Use of multitemporal satellite data for vegetation change detection
in Namibia
H. Wagenseil & C. Samimi
Institute for Geography, Friedrich-Alexander-University Erlangen-Nuremberg, Germany
hwagense@geographie.uni-erlangen.de
Keywords: time series analysis, change detection, rainfall variability, NDVI, Namibia
ABSTRACT: As water supply strongly influences plant growth in semiarid areas, precipitation
events in their spatio-temporal variability are reflected in NDVI time series (AVHRR, MODIS)
of individual rainy seasons and complicate a phenological delimitation of vegetation units.
To overcome this problem and to develop a “near-to-real-time”-vegetation monitoring, relation-
ships between NDVI and previous rainfall events are investigated, as it is assumed, that the dif-
ferent vegetation units (e.g. grass savanna, tree savanna) show specific reactions on water avail-
ability. Following a decision tree approach, homogenous rainfall-vegetation classes are separated
from seasonal AVHRR-NDVI data and correlations to rainfall data including recent and past
multi-day-sums are computed for each class. A supervised classification and a change map from
two sets of Landsat data are used for final validation and class labeling.
1 BACKGROUND AND OBJECTIVE
Signs of degradation in semi-arid regions of the world have long been the object of geographical
research. The strong susceptibility of these areas to land degradation is attributed to the high vari-
ability of the environment so that even insignificant changes in soil and climatic determinants cause
marked changes in the density and composition of vegetation (Kempf 1996). An excessive increase
in certain woody plants like Acacia mellifera and Dichrostachys cinerea can thus be observed in
particular in rangeland but also in national parks as a result of overgrazing, or selective grazing and
browsing. This process is termed bush encroachment and has already been recognized as a problem
by Walter (1964), Strang (1973), and others. Moreover, vegetation cover and composition is a lead-
ing and dominant factor for soil erosion, which in turn is the second relevant process in landscape
degradation (Beugler-Bell 1996).
As signs of degradation are indicated by the spatial composition of vegetation or conversely
changes in plant cover could indicate future degradation processes, a continuous and “Near-to-
Real-Time” monitoring of vegetation based on time series data of the Normalized Difference Vege-
tation Index (NDVI) can therefore serve to detect such developments and could allow appropriate
counter-measures to be adopted. Satellite images have routinely been used for such large-scale,
quick, and cost effective analyses of vegetation.
Vegetation studies of semi-arid areas have shown that restricted water supply is a limiting factor
and strongly influences plant growth. Therefore precipitation events are reflected in the NDVI de-
velopment in its variability over time and space (Nicholson & Farrar 1994, Grist et al. 1997, Rich-
ard & Poccard 1998, Eiden 2000) and complicate the phenological delimitation of vegetation units
183
New Strategies for European Remote Sensing, Oluiþ (ed.)
© 2005 Millpress, Rotterdam, ISBN 90 5966 003 X