In: Wagner W„ Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
Image ^ 
Name 
sor 
n Date 
IgfS" 
an 
d 
J ^of ul 
ution 
Range 
(um) 
Area 
(km 2 
) 
Topograp 
hie Map 
MLSL 
Lagos 
- 
1966 
- 
- 
- 
- 
Landuse 
Map 
NCRS, 
Jos 
1995 
- 
Land sat 
TM 
GLCF 
TM 
1986 
1- 
7 
30m 
0.45-0.90 
185 
xl85 
SPOT 
FORM 
ECU, 
Abuja 
MS 
s 
1995 
1- 
3 
10m 
0.45-0.69 
60x 
80 
NigeriaSa 
t-1 
NCRS, 
Jos 
I ma 
ger 
2007 
1- 
3 
32m 
0.52-0.90 
600x 
580 
3.2 Landuse Classes 
Based on the knowledge of the study area, reconnaissance 
survey and additional information from previous studies in area, 
a classification scheme was developed after Anderson et al., 
(1976). The scheme gives a broad classification where each of 
the land use/ land cover was identified by a class (Table 2). 
These classes are apriori well defined on the three images used 
for the study. 
Table. 2 Landuse classification scheme (after Anderson 
et al 1976) 
LANDUSE/LANDCOVE 
CATEGORIES 
DESCRIPTION OF THE 
LANDUSE/LANDCOVER 
Built-up Area 
Roads, buildings, open spaces 
Bare Rock 
Bare soil, bare land 
Farm Land 
Shrubs, fallow, cropped land. 
Secondary forest 
Agro forest, riparian forest, 
advanced bush re-growth 
Water Body 
Dam, rivers streams. 
3.3 LandCover/ Landuse Analysis 
For Landuse/Landcover analyses, the satellite images were 
classified using the supervised classification method. The 
combined processes of visual image interpretation of 
tones/colours, patterns, shape, size, and texture of the imageries 
and digital image processing were used to identify 
homogeneous groups of pixels, which represent various land 
use classes already defined. This process is commonly referred 
to as “training” sites because the spectral characteristics of 
those known areas are used to “train” the classification 
algorithm for eventual land use/ cover mapping of the 
remaining parts of the images. 
A Map of the study area was produced and was used to 
locate and identify features both on ground and on the image 
data. The geographical locations of the identified features on 
the ground were clearly defined. These were used as training 
samples for supervised classification of the remotely sensed 
images. The five categories of land uses/ land covers were 
clearly identified during ground truthing. Locations were 
tracked with the GPS to facilitate transference of the field 
information onto the images. 
3.4 Classification 
In this study, the satellite images were classified using 
supervised classification method. The combined process of 
visual image interpretation of tones/colours, patterns, shape, 
size, and texture of the imageries and digital image processing 
were used to identify homogeneous groups of pixels, which 
represent various land use classes of interest. The study 
engaged in ground truthing to the four Local Government Area 
of the study area. These are Ekiti west, Ado-Ekiti, Irepodun/ 
Ifelodun and Ekiti south-west Local government areas in Ekiti 
State (Figure 1). Before the ground truthing, map of the study 
area was printed and was used as guide to locate and identify 
features both on ground and on the image data. The 
geographical locations of the identified features on the ground 
were clearly defined. These were used as training samples for 
supervised classification of the remotely sensed images. Five 
categories of land uses and land covers were clearly identified 
during ground truthing. These are secondary re-growth forest, 
water body, bare rocks, built-up areas and farm land. The 
processed images were subject to band correlation analysis to 
assess the nature and strength of the relationship among the 
bands in the imageries. 
4 RESULTS 
4.1 Comparison of Basic Features among the Three 
Sensor Data 
Table 3 summarizes the correlation analysis of bands with 
each other within each of the three sensors. In the NigeriaSat-1 
image, the Near-Infrared (NIR) band was negatively correlated 
with the visible bands (Green and Red) (-0.16, < r > -0.04; p < 
0.05). In the Landsat TM image, the NIR band positively 
correlated with visible bands (0.02 < r > 0.22; p < 0.05). For 
the SPOT image, the NIR band also positively correlated with 
the visible bands (0.53 < r > 0.63; p < 0.05). The relationship 
between the visible bands were strongest in SPOT (r = 0.98), 
relatively strong in NigeriaSat-1 images (r = 0.53) and 
relatively low in Landsat TM (r = 0.22). 
Table. 3 Correlation matrix analysis results for the three sens 
data 
Sensor 
Bands 
Green 
Red 
NIR 
Landsat 
Green 
1.00 
0.22 
0.02 
Red 
0.22 
1.00 
0.93 
NIR 
0.02 
0.22 
1.00 
NigeriaSat-l 
Green 
1.00 
0.95 
-0.04 
Red 
0.95 
1.00 
-0.16 
NIR 
-0.04 
-0.16 
1.00 
SPOT 
Green 
1.00 
0.98 
0.63 
Red 
0.53 
1.00 
0.98 
NIR 
0.63 
0.53 
1.00 
Level of significance (p) <0.05 
The results imply that the SPOT image is likely preferable 
to either of the other image types for the study of earth base 
features at the Visible and Near Infrared portions of the 
Electromagnetic Spectrum. On the other hand, NigeriaSat-1 
imageries could give better information at the visible portion 
while Landsat imageries could be better in the Visible and 
Near Infrared portions of the spectrum. The results indicate that 
the strength of the correlation among the bands increases with 
increase in the spectral resolution of the imageries. This 
corresponds with what many authors have observed. For 
example, Kuplich et al. (2000) have suggested based on their 
studies, that high correlation between spectral bands is 
indicative of high degree of information. Spectrally adjacent 
bands in a multispectral remotely sensed image are often highly 
correlated. Multiband visible/near-infrared images of landuse 
areas will show negative correlations between the near-infrared