Finally, production specialization for each municipality
was obtained using data from the agrarian
census. Four variables (areas of croplands, pastures,
forest and others) were reduced to two factors using
a principal component analysis. These data were
selected because from our perspective they provide
a better synthesis of the spatial effect of economic
activities than population or the number of fiscal
licences in different productive sectors. Factor 1
synthesized 51 per cent of total variance, highlighting
the transition from highlands that are dominated
by grazing and forest activities (positive values of
factorial scores) to lowlands devoted to agriculture
accessibility and its influence on planning for the
location of different land-use types, urbanization and
economical specialization within a global market
(Klijn and Vos 2000). Finally, two different dependent
variables were used: land-cover diversity for the
original 220 classes (LCD220) and a reclassified
map comprising just 24 classes (LCD24) (Table 1).
Thus, two different models were developed to assess
the effect of the number of land-cover classes on
diversity.