FISHERY BULLETIN: VOL. 84, NO. 1 



tent across the course of this study, so that com- 

 parisons should be unaffected. In addition, two loca- 

 tions on the marsh were sampled with a 10 x 1 m, 

 6 mm mesh, seine, on an irregular basis. An effort 

 was made to seine every month but it was often not 

 possible, as the sites were difficult to seine at ex- 

 treme high or low tides. 



Fishes from each trawl were placed in washtubs 

 of water to minimize mortality and then identified, 

 measured to the nearest millimeter (standard 

 length), and returned to the water as quickly as 

 possible If more than 100 fish of any one size class 

 of a species were captured, only the first 100 were 

 measured; the rest were counted. Early in the study, 

 samples of all fishes were weighed (wet weight, in 

 gram), and a length/weight relationship developed 

 for each species. This was later used to estimate the 

 biomass of fish in each trawl. The shrimps Crangon 

 franciscorum and Palaemon macrodactylus in each 

 trawl were also counted. For the oppossum shrimp, 

 Neomysis mercedis, an index of abundance was used, 

 based on a l-to-5 scale, where "1" represented <3 

 individuals; "2", 3-50 shrimp; "3", 50-200, "4", 

 200-500, and "5", >500. The index was necessary 

 because most N. mercedis probably passed through 

 the net due to their small size (3-5 mm). Neverthe- 

 less, they were present seasonally in most hauls, at 

 times in enormous numbers. 



At each location, salinity and temperature were 

 taken with a YSI S-C-T meter and transparency was 

 measured with a Secchi disk. Tidal height was deter- 

 mined from a tide tabla An index of monthly fresh- 

 water outflow from the combined Sacramento and 

 San Joaquin Rivers at Chipps Island was obtained 

 from the California Department of Water Resources 

 (unpubl. data). 



For analysis, all the data were summarized by site 

 and month. A Spearman rank correlation analysis 

 using data ranked by month (N = 52) was used for 

 the initial analysis because many of the variables did 

 not conform to a normal distribution. Because no 

 single transformation could be applied to all the 

 variables, nonparametric statistics were used as the 

 most conservative method. We used 13 variables for 

 the analysis (Table 2). In addition, rank abundance 

 (by numbers) by month for the following species 

 categories was used: 1) total striped bass, 2) year- 

 ling and older striped bass, 3) young-of-year striped 

 bass, 4) total splittail, 5) yearling and older split- 

 tail, 6) young-of-year splittail, 7) total tule perch, 

 8) tule perch adults, 9) tule perch young-of- 

 year, 10) total prickly sculpin, 1 1) yearling and older 

 prickly sculpin, 12) prickly sculpin young-of- 

 year, 13) carp, 14) longfin smelt, 15) delta 



smelt, 16) staghorn sculpin, 17) starry 

 flounder, 18) threadfin shad, 19) Sacramento 

 sucker, 20) yellowfin goby, and 21) threespine 

 stickleback. Because only minor differences were 

 found among the correlations associated with adult 

 and juvenile striped bass, tule perch, splittail, and 

 prickly sculpin, only the results for the totals for 

 these species will be presented. 



Analyses were also run using the data from each 

 trawl separately. Species were analyzed using both 

 numbers and grams. Because these data were all of 

 species abundances, a log-normal transformation 

 was used to normalize them. The results were similar 

 in most respects to the analyses using ranks so are 

 not presented here However, because we were uncer- 

 tain as to the validity of using ranked data for prin- 

 cipal components analysis (PCA), we based our 

 discussion on cautious inspection of the correlation 

 matrix as generated. A principal components 

 analysis was run using the correlation matrix (Dix- 

 on and Brown 1977) of 1„ numbers of fish per trawl 

 (N = 1,238), to produce groups of species that 

 presumably were responding to the environment in 

 the same general ways. 



Table 2. — Environmental variables used in the correlation analyses. 



RESULTS 



Environmental Variables 



Salinity and temperature were negatively corre- 

 lated with river outflows (Table 3, Fig. 2). Salinity 

 had a strong (P < 0.01) positive correlation only with 

 Secchi depth. River outflows generally peaked in 

 February, March, or April, as the result of run-off 

 from melting snow in the Sierra Nevada. Lowest 



108 



