18 



Fishery Bulletin 101(1) 



growth rates have involved comparisons in the vicinity of 

 the Mississippi River discharge plume (Govoni et al., 1985; 

 DeVries et al., 1990; Lang et al., 1994; Allman and Grimes, 

 1998). The observed variability in larval Atlantic bumper 

 and vermilion snapper growth rates reported in our study 

 was not associated with conspicuous hydrographic fea- 

 tures (e.g. hydrographic convergence zones) and suggests 

 the existence of less-recognizable regions where condi- 

 tions for growth vary. 



Cruise estimates of mortality were determined to as- 

 certain a realistic level about which the effects of small 

 variations in growth rates on the cumulative survival of 

 larvae could be assessed. In order to do this, data from 

 all stations sampled during a cruise were pooled. This 

 provided the most reliable general estimate of mortality 

 for each cruise despite likely site-specific differences in 

 mortality rates that are extremely difficult to measure. 

 Such pooling of data is not unusual; in fact Morse (1989) 

 suggested that samples should be summed over the larval 

 production cycle. Essig and Cole (1986) estimated mortal- 

 ity rates of larval alewives iAlosa pseudoharengus) by us- 

 ing both converted length-frequency distributions, as we 

 did, and actual age-frequency distributions. They found no 

 statistical difference between the two methods. Pepin and 

 Miller ( 1993 ), however, warned that because variability in 

 observed length-at-age increases with larval age (Cham- 

 bers et al., 1988), analyses that use size in older fish to rep- 

 resent age may yield biased estimates of mortality rates. 

 Yet, Pepin and Miller (1993) observed that their mortality 

 rates, which were estimated by using size as a proxy for 

 age, were consistent with mortality rates reported from 

 other environments and species. Ideally, all fish would be 

 aged, but for our study this was not possible because of the 

 large sample sizes, multiple cruises, and the labor-inten- 

 sive nature of otolith preparation for age determination. 



Atlantic bumper lai-vae were extremely abundant in = 

 32,241 for six cruises), and cruise estimates of age-fre- 

 quency distributions showed consistent, well-defined de- 

 scending limbs. Estimates of mortality coefficients (Z) for 

 Atlantic bumper larvae were similar for September cruis- 

 es conducted in the same year. For example, in 1990 the 

 two cruise estimates of Z were 0.37 and 0.30, in 1991 the 

 two Z estimates were 0.20 and 0.28, and in 1993 estimates 

 of Z were 0.30 and 0.32. These mortality rates are similar 

 to estimates reported by Leffier and Shaw (1992) during 

 four September cruises in the same area during 1986-87 

 (Z=0. 17-0.35) and by Sanchez-Ramirez and Flores-Coto 

 (1998) in the southern Gulf (0.15-0.30). In addition, stan- 

 dard errors of the mortality estimates from our study were 

 low, ranging from 0.02 to 0.05. 



Cruise estimates of mortality rates for vermilion snap- 

 per were determined during four cruises when larvae 

 were relatively abundant («=2581). The descending limbs 

 of three of the size-frequency distributions uniformly 

 spanned all seven size classes, but during one cruise the 

 middle size class was most abundant and the descending 

 limb of this size-frequency distribution was restricted to 

 four size classes. However, mortality rates were quite simi- 

 lar during all cruises (Z=0.19 to 0.30) and each had a low 

 standard error {SE=0.02 to 0.05). 



Collections of Atlantic bumper and vermilion snapper 

 larvae were taken when water temperatures ranged from 

 25° to 30°C, and the mortality coefficients estimated from 

 these collections were similar to those reported for other 

 species under similar temperature regimes. Houde (1989) 

 summarized vital rates of six species of larval fish as 

 reported in seven studies where the mid-points of water 

 temperatures at the time of collection ranged from 26° to 

 28°C. Most of these studies generated a range of mortality 

 estimates, and the mid-points of the ranges reported in six 

 of these studies varied from 0.21 to 0.38, values that are 

 consistent with the mortality estimates (0.19 to 0.39) that 

 we observed. 



Our primary reason for estimating mortality rates was 

 to ascertain a realistic level about which small variations 

 could be assessed for potential effects on the cumulative 

 survival of larvae, particularly in conjunction with vari- 

 ability in larval growth rates. Our method assumes a con- 

 stant birth rate, or recruitment rate into the population, 

 and assumes that fish leave the population only through 

 death. There is clearly some expected variability in the de- 

 gree to which these assumptions were met; however, based 

 on the similarity of mortality estimates, not only between 

 cruises but also to previously published estimates, it is 

 concluded that our mortality estimates are biologically 

 meaningful. 



The well-accepted fisheries paradigm holds that changes 

 in year-class strength are determined by variability in mor- 

 tality during early life stages (Sissenwine, 1984; Houde, 

 1987; Bailey and Houde, 1989; Gushing and Horwood, 

 1994). Despite extensive efforts to understand the causes 

 of recruitment variability, significant questions remain 

 because the operant factors are likely to be interrelated 

 parts of the ecosystem dynamics that comprise a multidi- 

 mensional system (Ellersten et al., 1995). For example, it 

 is not the mortality or growth rate alone that determines 

 survival during the early life-stages, but the ratio MIG, 

 the stage-specific mortality rate (Pepin, 1991). Examin- 

 ing previously published information, Houde (1989) found 

 an exponential increase in predicted lai-val-stage dura- 

 tion with decreasing water temperature for 26 species of 

 larval fishes and surmised that when temperature is low, 

 small changes in growth rates can induce large changes 

 in larval-stage duration that may significantly affect the 

 recruitment process. 



To determine the potential effects that variability in vi- 

 tal rates might have on the cumulative survival of larvae, 

 hypothetical numbers of newly hatched Atlantic bumper 

 were projected to a size of 6 mm under the influence of 

 the growth and mortality rates we observed in the study 

 area (Table 1). According to these vital rates, and a hypo- 

 thetical initial cohort size of 1 x lO*" individuals, 124,930 

 larvae survive to a length of 6 mm under the scenario of 

 relatively fast growth and low mortality (G=0.61 mm/d; 

 Z=0.20). If the growth rate is slowed (G=0.45 mm/d) and it 

 takes approximately three days longer to reach a length of 

 6 mm, the number of larvae that sui-vive to this length is 

 reduced by 44*7^. If the slower growing lai-vae are exposed 

 to the higher mortality rate (Z=0.37), cumulative survival 

 of larvae decreases by an order of magnitude, and only 



