274 



Fishery Bulletin 102(2) 



West Bay 

 ME 



West Bay 

 SNB 



Xmas Bay 

 ME 



Xmas Bay 

 SAV 



Xmas Bay 

 SNB 



Figure 10 



Brown shrimp (F. aztecus) observed log density and standard deviation for bottom 

 types in Christmas Bay and West Bay. ME = marsh edge: SAV = submerged aquatic 

 vegetation; SNB = shallow nonvegetated bottom. 



Galveston Bay. Our data suggest that brown shrimp select 

 SAV over ME when these habitats co-occur (Christmas 

 Bay) and select ME when grassbeds are absent (West Bay) 

 ( Fig. 10 ). Habitat submergence time may explain high SAV 

 use in Christmas Bay (Rozas and Minello, 1998). Subtidal 

 grassbeds may provide more continuous refuge and food 

 supply at both low and high tides than the marsh surface, 

 which can be accessed only during high tides. Additionally, 

 brown shrimp were significantly smaller in SAV (,v = 17 

 mm) than in ME (5=25 mm) (t-test, P<0.001>. which may 

 imply ontogenetic changes in habitat or trophic require- 

 ments (Conrow et al., 1990; Thomas et al., 1990; Rozas and 

 Minello, 1999). Differences in the use of bottom types may 

 correspond with the population's size distribution at the 

 time of sampling. Additional research is needed to reveal 



ontogenetic habitat shifts and relationships among shal- 

 low estuarine bottom types (Mclvor and Rozas, 1996). 



Assessment of the model performance was based on 

 FWS HSI theory where there is a positive relationship 

 between HSI value and the carrying capacity of the avail- 

 able habitat. In the present study, the relationship equates 

 high brown shrimp densities with optimal habitat condi- 

 tions that promote high carrying capacity. Therefore, low 

 densities would reflect a low suitability or a low capacity 

 to support the population. Comparisons of predicted den- 

 sity with that of observed values from Galveston Bay, and 

 other Texas bays (Figs. 7 and 8) agree with FWS theory 

 by exhibiting a strong relationship between density and 

 suitable habitat as determined from the model. Model per- 

 formance and transferability were examined by applying 



