Distribution Patterns of Three Dominant Tuna Species in the Indian Ocean

We analyzed fishery data provided by Taiwanese fishermen between 1967 and 1996 to investigate monthly and spatial distribution, abundance patterns, and environmental characteristics of three dominant tuna species (bigeye tuna, Thunnus obesus; yellowfin tuna, T. albacares; and albacore, T. alalunga) inhabiting the Indian Ocean using a GIS approach. Environmental preferences of these species were characterized by the monthly composites of sea surface temperature (SST) measured from AVHRR images and chlorophyll concentration derived from the Coastal Zone Color Scanner (CZCS) instrument flown aboard the Nimbus-7 satellite. Results indicate that these species show distinct distribution patterns and minor monthly variations. The peak abundance regions exhibit unique environment conditions. SST mean and range values for the three species occurred in the peak abundance regions are different. Chlorophyll concentration index indicates that the yellowfin tuna occur in regions with higher values than the other two species. the use of discriminant function analysis to predict four abundance classes on the three species' monthly distribution has overall prediction accuracy range from 62.4 to 76.1% for albacore, 52.6-68.0% for bigeye tuna and 58.6-70.9% for yellowfin tuna. With the help of remote sensing data and analyses conducted in a GIS environment, our results show some promise on predicting marine species' abundance pattern in the Indian Ocean.

Introduction

Tunas are among the largest, most specialized and commercially important of all fishes (Collette and Nauen, 1983). Belonging to the genus Thunnus of the family Scombridae, they are found in temperature and tropical oceans around the world and account for a major proportion of the world fishery products. Biologically, tuna species have complex life history. They have streamlined bodies and vary extensively in size, color and fin length. Tunas are unique among fishes because they possess body temperature several degrees higher than the ambient waters and have high metabolic rates that enable them to exhibit extraordinary growth patterns. Tunas are fast swimmer and capable of traveling more than 48 km/h (Collette and Nauen, 1983). They are migratory and have few predators.

Tunas are in great demand throughout the world market due to their excellent meat quality (Chang and Lin, 1995; FAO, 1997). The tuna industry has been a successful program in the past. However, problems on the fishing stock status have occurred due to the recently increased intensity.

There is an urgency to conserve the tuna resources due to the awareness of decreasing tuna stocks (Fonteneau, 1995, 1997; Hsu, 1994, 1998). Although the data for the principal market tuna species have been accumulated for a long time, uncertainties exist in the basic distribution of these species. There have been few studies on large-scale distribution of tuna species (Hsu, 1994). General distribution description indicates that the tunas are widely distributed among the three major oceans, though they never extend to the polar region (Collette and Nauen, 1983; Kikawa and Ferraro, 1966; Laevastu and Rossa, 1962). Information on where and when tunas occur can be critical for resource management and practical for fishery vessels. Despite the abundant literatures on the tuna resources in the Indian Ocean, there appears to have a lack on the study of distribution pattern of these tunas (Lee, 1990, 1995; Kikawa and Ferraro, 1966; Lee and Liu, 1995; Petit et al., 1995; Wu and Chen, 1994; Yeh et al., 1995 and see Hsu, 1994 for review). To better conserve and maintain a sustainable yield of these species, there is a need to understand large-scale patterns of tuna stocks in space and time.

Recently large-scale data obtained by remote sensing technique have become available to the scientific community through the Internet. Data, such as sea surface temperature (SST) and chlorophyll concentration, can be downloaded from NASA's web site or a CD-ROM can be requested. With the help of a GIS, this distribution mechanism has greatly enabled scientific researches to incorporate large volume of important data in large-scale analyses.

In this paper, we documented the distribution patterns of three dominant tuna species, i.e., albacore (Thunnus alalunga), bigeye tuna (T. obesus) and yellowfin tuna (T. albacares), in the Indian Ocean based on the catch data recorded by Taiwanese vessels between 1967 and 1996 using a GIS-approach and investigated the characteristics of high abundance (yield) regions with SST and chlorophyll concentration. Finally, we applied a discriminant function analysis to predict monthly distribution pattern using these variables.

Methods

The Indian Ocean, having about 20% of the global tuna production, is the second largest proportion of principal tuna market in the world (FAO, 1997). Japan, Taiwan and Korea are the major fishing countries in the Indian Ocean. In recent years, tuna fisheries are growing in this ocean, partly due to the catching of small tunas by the non-traditional tuna catching countries, such as Franch, Ivory Coast, Spain, and etc (FAO, 1997).

Tuna database

Based on fishery statistics, Taiwan is the second largest tuna fishing countries in the world (FAO, 1997). Fishery data in the Indian Ocean are recorded by Taiwanese tuna longline fishing vessels between 1967 and 1996. The data set represent the best available information exist to date on the tuna resources found in the Indian Ocean because of its extensive cover in terms of both space and time. These data were compiled by the Oversea Fisheries Development Council (OFDC) of the Republic of China and include monthly summary of number of catch, total weight, and number of hooks of several tuna species in specific geographic locations. All catch data were georeferenced in a latitude and longitude system (Figure 1). Each grid cell is 5-degree x 5-degree in size. Based on the total catch data, we identify three dominant species: albacore, bigeye tuna and yellowfin tuna. The total production of these three species comprises of more than 75% of all fishery products. Yellowfin and bigeye tunas are the second and third important commercial species of tuna, respectively, while the production of albacore over the past 30 years has been fluctuating (FAO, 1997).

Figure 1. A 5-degree x 5-degree grid cell system for recording the tuna catch in the Indian Ocean

Environmental database

To correlate the oceanographic environment, such as the distribution of SST and chlorophyll concentration, with the fishery data, we obtained these data sets from NASA's web site. Monthly mean distributions of SST and chlorophyll concentration are derived from the Advanced Very High Resolution Radiometer (AVHRR) carried aboard the NOAA-series polar-orbiting satellites and from the Nimbus-7 Coastal Zone Color Scanner (CZCS), respectively. The data were converted into Arc/Info format. All data were converted into 5-degree x 5-degree grid format before performing analyses.

Data Analysis

For each grid cell, we determined three characteristics of tunas' distribution in the Indian Ocean, i.e., CPUE (number of catch per unit of effort, total catches/total hooks), biomass (total weight in kg per unit of effort, total weight/total hooks) and average weight (in kg per catch, total weight/total catches). Because the data were spatially scattered and not evenly distributed, we combined all of the data in different years. Monthly data were obtained. To investigate monthly distribution pattern, we calculated mean values to represent the characteristics of the tuna resources in grid cells. The data treatments for SST and chlorophyll concentration are the same.

ArcView was used to create monthly distribution maps for the three species. Patterns were exploited visually. We then identified the peak abundance regions. The peak abundance regions were defined as the regions that have CPUE or biomass records exceeding a certain limit. We pooled monthly data for each species and calculated the grand mean and its standard deviation. The threshold is equal to the grand means + 0.5 * standard deviation. This value for each species is roughly close to the third quartile of the pooled data. These peak abundance regions are a representation of "productive fishing grounds". All the monthly peak abundance maps were summed to reveal a yearly summary.

We characterized the environmental preferences of the species by overlaying SST and chlorophyll concentration maps with the peak abundance layers in each month. The ranges and mean values of SST and chlorophyll concentration for each species in a particular month were summarized.

A discriminant function analysis (DFA) was performed to predict the CPUE distribution using environmental variables. Monthly CPUE data were grouped into four classes based on the quartiles. Since the monthly SST and chlorophyll concentration data are highly correlated, we performed two separate principal component analyses (PCA) for each data set before using DFA. The PCA results show that there are two PC axes for SST (accounts for 99.7% of the variations) and five axes for chlorophyll concentration (accounts 94.8% of the variations). These axes were combined with topography data, distance to the nearest coastal lines, longitude and latitude as the environmental variables. A stepwise DFA was then performed using these variables.

We used SAS for Windows to perform the calculation and SYSTAT for Windows to generate the statistical graphs. ArcView was used to derive a new variable (distance to the nearest coastal line) and to display the distribution maps. In the figures presented in this paper, four colors were used to represent the abundance classes where quartiles were applied. Darker colors represent higher values of CPUE or weight indexes.

Results

Distribution

Since the results using CPUE and biomass indexes show similar distribution patterns for the three tuna species, CPUE was used to represent the results.

Albacore has the highest CPUE, while bigeye tuna has the lowest. The monthly trends of CPUE for albacore and yellowfin tuna show some degree of fluctuation, but no pattern was detected in bigeye tuna (Figure 2). For albacore, during October-November, the CPUEs are lower than the other months. In contrast, yellowfin tuna shows higher abundance in April and May.

Figure 2. Monthly mean CPUE for albacore, bigeye and yellowfin tuna in the Indian Ocean

The three species are distributed in the entire Indian Ocean, but show distinct abundance patterns. Albacore concentrated in the Southern Hemisphere, especially in the 30-45 degree S regions and across the ocean (Figure 3). Monthly CPUE distributions show some minor shifts of the abundance pattern, but the general trend is the same. Bigeye tuna is more abundant in the regions close to the Equator, especially the regions close to Indonesia. The monthly trend stays roughly the same (Figure 4). The distribution of yellowfin tuna mostly concentrates on the western coast of India (the Arabian Sea), and the regions around Madagascar, especially the regions above Mozambique Channel (Figure 5). In some months, the higher abundance regions extend to the eastern coast of India (Bay of Bengal).

Figure 3. Distribution pattern (CPUE) of albacore in the Indian Ocean. Due to space limitation, only January, April, July and October maps were shown.

Figure 4. CPUE distribution pattern of bigeye tuna in the Indian Ocean

Figure 5. CPUE distribution pattern of yellowfin tuna in the Indian Ocean

An example of the peak abundance map for the species studied are shown in Figure 6. The monthly total maps, based on the peak abundance maps for each species indicate where the high catch regions occur (Figure 7). It is clear that none of the cell values exceed 12, indicating that no regions are always within the peak abundance regions. These maps clearly indicate the locations of the high productivity fishing grounds in the Indian Ocean for the species studied.

Figure 6. Examples of peak abundance regions (yellow color) for albacore, bigeye and yellowfin tunas in the Indian Ocean

Figure 7. "Hotspot" (monthly summary of peak abundance regions) for albacore, bigeye and yellowfin tunas in the Indian Ocean. Color code: blue = 0; yellow=1-2 for albacore, 1-3 for bigeye tuna, and 1-4 for yellowfin tuna; orange=3-4 for albacore, 4-6 for bigeye tuna, and 5-8 for yellowfin tuna; red=5-8 for albacore, 7-10 for bigeye tuna, and 9-11 for yellowfin tuna.

The average weight distribution maps (Figure 8) show that the high productivity fishing grounds don't always have the largest size of individual tuna caught. In most of the cases, low abundance CPUE regions have larger size of fishes. This is particular true for albacore and bigeye tuna.

Figure 8. Average weight distribution for albacore, bigeye and yellowfin tunas in the Indian Ocean.

Environmental characteristics

Using SST and chlorophyll concentration to characterize the peak abundance regions for each of the three tuna species show that they have unique patterns. In SST, albacore is mostly found in the range of 17-26 C (Figure 9). Bigeye tuna distributes in the regions where SST all greater than 25 C. Yellowfin tuna shows similar pattern to bigeye tuna, but has SST values that are lower than bigeye tuna and greater than albacore.

Figure 9. Monthly mean fluctuation of sea surface temperature for albacore, bigeye and yellowfin tunas in the Indian Ocean

The SST differences among these species can also be viewed in the monthly histograms (Figure 10). These figures all indicate certain differences in the SST ranges that are also evident in the spatial distribution maps (Figures 3-5). SST ranges for albacore in each month are larger than those of bigeye and yellowfin tunas, while bigeye and yellowfin tunas show similar results, but not in the same months.

Figure 10. Distribution of sea surface temperature for the peak abundance regions of albacore, bigeye and yellowfin tunas in January, April, July and October in the Indian Ocean

Figure 11. Monthly mean fluctuation of chlorophyll concentration for albacore, bigeye and yellowfin tuna in the Indian Ocean.

The monthly means of chlorophyll concentration for albacore are similar to those of bigeye tuna, while the concentration for yellowfin tuna shows higher values and greater variabilities than the other two species (Figure 11). These results also indicate that yellowfin tuna occur in more productive regions, while the other species distribute in open waters.

Spatial prediction

Discriminant function analysis used to predict the CPUE abundance classes for the three species' monthly distribution indicate that the prediction accuracy range from 62.4 to 76.1% for albacore, 52.6-68.0% for bigeye tuna and 58.6-70.9% for yellowfin tuna (Figure 12).

Figure 12. Overall classification accuracy for predicting the abundance classes of albacore, bigeye and yellowfin tunas in the Indian Ocean using discriminant function analysis

The classification tables for each month's prediction are similar. In the followings we use one example from each species to present the results. For albacore, the regions that are either the lowest or the second highest abundance CPUE class have more cells correctly classified than the other categories (Table 1). The overall accuracy for this example is 74.6%. The case for bigeye tuna is similar with 68% overall classification accuracy. In the regions where the CPUE abundance are either the lowest and the highest, we have better prediction power than the others. In yellowfin tuna, except for the second lowest abundance class, which has only 37.5% of the cells correctly predicted, the others all have accuracy greater than 70.5%. The overall accuracy for this case is 70.9%.

Table 1. Classification table of discriminant function analysis for albacore, bigeye and yellowfin tunas in the Indian Ocean

Albacore

Bigeye Tuna

Yellowfin tuna

Discussion

In the ocean, large-scale distribution patterns in the pelagic environment are predominantly determined by the global thermohaline circulation, seasonality of production cycles and interactions with meso-scale eddies (Angel, 1994). The spatial and temporal scales of tuna data used in this study may limit the exploitation of these patterns in great details. However, our results show some insights on the distribution pattern of tuna in the Indian Ocean. The results greatly improve our current knowledge of the tuna distribution. It provides information for fishery vessels on where and when to catch the resources more efficiently. More importantly, it suggests insights on how to maintain a sustainable yield in the future. With proper population data, such as age structure, migration pattern and recruitment variability, reliable forecasts of abundance classes would greatly help with the setting of quota prior to each fishing year and conservation protocols can be appropriately scaled in order to insure the sustainability of these species.

The studies on the distribution of tuna were complicated by the fact that the ocean is three-dimensional, the targeted species may have different vertical preferences (Hanamoto, 1987) and the methods used to catch the tuna were frequently changed (Hsu, 1998). Current studies on the distribution have generally depended on longline catch data. However, Hanamoto (1987) showed that the vertical distribution for bigeye tuna can be as deep as 600 m, the regions do not receive exploitation using current catching methods. During the past 30 years, many methods have been applied to catch the tunas in the Indian Ocean. Conventionally the resources were exploited using longline gear, surface gear, gillnet, and purse seiner. Japan, Taiwan, Korea, France, Ivory Coast, Australia, Spain, Kenya, Mauritius, Mozambique and Russia took part in the fishery (Hsu, 1994).

Further studies are expected to differentiate the impact of different catching methods on the CPUE calculation and to model the distribution pattern by incorporating more environmental variables. The adjustment of nominal CPUE has been an active research subject for many years (Hsu, 1998). In our analyses, data were aggregated over 30 years to reflect monthly distribution patterns and to alleviate this effect. However, this treatment may be too general to ignore the effect of large-scale changes, such as El Nino. We expect to include variables in DFA prediction that will include salinity, dissolved oxygen, and many other measurements derived from satellite images and spatial autocorrelation. In addition, distribution of food resources for tunas may be an important and direct factor that contributes to the understanding of the distribution. No data were available at this time to explore these relationships (Caddy and Rodhouse, 1998).

Acknowledgments

This study was funded by the Council of Agriculture, Republic of China. Fishery data were provided by the Oversea Fisheries Development Council of the Republic of China. The authors wish to thank the Distributed Active Archive Center (Code 902.2) at the Goddard Space Flight Center, Greenbelt, MD, 20771, for producing the CZCS data in its present format and distributing them. The original data products were produced by the Nimbus Project Office in collaboration with the NASA Goddard Space Flight Center Space Data and Computing Division, the NASA GSFC Laboratory for Oceans, and the University of Miami Rosenstiel School of Marine and Atmospheric Science. Goddard's share in these activities was sponsored by NASA's Mission to Planet Earth program. The SST data were obtained from the NASA Physical Oceanography Distributed Active Archive Center at the Jet Propulsion Laboratory / California Institute of Technology.

References

E-mail: leepf@ccms.ntu.edu.tw
Department of Zoology
National Taiwan University
Taipei 10617
TAIWAN
886-2-23623501
886-2-23636837 (Fax)