In last decades, numerous efforts have been carried out by the scientific community to assess the multifunctional role of forests in supporting the forest decision-making processes. Recently developed in Italy, the Forest Landscape Management Planning seems to be the most suitable tool to maintain the sustainability balance between forests and local populations. This research identifies and maps different forest ecosystem functions in a mountain Natura2000 Network site by performing a spatial estimation of forest inventory parameters, and implementing a multi-criteria and multi-level approach. Final results on mapping of forest functions show acceptable values of classification accuracy. © 2014 University of Newcastle upon Tyne.

Forest ecosystems for their important multifunctional value, need a complex and increasing amount of descriptive information to support their management. Ecological and environmental related attributes have became nowadays important as traditional ones, such as wood growing stock and basal area. The correct application of Sustainable Forest Management criteria is boosted by spatial contiguous knowledge of such attributes. For such a reason in the last years a huge number of scientific experiences in the forest area have been concentrated to study the relationship between data acquired in the field and remotely sensed multispectral images. Models based on such relationships can be used to estimate and map forest attributes acquired in the field on the basis of a statistical sampling design. can be sucould not take in consideration spatially structured data. In last years many researches have focused on possible relationships between field data and remote sensed informations derived from multispectral imagery. Modeling these relationships allows to extend inventory data to not explored surfaces. In this paper were discussed results on spatializing forest biometrical attributes, tree heterogeneity and dimensional heterogeneity assessed during an inventory of Mountain Community "Alto Molise" (IS) throw Spot 5 and Lansat TM 7 imagery. For this purpose a multilinear regression and a k-Nearest Neighbor classifier were used.