A NOVEL METHODOLOGY FOR THE AUTOMATIC ACQUISITION OF REAL FOREST FIRE DATASETS OVER LONG PERIODS OF TIME

Abstract. Forest areas or green infrastructure have become a fundamental economic and social factor which has made it possible to generate new sources of employment and to maximize the use of basic resources. Nevertheless, good conservation of this type of infrastructure is a challenge. This is due to the problems deriving from, on the one hand, the increase of its physical scope and, on the other, poor management or no management at all. The Spanish region of Galicia is a historic place of natural wealth, especially concerning forest resources, wherein 2/3 of its territory is forest area from where more than half of the Spanish wood supply comes.

This paper seeks to create a mapping of major forest fire disturbances on the Galician territory over extended periods of time. To achieve this, an automated multitemporal detection process based on vegetation indices and unsupervised learning is developed. The objective is to obtain data heterogeneity in terms of vegetation state, land use, and image properties, allowing a better understanding of forest land disturbances and improving their management.