Question: Can we improve the knowledge of urban vegetation
using data from ongoing floristic and management projects
with a data mining approach? We have two questions: 1. How
strong is the relationship between land cover pattern and the
species composition of vegetation? 2. What is the relationship
between land cover pattern and species richness?
Location: Trieste, northeastern Italy.
Methods: Using land cover maps and GIS we characterized
the cells of a floristic project grid by percentage cover of land
cover types. We applied Canonical Correlation Analysis to
test the correlation between floristic composition of the cells
and land cover. We classified the cells by clustering methods,
based on land cover description. With these clusters, we analysed
the variation of species composition of urban vegetation
along a gradient of urban density. We used Jaccardʼs similarity
index to compare floristic composition of the clusters with the
floristic composition of the homogeneous cells with respect to
the land cover types. To answer question 2, we calculated land
cover heterogeneity with the Shannon index and correlated the
number of species in clusters with land cover heterogeneity
and urban density.
Results: Each land cover type contributes to species richness
and species composition of the clusters. Species richness decreases
significantly and linearly as urban density increases and
land cover heterogeneity decreases in the clusters.
Conclusions: A data mining approach can combine different
existing projects to improve knowledge of the urban vegetation
system. The methods we have applied offer tools to answer the
specific questions mentioned above.
