The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material. http://http://dspace.unitus.it:80 2013-05-22T01:03:55Z 2013-05-22T01:03:55Z Corona, Piermaria Dettori, Sandro Filigheddu, Maria Rosaria Maetzke, Federico Scotti, Roberto http://hdl.handle.net/2067/2116 2011-06-10T00:30:46Z 2004-12-31T23:00:00Z

Title: Site quality evaluation by classification tree: an application to cork quality in Sardinia Authors: Corona, Piermaria; Dettori, Sandro; Filigheddu, Maria Rosaria; Maetzke, Federico; Scotti, Roberto Abstract: Cork harvesting and stopper production represent a major forest industry in Sardinia (Italy). The target of the present investigation was to evaluate the ‘‘classification tree’’ as a tool to discover possible relationships between microsite characteristics and cork quality. Seven main cork oak (Quercus suber) producing areas have been identified in Sardinia, for a total of more than 122,000 ha. Sixty-three sample trees, distributed among different geographical locations and microsite conditions, were selected. A soil profile near each sample tree was described, soil samples were collected and analysed. After debarking, cork quality of each sample tree was graded by an independent panel of experts. Microsites where trees had more than 50% of the extracted cork graded in the best quality class, according to the official quality standard in Italy, were labelled as prime microsites, the others as nonprime microsites. Relationships between a binary dummy variable (0 for nonprime microsites, 1 for prime microsites) and site factors were investigated using classification tree analysis to select the relevant variables and to define the classification scheme. Prime quality microsites for cork production proved to be characterised by elevation, soil phosphorus content and sandiness. Results have been compared with those of the more conventional parametric approach by logistic regression. The work demonstrates the advantages of the classification tree method. The model may be appropriate for classifications at landscape and stand mapping levels, where it is possible to sample a number of microsites and to evaluate distributional characteristics of model output, while its precision is only indicative when estimating the prime quality of single microsites. Description: L'articolo è disponibile sul sito dell'editore www.springerlink.com

2004-12-31T23:00:00Z