Ecological Archives E086-043-A1

Peter Ward and Ransom A. Myers. 2005. Shifts in open-ocean fish communities coinciding with the commencement of commercial fishing. Ecology 86:835–847.

Appendix A. Estimation of relative abundance and biomass using longline data from surveys and observers.

INTRODUCTION

This appendix provides details of the data that we used in models to develop indices of abundance and biomass, which are presented in our Ecology article. We also examine the robustness of abundance indices and consider whether indices from different periods are indeed comparable. Appendix B presents parameter estimates and diagnostic statistics for each model.

DATA SOURCES

1950s survey

In 1950, the US embarked on an ambitious program of fishery monitoring and scientific surveys of Pacific tuna resources. The Pacific Oceanic Fisheries Investigations (POFI) was a response to interest in harvesting the tuna resources of newly acquired US territories and possessions in the region. The U.S. fishing industry sought large quantities of yellowfin tuna (Thunnus albacares) for canning. Using longline fishing gear and techniques adopted from Japan, POFI conducted 24 longline fishing trips each of about two months duration in the tropical Pacific Ocean during 1951–1958 (Murphy and Shomura 1972). Most of the activity was during 1951–1954. We refer to the POFI program as the “1950s survey”.

The 1950s survey was conducted as a controlled experiment where fishing gear and techniques were held constant throughout the study. Murphy and Shomura (1972) and reports that they cite provide details of the fishing gear and techniques. Longlines were deployed in a grid at pre-determined stations, e.g., at each one-degree of latitude. They were deployed at dawn each day and retrieved in the afternoon. Usually, six hooks were attached to the mainline between each buoy, amounting to about 342 hooks in each daily fishing operation (Table A1). Within the same operation, the performance of the standard gear was sometimes compared with that of gear with different types of bait, methods of bait attachment, or longline depth (Murphy and Shomura 1972). After the mid 1950s, POFI chartered commercial longliners that were encouraged to search for tunas for US canning markets.

1990s observer data

Our Ecology article examines changes in the pelagic fish community by comparing estimates from the 1950s with those from data collected by observers on Hawaii-based longliners in the tropical Pacific Ocean during 1994–2002. In the study region, the 1990s longliners targeted bigeye tuna (Thunnus obesus) and yellowfin tuna for commercial sale. Their catch of other species, such as broadbill swordfish (Xiphias gladius), striped marlin (Tetrapturus audax), mahi mahi (Coryphaena hippurus), and wahoo (Acanthocybium solandri) was also valuable. The markets pay a premium for large individuals of most species (T. Swenarton, personal communication).

For each period, we used data that overlapped in terms of deployment time (02:00–08:00 local time) and month (January–November), within a broad region of the Pacific Ocean (10°S–11°N, 175°E–115°W; Fig. A1). Table A1 highlights several similarities between the fishing gear and techniques used in the 1950s and 1990s. However, there are also differences such as the depth range, soak time, number of hooks per operation and areas fished.

Species identification

Our analyses revealed several species that were caught in the 1990s, but were not reported in the 1950s, e.g., pelagic stingray (Dasyatis violacea). Differences in sample size, catchability, and the level of species identification prevented us determining the exact number of extra species that appeared in the 1990s. We were confident, however, that the extra species would have been reported if they were caught in the 1950s. Scientists involved in the 1950s survey had access to taxonomic keys that covered the extra species. Furthermore, survey activities outside the study region reported several of the extra species, e.g., pelagic stingray. We contacted a scientist involved in the survey, R. Shomura (personal communication), who confirmed that they would have identified and reported the extra species if they had been caught in the study region.

Data for both periods included fish that were not identified because they were released or broke free before being identified or, more rarely, fish that were damaged beyond recognition. In the 1950s, 1883 (9%) of the 17 439 fish caught were not identified, compared to 482 (2%) of the 24 208 caught in the 1990s. A check of the original data sheets revealed that 99% of the unidentified fish in the 1950s were unidentified sharks. Few, if any, unidentified fish could have been the extra species.

We used generalized linear models to derive indices of abundance and biomass for each species or species group in the study region. For brevity, we use the term “species” in this appendix to refer to species groups as well as individual species. The models adjusted estimates for operational differences and spatial and temporal variations among the longline operations and periods. Following is a description of the procedures used to estimate hook depth, soak time, and fishing-effort for the models.

DATA PROCESSING

Depth and soak time estimation

We used the formula for a catenary curve to estimate the maximum depth reached by each longline hook (Suzuki et al. 1977). The formula uses , the angle made between a horizontal line drawn between the tangential line of the mainline and the connecting points of the buoy line and mainline. Suzuki et al. and many other researchers selected a value of 72° for because they did not have data on the sagging rate k. The sagging rate is the ratio of the distance between buoys and the length of mainline between buoys, measurements that were available for the 1950s survey and 1990s longlining activities. We used a derivation of the formula presented by Yoshihara (1954) to calculate from k for each longline operation:

(A.1)

We assumed that the shape of the curve formed by the longline (and therefore the corresponding depth of hooks) does not systematically vary over the entire operation. Local conditions (e.g., the direction and velocity of wind and current), the buoyancy of the gear, the pull of hooked fish, and shortening of the distance between buoys over time are factors known to affect the actual depth range of longline hooks. Bigelow et al. (2002), for example, estimated that hook numbers 3 and 10 of longline gear with 13 hooks per buoy, shoaled by about 20% when subjected to a current velocity of 0.4 m/s. Strong equatorial currents characterize our study region. To represent the likely shoaling of the longline, we reduced all depths predicted by the catenary curve by 25%. The application of a 25% constant would not affect comparisons between periods because hook depth was used as a relative index in our analyses.

The data for both periods included a unique identifier—the hook number—for each longline hook deployed. The 1950s survey reported the hook number for each fish caught, allowing the estimation of its likely depth of capture. The 1990s observers reported the hook number for most catches of tunas, billfishes, and sharks, but not for other species. For the other species we used information on depth from a wider study (see “Depth”) or the number of hooks per buoy as a measure of the depth range of the entire longline operation.

We estimated the soak time of each hook from records of the time when each hook was retrieved, combined with the start and finish times of longline deployment and retrieval. The speed of longline deployment and retrieval was assumed constant throughout each operation (the analyses were limited to operations that had no evidence of stoppages due to line breaks or mechanical failure). The data were binned into 40-m depth – 1 h soak time strata. However, we used the median soak time for species where the hook number was not reported. The analyses included soak time – depth strata where hooks were deployed but a zero catch was reported.

Estimation of fishing-effort

The total number of hooks deployed is often used as a measure of fishing-effort in longline fisheries. However, the measure of fishing-effort should be adjusted for the number of hooks occupied by fish during the operation because occupied hooks are no longer available for fish that might subsequently encounter the longline (Rothschild 1967). Gear saturation would have affected our indices of abundance and biomass, particularly for the 1950s when total catch rates often exceeded 100 fish per 1000 hooks (Fig. A2).

To adjust fishing-effort for gear saturation, we subtracted half the number of fish caught from the number of hooks deployed in each soak time – depth stratum of each operation. Information was not available on the depth and soak time for all species. Therefore we also subtracted half the mean number of those species per longline stratum from the number of hooks deployed. This approach is likely to overestimate the number of vacant hooks because of local saturation and bait loss (Rothschild 1967).

Body mass and biomass estimation

We estimated the biomass index of each species by multiplying its abundance index by its mean mass in each period. The 1950s survey weighed 25% of all fish caught in the study region. The observers did not routinely weigh fish in the 1990s. However, they measured 57% of all fish, so we estimated the mass of each fish by applying length–mass relationships to length measurements (Strasburg 1957, Uchiyama and Kazama 1999, Froese and Pauly 2003). For ten species that were not measured in the 1990s, we used length data collected by Secretariat of the Pacific Community (SPC) observers on commercial longliners fishing in adjacent waters of the western Pacific during 1990–2001 (P. Williams, personal communication). For those species we used 1950s estimates from a wider region (10°S–15°N, 175°E–115°W).

The 1950s survey attempted to land and weigh all fish, regardless of species or body mass (R. Shomura, personal communication). By contrast, the 1990s longliners sometimes released fish before observers could measure them. Fortunately, observers visually estimated the length before the fish was released or broke free from the longline. The visual estimates were to the nearest 12 inches (25.4 cm). We included both retained and released fish in our indices of abundance and biomass. For most species, retained individuals tended to be larger than those released (Fig. A3). Our interpretation is that crewmembers released valuable species that were below a marketable size. There might be an element of bias in the visual estimates of length, although a study by Gillanders et al. (2001) concluded that recreational fishers were just as likely to underestimate the lengths of fish alongside their boat, as they were to overestimate them.

The inclusion of visual estimates made little difference to our indices of biomass and body mass for most species because few were released (Table A2). However, more than half of the length data on sharks came from visual estimates. The declines in body mass and biomass indices presented in the Ecology article would be much larger if the mass of released sharks had not been included.

Our index of community biomass excluded species that were caught in the 1990s as a result of longlining near land masses (Table A3). Those species would have accounted for less than 1% of the 1990s biomass. The index of community biomass was based on estimates of 21 species with well-determined indices of biomass. For each period, we raised the biomass index by the proportion of species with unreliable indices of abundance:

(A.2)

where Cw is the catch (kg) and Bw is the biomass index (kg) of well-determined species, CT is the catch of all species, and BT is the index of community biomass. For the 1950s, the well-determined species accounted for 90% of the total catch, compared to 89% for the 1990s.

ROBUSTNESS OF ABUNDANCE INDICES

Model selection

We derived abundance indices from generalized linear models that accounted for bias caused by the effects of hook depth, soak time, location, season, and fishing-effort. The models were implemented in SAS (version 8.0) using the GENMOD function and S-Plus (version 6.1) using glm and glm.nb from library MASS (Venables and Ripley 1999). Model building examined the type of error distribution (e.g., Poisson or negative binomial model), the inclusion of variables (e.g., latitude and longitude), and functional relationships for each variable (e.g., linear or quadratic). We used Akaike's information criterion (AIC), deviance residuals, and the negative binomial model parameter  to check the performance of the various model formulations. We also used stepwise model selection using the AIC criterion holding the negative binomial parameter  f ixed, but they did not produce estimates of change in abundance that were significantly different to the estimates from the model with all variables included (Table A4). We presented indices for the models with the lowest standard error, excluding indices for species where the standard error of the estimate was greater than one or where the change in biomass was not significantly different from one (Table 5). For 19 species, the performance of models with negative binomial error was superior to the same models with a Poisson error distribution. The Poisson model was recovered for several rare species, e.g., striped marlin. The inclusion of quadratic terms for depth, soak time, latitude, and longitude provided the best fit for most species (Appendix B).

Area

Commercial longliners often concentrate over deep-sea canyons and seamounts where target species are abundant (Bigelow et al. 1999). The 1990s longline activity was in the vicinity of seamounts and islands of the Christmas Ridge in northwestern waters of the study region. By contrast, the 1950s activity had a much wider distribution, including relatively featureless parts of the eastern study region (Fig. A1). We investigated the sensitivity of abundance indices to variations in the boundaries of the study region. Variations in the longitudinal boundary made little difference to abundance indices, whereas latitude had a greater effect (Table A6). Estimates of model coefficients (s) confirmed that latitude had a stronger effect on abundance indices for most species than did longitude. We drew on the power of a larger data set by using a study region that had a wide longitudinal range and a narrower latitudinal range.

Figure 1 of the Ecology article shows considerable activity in northern waters outside the study region during the 1990s. The exclusion of those activities from analyses did not affect our indices because the mean catch rate for 1990s activity in the study region (24.5 animals per 1000 hooks) was similar to that in northern waters (26.7 per 1000 hooks). 

We also explored the joint effects of latitude and longitude on abundance indices by including a latitude–longitude interaction term in the hook model.  The interaction term was significant at the 95% level for six of the nine species investigated (Table A7). Regardless, inclusion of the interaction term did not make a significant difference to estimates of the change in abundance between periods. For example, the model that included the interaction term estimated that the abundance of blue shark in the 1990s was 0.14 times that in the 1990s. By contrast, it was 0.13 times that for the model without the interaction term. We concluded that adding a joint measure of latitude and longitude does not significantly alter estimates of change in abundance.

Soak time and timing

Soak times ranged up to 8 hours in the 1950s compared to 12 hours or longer in the 1990s. Parameter estimates indicated that catch rates increased with soak time for many species, particularly sharks and billfishes. The catch rates of several species, such as skipjack tuna (Katsuwonus pelamis), decreased with soak time.

We adjusted estimates for the effects of soak time, but did not correct them for variations in exposure to dawn or dusk. Ward et al. (2004) found that exposure of longlines to dawn and dusk produced elevated catch rates of many pelagic species. In our study, the long duration of 1990s operations resulted in about 80% of longline hooks being available at dusk as well as at dawn. By comparison, 1950s hooks were only available at dawn. Therefore the abundance and biomass of several species may have been underestimated in the 1950s.

Depth

By covering a wider depth range than shallow longlines (<200 m), deep longlines (<600 m), produce elevated catch rates of several species, e.g., bigeye tuna (Bigelow et al. 2002). For some epipelagic species, such as blue marlin (Makaira nigricans), they produce lower catch rates than shallow longlines (Hinton 1996). We explored the limited overlap in depth range between periods by adding an interaction term for fishing period – hook depth and adding a dummy variable to the hook model for 12 species (Table A8). The dummy variable indicated whether hook depth was less than 200 m. The interaction term was significant at the 95% level for 5 of the 12 species, indicating that hook depth modifies the estimate of change in abundance (i.e., fishing period) for those five species. However, restricting input data to strata where depth was less than 200 m made only a small difference to estimates of the change in abundance between periods (Table A9). For example, the model fitted to data with hook depths less than 200 m estimated that the abundance of blue shark in the 1990s was 0.15 times that in the 1950s. By contrast, the change in abundance was 0.13 times for the model that included all depths.

We applied the “offset model” to ten species where an estimate of the depth of capture was not available (Table A10). It used information on depth distributions from the wider study by Myers and Ward (public communication, <http://as01.ucis.dal.ca/ramweb>). Their analyses included the 1950s and 1990s data, and data from 1990s longline operations that targeted broadbill swordfish in the North Pacific, and those that targeted bigeye tuna and yellowfin tuna in the western Pacific. The offsets we used were those for longlines deployed in the morning.

Verification of catch rates

We compared 1950s survey catch rates with those from commercial longline operations during 1952–1954. The commercial operations involved several large processor vessels or “motherships”, each accompanied by 12–25 smaller catcher vessels. Japanese companies were permitted to undertake nine mothership expeditions in a restricted area of the tropical Pacific (Shapiro 1950). The mothership expeditions targeted tunas for canning markets (Van Campen 1952). The fishing gear and techniques used by catcher vessels were almost identical to those of the 1950s survey (Niska 1953). The 1950s mothership data were available for nine commercially valuable species of tunas and billfishes, aggregated by month and five-degree square area.

We compared mothership catch rates in the area bounded by 20°S–10°N and 150°E–130°W with 1950s survey catch rates in the study region (Table A11). Overall, the 1950s survey reported catch rates of 48 tunas and billfishes per 1000 hooks compared to 47 per 1000 hooks for the motherships. The catch rates of two species (skipjack tuna and yellowfin tuna) were higher in the 1950s survey, whereas the catch rates of bigeye tuna and blue marlin were lower. We might expect the mothership catch rates to have been significantly higher than survey catch rates because the survey was restricted to often predetermined stations along a grid. However, observers on the motherships reported that external factors kept catch rates below levels that true commercial longline operations could have accomplished. Those factors included restrictions on areas of operation, catcher vessels having to remain in close proximity to their mothership, and poor bait quality (Van Campen 1952).

POSSIBLE CAUSES OF VARIATIONS AMONG INDICES

Competition among and within operations

Our comparison of abundance indices that were derived from catch rates indicated marked differences between the pelagic fish community of the 1950s and that of the 1990s (Fig. A2). Other than a massive decline in abundance, we can imagine few plausible explanations for 1990s catch rates being so much lower than 1950s catch rates. One alternative explanation is competition within longline operations. The 1950s survey deployed fewer longline hooks in each operation than were deployed in the 1990s. Individual hooks might not fish independently within a longline operation (Rothschild 1967). However, competition might not be a concern because commercial longliners would not increase the number of hooks deployed, or shorten the distance between hooks, if it reduced the day’s catch of target species (Hirayama 1969,  Hamley and Skud 1978).

A second possible explanation of the decline in catch rates is competition among fishing vessels (Hilborn and Walters 1987). The increased number of longliners and other vessels fishing for pelagic species in the study region after the 1950s might have increased competition for the most productive areas, resulting in the displacement of some longliners to less productive waters. However, it is noteworthy that longliners rarely sight each other, let alone physically interact in the open ocean. Longline operations are measured on scales of hours and tens of kilometers whereas pelagic fish distributions are measured on much broader scales. Angel (1993) suggests that 1000 km is an appropriate horizontal scale for the distribution of pelagic communities of the open ocean. In the next section, we discuss the ability of 1990s longliners to find concentrations of target species. This mobility would have enabled them to maintain catch rates by moving to alternative areas if competition was a problem. We did not investigate competition between fishing vessels, which would require accurate estimates of local fish abundance, mixing rates, and fine-scale data on catch and fishing-effort of all vessels in the region.

Searching, fishing gear, and techniques

There is strong evidence that searching and technological improvements resulted in the underestimation of 1950s abundance relative to 1990s abundance. The 1950s survey adopted longline gear and techniques that had been used in Japan and Formosa (now Taiwan) during the 1930s (Niska 1953). They were not modified or refined during the survey, except for tightly controlled experiments with bait type, attachment methods, and longline depth (Murphy and Shomura 1972).

In contrast to the 1950s survey, the longline gear and techniques used in the 1990s were the product of 40 years of practical experience and innovation. Refinements ranged from the adoption of new technology (e.g., color sounders, Doppler current meters, and satellite imagery) to more subtle changes, such as the materials used for branchlines and mainlines. The 1990s longliners also had the ability to modify their fishing techniques to suit conditions while on the fishing grounds, e.g., longline depth, time of deployment, and time of retrieval (T. Swenarton, personal communication).

Quantity, species composition, body mass, and quality largely determine the value of the catch landed by commercial longliners. The incomes of masters and crew were based on a proportion of the value of the catch. They would use every available piece of information and equipment to maximize the value of the catch (T. Swenarton, personal communication).

Perhaps the most important difference between the data sets is that the 1950s survey often deployed longlines at predetermined stations along a survey grid, whereas the 1990s longliners actively searched for concentrations of target species. The 1990s longliners remained in areas of high catch rates or followed the fish as the concentrations moved. Searching also involved the use of past experience to select fishing areas and communication with other longliners to locate concentrations of target species (T. Swenarton, personal communication). In estimating abundance, we were unable to quantify the effects of searching or improvements in fishing gear and techniques. One option would be to repeat the 1950s survey using exactly the same survey grid, fishing gear, and techniques.

Removals by sharks

There is evidence of a long-term decline in the proportion of fish that is damaged by sharks while the fish are hooked on the longline. The 1950s data indicate that sharks damaged 20–30% of tunas caught by survey longliners in the study region. By comparison, 1990s observers reported damage rates of about 4% (which is consistent with the decline in shark abundance shown by our indices ). If shark-damage rates reflect the rate at which hooked fish are removed from longlines by scavengers, then loss rates might have been higher in the 1950s, further adding to the underestimation of early abundance.

Influence of oceanographic conditions

Various authors have highlighted the effects of broad-scale oceanographic events on ocean productivity (e.g., Mantua and Hare 2002, Chavez et al. 2003) and the distribution of pelagic fish species (e.g., Polovina 1996, Bigelow et al. 1999, Rodriguez-Sanchez et al. 2002, Ravier  and Fromentin 2004). Strong equatorial currents and undercurrents that flow eastwards or westwards characterize our study region. They produce areas of divergent upwelling where relatively cold, nutrient-rich water is drawn to the sea surface. West of about 180°E lies the “warm pool” of less productive waters. The location, strength and direction of currents, and associated areas of upwelling vary with El Niño – Southern Oscillation events (Lehodey 2001).

We checked oceanographic conditions that are known to strongly influence the productivity of pelagic fish or their availability to longline gear (Table A12). During the 1950s study period, the mean monthly Pacific Decadal Oscillation (PDO) for November–March was –0.98 (±0.89 SD). With a mean PDO of –0.28 (± 1.07 SD), conditions in the 1990s were not markedly different to those in the 1950s. By comparison, the PDO ranged between –3.60 and 2.65 during November–March in the twentieth century (mean 0.00 ± 1.04 SD).

The study periods spanned months of high and low values of the Southern Oscillation Index corresponding to a mixture of La Niña and El Niño conditions. The 1990s featured stronger El Niño conditions than were experienced in the 1950s. The study periods spanned six years or more, which may smooth out the short-term effects of oceanographic events on fish availability and abundance. Furthermore, unusually high recruitment in the early years would need to be synchronized across various species with quite different life histories. Exceptionally high abundance of large predators in 1953, for instance, would require strong cohorts of tunas from the early 1950s combined with strong cohorts of sharks from the mid 1940s.

On a finer scale, water temperature and oxygen concentrations are known to influence the vertical distribution of many pelagic fish species (Bigelow et al. 2002). We were unable to obtain data on oxygen concentrations in the study region during the 1950s. The mean depth of the thermocline, as indicated by the 20°C isotherm, was 135 m (± 28 SD) in the 1950s compared to 131 m (± 22 SD) in the 1990s. In conclusion, there was no obvious difference between periods in the oceanographic conditions that we examined.

Temporal and spatial variation

Our indices of relative abundance and biomass represent a snapshot of the pelagic fish community in space and time; they might not represent trends over longer time periods or more broadly in areas outside the study region. The fish community is unlikely to have been static in the study region before fishing commenced. Variations in abundance caused by fishing may be magnified or moderated by broad-scale oceanographic events that affect the species’ distribution, recruitment, and vulnerability to the gear (Steele 1998, Jennings and Kaiser 1998). The periods that we analyzed might have coincided with years of particularly high (or low) abundance in the study region. The study region is at the margins of the distribution of several species, e.g., albacore tuna (Thunnus alalunga), striped marlin and blue shark (Prionace glauca). Estimates of those species might be confounded by geographical contractions or expansions of the populations that are driven by oceanographic conditions as well as population abundance (MacCall 1990). Nevertheless, the study area is a significant portion of the core habitat of most of the other pelagic fish species (Collette and Nauen 1983, Nakamura 1985, Last and Stevens 1994). The uncertainties outlined here might contribute to underestimation or overestimation of community changes.

Exploitation

The 1950s pelagic fish community was markedly different to that of the 1990s in terms of indices of abundance and biomass, and size composition. We believe that fishing is the most likely cause of the community changes. Industrial-level fishing began in the open ocean with open-boat whaling in the seventeenth century (Whitehead 2002). A second wave of expansion began in the latter half of the twentieth century when Japanese fishers combined pelagic longline gear with new technology for preserving fish. Low-temperature refrigeration systems enabled longliners to remain at sea for several months and to harvest highly mobile tunas and billfishes thousands of miles from ports (Sakagawa et al. 1987).

The study region is a relatively large section of the Pacific Ocean that is central to the distribution of many of the species that we analyzed (Collette and Nauen 1983, Nakamura 1985, Last and Stevens 1994). Therefore, fluctuations in the mean mass of those species should not be due to size-related expansions and contractions in range linked to variations in ocean productivity and environmental conditions. The mean mass of only one large predator, albacore tuna, increased between periods. This pattern might reflect the fact that the study region is at the margins of this species’ distribution. An alternative explanation is that heavy exploitation of albacore tuna by a longline fishery that had exclusively targeted the species in the North Pacific since the 1920s (Nakamura 1950) had reduced the abundance of large albacore tuna well before the 1950s.

Pelagic fish species, such as albacore tuna and yellowfin tuna, have been commercially harvested in the Pacific Ocean off California since the 1920s, around Hawaii since 1917, and in the northwestern Pacific and southeastern Asia since the 1920s. By the late 1940s, catches of yellowfin tuna amounted to at least 136 708 t per year (Nakamura 1950, Otsu 1954, Mais and Jow 1960). Fishing would have reduced the abundance of these highly migratory species over a wide area of the Pacific Ocean before the 1950s. Many of the masters of commercial longliners in the early 1950s believed that tuna abundance was substantially lower than it was in earlier years (Shimada 1951).

Soon after the 1950s survey, longlining expanded rapidly to high levels over large geographical scales (Sakagawa et al. 1987). Catches of most pelagic species have been high in the Pacific Ocean since the 1970s. Longline catches of yellowfin tuna in the study region, for instance, reached about 50 000 t per year by 1979, with few caught by other fishing gears (Secretariat of the Pacific Community public communication, <http://www.sidsnet.org/pacific/spc/OceanFish/html/statistics>, accessed 10 June 2003). In subsequent years, the longline catch of yellowfin tuna averaged 30 000 t in the study region, with an additional 20 000 t of smaller yellowfin tuna caught by pole-and-line and purse-seine gear each year. We expect future analyses to demonstrate that, by the early 1960s, those levels of exploitation were sufficient to have caused the changes in the pelagic fish community identified by our study.

LITERATURE CITED

Angel, M. V. 1993. Biodiversity of the pelagic ocean. Conservation Biology 7:760–772.

Bigelow, K. A., C. H. Boggs, and X. He. 1999. Environmental effects on swordfish and blue shark catch rates in the US North Pacific longline fishery. Fisheries Oceanography 8:178–198.

Bigelow, K. A., J. Hampton, and N. Miyabe. 2002. Application of a habitat-based model to estimate effective longline fishing effort and relative abundance of Pacific bigeye tuna (Thunnus obesus). Fisheries Oceanography 11:143–155.

Chavez, F. P., J. Ryan, S. Lluch-Cota, and M. Niquen C. 2003. From anchovies to sardines and back: multidecadel change in the Pacific Ocean. Science 299:217–221.

Collette, B. B., and C. E. Nauen. 1983. Scombrids of the world. An annotated and illustrated catalogue of tunas, mackerels, bonitos and related species known to date. United Nations Development Programme, Food and Agriculture Organization of the United Nations, Rome.

Froese, R., and D. Pauly. 2003. FishBase. World Wide Web electronic publication. WorldFish Center, www.fishbase.org.

Gillanders, B. M., D. J. Ferrell, and N. L. Andrew. 2001. Estimates of movement and life-history parameters of yellowtail kingfish (Seriola lalandi): how useful are data from a cooperative tagging programme? Marine and Freshwater Research 52:179–192.

Hamley, J. M., and B. E. Skud. 1978. Factors affecting longline catch and effort: II. Hook-spacing. International Pacific Halibut Commission, Scientific Report No. 64, Seattle.

Hilborn, R., and C. J. Walters. 1987. A general model for simulation of stock and fleet dynamics in spatially heterogeneous fisheries. Canadian Journal of Fisheries and Aquatic Sciences 44:1366–1375.

Hinton, M. G., and Nakano, H. 1996. Standardizing catch and effort statistics using physiological, ecological, or behavioural constraints and environmental data, with an application to blue marlin (Makaira nigricans) catch and effort data from the Japanese longline fisheries in the Pacific. Inter-American Tropical Tuna Commission 21:169–200.

Hirayama, N. 1969. Studies on the fishing mechanism of tuna longline: I. Relation between catch and size of the gear. Nihon Suisan Gakkai-shi [Bulletin of the Japanese Society of Scientific Fisheries] 35:546–549.

Jennings, S., and M. J. Kaiser. 1998. The effects of fishing on marine ecosystems. Advances in Marine Biology 34:201–352.

Last, P. R., and J. D. Stevens. 1994. Sharks and rays of Australia. CSIRO Australia, Hobart.

Lehodey, P. 2001. The pelagic ecosystem of the tropical Pacific Ocean: dynamic spatial modelling and biological consequences of ENSO. Progress in Oceanography 49:439–468.

MacCall, A. D. 1990. Dynamic geography of marine fish populations. Washington Sea Grant, University of Washington Press, Seattle.

Mantua, N. J., and S. R. Hare. 2002. The Pacific Decadal Oscillation. Journal of Oceanography 58:35–44.

Murphy, G. I., and R. S. Shomura. 1972. Pre-exploitation abundance of tunas in the equatorial central Pacific. Fishery Bulletin 70:875–913.

Nakamura, H. 1950. The Japanese long-line fishery for tunas. U.S. Fish and Wildlife Service. Commercial Fisheries Review 12:1–26.

Nakamura, I. 1985. FAO species catalogue. Vol. 5. Billfishes of the world. An annotated and illustrated catalogue of marlins, sailfishes, spearfishes and swordfishes known to date. United Nations Development Programme, Food and Agriculture Organization of the United Nations, Rome.

Niska, E. L. 1953. Construction details of tuna long-line gear used by Pacific Ocean Fisheries Investigations. U.S. Fish and Wildlife Service. Commercial Fisheries Review 15:1–6.

Polovina, J. J. 1996. Decadal variation in the trans-Pacific migration of northern bluefin tuna (Thunnus thynnus) coherent with climate-induced change in prey abundance. Fisheries Oceanography 5:114–119.

Ravier, C., and J.-M. Fromentin. 2004. Are the long-term fluctuations in Atlantic bluefin tuna (Thunnus thynnus) population related to environmental changes? Fisheries Oceanography 13: 145–160.

Rodriguez-Sanchez, R., D. Lluch-Belda, H. Villalobos, and S. Ortega-Garcia. 2002. Dynamic geography of small pelagic fish populations in the California Current System on the regime time scale. Canadian Journal of Fisheries and Aquatic Sciences 52:1980–1988.

Rothschild, B. J. 1967. Competition for gear in a multi-species fishery. ICES Journal of Marine Science 31:102–110.

Sakagawa, G. T., A. L. Coan, and N. W. Bartoo. 1987. Patterns in longline fishery data and catches of bigeye tuna, Thunnus obesus. Marine Fisheries Review 49:57–66.

Shapiro, S. 1950. The Japanese longline fishery for tunas. U.S. Fish and Wildlife Service. Commercial Fisheries Review 12:1–26.

Steele, J. H. 1998. Regime shifts in marine ecosystems. Ecological Applications. Supplement 8:S33–S36.

Strasburg, D. W. 1957. Distribution, abundance, and habits of pelagic sharks in the Central Pacific Ocean. U.S. Fish and Wildlife Service. Fishery Bulletin 58:335–361.

Suzuki, Z., Y. Warashina, and M. Kishida. 1977. The comparison of catches by regular and deep tuna longline gears in the western and central equatorial Pacific. Bulletin of the Far Seas Fisheries Research Laboratory 15:51–89.

Uchiyama, J. H., and T. K. Kazama. 1999. Length–weight relationships of Pelagic Management Unit Species from Hawaiian waters. In Ecosystem-based Fishery Management. A report to Congress by the Ecosystem Principles Advisory Panel as mandated by the Sustainable Fisheries Act amendments to the Magnuson-Stevens Fishery Conservation and Management Act of 1996. National Marine Fisheries Service, Washington, D.C.

Van Campen, W. G. 1952. Japanese mothership-type tuna fishing operations in the western equatorial Pacific, June – October 1951 (Report of the seventh, eighth and ninth expeditions). U.S. Fish and Wildlife Service. Commercial Fisheries Review 14:1–9.

Ward, P., R. A. Myers, and W. Blanchard. 2004. Fish lost at sea: the effect of soak time and timing on pelagic longline catches. Fishery Bulletin 102:179–195.

Whitehead, H. 2002. Estimates of the current global population size and historical trajectory for sperm whales. Marine Ecology Progress Series 242:295–304.

Yoshihara, T. 1954. Distribution of catch of tuna longline. IV. On the relation between k and f with a table and diagram. Nihon Suisan Gakkai-shi (Bulletin of the Japanese Society of Scientific Fisheries) 19:1012–1014.


 


TABLE A1. Comparison of 1950s and 1990s fishing gear and techniques. This table is based on survey and observer data, supplemented with information from Niska (1953) and Murphy and Shomura (1972).


Characteristic

1950s

1990s


Source

Pacific Oceanic Fisheries Investigations (1951–1958),
US National Marine Fisheries Service (NMFS), Pacific Islands Regional Office

Observers on Hawaii-based longliners (1994–2002),
US National Marine Fisheries Service (NMFS), Pacific Islands Regional Office

Target species

No targeting

Bigeye tuna and yellowfin tuna

Mainline material

Hard-lay cotton twine

Monofilament

Branchline material

12-strand cotton twine with wire leader

400 kg breaking strain monofilament, 92% of operations used wire leaders

Level of fishing-effort

880 operations
301 520 hooks

505 operations
1 002 637 hooks

Hooks per operation

343 (±269 SD) hooks

1985 (±361 SD) hooks

Hook type

9/0 or 8/0 “Mustad flattened tuna” hooks

“Asian ring” hooks

Bait

Frozen sardine, occasionally herring, milkfish, or squid

Frozen sardine or saury

Lightsticks

No

No

Buoyline length

19.2 (±6.42 SD) m

22.3 (±5.2 SD) m

Branchline length

20.7 (±7.02 SD) m

13.3 (±3.72 SD) m

Hooks per buoy

Usually 6, ranging up to 21

26–30, ranging from 12 up to 38

Line shooter

No

Yes

Depth range

26–200 m

27–600 m

Deployment time

One hour before dawn

About one hour before dawn

Median soak time

7 hours

12 hours

Retrieval time

One hour after noon

About two hours before dusk


 

 

TABLE A2. Sources of 1990s length data. The number of individuals measured (“retained”) and the number visually estimated (“released”) in the study region are shown for the 21 species that had reliable biomass indices. Also shown are the number of each species measured by Secretariat of the Pacific Community observers in the wider area (10°S–15°N, 175°E–115°W).


Common name

Study region

Wider
area

Total

Retained

Released


Tunas and tuna-like species

 

Albacore tuna

30

0

0

30

 

Bigeye tuna

2 627

12

0

2 639

 

Skipjack tuna

1 133

17

0

1 150

 

Wahoo

0

0

41

41

 

Yellowfin tuna

6 241

41

0

6 282

Billfishes

 
 
 
 
 

Black marlin

1

1

12

14

 

Blue marlin

411

5

0

416

 

Sailfish

30

0

0

30

 

Shortbill spearfish

50

3

0

53

 

Striped marlin

191

5

0

196

Other teleosts

 
 
 
 
 

Great barracuda

0

0

3

3

 

Longnosed lancetfish

0

0

476

476

 

Mahi mahi

0

0

2

2

 

Pomfrets

0

1

0

1

 

Snake mackerels

20

0

1

21

Sharks and rays

 
 
 
 
 

Blue shark

317

252

0

569

 

Mako sharks

31

23

0

54

 

Oceanic whitetip shark

281

231

0

512

 

Pelagic stingray

0

1

0

1

 

Silky shark

431

214

0

645

 

Thresher shark

272

126

0

398


 


TABLE A3. Scientific name, numbers caught, numbers of body mass measurements, and mean mass of each species. Unless otherwise indicated, the number measured is for the study region. For the 1990s, the number measured includes visual estimates of the mass of released fish (Table A2). Habitat categories are based on Froese and Pauly (2003) and are listed in decreasing order of importance for each species.


Common name
Scientific name
Habitat
Caught
Body mass measurements
 
 
 
1950s (no.)
1990s (no.)
1950s (no.)
1990s (no.)

Tunas and tuna–like species

 
 

Albacore tuna

Thunnus alalunga

I, II

323

31

208

30

 

Bigeye tuna

Thunnus obesus

I, II

694

2 975

291

2 639

 

Pacific bluefin tuna

Thunnus orientalis

I

0

1

0

0

 

Skipjack tuna

Katsuwonus pelamis

I

438

1 668

163

1 150

 

Wahoo

Acanthocybium solandri

I

100

683

29

41

 

Yellowfin tuna

Thunnus albacares

I, II

10 636

10 625

3 358

6 282

 

Unidentified tunas

0

191

0

0

Billfishes

   
 

Black marlin

Makaira indica

I

38

2

19

14

 

Blue marlin

Makaira nigricans

I

325

459

70

416

 

Broadbill swordfish

Xiphias gladius

I, II

11

159

1

139

 

Sailfish

Istiophorus platypterus

I

25

31

8

30

 

Shortbill spearfish

Tetrapturus angustirostris

I

15

55

11

53

 

Striped marlin

Tetrapturus audax

I

55

215

15

196

 

Unidentified billfishes

11

54

5

44

Other teleosts

 
 

Crestfish

Lophotus lacepede

II

0

4

0

0

 

Great barracuda

Sphyraena jello

I, A

19

97

5

3

 

Longnosed lancetfish

Alepisaurus ferox

II, I

100

1 699

13

476

 

Mahi mahi

Coryphaena hippurus

I

53

190

9

2

 

Opah

Lampris guttatus

I, II

0

5

0

1

 

Pomfrets

 

0

637

   

Bigscale pomfret

Taractichthys steindachneri

III

0

559

0

0

   

Dagger pomfret

Taractes rubescens

III

0

48

0

0

   

Pacific pomfret

Brama japonica

I

0

13

0

1

 

Rainbow runner

Elagatis bipinnulata

A, I

0

17

0

2

 

Snake mackerels

 

23

933

   

Escolar

Lepidocybium flavobrunneum

II, I, A

0

171

0

1

   

Oilfish

Ruvettus pretiosus

A, II, III, I

0

66

0

19

   

Unid. snake mackerels§

F. Gempylidae

II, I

23

696

1

1

 

Sunfishes

 
   

Common sunfish

Mola mola

I

0

4

0

25

   

Pelagic puffer

Lagocephalus lagocephalus

I

0

3

0

0

   

Slender sunfish

Ranzania laevis

I

0

12

0

0

   

Unidentified sunfishes

F. Tetradontidae

I

4

0

2

0

 

Yellowtail amberjack

Seriola lalandei

A

0

4

0

0

 

Unidentified teleosts

0

7

0

0

Sharks and rays

 
 

Blacktip shark

Carcharhinus limbatus

A, I

5

1

0

1

 

Blue shark

Prionace glauca

I, II

696

1 081

32

569

 

Cookiecutter shark

Isistius brasiliensis

I, II

0

3

0

2

 

Crocodile shark

Pseudocarcharias kamoharai

I, II

0

103

0

76

 

Dusky shark

Carcharhinus obscurus

A

0

53

0

25

 

Galapagos shark

Carcharhinus galapagensis

A, I

0

2

0

3

 

Hammerhead sharks

 
   

Scalloped hammerhead

Sphyrna lewini

A, I

0

4

0

3

   

Smooth hammerhead

Sphyrna zygaena

A, I

0

2

0

2

   

Unidentified hammerhead

Sphyrna spp.

5

1

0

0

 

Mako sharks

 

51

72

   

Longfin mako

Isurus paucus

I

21

32

5

23

   

Shortfin mako

Isurus oxyrinchus

I

0

37

0

31

   

Unidentified mako||

Isurus spp.

30

3

2

0

 

Oceanic whitetip shark

Carcharhinus longimanus

I

1 149

794

37

512

 

Rays

 
   

Manta ray

Manta birostris

A, I

0

12

 0

1

   

Pelagic stingray

Dasyatis violacea

I

0

302

 0

1

   

Unidentified rays

0

1

 0

0

 

Silky shark

Carcharhinus falciformis

I, II

2 210

1 080

23

645

 

Thresher sharks

 

112

511

   

Bignose thresher

Alopias superciliosus

A, I, II

0

333

0

293

   

Pelagic thresher

Alopias pelagicus

I, II

0

105

0

94

   

Thintail thresher

Alopias vulpinus

A, I, II

0

18

0

11

   

Unidentified thresher

Alopias spp.

112

55

2

0

 

Tiger shark

Galeocerdo cuvier

A

0

6

0

3

 

Unidentified sharks

1 482

226

0

0

Other unidentified species

8

4

0

0

Total

   

17 439

24 208

4 309

13 860


Habitat categories:

   I = epipelagic zone of the open ocean (0–200 m)

   II = mesopelagic zone of the open ocean (200–1000 m)

   III = benthopelagic zone of the open ocean (immediately above the sea floor)

   A = associated with land masses, such as seamounts, reefs, and islands

Include measurements reported by Secretariat of the Pacific Community observers in the western Pacific Ocean in the 1990s. 1950s data from a wider area (10°S–15°N, 175°E–115°W).

§Most unidentified snake mackerels were probably Gempylus serpens.

||The 1950s survey originally identified these mako sharks as salmon shark (Lamna ditropis). That identification is questionable because salmon shark are distributed between 22°N and 66°N, generally over shelf waters (Froese and Pauly 2003).

 


TABLE A4. Stepwise regression of hook models for nine species. Statistics for the most parsimonious model derived from the MASS function stepAIC are compared with those from the full model that included all variables. The full model was the most parsimonious model for yellowfin tuna and blue shark.


Common name

Variables dropped in
parsimonious model

Statistic

Parsimonious
model

Full
model

     

Tunas and tuna–like species

     
 

Albacore tuna

E

A

0.040

0.053

   

SE

0.351

0.402

   

AIC

1 384

1 385

   
 

Bigeye tuna

E

A

0.413

0.410

   

SE

0.093

0.094

   

AIC

10 771

10 773

   
 

Skipjack tuna

D2

A

2.445

2.415

   

SE

0.118

0.118

   

AIC

7 706

7 707

   
 

Wahoo

abundance

   

SE

0.298

   

AIC

   
 

Yellowfin tuna

none

A

0.166

0.166

   

SE

0.052

0.052

   

AIC

24 253

24 253

Billfish

 

Black marlin

abundance

   

SE

1.393

   

AIC

   
 

Blue marlin

D2+N+E2

A

0.529

0.591

   

SE

0.135

0.148

   

AIC

3 873

3 874

   
 

Sailfish

-

abundance

   

SE

0.208

   

AIC

   
 

Shortbill spearfish

abundance

   

SE

0.275

   

AIC

   
 

Striped marlin

abundance

   

SE

0.131

   

AIC

Other teleosts

 

Great barracuda

abundance

   

SE

0.649

   

AIC

   
 

Longnosed lancetfish

abundance

   

SE

0.325

   

AIC

   
 

Mahi mahi

abundance

   

SE

0.666

   

AIC

   
 

Pomfrets

abundance

   

SE

   

AIC

   
 

Snake mackerels

abundance

   

SE

0.569

   

AIC

Sharks

 

Blue shark

none

A

0.134

0.134

   

SE

0.098

0.098

   

AIC

7 878

7 878

   
 

Mako sharks

abundance

   

SE

0.669

   

AIC

   
 

Oceanic whitetip shark

N2

A

0.266

0.267

   

SE

0.095

0.099

   

AIC

7 503

7 505

   
 

Pelagic stingray

abundance

   

SE

   

AIC

   
 

Silky shark

D2+N

A

0.077

0.078

   

SE

0.079

0.080

   

AIC

9 317

9 319

   
 

Thresher sharks

N

A

0.205

0.173

   

SE

0.218

0.219

   

AIC

2 602

2 602


The full model was: 

where is the mean catch of the species, Pi is the fishing period, Ni is the latitude, and Ei is the longitude of longline operation i; and Ti,s is the soak time, Di,s is the depth below the sea-surface, and hi,s is the number of vacant hooks of each stratum s of operation i. The j are estimated parameters.

Symbols and abbreviations:

A = Estimated change in abundance between the 1950s and 1990s

SE = Standard error

AIC = Akaike's information criterion (smaller is better).

 

 

TABLE A5. Input data and parameter estimates for 21 species that had well-determined indices of abundance and were not associated with land masses. The negative binomial model parameter  is the amount of overdispersion in the data relative to the Poisson distribution (very large values of  indicate a Poisson distribution).


Common name

Model
type

No. modeled

Abundance change

 

1950s

1990s

Estimate

SE

Estimate

SE


Tunas and tuna–like species

           
 

Albacore tuna

hook

 262

 31

0.053

0.402

0.364

0.628

 

Bigeye tuna

hook

 599

2 573

0.410

0.094

0.237

0.239

 

Skipjack tuna

hook

 311

1 570

2.413

0.118

0.190

0.420

 

Wahoo

offset

 99

 646

1.514

0.298

1.285

0.162

 

Yellowfin tuna

hook

6 284

9 739

0.166

0.052

0.254

0.080

Billfishes

 
 

Black marlin

offset

 32

 1

0.050

1.393

0.698

0.881

 

Blue marlin

hook

 234

 421

0.591

0.148

0.154

1.276

 

Sailfish

offset

 16

 26

0.062

0.208

0.261

0.179

 

Shortbill spearfish

offset

 15

 51

0.015

0.275

0.345

0.144

 

Striped marlin

offset

 47

 179

0.189

0.131

27.473

0.047

Other teleosts

 
 

Great barracuda

offset

 16

 93

0.556

0.649

0.732

0.969

 

Longnosed lancetfish

offset

 91

1 618

2.596

0.325

1.016

0.136

 

Mahi mahi

offset

 42

 177

0.501

0.666

0.223

1.144

 

Pomfrets

operation

 0

 561

>1.000

0.351

0.355

 

Snake mackerels

operation

 22

 870

4.212

0.569

0.712

0.153

Sharks and rays

 
 

Blue shark

hook

 618

 942

0.134

0.098

0.398

0.325

 

Mako sharks

offset

 37

 65

0.319

0.669

0.193

2.367

 

Oceanic whitetip shark

hook

1 031

 713

0.266

0.099

0.206

0.404

 

Pelagic stingray

offset

 0

 301

>1.000

2.366

0.222

 

Silky shark

hook

1 872

 972

0.078

0.080

0.247

0.241

 

Thresher sharks

hook

 108

 468

0.183

0.219

0.125

1.221


Hook models used information on the depth and soak time of each hook; offset models used the median soak time of each operation and an offset based on the number of hooks per buoy and the species’ depth distribution (Table A10); operation models used the median soak time and the number of hooks per buoy as an indicator of the operation’s depth range.

We limited modeling to records that had complete input data, e.g., hook depth was required for the depth model. Consequently, the number modeled is often less than the number caught.

 


TABLE A6. The effect of latitude and longitude on abundance indices of ten species. We fitted the hook model to catch and effort data from seven progressively smaller areas of the tropical Pacific. Values are the change in abundance between the 1990s and 1950s (standard errors in parentheses). Shading indicates the designated study region.


Common name

Area


   

10°S-11°N

10°S-11°N

10°S-11°N

10°S-11°N

10°S-11°N

0-11°N

3-8°N

   

175-245°E

175-235°E

175-225°E

175-215°E

175-205°E

183-217°E

194-200°E

Tunas

 

Albacore tuna

0.053

0.049

0.051

0.059

0.046

0.047

     

(0.999)

(0.984)

(0.965)

(0.857)

(0.904)

(0.537)

 

Bigeye tuna

0.410

0.397

0.362

0.331

0.683

0.200

0.688

   

(0.119)

(0.122)

(0.127)

(0.140)

(0.309)

(0.176)

(0.361)

 

Skipjack tuna

2.415

2.412

2.465

2.427

2.263

3.406

2.837

   

(0.230)

(0.226)

(0.225)

(0.266)

(0.451)

(0.469)

(3.734)

 

Yellowfin tuna

0.166

0.163

0.165

0.122

0.140

0.100

0.179

   

(0.068)

(0.069)

(0.070)

(0.075)

(0.100)

(0.086)

(0.195)

Billfish

 

Blue marlin

0.591

0.583

0.624

0.643

0.236

0.432

0.297

   

(0.179)

(0.182)

(0.182)

(0.185)

(0.239)

(0.222)

(0.648)

Sharks

 

Blue shark

0.134

0.132

0.133

0.135

0.143

0.106

0.100

   

(0.102)

(0.103)

(0.105)

(0.114)

(0.144)

(0.143)

(0.179)

 

Oceanic whitetip shark

0.267

0.269

0.277

0.322

0.354

0.402

0.418

   

(0.117)

(0.118)

(0.121)

(0.121)

(0.159)

(0.168)

(0.833)

 

Silky shark

0.078

0.080

0.082

0.060

0.034

0.052

0.027

   

(0.087)

(0.088)

(0.091)

(0.105)

(0.109)

(0.103)

(0.196)

 

Thresher sharks

0.138

0.138

0.139

0.106

0.125

0.117

0.347

   

(1.722)

(1.920)

(1.821)

(0.697)

(0.864)

(0.687)

(7.568)


 


TABLE A7. The joint effects of latitude and longitude on abundance indices. We applied hook models that included a latitude–longitude interaction term to nine species.  The table shows statistics for the interaction term (“LatxLon”) and estimates of the change in abundance between periods for models with and without the interaction term.


Common name

Estimate of
LatxLon

Standard
error

Confidence limits

Chi-
square

Pr>
Chi-square

Estimated change
in abundance

lower

upper

with
LatxLon

without
LatxLon


Tunas

                 
   

Albacore tuna

0.0027

0.0021

–0.0015

0.0068

1.5900

0.2069

0.058

0.053

 

Bigeye tuna

0.0002

0.0008

–0.0015

0.0019

0.0500

0.8158

0.412

0.410

 

Skipjack tuna

–0.0019

0.0011

–0.0041

0.0004

2.6200

0.1055

2.303

2.413

 

Yellowfin tuna

–0.0037

0.0006

–0.0048

–0.0026

40.1000

<.0001

0.152

0.166

Billfishes

 
 
 
 
 
 
 
 
 

Blue marlin

0.0058

0.0011

0.0036

0.0080

26.7200

<.0001

0.732

0.591

Sharks

 
 
 
 
 
 
 
 
 
 

Blue shark

0.0017

0.0007

0.0003

0.0031

5.9400

0.0148

0.144

0.134

 

Oceanic whitetip shark

–0.9311

0.0899

–1.1073

–0.7550

107.3400

<.0001

0.394

0.266

 

Silky shark

–0.0039

0.0010

–0.0059

–0.0020

16.0600

<.0001

0.072

0.078

 

Thresher sharks

–1.8080

0.2153

–2.2299

–1.3860

70.5100

<.0001

0.173

0.183


For models that included the interaction term, the convergence criterion exceeded the limit. Therefore, parameter estimates may be unreliable for oceanic whitetip shark and thresher sharks.

 


TABLE A8. The effects of hook depth on abundance indices. We applied hook models that included a fishing period – hook depth interaction term and a dummy variable to 12 species. The dummy variable indicated whether hook depth was less than 200 m. The table shows statistics for the interaction term (“Period × Depth”). 


Common name

Estimate of
Period × Depth

Standard
error

Confidence
limits

Chi-
square

Pr>
Chi-square

 
 
 
 

lower

upper

 
 

Tunas

 
 
 
 
 
 
 
 

Albacore tuna

1.6416

0.4976

0.6664

2.6168

10.8900

0.0010

 

Bigeye tuna

0.4363

0.0809

0.2779

0.5948

29.1200

<0.0001

 

Skipjack tuna

0.3755

0.1064

0.1670

0.5839

12.4600

0.0004

 

Yellowfin tuna

0.2486

0.0504

0.1497

0.3474

24.2700

<0.0001

Billfishes

 
 
 
 
 
 
 

Blue marlin

0.1981

0.1791

–0.1530

0.5492

1.2200

0.2687

 

Shortbill spearfish

–0.5620

0.6571

–1.8498

0.7258

0.7300

0.3924

 

Striped marlin

0.1972

0.2769

–0.3456

0.7400

0.5100

0.4764

 

Sailfish

–0.2744

0.7031

–1.6525

1.1037

0.1500

0.6963

Sharks

 
 
 
 
 
 
 

Blue shark

0.1585

0.1111

–0.0593

0.3762

2.0300

0.1537

 

Mako sharks

0.5945

0.4196

–0.2280

1.4170

2.0100

0.1566

 

Silky shark

0.2549

0.1144

0.0307

0.4790

4.9700

0.0258

 

Thresher sharks

–1.7228

1.3611

–4.3904

0.9448

1.6000

0.2056


 


TABLE A9. The effects of hook depth on abundance indices. We applied hook models to 10 species, restricting input data to strata where depth was less than 200 m. The table shows statistics for the fishing period term, which represents the change in abundance between the two periods. Also shown are estimates of abundance change derived from the same models that used all depth strata. 


Common name

Estimate of
fishing
period

Standard
error

Confidence
limits

Chi-
square

Pr>
Chi-
square

Estimated change
in abundance

lower

upper

depth < 200 m

all depths


Tunas

 
 
 
 
 
 
 
 
 
 

Albacore tuna

–1.7889

0.4243

–2.6206

–0.9573

17.77

<.0001

0.136

0.053

 

Bigeye tuna

–0.7895

0.1041

–0.9935

–0.5855

57.54

<.0001

0.453

0.410

 

Skipjack tuna

1.1928

0.1228

0.9520

1.4335

94.3

<.0001

2.487

2.413

 

Yellowfin tuna

–1.2389

0.0523

–1.3414

–1.1365

561.54

<.0001

0.187

0.166

Billfishes

 
 
 
 
 
 
 
 
 

Blue marlin

–0.4064

0.1343

–0.6696

–0.1431

9.15

0.0025

0.523

0.591

Sharks

 
 
 
 
 
 
 
 
 

Blue shark

–1.8911

0.1044

–2.0957

–1.6865

328.31

<.0001

0.154

0.134

 

Mako sharks

–1.3492

0.3701

–2.0746

–0.6238

13.29

0.0003

0.208

0.319

 

Oceanic whitetip shark

–1.2059

0.1030

–1.4077

–1.0041

137.17

<.0001

0.299

0.266

 

Silky shark

–2.4797

0.0829

–2.6422

–2.3172

894.85

<.0001

0.079

0.078

 

Thresher sharks

–1.4610

0.2517

–1.9544

–0.9677

33.69

<.0001

0.244

0.183


 


TABLE A10. Parameters used in the estimation of depth effects. We used the parameters in offset models to derive abundance indices of ten species. Myers and Ward (public communication) estimated the parameters from four longline datasets.


Common name

Coefficient

1

2

3


Tuna-like species

       
 

Wahoo

–5.68

–6.88

4.46

4.95

Billfishes

 

Black marlin

–6.06

–9.48

22.77

–16.81

 

Sailfish

–5.81

–8.58

8.61

2.61

 

Shortbill spearfish

–4.79

–7.95

4.31

3.34

 

Striped marlin

–5.07

–3.99

–2.14

6.38

Other teleosts

 

Great barracuda

–4.91

–8.86

–45.90

178.29

 

Longnosed lancetfish

–6.27

–20.84

93.31

–74.91

 

Mahi mahi

–4.41

–25.17

86.37

–108.34

Sharks and rays

 

Mako sharks

–6.14

–9.11

26.32

–22.57

 

Pelagic stingray

–5.85

–9.98

28.79

–24.50


For each species, the coefficients were used to estimate the depth effect  in operation i.  is the mean effect of hook depth D on catch rate relative to that at a depth of 0.175 km over all hooks deployed in the operation:

 


TABLE A11. Comparison of mean catch rates of tunas and billfishes reported by the 1950s survey in the study region (10°S–11°N, 175°E–115°W; 1951–1958) and mothership expeditions during 1952–1954 in the area bounded by 20°S–10°N and 150°E–130°W. Standard deviations are in parentheses.


Common name

Mothership

Survey

 

(no./1000 hooks)

(no./1000 hooks)


Tunas

   
   

Albacore tuna

4.0

2.1

   

(10.5)

(8.1)

   
 

Bigeye tuna

9.0

3.9

   

(8.2)

(10.3)

   
 

Skipjack tuna

0.4

1.7

   

(1.1)

(4.5)

   
 

Yellowfin tuna

25.8

35.8

   

(19.6)

(48.5)

   

Billfishes

 

Black marlin

0.3

0.2

   

(0.9)

(1.1)

   
 

Blue marlin

6.0

3.4

   

(3.5)

(9.9)

   
 

Broadbill swordfish

0.1

0.1

   

(0.2)

(0.7)

   
 

Sailfish and spearfish

0.7

0.2

   

(1.1)

(1.5)

   
 

Striped marlin

0.4

0.3

   

(2.3)

(1.6)

Total

46.8

47.6

   

(20.8)

(52.8)


Standard deviations may not represent the true variance among operations because the original data were aggregated by month and five-degree square.

 


TABLE A12. Sources of information on oceanographic conditions in the study region. We derived statistics presented in the text from oceanographic data and indices for 1951–1958 (1950s study period) and 1999–2002 (1990s study period).


Index name

Description

Source


Pacific Decadal Oscillation

An index based on anomalies in sea-surface temperature and sea-level pressure in the North Pacific.

Joint Institute for the Study of the Atmosphere and Oceans
ftp://ftp.atmos.washington.edu/mantua/pnw_impacts/INDICES/PDO.latest
accessed 9 April 2003

Southern Oscillation Index

An index of the difference in air pressure between Tahiti and Darwin.

Climate and Global Dynamics Division
http://www.cgd.ucar.edu/cas/catalog/climind/SOI.signal.annstd.asciiaccessed accessed 9 April 2003

Thermocline depth (1950s)

Depth of the 20°C isotherm derived from bathythermographs taken by survey longliners during 1950–1953.

NOAA Satellites and Information
http://www.nodc.noaa.gov
accessed 16 May 2003

Thermocline depth (1990s)

Depth of the 20°C isotherm from nine Tropical Atmosphere Ocean moorings in our study region.

NOAA Pacific Marine Environmental Laboratory
http://www.pmel.noaa.gov/tao/
accessed 23 May 2003


 

 
   FIG. A1. Bathymetric map of the central tropical Pacific, indicating the study region and the extent of 1950s (broken line) and 1990s (solid line) fishing activity.


 
   FIG. A2. Catch rates of the 21 species most frequently caught in the study region. Horizontal bars are 95% confidence intervals for the mean catch rate in each period. Catch rates are not standardized for variations in fishing activities.

 

 
   FIG. A3. Comparison of the size composition of species measured and estimated by observers in the 1990s in the wider study area (10°S–15°N, 175°E–115°W). Visual estimates of the body mass of released fish (“Rel.”) are compared to measurements of fish that were brought on board (“Ret.”). We predicted mass from length–mass relationships applied to length measurements. The boxplots show the interquartile range (IQR), which is the difference between the first and third quartiles. Boxplots are inside-out where the sample size is small and the confidence interval (CI) is wider than the interquartile range.

 



[Back to E086-043]