É suit 
40 so 
  
1982 CMS. 
easing popu- 
‘ace of 
tive approaches 
-otecting 
d by the 
1s well as 
jes from sea 
:limatic 
and consti- 
| former type 
ar montane are 
ine forests 
The dis- 
/ processes 
Inchecked 
f water 
iquatic eco- 
tural lands 
je Weekly 
above in the 
iated with 
TY. 
ry to apply 
js not'uns- 
rts of Europe 
uming tech- 
ons such as 
manage- 
and coupled 
verage, 
alternative 
t management 
Hi 
Em | 
E UN 
Bl gg en te a arr re LUNDI a a ee ee ee Up eno quinte e 
  
  
  
of such areas as Kenya's forests. Since 1972 the LANDSAT system has made data 
avai lable that have the potential to be used for the purpose of forest manage- 
ment and monitoring of forest cover changes in such regions as the Mau Forest 
(Miller et al.,1978). 
The objectives of this study are to demonstrate that: 1) Excision in 
forested areas are discernible with LANDSAT data; 2) LANDSAT data can be used to 
discriminate landcover types generally associated with land use of forest change; 
3) forest change, using temporal LANDSAT data, is discernible on a regional basis 
and 4) the Earth Resources Data Analysis System (ERDAS) 400 is a compact, viable 
data processing system amenable to limited training and budgets of developing 
nations for natural resources management in a cost effective and timely fashion. 
METHODOLOGY 
A. Study Site Selection: The study sites include the area between Lake 
Nakuru and the town of Lumbwa (See Figure 2). This geographic region is still 
heavily forested but plagued by some of the problems discussed earlier. One 
prime reason for the selection of this particular area for this study is directly 
related to the fact that one of the authors is intimately familiar with this part 
of the world. Additionally, air photos of scattered parts of the study sites 
were made available to serve as ground truth. However, the photos did not well 
correlate with the LANDSAT data since the former did not represent one specific 
season or even one specific year. The related problems will be discussed later. 
B. Digital Analysis Procedures: Temporal LANDSAT data for the study site 
was obtained for January 3, 1973 and December 31, 1978 (Path 181, Row 60). 
These particular data sets were deemed suitable because they were virtually 
cloud free, showed good vegetation vigor, coincided seasonally and the six year 
interval was accepted by the authors as appropriate to adequately demonstrate 
encroachment on forested lands. The data interval was selected on the basis of 
readily available data and experience by the authors of monitoring revegetation 
although in a very different geographic setting, namely land reclamation progress 
in surface mine areas of east central Ohio (Bloemer et al., 1981; Witt et al., 
1982; Brumfield et al., 1981). 
The areas of the previously described study sites were subset from both data 
tapes (Mau West and Mau East) and reflect comparable areal extents. These sub- 
scenes were transferred to floppy discs for digital data processing on the ERDAS 
400. The ERDAS 400 is a complete, self-contained image processing and Geographic 
Information microcomputer system based on the Z80 microprocessor. The hardware 
of this system has dual eight-inch floppy drives, a high resolution color monitor 
(512 by 480), and a matrix printer for hard copy output. The software package 
consists of a basic operating system, a digital image processing sub-system 
integrated with a geographic information system based IMGRID (ERDAS 400, 1982). 
The programs are interactively menu driven to provide a relatively user friendly 
environment. The digital data processing capabilities of the ERDAS ^00 provide 
both supervised and unsupervised classification procedures. Due to lack of 
timely ground truth, this study necessitated the application of a cluster algorithm 
based unsupervised classification approach.  (Bloemer et al., 1981) The cluster 
algorithm was set to develop twenty-seven (27) spectral groups in the data. These 
groups were correlated with land cover information and finally combined into three 
land cover categories, namely forested, non-forested, and water. 
C. Procedure for Monitoring Forest Change: The prime purpose of this study 
is to determine the change of forest cover for the selected sites. For each of 
the temporal data sets pixel counts were established for the respective land 
cover categories and converted to acreages (pixel x 1.1). These acreages were 
then compared for each land cover category for the two dates. The geographic 
323