Manipulation of egg distribution patterns produced clear differences in the spatial distribution of under-yearling salmon juveniles. We detected a significant interaction between section and nest treatment (patchy vs. dispersed) on density of under-yearling salmon (F = 2·36, d.f. = 6, 98, P = 0·036, Fig. 2). Specifically, going upstream from the –50 section, there was a more or less continuous, gradual decrease in densities in the dispersed treatments. As a result, there was a significant spatial autocorrelation in normalized densities in this treatment (density in each section against density in the section above, n = 6, r = 0·83, P = 0·042). In contrast, in the patchy nest treatments, there was a decrease in density at the first section upstream of the lowermost nest (Fig. 2), and no spatial autocorrelation (n = 6, r = 0·40, P = 0·434). Furthermore, post-hocanovas demonstrated a significantly lower relative density in the two- than in the 10-nest treatment at the 50 m section (Fig. 2). The same was true even when comparing absolute densities in this section (mean densities 0·19 vs. 0·42 ind. m−2, pairwise t-test, t = 3·19, d.f. = 7, P = 0·015), but not in the other sections (all P-values > 0·1). However, these analyses of absolute densities may be influenced to some extent by annual variation in survival rates, which could not be controlled for with the present design, and should therefore be treated more cautiously. The deficiency of juveniles just above the lowermost nest in the two-nest treatment suggests successful dispersal from nests to occur predominantly in the downstream direction. This is also indicated by the gradual increase in density when going downstream in the 10-nest treatment, suggesting that juveniles originating from an increasing number of nests are present. The two different treatments also resulted in significantly different frequency distributions of normalized under-yearling densities (two-sample Kolmogorov–Smirnov test, n = 112, Z = 1·42, P = 0·036). The 10-nest treatment resulted in a distribution of densities that closely approximated a normal distribution (Kolmogorov–Smirnov test for deviation from normal distribution, n = 56, Z = 0·52, P = 0·947, Fig. 3a). In contrast, the two-nest treatment had a high frequency of low-density sections, a pronounced tail towards high densities and, as a result, differed significantly from a normal distribution (Kolmogorov–Smirnov, n = 56, Z = 1·94, P = 0·001, Fig. 3b).
To test whether the distribution of eggs also influenced patterns of net movement among sections over the period from the first to the second sampling, we plotted the normalized density of under-yearlings for the different sections in 2005 against the normalized density of 1-year-olds the following year. In the 10-nest treatment, there was a highly significant positive correlation between these two measures (Fig. 4a). Thus, in this treatment, a high density of under-yearlings resulted in a corresponding high density of 1-year-olds the following year. In contrast, within the two-nest treatment, no significant correlation was observed (Fig. 4b).
Figure 4. Relationship between density of under-yearling salmon and the corresponding density of 1-year-old fish the following year among stream sections for (a) the 10-nest [1 year = 0·00 (0·14 SE) + 0·45 (0·16 SE) under-yearling, n = 35, r2 = 0·20, P = 0·007] and (b) the two-nest treatment [1 year = 0·00 (0·20 SE) – 0·30 (0·22 SE) under-yearling, n = 21, r2 = 0·09, P = 0·192].
Download figure to PowerPoint
To test directly whether the divergence in spatial distribution of under-yearling densities influenced the relative spatial distribution of body sizes, we first calculated the difference in normalized density between the two treatments for each section and stream, and similarly the difference in normalized body size. These were closely negatively correlated (Fig. 5). Thus, sections that had high relative densities in the two-nest treatment compared to those in the 10-nest treatment also had a relatively small juvenile body size, and vice versa.