The spatial and temporal distribution of wildfires in mainland Portugal can help identifying (dis)similar characteristics between regions. We used specific fire metrics, from historical fire data between 2000 and 2021, to identify groups of municipalities based on their pyro similarities: (a) cumulative percentage of total burned area, (b) cumulative percentage of burned area in the summer months, (c) mean annual number of fires and (d) GINI index applied for burned area over time. We apply clustering methods in two steps., The first step identifies groups of municipalities (pyro regions) of mainland Portugal (n=277), based on four algorithms: k-means, k-medoids, ward.D2 and Affinity Propagation, using Geographic Information System tools (ArcGIS Pro) and specific packages in R software. This step divided the country in 4 clusters, and despite the different methods, 77% of the municipalities were always placed in the same group. Afterwards, we joined the results of the 4 methods, based on majority of placement, to obtain a common classification for the pyro regions. In the second step, we chose two clusters: A (n=37) located in central Portugal, with a high percentage of cumulative burned area over the years, but with low mean number of fire occurrences and B (n=53), which covers the municipalities in the north and is characterised by a high mean number of fires dispersed over the years (low GINI index). For each cluster, we applied the Affinity Propagation algorithm to verify differences within each cluster. Both clusters were divided in two new groups: in A, the main differences were in mean number of fire occurrences and the cumulative percentage of burned area in the summer, and in B, the principal differences were in the cumulative percentage of total burned area. The next step will be analyse the explanatory variables that condition these fire patterns.

Bruno Barbosa
Centre of Geographical Studies, Institute of Geography and Spatial Planning – University of Lisbon

 
ID Abstract: 381