RENR690 - Data

Data

Data Table Processing

Data for this project was collected in 2023 and 2024 in Weyerhaeuser Grande Prairie, Saddle Hills FMA (GP in tables and graphs), and 2024 in Weyerhaeuser Pembina Timberlands FMA (PT in tables and graphs). Field data was input in Excel and separated by research block. All data was compiled into a single sheet, with added columns for year and location, and simplified for analysis in R. Table 1 shows a few rows of one raw data table. Table 2 shows the simplified data tables for import and analysis in R Studio. For height and percent cover of each vegetation group, data needed to be standardized by replacing blanks with zeros to represent the absence of vegetation.

Key response variables include seedling height (TREE_HT), fresh and dry weights (FRESH_WT, DRY_WT), and percent cover and height for each vegetation group (e.g. DECID_COV, DECID_HT). Moisture content and Comeau’s Competition Index variables were not included in Excel, as these were calculated in R for the entire dataset.

Table 1. Raw Excel file received from Apical Forestry in 2024, showing data collected in the Saddle Hills FMA. Pembina Timberlands data follows the same table format. Column "..." indicates species, height, and cover data for vegetation groups (shrubs, forbs, and grasses), in the same format as shown for deciduous data.

Table 2. Simplified and filtered data table for analysis. This shows the first rows of the data table, which includes data from Saddle Hills FMA in 2023 and 2024 and Pembina Timberlands FMA in 2024. Column "..." indicates species, height, and cover data for vegetation groups (shrubs, forbs, and grasses), in the same format as shown for deciduous.

Data Exploration

Comeau's Competition Index

CCI was calculated in R using the formula outlined in the methods section. Competition index data was fairly zero-inflated (Fig 1) due to some plots having no vegetation in certain groups, as well as the bulk of grazed plots having lower CCI. Additionally, CCI data contained many outliers (Fig 3), though these points are less likely due to measuring errors than to some plots simply having much larger deciduous trees or other vegetation which increased overall CCI. Though visualizing CCI by cutblock (Fig 4) is visually overwhelming, plotting it this way allowed for better visualization of data spread and easier identification of outliers than in Fig 3. CCI was also calculated for each vegetation group and visualized with a stacked barchart (Fig 2) to observe how grazing influences specific vegetation types in grazed and ungrazed plots.

Fig 1. Competition Index histogram illustrates data distribution and zero-inflation.

Fig 2. Mean CCI visualized with a stacked bar chart to show differences between grazed (G) and control/ungrazed (U) plots at both Grande Prairie/Saddle Hills (GP) and Pembina Timberlands (PT) sites in 2023 and 2024.

Fig 3. Competion Index boxplots show distribution of data across the study in grazed (G) and control/ungrazed (U) plots. Many outliers are present in the CCI dataset.

Fig 4. Competition Index boxplots by cutblock and treatment. Separating by block better visualized the location outliers present in the dataset, but led to messy labels.

Vegetation Biomass

Vegetation biomass was analyzed in R with both fresh sample weights and the subset of dried sample weights. The relationship between fresh and dry weights was visualized to check for trend and consistency (Fig 5). Moisture content was calculated in R for samples with fresh and dry weights and visualized in a histogram (Fig 6). This immediately indicated a few errors in the dataset from either weighing vegetation or entering data, since negative moisture content should not be possible. We know these outliers are the result of an error, not a true representation of values measured, so the weights for these data points were removed. Given the large sample size, approximately 700 fresh and 230 dry samples, removing these values should not impact the statistical power of the analysis.

Fig 5. Scatterplot showing the relationship between fresh and dry weight of clipped biomass samples. This relationship follows a positive trend, with higher fresh weights correlated to higher dry weight.

Fig 6. Histogram of moisture content of clipped samples. Weights for samples with negative moisture content percent will be removed, as this indicates a clear error in either measurement or data entry.

Exploration of vegetation biomass (both fresh and dry weights) with boxplots indicated that there were some outliers (Fig 7). Though there are outliers in both, these large outliers are due to some samples being in very dense plots, as with the large competition index values, and represent a true portion of the data. Using robust ANOVA techniques for statistical analysis methods should allow for outliers and non-normal data distributions without significantly impacting results.

Fig 7. Boxplots showing the spread of clipped biomass data for both dry and fresh weights in grazed (G) and control/ungrazed (U) plots at Grande Prairie/Saddle Hills (GP) and Pembina Timberlands (PT) sites in 2023 and 2024.

Direct influence of grazing on tree health

As part of data collection, it was indicated by Y (yes) or N (no) if a seedling was trampled to some degree by grazing. Anything from 10% lean to 100% dead and down was observed. The percentages of seedlings trampled to any degree from each dataset are shown in Table 3. No browsing was observed on any of the seedlings in this study, so only damage from trampling was recorded.

Table 3. Percent of seedlings trampled. Grande Prairie/Saddle Hills is represented with GP and Pembina Timberlands with PT.

Fig 8. ANOVA residuals for linear mixed effects model evaluating treatment as a fixed effect and block as a random effect on Comeau Competition Index.

Fig 9. ANOVA Q-Q residual plot for linear mixed effects model evaluating treatment as a fixed effect and block as a random effect on Comeau Competition Index.

Fig 10. ANOVA residuals for linear mixed effects model evaluating treatment as a fixed effect and block as a random effect on biomass of competing vegetation.

Fig 11. ANOVA Q-Q residual plot for linear mixed effects model evaluating treatment as a fixed effect and block as a random effect on biomass of competing vegetation.

ANOVA residual plots for the Comeau Competition Index (Fig. 8 and Fig. 9) and biomass (Fig. 10 and Fig. 11) with linear mixed-effects models, with treatment as a fixed effect and block as a random effect to account for within-block variability in vegetation. While the residuals show some right skew and the presence of outliers, the ANOVA approach is generally robust to some deviation from normality.

Page updated

Report abuse