Spatial analysis and investigation of fire events occurrences in the Valencian Community, Spain

Dissertation submitted in partial fulfillment of the requirements for the Degree of Master of Science in Geospatial Technologies.Fires have been affecting on average half a million hectares of forests, shrubland and crops every year. During the second half of the 20th century with socio economic development people abandoned unproductive land and overpopulated more fertile areas and cities. Landscapes started to be covered with natural vegetation or new plantation, often with highly flammable flora (conifers, olive trees, fruit trees, etc.) causing more frequent fire occurrences. Spain follows this trend with high incidence of fires in recent years, underling and emphasising the importance of understanding the causes and spatial distribution of these phenomena. In order to evaluate main characteristics of fires and the distribution of ignitions, 3292 fire events detected in Valencian Community during the period 2000 – 2006 are analyzed. GIS and spatial point process modelling approach are used to quantitatively study the fire effects in relation to variables such as cause, burnt area, proximity to urban areas and roads, population density, land cover and geographic elements. Point pattern analysis was performed using the library SPATSTAT with the statistical package R to determine the spatial intensity of fire ignition distribution and how covariates affect the pattern. Results showed that humans are the leading cause of fires in this region, but as well that the Valencian Community has significant number of lightning caused fires. Fire location are spatially clustered and high fire occurrences was found within areas 1 – 2 km from urban areas and roads, highly populated areas, in agricultural and shrubland cover, lower elevations and tender slopes. Results suggested that there is no simple fire regime for Valencian Community. The Akaike information criterion method is used to select the best inhomogeneous Poisson process model from a set, to best fit the data. The fitted model was diagnosed using simulation envelopes of K function and residual analysis. The model turned out to be inadequate because the fitted intensity function failed to capture the dependence of intensity on covariates. Regardless that a satisfactory model was not found, the study emphasizes the importance of understanding where fires occur and how they interact with socio-economic and environmental factors