Preharvest Monitoring of Biomass Production - Engineering and Science of Biomass Feedstock Production and Provision - page 71

stages, and the switchgrass images refl ected more NIR and absorbed red during
September; the NIR refl ectance decreased during November and December. The
reference data was collected from the ground reference points during this period.
The stand-alone tower images for the ground reference points present the crop
response and physiological changes. In September, the crops were green, and the
NIR refl ectance and red band absorption were higher. The RGB images also repre-
sent the changes of canopies. The perennial crops are in the fi rst year of their growth.
The canopies would be denser as the year increases.
The images were captured from 38-m height during 12 to 3 pm daily during the
growing season. The daily high temporal resolution is the major advantage of the
image database. In the ground sampling, the reference points were selected to keep
track of the vegetation index and intercepted solar radiation by the canopy. The
spectrometer response from Miscanthus canopy was analyzed for NDVI and
GNDVI (see Fig. 4.9 ). NDVI are related with red and NIR band and chlorophyll
absorption. On the other hand, the GNDVI was related with green band. As indi-
cated by Fig. 4.9a , the NDVI value decreased from September to November and
December; that is to say, the CIR image of Miscanthus in September had more NIR
information than red and gradually decreased during November and December. The
GNDVI index value observed of Miscanthus was closer during the 3 months.
The NDVI and GNDVI trajectories are depicted for switchgrass during
September, November, and December as indicated in Fig. 4.9b . In September, the
NDVI value of switchgrass was higher than the value for November and December.
On the other hand, the GNDVI was closer during September, November, and
December. In the prairie fi eld, the NDVI value for September at point 4 had noise
and did not represent the regular response (Fig. 4.9c ). This could have occurred due
to measurement error or irregular canopy structures.
Based on the daily images from the established biomass energy crop remote
sensing system, we can easily monitor the daily growth condition of biomass energy
crop. The daily NDVI value, which represents the growth condition, can be calcu-
lated, and therefore, the growth pattern of different bioenergy crop in 2012 can be
recognized as indicated in Fig. 4.10 .
The daily NDVI value can be accumulated during the whole season for predict-
ing the biomass accumulated in the energy crop. To verify the feasibility of biomass
yield prediction based on remote sensing data, the accumulated NDVI value based
on the remote sensing image of Miscanthus was correlated with the actual harvested
biomass from ground truth data of year 2011 as indicated in Fig. 4.11 . The results
showed that the fi tting accuracy ( R 2 ) of the correlation model was 64.4 %. Therefore,
there is great potential for predicting biomass yield based on the near-real-time
remote sensing image after recalibration with the ground truth data.
The biomass yield of Miscanthus in 2012 was predicted based on the correlated
model derived in 2011 as indicated in Fig. 4.12 . Additionally, large-scale biomass
yield prediction based on the near-real-time remote sensing image after recalibra-
tion with the ground truth data becomes possible and so that the decision support
tool with data to knowledge can be achieved for the BFP industry.