by Madison Downer-Bartholomew
Forest fires are a frequent and natural disturbance within the boreal forest. The diversity of the boreal is largely the result of these fires that are varied in frequency, intensity, size, shape, and season. Fire behavior in the boreal varies from intense crown fires to slow moving ground fires, depending on factors such as the reason for ignition, wind and species composition and forest moisture. Each fire differs in its intensity of burn that overall influences the scar on the landscape that remains visible. Fire mapping tends to neglect analyzing the complexity within the boundary of where forest is burned and unburned. Yet, increased knowledge of fire boundary structure can provide foresters and land managers the ability to replicate conditions like natural fires to sustain the diversity of the boreal forest.
For my EUC Undergraduate Research Award (EUCURA) project with Professor Tarmo Remmel, I examined and quantified boreal fire boundaries using ArcGIS Pro. I examined 11 different fire footprints in 5 different spatial resolutions (4, 8, 16, 32, 64). My task in this project was to use tools in ArcGIS Pro to identify 1,2,3,4, and 5 rings of pixels inside and outside the perimeter of the fire footprint and then analyze the frequency of land cover type in each ring using land cover data from before and after the area was burnt. This process tells us what is within the boundary of where forest is burned and unburned in terms of land cover.
With this project I was tasked with figuring out how to get pre and post burn land cover data from 11 different fires in 5 different spatial resolutions. I very quickly learnt that I would need to automate the process in a format where I can do multiple tasks repetitively in the most efficient way possible.
Therefore, in ArcGIS Pro, I used ModelBuilder to create a workflow that strings together multiple geoprocessing tools. With the model I created, I was able to take a binary map of each of the 11 fire footprints in each spatial resolution, expand and shrink the footprint by 1 to 5 pixels and use the raster calculator tool to subtract the expanded/shrunk layer from the binary footprint resulting in ring layers of the expanded/shrunk pixels. Next, I performed a conditional evaluation with the ring layers and both pre-burn and post-burn land cover data to identify land cover type within each expanded/shrunk ring. Lastly, the data on the frequency of each land cover type within each ring was exported to excel. The excel files were then imported into R where the land cover data of all the fires in each spatial resolution can be analyzed and used to create histograms.
In the next phase of this research, perpendicular transects need to be used around the perimeter of the fire to extract land cover values along each transect to determine how a gradient exists within fire boundaries. This research, once published, could fill a gap and improve knowledge on boundary structures and can improve forest managers' ability to mimic natural disturbance conditions.