CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Clustering analysis of wildfires: A classification according to land cover types in the Valencian Community Authors:  Laura Serra - University of Girona (UdG) (Spain) [presenting]
Abstract: Typically forest fires are spread reflecting a particular pattern, which can be justified by the features of the specific place of their ignition, either from the elements of the ground that are set on fire or from their geographical location (near roads or urban centres). Within this context, the current study explains how the forest fires that occurred in the Valencian Community during the last years (from 2016 to 2020) are grouped spatially. The same work also attempts to characterise each group (cluster) in terms of location and land cover. We have performed a brief exploratory analysis of environmental covariates related to forest fires. Finite Gaussian mixture models are used to carry out the analysis. More concretely, the R package Mclust, which offers many functions to understand clustering is used. The main outcomes of this research are to offer a set of tools with which to obtain better knowledge about the type of forest fires in each studied spatial zone. Some interesting cluster patterns in certain geographical locations like river basins have also been reported. Besides, they allow measures to be applied to decrease the impacts of forest fires and help the extinction helped by the features of all forest fires grouped using spatial and land cover dimensions.