A total of 274 plots of 200 m2 were sampled in which only woody plants more than 5 cm dbh (diameter at breast, 1.3 m height) were registered, hereafter called adults. Woody plants 1-5 cm dbh, hereafter called saplings, were sampled in a nested 50 m2 subplot. All individuals were taxonomically identified at species level or the closest possible taxon, and their height and the dbh of each stem were measured (see Dupuy et al. 2011 for more details). We calculated various vegetation variables but only ecologically relevant and statistically nonredundant variables were used in the analysis: stand age (AGE), total species richness (SPPT), species richness of adults (SPPa), species richness of saplings (SPPs), abundance of adult individuals (ABUNa), abundance of saplings (ABUNs), total basal area (ABT), total basal area of adults (ABa), total basal area of saplings (ABs), mean height of adults (Ha), total density (TD), mean height of saplings (Hs), density of adults (Da), density of saplings (Ds), and Simpson (SIMP) and Shannon (SHAN) diversity indices.
11.2.2 Landscape Variables
We used a classified map derived from a SPOT5 satellite imagery acquired on January 2005 to calculate the landscape metrics for the 23 landscape units, using the software Fragstats (McGarigal et al. 2002). The classified map had an overall accuracy of 75.6 % and Cohen’s k statistic of 0.7 (Herrnndez-Stefanoni et al. 2011).
The landscape metrics considered for the analysis are based on patch-type indices and were selected considering variables that quantify different aspects of landscape configuration, that are ecologically relevant, and are commonly used in the landscape ecology literature for the taxonomic group of interest (Mazerolle and Villard 1999). Based on these criteria, six indices were selected in this study: percentage of land of each patch type (PLAND), patch density (PD), edge density (ED), mean area weighted shape index (SHAPE), mean area weighted Euclidean nearest neighbor distance (ENN), and total edge contrast index (TECI). (For details of how metrics were calculated, see Hernandez-Stefanoni et al. 2011).
