EDAFOLOGÍA, Vol. 12 (1), pp. 65-73, 2005
MODELIZACIÓN DEL HÁBITAT POTENCIAL DE FORMACIONES FORESTALES EN LA PROVINCIA DE HUELVA
M. ANAYA-ROMERO (1), R. PINO (2), A. JORDÁN (1), L. MARTÍNEZ-ZAVALA (1), N. BELLINFANTE (1)
(1)Departamento de Cristalografía, Mineralogía y Química Agrícola (Universidad de Se- villa). Autor para correspondencia, ajordan@us.es
(2)Departamento de Estadística e Investigación Operativa (Universidad de Sevilla)
Three methods for modelling the potential distribution of forest formations in protected areas in southern Spain are discussed in this paper: logistic multiple regression, artificial neural net and de- cision tree. For every model, the potential habitat of every forest type has been designed using environ- mental information (soil, lithology, geomorphology and climate). The results were compared with the current area corresponding to every forest type, to determine the relative accuracy of every method. The multiple logistic regression seems to offer the best results.
Key words: Artificial neural net; decision tree; logistic regression; relationship vegetation - environ- ment; Mediterranean forest.