1tial
nes,
land
ands
ve
eans
e
| to
'ch
.assi-
1e
entory
1
le
opriate
classi-
ore
al
pe of
t has
humid
ban
cially
ose
ather
nclu-
lage
e of
station
lenti-
scslhe
that
|tera-
lopt
opical
MAL A n
«c
The land uses present in the classification are characteristic of a coastal
tropical city having some 125,000 inhabitants and covering approximately 200
square kilometres. Calabar city (Figure 1) has a wide diversity of industry
including the manufacture of cement and metal sheets, as well as the processing
for export of timber, rubber and palm oil. The new port complex, completed in
1979, provides a stimulus for growth as a centre for industry and commerce to-
gether with, as yet peripheral but important, developments in finance and
tourism.
MEASUREMENT AND ANALYSIS METHODOLOGY
The most precise method of recording and measuring land use information
from aerial photographs is to compile interpretation overlays, to transfer the
land use boundaries from these to maps, and then to measure individual cate-
gory areas. This technique is extremely time-consuming and requires trained
personnel as well as sophisticated equipment such as electric digitizers. If
land use information is to be obtained of extensive areas, or speedily, then
such a total survey must be abandoned in favour of some method of sampling.
It has been shown (Emmott and Collins, 1980) that the use of systematic
sampling, directly off air photographs, provides a simple method for the
acquisition of land use data. This system was tested for applications to the
survey of land use in Calabar.
The classification scheme was used in the compilation of land use air
photo interpretation overlays for the study area. Seven of these overlays were
selected as samples representing a cross-section of urban land use types, each
sample having a dominant land use. The category areas were measured for each
sample by three methods:
(a) by digitizing the boundaries of each land use parcel, computing parcel
areas from these co-ordinates and compiling category total areas,
(b) by recording the dominant land use in each cell of a range of square
grids placed upon the overlays,
(c) by recording the actual land use at each point on a range of ortho-
gonal dot grids placed over each overlay.
RESULTS
The discrete boundary areas obtained by digitization were accepted as
correct and these were compared with the values obtained from the various den-
sities of grid square and dot grid samples.
Multiple correlation and regression analyses were used in a mathematical
model which related density of observation, category size (as a percentage of
sample area) and accuracy (as a percentage of category area).
Figures 2 and 3 illustrate the relationship between percentage error and
density of observations when the category area is a specific proportion of the
sample area. For both grid squares and dot grids there is an increase in
accuracy with increase in density of observations and with increase in category
percentage area. Where the category percentage area is less than 5% then at all
densities of observation the grid of squares yields slightly higher accuracies.
However where the category percentage area is5 percent or above then at all den-
sities of observation dot grids yield marginally higher accuracies.
Figures 4 and 5 show, for specified grid densities, the relationship
827
-— UA
da al WEE COMME, te - N le
: UND HR MR ———— = - n
iiM n: * = es ee A