Since landscape attributes show different patterns at different spatial extents, it is fundamental to identify how the relation between landscape structure and plant species diversity at local scale varies with scale. Then, it is fundamental to assess the appropriate extent at which landscape factors affect plant species richness at the local scale. To investigate this relation, data on plant species richness of forest communities at plot scale were extracted from a large data set and landscape metrics were calculated around the same plots for a range of extents (250-3000. m). Then, multiple regression models and variance partitioning techniques were applied to assess the amount of variance explained by the landscape metrics on plant species richness for a range of extents. In general, we found that increasing extent of the surrounding landscape analyzed, improved the strength of relationship between the landscape metrics and the properties of plant communities at plot scale. The medium-large extent was most informative as it combined a decent total variance explained with high variance explained by the pure fractions of complexity, fragmentation and disturbance and the minimum of collinearity. In conclusion, we found that it is possible and beneficial to identify a specific extent, where the redundancy in the predictor variables is minimized and the explanatory power of the pure fractions (or single groups) maximized, when examining landscape structure effects on local plant species richness.
