1l 2004
ges;
tial
ted
an
1€
2. Decision-makers at different decision-levels can
make sensible consistent decisions on forest area
change, because overall the two images responded
quite similar to the LCM parameter spatial size (with
restriction of conclusion 3).
Decision-makers at different decision-levels need
temporal maps that are produced with identical
parameter settings to avoid comparing apples and
oranges, because the LCM parameter spatial size
proved not to be consistent in the arrangement of
forest cover.
m
In addition, regarding the sensitivity estimators it can be
concluded that the composition and configuration measures as
used at both the class and landscape level are very useful for
remote sensing based classifications in addition to conventional
KHAT statistics and cover values (as measure by PLAND). In
this respect, the indices LSI and KHAT seem to be a useful
combination to indicate the levels of disaggregation at which
classification results differ significantly. The indices NP and
KHAT seem to be a useful combination to indicate which
classes are causing significant differences between
classification results. The indices LSI and PLAND, or NP and
PLAND (a configuration and a composition measure) seem to
be a useful combination to indicate suitable combinations of
parameter settings for the LCM parameter size to compare
temporally different images. Moreover, the indices SIDI and
PLAND (both composition measures) seem to be a useful
combination to indicate underlying change processes in forest
areas. SIDI provided the area range (35-100 ha) at which a
striking diversity change occurred, whereas PLAND indicated
the vegetation class most likely involved (shrub vegetation).
Combining both measures, it can be concluded that shrub,
which is an intermediar vegetation in the transition from forest
to agriculture and vice versa, contributed significantly to the
change in SIDI and thus plays an important role in the
underlying change process. This means that between 1990 and
1996 forest is depleted not due to logging practices, but due to
agricultural practices. This type of information can be very
useful for decision-makers at local, provincial and national
level, who are urged to preserve rainforest for future
generations.
REFERENCES
Bandibas, J.C., Bruce, R.C. & Daels, L., 1995. Using frequency
based classifier for remote sensing of spatially heterogeneous
land-use/land-cover systems in the tropics. Asian-Pacific
Remote Sensing Journal 7(2), pp. 65-70.
Beurden| A.U.CJ. van & Douven, W.LA.M. 1999.
Aggregation issues of spatial information in environmental
research. /nternational Journal of Geographical Information
Science, 13(5), pp. 513-527.
Burnett, C. & Blaschke, T., 2003. A multi-scale segmentation/
object relationship modelling methodology for landscape
analysis. Ecological Modelling, 168(3), pp. 233-249.
Carvalho, L.M.T. de, Fonseca, L.M.G., Murtagh, F, & Clevers,
J.G.P.W., 2001. Digital change detectio with theaid of
multiresolution wavelet analysis. International Journal of
Remote Sensing, 22(18), pp. 3871-3876.
Congalton, R.G., Oderwald, R.G. & Mead, R.A., 1983.
Assessing Landsat classification accuracy using
discrete multivariate analysis statistical techniques.
Photogrammetric Eng.&Remote Sensing, 49(12), pp.
1671-1678.
795
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
Cortijo, F.J. & Perez de la Blanca, N., 1998. Improving classical
contextual classifications. /nternational Journal of Remote
Sensing, 19 (8), 1591-1613.
FAO, 2001. The Global Forest Resources Assessment 2000
(FRA 2000). FAO Forestry Paper 140,
http://www.fao.org/forestry/fo/fra/main/index.jsp (accessed 5
Nov. 2003).
Forman, R.T.T. 1995. Land mosaics: the ecology of
landscapes and regions. Cambridge University Press,
Cambridge, pp. 317-321.
Gustafson, EJ, 1998. Quantifying landscape spatial
pattern:What is the state of the art. Ecosystems, 1(2), pp.
143-156.
Hill, R.A., 1999. Image segmentation for humid tropical forest
classification in Landsat TM data. [Internation Journal of
Remote Sensing, 20(5), pp. 1039-1044.
Hudson, W.D. & Ramm, C.W., 1987. Correct formulation of
the Kappa coefficient of agreement. Photogrammetric
Engineering and Remote Sensing, 53(4), pp. 421-422.
Kushwada, S.P.S, Kuntz, S. & Oesten, G., 1994. Applications
of image texture in forest classification. /nternational
Journal of Remote Sensing, 15 (11), 2273-2284.
Lund, H.G., 1999. Forest classification: a definitional
quagmire. In: The World's Natural Forests and their Role in
Global | Processes, Khabarovsk, Russia, 17; p.
http://home.att.net/~gklund/quagmire.htm (accessed 7 Nov.
2003).
McGarigal, K., S.A. Cushman, M.C. Neel, and E. Ene (2002),
FRAGSTATS: Spatial Pattern Analysis. Program for
Categorical Maps. University of Massachusetts, Amherst.
www.umass.edu/landeco/research/fragstats/fragstats.html
(accessed 23 Feb. 2004).
Molenaar, M., 1998. An introduction to the theory of spatial
object modelling for GIS. Taylor & Francis, London, pp. 20-
31.
Obbink, M.H., Clevers, J.G.P.W. & Viergever, K.M., 2002.
Multi-level segmentation of remote sensing imagery for
mapping forest transitions. In: Operational tools in forestry
using remote sensing techniques Edinburgh, Scotland, 10 p.
Palubinskas, G., Lucas R.M., Foody, G.M. & Curran, P.J.,
1995. An evaluation of fuzzy and texture- based classification
approaches for mapping regenerating tropical forest classes
from Landsat-TM data. /nternational Journal of Remote
Sensing, 16 (4), 747-759.
Schaijk-Obbink, M.H. van, Molenaar, M. & Gier, A. de, 2000.
Tropical forest change assessment: the use of aggregates, a
semantic approach. In: International Archives of
Photogrammetry and Remote Sensing, Amsterdam, The
Netherlands, Vol. XXXIII, Part B7, p. 1601-1608.
Whitmore, T.C., 1998. An introduction to tropical rain forests
— 2" ed. Oxford University Press, Oxford, 282 p.
Woodcock, C.E. & Strahler, A.H., 1987. The factor of scale in
remote sensing. Remote Sensing of Environment, 21, pp.
311-332.
ACKNOWLEDGEMENTS
SRON, Wageningen University and ITC (The Netherlands) are
acknowledged for supporting this research.
