Data

Data Table

For our study, we used data compiled by Ogden et al (2015) who focused on developing a "single productivity quality rating" for 3 target species within BC: Sockeye Salmon (Oncorhynchus nerka), Pink Salmon (Oncorhynchus gorbuscha), and Chum Salmon (Oncorhynchus keta). Ogden et al. (2015) compiled data throughout BC from 1950-2013 (simplified and reduced to just 3 rows in Table 1). Their dataset includes figures on salmon catch, release, and escapement, which were collected through database research and estimates derived from local fisheries. They also gathered spawning data from 29 B.C Sockeye Salmon (Peterman et al.,1998), 43 Pink Salmon (Pyper et al. 2001), and 67 Chum Salmon (Pyper et al., 2002). This dataset split the salmon productivity metrics for Pink salmon (Oncorhynchus gorbuscha) into 'Pink odd' and 'Pink even' based on the year they returned to spawn. Therefore, this analysis involved 4 'species.'

As we anticipate climate to have an effect on salmon productivity, our climate variables are the predictor variables, and salmon returns and recruits per spawner are the response variables (Table 1). As Minimal Annual Temperature and Maximum Annual Temperature are encompassed in the Mean Annual Temperature, for simplicity, Mean Annual Temperature (MAT) and MAP were the main response variables analyzed (Figures 8,9, 10, and 11)

Table 1: Simplified data table illustrating the Experimental units (ID, Year, Name, Species, Area); the Predictor Variables: "Maximum average annual temperature", "Minimum Average Annual Temperature", "Mean Annual Temperature", and "Mean Annual Precipitation" (MAMax_T, MAminT, MAT, and MAP), and the Response Variables: Salmon Returns and Recruits per spawner (R.S).

Data Visualization

Initial visualization of our data revealed the assumptions of the homogeneity of variances and of normality were not met for our two response variables as they were highly skewed to the left. However, taking the natural logarithm of both predictor variables and using this in subsequent analyses and display allowed for these assumptions to be met (Figures 8 and 9).

Figure 8. Residual plots for each Response Variable (Recruits per Spawner and Returns) as a function of the 2 main Predictor Variables - Mean Annual Temperature (top 2 plots) and Mean Annual Precipitation (bottom 2 plots).

Figure 9. Boxplots displaying Recruits/spawner (left) and Returns (right) for each of the 4 salmon species included in the study by Ogden et al. (2015). Note that 'Pink even' and 'Pink odd' refer to the same species of Pink salmon separated by the year they returned to spawn.

Our main Predictor Variables were visualized in Figures 10 and 11 before any statistical tests were performed. Clearly, there are regional differences in temperature trends and fluctuations, with the North and Central coast experiencing more variable average temperatures and precipitation compared to other regions (Figures 10 and 11). As well, the Southern-inside (non-Fraser) region received far less precipitation compared to other regions over this time period (Figure 11). As we are not particularly interested in the regional differences in salmon productivity in response to climate in this analysis, these regions became blocks for a linear model performed.

Overall, the climate data is well correlated aside from some minor deviations, particularly for the Mean Annual Precipitation, which may have resulted from human error in climate data collection (Figures 10 & 11). In our judgment, these minor deviations should not impact our analysis and therefore, none of the climate data was omitted.

Figure 10. Mean Annual Temperature Variation from 1950 to 2009 grouped by the four general areas studied.

Figure 11. Mean Annual Precipitation variation from 1950 to 2009 grouped by the four general areas studied.