parameters and coefficients are presented in table 9. For all the 
indices, the best fit empirical relation with biomass was 
exponential, except for NDVI (NIR, Red), where the best fit was a 
quadratic form. The R? values for these non-linear relationships, 
ranged from 0.268 for MSI to 0.427 for RVI (NIR, Red). These R 
values are significant at 5% level, considering the number of data 
points being 28. 
Table 9. Parameters of best-fit regression equation between above 
ground biomass, at harvest, and VIs derived from 
hyperspectral data 
IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India,2002 
  
  
  
Index | Equation | R? | N F b0 bl 
Form 
NDVI | Expon- | 0.427 | 28 | 20.88 | 327.019 | 2.5081 
(552- ential 
687) 
NDVI | Quadra- | 0.355 | 28 | 7.44 | 31981.9 | -74353 
(927- tic b2 
687) 44084.5 
  
  
  
     
  
  
       
RVI Expon- | 0.427 | 28 | 20.84 | 257.4 0.5354 
(552- ential 
  
  
687) 
RVI Expon- | 0.308 | 28 | 12.46 | 506.184 0.0226 
(927- ential 
687) 
MSI Expon- | 0.268 | 28 | 10.25 | 1522.75 -1.9163 
  
  
  
  
  
  
ential 
  
  
A similar correlation analysis was also carried out between soil 
colour spectral indices and the soil nutrient parameters (Table 10). 
All the colour related indices were negatively correlated with the 
soil parameters except for brightness index (BI). However, except 
for the correlation between available potassium and spectral 
indices, the correlation were found to be generally not significant. 
The available K is dependent upon clay content (type of clay), 
thus to soil texture. Since soil texture is a soil colour imparting 
character, this might be the reason of high correlation between 
available K and colour indices. Principal components of 
hyperspectral data have been found to be correlated to soil 
properties such as texture, OM and iron content (Palcios-Orueta 
and Ustin, 1998). Hence this will be followed as further 
improvement to this analysis. 
Table 10. Correlation study of spectral indices and soil parameters 
  
  
  
  
  
  
O.C.(%) | Available N Available P Available K 
(ppm) (ppm) (ppm) 
BI 0.10 0.07 0.16 0.13 
SI -0.20 -0.20 -0.09 -0.67 
HI -0.23 -0.19 -0.12 -0.59 
CI -0.21 -0.19 -0.10 -0.64 
RI -0.07 -0.14 -0.09 -0.34 
  
  
  
  
  
  
  
4.4 Study of variability maps 
Variability maps were generated from to know the spatial 
distribution of the crop and soil variability. Figure 3 shows the 
variability of cropped field as generated by NDVI thresholding of 
satellite based remote sensing data and interpolation (krigging of 
above ground biomass, at harvest, and NDVI (552 - 687). The 
variability distribution pattern of NDVI (552-687) and biomass 
are mostly similar, the direction of increase in the value of the 
parameter being from north to south of the field. However, the 
satellite based NDVI map does not show, a similar pattern 
indicating limitation of 23-m resolution for studying small fields. 
Similarly figure 4 shows the variability map generated for soil by 
interpolation (krigging) of available K and Redness Index (RI). As 
shown in the correlation analysis the variability maps also show 
opposite pattern, high RI and low available K towards northern 
side and low RI and high available K towards eastern side. 
im 
  
  
   
© 
a NDVI ras b) NDVT (882-687 cHioruass si harvest 
Figure 3. Variability of cropped field as shown by a) NDVI 
thresholding of satellite based remote sensing data and 
interpolation of b) NDVI (552-687) and c) above 
ground biomass at harvest 
» ® = æ "a "I ww = » ® Kk wm a 
a Redmess Index b; Avattahle k ippmi 
Figure 4. The variability maps generated for soil by interpolation 
of a) Redness Index and b) available K. 
    
Th 
am 
sm 
by 
nal 
crc 
spe 
sat 
use 
sat 
Fic 
Sp 
Ha 
oa! 
In 
Te 
of 
so 
Re 
sir 
an 
an 
m: 
De 
fai 
Ve 
an