International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
locate. Field work is time consuming and costly as well.
Therefore, airphotos are used as complementary or in
place of field trips.
Several authors experienced the air photos usefulness
for digital processing assistance, either for forests or
natural landscapes (Fontes et al. 1998, Hotyat 1998). In
previous studies, photo-interpretation helps in
characterizing cartographic units (Codjia et Domingo
1998). Airphotos, sometimes orthophotos, are often
used to locate sampling plots before visiting them in the
field (Hotyat 1998). In the present study, orthophotos
are used as ground data to validate training areas on
satellite products.
2.2 Study Area
It lies on the coast of Saronic Gulf, 20km East of
Korinth city, on about 37°S5’N and 23°07’E (Fig.1).
From the geomorphologic point of view, this 5km x
5km area is located on the southern slopes of Yerania
Hill (1351m). The studied zone covers the lower part of
these slopes, below 500 m elevation. Soils are mainly
clayey with some outcrops of sandstones. In the
southern centre part of the scene, the most conspicuous
components are Agioi Theodoroi harbour on alluvial
deposits, on a small river mouth, and an important
refinery. This region fringing Saronic Gulf belongs to
the * thermo-Mediterranean zone * (Ozenda 1975). It is
one of the driest parts of Greece, characterized by poor
rainfall (less than 400mm/year), with the main rainy
season in autumn and winter (October-December) and a
severe dryness in summer. In summer, mean monthly
temperature is between 21? and 26?C with maxima
around 32°. In winter, mean monthly temperature is
between 10? and 12?C and minima are not lower than
7? near the coast. Landward, temperatures noticeably
decrease with increasing elevation. The warmer fringe
of this country does not extend beyond 200m elevation.
The characteristic vegetation cover growing from the
sea-side up to 200m should be a low broadleaved
evergreen forest (Debazac et Mavrommatis 1971.
Barbero et Quézel 1976) typically comprising Pistacia
lentiscus, Olea europea, Ceratonia siliqua, Juniperus
phoenicea, Myrtus communis ; Pinus halepensis is also
present. Today, this original forest type has practically
disappeared, replaced by human settlements and crops,
roads, plants (refinery). Only an important open pine
forest (Pinus halepensis) still remains, greatly
threatened by human activities.
Above 200m, the typical spontaneous vegetation type is
a low (6-8m) broadleaved evergreen forest with
Quercus coccifera and Phyllirea media. Some remnants
of this forest type are still preserved on steep slopes,
cool bottom valleys and to the upper part of the scene.
However, due to ancient human pressure, it often gives
way to a bush, 2 or 3m high, characterized by the
shrubs Arbutus unedo, A. andrachne and Erica arborea.
But the largest area, behind the urbanized coastal
fringe, is covered by * phrygana *, which is the ultimate
degradation stage of forest, due to overgrazing and
fires. Whereas in Greece, these practises have given
way to three kinds of degradation stages : ' garrigue ‘
mainly corresponding to a quantitative impoverishment
of Pistacia lentiscus and Quercus coccifera formations ;
bush and moor with predominant Ericaceae ; phrygana
when other shrubs are predominant. These phryganas
are poor shrubby vegetation types, more or less low and
scarce, with various facies. The main dwarf scrubs
constituting phryganas are deserted by cattle. The
commonest facies in the study area, deriving from bush,
is characterized by several species of Cistus (C.
monspeliensis, C. villosus, C. salviaefolius). Other
common facies are those with Phlomis fruticosa on
rocky soil, or with Poterium spinosum and
Coridothymus capitatus on stony soils (Debazac. et al.
1971).
2.3 Data
An IKONOS imagery was acquired from ‘Space
Imaging Europe‘ (Fig.1). The characteristics of this
CARTERRA Geo Product are given below :Area-
around 25km? (5km x 5km), Location- Agioi Theodoroi
, Ground resolution- 1m, Image size- 6648 columns,
6296 rows, Data type- Pan-sharpened multispectral,
Four channels- blue (0.45-0.52um), green (0.52-
0.60pm), red (0.63-0.69um), near-infrared (0.76-0.90),
Processing level- Standard geometrically corrected,
Datum- WGS 84, Map projection- UTM 34, Date- 14
July 2000. The processing software is IDRISI 32.
Aerial photographs on 1/15 000 scale and topographic
maps on 1/5000 scale were acquired from the * Hellenic
Military Geographical Service *. Aerial photographs (15
June 1998) have been geometrically corrected in order
to produce orthophotos.
2.4 Method
Based on general knowledge of plant formations in this
region, a preliminary image FCC interpretation allows
to select 12 items or classes and delineate
corresponding training areas: 1.Pine forest, 2.Mixed
forest, 3.Broad-leaved forest, 4.Broad-leaved forest
under smoke, 5.Phrygana, 6.Phrygana under smoke,
7.Fire scars, 8.Crops, 9.Bare cultivated soils, 10.Roads,
11.Settlements, plant, 12.Sea. These classes are
provisional and should be defined again when getting
airphotos. For each class, at least one training polygon
is selected and delineated on the imagery. Then, an
automatic classification is produced using the
maximum likelihood algorithm (Fig. 3). At this stage, a
first estimation regarding the size and location of each
class is available. This first classification will be
modified and improved after the study of airphotos and
ground truth data.
Ground data are given by scanned panchromatic
orthophotographs. The 12 classes used for the above
classification of IKONOS imagery are interpreted in
selected points derived from a systematic sampling. So,
1286 points were located on the airphotos (Fig. 2). One
of the classes 1 to 12 (excluding 4 and 6, i.e. smoke)
was affected to the ground data sampling points,
avoiding to consider borders of landcover units. So, the
random sampling has been locally slightly modified in
order to avoid the *edge effect'. This means that when a
border was in the middle of the small sampling
square/circle we sampled few pixels further. Then, each
class on the classified image is compared to the
632
pe
cl
nu
th
nu
CO
thi
im
ar
the
2,1
Tt
ort
IK
int
(cl
°P]
(cl
be
wa
the
un
anc
to
Hel
froi
[K(
ord
con
the
agr
The
acci
soni
orth
revi
hale
mec
4.Bi
smo
7.Fi
tracl
12.S
imag
CH
mer;
clas:
plun
resp
mas]
conf
num
pixe
80.0:
Cons
may
beco
