quently
ploited
as wind
sweet .
y by the
and falls
'asonally
; may be
t periods
of flood
nificance
(t season
g are not
ultivated
flooded,
a tudu).
ection of
retention
ocultures
ows only
vn
Farm sizes are usually small (less than one hectare), and typically include a mix of both
fadama and tudu. (Goddard et al., 1971; Goddard, 1972). The viability of these farms is further
compromised by the severe fragmentation which is the norm. This is largely attributable to the
Islamic law of inheritance, and the threat which this represents to the future viability of farm
operations is exacerbated by the continuing high rates of population growth.
PRELIMINARY RESULTS
The project involves mapping the extent and character of wet and dry season land use/cover,
using recent Landsat imagery. In addition, a comparison of land use/cover before and after
construction of the irrigation project is being undertaken. In this section of the paper, prelimi-
nary results of land use/cover mapping for the dry season using December 1984 imagery, are pre-
sented. Results of change detection studies for the dry season, using 1984 and 1975 data, are also
reported.
METHODOLOGY
The procedures used in this study may be summarized as follows:
1. Preparation of unsupervised classifications using December, 1984, Landsat data.
2. Field studies to check unsupervised classifications and develop suitable categories for super-
vised classification,
3. Preparation of supervised ce fa tions using 1984 and 1975 Landsat data,
4. Comparison of 1984 and 1975 using both image enhancement and classification pine
5. Field checking of results and comparison with other information on the study ren. EN UT
Prior to fieldwork, unsupervised classifications were produced using the Dipix ARIES II
image analysis system. The number of classes generated was based on the statistical characteristics
of the Landsat data and general knowledge of the different land use/cover types that occur in
the study area. Original images and the result of classifications for each study site were produced
as 8 x 10 inch colour prints and transparencies. ;
Field studies involved comparing the images and classifications with ground cover at selected
points in each study site. in this way, an urderstanding of the capabilities of the Landsat imagery
for displaying different land covers and land uses in the area was developed. Field interviews with
farmers were also carried out.
The third stage was the development of a classification system and the production of super-
vised classifications for each study site. The development of a classification system ( Table 3) was
based upon a number of considerations, including the complex ity of the local agricultural system
and the scale at which final mapsare to be prepared. Since most loca! farmers practice inter-crop
ping (Norman, 1975), it is not possible to identify individua! crops from Landsat data, except in
the case of those crops planted as monocultures. This constrain is acceptable since the change
information needed for the study does not require identification of individual crops.