FISHERY BULLETIN: VOL. 86, NO. 2 



relatively insensitive to the value of the cluster- 

 ing parameter. However, the fact that a large per- 

 centage of the clusters tend to be dominated by 

 either dolphin or non-dolphin sets is not a neces- 

 sary consequence of the clustering algorithm and 

 suggests that the two types of methods for locat- 

 ing and catching tuna tend to be spatially and 

 temporally distinct. Encounter rates are substan- 

 tially lower in clusters dominated by non-dolphin 

 sets than when a vessel is searching between 

 clusters (Table 5). This result is also not a neces- 

 sary consequence of the clustering procedure and 

 suggests that these clusters not only define areas 

 of high densities of spotted dolphin schools but 

 also areas of low densities. 



Encounter rates could be lower in non-dolphin 

 areas because of differences in detectability not 

 related to the density of schools. For example, 

 Hammond'* suggested that the crew may scan 

 closer to the vessel when searching in non- 

 dolphin areas (see also Polacheck 1983). It seems 

 unlikely that such factors could account for all of 

 the differences between the encounter rates in 

 Table 5. 



Differences in detectability due to differences 

 in weather conditions between and within clus- 

 ters could also affect the results in Table 5. En- 

 counter rates do decrease at higher Beaufort sea 

 states (Polacheck 1983). However, little search- 

 ing occurs above Beaufort state 4. The difference 

 in encounter rates at Beaufort 0-2 compared with 

 Beaufort 3-4 (about a factor of 1.28) is insuffi- 

 cient to explain the difference in Table 5 

 (Polacheck 1983). Moreover, areas of non-dolphin 

 sets tend to be in nearshore areas with calmer 

 seas and fishermen do not consider Beaufort 4 

 conditions as being too rough to fish. 



Do Tuna Catches Differ Between 

 and Within Clusters? 



The average tons of tuna caught per set tend to 

 be greater for sets which occur within clusters 

 than sets between clusters (Fig. 9). For all values 

 of the clustering parameter, the average catch per 

 set was greater within than outside of clusters in 

 approximately 70% of the cruises (Table 6). Para- 

 metric statistical comparisons of the average tons 

 per set are not appropriate because of the large 

 differences in the average catch per set among 



vessels (Fig. 9). A nonparametric sign test (Sned- 

 cor and Cochran 1966) suggests that the differ- 

 ences in catch per set between and within clusters 

 are significant at least at the 0.05 probability 

 level for all values of the clustering parameters 

 that were considered. 



60 t- 



56 



CO 

 QC 

 liJ 



t- 

 V) 



u 

 _l 



o 



z 



UJ 

 UJ 



I- 



LLI 



UJ 

 Ui 



QC 



LU 



a 



I- 



X 



o 



< 

 O 



</) 



z 

 o 



I- 



z 

 < 



LU 



2 



4 8 12 16 20 24 28 32 



MEAN TONS CAUGHT PER SET WITHIN CLUSTERS 



Figure 9. — The average tons of tuna caught per set while a 

 vessel was travelling between clusters versus the average tons 

 per set within clusters with at least three members. The dashed 

 line represents the expected value if the catch rates were equal. 

 The value of the cluster parameter equals 100. 



Table 6. — The ratio of the tons of tuna caught per set within clus- 

 ters with at least 3 members compared to tons caught per set 

 between clusters. The means are the average values for the ratio 

 within a cruise. Cruises in which 100% of the sets were within 

 clusters are not inlcuded in the results. 



■♦Hammond, P. S. 1981. Some problems in estimating the 

 density of dolphin populations in the eastern tropical Pacific 

 using data collected aboard tuna purse seiners. Inter-Am. 

 Trop. Tuna Comm. Intern. Rep. 



' P < 05 of getting the observed number if the ratio equaled 1 by a sign test 

 (Snedecor and Cochran 1966). 



362 



