Wildfires in Python Using Pandas, Numpy, and SciKit
Two Seaborn Heatmaps of Average Fire Size by Latitude-Longitude I wanted to try heatmaps for this and hoped there would be enough data to trace out the borders of the contiguous US, fortunately this turned out to be the case. Because heatmaps use aggregation we see more contrast in the plot with bigger bins(LEFT). This is because bigger bins can catch more occurrences. The image on the RIGHT has bins one-tenth the size of those on the LEFT, which affords it a more detailed outline. However, because very large fires in little bins are not being pulled down by enough smaller fires via the numpy.mean aggregator function the scaling throws off the potential for better contrast. My solution to this issue is below.
To generate this heatmap I applied np.log to the values of the pandas pivot_table used above, but only after adding 1 to the fire size column before pivot table aggregation; this allowed us to avoid negative numbers from applying a log to fire sizes less than 1 acre.
Fires by Bin Suggests Occurrences are Skewed toward Drought(negative PDSI) and Higher Temperatures
Linear Regression of Fire Size on Years using the Seaborn Package Suggests Fires are Getting Bigger