International Journal of Applied Earth Observation and Geoinformation xxx (2003) xxx–xxx A Multiscale Object-Specific Approach to Digital Change Detection Ola Hall a,∗ , Geoffrey J. Hay b a Department of Physical Geography and Quaternary Geology, Stockholm University, SE-106 91 Stockholm, Sweden b Geocomputing Laboratory, Département de Géographie, Université de Montréal, C.P. 6128, Succursale Centre-Ville, Montréal, Qué., Canada H3C 3J7 Received 20 December 2002; accepted 16 May 2003 Abstract Landscape spatial pattern is dependent not only on interacting physiographic and physiological processes, but also on the temporal and spatial scales at which the resulting patterns are assessed. To detect significant spatial changes occurring through space and time three fundamental components are required. First, a multiscale dataset must be generated. Second, a change detection framework must be applied to the multiscale dataset. Third, a procedure must be developed to delineate individual image-objects and identify them as they change through scale. In this paper, we introduce an object-specific multiscale digital change detection approach. This approach incorporates multitemporal SPOT Panchromatic (Pan) data, object-specific analysis (OSA), object-specific up-scaling (OSU), marker-controlled watershed segmentation (MCS) and image differencing change detection. By applying this framework to SPOT Pan data, image-objects that have changed between registration dates can be identified and delineated at their characteristic scale of expression. Results illustrate that this approach has the ability to automatically detect changes at multiple scales as well as suppress sensor related noise. This study was conducted in the forest region of the Örebro Administrative Province, Sweden. © 2003 Elsevier B.V. All rights reserved. Keywords: Digital change detection; Multiscale object-specific analysis; Watershed transformation; Remote sensing; SPOT; Hierarchy; Scale domain; Image differencing; Multiscale; Critical landscape threshold 1. Introduction Ecological theory, in particular hierarchy theory (Allen and Starr, 1982), predicts that changes in landscape spatial pattern and structure are depen- dent on the spatial and temporal scales at which they are assessed (Meentemeyer and Box, 1987; Turner, 1989; Malingreau and Belward, 1992). This theory also predicts that landscapes are composed ∗ Corresponding author. Tel.: +46-8-6747852; fax: +46-8-169629. E-mail address: ola.hall@humangeo.su.se (O. Hall). of separated levels (scale domains). Scale domains are separated by scale thresholds, which represent a break in the relative importance of process variables (Meentemeyer, 1989; Wiens, 1989). Therefore, any methodological framework for analyzing landscape change needs the capacity to explicitly handle scale. In general terms, ‘scale’ corresponds to a ‘window of perception’. As the size (i.e. grain and extent) of the window is changed, new patterns and struc- tures emerge; thus, the conclusions drawn by the observer are strongly influenced by the scale of ob- servation. The term ‘grain’ refers to the smallest distinguishable component, i.e. spatial resolution, 0303-2434/03/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0303-2434(03)00010-2