Results 



The size structure of tilefish caught off Georgia 

 was typical of unexploited to lightly exploited tilefish 

 stocks (Grimes et al. 1980; Turner et al. 1983). This 

 size structure remained relatively constant for ~ 10 

 mo, after which a slight decrease in catch rates and a 

 possible truncation of size structure were observed 

 (authors' unpubl. data). These results confirm verbal 

 reports that little exploitation has occurred off 

 Georgia (Harrington footnote 1). Hence, the data 

 used in this analysis were probably not influenced by 

 prior exploitation. 



A total of 323 tilefish were taken on 19 longline 

 sets (Table 1). Catch rates ranged from to 5.34 tile- 

 fish/100 hook-h. Parameter estimates for linear and 

 quadratic terms of the polynomial regressions were 

 significantly different from zero (Table 2). Inclusion 

 of a cubic term, however, did not significantly im- 

 prove {F = 0.75, P > 0.40) the fit which was obtained 

 using a second-degree polynomial. The second- 

 degree polynomial yielded a higher K~ value than the 

 nonlinear exponential model (Table 2) and hence was 

 deemed to be the model of best fit. The ?/-intercept of 

 this model also was not significantly different than 

 zero (Table 2, Fig. 1) which contributes to its biolog- 

 ical realism. Using this equation, 74% of the varia- 

 tion in catch rate could be accounted for by substrate 

 composition alone. 



(O 



Table 2.— Comparison of regression models. 

 Either F-tests (b-,), f-tests (too), or asymptotic 

 confidence intervals (exponential model) were 

 used to test the significance of parameters. 



Model 



'1 



'0 



/?2 



y = 0.087X - 1.496 ** * 0.64 



y = 0.155{e0058X) • ps 0.68 



y = 0.002X2 _ 0.050X + 0.122 *** ns 0.74 



ns = nonsignificant 



* = P < 0.05 



** = P< 0.001 



*** = P< 0.0001 



Figure 1.- Relationship between the silt-clay fraction of the 

 sediments and tilefish catch rates off Georgia, U.S.A. 



Discussion 



Tilefish abundance, as estimated by catch rates off 

 Georgia's continental slope, was strongly correlated 

 with the silt-clay fraction of the substrate. This rela- 

 tionship was nonlinear, and based on W- values, a 

 second-degree polynomial regression provided the 

 best fit to the data. Off the northeastern United 

 States, tilefish also were most abundant on fine- 

 grain sediments (Able et al. 1982), although they 

 were also found in horizontal burrows in the sides of 

 submarine canyons (Warme et al. 1977), and in 

 boulder fields (Valentine et al. 1980). Because tilefish 

 construct vertical burrows in the substrate (Able et 

 al. 1982), they require sediments which possess suffi- 

 cient stability to prevent the collapse of their bur- 

 rows. It would appear that bottom areas off Georgia 

 which contain a sand fraction > 60% do not support 

 substantial tilefish densities (Table 1, Fig. 1). It is 

 likely that such substrates are not stable enough to 

 allow tilefish to construct burrows. Thus, the ob- 

 served correlation between catch rate and substrate 

 composition has a biologically realistic explanation: 

 substrates with high silt-clay fractions are conducive 



to the construction and maintenance of tilefish bur- 

 rows, while substrates with high sand fractions are 

 not. A similar explanation, based on submarine 

 observations, has been proposed by Able et al. (1982) 

 to explain tilefish distributions off the northeastern 

 United States. Although we have not observed tile- 

 fish burrows off Georgia, they have been identified in 

 soft bottom areas off South Carolina (R. Jones'*). 



While the relationship between catch rates and 

 sediment composition is quite strong, several poten- 

 tial sources of error exist in our data. First, catch 

 rate data were collected from two different vessels 

 using different gear. Pooling data from the different 

 vessels, however, would tend to obscure the relation- 

 ship between catch rates and sediment composition. 

 Hence, if differences in sampling methods did have 

 an effect on our data, it would make the estimates of 

 the catch rate-sediment relationship conservative. 



Second, only one substrate sample was collected 

 with each longline set. While quantification of 



*R. Jones, Harbor Branch Foundation, Fort Pierce, FL 33450, 

 pers. commun. 1983. 



445 



