ul 2004
————
Iped
tudy
Data
aphs
was
field
field
the
] use
erton
en in
erose
wtch
erous
area.
the
tudy.
t the
fence
hows
Rift
id the
ch is
|-rich
E (he
s and
'es Or
engai
tream
many
parts of the image. The bright yellowish areas have
high reflectance in all wave-lengths. These are areas of
open pasture and grassland, bare soil, overgrazed, just
harvested, roads or built-up areas. Patches of dark red
represent trees or scrub vegetation. In agricultural
lands, the bright tones are an indication of cleared or
fallow fields at this time of the year in January. The
greenish tones are ploughed fields. Water bodies like
lake Nakuru, Lake Elmentaita, Lake Bogoria and
Sasumua dam are indicated in black tones. Lake
Baringo appears as bright green because of high
sedimentation. At the middle of Lake Nakuru, the
bright red indicates aquatic vegetation. Woodlands in
the image appear as red mottled tones.
Landsat TM (July 1989)
Land use cover change was detected on the image
after analyzing it and comparing the results with the
photograph of 1969. The colour key shows a
significant change of land use from the dominant
forest and large scale farms seen on the photograph. It
can be seen from figure 8 that the forest which was at
the confluence of the little shuru and enjoro has been
cleared. The blue patches within the forest indicate
new-clear lands now under subsistence cultivation.
Configuration and Composition of Land cover/Land
Use
It was ascertained through remote sensing and time
series analysis that there has been a significant land
use cover change, particularly the conversion of forest
and woodlands to arable land as shown in table 2 and
figures 1 and 2. The air photographs of 1969 showed
large scale farming, forest conservation and little built-
up lands. The pre-change geo-topes of planted forests
and farms showed clear corridors, sizes, shapes and
connectivity. The dominant land use cover types were
forests (75%), woodlands (12%) and farms (13%). By
140
= 120
=
= 100
S
= 8
S
z 60
=
5 40
Zia |
0 Lm
Figure 1: Change in woodland
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
1989, the landscape had changed tremendously as
showed by the landsat sub-scene image which showed
60 % forest and woodland and 40 % agriculture and
built-up area. The dominance of mixed agriculture and
subsistence farming in much of the former Nessuiet
forestlands and Ngondu large scale farm is shown by
the different shades of blue and whitish patches. Due
to decrease of spatial resolution on the images, the
dominant land use cover types of forest and mixed
farming exhibit a rapid increase and overestimation at
the coarse levels of spatial aggregation at the expense
of non-dominant land use cover types like rural built-
up lands. The dominant land use cover types are
overestimated due to spatial mixing of covers hence
the high variances. Settlements like Ngondu and
Njokerio now stand on former large scale farms and
grasslands. Ngondu formerly known as Wright farm
produced wheat and dairy products while Njokerio
formarly Sebiens farm for horses and flower
production. Forest lands in Beeston area have been
converted into arable and built-up lands. Expansion
into forest and riverine areas has been taking place for
quite a while now. From ground survey of 2003, the
dominant land cover types were 5% forest, 7%
woodland, 82% agriculture and 6% built-up area. It
was evident that loss of forest cover in the reserves
was by both clear cut and progressive thinning due to
poaching by local residents. There are many small
land use patches of maize, beans, potatoes, wheat,
pasture and homesteads with spectrally similar land
cover which tend to clump to the bigger entities
contaminated by mixed pixels in remotely sensed data.
The EU-LUCAES method used for this study provided
a simple calibration which has been validated to assist
with interpreting images to determine land use cover
types of heterogeneous type, especially subsistence
agriculture and rural built-up land on former forests
Woodland per segment/Imagette
and large scale farms.
NS Q
qe Q9 RTS SAN à > 0 oy
Segments
22003 mM 1989 [11969
407