nbul 2004
IEEE
[KC 5| 10
Xes
and Spot
icentrated
igher than
d ED are
trial area
ium-wide
similar to
MA
1.62
0.93
0.15
0.92
1.62
CS
KONOS 10 —
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
T
[SPOT TRONOS
E
[LANDSAT — —
A LS
-— Hh B +
[IKONOS 10
| IKONOS 5
"T
i
|
Figure 17. Normalized graphs of edge density — "s
FLANDSAT — — vi SPOT ]IKONOS IKONOS 5 IKONOS 10
|
4
|
|
|
|
{
|
« |
|
|
|
|
Figure 18. Normalized graphs of mean patch area
7. CONCLUSIONS
With regard to the questions formulated in introduction, the
experiences that we have carried out allow us to draw the
following conclusions:
l. Landscape indexes analysed display values that are not
enough different to be assumed as variables characteristic of
various urban typologies. A reason for that could be an
inadequate or incomplete correspondence between conceptual
and cartographic definition of sub-areas. What are the limits of
concentrated and diffuse city? Conversely, the analysis of the
evolution of every sub-area, made using the same spatial
definitions, enables to identify a trend of development both
stable and characteristic of the various typologies.
2 Although calculated indexes were not characteristic, regarded
as a whole (ie. rejecting anomalous values) they offer a
comparative description among various sub-areas. Spatial
configurations can be analysed in detail with reference to the
specific meaning of indexes: for example edge density together
with mean size of patches gives information about
fragmentation, etc...
3. As supposed and foreseen, working at different scales
(IKONOS vs Spot and Landsat) entails working on different
phenomena. To sum up the results drawn from the second work:
e Density function shows great diversity with regard to the
attribution of pixels to a given class of density.
In IKONOS: low density surface is 36% as against 10% of
Landsat, while high density surface is 15% in Landsat as
against 8% in IKONOS. Such diversity is overcame through
a buffer applied to IKONOS, this mainly for high density
classes.
e Through the overlay of density classes maps a systematic
migration from a given density class of IKONOS into the
next class at higher density of Spot can be observed. Such
migration is recovered in IKONOS 5, whose density classes
distribution is near to Spot.
What happens is that in IKONOS the single building is detected
individually, while in Landsat and Spot either it is missing or it
is aggregated to the adjoining buildings. This explains the
maximum of built area in Landsat, the increase of low density
surface in IKONOS and the migration of density classes of
IKONOS towards higher density classes of LANDSAT. Buffers
allow to recover Spot and Landsat values with regard to :
- total built surface
47]
1
eed e E ERO RD M
- high density classes (A and B) since buffer fills the voids
between adjoining buildings.
Conversely, in low density values (C, D, E) the Spot and
LANDSAT values cannot be recovered because of the
dispersion of buildings and the variability of configurations.
In the third work, tables and their graphical representations
clearly show how indexes calculated over the 5 input images
display normalized values always opposite to IKONOS ones
in comparison with other images.
This because meaning of landscape indexes is completely
different when applied at different resolutions.
It is worth noting that ED is an index of dispersion and
fragmentation only at Spot and Landsat (Herold, 2001) and
not at IKONOS resolution. In IKONOS, ED is higher in
concentrated areas compared to scattered settlements, since a
higher number of buildings - and consequently a bigger
perimeter value — is associated to the same surface.
In this last instance, too, buffers applied to IKONOS allow us
to recover the behaviour of indexes at lower resolution for
concentrated urban areas.
BIBLIOGRAPHICAL REFERENCES
Fregolent L., 2004, Valutazione, pianificazione e sostenibilità
ambientale. Relazioni e metodi: la VAS di piani urbanistici
per sistemi territoriali complessi, Tesi di Dottorato in Scienze
e Metodi per la Città e il Territorio Europeo, XV ciclo,
Facoltà di Ingegneria, Università di Pisa.
Herold M. , 2001, Remote sensing and Spatial metrics-a new
approach for the description of structures and change in urban
areas, Proceeding of international Remote Sensing and
Geoscience Symposium (IGARGASS), Sydney.
Indovina F., Matassoni F., Savino M., Sernini M., Torres M.,
Vettoretto L., 1990, La città diffusa, IUAV DAEST, Venezia
Lacoste Y., 1980, Les objets geographiques, Cartes et figures
de la terre, Centre Georges Pompidou , Paris, Paris, pp. 16-23.
Megarical K., Cushman S.A.,, Neel M.C., Ene E., 2002,
FRAGSTATS: Spatial Pattern Analysis Program for
Categorical Maps, Computer software program produced by
the authors at the University of Massachusetts, Amherst.
Available at the following web site:
www.umass.edu/landeco/research/fragstats/fragstats.html
Pesaresi M., 1993, Analisi numerica dello spazio edificato
nella città diffusa Techical Report, IUAV DAEST, Venice.
Pesaresi M. and Bianchin A., 2001, Recognizing settlement
structure using mathematical morphology and image texture.
In Donnay J.P., Barnsley M. J. and Longley P.A., eds. Remote
Sensing and Urban Analysis, Taylor & Francis, pp. 55- 67.
Racine J.B., 1981, Problematiques, et metodologie: de
l'implicite à l'explicite, In : Isnard H., Racine J.B., Raynard
H., Problematiques de la geographie, PUF, Coll. Le
geographe, Paris.
Ruas À. and Bianchin A., 2002, Echelle et niveau de detail.
In: A. Ruas, ed. Generalisation et representation multiple,
Hermes, Paris, pp. 25-44.
