PPS Se ha So 8 SO 2 
MAPPING SEMI-ARID VEGETATION IN NORTHERN KENYA 
FROM LANDSAT DIGITAL DATA 
  
  
G.H. GRIFFITHS and W.G. COLLINS, 
Remote Sensing Unit, 
Civil Engineering Department, 
University of Aston, 
Birmingham B4 TET, 
ENGLAND. 
ABSTRACT 
A 30 X 50 km test area of semi-arid vegetation in Northern 
Kenya was selected for an investigation into the relationship 
between vegetation spectral response and LANDSAT radiance values. 
Digital radiance values for six mapped vegetation classes ‘were 
read from a dry-season LANDSAT CCT tape using an Interactive 
Image Analyser. A scattergram of MSS band 5 (multi-spectral 
scanner) values against band 7 values showed that increasing 
vegetation canopy cover was strongly correlated with decreasing 
radiance values. This paper attempts to demonstrate that the main 
problem in mapping semi-arid vegetation within this region is to 
determine the extent to which vegetaton of varying canopy cover 
and species composition modifies a uniform soil response. Two 
interpretations of the data are proposed and it is tentatively 
suggested that variations in the percentage of vegetation canopy 
cover result in changes in the intensity of reflectance, whilst 
variations in species composition are accompanied by changes in 
Spectral content. 
  
  
1. INTRODUCTION 
Figure 1 shows the position of the study area which is 
Situated to the east of Lake Turkana in Marsabit District. A test 
area of approximately 30 x 30 km (512 x 512 LANDSAT pixels) 
Surrounding the small Rendille settlement of Kargi (Long. 37° 35! 
E., Lat. 2° 30'N), was selected for detailed investigation. A 
comprehensive description of the environmental characteristics 
of this area can be found in an earlier paper (Griffiths and 
Collins, 1981), but a brief account is given below. 
A low and unreliable rainfall («1500mm/annum) combined with 
very high evapotranspiration rates, supports only a sparse cover 
of Acacia shrub, with Indigofera dwarf shrub and Aristida 
grassland species. The vegetation cover is susceptible to 
overgrazing, regeneration being slow under the unfavourable 
climatic conditions found in Northern Kenya. A generally uniform 
soil type across flat relief and a vegetation cover of structural 
simplicity and low species diversity, suggested that this would 
be a suitable area in which to test the use of LANDSAT data for 
mapping and monitoring semi-arid vegetation. 
2. METHODOLOGY 
An earlier paper (Griffiths and Collins,1981) showed that 
only two vegetation types covering a small part of the study area 
could be accurately mapped from the 1:250 000 scale standard NASA 
false colour composite imagery. Dense thickets of A. reficiens 
and A. senegal shrubland were clearly visible along the bottom of 
381 
; TEE RARE LE aa DELE 
nad gerere is men a iie pite terms ripa n mieten erre be ie i el Sl fel :