. Istanbul 2004 
ected from the 
alyzed for soil 
ym), available 
ind soil texture 
thodology for 
rs is given in 
ational Pipette 
suming process 
alyzed. 
ical parameters 
rence 
ley and 
(1934) 
iah and 
(1956) 
net al. 
54) 
r et al. 
65) 
  
ilculated from 
| numbers into 
related indices 
HI), Saturation 
ndex (RI). The 
table 3. Apart 
PC2 and PC3) 
nalysis (PCA). 
reduction of 
cations and for 
| from a large 
. hyperspectral 
from Mathieu 
  
Property 
ze reflectance 
ude 
a Slope 
  
y colours 
olour 
  
ite Content 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
parameters, where correlation was significant. These empirical 
equations were used to generate soil fertility variability maps 
from RS data. 
4. RESULTS AND DISCUSSION 
The IKONOS multispectral data was used to compute 
radiometric indices and the three principal components. The 
principal component analysis for the study field showed that the 
first principal component accounts for 91.7 percent of total 
variance in the dataset (Table 4). Combined, first and second 
and first, second and third contained 96.2 and 98.5 percent of 
total variance, respectively. Here the first three principal 
components were used in this study. 
  
  
4.2 Analysis of Interrelationship of Variability 
Among the soil fertility and textural parameters soil organic 
matter (OM), available N and silt content were significantly 
correlated with spectral indices (Table 6). The number of data 
points was 35 for OM, available N, Available P and available K 
and 12 for sand, silt and clay content. For soil organic matter 
highly significant correlations were found with the radiance of 
blue and green bands, brightness index and redness index. The 
red and VNIR band and the PCI also produced significant 
correlations. Most of these correlations, except for RI and PCI 
were negative. All the spectral bands and indices such as BNI, 
SI, HI and CI had significant negative correlations with 
available nitrogen. Available P and K did not have any 
significant correlation. Among the soil textural parameters, silt 
content had significant correlation with maximum number of 
spectral parameters. The sand and clay content had significant 
correlation with RI and CI, respectively. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
— ne 
s analyzed by 
> soil and the 
s carried out to 
al parameters. 
ising stepwise 
from RS data. 
ly for those 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Table 4. The characteristics of the principal components from 4 
band IKONOS MS data Table 6. Correlation study of spectral parameters derived from 
Principal Eigen- Deviation | Variance Cum. IKONOS data and soil parameters 
Compo- value (%) Variance Spectral 
nent (%) Para- O.M. | Avl.N | Avl.P | Avi. K | Sand Silt Clay 
1 646.2826 25.4221 91.69 91.69 meters (%) (ppm) | (ppm) | (ppm) (95) (96) (9/0) 
2 31.6867 5.6291 4.50 96.19 : 
= = Bl -0.584***|-0.295*| -0.160 | -0.035 | 0.396 |-0.600*| 0.100 
3 16.5773 | 40715 2.35 98.54 x : Lp 
4 10.3307 3.2141 1.47 100.00 Green |-0.477** |-0.325*| -0.090 | 0.022 | 0.390 |-0.505*| -0.011 
Red -0.365* |-0.367*| -0.096 | -0.014 | 0.300 |-0.512*| 0.149 
. The Variability Analysis 
di me vertability Analysis VNIR | -0.422* |-0.301*| -0.072 | 0.025 | 0.350 |-0.530*| 0.089 
The mean, standard deviation and coefficient of BNI |-0.474** |-0.344*| -0.097 | 0.010 | 0.420 |-0.569*| 0.021 
variation of the soil and spectral parameters for 35 locations (12 
locations for soil texture) are presented in table 5. Overall the SI 0.136 [-0.324*1-0.048 | -0.030 [ 0.024. | -0.262 | 0.294 
soil has sandy loam texture with low organic matter, low 5 
z e , HI -0.266 |-0.331*| -0.007 | -0.019 | 0.034 | -0.192 | 0.188 
available N, high available P and low available K. The : 
variability analysis, as reflected by the coefficient of variation, CI 0.158 | -0.209 | -0.029 | -0.080 | -0.125 | -0.222 | 0.496* 
showed that, among soil parameters the variability was highest o : oo 
© e^ * C 3 3 42 = A] # * GC 
for available P (CV=29.9%), followed by silt percentage (CV= RI 0.244 B 194 0-065. Re Ml a RT 
20.8%). Among the spectral parameters the CV was highest for PCI 0.328*..1.0.047. ]. 9.103.1.-0.139 1:0.245. 1 -0.306 ] -0.021 
PC3 (161.9%), followed PC2 (101.4%) and PC1 (84.0%). : 
PC2 -0.194 | -0.173 | -0.002 | 0.043 | 0.133 | -0.384 | 0.268 
Table 5. Variability of field and special Parameter or the soil PC3 -0.251 0.125 | -0.004 | 0.073 | 0.390 | -0.169 | -0.441 
Parameter Mean Std. Dev. C V. (95) *0.01«p«0.1, **0.001«p«0.01, ***p«0.001 
O.M.(%) 0.25 0.057 22.98 
Available N (ppm)| 103.23 14.22 13.77 
Available P (ppm) 28.54 8.61 30.16 The multiple regression equations were generated between soil 
Available K (ppm) 100.34 20.70 20.62 and spectral parameters using stepwise regression technique. 
; - Empirical equations were generated only for those parameters 
and (% 2 2.208 2 : : ee. eu ap T 
Sand SS 82.58 2208 Sr for which the correlations were significant (Table 7). The 
Silt (%) 8.05 1.678 20.84 empirical relation between OM and spectral indices was highly 
Clay (%) 9.37 1.310 13.93 significant with coefficient of determination R = 0.733 and F = 
BI 0.61 0.016 2.65 SAT RE sand, silt and clay DE icd ined Sy 
nultiple regression equations with spectral parameters. In each 
S 0.12 0.011 8.97 I cn 
of these equations only one spectral parameter came into the 
HI -3.92 0.498 -12.73 equation. The multiple R for these equations ranged from 0.495 
CI -0.19 0.006 -3.34 to 0.599 and F value ranged from 3.3 to 5.6. However, 
RI 1.07 0.056 $.2] Available N, though individually had significant correlation 
— E : with spectral parameters, did not form a significant multiple 
PCI -0.18 9.155 :83:96 regression equation. 
pc? 0.007 0.007 101.38 
PC3 -0.003 0.005 -161.95 
  
  
  
  
  
  
Table 7. Empirical equations between soil and spectral 
parameters derived using stepwise regression 
technique.