Vaughan and Prager: Decline in abundance of Pagrus pagnis off the soutfneaslern United States 353 



Vaughan et aL (1992), we adjusted the data to avoid dupli- 

 cation from inclusion of headboat landings in charter-boat 

 landings reported by MRFSS, 1979-85. 



Mean recreational landings by mode were calculated for 

 1979-97 and used for 1972-78, a period of minimal ex- 

 ploitation before MRFSS data were collected. Because rec- 

 reational landings are a small fraction of the total, this 

 procedure, while approximate, was better than assuming 

 landings of zero in that period. 



Recreational length-frequency distributions from MRFSS, 

 1979-97, were weighted by catch (A-t-Bl) in number by 

 mode, wave (2-month interval), and state. The minimal 

 catch associated with shore-based fishing was pooled with 

 the private-boat mode. Headboat length-frequency distri- 

 butions from MRFSS ( 1979-85) were not used in develop- 

 ment of catch matrices for the recreational fishery because 

 more complete data were available from the headboat 

 sampling program. 



Headboat fishery Estimates of headboat catch in num- 

 bers and weight and fishing effort in angler-days (Table 

 1) were obtained from the NMFS (Beaufort, NC) head- 

 boat sampling program (Huntsman, 1976; Huntsman et 

 al.. 1978). Zero-intercept hnear regression analyses on NC 

 and SC data were used to estimate landings from GA to 

 northeast FL and landings from southeast FL. and fishing 

 effort back to 1972. Catch per effort (CPE) in numbers was 

 calculated by dividing annual catch in numbers by annual 

 effort. 



Length samples from headboats were not available from 

 GA or FL before 1976. nor from southeast FL before 1981. 

 We used mean length-frequency distributions for 1976-80 

 to fill in for GA and FL, 1972-75, and length-frequency dis- 



tributions from GA to northeast FL to fill in for southeast 

 FL. 1976-80. Annual headboat length-frequency distribu- 

 tions from the headboat sampling program. 1972-97. were 

 weighted by catch in numbers caught during that season 

 (January-May. June-August, September-December), and 

 geographic area (NC. SC. GA to northeast FL, southeast 

 FL) to obtain catch in length by season and area. 



Development of catch-at-age matrices 



Estimates of catch at age m numbers were made by using 

 an approach similar to that of Vaughan et al. (1992). Total 

 catch in numbers (n. scalar) for each combination of time, 

 gear, and area was multiplied by an age-length key (A, 

 matrix of dimension a x 6); the product was multiplied by 

 the corresponding length-frequency distribution (L, vector 

 of length b) to obtain catch in numbers at age (N, vector of 

 length a ): 



^ = n -A-L. 



(1) 



Here, a is the number of ages ( for this study, ages to 8* ) 

 and b is the number of length intervals (for this study, 15 

 intervals of 25 mm each from 200 mm to 550+ mm). 



Age-length keys Separate age-length keys were devel- 

 oped from fishery-dependent and fishery-independent 

 data. Fishery-dependent keys for 1972-74 were those of 

 Manooch and Huntsman ( 1977); for 1986. those of Vaughan 

 et al. (1992); for 1996-97, those of Potts and Manooch 

 (2002). Keys for 1975-85 and 1987-95 were approximated 

 by linear interpolation from available keys (method of 

 Vaughan et al., 1992). 



