ometer was used
les of each of the
' was specifically
ls that correspond
dicated in Table 7.
d with TM from
>
iD
;
NIR | NR
(middle) |(high)
840 nm R90 nm
e from Landsat in
indsat would have
) that of a Landsat
rea with the
rdaan, 2011)
es for all four the
Ocess, two images
in Figure 9. Note
mass) in the later
r lying Marakele
actly with the area
tion type (Mucina
vonden February 2005
d 2005.
ne using a change
' the newer image
compared to the
corresponding pixels in the old image (before elephant impact
image = 1995). The comparison is based on a 25% change in
the NDVI value per pixel and if the was a change of more than
25% the relevant pixel is green (if there was an increase) and
red (if there was a decrease).
À similar process was done on the digital numbers of the
spectral bands of the satellite images directly (Landsat bands 2,
3 and 4 were used) and again if a change of more than 25% was
detected per corresponding pixel, it was coloured green for an
upwards increase in value and red if the was a decrease. Figure
10 shows the changes detected using Landsat band 4 as NIR
was found to be the most predominant indicator of vegetation.
Figure 10 illustrates the changes detected (25% or more per
pixel) when the two NDVI images were compared and also
when the two satellite images using spectral band 4 were
compared.
Change 1995-2005 using NDVI
(Change 1995-2005 using spectral signatures
Legend Scale
Kilormeters
Class_Names
Fa Decreased
Increased
Figure 10: Changes by NDVI versus Changes by reflectance
value
Marakele National Park does not show much change with
exception to some uniform black areas (figure 9) that may
indicate fire events rather than a decrease in vegetation.
Welgevonden, Tembe and Ithala also indicated patches of
darker area but these are not likely to be fire related as it is not
uniform. These darker areas are suspected to be caused by
elephant browsing. Overall the later image (figure 9) appears
lighter which would indicate an increase in vegetation over the
area rather than a decrease. Rainfall data indicated both years to
have been relatively dry but the preceding year has been a
relatively wet year for Welgevonden. As commonly understood
NDVI is used indicator of biomass and of vegetation production
which is directly linked to rainfall, so the rainfall figures for the
areas surrounding the four reserves were scrutinized to
determine possible NDVI increases.
5 DISCUSSION
Wessels et al. (1999) indicates that the relationship between
remotely sensed measures of landscape and biological diversity
15 valuable in terms of biodiversity assessments and is
beginning to be used as an important tool in integrated
approaches to biodiversity assessment and conservation. The
use of remote sensing has been found to be a successful method
of determining vegetation change over time while also
providing reliable information on vegetation composition (Liu
et dl, , 2008, Brink & Eva, 2008 and 2005 Akasheh et al.).
However the results from this study initially appear
inconclusive. The change detection image generated suggests
that the vegetation in Welgevonden has decreased over a ten
year period, but there has been little change in Marakele.
Change detection for Tembe indicates changes in the Coastal
bushveld and shrubveld but very little in the threatened ‘Sand
Forest’. Ithala however has had severe changes but this can be
contributed to their dedicated fire regime (Bigwood, T. personal
communication, December 04, 2010)
A similar study done by Guldemond & Van Aarde (2007) who
conducted a study looking at what they classified ‘Woodlands’
(basically ‘Tembe Sandy Bushveld’ and ‘Sand Forest’)
comparing these woodlands inside ‘Tembe Elephant Park’ and
at intact vegetation communities outside the Park. Three
representative sampling sites were selected inside and outside
the Park where quadrants (16*16 m) were used to record the
number, species and structure of vegetation. Their study
concluded that the elephants in Tembe Elephant Park fed on
seven main species (Acacia burkei, Afzelia quanzensis, Albizia
adianthifolia, | Dialium schlechteri, Maniklara discolour,
Sapium integerrimum and Spirostachys africanum). These
species occur at low numbers within the Sand Forest and are
more common in ‘Tembe Sandy Bushveld’. Using remote
sensing alone as instrument to determine elephant impact can
also be misleading in that at Tembe the elephant do not impact
severely on the endangered Sand Forest but they do create
openings for the smaller antelopes such as Nyala (Nyala angasi)
and Kudu (Tragelaphus strepsiceros). (Mathews, W. personal
communication, April 2009). These antelopes then destroy the
undergrowth and lower branches but not the larger trees and tree
canopy. Therefore there is little immediate change to the
vegetation but in the long run there is no succession as al the
seedlings have been foraged by the antelopes.
This study indicates (1) the value of NDVI variability as an
indicator of landscape heterogeneity and biological diversity,
and (2) the potential of integrating vegetation data from
vegetation maps with the NDVI information to indicate
landscape patterns of biological diversity.
At an annual increase of 7% the populations in the four reserves
would have a density of +2.15 elephants per km?. This would
lead to huge losses in biomass and likely diversity as well, as
shown in Sengwa Wildlife Research Area where there was a
reduction of 45% in biomass of woodland species over a five
year period (Cumming 1981) when the elephant population
density was at a similar level to what the population could be in
the four reserves by 2020, if the elephant population grows at
7% per annum.
Management of the four reserves used in the study, in particular
but all reserves conserving elephants, will need to decide
whether they want to manage according to biodiversity or
whether they want to manage to the elephants needs. Various
management actions could include culling, contraception of
female elephants or possible vasectomies of elephant bulls (this
has been done successfully at Welgevonden Private Nature
Reserve).
6 REFERENCES
Akasheh, O.Z., Neale, C.M.U. and Jayanthi, H., 2008. Detailed
Mapping of Riparian Vegetation in the Middle Rio Grande
River using High Resolution Multi-Spectral Airbourne Remote
Sensing. Journal of Arid Environments, 72, 1734-1744.
Bigwood, T. 2010. Personal communication, 04 December
2010,