HIGHTOWER and GROSSMAN: TILEFISH FISHERY 



RESULTS AND DISCUSSION 



Our approach for estimating virgin recruitment 

 level was similar to stock reduction analysis (SRA) 

 (Kimura et al. 1984; Kimura 1985) except that we 

 used a more general model to represent stock 

 dynamics. In both approaches, a model is fully spe- 

 cified and the sequence of J^'s used to drive the model 

 are those that would have produced the observed 

 sequence of catches. A range of solutions can be ob- 

 tained corresponding to a range of virgin recruit- 

 ment levels, but the solution set can be restricted 

 by comparing model predictions to auxiliary data. 



We obtained similar 9-yr patterns of F at differ- 

 ent levels of virgin recruitment, particularly at 

 higher recruitment levels where Fs were low in all 

 years (Fig. 1). For that reason, correlation coeffi- 

 cients were similar over a wide range of recruitment 

 levels (Fig. 2). Stronger conclusions about the true 

 level of virgin recruitment may be possible once 

 additional years of catch and CPUE data become 

 available. 



Based on commercial snapper reel CPUE data, 

 the virgin recruitment level that maximized the cor- 

 relation between estimates of F and fishing effort 

 ranged from 30,000 to 100,000, depending on the 

 assumed selectivity parameters and level of natural 

 mortality (Table 4, Fig. 2). Correlations were high 

 at all virgin recruitment levels, with maximum 

 values obtained at the lowest recruitment levels 

 capable of sustaining the 1978-86 observed catches 

 (in order to match the sharp decline in CPUE from 

 1980 to 1982). Estimates based on research cruise 

 CPUE data were higher, ranging from 40,000 to 

 180,000 (Table 5, Fig. 2). In both cases, the results 

 were much more sensitive to M than to the LSOg^; 

 parameter of the selectivity function (Tables 4, 5). 



Based on these estimates of the virgin recruitment 

 level, the adult population prior to fishing would 

 have ranged from 287,000 to 452,000 fish (commer- 

 cial CPUE) or 420,000 to 814,000 fish (research 

 CPUE). Assuming 476 km^ of tilefish habitat off 

 South Carolina and Georgia (Low et al. 1983), the 

 estimated density prior to fishing would have been 

 603-950 (commercial CPUE) or 883-1,710 (research 

 CPUE) per km^. We are not aware of other esti- 

 mates of tilefish density prior to fishing. Submer- 

 sible dives were made on the South Carolina tile- 

 fish grounds after the period of (assumed) heavy 

 exploitation; unfortunately, no density estimates are 

 currently available. Comparisons with the exploited 

 MAB stock are of some interest because MAB catch 

 rates in the late 1970s were similar to initial catch 

 rates off South Carolina and Georgia (Low et al. 



0.35 



1978 1979 1980 1981 1982 1983 1984 1985 1986 



2000 



1500 



■^ 1 000 



500 



1978 1979 1980 1981 1982 1983 1984 1985 1985 

 Year 



Figure 1.— Upper panel: estimated fishing mortality rates (F) 

 from 1978 to 1986 at four (arbitrarily selected) levels of virgin 

 recruitment (30,000-120,000 age-6 tilefish), assuming a natural 

 mortality rate (M) of 0.10. Year-to-year changes in F were similar 

 at other levels of M. Lower panel: estimates of fishing effort 

 (hooks fished) based on commercial (1980-81) and research 

 (1982-85) CPUE data. 



1983). Our density estimates were similar to the bur- 

 row density estimates from the MAB. Grimes et al. 

 (1986) reported that biu*row density in Hudson Can- 

 yon ranged from 1,815 per km' in 1980 to 1,132 in 

 1982. Estimates for Veatch Canyon ranged from 

 772 per km^ in 1981 to 1,531 in 1984. Tilefish den- 

 sity in the MAB may be lower than these estimates 

 because not all burrows may be occupied (Able et 

 al. 1982). In addition, burrow density is highly vari- 

 able (Able et al. 1987) and some burrows may be in- 

 habited only during certain seasons, depending on 

 water temperature (Grimes et al. 1986). Neverthe- 

 less, these comparisons suggest that the density 

 estimates we generated from catch data were 

 reasonable. 



181 



