu (140°–170°E).

4. AGE AND GROWTH

Historically, two methods have been applied to the problem of directly estimating the ages of skipjack tunas. These are the use of growth rings, or marks, on vertebrae (Aikawa, 1937; Chi and Yang, 1973; Sosa-Nishizaki et al., 1989) or dorsal spines (Shabotiniets, 1968; Batts, 1972a; Chur and Zharov, 1983; Sosa-Nishizaki et al., 1989) and the use of daily increments on otoliths (Uchiyama and Struhsaker, 1981). In their review of these methods, Josse et al. (1979) pointed out that estimates prepared from vertebrae and spines were unreliable because the results cannot be verified. Conflicting interpretations could therefore not be resolved, such as the view that two vertebral rings were deposited each year (Chi and Yang, 1973) rather than one (Aikawa, 1937). On the other hand, otolith increment counts were regarded as more reliable, provided that the sequence of daily increments did not contain any interruptions. This reservation is critical with respect to skipjack tuna. In the eastern Pacific, Wild and Foreman (1980) found that 21 out of 26 skipjack failed to deposit an increment (I) each day (d) during their period at liberty in a mark-recapture experiment conducted at sea. The fish were initially captured, tagged, injected with oxytetracycline in order to label their otoliths, and released. For fish in the length range of 42–64 cm, the average deposition rate was only 0.76 I/d (range: 0.44–1.09 I/d) during periods of liberty from 17–249 days. These results contrast with the expected rate of one I/d and the rates for other tunas, i.e.,

The low deposition rate for skipjack could have resulted from counting increments in the wrong location on the otolith, the sex of the fish, or because some of the increments were too small to be detected by the light microscope. Therefore, on fish recovered from a second tetracycline-injection experiment carried out in 1980–81, additional counting sites and the enhanced magnification (3,000–3,500x) of the scanning electron microscope were applied to the otoliths of each sex. Despite these refinements, the deposition rates for the sexes did not differ, and their pooled average rate of 0.83 I/d (n = 41) was not significantly different from the previous result (Wild and Wexler, IATTC; unpubl. data). Consequently, otolith increments are presently unreliable for estimating the age of skipjack tuna in the eastern Pacific. A count of the total number of increments on an otolith will underestimate the actual age and overestimate the growth rate.

For skipjack in the central Pacific it is known that increments are deposited daily for at least the first five days after hatching (Radtke, 1983). To develop their growth model for the central region, Uchiyama and Struhsaker (1981) concluded that captive skipjack (n = 4) in the length range of 45–48 cm deposited daily increments over a period of 5–30 days following the onset of satiation feeding. A check mark on the otolith, induced by the stress of capture and adjustment to the holding tanks, was used to locate the beginning of the period of deposition of wider increments caused by full feeding. Because such stress marks are difficult to distinguish from other, similar marks on the otolith, the authors noted that the assumption of daily deposition would eventually have to be validated. With this proviso, the growth-equation parameters and the sizes-at-age of skipjack in the central Pacific are recorded in Table 1. Selected examples of growth curves developed for the Pacific Ocean also appear in Figure 2.

The measurement of skipjack growth is seemingly a more tractable problem than direct age determination, but the two most popular methods of measuring growth, tagrecapture experiments and modal progression of length frequencies, are not without limitations. Joseph and Calkins (1969), for example, drew attention to the subjective aspect of modal analysis after applying the method to skipjack in the eastern Pacific. Based on data collected from the South Pacific, Josse et al. (1979) also characterized the method as least reliable because of the uncertainty of demonstrating a “…clear progression of modes for more than a few months. Apparent mode progressions may give in a single region, in different years, growths which are rapid, slow, nil and even negative.” With respect to tagging, one of the most persistent and troublesome questions is whether the procedure affects subsequent growth. The rapid buildup of blood lactate due to the stress of capture and tagging probably increases mortality (Barrett and Connor, 1962), and tagging has been shown to reduce condition in southern bluefin tuna (T. maccoyii) for a short period after tagging (Hampton, 1986). However, the longterm effect on tuna growth is unknown. Schaefer (1961) concluded that, compared to untagged fish in modal progressions, the growth rate in the EPO of yellowfin tagged with loop tags was somewhat lower and their attrition rate higher. More recently, the average growth rate of untagged yellowfin, caught north of the equator and in the length range from 40–135 cm, was estimated from otolith increments to be 1.06 mm/d with a 95-percent confidence interval of 1.03–1.08 mm/d (Wild, 1986). For yellowfin tagged with dart tags, Bayliff (1988) found that their overall growth rate in the EPO was 0.85 mm/d (variance: 0.07). Considering that skipjack are more sensitive to handling than yellowfin, it is unlikely that their response to tagging would be dissimilar (Forsbergh, 1980). Through simulation studies, Sibert et al. (1983) also noted the effect of measurement errors on the parameter estimates of the equation (von Bertalanffy, 1938) used most frequently to analyze tagging data, i.e.,

where l₂ is the length at recapture, l₁ is the length at tagging, L∞ is the asymptotic size, K is a growth parameter and Δt is the elapsed time. An increase in the magnitude of the error in l₂ increased the confidence limits of the parameter estimates of L∞ and K, but they were unbiased. However, small changes in Δt or the least reliable length measurement, l₁, caused large changes in estimates of L∞ and K, and frequently their joint confidence region excluded their true values.

TABLE 1. Growth parameters of the von Bertalanffy (1938) equation for skipjack from major oceanic regions, and estimates of lengths at different ages prepared by various methods. Bracketed lengths represent extrapolations of the data. The range of the L and K values for the South Pacific include the island nations affiliated with the South Pacific Commission as well as New Zealand. Legend: f= fixed parameter; r = relative sizes; a = average.

Figure 2. Examples of three, von Bertalanffy-type growth curves for skipjack tuna. Curve: 1) from tagging data in the EPO from Bayliff (1988); 2) from otolith increments in the central Pacific from Uchiyama and Struhsaker (1981); and 3) based on average parameter estimates derived from tagging data from the South Pacific Commission area. (After Forsbergh, 1989).

From the foregoing remarks it is clear that, in the absence of a validated method of determining the age of skipjack, the predicted sizes-at-age (Table 1 and Figure 2) are approximations. This conclusion, drawn earlier by Matsumoto et al. (1984) remains unaltered at present. Nevertheless, several comments can be made on skipjack growth, and the first one concerns its variability. Sibert et al. (1983) found that there were significant differences in the amount of growth and the parameters of the von Bertalanffy (1938) equation estimated for the same regions at different times (see Table 1). The differences between regions were also significant. Therefore, these variations in growth and unstable parameters were likely due to the temporal and spatial variation of environmental conditions. The broad range and values of K, i.e., 1.54–3.13, in the Senegal and Cap Vert regions (Cayré et al., 1986) also emphasize the variability in growth over short-term recovery periods (≤ 60 d). Based on the previous comment that a short (1 cm) error in the release length of tagged fish can seriously bias parameter estimates, the results of many of the earlier experiments, e.g., Schaefer et al., 1961; Rothschild, 1967; Joseph and Calkins, 1969, may be compromised because the fish were measured only to the closest 5 cm. Bayliff (1988) eliminated such data from his study in the EPO and found that, overall, skipjack east of 100°W grew less rapidly than those west of 100°W. Provided that there are no interruptions in the record of otolith increments, the growth curves in the central (Uchiyama and Struhsaker, 1981) and eastern Pacific (Bayliff, 1988) appear to be similar (Figure 2), but the average growth rate is more rapid during the early stages in the western Pacific (Sibert et al., 1983). The data in Table 1 represent but a small selection of the available information; a more extensive list appears in Bayliff (1988). The difficulties in estimating skipjack age, and in relating age to length, have substantial implications for the assessment of stocks. In view of these difficulties, the application of traditional age-structured models may be limited for skipjack, and alternative methods of assessment may be required.

The length-weight relationships in Table 2 are reproduced, in part, from Matsumoto et al. (1984), following the original outline by Nakamura and Uchiyama (1966). To facilitate parameter comparisons, all lengths and weights are expressed in centimeters and kilograms, respectively. The detailed statistics needed to set confidence limits about selected sizes are generally not available in the literature. It appears however that, with the exceptions of Evans et al. (1978) and Batts (1972a), the weights of 40-, 50- and 60-cm skipjack from different oceanic regions are similar. Skipjack from the western Atlantic (Batts, 1972a) in the 50–60 cm range appear to be relatively heavier than elsewhere and progressively so as length increases.

5. MATURATION, SPAWNING, AND SEX RATIO

5.1 Maturation and Fecundity

Matsumoto et al. (1984) concluded that the minimum size of female skipjack at maturity is 40 cm, and that first spawning may occur in fish between 40–45 cm, or larger. They based their interpretation on several sources of information in which, for the Pacific, “… Marr (1948) recorded skipjack tuna as small as 40 cm fork length with spent ovaries from the Marshall Islands; Wade (1950) recorded fish in the 40.0 to 40.9 cm size class with ripe and spent ovaries from Philippine waters, and reported … a female in the 34.0 to 34.9 cm size class having ripe ovaries; and Brock (1954) noted that the smallest fish that possessed maturing ova during the spawning season in Hawaiian waters were around 40 to 45 cm.” In the eastern Pacific, Schaefer and Orange (1956) estimated that the minimum size at maturity in the vicinity of the Revillagigedo Islands, offshore from Baja California, was 55 cm and nearer 50 cm off Central America. In the Indian Ocean, Raju (1964a) estimated the size at sexual maturity around Minicoy Island at about 40 to 45 cm. His conclusion was based on the observation that remnants of mature ova were present in fish above 40 cm, but none in fish below that size. Similarly, on the basis of fish with spawned or spent ovaries, Simmons (1969) reported that the minimum size at maturity in the Atlantic Ocean was 41 cm. Stequert (1976) also noted the first spawning of skipjack tuna at 41 to 43 cm off the northwest coast of Madagascar.

In contrast to this rather narrow range at first maturity, the fecundity of female skipjack is quite variable in terms of both oceanic region and fish size. For example, although 50-cm and 60-cm fish in the eastern Atlantic (Batts, 1972b) and Pacific (Joseph, 1963), respectively, produce nearly equal numbers of ova (300 × 10³), at a length of 70 cm the fecundity in the Pacific (970 × 10³) exceeds that in the Atlantic (600 × 10³) by a considerable margin (Figure 3). Within the Atlantic, increasing differences in fecundity with the size of fish are also demonstrated by regressions 4 and 5 in Figure 3. Caution is needed in interpreting this information, however, since Simmons' (1969) data represents only 13 fish from both sides of the Atlantic, whereas Batts' (1972b) data includes 30 fish from only the western Atlantic. The parameter estimates for most of the regressions in Figure 3 are reproduced in Table 3.

TABLE 2. Length-weight relationships for skipjack tuna. The parameters refer to the allometric relationship, weight = a(FL)^b, where FL is the fork length, or the distance from the tip of the upper jaw to the cartilaginous median of the caudal fork. This is the definition given by Marr and Schaefer (1949) to total length (TL), and in the table TL probably refers to FL, the most reasonable length measurement on tunas. (After Matsumoto, et al., 1984).

¹ Evans, L.C., A.D. uy and D.D. Tandog. 1978. The abundance, biology and distribution of tuna (Family Thunnidae) in Camiguin and nearby waters. Philippine Dept. Nat. Resour., Bureau Fish. Aquatic Resour., Prelim. Rep., 55 p.

Figure 3. Regressions of fecundity on length and observed fecundity estimates of skipjack tuna above 69 cm. 1) Joseph (1963), eastern Pacific; 2) Raju (1964b), Indian Ocean; 3) fecundity-length³, Raju (op. cit.); 4) Simmons (1969), eastern and western Atlantic combined; and 5) Batts (1972b), western ATlantic. (After Matsumoto et al., 1984).

TABLE 3. A comparison of fecundity-length and fecundity-weight relationships of skipjack tuna from various locations. The relationships are all of the form Y = a + b(FL)^x or Y = a + bW, where Y is the number of ova in thousands, FL is the fork length (mm) raised to the power x, and W (lbs) is the weight. The number in brackets after the relationship refers to the corresponding numbered curve in Figure 2. (After Matsumoto et al., 1984).

5.2 Spawning

“Gonadal studies in the Pacific, Atlantic and Indian Oceans indicated that skipjack tuna spawn throughout the year in tropical waters near the equator and from spring to fall in subtropical waters” (Matsumoto et al., 1984). The collection of spawning adults for such studies have taken place in association with commercial and experimental fisheries in diverse locations such as waters south of Japan, the Philippine, Marshall and Hawaiian Islands, New Caledonia, Marquesas-Tuamotu Archipelago, and off central America and Baja California. Since the larvae are generally found in waters with sea-surface temperatures > 24°C (Ueyanagi, 1969), the spawning period necessarily becomes shorter with increasing distance from the latitude of greatest temperature. As mentioned earlier (Section 3), the distribution of warm waters, and hence that of spawning and larvae, is affected by current transport. Warm waters extend to higher latitudes in the western Pacific because the water is transported poleward along the western shores. In the eastern Pacific, however, the warm water is restricted to lower latitudes because currents transport colder water toward the equator along the eastern shores (Forsbergh, 1989). On the basis of gonad indices, Naganuma (1979) decided that in the western Pacific between 15°N and 20°S smaller adults spawn during the warm season from October to March, and larger fish (> 60 cm) spawn north of 15°N in the northern summer. In the EPO, skipjack spawn off Central America at least during the late winter and spring months (Schaefer and Orange, 1956). Off Baja California near the Revillagigedo Islands, females with maturing ovaries were collected from April through November, although the main spawning season is indicated to be from July through November (Orange, 1961).

The use of gonad indices (GI), or the ratio of gonad weight to body weight, is a relatively rapid procedure to gauge the degree of maturity and it avoids the difficulty of measuring ova diameters. The accuracy of the index, however, has been questioned. Yoshida (1966), for example, found that the diameter was not related to GI at higher indices. Matsumoto et al. (1984) also concluded that the GI “…does not seem to be a sound measure of maturity of skipjack…”, because of the wide range of values for the same developmental stage. The index is also biased because larger females develop larger ovaries in proportion to their body weight than do smaller females (de Vlaming, 1982). Schaefer (1987) suggested that the GI be validated with histology or öocyte diameter and adjusted for the size of individuals if it is to be used to interpret reproductive activity correctly.

In the vicinity of the Philippines (Buñag, 1956) and in the Indian Ocean (Raju, 1964a), multiple spawning of skipjack has been indicated by following the development of ova-diameter modes over time. Earlier, Brock (1954) concluded that skipjack near Hawaii may spawn several times during a season. Multimodal distributions of ova diameters during the spawning season, and the absence of spawned-out fish until after the season were used to support this idea. Based on data from Brock (1949) and Buñag (1956), Matsumoto et al. (1984) estimated that skipjack in Hawaiian waters spawned at least twice during a year: in late April or early May and again at the earliest in late July or early August. The first spawning occurred after a 2-month development of ova from the immature to the ripe stage (April). Following a reversion of the ovaries to an intermediate mature stage, at least 6 to 7 weeks were then required for a second ripening in September. Thereafter, additional spawnings were not ruled out provided that the fish sought warmer waters. Histological studies of the ovaries of 87 skipjack captured in November and December in the South Pacific indicated that spawning occurred about every 1.18 days (Hunter et al., 1986). Therefore, it is possible that individuals may spawn repeatedly over the duration of a seasonal spawning period.

5.3 Sex Ratio

The sex ratios of skipjack tuna obtained from various locations display a broad range of values (Table 4). Matsumoto et al. (1984) concluded that, in general, fisheries that rely on young, immature fish have ratios dominated by females, whereas those that capture mostly older and spawning-age fish have sex ratios that are mostly male. Due to the uncertainty of spawning frequency and sex ratios, the variation in numbers of ova and fish size, the possible reduction in fecundity after successive spawnings (Joseph, 1963), and the lack of information on the size composition and abundance of spawning stocks, the total egg production of skipjack tuna would be difficult to estimate (Matsumoto et al., 1984).

6. STOCK STRUCTURE, DISTRIBUTION, AND MIGRATION

6.1 Subpopulations

Fujino et al. (1981) found significant differences in the serum naphthyl esterase (Es¹) allele frequencies between skipjack from the Indian and Atlantic Oceans, and between those from the Indian Ocean and the western Pacific (Figure 4). They proposed that skipjack originated in the Indian Ocean, and by dispersal to other oceans together with reproductive isolation, they formed a total of four subpopulations located in the Indian, Atlantic, western Pacific, and central-eastern Pacific. They further demonstrated that in the Pacific the Es¹ frequency remained relatively low and constant from 80°W to 175°E, and thereafter in the southern hemisphere it exhibited a “step structure” by increasing from about 0.50 at 175°E to about 0.67 at 160°E (Figure 5). Based on these results they concluded that the population structure, as evidenced by the trend in Es¹ frequencies, was discontinuous rather than clinal, thereby supporting Fujino's (1970a,b; 1972) earlier statements regarding the existence of at least two subpopulations in the Pacific. In the northern hemisphere between 140°E and 175°E the Es¹ frequencies were both high and low. This variability was also interpreted to mean that fish from the central-eastern subpopulation also occurred off the east coast of Japan (Forsbergh, 1988). Richardson's (1978) electrophoretic studies supported the idea of two major subpopulations, but he also proposed that a third exists off New Zealand. Sharp (1978) hypothesized that there were at least five genetically-different subpopulations in the Pacific.

Schaefer (1963) proposed that skipjack in the eastern Pacific originate in the central Pacific. Morphometrics, blood types, relatively few tag returns, and the fact that little spawning occurs in the fishing areas of the eastern Pacific were used to support this idea (Forsbergh, 1988). Rothschild (1965) and Joseph and Calkins (1969) also noted that skipjack were infrequently captured in an area of warm water off southern Mexico at about 15°N. Historically, the fish to the northwest and southwest of this area were thought to be from two different subpopulations. This idea was revised in favour of a single subpopulation following the capture of a large amount of skipjack in the area in late 1955 and in 1956 (Broadhead and Barrett, 1964), and to the west of the area between 10°–20°N in 1980 and 1981 (IATTC 1981, 1982). Events such as these demonstrated that the distribution of skipjack could be continuous throughout the area. It was further suggested that there is a single group of skipjack distributed in an arc-shaped area surrounding the region of warm water off southern Mexico (IATTC, 1983; Bayliff, 1984). By means of tag returns, the details of which appear in Bayliff (1988, Appendix), it was also evident that the fish at the ends of the distribution mix to at least some extent on the spawning grounds of the central and western Pacific (IATTC, 1984).

TABLE 4. Sex ratios for skipjack tuna from various locations. (After Matsumota et al., 1984).

Figure 4. Rejection limits, at 5% significance level, of serum esterase allele frequency distributions for the populations from the Atlantic, Indian and Pacific Oceans. The two rejection limits at the right represent those for the two subpopulations identified by Fujino (1972). (After Fujino et al., 1981).

Figure 5. Gene frequenciies for serum esterase alleles (Es¹) versus longitude. Small dots and open circles represent individual lots of samples from the northern and southern hemispheres, respectively (from Fujino et al., 1981). Crosses represent samples from the South Pacific Commission (1981a). Bars on the right are the 95% confidence limits of Fujino's (1972) western Pacific (upper) and central-eastern Pacific (lower) subpopulations. (After Hunter et al., 1986).

The problem of whether there is more than one subpopulation of skipjack in the Pacific was also considered during workshops sponsored by the SPC (SPC, 1980, 1981a). The material examined for this purpose included the results of tagging experiments and data on the gene frequencies of Es¹, serum transferrin and erythrocyte guanine deaminase (Gda¹) in skipjack blood. Because of the large geographic distances involved, the conclusion was drawn that skipjack in the Pacific are not panmictic because all adults do not have an equal opportunity of breeding with each other throughout their oceanic distribution within a generation. In contrast to panmixia, the distribution of the Es¹ data in Figure 5 was interpreted as either a continuous or a stepped cline, demonstrating that skipjack follow, “…some form of population structuring across the Pacific…” (SPC, 1981a). To describe this structure, two major hypotheses were recognized as the most feasible, and the first one was Fujino's (1972) discrete subpopulation model mentioned above. Although the workshop could not identify any genetically-isolated subgroups separated by stable geographic boundaries, neither could the extent of the neighbouring regions be defined on the basis of existing genetic data. Consequently, the possibility of discrete subpopulations that reflect a stepped cline could not be ruled out. The second, or the continuous cline hypothesis, accepts that Pacific skipjack belong to a single population, but the chance of any two fish breeding is inversely proportional to the distance separating them. An isolation-by-distance model is therefore an integral part of this hypothesis.

Richardson (1983) examined 42 loci for protein variation in skipjack tuna by utilizing material from previously unsampled sites in the South Pacific (Figure 6). Having found only one new polymorphism, he concluded that “…the discriminatory power of the genetic approach to the identification of skipjack subpopulations…is still very limited.” By combining these new samples with the previous results of Fujino (1972, 1976) and Sharp (1978), the important observation was made that “…the gene frequency for any well-studied stock in the western Pacific is heterogeneous.” Heterogeneity was detected within and between stocks in areas A and B of Figure 6, and the discrete-subpopulation hypothesis neither accounts for these results nor the dispersion of tagged skipjack between areas (SPC, 1981a). The mobility of skipjack and the genetic heterogeneity within a stock also signify a degree of overlapping that is compatible with Sharp's (1978) view of multiple subpopulations. No evidence of homing for breeding purposes has been detected, and the absence of this requirement of the discrete-subpopulation hypothesis further detracts from its acceptance. Samples collected in the New Zealand fishery from single schools over a four-year period also demonstrate heterogeneity for the Es¹ allele among and between years, and the allele frequency also declined significantly with an increase in the average size of fish in the school (Richardson and Habib, 1987).

The isolation-by-distance model proposes that skipjack exist in a series of semi-isolated genetic neighbourhoods, each of which encloses a group of randomly-breeding adults (Richardson, 1983). Although such a model is consistent with the observed heterogeneity within an area as well as the clinal changes in gene frequency, it is difficult to estimate reliably the size of a neighbourhood or to know whether the size is the same in different locations. Amongst several other untested assumptions noted by Richardson (1983), the model also assumes that there are no barriers to movement other than distance. Esterase and, to a lesser extent, Gda¹ clines of different magnitude exist to the north and south of the North Equatorial Counter Current, however, and this suggests a restriction in genetic flow between the regions (Richardson, 1983). It is also known that sea-surface temperature and salinity show gradients across the Pacific, and the longitudinal clines exhibited by Es¹ and Gda¹ may be related to such an environmental variable (Endler, 1977; Lewis, 1981; SPC, 1981a). Consequently, the difficulties that are encountered in applying either the isolation-by-distance or discrete-subpopulation hypotheses prevent the choice of a single, descriptive model of the skipjack population at this time (Richardson, 1983; IATTC, 1984).

Figure 6. Sampling stations and area boundaries used to study protein variation in Pacific skipjack tuna. (After Richardson, 1983).

6.2 Distribution

In general, skipjack occur within the 15° C or warmer isotherm of the world oceans (Matsumoto et al., 1984). In the western Pacific the widespread distribution of longline fishing demonstrates that skipjack have been captured as far north as 44° N off Japan and as far south as 37° S off Australia. The catches were highest from May to August off Japan, and from October to March off New Guinea (Forsbergh, 1980). Skipjack occur from October to April off northern New Zealand (Clement, 1976; 1978), while large concentrations have been seen as far south as southeastern Tasmania (Robins, 1952). In the eastern Pacific, skipjack have been fished along the west coast of the Americas from 34° N off southern California to 27° S off northern Chile. In the north, their range fluctuates with sea temperature from 25° N in February through April to 34° N in August through October (Williams, 1970). As the distribution of skipjack approaches the western Pacific it diverges from the equator to a greater extent than in the eastern Pacific. As mentioned earlier, this is due to the westward transport of warm surface waters and their poleward displacement along the western shores. The average distribution of skipjack in the EPO during 1979–1987 for all purse-seine sets is depicted in Figure 7a. The average distribution of skipjack catch by purse seiners and baitboats in 1988–90 in the WPO, as indicated by data submitted to the SPC¹ is shown in Figure 7b.

Skipjack physiology and morphology play a major role in the determination of suitable habitat, and by extension, their distribution. In common with most small scombrids, skipjack lack a swim bladder. While this permits rapid vertical movements within the near-surface habitat, it also increases the minimum swimming speed required to maintain hydrostatic equilibrium. Sharp (1978) calculated that a 50-cm skipjack must swim 60.5 km/d just for hydrodynamic stability and respiration. Skipjack have a large red muscle mass that functions aerobically to sustain this basal swimming requirement. In addition, white muscle tissue, which can function both aerobically and anaerobically, enables short bursts of very high swimming speeds for pursuit of prey and escape from predators. High oxygen demands due to high metabolic rates and the resultant production of large amounts of metabolic heat by the red muscle mass are of primary importance in determining skipjack distribution. A minimum dissolved oxygen level of about 2.45 ml/l is required by skipjack to maintain the basal swimming speed (Sharp, 1978), and a somewhat higher level would be required (about 3.0–3.5 ml/l) to sustain their survival over extended periods. Dissolved oxygen usually approaches saturation (4.5 ml/l) in surface waters, but is often less than the minimum requirement in waters in and below the thermocline (Toole et al. 1988). This generally restricts skipjack to the mixed layer above the thermocline. Despite some ability to thermoregulate, large skipjack encounter the problem of retention of excessive metabolic heat, and thus need to remain in the vicinity of cooler water. In the tropics, suitable temperatures for these large fish would be found only at the thermocline and below. However, as noted above, such water may be an unsuitable habitat because of low dissolved oxygen. Using physiological data from tank experiments, Barkley et al. (1978) defined the hypothetical limits of skipjack habitat. They suggested that only small (<4 kg) skipjack can inhabit most surface tropical waters and that the habitat of large (> 6.5 kg) skipjack in the tropics is in the vicinity of the thermocline, adjacent to cooler water. Where this is poorly oxygenated, large skipjack would be excluded. The gross features of skipjack distribution appear to broadly fit this hypothesis.

A long list of oceanographic and biological features is known to influence, either directly or indirectly, the distribution of skipjack within their overall limits. These include temperature, salinity, dissolved oxygen, thermocline structure, bottom topography, water transparency, current systems, water masses and biological productivity. Within the areas of greatest abundance, i.e. the tropics, temperature may play only a minor role in determining small-scale distribution, as thermal gradients in these areas are generally weak. In subtropical areas, skipjack fisheries exhibit seasonality which correlates well with surface temperature, e.g. in New Zealand and Australian waters. In the tropics, the distribution of suitable forage would seem to play a more important role in determining skipjack distribution. In fact, the patchy distribution of skipjack may be best correlated with the similar patchy distribution of zooplankton and micronecton (Lewis, 1981). Increased skipjack abundance around islands, seamounts, banks etc., may well result from increased food supply near these areas. Similarly, the frequent occurrence of skipjack in the vicinity of upwellings and convergences may result from the increased productivity associated with deep-origin, nutrient-rich water and the concentration of drifting or weakly-swimming biota, respectively.

Figure 7. a) Average annual catches of skipjack in the EPO during 1979–1987 for all purse-seine trips supported by logbook data, (after IATTC, 1989); and b) Average distribution of skipjack in the WPO purse-seine and baitboat fisheries, 1988–1990. The maximum circle size represents a catch of 500 mt or more. (From the SPC Regional Fisheries Database).

The small-scale distribution of juvenile and adult skipjack in the western Pacific is also influenced by their attraction to floating logs and other naturally-occurring flotsam. To an increasing extent, this is also true of artificial fish aggregating devices (FADs). Presently, FADs are moored in coastal waters of many Pacific Island countries to increase fishing success and minimize fuel costs for artisanal fishermen. In other locations, FADs play an important role in industrial baitboat and purse-seine fisheries, e.g. the Solomon Islands, Philippines, and Indonesia. Increasingly, large, industrial purse seiners are deploying moored or drifting FADs throughout the major fishing areas of the western tropical Pacific. Little is known regarding the dynamics of tuna attraction to and retention at FADs, both artificial and natural. As FAD deployment increases, changes in the small-scale distribution and vulnerability of skipjack and other small pelagics might be expected in some areas.

¹ Data provided to the SPC result mainly from bilateral access agreements between Pacific island countries and distant-water fishing nations, and therefore catches in international waters may be under-represented. The exception to this is that high-seas catches by USA purse seiners are provided as a condition of the Mulitlateral Treaty on fisheries.

6.3 Migration: Eastern Pacific

Based on a variety of data sources, including genetic studies and tag returns, Rothschild (1965) hypothesized that a large fraction of the pre-recruit (35–45 cm) skipjack that disperse from the central Pacific migrate towards the coast of the Americas. One portion, the northern group, enters the Baja California-Revillagigedo Islands fishery, while the southern group proceeds to the fishery off Central America to northern Chile. The fish arrive in the EPO when they are 1 to 1.5 years old, and return to the central Pacific after several months, or when they are 2 to 2.5 years old (55 – 65 cm) (IATTC, 1979). It is also of interest to note that none of the 35,534 skipjack tagged in French Polynesia by the SPC and IATTC have been recovered in the eastern Pacific (Gillett and Kearney, 1983); it was considered that the fish were generally too large when tagged for them to have had sufficient time to undertake the migration and still be of a size vulnerable to the EPO skipjack fishery.

From tagging experiments, Fink and Bayliff (1970) noted that the fish in the northern group migrate from the Revillagigedo Islands to Cabo San Lucas, Baja California, during May and June, and then up the west coast of the peninsula from May to September (Figure 8a). The fish may penetrate as far north as 30°N depending, amongst other factors, on the temperature. During September to November or December they then migrate southward along the Baja California peninsula, and although most of the fish turn westward towards the central Pacific, the remainder continue towards the Revillagigedo Islands and arrive at year's end. Some fish that overwinter in the area are thought to either repeat the northern migration in the following year, return to the central Pacific or travel to waters near the southern tip of Mexico.

“Recruits to the southern region appear mostly in or near the Panama Bight (Figure 8b). Those which enter the fishery in the northern Panama Bight migrate rapidly northwest to Central America and south to Ecuador. The proportions which migrate in these two directions vary considerably from year to year, and may depend on differences in sea-surface temperatures. The fish appear to migrate from the Gulf of Guayaquil to the Ecuador coast during the first half of the year, and return during the second half. Some fish migrate further south to central Peru and to southern Peru and northern Chile” (Fink and Bayliff, 1970). There is evidence from more recent tagging experiments that skipjack from the southern group also return to the central Pacific (Bayliff, 1984). As mentioned above (Section 6.1), the area between the northern and southern groups is usually devoid of skipjack (Fink and Bayliff, 1970).

Williams (1972) supported Rothschild's (1965) hypothesis by proposing three migration models for skipjack from the central to the eastern Pacific. In the first one the fish actively swim eastward against the westward-flowing North and South Equatorial Currents (NEC and SEC) (Figure 9). The second model involves passive migration in which the fish are carried eastward by the North and South Equatorial Counter Currents (NECC and SECC). The third includes a pattern in which fish in the northern group are carried in the counterclockwise gyre of the northern equatorial water mass in the EPO, and most of the fish in the southern group are transported by a clockwise gyre formed by the NECC and SEC.

To evaluate these possibilities, Forsbergh (1988) reviewed data on geostrophic flow, current velocities, larval distribution, oxygen requirements, and swimming speeds in relation to fish length, as well as temperature profiles of the Cromwell, or Equatorial Undercurrent (EUC). From this information he drew several conclusions regarding the relationship between current patterns and their role in migration. For example, drogues placed in the NECC moved eastward for 4,500 km (2,428 mi) in 4 months (Wyrtki et al., 1981), indicating that the NECC could be a major, passive migration route. This current also extends eastward to about 85°W from May to December, but reaches to only about 120°W in February and March (Wyrtki, 1967). However, the average monthly width of the NECC at 130°W is only 3° –6° of latitude (Wyrtki, 1965), whereas the latitudinal distribution of skipjack larvae in the central Pacific is about 40° (Nishikawa et al., 1985). Consequently, only about 7.5–15 percent of the larvae could be carried eastward by the NECC, and the remainder, or major portion, could be transported westward by the NEC and SEC. More recently, Forsbergh (1989) found that estimates of relative apparent abundance of skipjack in the EPO were not correlated with the velocity of the NECC at earlier times. Although the latitudinal concentration of juveniles east of 180° appears to be similar in the NECC, NEC and SEC, there are insufficient data to determine the concentration of skipjack in relation to zonal currents (Nishikawa et al., 1985; Mori, 1972). In contrast to the NECC, the eastward transport of juveniles by the SECC is negligible because the current does not always exist and it is weaker than the NECC. As another means of active transport, the EUC also flows eastward beneath the surface at the equator at depths between 25–300 m, but it probably contains few juveniles because the temperature is too low and the oxygen concentrations appear to be too low for skipjack > 35 cm (Forsbergh, 1988).

With regard to the active migration model, Forsbergh (1988) estimated that juveniles in the length ranges of 13–17, 26–34, and 38–51 cm could swim at 1, 2, and 3 kt, respectively, if they could maintain cruising speeds of 3–4 body lengths/sec. These velocities would be adequate to overcome the 0.3 kt average speed of the NEC, but probably not that of the SEC (0.6—1.8 kt) on a consistent basis. From these and the previous observations, Forsbergh (1988) concluded that, “…it is unclear whether either of William's (1972) active or passive migration models is more realistic than the other.” From these arguments it is possible to conclude that passive migration may take place in the NECC and that active migration may take place in the NEC; both migratory modes could result in the transport of juvenile skipjack towards the eastern Pacific.

Figure 8. Migratory paths of skipjack entering (a) to the south and (b) to the north in the eastern Pacific. (After Fink and Bayliff, 1970).

Figure 9. The major surface currents of the Pacific Ocean between 50°N and 50°S. Stippled areas indicate subsurface currents that are also found at the surface in some months. (After Suzuki et al., 1978).

Seckel (1972) developed a quantitative model to indicate how skipjack in the EPO could accomplish the return migration to Hawaiian waters or beyond. The components of the model included the geostrophic flow of the NEC and the northward component of surface, wind-driven current resulting from the trades. According to this drift model, fish schools distributed between 10° –20° N and 128°W would be concentrated by the longitudinal component of the current flow as they drifted westward. The minimum time to reach the Hawaiian Islands under these conditions was 21–23 mo. Of 23 skipjack tagged in the EPO and recovered off Hawaii, 17, or 78 percent, completed the journey in less than 20 mo (range: 8–15 mo; Bayliff, 1988, Appendix). The differences in timing may be explained by net swimming with the flow of the NEC (Forsbergh, 1988).

6.4 Migration: Western Pacific

A substantial body of information on skipjack migration, and movement in general, in the WPO was collected by the SPC during its Skipjack Survey and Assessment Programme (SSAP), carried out during 1977–80. Most of this information was provided by the return of 7,000 skipjack tags from approximately 140,000 releases. Analysis of these data has been hampered by the incompleteness of catch and effort data from the fisheries that recovered the tags; therefore the conclusions drawn to date with respect to movements have been largely qualitative in nature. One prominent feature of the SSAP data was an apparent eastward trend in the movement of tagged skipjack in Micronesian waters between 5°N and 10°N (SPC, 1981b), although the authors cautioned that this trend may have been an artifact of the largely unknown distribution of the Japanese baitboat fishery at the time. Another feature that emerged was the lack of apparent movement out of certain areas, notably Papua New Guinea, the Solomon Islands, Fiji, Western Samoa, the Society Islands and the Tuamotu Archipelago (SPC, 1981b). For the latter four areas, this may have been an artifact due to the lack of active fisheries in the neighbourhoods.

While numerous long-distance movements of individual tagged skipjack were observed, the overall percentage of recoveries having displacements of greater than 200 nmi was only 17 percent (SPC, 1981b). This result led some authors (Hilborn and Sibert, 1987) to suggest that long-distance movement may in fact be uncommon. The alternative interpretation is that the observed spacio-temporal pattern of tag recoveries is a result of most recoveries having been made from releases in the vicinity of locally-intense fishing effort (Mullen, 1989; Polacheck, 1990). The SPC's current Regional Tuna Tagging Project (RTTP) is providing new information on skipjack movement by tagging throughout the area 10°N–10°S, 120°E–180°. In these experiments, there is currently a higher percentage (27 percent at 1 June, 1991) of returns that have displacements of > 200 nm, and this percentage is likely to increase as longer times at liberty are sampled (SPC, unpubl. data). It is likely that this higher proportion of long-distance returns has resulted because of a more homogeneous distribution of fishing effort (mostly purse seine) than existed in 1977–80 (mostly baitboat), which would tend to support the views of Mullen (1989) and Polacheck (1990). This issue, which has important international management implications, will hopefully be resolved by the application of new, spatially-disaggregated movement models to the SSAP and RTTP tag data, along with appropriate data on the distribution of fishing effort and catch.

7. NATURAL MORTALITY

It is a difficult problem to estimate the instantaneous mortality rate (M) of most commercial species of fish, but it is particularly troublesome in the case of skipjack tuna. Part of the difficulty stems from the fact that, regardless of the method of evaluation or the fishery, an emigration term (E) must be factored into the overall rate of attrition. Particularly if tagging experiments are used to estimate mortality rates, several additional mortality components must be recognized besides the instantaneous fishing mortality rate, F. Bayliff and Mobrand (1972) noted that in a tagging experiment mortality can occur in two modes, actual and effective, as well as in two different time periods. Fish can either immediately die or shed their tags as a result of the tagging operation (Type 1 mortality), and the proportion surviving in each case is π and ρ, respectively. The proportion of fish recaptured with useful information (β) is frequently less than anticipated because recapture dates may be unknown or known imprecisely. Over the long term, fish may also die at continuous, instantaneous rates (Type 2 mortality) from carrying the tags (G), and through tag shedding (L). While the total mortality rate (Z) of untagged fish in a closed fishery is Z = F + M, in a fishery open to emigration the total attrition is expressed as Z ′ = Z + E. The total attrition rate for tagged skipjack (Z*) in an open fishery is therefore Z* = F + X, where X = M + E + G + L (Joseph and Calkins, 1969; Bayliff, 1977). Estimates of F can usually be obtained from information on catch and effort and the recovery rate of tagged fish, but the best that can be achieved usually with the mortality components of M, E, G, and L is a partial separation.

From tagging experiments initiated off Baja California, Fink (1965) estimated that the proportion of yellowfin (y) surviving immediately after tagging was (πρ)_y = 0.85, and Type-2 mortality² was (G + L)_y = 1.06. In the absence of any other information for skipjack (s) at the time, subtracting this last sum and F = 2.10 from Z* = 6.98 yielded an estimate of (M + E)_s = 3.82. For the skipjack fishery off northern Peru, a similar estimate of (M + E)_s = 3.65 was obtained from values of Z* = 5.02 and F = 0.31. Subsequently, Joseph and Calkins (1969) estimated from a skipjack tagging experiment that (πρ)_s = 0.68. Bayliff and Mobrand (1972) utilized single and double-tagged experiments on yellowfin in the EPO to obtain estimates of ρ_y = 0.9 and L_y = 0.28. Based on Fink's (1965) result for (G + L)_y above, one estimate of the long-term mortality rate from carrying tags is therefore G_y = 1.06–0.28 = 0.78. Since the last-mentioned value of ρ is greater than the previous product of πρ for skipjack, there is some indication that the immediate loss of tags and mortality of skipjack is greater than that of yellowfin. Joseph and Calkins (1969) were also able to estimate the combined rate of mortality, X_s = 2.76, from a single tagging experiment. Bayliff (1977) reassessed the data of Schaefer et al. (1961), Fink (1965) and Joseph and Calkins (1969), as well as more recent information, and concluded that the value of X_s was probably less than 3.0, thereby supporting the earlier result. However, Bayliff (1977) was unable to establish a reliable difference between Z*, which includes E, and attrition rates that presumably excluded E because the data base was limited (truncated) to tag returns of ≤ 6 mo. While it is reasonable to expect that Z* would exceed the attrition rate from the truncated data, this result was achieved in only half of 26 experiments, and the differences between the two rates were often negligible. The test was not considered valid because of too few returns of fish at liberty in excess of 6 mo. It was also expected that the attrition rate of single-tagged skipjack would exceed that of double-tagged fish, but this result occurred in only one out of two such experiments. Bayliff (1977) concluded that irregularities in the semilogarithmic plots of tag returns per unit effort against time precluded the reliable estimation of attrition rates that ranged from 3.48 to 14.93 with an average value of Z*_s = 6.86. These irregularities, in turn, were apparently caused by the failure of tagged and untagged fish to mix completely, and the uneven distribution of effort with respect to the fish.

² All instantaneous rates are expressed in annual rates.

Forsbergh (1987) drew attention to the small numbers of returns in most of the tagging experiments examined by Bayliff (1977) and the large variation in Z_s* values. He therefore tried to estimate Z ′_s directly from the catch rates of untagged fish of different age groups, thereby avoiding the additional mortalities attributed to tagging. By applying Robson and Chapman's (1961) method of catch-curve analysis to eight, consecutive quarterly intervals of catches from 1961–83, an average value was obtained of Z ′_s = 3.35 (range: 1.56–5.56). This result was based on an assumed linear growth rate of 24 cm/yr (Forsbergh, 1989), which in turn was used to estimate the ages of fish in the length-frequency distribution. If growth was assumed to follow the von Bertalanffy (1938) model with L∞ = 86.0 cm and K = 0.79 (IATTC, 1984) then Z ′_s = 3.25. These rates for Z ′_s are considerably less than Bayliff's (1977) value of 6.86, but Forsbergh (1987) attributed greater accuracy to the latter result because it was based on tagging experiments. Currently, the IATTC assumes that 75–90 percent of the fishable stock is lost through attrition each year (IATTC, 1989), and this is equivalent to Z ′_s = 1.39–2.30. At this point, by assuming that either G_y = G_s = 0.1 (Bayliff, 1971), L_y = L_s = 0.28, or that (G + L)_y = (G + L)_s = 1.06, the following estimates of (M + E)_s can be obtained by subtraction from the preceding information:

Since so little is certain about skipjack in the EPO, Forsbergh (1987) also attempted to apply cohort analysis to quarterly catch data even though the assumption of a closed system is thought to be violated. Estimates of the recruits entering the fishery frequently failed to converge and the results were considered unreliable. The lack of convergence may have been due to variable immigration and recruitment during the first three quarters of the year and variable emigration during the last quarter.

As a result of the SSAP tagging carried out during 1977–1980 in the southwestern Pacific, Kleiber et al. (1983, 1987) estimated an aggregate value of L_s = 0.09 for the individual countries and territories participating in the South Pacific Commission's experiments. The results of double tagging also indicated that Type-2 tagging-mortality loss (G_s) was negligible, and therefore assumed to be zero. The product of Type-1 mortality and the midpoint between the worst and best estimates of β was (πρβ)_s = 0.60 in the aggregate. By subtracting F_s = 0.08 (± 0.014) and L_s from Z_s* = 2.04 (± 0.24), the approximate, aggregate value of (M + E)_s is 1.87. This value is not substantially different than those shown above for the EPO. Kleiber et al. (1987) cautioned that, with the exception of Z_s* the confidence limits associated with such estimates would be larger if the uncertainties in πρ and β were included. To augment these comparisons, it should be noted that for the Gulf of Guinea in the eastern Atlantic, Bard (1986) obtained estimates of (πρ)_s = 0.60 and β_s = 0.80 from fish tagged in 1980–1981. By further assuming that G_s = 0.10–0.20, and L_s = 0.10, a value similar to that obtained in the South Pacific experiments by Kleiber et al. (1983), the sum of (M + E + G)_s fell in the range from 1.78–1.88. Fonteneau (1986) estimated that M_s = 0.60–1.00 for eastern Atlantic skipjack, and subtracting these values from the previous sum yields (E + G)_s = 0.78–1.28.