The four reserves used for the research were selected based on 
similar characteristics which are fundamental for comparative 
research studies. The similarities of the four reserves used in the 
pilot study are as follows: 
e Firstly, the size of each of the conservation areas are + 
300km? classing them all into the medium sized 
category for conservation areas (250-500 km”). 
° The second similarity is the elephant population 
numbers ranging from 100+ and with proper elephant 
management plans in place these numbers are set to 
stabilize and remain within the prescribed parameters 
set out by the National Norms and Standards policy 
(DEAT, 2008). This relates to an average elephant 
density of 0.39 elephants/ha. 
e The third similarity is the duration of elephant 
occupation. This is determined by the date of relocation 
of the elephants to each individual reserve and for 
Tembe, the date of enclosure from Mozambique. All the 
reserves have thus been exposed to the impact of 
elephant forage for an extended and continuous period 
of 15 years or more. 
Comparing the impact of elephants are further extended by the 
fact that all four the conservation areas have basically similar 
vegetation types. There are also some tremendous differences 
in that Marakele and Ithala are both extremely mountainous 
with deep valleys and steep cliffs (figures 4&5). Tembe is 
however the opposite, in that it is very flat with only a slight 
undulations. Welgevonden is also in the mountains but does not 
have the steep cliffs and deep valleys and is situated on the 
higher more central region of the Waterberg. 
3 ANALYSIS 
3.1 Rationale 
“Most conservation practitioners rely on experience-based 
information for management decisions. In South Africa, 
government policy thus directs managers to base decisions for 
elephant management on the best available scientific 
information”. (Young & Van Aarde 2011, p 876). 
The research is rooted in the requirement of these smaller 
reserves to acquire much needed information on a regular basis 
over the whole extent of the reserve. Remote sensing and the 
use of satellite images are therefore ideal for this endeavour 
except for the fact that the reserve managers usually do not have 
the expertise to do the satellite image analysis and to use GIS 
for further enhanced spatial analysis. 
The first stage and pilot project of the research had the objective 
of illustrating to the reserve managers and conservation decision 
makers, the potential of satellite imagery and to demonstrate the 
advantages (and limitations) of imagery as a source of 
management information. These included the capabilities 
related to the temporal, spectral and to a smaller extend the 
spatial resolutions associated with universally available satellite 
imagery. 
3.2 Method 
For the pilot phase of this research Landsat 5 and 7, SPOT 4, 
Quickbird and where possible Sumbandilesat images for each 
reserve were collected. Two of the images were from the date 
when the elephants were first relocated (for Tembe when first 
fenced in). The one image was selected to be from the summer 
(rainfall season) and one from winter (dry season). 
To start the research spectral scans were done on those 
vegetation types identified previously as favourite forage for 
    
   
   
     
   
    
   
   
      
  
elephants. A hand-held multispectral radiometer was used 
(figure 8) to scan individually selected samples of each of the 
tree species. This multispectral radiometer was specifically 
designed and calibrated to scan spectral bands that correspond 
exactly with those of Landsat and SPOT as indicated in Table 7. 
Table 7: Wavelengths used to correspond with TM from 
Landsat and SPOT 
WAVELENGTHS SELECTED 
  
  
Green |Green [Yellow | Red | Red Near NIR NIR 
(low) | (low) (low) |(high) | Infrared |(middle) |(high) 
(low) 
530nm |560nm |580nm | 630 p70nm| 800nm |840 nm B90 nm 
nm 
  
  
  
  
  
  
  
  
  
The scanning was done to resemble a passage from Landsat in 
that it was done at a time similar to when Landsat would have 
passed and from a direction corresponding to that of a Landsat 
orbit angle. 
Figure 8: Scanning Dichrostachys cinerea with the 
multispectral radiometer. (Source M Jordaan, 2011) 
4 RESULTS 
NDVI was run on all the multispectral images for all four the 
reserves. Using Marakele to illustrate the process, two images 
taken over summer ten years apart are shown in Figure 9. Note 
that the areas with higher values (more biomass) in the later 
image are predominantly within the lower lying Marakele 
National Park. Interestingly this correlates exactly with the area 
classified as Western Sandy Bushveld vegetation type (Mucina 
& Rutherford, 2006). 
  
i NDVt Welgevonden February 1995 
Dl Welgevonden February 2005 
    
   
  
  
  
Figure 9: NDVI maps for 1995 and 2005. 
The comparison of the two images were done using a change 
detection process where the pixel values of the newer image 
(after elephant impact image = 2005) is compared to the 
     
    
corresp( 
image = 
the ND' 
25% the 
red (if tl 
A simil 
spectral 
3 and 4 
detectec 
upward: 
10 sho: 
was fou 
Figure 
pixel) ^ 
when t 
compar 
| |= 
  
  
Figur 
Marake 
excepti 
indicat 
Welges 
darker 
uniforn 
elephai 
lighter 
area ra 
have t 
relativ: 
NDVI 
which 
areas 
determ 
Wesse 
remote 
is val 
beginn 
approa 
use of 
of de 
provid 
et al, , 
Howe 
inconc 
that tl 
year |