Natural variability within corn fields is a common observance, but the resolution at which this variability occurs and the methods of nitrogen (N) fertilization currently employed fail to factor in the spatial difference among corn plants. The objective

1
By-Plant Prediction of Corn (Zea Mays L.) Grain
Yield Using Early Season Sensor Measurements
Plant and Soil Sciences-OSU USDA-ARS
Lincoln, Nebraska USDA-ARS NSTL,
Iowa Biosystems and Agricultural Engineering
K.L. Martin, P.J. Hodgen, J.L. Hatfield, W.R.
Raun, K.W. Freeman, J.B. Solie, M.L. Stone, R.K.
Teal, D.B. Arnall, K. Desta
Abstract
Results
Natural variability within corn fields is a
common observance, but the resolution at which
this variability occurs and the methods of
nitrogen (N) fertilization currently employed
fail to factor in the spatial difference among
corn plants. The objective of this study is to
demonstrate that differences in plant yield occur
on an individual plant basis and to estimate the
yield of each plant by collecting the normalized
difference vegetative index (NDVI) using a
GreenSeeker optical sensor. NDVI readings were
collected every 1.1 cm in a 30 m row. Knowing
the exact location of each plant, the average
NDVI value was calculated by assuming that each
plant occupies half the distance between it and
its neighbors. As NDVI values were collected, it
was clear that differences in plants could be
detected using an optical sensor. Yield was then
obtained for each plant and was used to find a
relationship with NDVI. Using NDVI alone
deciphers plant biomass, while a second
dimension, height, can be added to the yield
prediction. It is clear from this work that
precision solutions for nutrient management must
be made at the by-plant level, since 10X
differences in yield were observed over very
short distances (15m) and that could be predicted
at V8.
Conclusions
Materials and Methods

• The GreenSeeker sensor consistently detected
  differences, by-plant
• 10X differences in corn grain yield were
  found over 15m (Iowa and Oklahoma)
• By-plant NDVI readings at V8 could be
  successfully used to predict final grain yield
• Incorporation of plant height (sonar) may
  improve prediction of grain yield
• Precision sensing solutions for nutrient
  management should be made at the by-plant level
• since 10X differences in yield were
  observed over very short distances (15m) and that
  could
• be predicted using NDVI readings at the V8
  corn growth stage.

Two site-years were established in the spring of
2003 EFAW Research Station at
Stillwater, Oklahoma Lake Carl
Blackwell Research Station west of Stillwater,
Oklahoma Five site-years were established in 2004
EFAW Research Station at Stillwater,
Oklahoma Lake Carl Blackwell Research
Station west of Stillwater, Oklahoma
Hennessey Research Station at Hennessey,
Oklahoma USDA-ARS MSEA site near
Shelton, Nebraska USDA-ARS site at Ames
Iowa The location of each plant was measured from
the start point or beginning of each row. At
various growth stages, a GreenSeeker optical
sensor equipped with a shaft encoding device and
mounted on a bicycle was used to measure the NDVI
of each individual plant. The resolution at
which NDVI values were measured was about 1.02
cm, that was then averaged based on half the
distance from the plant in question to its
nearest neighbor. In 2004, the height of each
plant was measured by extending the uppermost
fully developed leaf up and measuring the height
from the ground. A sonar unit was also used to
accumulate canopy height while sensing. NDVI
and yield was calculated based on the area that
each plant occupied using the following
equation Area of plant 2 (½ distance
between 1 and 2 ½ distance between 2 and 3)
row spacing
Oklahoma State University
Cowboys