In summary, the present study proposed a methodological
approach that is suitable for use at
regional scales and has been used to detect a robust
signal supporting the hypothesis proposed by previous
authors (García-Ruiz and Lasanta 1990) from a
descriptive perspective without any numerical verification:
traditional management in European mountains,
especially in Mediterranean mountains, has
resulted in an increased diversity of land cover,
although new homogenization trends have emerged.
Results obtained herein also show that land-cover
diversity is the final result of synergies among environmental
and anthropogenic factors, but that the
effect of human management on landscape trends
(diversification–homogenization) differs in relation
to geographical context. The availability of the data
used herein for many European countries supports
the applicability of this approach to other landscapes
in order to gain a better regional-scale insight into
diversification–homogenization trends where detailed
information on land-cover changes based on aerial
photographs or remote-sensing images is unavailable.
It is important to emphasize that the results contain
a degree of uncertainty. I have assessed only two
environmental variables (climatic variability and
The final model explains less of the total variance
than the climatic factor when regressed separately
against land-cover diversity. This result is due to the
use of penalized splines to reduce the complexity
of the curves and control the estimated degrees of
freedom. To obtain statistical significance, curve
complexity was reduced and the estimated degrees
environmental model developed with 24 land-cover
classes, while the other variables show weak
correlations and low levels of statistical significance
(Table 4). Positive values of productive specialization
factor in areas traditionally devoted to forestry and
grazing are related to positive residuals, while
negative values for areas devoted to cropping in
lowland areas are related to negative residuals
(Figures 4 and 5). Only productive specialization
showed statistical significance when regressed
against residuals of the second natural sub-model
with 220 land-cover classes (Table 4).