Applying Binary Structured Additive Regression (STAR) for Predicting Wildfire in Galicia, Spain
Zitierfähiger Link (URL): http://resolver.sub.uni-goettingen.de/purl?gs-1/13649
Studies on causes and dynamics of wildfires make an important contribution to environmental. In the north of Spain, Galicia is one of the areas in which wildfires are the main cause of forest destruction. The main aim of this work is to model geographical and environmental effects on the risk of wildfires in Galicia using flexible regression techniques based on Structured Additive Regression (STAR) models. This methodology represents a new contribution to the classical logistic Generalized Linear Models (GLM) and Generalized Additive Models (GAM), commonly used in this environmental context. Their advantage lies on the flexibility of including spatial and temporal covariates, jointly with the other continuous covariates information. Moreover, these models generate maps of both structured and the unstructured effects, and they plotted separately. Working at spatial scales with a voxel resolution level of 1Km x 1Km per day, with the possibility of mapping the predictions in a color range, the binary STAR model represents an important tool for planning and management for the prevention of wildfires. Also, this statistical tool can accelerate the progress of fire behavior models that can be very useful for developing plans of prevention and firefighting.