94
There are 4 main ecological regions:
1. the Low Karst in the west shows a widespread
natural afforestation of abandoned farmland,
2. the Nanos and Javorniki mountains in the SE-NW
direction are heavily forested,
3. the continental part in the northeast features
fragmented forest cover and 2 large marshy
depressions,
4. the tip of the Vipava valley in the northwest and the
Brkini hills in the south represent an intensively
cultivated sub-mediterranean landscape.
THE DATA
The continuous process of widespread natural
afforestation renders the drawing of a clear line between
forest and non-forest impossible. There are several
transitional vegetation successions from managed
farmland to full-grown forest, but not all of them can be
identified from Landsat TM and GIS data, nor are they all
interesting for most users. It was decided to distinguish
the following 4 land cover classes in the final map:
1. Non-forest,
2. Abandoned pasture,
3. Shrub,
4. Forest.
Because of the heterogeneity of the non-forested areas
and in order to reduce the within-class variability in the
satellite image, the "Non-forest" class was additionally
partitioned into the "Unvegetated", "Water", "Marsh" and
"Farmland" subclasses (see also Table 1). In the
intermediate stages of our study we thus considered 7
classes (termed CLC_G classes). Only the final result of
classification was reclassified back into the 4 main
classes.
Table 1: Aggregation of the original CLC classes (EC 1993) into the more general CLC_G classes of the reference data
set
Attribute / GIS layer
Content
CLC_G
The generalized CLC classes for training (see Table 1)
UNSUPERVISED_RESULT
Result of the unsupervised classification
FOREST81
Presence / absence of forest in 1981 (SMAS 1995b)
DEM
Digital elevation model, 100 m resolution (SMAS 1995a), interpolated to 25 m
resolution
SLOPE
Slope in degrees, derived from DEM
LAPLACE
Surface roughness, derived by the Laplace filter from DEM
INSOLATION
Yearly global insolation in MJ/rn^ (Gabrovec 1996)
POP_DENSITY
Population density at a geographical subregion level (Perko et al. 1998)
POP_AGE_INDEX
Ratio of population aged under 20 to aged over 60 (adapted from Perko et al. 1998)
PROX_SET_HW
Proximity to the nearest settlement or highway in meters, based on a digized maps of
settlements (SMAS 1995b); the highway was digitized on-screen from the Landsat TM
image
PROX_WATER
Proximity to the nearest stream or lake, based on a digitized hydrology map (SMAS
1995b)
PROX_LAKE
Proximity to the nearest lake, based on a digitized hydrology map (SMAS 1995b)
TM2
Ortorectified, topographicaly normalized image channels of Landsat TM, acquired July
26, 1995 (EC 1997, NLR 1997). TM1 was not used because of a considerable
stripping.
TM3
TM4
TM5
TM7
NDVI
NDVI = (TM4 - TM4) / (TM4 + TM4); Normalized difference vegetation index was
computed from the Landsat TM image channels prior to topographic normalization.
Table 2: Attributes of each pixel