Data

Dataset

For this study, transect data was taken from ArcGIS Pro and put into a data table (Table 1). The total number of observations was 2335.

Design variables: ID, FIRE#, and Transect.

Independent variables: Fuel Type, Location, and Time of year (TOY). They are observed and categorical.

Dependent variable: Burn Code, which is categorical.

Fuel Type is represented in the CWFIS FBP Categories:

C=Coniferous

D=Decidious

M=Mixed-wood

O=Grass

S=Logging Slash

For further analysis, all fuel types were combined into three categories:

Seasonal- Deciduous, Grass, Mixed-wood <50% conifer content

Conifer- Conifers, Slash, and Mixed-wood >50% conifer content

Non-Fuel- Vegetated Non-Fuel, Non-Fuel, Water

Raw data was transformed into count data of Fuel Types split by TOY, Burn Code, and Fire# (Table 2). Similar tables were also made for Fuel Category Counts and for Fuel Category counts by Location instead of Burn Code.

Table 1. A partial data table showing the raw data collected. Includes fire number, transect number, burn status, fuel type, position on the transect, and time of year for the fire.

Table 2. Count data for each fuel type at different times of year and burn status.

Data Exploration

Full Observation Dataset

To start, the full dataset was explored for eventual testing of hypothesis 1. Data was long-tailed with zero values, so a log1p translation was performed. Clear trends toward higher burnt observations were seen in the C-2 and O-1 fuel types (Figure 4). Many outliers were present in the dataset, and many fuel types, such as S-1 and S-2, had very few observations (Figure 4).

Figure 4. A boxplot that represents each fuel type's count data per fire at different burn codes. The x-axis is the mean counts on a log scale. Orange dots represent mean values, and black dots represent outliers.

Fuel types were then combined into the three categories "Conifer," "Seasonal," and "Non-Fuel," as described at the top of this page, and then re-plotted. There are a few notable trends. There are more conifer observations in the burnt category (Figure 5). There are slightly fewer mean burnt observations in Non-Fuel. The most notable trend for this study is that there is slightly more burnt in the Seasonal boxplot (Figure 5). The outliers were largely reduced through this fuel type combination into categories.

Figure 5. A boxplot representing counts of Fuel Categories per fire. The x-axis is the mean counts on a log scale. The orange dots represent means.

Observations Split by Time of Year

Next, the observations were faceted by time of year to test hypothesis 2. A log1p translation was again applied. One notable trend is the higher total observations in early fires driven by the large SWF063 wildfire (Figure 6).

Figure 6. A boxplot that represents each fuel type's count data per fire at different burn codes and times of year. The x-axis is the mean counts on a log scale. Orange dots represent mean values and black dots outliers.

Then, just like the combined data, the data was replotted with fuel categories. Seasonal fuels are more involved in the burnt observations of early fires and more in the unburnt observations of late-season fires (Figure 7). Conifers have more burnt observations than unburnt at both times of the year, and non-fuel has fewer burnt observations compared to unburnt (Figure 7). Once again, outliers were largely reduced.

Figure 7. A boxplot that represents each fuel category's count data per fire at different burn codes and times of year. The x-axis is the mean counts on a log scale. Orange dots represent mean values.

Location on the Transect

The locations on the transect were plotted to look for notable differences and decide what to utilize for the statistical tests. Overall trends seem mostly similar across the different transect locations (Figure 8). For example, the 100m External in the early fires has a similar distribution as the 500m External in the early fires despite them being 400m apart (Figure 8). There are some visible differences, but all tests were done using the combined dataset, as differences were mainly insignificant.

Figure 8. A comparison of observations of fuel categories at different locations along the transect. The y-axis is mean counts on a log scale. Internal observations were located inside the fire; external were located outside the fire. The distances represent how far away the point was from the fire's edge. Orange dots represent means; black dots represent outliers.

Page updated

Report abuse