Oklahoma State University
Determining the Most Effective Growth Stage in Corn Production for
Spectral Prediction of Grain Yield and Nitrogen Response
Department of Plant and Soil Sciences
Department of Biosystems and Agricultural Engineering
0.7
0.8
0.6
0.7
51,870
66,690
81,510
0.6
0.5
0.4
0.3
0.5
0.4
0.3
0.2
0.2
0.1
0.1
0
V8
V9
V11
R1
R2-R3
R4
R5
V5
V6
V8
V9
V11
R1
R2-R3
R4
Influence of plant population on CV from Green and Red NDVI at V8 in the 99-day hybrid with sufficient
nitrogen, Haskell, OK
37,050
51,870
66,690
81,510
60
60
50
50
37,050
51,870
66,690
81,510
40
Red CV
40
30
30
0.6
0
0.8
0.2
20
10
10
0.6
0.8
1
0.8
1
14000
10000
8000
6000
4000
10000
8000
6000
4000
2000
2000
0
0
0.2
y = 1349.7e2.6383x
R2 = 0.659
12000
y = 1033.8e3.5346x
R2 = 0.6749
12000
0.4
0.6
0.8
0
0.2
0.4
0.6
RNDVI
GNDVI
Exponential regression, NDVI and grain yield
Linear regression, RINDVI and RIHARVEST
0
V5
V6
V8
V9
V11
R1
R2-R3
R4
R5
V5
V6
V8
V9
V11
R1
R2-R3
R4
R5
Coefficient of determination (R2)
Coefficient of determination (R2)
Influence of N rate on Green and Red NDVI at V8 in the 99-day hybrid with high plant population,
Haskell, OK
0N
0.9
84 N
168 N
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.4
0.2
0.2
0.1
0.1
0
0
V8
V9
V11
R1
R2-R3
R4
Green
NDVI
168 N
Red
NDVI
Green
NDVI
Red
NDVI
R5
V5
V6
V8
V9
V11
R1
R2-R3
R4
R5
84 N
168 N
0N
50
50
84 N
Red
NDVI
168 N
45
45
V8
V9
99-day
0.368
0.258
0.399
113-day
0.344
0.319
0.433
99-day
0.179
0.254
0.500
113-day
0.412
0.221
0.467
V7
V8
V9
99-day
NA
0.751
0.679
113-day
NA
0.644
0.673
99-day
NA
0.745
0.671
113-day
NA
0.598
0.558
V7
V8
V9
99-day
0.429
0.502
0.548
113-day
0.163
0.322
0.250
99-day
0.545
0.549
0.529
113-day
0.143
0.281
0.273
Lake Carl Blackwell
Green
NDVI
Influence of N rate on CV from Green and Red NDVI at V8 in the 99-day hybrid with high plant
population, Haskell, OK
0N
V7
Haskell
0.4
0.3
V6
84 N
Greenlee Farm
0.5
0.3
V5
0N
0.9
RNDVI
GNDVI
0.4
RNDVI
14000
0
20
0
0.4
y = 3067.3e1.5258x
R2 = 0.3734
Relationship between grain yield and NDVI at V8, 99-day hybrid over three locations
R5
Grain yield (kg/ha)
V6
0.2
20000
18000
16000
14000
12000
10000
8000
6000
4000
2000
0
GNDVI
0
V5
Green CV
y = 3060.5e1.7866x
R2 = 0.3567
Grain yield (kg/ha)
0.9
20000
18000
16000
14000
12000
10000
8000
6000
4000
2000
0
Grain yield (kg/ha)
37,050
81,510
Grain yield (kg/ha)
66,690
RNDVI
GNDVI
51,870
0.8
0
Greenlee Farm
Green
RINDVI
Red
RINDVI
V7
V8
V9
99-day
0.039
0.010
0.183
113-day
0.434
0.255
0.614
99-day
0.069
0.003
0.086
113-day
0.325
0.541
0.556
V7
V8
V9
99-day
NA
0.235
0.056
113-day
NA
0.092
0.251
99-day
NA
0.161
0.239
113-day
NA
0.396
0.529
V7
V8
V9
99-day
0.764
0.819
0.757
113-day
0.381
0.823
0.625
99-day
0.593
0.603
0.686
113-day
0.671
0.681
0.817
Haskell
Green
RINDVI
Red
RINDVI
Lake Carl Blackwell
Green
RINDVI
Red
RINDVI
40
40
35
Red CV
35
Green CV
30
25
20
30
20
15
10
10
5
5
0
0
V5
V6
V8
V9
V11
R1
R2-R3
R4
Conclusions
25
15
V5
R5
V6
V8
V9
V11
R1
R2-R3
R4
R5
Relationship between plant population and CV from Green and Red NDVI at V8 in the 99-day hybrid with
sufficient N over three locations
70
70
60
y = 27.786e-1E-05x
R2 = 0.5138
60
50
y = 54.352e-2E-05x
R2 = 0.5748
50
Red CV
•Three experimental sites were established in the spring of 2004
•Eastern Oklahoma Research Station near Haskell, OK on Taloka silt loam
soil (fine, mixed, thermic Mollic Albaqiustoll)
•Lake Carl Blackwell Agronomy Research Farm near Stillwater, OK on
Pulaski fine sandy loam soil (course-loamy, mixed, nonacid, thermic Typic
Ustifluvent)
•Greenlee Farm near Morris, OK on Taloka silt loam soil (fine, mixed, thermic
Mollic Albaqiustoll)
•Ammonium Nitrate (34-0-0) was broadcast at 0, 84, and 168 kg N ha-1 by hand
and incorporated in the soil shortly before planting
•Two Bacillus thuringiensis (bt) gene enhanced corn hybrids identified by their
maturity date (99-day and 113-day) were planted at each site in 2004
•Four seeding rates were evaluated in 76 cm rows
•37,050, 51,870, 66,690, and 81,510 plants ha -1
•Sensor readings were taken with a GreenSeeker Hand Held optical reflectance
sensor (Ntech Industries, Ukiah, CA), measuring Red and Green, normalized
difference vegetation index (NDVI) at different vegetative and reproductive
growth stages
•Corn grain was harvested by hand, removing 2 rows x 9.14 m from the center of
each plot
•Grain yield from each plot was determined and a sub-sample was taken for total
N analysis
•Red NDVI = [(NIRref/NIRinc)-(Redref/Redinc)] /
[(NIRref/NIRinc)+(Redref/Redinc)]
•Green NDVI=[(NIRref/NIRinc)-(Greenref/Greeninc)] /
[(NIRref/NIRinc)+(Greenref/Greeninc)]
•Response indices (RI)
•Vegetative = calculated by dividing the highest N treated NDVI average by
the check (0 N rate) average
•Harvest = highest N treated grain yield average divided by the check (0 N
rate) average
37,050
Green CV
Materials and Methods
Relationship between grain yield and NDVI at V8, 113-day hybrid over three locations
Influence of plant population on Green and Red NDVI at V8 in the 99-day hybrid with sufficient N,
Haskell, OK
Abstract
With the escalation in environmental concern and cost of production,
researchers have recently focused on investigating more efficient means of
increasing grain yield while reducing fertilizer use. This study evaluated spectral
reflectance, measuring the normalized difference vegetation index (NDVI) with a
GreenSeeker® Hand Held optical reflectance sensor as a function of corn (Zea
mays L.) hybrid, plant population, and fertilizer N rate. Initial investigation of these
variables in 2002 and 2003 concluded that higher plant populations (>49,400 plants
ha-1) caused early canopy closure, resulting in NDVI peaks at V10, where as NDVI
did not peak at lower plant populations (35,568 plants ha-1) until R1. In the spring
of 2004 with the addition of a third site and the availability of a green NDVI sensor,
the trials were reconfigured removing one hybrid and imposing two more plant
populations and the utilization of both green and red NDVI. Green NDVI values
peaked between V7 and V8 when compared to red NDVI (peaked at V11) and green
NDVI was not affected by plant population in the vegetative stages, as was red
NDVI. Plant population increased NDVI measurements and reduced coinciding
coefficient of variation (CV) measurements significantly as population increased
from 37,050 to 66,690 plants ha-1, but no differences occurring between 66,690 and
81,510 plants ha-1. Green NDVI, Red NDVI, and CV were all highly correlated at V7,
V8, and V9 growth stages. Coefficient of Variation data from V8 showed a
relationship with measured plant population at sufficient N levels. Grain yield
correlated well with both green and red NDVI at V8 and V9 growth stages.
Vegetative response index (RINDVI) peaked between V8 and V9 at responsive
locations, however correlation with final RI (RIHARVEST) was limited. Regression
analysis indicated that early-season grain yield prediction and vegetative RI
measurement was hybrid and site sensitive and needs further refining to improve
accuracy. Nevertheless, this study revealed that N response could be determined
at early growth stages using either Green or Red NDVI and that the potential exists
to predict grain yield using either band.
R.K. Teal, K.W. Freeman, W.R. Raun, J. Mosali, K.L. Martin , G.V. Johnson, J.B. Solie, and H. Zhang
40
30
40
30
20
20
10
10
0
0
0
20,000
40,000
60,000
Plant pop. (plants/ha)
80,000
100,000
0
20,000
40,000
60,000
Plant pop. (plants/ha)
80,000
100,000
Plant population can influence NDVI and grain yield prediction
CV can be used to predict plant population (improve yield prediction)
Green and Red NDVI from V8 and V9 growth stages was highly correlated with
grain yield
Green and Red NDVI worked equally well for predicting grain yield from V7 to V9
Different yield prediction curves will be necessary for Green and Red NDVI
Vegetative response index (RINDVI) to N peaked between V8 and V9 at responsive
locations
Need for added N can be determined early in season while the crop is small
enough for side-dress N applications
Regression analysis indicated that early-season grain yield prediction and
vegetative RI measurement was hybrid and site sensitive and needs further refining
to improve accuracy