950nm), the average of o is 0.45 and that of z is 0.49. 
These values and measured soil reflectance op ,=0.055 
are used for the simulation. Coefficient of variation of 
p is about 5%. In the other regions, coefficient of 
variation is higher than 15%. This shows that NIR 
region is the most optimal to use as the indicator of the 
3-D structure of rice canopy that is not withered. 
4. SIMULATION OF CANOPY 
BIDIRECTIONAL REFLECTANCE 
4.1 condition of the simulation 
The simulation is carried out under the following 
two conditions. 
Condition A. Experimental field: this is the condition 
that is similar to the experimental field measurement 
condition. In the other words, direction - intensity 
characteristics of the incident light is approximated by 
a function of yzl(0)cos7.168. Where I(6) shows the 
intensity to the direction 0. Basic unit is situated 
based on the measurement situation. 
Condition B. Wide spread field: This is the condition 
that specular light enters homogeneously to a scene 
composed of repeat of basic unit which spread enough 
widely to the vertical direction. IFOV of the sensor can 
be any combinations of basic unit. In order to account 
for the air effect, it is necessary to combine air 
reflectance models with this model. The probability 
concerning to photon fate remarked in 2.3 is given by 
causing a suspected random number. Individual leaf 
reflectance and soil reflectance are given by in situ and 
field measurement respectively. 
B-1. Leaf reflectance o, absorbance «œ and 
transmittance c are constants. 
B-2. p, a and r are given as follows. 
p7 p.* p, 
a=a ta, (1) 
171-p-a 
Where p and a. are constants. oy and a, are 
10.050, and £0.05 respectively. 
4.2 comparison between simulated and measure 
NIR bidirectional reflectance distribution qj 
thirteen kinds of rice canopy is simulated under 
condition A. Fig. 4. Shows the comparison of the 
calculated reflectance factor to the view azimuth 1g 
deg., view zenith 0, 45, 60 deg. with the measured one 
in an experimental field. RMSE means Root Mea 
Square Error. This shows the simulation is effective for 
valuating the rice canopy bidirectional reflectance. 
4.3 simulation of rice canopy reflectance in wide 
spread fields 
Fig. 5. shows simulated NIR bidirectional 
reflectance distribution of thirteen kinds of rice canopy 
under condition B-1. Incident zenith angle is 15 deg. 
2.5 x 106 photons are traced. Following results are 
obtained by the simulation. 
Basically, reflectance factor inclined to be higher 
to each direction as rice grows up. This shows that the 
effect of the total leaf area increasing on bidirectional 
reflectance is more dominant than that of leaf 
inclination in each direction.However, comparing the 
dense plot in July 16th with that of July 21th, the 
former has higher reflectance to the nadir direction, and 
has lower reflectance to greater than 30 deg. zenilh 
angle. This is the case of either 0, 45, 90 deg. azimulh 
angle. This seemed to be caused by the difference oi 
leaf inclination. 
Dense plot:e ,Sparse plot:o 
(a) V.Z.-60deg. 
(b) V.Z.-45deg. 
(c) V.Z=0deg. 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
25; - | 
8 20! RMSE=4.0 | < | RMSE=3.5 | RMSE=1 5 
8 15] 14 ; 
© ] o Me | 
> 10- Z^ À 
oi 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
0 510152025 0 5 10152025 0 5 10 1520 25 
  
Measured (%) 
Fig. 4. Comparison between measured and calculated reflectance factors 
(condition A, view azimuth = 0° ) 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996