inbul 2004
oint out a
itecture is
it and (b)
ove).
! in Figure
eb service
he request
request is
ending on
link to a
will either
nform the
k that the
ds to the
ast vertex
estor gets
d how to
1e biggest
ained. By
'Ssibilities
] diverse
y forcing
ng in that
n Source)
— mostly
ation and
in are not
cr se. For
d (and in
ogies are
ice or the
lications.
iostic and
ery and
rvices in
or more
arm and
work for
es.
e) (W3C,
! describe
changing
yrocedure
C, 2004),
choice
> biggest
is XML
used) is
tocols as
HTTP,
amlessly
id usc of
ise of the
; or the
e service
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
can be used by both desktop and web based applications, one
can see that the role of the description, discovery and
invocation of a web service is equally important to the service
itself.
2.3 GIS Web Services
Geographic Information Science has a lot to benefit by the
adoption of the service computing model. As mentioned also in
the introduction geographic information comes from different
and diverse sources and in different formats. This is especially
true for the environmental related information which has to
combine not only data form different sources but also models
and software. The GIS web services initiative is mainly driven
by the Open GIS Consortium (OGC) (OGC, 2004). OGC has
provided specifications for many web based applications like
the OpenGIS Web Map Server and data descriptions like GML
(Geography Markup Language). But this only makes the
integration easier. GIS web services require the use of
technologies described in section 2.2 in order to become truly
available to users.
Environmental monitoring applications in particular, like the
one described in this paper, are extremely suitable to be
developed under the web services model. These applications are
based upon the availability of large volumes of data sets in the
form of both a large enough number of satellite images, each
one having a size of several megabytes, and a number of
intermediate images produced by the specific algorithms used.
In a application developed under the traditional software model,
one (but everyone) would have been obliged either to download
and keep all these data in the local machine (in the case of a
desktop application) or use a web-based application with a more
or less restricted functionality.
VEGETATION
3. SATELLITE AND
INDICES
IMAGERY
Considering vegetation as a functional equivalent of terrestrial
ecosystems, it follows that changes in vegetation structural
dynamics provide important indications for physical process in
space and time. A widely used parameter to investigate
vegetation conditions is the Normalized Difference Vegetation
Index (NDVI), which uses the strong reflection of living
vegetation in the near infrared region of the electromagnetic
spectrum and the relatively low reflection in the visible red
wavelength (Lillesand & Kiefer, 1987). Thus, the NDVI is the
difference of near-infrared (NIR) and visible (VIS) reflectance
values, derived from multispectral satellite images, normalized
over the sum of the respective image channels: (NIR-
VISJ(NIR-VIS). Several works have processed and made
available global or regional NDVI data sets for subsequent use,
either directly in modelling applications or in studies to extract
land cover information (Los et al., 1994; Sellers et al., 1994; El
Saleous et al., 2000).
3.1 Satellite images
The AVHRR (Advanced Very High Resolution Radiometer)
onboard NOAA satellites is a radiation detection imager that
can be used for remotely determining the vegetation status of
the earth surface (Cracknell, 1997). AVHRR was a 4-channel
radiometer, first carried on TIROS-N (1978). This was
subsequently improved to a 5-channel instrument (AVHRR/2)
that was initiallv carried on NOAA-7 (1981). The latest
instrument version is AVHRR/3, first carried on NOAA-I5
675
(1998). AVHRR/3 records the incoming radiation in 6 spectral
bands: 0.580 - 0.680 um (Channel 1), 0.725 — 1.100 pm
(Channel 2), 1.580 — 1.640 um (Channel 3A), 3.550 — 3.930 um
(Channel 3B). 10.300 - 11.300 um (Channel 4) and 11.500-
12.500 pm (Channel 5). Channels 1 and 2 are the VIS and NIR
chnnels, respectively, whereas both 4 and 5 are thermal infrared
(TIR) channels. Each AVHRR pass provides a 2399 Km wide
swath with a ground resolution of 1.1 Km at nadir from the
nominal orbit altitude of 833 Km.
NOAA satellites orbit the Earth 14 times per day. The AVHRR
provides on board collection of data from all spectral channels.
These data are coded in 10 bits and transmitted to ground
receiving stations. The FORTH (Foundation for Research and
Technology - Hellas) ground receiving station provides all
required data to the developed GIS web services. FORTH
station carries the advantage of near real time AVHRR image
acquisition from NOAA 12, 14, 15, 16 and 17 satellites. All
images used by the GIS web services are pre-processed using
station’s capabilities (Dartcom, 2002). Pre-processing includes
the following three steps: a) Polar navigation and re-projection
toa 1.1 x 1.1 Km cell grid (Hugget and Opie, 2002). b) Subset
and conversion to a S-layer raster with no change to the original
pixels! Digital Numbers (DN). c) Export to the application's
database.
3.2 The NDVI algorithm
The NDVI is calculated for cloud free land areas taking into
account the received radiation in AVHRR channels | and 2:
Nols nme Champ
Channel2+Channell
(D
Eq.(1) produces values in the range of -1 to 1, where increasing
positive values indicate increasing green vegetation and
negative values indicate non-vegetated 'surface features.
AVHRR channels 1, 2 and 5 used by the NDVI algorithm.
Channels | and 2 are used to implement Eq. (1), whereas
channels 1 and 5 are used for cloud masking.
In practice, the NDVI algorithm consists of the following steps:
a) Calibration of channels I, 2 and 5 to convert pixel's
DN to radiance values (Wm um’ 'sr™'), using the pre-
launch calibration coefficients issued by NOAA
(Kidwell, 1998; Goodrum et al., 2000).
b) Production of a cloud mask (Chrysoulakis and
Cartalis, 2002).
c) Production of an intermediate NDVI image using the
Eq. (1).
d) Production of the final NDVI image combining steps
b) and c).
3.3 The Implementation of the NDVI algorithm
The implementation of each algorithm step is performed by a
different module of a Java-based in house software. In
particular, each module consists of one or more programme
classes, which have been designed using the Java2
programming language. AVHRR image manipulation is
achieved with the use of the immediate mode imaging model of
Java 2D API (Application Programming Interface.), which is a
set of classes for advanced two-dimensional graphics and
