Power and May: Sea-surface temperatures and Thunnus albacares catch and effort 



435 



the expected seasonal progression of SST (Fig. 7). Of 

 note is the more rapid warming of SST in spring of 

 1987 compared with the spring of 1988. The SST coef- 

 ficient of variation can be viewed as a broad index of 

 how "structured" the sea-surface temperature is (Fig. 

 8). There was a clear seasonal trend to this statistic 

 during 1986-87, indicating that considerable spatial 

 variation in SST was present from the winter through 

 early spring. The warming of SST during the spring 

 months to more isothermal conditions, mentioned 

 previously, is coincident with a decline in the SST coef- 

 ficient of variation. 



Image and yellowfin tuna 

 CPUE relationships 



Regions of rapid sea-surface temperature change 

 appear as lighter lines in the gradient image (Fig. 9). 

 Superimposed on this image are the locations of long- 

 line sets (crosses), and the circular region encompassed 

 by the length of the longline set. Plots of yellowfin tuna 

 CPUE versus mean circular polygon SST, SST coeffi- 

 cient of variation, mean polygon gradient, and polygon 

 gradient coefficient of variation, respectively, indicated 

 no apparent relationship between CPUE and these 

 statistics (Figs. 10-13, computed for polygons with 

 radii equivalent to set length). This result was also true 

 when examining plots of yellowfin tuna CPUE versus 

 the polygon statistics partitioned by month and per- 

 centage cloud cover, and for polygon statistics com- 

 puted using the more restricted regions encompassed 

 by one half and one quarter of the set length. 



Discussion 



Although the results of other studies support the 

 hypothesis that tuna are more abundant near thermal 

 fronts (Laurs et al. 1984, Maul et al. 1984, Fiedler and 

 Bernard 1987), we were unable to detect any relation- 

 ship between yellowfin tuna CPUE and SST structure 

 in the northwestern Gulf of Mexico during 1986-87. 

 Our results therefore seem to contradict, at least for 

 the northwestern Gulf of Mexico, the belief among 

 longline fishermen that tuna and other oceanic fish 

 aggregate in regions of rapid temperature change. The 

 perception of increased fishing success near fronts has 

 apparently been incorporated into the fishing strategy 

 used by the longline fleet, since fishermen monitor SST 

 and other environmental indicators to decide where to 

 set the gear. However, our data represent the initial 

 stages of this developing fishery, and the longliners 

 may have employed this strategy in the absence of 

 alternative information concerning where best to locate 

 their gear. 



APRS6 JUL86 OCT86 JAN87 APR87 JULB7 OCT87 JAN8B APR88 JUL88 OCT88 



Date 



Figure 8 



Coefficient of variation of sea-surface temperature (SST) for 

 the entire Gulf of Mexico derived from Advanced Very High 

 Resolution Radiometer imagery, 1986-88. Curve is fit using 

 robust locally-weighted regression. 



We nonetheless accept that under appropriate cir- 

 cumstances, oceanic fish orient to and aggregate at 

 thermal features. Hence, there may be several explana- 

 tions why we did not detect any associations. Only one 

 set of geographic coordinates was recorded for each 

 longline set, and it was not known whether the loca- 

 tion represented the beginning, midpoint, or end of the 

 set. By comparison, the positive albacore-front associa- 

 tions reported by Laurs et al. (1984) were obtained 

 using data from trolling vessels. In that case, fishing 

 effort could be located more precisely, both in terms 

 of the fisherman's strategy and with respect to analyz- 

 ing the resultant CPUE relative to SST patterns. Since 

 longlines used by the Gulf fleet may exceed 50 miles 

 in length, the uncertainty associated with the unknown 

 orientation of the set could have masked a relationship 

 between yellowfin tuna CPUE and the satellite-derived 

 SST structure. Although the actual orientation of a set 

 would be difficult to determine, given the effects of 

 wind and currents during the time the gear was fished, 

 information on the coordinates of each end of the 

 longline during payout and haulback would provide 

 some insight into the spatial orientation of the set, and 

 in turn enable a more refined analysis of tuna-tem- 

 perature associations. Finally, information on the times 

 of payout and haulback would be valuable for refining 

 estimates of fishing effort and selecting satellite images 

 nearer the actual times of fish capture. 



An alternative analytical approach would be to define 

 specific fronts in the imagery, and examine CPUE 

 versus distance from an identified front as the measure 

 of the association between fish and front. This is the 



