MULTI-SCALE APPROACHES FOR GEODATA
Martien Molenaar
Centre for Geo-Information Processing (CGI)
Wageningen Agricultural University, The Netherlands
E-mail: MARTIEN.MOLENAAR@WETENSCH .LMK.WAU.NL
Tel: +31(0)8370-82910 Fax: +31(0)8370-84643
ISPRS Intercommission Working Group llI/IV
KEY WORDS:
object aggregation.
ABSTRACT:
Geo-data models, generalization, multi-scale approach, spatial objects, topology, class hierarchies,
Topological object relationships in combination with object classification hierarchies appear to be fundamental in the
definition of the aggregation rules for spatial objects. Such rules are essential building blocks for the construction of
generalization procedures in spatial databases. A model for spatial database generalization can be formulated based
on the syntax of the Formal Data Structure (FDS) as proposed in (Molenaar 1989). The syntax of the FDS willbe formalised
first, then database generalization procedures will be formulated with this syntax.
Four strategies will be explained for generalization, these are:
- geometry driven generalization, the change of geometric resolution cells determines the transition from entities
at a large scale to new entities at a smaller scale,
- class driven generalization, spatial objects at a large scale forming a region under one thematic class are merged
for representation at a smaller scale,
functional generalization links objects that are considered as response units in processes defined at different
scale levels,
- structural generalization gives a stepwise simplification of a spatial process description in an area.
These different strategies will be explained and compared. In the final discussion spatial data generalization will be
presented as data transformation processes. For such a transformation we should specify which aspects of a terrain
description should be invariant after generalization, this may have the effect that other aspects will not be invariant.
1. INTRODUCTION
1.1 Spatial Processes at Multi-Scale Levels
Multi-scale approaches are at present one of the focal
points of the GIS research community. This is due to the
rising awareness that many processes on the earth surface
can only be monitored and managed if they are understood
in their geographical context. Part of this context is
defined by the scale range at which these processes work.
If we consider for instance the development of land use
in a district, then we see that this is driven by actors at
a lower aggregation level such as farmers, residents and
companies. Their activities are constrained however by
the socio-economic conditions and the infrastructure at
regional level and by the macro economic planning at
national and supra national level. An other example is the
development of the natural vegetation cover of certain
regions. The actual state of such a vegetation cover is
defined by the co-occurrence of species that form
vegetation types, which are part of eco-systems. Their
development will be constrained by climatic conditions,
the geologic and soil condition of the region and its
542
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
hydrology. Here too we find hierarchical levels of
organization.
The monitoring and management of such processes requires
information at different scale levels. The research problem
for the GIS community in this context is:
- to decide for each type of process which information
should be handled at each scale level,
- to develop methods for transferring information
between the different scale levels so that duplication
of expensive data acquisition can be avoided as
much as possible, and so that the consistency
between data at the different scale levels can be
maintained.
This second item is strongly related to the long standing
research problem of map generalization, that is why it is
often seen from that perspective. Researchers in this field
become more and more aware of the fact, however, that
multi-scale approaches in a GIS environment can be dealt
with by data base generalization operations. These allow
approaches that are quite different form the procedures
applied in map generalization and they are more flexible.
This new research field derives it terminology and many
of its c
oriente«
of the
confusi
terminc
researc
solve r
perspec
empha:
cal) dai
The diff
of spat
strateg
problen
1.2 Sp
A spati
elemen
elemen
objects
operati
contair
the sta
ships. 1
of a hic
should
mappe
relatior
structu
level tt
to relat
standir
levels
genera
Aggrec
basic t:
the co
abstrac
be bas
valued
aspect
Point-,
geome
tation :
ships,
object
class |
relatior
hierarc
hierarc
of spa
theses
for mc
differe
Inform
topolo(
hierarc
tion of
In a ca
as the
