—plot 101 
m"plot 120 
* plot 210 
plot 300. 
  
es 
— plot 101 
—plot 120 
—plot 208 
— plot 210 
— plot 300 
  
Clevers, Jan 
  
4.3 Calibrating Crop Growth Model 
The simulation runs with the crop growth model were compared to the results of the two methods of introducing the 
remote sensing data. Figure 6 and 7 illustrate the three simulation methods for winter wheat field 120. Table 2 gives the 
obtained results for the comparison of LAI and grain yield values against the field observations. 
  
  
LAI Simulated Simulated Grain Yield 
  
~~ 
aan 
  
  
  
  
—"M+RESET 
——— SIM+RESET+INTERP 
  
  
  
SM+RESETHNTERP 
  
  
  
Grain Yield (kg ha-1) 
  
  
  
   
  
  
Day of Year Day of Year 
  
  
  
  
  
Figure 6: The 3 simulation methods applied for LAI Figure 7: The 3 simulation methods applied for grain 
simulation as a function of the day of year in 1997 on yield simulation as a function of the day of year in 
winter wheat field 120 at the Alpilles test site. 1997 on winter wheat field 120 at the Alpilles test 
Table 2: Root mean square errors of LAI (ha ha) and grain yield (t ha) for the 3 simulation methods. Numbers in 
brackets point to the number of observations. 
  
  
  
  
  
  
  
  
LAI Grain Yield 
Field-id | Simulation Simulation 4 Simulation, Simulation Simulation + Simulation, 
reset reset + reset reset + 
interpolation interpolation 
101 0.888 (17) 0.617 (17) 0.362 (17) 1.652 (10) 1.379 (10) 1.436 (10) 
120 1.010 (13) 0.907 (13) 0.577 (13) 0.816 (10) 0.679 (10) 0.673 (10) 
208 0.706 (10) 0.441 (10) 0.522 (10) 1.734 (7) 2.057 (7) 1.900 (7) 
210 0.571 (8) 0.735 (8) 0.712 (8) 1.936 (3) 2.201 (3) 2.315 (3) 
300 0.756 (7) 0.654 (7) 0.685 (7) 1.124 (6) 1.379 (6) 1.314 (6) 
average | 0.786 (55) 0.671 (55) 0.572 (55) 1.452 (36) 1.561 (36) 1.528 (36) 
  
  
  
  
  
  
  
  
  
As the results in table 2 indicate, the introduction of remote sensing data increases the accuracy of the LAI estimates, 
although the yield component of the biomass simulation does not increase likewise: estimations of the grain yield 
become less accurate. This is due to the fact that leaf area estimates are specifically well estimated in the beginning of 
the growing season, while for yield formation, the values at the grain filling period are more important. 
Figure 6 shows the effect of the three simulation methods on the LAI. As can be seen, the error in estimating LAI from 
SPOT data is introduced directly into the crop model. Nevertheless, it is clear that LAI estimates become more accurate, 
compared to the observed values. The relatively high values of the rmse for grain yield, are merely due to deviations in 
the first phase after flowering, while the final results of the simulation methods are more accurate with the final 
observed yields. Table 3 gives the deviations (t ha) in final yield simulations for the three methods. Results show that 
the third method, mainly based on remote sensing, as an average yields best results. Results are not improved 
significantly by using remote sensing information, because the growth model used was initially calibrated on field data 
from the study area including the year of study and the actual sowing date was included. 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 277