Disclaimer: This project website has been created for a university course. It contains partial and unpublished data that should not be used for any other purposes.
Sampling effort
In 2014, 18 patches and 135 individual sites were sampled over the 3 sampling periods with Arctic Grayling found in 19 sites (14% of total sites sampled). A summary of the total number of sites sampled and repeat sampled and the number of sites with Arctic Grayling are summarized below in Table 4 for each sampling period. Figure 8 demonstrates the distribution of sites sampled throughout the watershed and the areas with concentrations of Arctic Grayling presence. Figure 9 identifies the continuous temperature data loggers that were deployed in 2014 which will play an important role in understanding thermal regimes in the watershed.
In 2014, 18 patches and 135 individual sites were sampled over the 3 sampling periods with Arctic Grayling found in 19 sites (14% of total sites sampled). A summary of the total number of sites sampled and repeat sampled and the number of sites with Arctic Grayling are summarized below in Table 4 for each sampling period. Figure 8 demonstrates the distribution of sites sampled throughout the watershed and the areas with concentrations of Arctic Grayling presence. Figure 9 identifies the continuous temperature data loggers that were deployed in 2014 which will play an important role in understanding thermal regimes in the watershed.
Arctic Grayling
Table 5 below summarizes the total number of Arctic Grayling detected (captured and observed) during each sampling period. Three sites were included in the table to provide examples of the difference abundances of fish detected at different sites and how those number changed during the sampling periods. The majority of Arctic Grayling were captured in Patches 128 and 67. Figure 10 below shows the relative lengthfrequency distributions for Arctic Grayling caught and measured over the 3 sampling periods in 2014. The fork length values reported in these figures represent a range of lengths measured between those values. The size distributions are similar between sampling periods 1 and 2, with the majority of fish caught being youngofyear (<~80mm). Larger Arctic Grayling (juvenile and adult) were captured more frequently during sampling period 3 which could suggest that they utilize these stream habitats for longer periods of time prior to migration to overwintering areas. But it may also be a factor of several new sites that were sampled during that period, as identified above in Table 4.
Figure 10. Relative frequencies of Arctic Grayling fork lengths measured during each Sampling Period. Sampling Period 1 (June 27July 4, 2014), n=78. Sampling Period 2 (July 27August 4, 2014), n=131. Sampling Period 3 (August 2531, 2014), n=24). Fork length values represent a range of lengths measured between the values shown.
Site Characteristics
A total of 15 habitat variables were measured at each site in sampling periods 1 and 3 and 13 variables (no water temperatures or velocity) were collected in sampling period 2. Table 6 below provides mean values of select habitat variables from all 3 sampling periods. Sites with Arctic Grayling had average water temperatures of >9oC, average elevation <1100m and an average slope of ≤2.7%. In contrast Arctic Grayling were absent from sites with average water temperatures of <6oC, elevations of >1250m and slopes of ≥6%. This information is valuable as it starts to reveal potential habitat thresholds that may drive Arctic Grayling distribution in the watershed and can help to evaluate the biophysical criteria being used to identify suitable habitat patches using distributional monitoring methods.
Figures 11 and 12 below graphically show the data collected on some of the environmental variables that were used as criteria to identify suitable habitat patches for sampling in the Little Nahannni watershed. Figure 11 plots site elevation and water temperature values for sites collected during sampling period 1. The results show that the average water temperature was higher and the average elevation was lower for sites with Arctic Grayling than for all sites sampled during sampling period 1. Figure 12 shows the differences measured between stream width and slope for all 3 sampling periods. Arctic grayling sites had a similar average water depth, but a lower average slope than all sample sites. These preliminary results are informative as potential patterns and differences emerge between sites available broadly in the watershed and sites with grayling occupancy.


Stream habitat types were evaluated (out of 100%) to characterize the 100m length of each sampling site. Percent composition of substrate type, stream habitat type and riparian vegetation were recorded. Figures 13 and 14 below show the average composition of the four categories of substrate and stream habitats that were used to describe the conditions of the sampling sites. The graphs show that there is variation between sites with and without Arctic Grayling. Arctic Grayling sites were approximately 80% comprised of silty sand and gravels, as well as 80% run and pool habitat, compared to approximately 50% at sites without Arctic Grayling. This type of information demonstrates a pattern and assists in describing potential species preferences for certain habitat characteristics. Photos 7 and 8 above show examples of the differences between stream habitats available in the watershed that were sampled in 2014.
Statistical Analysis
Select variables for the analysis
Table 7 below shows the results of the individual logistic regressions run on each of the 15 habitat variable to evaluate the initial strength of relationships. Almost half of the variables showed significant pvalues (<0.05) but water temperature reported the strongest AIC value.
Test different variable combinations to build a model
Logistic regressions were developed for 2 separate data sets, one for sampling period 1 which included water temperature and velocity as variables and the other for all sampling periods but without water temperature and velocity as available predictor variables. The logistic model for Arctic Grayling presence from sampling period 1 that included water temperature and velocity (Model 1) produced the best fit for predicting occurrence, with water temperature having the most significant effect (p<0.01) in both the regression output and the result of the ANOVA analysis.
Table 8 below shows results from 5 representative logic models that were run. Model 2 had very similar results to Model 1 using elevation, slope and water temperature as factors. Model 3 includes slope velocity and stream width, which produced the best fit for a model without water temperature included. For the variables analyzed from all 3 sampling periods, Model 4 which included elevation, slope, depth and run produced the best fit for predicting Arctic Grayling occupancy. Model 5 shows the output using elevation, slope and stream order which are the 3 main habitat criteria where values can be accessed using GIS data. These were the variables used to define suitable habitat in the distributional monitoring approach.
Logistic regressions were developed for 2 separate data sets, one for sampling period 1 which included water temperature and velocity as variables and the other for all sampling periods but without water temperature and velocity as available predictor variables. The logistic model for Arctic Grayling presence from sampling period 1 that included water temperature and velocity (Model 1) produced the best fit for predicting occurrence, with water temperature having the most significant effect (p<0.01) in both the regression output and the result of the ANOVA analysis.
Table 8 below shows results from 5 representative logic models that were run. Model 2 had very similar results to Model 1 using elevation, slope and water temperature as factors. Model 3 includes slope velocity and stream width, which produced the best fit for a model without water temperature included. For the variables analyzed from all 3 sampling periods, Model 4 which included elevation, slope, depth and run produced the best fit for predicting Arctic Grayling occupancy. Model 5 shows the output using elevation, slope and stream order which are the 3 main habitat criteria where values can be accessed using GIS data. These were the variables used to define suitable habitat in the distributional monitoring approach.
Use model outputs to evaluate predictive strength of variables
Table 9 below shows the results of the binomial ANOVA conducted on the model outputs for the 2 best fit models for sampling period 1 and the best fit model for all 3 sampling periods. Water temperature came out as the strongest habitat variable having a significant effect on the probability of Arctic Grayling presence. Elevation had the most significant effect for the best fit logistic model that did not include water temperature as a factor.
Table 9 below shows the results of the binomial ANOVA conducted on the model outputs for the 2 best fit models for sampling period 1 and the best fit model for all 3 sampling periods. Water temperature came out as the strongest habitat variable having a significant effect on the probability of Arctic Grayling presence. Elevation had the most significant effect for the best fit logistic model that did not include water temperature as a factor.
Use model probabilities and plot against key variables
The resulting predicted probabilities from Models 1 and 4 were plotted against the strongest predictor variables, water temperature and elevation respectively. Figures 15 and 16 below show the relationships between between temperature, elevation and the predicted occupancy of Arctic Grayling from each model. The results of Model 1 (Figure 15) show that the probability of Arctic Grayling occurrence has a strong response to water temperature at approximately 8 degrees Celsius. Model 4 (Figure 16) also shows a response to elevation at approximately 1000m.
Fit a relationship
The results of the multiple logistic regression models show that water temperature has the strongest effect on the probability of Arctic Grayling occurrence. A response curve was plotted to fit a relationship between water temperature and predicted occupancy (Figure 17). Response curves allow for the establishment of thresholds for temperature and probability of Arctic Grayling occurrence.
The results of the multiple logistic regression models show that water temperature has the strongest effect on the probability of Arctic Grayling occurrence. A response curve was plotted to fit a relationship between water temperature and predicted occupancy (Figure 17). Response curves allow for the establishment of thresholds for temperature and probability of Arctic Grayling occurrence.
The information collected in 2014 provides direct observation of Arctic grayling distribution and suitable habitats within a northern watershed. It provides an improved understanding of the relationships between biophysical parameters and the presence of Arctic Grayling in this watershed, which will allow for the establishment and identification of ecological thresholds for the prediction of Arctic Grayling occurrence.
Next Steps
Further research will be conducted in the Little Nahanni River watershed in 2015 to:
1. Refine the distributional monitoring approach and criteria for identifying suitable habitat for Arctic Grayling in the north;
2. Study juvenile Arctic Grayling to identify habitat and ecological thresholds;
3. Examine thermal regimes across the subwatersheds in the Little Nahanni watershed;
4. Develop occupancybased models that can be used to predict Arctic Grayling occurrence in unsurveyed areas of the north.