AUTOMATED MAPPING WITH LiDAR AND SPECTRAL CHARACTERIZATION IN MEDITERRANEAN FOREST AGROECOSYSTEMS

Abstract

Mapping with LiDAR data is not a standardized practice, though LiDAR databases
are increasing in all countries in Europe. We develop and test a simple method for
automated land-cover mapping. The study area was a farm located at a natural park
of southern Spain. It comprises 502 ha covered by Mediterranean forest
agroecosystems, like dehesa (a very open woodland of scattered evergreen trees
used by grazing animals), woodland and scrubland, and transitions among them,
composing a heterogeneous landscape. This heterogeneity is caused by variations
in holm and cork oak tree density and a sclerophyllous shrub cover, i.e., 3D
structure of woody vegetation. Using aerial photographs digitization, Landsat
image classification, and image segmentation of tree crowns, land-cover maps were
generated. Besides, other maps were produced from LiDAR-derived canopy cover
and height of tree vegetation and shrub stratum. These 3D variables allowed to a
wall-to-wall characterization of woody vegetation land-cover classes in the study
area, that was completed with a NDVI assessment. The results show that automated
mapping with LiDAR is reliable and accurate enough in comparison with other
mapping techniques. It outperforms them because its higher spatial resolution, and
can be combined with other remote sensing methods to provides an improved
understanding of forest landscapes.