International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
  
used within KNP for this purpose. A rich remote sensing (and were then masked out, before simple land cover classification A 
associated geographical information system) data resource is was performed (Figure 3b) to gain an understanding of the re 
; available for the park, including images from Landsat, SPOT, spectral separability of classes, particularly vegetation types. J 
: IKONOS and QuickBird instruments, and numerous airborne Sometimes, though, fine spatial resolution imagery can be too 
images and photographs (Aplin 2003b). detailed for accurate classification analysis, particularly in very 
mixed semi-natural environments such as KNP. For instance, 
The strengths of fine spatial resolution satellite sensor imagery where pixel size (e.g, 2.8 m spatial resolution QuickBird 
for ecological studies have been described above. In simple imagery) is markedly smaller than features of interest (e.g., 10 
terms, fine spatial resolution imagery such as QuickBird m tree canopies), classification algorithms may erroneously 
enables more detailed observation than coarser spatial classify within-feature variation (Aplin et al. 1999). The key 
resolution imagery such as Landsat ETM+. For instance, while consideration here is to select classes with regard to the 
individual trees and even airport runway markings can be discriminatory capabilities of the spatial resolution of the data, 
identified using a 2.8 m spatial resolution QuickBird image, the and/or vice versa. 
runway and service road can barely be discerned using a 30 m 
spatial resolution Landsat ETM+ image (Figure 2). In A technique that makes full use of the detail present in fine 
  
  
ecological terms, QuickBird provides the benefit of observing at spatial resolution imagery (sometimes used in combination 
the individual plant level, while Landsat ETM+ enables only with, or as an input to, classification) is texture (Figure 3c). 
aggregated measurements of vegetation (or other land cover) Texture measures, of which there are many, exploit the spatial 
patches. variation between pixels, enabling features or areas to be 
characterised according to how homogeneous or heterogeneous 
they are. For instance, while tree canopies or parts of tree 
canopies may be misclassified using land cover classification, 
they may have characteristic (smooth or rough) textures, 
enabling accurate identification. 
Given the difficulties of allocating pixels to specific land cover 
classes in mixed environments like KNP, it can be useful to use 
continuous rather than thematic representations. A common 
means of characterising the presence of vegetation, for instance, 
is through NDVI analysis (Figure 3d). This has the benefit that 
pixels are not forced into (sometimes inappropriate) classes, but 
instead provide a proportional measure of vegetation. Further, 
analyses such as NDVI can contribute to the derivation of 
biophysical variables such as LAI and NPP. 
Finally, habitat monitoring can be performed by comparing 
analyses such as land cover classification and NDVI over time. 
Aplin (2003b) provides an example of vegetation change 
analysis in KNP conducted using Landsat ETM+ imagery. 
  
Research remains ongoing to exploit fine spatial resolution 
(b) satellite sensor imagery, in combination with coarser spatial 
Figure 2. Skukuza airport runway, Kruger National Park, resolution imagery, for ecological investigation in KNP. In 
showing the finer feature delineation provided by (a) 2.8 m particular, vegetation information derived from remote sensing 
spatial resolution QuickBird imagery compared to (b) 30 m is integrated with animal population data to further our 
spatial resolution Landsat Enhanced Thematic Mapper Plus understanding of the dynamic relationship between the two. 
imagery. (North is to the top of the page and the image extends 
approximately 1 km east-west.) REFERENCES 
Andréfouet, S., Kramer, P., Torres-Pulliza, D., Joyce, K.E., 
Hochberg, E.J., Garza-Pérez, R., Mumby, P. Riegl, B. 
Yamano, H., White, W.H., Zubia, M., Brock, J.C., Phinn, S.R., 
Naseer, A., Hatcher, B.G. and Muller-Karger, F.E., 2003. 
Multi-site evaluation of IKONOS data for classification of 
tropical coral reef environments. Remote Sensing of 
Environment, 88, pp. 128-143. 
In KNP, fine spatial resolution satellite sensor imagery has 
application for each of the three main areas of ecological 
remote sensing outlined by Kerr and Ostrovsky (2003): land 
cover classification, integrated ecosystem measurement and 
change detection. Both IKONOS and QuickBird imagery are 
available for analysis, although these data are usefully 
combined with coarser spatial resolution imagery to extend 
analysis over relatively large areas. For instance, spatially 
detailed QuickBird data can be linked to, and extrapolated over 
the wider areal coverage of, Landsat ETM- data. 
Aplin, P., 2003a. Remote sensing: base mapping. Progress in 
Physical Geography, 27, pp. 275-283. 
Aplin, P., 2003b, Vegetation monitoring in southern Africa: a 
consideration of natural and anthropogenic factors, 
Proceedings, 30th International Symposium on Remote Sensing 
of Environment: Information for Risk Management and 
Sustainable Development, Honolulu, 10-14 November 2003, 
International Center for Remote Sensing of Environment, CD. 
Examples of fine spatial resolution image analysis are provided 
in Figure 3. Specifically, a 2.8 m spatial resolution multispectral 
QuickBird image, acquired on 15 November 2002, was 
atmospherically corrected using dark object subtraction and 
geometrically registered to the local map projection system. 
(Figure 3a). Cloud and cloud shadow, prevalent in the image,