Rilling and Houde: Variability in growth and mortality of Anchoa mitchilli 



557 



the head of the Bay (39°25'N) to near the Bay mouth 

 (ST^OS'N). Data were analyzed and compared in three 

 regions: upper bay— transects 1-5 (39°25'N-38°45'N); 

 mid bay— transects 6-10 (38°45'N-37°55'N); and 

 lower bay— transects 11-15 (37°55'N-37°05'N). 



At each station, ichthyoplankton was collected in 

 one net of an opening-closing, 60-cm bongo sampler 

 with 280-(im meshes and preserved in ethanol. Two 

 tows, of 2-min duration, were made at each station. 

 The first tow was from within 1 m of bottom to the 

 pycnocline, and the second was made from the 

 pycnocline (or middepth when no pycnocline was 

 present) to the surface. Sampling protocols and meth- 

 ods to estimate densities and abundances of organ- 

 isms are detailed by Rilling and Houde, manuscript 

 in review); only brief descriptions are given here. 



Immediately before each tow, a conductivity-tem- 

 perature-depth (CTD) cast was made from within 

 1.0 m of bottom to within 1.0 m of the surface to pro- 

 vide depth profiles of temperature, salinity, and dis- 

 solved oxygen. To make results comparable to those 

 of Dorsey et al. (1996) and MacGregor and Houde 

 (1996), temperature, salinity, and dissolved oxygen 

 were examined at 3-m depth. Zooplankton from ei- 

 ther three or four designated depths was sampled in 

 10-L Niskin bottles and collected on 35-/;m mesh. 

 Gelatinous zooplankters from each ichthyoplankton 

 tow were counted and their biovolumes recorded. 



In the laboratory, anchovy lar\'ae were measured 

 to the nearest 0.1 mm standard length (SL). Lengths 

 were corrected for shrinkage during collection and 

 preservation (Theilacker, 1980; Leak, 1986). Small 

 larvae of bay anchovy may be extruded through net 

 meshes (Leak and Houde. 1987). Therefore, we ap- 

 plied a regression method to adjust abundances of 

 <5.5 mm SL larvae collected in the 280-|im net 

 meshes. The regression, which adjusted abundances 

 of larvae upward by factors of 2.3 (at 2.0 mm), 1.9 

 (at 3.0 mm), 1.6 (at 4.0 mm), and 1.2 (at 5.0 mm lar- 

 vae), was derived from comparisons of length-specific 

 abundances in paired tows of 53-)im and 280-|.im mesh 

 bongo nets made in Chesapeake Bay under condi- 

 tions similar to those during this survey (MacGregor, 

 1994; Rilling and Houde, manuscript in review). 



Otolith analysis 



Otolith microstructure was analyzed to estimate age, 

 growth, and mortality. In the present study, sagittal 

 otoliths from 509 larvae were examined. Otoliths 

 from representative samples of larvae from each re- 

 gion of the Bay were examined for each cruise. Each 

 larva in the otolith analysis was measured to the 

 nearest 0.1 mm SL. Otoliths from larvae of 2.0 to 

 25.0 mm SL were mounted in "Epon" under a cover 



slip, and heated for 24 h at 60°C to harden the epoxy 

 (Secor et al., 1991 ). Otolith increments were counted 

 on two separate occasions under a compound light 

 microscope at 600 to lOOOx magnification by one 

 reader (Rilling). The mean of the two increment 

 counts plus two days was the estimated age. 



Growth rates 



Growth in length of larvae (mm/d) was estimated 

 from the slopes of the linear regressions of shrink- 

 age-adjusted lengths (SL) on ages from daily otolith- 

 increment analysis: 



L, = a + gt. 



where L, = standard length (mm) at age t (d); 



t = age (d) = otolith increment count plus 



two days; 

 g = growth rate (mm/d); and 

 a = y-intercept, the estimated length (mm) 



SL at hatch. 



Gompertz growth models (Bolz and Burns, 1996) also 

 were fitted to the data for each region and cruise. 

 The fits were no better than those for the linear model 

 and were not considered further in our analysis. 



Larval lengths were converted to dry weights (g) 

 from a weight-length relationship: 



W=0.1550xL3 5307_ 



where W = dry weight (g); and 

 L = mm SL. 



Rates of growth in weight then were estimated 

 from an exponential model, fitted by regressing log^- 

 transformed dry weights on age: 



where W^ = dry weight (g) at age Md); 



Wq = dry weight (g) at hatch (the y-intercept 



of the log-linear regression); and 

 G = weight-specific growth coefficient (/d). 



Coefficients in growth-model regressions were com- 

 pared among regions and between cruises (months) 

 m analysis of covariance ( ANCOVA). The ANCOVAs 

 tested for differences in slopes (growth rates) andy- 

 intercepts in the growth equations. When significant 

 differences were found, a multiple range test (Stu- 

 dent-Newman-Keuls) was applied to determine 

 which of the growth rates differed significantly. The 



