Given the nested approach of our model, there are several different outputs as a result of applying our model to data from WBNP. We will discuss the results of each submodel (Ignition Potential, Spread Potential, and Fuel Potential) in addition to the final wildfire risk overlay. We will also discuss our evaluation of our model using the 2016 fire data, and limitations of our research approach.
Figure 6 shows our ignition potential output. Zones considered at a high risk of ignition were concentrated in the elevated areas of the park, and are shown in red. Since most of the park os at a moderate elevation, it is considered at a low-medium risk of wildfire ignitions due to the potential for lightning strikes. Roads, trails, and buildings can be seen in this layer as regions of moderate risk of ignition due to human influence. Overall, there is not much heterogeneity in the data due to the low spatial variance of these criteria.
Figure 6. Ignition potential overlay
Figure 7 shows the spread potential output. This layer again shows a higher spread potential near elevated regions of the park due to the slope factor. Crown closure, however, is less focused near the southern area of the park. This factor corresponds with our fine fuel accumulation layer, as areas with low crown-closure represent areas where fires recently burned.
Figure 7. Spread potential overlay
Figure 8 shows the fuel potential output. Zones of high fuel potential were concentrated in areas of lower latitude. This is because the southern part of WBNP is drier and hotter compared to the northern parts. Areas of high elevation were generally cooler and wetter, and therefore were less at risk for fuel potential. The highest fuel potential was in the bottom left corner of the park due to high maximum monthly temperatures during the growing season and low precipitation.
Figure 8. Fuel potential overlay
Spatial Distribution of Wildfire
As shown in Figure 9 below, there is a distinct pattern in wildfire risk zones throughout the park. Northern areas of the park have a lower overall risk for wildfire than southern-lying areas. This can be attributed to the influence of climate on wildfire potential. Moreover, southern regions of the park experience greater temperature maximums, and lower amounts of precipitation than northern regions. We also see a concentration of wildfire risk near the roads and buildings located in the centre of the park. This is the result of the human influence buffer, where a greater potential for ignition is present due to proximity to human activity. There is a baseline risk for wildfire throughout the park due to the potential for a lightning ignition, which although has been predicted by elevation, can occur virtually anywhere within the park.
Figure 9. Map of wildfire risk zones within Wood Buffalo National Park
The greatest area of WBNP falls underneath 'Medium' risk (Figure 10), which accounts for 46% of the park area (Table 8). High risk areas account for 20% of park area, whilst low risk areas account for 26% of park area, and 8% falls under 'No Risk', due to the presence of several large waterbodies within the park (Table 8).
Figure 10. Chart of park area within each wildfire risk zone
Table 8. Areas of wildfire risk zones
The 2016 wildfire season was used to evaluate the effectiveness of our model. The 2016 wildfire season was smaller than typical years, accumulating only 72.7 km2 of area burned and with only 4 fires that were above 1km2. Since wildfire seasons can have a large variance of area burned from year to year, wildfire season extent and the lightning distribution underpin the unpredictable nature of wildfire occurrence. By overlaying the 2016 wildfire extent over our final output map, we were able to account for which zones of risk intensity the area burned fell into.
Figure 11 shows the percentages of area burned in each wildfire risk zone; no risk, low, medium and high risk. 76% of area burned in 2016 fell under what our model categorized as medium risk. 12% of the areas burned fell into low and high zones each, with a negligible area falling under no risk (Figure 11).
Figure 11. Modelled wildfire zone coverage of areas burned in 2016
The largest fire of the 2016 season was 58km2. This fire was also in the same area of a fire in 1999. We had used previous wildfire extent as a factor to determine fine fuel accumulation and did quantify this area as lower risk then compared to others. It is unlikely that a wildfire occurred in this area twice in the past 20 years due to fine fuel accumulation (Fauria and Johnson, 2008). This fire also occurred near the Peace River, located alongside a major roadway.
Given this, we can say our model predicted with some degree of effectiveness the likelihood for fires within WBNP. Given the large extent of the park, and the relatively large size of wildfire zones, it is difficult to pinpoint exact fire areas, however using the general spatial distribution mentioned above and evident in the figures, we can see a pattern of predictability amongst wildfires.
Through our evaluation of our results and model, we identified several limitations to our approach: