Fig. 1. Clustering dendogram. The end of each line on the left side of the dendogram represents one sampling data; a list of dates associated with each cluster is given in Table 2. The dendogram is scaled both by Wishart's (1969) objective function and by percentage of information remaining. Wishart's objective function is a measure of information loss as clustering proceeds, and is calculated as the sum of the error sum of squares from the centroid of each group to the items in that group.

Fig. 2. A time series of the value of the monthly Multivariate ENSO Index (from Wolter and Timlin, 1998; and http://www.cdc.noaa.gov/ENSO/enso.current.html/indices). The eight El Niño events since 1965 are numbered. During this study, we sampled during event numbers 2, 3, 5 and 8. The grey bars indicate the sampling periods.

Fig. 3. Frequency distribution of sampling dates by cluster group pooled by month. Data are proportion of samples in a given month that were classified into the indicated cluster group. Shading indicates samples collected in May–August 1997.

Fig. 4. Ordination of copepod communities for the 12-year data set. Each symbol is a sampling date, and the shape of each symbol identifies its corresponding cluster number, as in Table 2. Symbols adjacent to each other indicate the greatest similarity in community composition between two samples; the greatest separation between symbols indicates the greatest dissimilarity.

Fig. 5. Relationship between (arc sine transformed) proportion of sampling dates that grouped with Cluster 2A (Summer/Low Biomass cluster) and the PDO summed over the period January–September.

Fig. 6. Ordination of copepod communities with abundance of the dominant cold-water neritic species overlayed. Symbol indicates cluster group; size of symbol indicates relative abundance of the taxa.

Fig. 7. Ordination of copepod communities with abundance of the dominant warm-water species overlayed. Symbol indicates cluster group; size of symbol indicates relative abundance of the taxa.

Table 1. Average abundance (geometric mean) of the dominant species within each of the seven clusters with the upper and lower 95% confidence levels below. Arithmetic data were log (Y+0.1) transformed for averages and back-transformed for presentation; back-transformation sometimes resulted in negative lower confidence limits

Table 2. Chronological order of sample dates coded by cluster number

Breaks in sampling >3 months are indicated by dashed lines. Symbols for each cluster are the same as those in Fig. 4, Fig. 6 and Fig. 7.

Table 3. For each year of sampling: proportion of sampling dates in each of the four summer clusters (1A, 1B, 2A, 2B); the value of the Bakun upwelling index at 45°N (Bakun index) summed for May–September of each year; the value of the PDO summed over January–September of each year; the period of time between which the transition from a winter community to summer community occurred (indicated by the date the last winter community was observed/the date the first summer community was observed); and the dates between which the community transitioned from a summer to a winter community (last summer observation/1st winter observation)

Table 4. Indicator values (IV) for the most common copepod species collected at the station NH 05, arranged by cluster group; and Spearman's correlation coefficient (r) for correlation of the species abundance with the two axes of the NMDS ordination

The higher the IV, the better indicator that species is of that particular cluster group. Only those species whose IV was significant for at least one cluster are listed. For 204 degrees of freedom, a coefficient greater than 0.14 is significant at the p=0.05 level.