u 2004
da into
S were
led as
e et al.
hanges
Jtes on
en be
hanges
used a
ssisted
on the
st, the
ctions.
isis for
egions
| cities
f arcs
ffers a
sets.
of the
/namic
he end
6) and
these
cts are
Ma are
stances
sed on
in the
xisting
on of a
licated
orm of
ddress
policy
coastal
is aim
model
of the
cimum
M that
e, the
to the
| and
speed)
d may
d with
tion of
ligent"
text of
3 been
X the
livided
nts) in
| rise.
ording
nomic
se, the
in the
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
coastal zone should reflect the impacts and responses of the
coastal system to this factor. Based therefore on the assumption
that variations in vulnerability within the coastal zone are
controlled by primary variations in the human and physical
coastal interchange several critical parameters were identified.
These parameters were (i) the geomorphic structure of the
coastal environment, (ii) the potential for wetland migration,
(iii) the locations of major rivers and deltas, (iv) population
density classes and (v) administrative boundaries. According to
these parameters, the world's coastline was segmented into
homogeneous coastal segments on which all modelling and
analysis in DIVA were based. The theory behind the
segmentation methodology employed in DINAS-COAST is
discussed in detail by McFadden et al. (2003).
The choice of the data model also addresses the physical and
functional requirements of the project. The segmentation of the
coastline was used as a means to provide a series of spatial
reference units for the modelling tool of the project and to link
it to the GIS database and to the graphical output. Considering
the restrictions of storage space, imposed by the form in which
the DIVA tool has been produced and will be disseminated, the
database that would be included in the final product had to be
as lean and as well-structured as possible, without however
compromising the quality of the output results. À large database
would not only be impossible to store on a single CD-ROM but
would also substantially decrease the performance targets set
for DIVA, thus rendering the tool inefficient and user
unfriendly. Based on these facts and also due to the scale and
size of the project, the selection of the segmentation criteria
described in the previous paragraph was necessary for
achieving the target of a manageable database. In this way it
was ensured that the database would not be populated by “an
unmanageable profusion of impossibly small line segments ...
and degenerate into an unmanageable and grossly
overcomplicated assemblage” (Bartlett et al., 1997, p.141)
which would have severe impacts on the analysis and graphical
display performance of the tool.
3. GIS IMPLEMENTATION OF THE COASTLINE
SEGMENTATION - THE DIVA DATABASE
Due to the explicit spatial nature of the data and of the type of
operations required for populating the database, both tasks were
performed as pre-processing steps within a GIS, externally to
DIVA (Hinkel and Klein, 2003). For this purpose, a GIS
database containing individual data layers that have been
collected from various sources (table 1) was compiled. Due to
the scale of the project, data collection primarily relied upon the
contents of the vast archives of existing datasets. Datasets
derived from remote sensing data have been an important data
source as such data offer a globally coherent view of the earth,
are largely unaffected by national data collection practices, can
be aggregated comparatively easily and are readily available
(Rhind and Clark, 1988). These datasets were complemented
with analogue cartographic data from various sources which
were converted into digital form and were incorporated in the
database. According to Rhind and Clark (1988) the integration
of remotely-sensed and analogue cartographic data offers the
most promising approach to globally coherent, up-to-date and
scientifically valid databases. Finally, tabular datasets collected
from various organisations (e.g. UNESCO) as well as expert-
judgement assessments were employed in cases where gaps in
global coverage existed for certain parameters.
803
Dataset Format Source
Gridded Population Raster CIESIN
of the World
World Elevation Raster NGDC
and Bathymetry
Geomorphic type Analogue map McGill, 1958
Landform type Analogue map Valentin, 1952,
Tidal Range Raster IGBP-LOICZ
Wetlands database Tabular CCRU
Second Level Polygon DCW, ESRI
Admin. Boundaries
Table 1. Characteristic datasets, from different sources and in
different formats, that were included in the database
After the compilation of the GIS database, the segmentation of
the world's coastline was performed according to the criteria
described in section 2.2. This process produced. 12,148 coastal
segments (Figure 1).
Be,
à
X + 3
vu]:
M
rl
x^
Mt
~
1
À
f
47 com
nf
- - o
> >
} a } z
7 { A Cx T
= 4 Ana d SS A
a
e - T ? a ER
T T T
couw vote wrovt 2009 xeovt
Figure 1: An example of the coastline segmentation. Display of
segments for the European and North African coast.
Since the original data that were stored in the GIS database
were in various formats (e.g. rasters, polygons, points),
extensive GIS processing was undertaken in order to reference
the existing data to the linear-representation model of the coast
that was chosen for DIVA. The processing methodology varied
for each dataset, depending on factors such as the source, the
origin, the nature and the format of the data and the
requirements of the DIVA model. The methodology used for
each parameter is analytically described in the associated DIVA
metadata files.
The final outcome of this referencing process was a database
where each coastline segment has a unique numeric id and is
associated with its attribute data in a table (Figure 2). Attribute
data are included at different spatial levels (e.g. country,
administrative units, regions) depending on the scale and
accuracy of the original data and on the requirements of the
algorithms incorporated into DIVA. The DIVA database is
STOO
709
570
