Diversity was calcvilated as Nl = (exp) , where IT = 2 (p, In p), where p, = n, / n, where a, = abundance of 

 species i, and n = total abundance (Ludwig and Reynolds 1988). Evenness was calculated as El = ln(Nl) / 

 ln(NO), where NO = total number of species (Ludwig and Reynolds 1988). Nl is the number of abundant 

 species, and El is the familiar J . 



Statistical Analysis 



A two-way analysis of variance (AN OVA) was used to determine turbidity and velocity differences among flow 

 treatments and sampling dates. More complex ANOVA was used to test for differences in macrobenthic 

 response among flow, disturbance frequency, and sampling date treatments. All ANOVA models were 

 calculated using SAS GLM procedures (SAS 1985). 



A three-way, incomplete factorial, randomized block design was tised to test for community differences in over 

 all sampling dates (Fig. 1). Main effects (i.e., treatments) were flow, disturbance frequency, and sample 

 collection date. The randomized block was flow plot. Blocks are replicates that do not have interactions with 

 main effects. The design is incomplete factorial because distiorbance frequency treatments were not started or 

 ended on the same dates. Because some date cells are missing in the biweekly, monthly, and bimonthly 

 frequency levels, the frequency*date interaction and flow* frequency* date interaction do not exist in the 

 ANOVA model (Table 1). In this model, date is more like a block (controlling nuisance variation) than a main 

 effect. 



Full rank, three-way ANOVA models do exist for subsets of the data set (Fig. 3). The experiment was 

 replicated two different times: as starting dates and as ending dates. So, macrobenthic data was analyzed with a 

 three-way randomized block design twice: Once for trays deployed on two dates: 23 April and 4 June 1997, and 

 once for trays collected on two dates: 20 June and 1 August 1997. These analyses are referred to as "initiating" 

 and "ending" sampling dates respectively (Table 1). 



Interpretation of results from complex ANOVA designs is often obscured by significant interactions, because 

 the main effects tests are invalid. To simply interpretation of the present study, two-way, incomplete, 

 randomized block models were calculated by a treatment level. Examination of the simple main effects allows 

 testing and interpretation of the first main effect at all levels of the second main effect. The simple main effects 

 models were calculated by disturbance frequency for flow and sampling date treatments with plot as a 

 randomized block. Tukey multiple comparison tests were used to determine differences among levels of flow 

 or distiirbance frequency cell means. The implementation of Tukey uses the harmonic mean of cell sample 

 sizes when sample sizes are unequal (SAS 1985). \'ariance components analysis was used to estimate the 

 percent of variation attributable to each main and interaction effect in all ANO\''A models. 



Principal component analysis (PCA) was used (SAS 1985) to determine treatment effects on species 

 composition. The covariance matrix of log transformed species abundance, standardized to a normal 

 distribution was used for PCA. Using the covariance, ittstead of the correlation matrix, eliminates problems 

 encountered where many rare species with zero counts exist. The multivariate PCA method is a species 

 dependent analysis of community structure, unlike the species independent analysis of diversity indices. 



RESULTS 



Hydrography 



Hydrographic conditions at station C varied during the course of the experiment (Fig. 4). The small (about 

 5 %o) drop in salinity from 23 April - 22 May 1997 is due to local rainfall. Average salinity peaked on 20 June 

 1997 at 25.6 %o, but dropped to %o on 2 July. The fresh conditions correspond with a flood event that began 

 on 22 June. The flood resulted primarily from rain in the watershed northwest of the delta. Dissolved oxj'gen 

 was highest 23 April and 7 May, but lowest 4 June. On 4 June, dissolved oxygen data collected prior to 

 9:30 a.m. were 2.4 mg 1"' at weir 1, and 2.88 mg 1 ' at net 1, indicating hypoxic conditions probably occurred 

 during the previous night. Temperature generally increased with onset of summer. 



A significant (p = 0.0020) flow*date interaction was present because there were greater differences among flow 

 rates on 17 July than on 07 May (Fig. 5). On both dates, weir structures had increased flow velodries, and the 



F-4 ^ Effects of Temporality. Disturbance Frequency and Water Flow 

 on an Upper Estuarine Macroinjauna Community 



