Fig. 2: Discovering pyromes.
Self-organizing maps are useful for summarizing multidimensional fire data and for determining potential groups of similar characteristics. These data are reduced to a two-dimensional grid, and the samples are organized according to their Euclidean distance. Observations sharing similar characteristics are easily visualized in a topographic map (a) where warmer colors represent widely separated samples, and cooler colors depict closely related values. Using image processing algorithms (see Methods), we detect relevant potential pyromes/clusters (red circles). The number of observations belonging to each section of the map can be presented in a matrix known as a hit-map (b). As an example, we can easily observe the group of cells without fire activity as a large dark blue region (top) and white valley (bottom), representing a relevant percentage of the observations.
