The spatial and temporal variability in smolt emigration times is shown in Figure 2. There was no indication of non-linear effects in any of the models investigated. Following step down model selection, the final model for both S25 and S75 included Site as a random effect and Elevation and Year as fixed effects (Figure 3). This indicated that there was no significant spatial variability in the timing of smolt migration to sea (zero metres elevation), that could be predicted from simple GIS metrics. There was however, considerable inter-site variability that was unrelated to geographic location. There was also substantial inter-annual variability at individual sites and significant long-term trends in migration time with migration advancing by ca. 7 and 14 days for S25 and S75 respectively over the last 47 years. To identify a "sensitive window" for salmon smolt migration, predictions of S25 and S75 were made for zero metres Elevation and the Year 2014. The lower (S25) and upper (S75) 95% confidence intervals were then used to bound the sensitive period, considering uncertainty in the predictions. This results in a period that runs from day 103 to 145.
Figure 2: Temporal variability in smolt emigration times as indicated by counts of juvenile Atlantic salmon caught on each day of the year (DoY). Points show individual counts obtained from a site visit, lines provide a kernel density estimate of the distribution of emigration times. Individual years are indicated by separate colours.
Figure 3: Model predictions of the effects of Elevation (metres) (A,C) and Year (B,D) on the timing of S25 (A,B) and S75 (C,D). Predictions of the effects of elevation are made for year 2014. Predictions of the effects of Year are made for an elevation of zero metres. Points show individual observations, black lines indicate the estimate of an effect, purple shading indicates uncertainty in the fixed effects, while blue shading indicates uncertainty in the fixed effect, combined with variance from the random effect (i.e. includes inter-site variation).
There was substantial temporal and spatial variability in the size of emigrating salmon (Figure 4, Table 1). Across all sites and years, the proportion of emigrants >135mm ranged from 0-92%. Within sites, the proportion exceeding the threshold varied by as much as 92% between years. The mean percentage of emigrants exceeding the threshold at individual sites varied between 6 and 29% for sites with ≥ 10 sampling years. Unfortunately it was not possible to model the spatial and temporal variability in smolt sizes given the small number of sites with high precision (1mm) size data and the time available for analysis.
Table 1: The percentage of fish exceeding the 135mm size threshold required for tagging with a VEMCO V7 acoustic transmitter.
|Site||Maximum||Minimum||Mean||Number of Years|
Figure 4: Kernel density estimates of the distribution of smolt sizes. Colours indicate individual years. Vertical red line indicates the threshold size required for tagging with a VEMCO V7 acoustic tag.