stanbul 2004 
ipal public 
y in Urban 
1965 
bad and its 
ing House- 
Delhi. Inter 
n response. 
Manoharlal 
man Well- 
ouse, New 
nt. Rawat 
idition).BS 
QUANTIFYING THE ECOLOGICAL PATTERNS OF URBAN DENSIFICATION 
THROUGH MULTIPLE ENDMEMBER SPECTRAL MIXTURE ANALYSIS, 
LANDSCAPE METRICS, AND FUZZY LOGIC 
WG VII/4 Human Settlement and Impact analysis 
T. Rashed 
Department of Geography, University of Oklahoma, Sarkeys Energy Center 680, Norman, OK 73019, USA -- rashed@ou.edu 
KEY WORDS: Sustainability, Urban Change, MESMA, Fuzzy Logic, Landscape Metrics, Los Angeles 
ABSTRACT: 
This paper introduces an integrative methodology that has been developed to measure temporal changes in urban morphology based 
on the techniques of multiple endmember spectral mixture analysis (MESMA), landscape metrics, and fuzzy logic. In order to 
illustrate an application of the methodology, the paper uses two satellite images acquired in 1990 and 2000 for the metropolitan area 
of Los Angeles County, California, a megacity with complex morphological patterns that are rapidly changing due to a range of 
complex, interrelated forces of urbanization that are poorly understood. Through a wall-to-wall exercise, the paper discusses: (1) 
how the spatially continuous character of urban morphology in Los Angeles has been analyzed through the MESMA technique to 
capture and quantify within-class changes at the sub-pixel level; (2) how the magnitudes of changes have been assessed through 
fuzzy logic; and (3) how landscape metrics have been applied to quantify the ecological patterns of change at the census tract level. 
1. INTRODUCTION 
1.1 Background 
Satellite remote sensing has been widely recognized as one of 
the essential technological tools for sustainable development. A 
considerable number of recent studies has been conducted 
utilizing satellite sensor data in the analysis of urban change 
(e.g. Kwarteng and Chavez, 1998; Costa and Cintra, 1999; 
Chen et al., 2000; Ward et al., 2000; Batty and Howes, 2001; 
Madhavan et al., 2001; Yang and Lo, 2002; Herold et al., 2003; 
Weber and Puissant, 2003; Rashed et al., 2004). The findings of 
these studies have enriched our understanding of the physical 
and socioeconomic drivers of changes in urban land cover and 
the implications of these changes on land use practices and 
resource management in cities. Some of these studies went 
further beyond the characterization of change and its causes and 
attempted to integrate remotely sensed data with models of 
urban growth in order to project future changes in a given city. 
Looking back at how these studies have informed and been 
linked to sustainability policies, one can easily observe the sole 
focus on only one type of sustainable development, the so- 
called “smart growth,” “managed growth,” or “new urbanism.” 
These and other similar approaches direct attention to changes 
that occur at the urban fringe, as in the case of “edge cities” in 
the U.S. and decentralized suburban communities in Europe. 
However, few studies have directed attention to another 
important mode of urban land cover and land use change taking 
place within existing landscapes in cities (Rashed et al., 2004). 
Urban landscapes change over time as new urban fabric is 
added and also as the existing fabric is internally modified (e.g. 
new buildings replace old ones, plots are amalgamated or 
subdivided, street layouts are modified) (Knox, 1995; 
Cadwallader, 1996). These patterns of urban densification and 
internal modifications are of major concern to sustainable 
development because they represent the physical manifestation 
of a range of social, economic, cultural, and political 
dimensions associated with urban dynamics. Moreover, 
densification typically takes place locally within urban 
neighborhoods where the impact of sustainable policies is more 
spatially and socially manifested. Hence, a better 
characterization and quantification of densification patterns at 
the local level will both provide a rich understanding of the 
processes involved and challenge the credibility of policy, 
citizens’ preferences for sustainable living space, and 
503 
developers’ understanding of and attitude towards those 
preferences (Webster and Senior, 1999). 
This paper introduces a methodology for quantifying the 
ecological patterns of urban densification taking place at the 
local or microscale level based on satellite imagery. The 
suggested methodology builds on a rigorous framework of 
urban landscape ecology (Ridd, 1995) and integrates the 
techniques of multiple endmember spectral mixture analysis 
(MESMA), landscape metrics, and fuzzy logic. The proposed 
methodology brings these techniques together in a way that 
allows for: (1) a better decomposition of the urban landscape 
into its underlying land cover materials; (2) an improved 
assessment of urban land cover change that takes into account 
not only changes between land cover classes but also within 
land cover classes, and quantifies the actual magnitude of this 
change (e.g. high increase, lower decrease, no change, etc); and 
(3) a quantitative comparison of the ecological patterns of 
change in land cover between urban locales. In the following 
sections, the proposed methodology is described and its 
application is demonstrated through a case study that utilized 
Landsat data to quantify urban land cover change in Los 
Angeles County, California, between 1990 and 2000. 
2. METHODS 
2.1 Approach 
The proposed methodology for quantifying the ecological 
patterns of urban densification consists of three sequential 
phases. In the first phase, the MESMA technique is separately 
applied to individual images in order to derive per-pixel 
physical measures of urban land cover abundance at a given 
point of time. Land cover fractions of individual dates are then 
validated against test data to determine the accuracy of 
MESMA-derived measures. Once acceptable, multi-date 
fractions of corresponding land cover materials are used to 
calculate per-pixel temporal differences in these fractions. 
Hence, the resultant fractional differences will represent a direct 
measure of the changes that take place in the composition of 
urban morphological patterns over time due to such processes 
as urban densification and urban sprawl. In the second phase of 
the methodology, the magnitude of change in land cover 
fractions is estimated through fuzzy logic. A number of 
predefined fuzzy membership functions are applied to 
characterize the magnitude of change in each type of urban land 
EN 
É 
a