Subpixel Land Cover Change from Landsat:
A Case Study for Durban, South Africa (1985-2009)
Christopher Small, Lamont-Doherty Earth Observatory, Columbia University
Cristina Milesi, California State University Monterey Bay / NASA Ames Research Center
Introduction
Satellite-based land cover change characterization of urban areas and their
surrounding regions offers a valuable tool for investigating the effects of
socio-economic changes on the physical environment, and thus understand
the challenges for sustainable development.
However, land cover change characterization over urbanized areas is
complicated by the presence of spectrally heterogeneous land cover types
within the field of view of satellite sensors with spatial resolution greater
than a few meters. Multitemporal spectral mixture analysis (MSMA) derives
physical representations of the temporal changes within spectrally mixed
pixels as areal fractions of spectral endmembers.
Here we show an example of land cover change prepared in occasion of the
2011 COP 17 meeting in Durban, South Africa. Durban is an example of an
African city where changes in climate have the potential to pose significant
challenges on the local population that is already vulnerable in terms of
poverty, health, and secure access to food and water. Characterization and
maping of changes in vegetation, shadow and impervious fractions from the
Landsat archive offers an opportunity to examine the effects of significant
socio-economic changes during the pre- and post Apartheid period (before
and after 1994) on the urbanization of South Africa from 1985 to 2009.
Methods
References
Acknowledgment
This work is supported by a grant from the NASA Land
Cover and Land Use Change Program.
1985 2009 2000
2009-1985 2000-1985 2009-2000
Inanda Dam
Early construction
Inanda Dam
Inanda Phoenix
KwaMashu
Inanda Phoenix
KwaMashu
Pinetown
Chatsworth
Uhmalanga Gateway, largest
shopping mall on the African
continent. Built on a dump
that closed in the mid-1990.
The sixth largest electricity
consumer in the Durban
metro area.
Hillcrest
Mpumalanga
Cropland and forestry
changes
King Shaka Airport Hazelmere Dam
Umlazi Mpandwini
Nungwane
Dam
Sugar plantations
by Mpandwini
Illovo
Sugar plantations
by Illovo
Sugar plantations
by Illovo
Sugar plantations
by Mpandwini
Sugar plantations
by Mpandwini
Illovo Illovo Nungwane
Dam
Nungwane
Dam
Chatsworth Chatsworth
Umlazi Umlazi
Inanda Phoenix
KwaMashu
Durban
International
Airport
Durban
Intenational
Airport
Durban
International
Airport
Durban International
Airport losing traffic in
favor of Tambo
International Airport
in Johannesburg
Inanda Dam, Water supply
to Durban,
Completed in 1998,
Inanda Dam
Durban Durban
Durban
Abandoned
construction
at King
Shaka
Airport. Construction began
in 1973. The project
was halted in 1982
due to an economic
slowdown.
King Shaka Airport King Shaka Airport
Completion of King Shaka Airport
took place between 2000 and
2010), and now replaces Durban
Int. Airport.
Expansion of
Umlazi
township
Densification
of Umlazi
township
Pinetown Pinetown
Mpumalanga
Cropland and forestry
changes
Hillcrest
Construction boom in
1990’s and 2000’s
Hillcrest
Construction boom in
1990’s and 2000’s
Hillcrest
Verulam Verulam Verulam
Sugar cane
cultivation Sugar cane
cultivation
Sugar cane
cultivation
Densification of
Inland Durban
Densification of
Inland Durban
Mpumalanga
Cropland and forestry
changes
Results
RGB composites of Substrate, Vegetation & Dark fractions derived from Spectral Mixture Analysis for 3/23/1985, 3/25/2000, and 3/24/2009
RGB composites of change in fractions derived from Spectral Mixture Analysis for 2000-1985, 2009-2000, and 2009-1985. Color indicates increase in substrate, increase in
vegetation, and increase in dark fraction (water or shadow). Prominent changes include Increase in impervious+shadow, impounded water and Increase in fallow croplands.
Densification of
KwaMashu
W
We use the linear spectral mixture model (Adams et al. (1986, 1993) to
represent land cover as fractions of Substrate, Vegetation and Dark
surfaces. The Substrate endmember represents SWIR-bright rock, soil
and impervious surfaces. The Vegetation endmember represents
illuminated foliage. The Dark endmember represents water, shadow and
absorptive materials. We use a unity-constrained inversion of the 6x3
linear mixture model based on the Singular Value Decomposition given
by Boardman (1989, 1990). Calibration to exoatmospheric reflectance,
radiometric rectification, and use of generic global endmembers (Small,
2004) reduce, but do not completely eliminate spurious change., Modal
difference fractions of Substrate and Vegetation are 0.01 and 0.005
respectively, indicating that the calibration and rectification are
effective. The modal difference fraction for the Dark endmember is -
0.35; a result of actual changes in both atmospheric opacity and
turbidity in the coastal ocean.
RGB composites of the SVD fractions show the dominant endmember
combinations as additive colors. RGB composites of the fraction
differences show relative increases in SVD fractions. Relative
increases in one or two fractions may result from decreases in the
complementary fraction(s).
Adams, J. B., M. O. Smith, et al. (1993). Imaging Spectroscopy: Interpretation based on spectral mixture analysis. Remote Geochemical Analysis: Elemental and Mineralogical
Composition. C. M. Pieters and P. Englert. New York, Cambridge University Press: 145-166.
Adams, J. B., M. O. Smith, et al. (1986). "Spectral mixture modeling; A new analysis of rock and soil types at the Viking Lander 1 site." Journal of Geophysical Research 91: 8098-
8122.
Boardman, J. (1990). "Inversion of high spectral resolution data." SPIE - Imaging Spectroscopy of the Terrestrial Environment 1298: 222-233.
Boardman, J. W. (1989). Inversion of imaging spectrometry data using singular value decomposition. IGARSS'89 12th Canadian Symposium on Remote Sensing, Vancouver, B.C.
Small, C. (2004). The Landsat ETM+ spectral mixing space. Remote Sensing of Environment, 93, 1-17.
