Auth et dl.: DIel variation in vertical distribution of an offshore ichthyoplankton community 



315 



DHau!=I.D,r,/J^r^, 



(1) 



where r, = the depth range (mj of each depth-stratified 

 sample. 



Weighted mean depths iWMDs) of dominant larval taxa 

 were calculated according to the following equation 

 (Pearre, 1973): 



WMD = ^n,d,/'^n^, 



(2) 



where n, = the number of individuals in each depth- 

 stratified sample; and 

 d^ = the mean depth (m) of each sample. 



To facilitate vertical distribution analyses, the wa- 

 ter column was divided into seven depth strata: 0-10, 

 10-20, 20-50, 50-100, 100-150, 150-200. and 200-350 

 m. The weighted mean density of eggs and larvae in all 

 samples collected in each depth stratum per haul were 

 calculated as the strata densities for each haul. A type- 

 II ANOVA model was used to test the null hypothesis 

 that egg and larval densities did not differ between day 

 and night periods or between depth strata, and that 

 there was no interaction among these factors (Dunn 

 and Clark, 1974; Lough and Potter, 1993). ANOVA 

 and a Tukey's multiple range test were applied to the 

 logj,(H + 0.1)-transformed haul and depth-strata densi- 

 ties to test for significant differences between day and 

 night periods and depth strata. Weighted mean (based 

 on density) larval lengths of important species were also 

 calculated for each haul and depth stratum, and were 

 similarly tested for significant differences between day 

 and night periods and depth strata. 



Taxa diversity and evenness for day and night and 

 total diversity and evenness for each depth stratum 

 were analyzed for all identifiable egg (7;=11) and larval 

 (n=20) taxa. The Shannon-Wiener diversity index (H') 

 was used to measure egg and larval diversity, where 

 higher H' values denote greater diversity. Taxa even- 

 ness was assessed by using Pielou's evenness index 

 (J'), which ranges from zero to one, with the maximum 

 J' value indicating that all taxa are represented in 

 the same relative concentrations. Both H' and J' were 

 calculated according to the formulas of Shannon and 

 Weaver (1949) and Krebs (1989). 



On Field et al.'s (1982) recommendation, we performed 

 a hierarchical cluster analyses in conjunction with non- 

 metric multidimensional scaling (MDS) ordinations to 

 identify assemblages by potential taxa, depth-strata, 

 and diel egg and larval concentrations. For analyses 

 of taxa assemblages, only those egg (n = 4) and larval 

 (n=12) taxa found during more than 5% of the sam- 

 pling events were included, whereas all identifiable 

 egg («=10) and larval (?! = 19) taxa were included in the 

 other assemblage analyses. Concentrations for each egg 

 and larval taxon were averaged for each diel tow (eggs: 

 n = 8; larvae: n = 9) and each depth stratum from all 

 tows (eggs: n=22; larvae: 73 = 53), which constituted the 

 sampling units in the respective multivariate matrices. 



Sampling units for which no taxa were found were ex- 

 cluded from the analyses. 



Dendrograms of egg and larval taxa, depth strata, 

 and diel presence were created using hierarchical, 

 group-averaged clustering from Bray-Curtis ranked 

 similarities on standardized, fourth root-transformed 

 egg and larval concentrations (Clarke and Warwick, 

 2001). Dendrograms were cut to produce ecologically 

 interpretable clusters when they were apparent. In or- 

 der to verify our interpretations of the dendrograms, we 

 performed nonmetric MDS ordinations using similarity 

 matrices from the cluster analyses, with 20 random 

 restarts each to minimize stress levels. A two-dimen- 

 sional ordination approach was adopted because stress 

 levels were sufficiently low (<0.08) in all cases and were 

 not appreciably reduced by the addition of a third di- 

 mension, and the results were sufficiently interpretable 

 ecologically in two-dimensional space (Clarke and War- 

 wick, 2001). 



A nonparametric, multivariate procedure (BIO-ENV) 

 was used to analyze the relationship between select 

 environmental variables and egg and larval community 

 structures. The details of the BIO-ENV algorithm and 

 its suitability for use in analyzing the interactions of 

 biological and environmental data are described by 

 Clarke and Gorley (2001) and Clarke and Warwick 

 (2001). Two separate analyses were performed; one with 

 a similarity matrix of depth-stratified samples by egg 

 taxa (24 samples x 9 taxa), and the other with a simi- 

 larity matrix of depth-stratified samples by larval taxa 

 (61 samples X 19 taxa). These matrices were analyzed in 

 association with three environmental variables: mean 

 depth (m), mean temperature (°C), and mean salinity of 

 each depth-stratified sample. Both BIO-ENV analyses 

 were performed by using the Spearman rank correlation 

 method on the normalized Euclidean distance similarity 

 matrices of the log^(;!-i-0.1)-transformed, nonstandard- 

 ized environmental variables by depth-stratified sam- 

 ples (Clarke and Gorley, 2001). All diversity, evenness, 

 cluster, MDS, and BIO-ENV analyses were performed 

 by using PRIMER statistical software (PRIMER, vers. 

 5.2.9, PRIMER-E Ltd, Plymouth, UK). 



Pairwise correlation analyses were also conducted to 

 assess the relationship between concentrations of sev- 

 eral prominent egg iSardinops sagax [Pacific sardine], 

 Icichthys lockingtoni [medusafish], and Chauliodus ma- 

 couni [Pacific viperfish]) and larval (Sebastes spp. [rock- 

 fishes], Stenobrachius leucopsarus [northern lampfish], 

 Tarletonbeania crenularis [blue lanternfish], and Lyop- 

 setta exilis [slender sole] ) taxa as well as total eggs and 

 larvae and the environmental variables salinity and 

 temperature. Mean egg and larval densities, salinity, 

 and temperature per depth-stratified sample were used 

 as variable measures. Before inclusion in the analyses, 

 egg and larval concentrations were log^(n-i-0.1)-trans- 

 formed to normalize the data and homogenize residual 

 variances. Statistical significance was determined at 

 a = 0.05. All ANOVA and correlation analyses were 

 performed using the statistical software JMP (JMP, 

 vers. 5.1., SAS Inst., Inc., Cary, NC). 



