Transcript Document 7363831

Variable Rate Technology in Wheat
www.dasnr.okstate.edu/precision_ag
1993
Dr. Marvin Stone adjusts the fiber
optics in a portable spectrometer
used in early bermudagrass N rate
studies with the Noble Foundation,
1994.
Dr. John Solie, Dr. Marvin Stone, and Dr. Shannon Osborne collect
sensor readings at ongoing bermudagrass, N rate * N timing
experiments with the Noble Foundation in Ardmore, OK. Initial results
were promising enough to continue this work in wheat.
Variable N rates using an inverse N-rate, NDVI
scale were applied. N rates were cut in half
with no differences in grain yield compared to
fixed rates. Grain N uptake levels using VRT
across a 70 meter transect were less variable
when compared to the fixed rates (left).
Rep 2, Miller-2
Grain N uptake, kg/ha
60
50
40
30
20
10
0
0
10
20
30
40
50
60
70
Check
28.9 ± 6.4 Distance, m
Variable Rate 48.5 ± 4.2
Fixed Rate 44.1 ± 8.2
NDVI
NDVI == NIR
NIR -- Red
Red
NIR
NIR ++ Red
Red
Increased
Increased soil
soil background
background
Red
Red 620-750
620-750 nm
nm
NIR
NIR 750-1x10
750-1x1066 nm
nm
increased
increased red
red reflectance
reflectance 
 decreased
decreased NDVI
NDVI
00
10
10
20
20
30
30
40
40
50
50
60
60
70
70
80
80
0.2
0.2
50
50
40
40
30
30
20
20
10
10
0.3
0.3
0.4
0.4
0.5
0.5
0.6
0.6
0.7
0.7
N/ac)
(lb N/ac)
Uptake (lb
N Uptake
N
In
In this
this region,
region, NDVI
NDVI is
is
influenced
influenced by
by
(a)
soil
(a) soil
(b)
(b) %
% coverage
coverage
(c)
(c) color
color
N Rate (lb N/ac)
In
In this
this region,
region, NDVI
NDVI is
is
influenced
influenced by
by
(a)
(a) color.
color.
1994
0.8
0.8
NDVI
NDVI
Initial algorithms used to spatially treat
N deficiencies in wheat and
bermudagrass employed an inverse N
Rate-NDVI scale.
John Ringer and Shannon Osborne
collected sensor readings and later
applied variable N fertilizer rates
based on an initial bermudagrass
algorithm.
Samples were collected from every 1 square foot. These
experiments helped to show that each 4ft2 in agricultural
fields need to be treated as separate farms.
1995
New ‘reflectance’ sensor developed.
Extensive field experiments looking at changes in
sensor readings with changing, growth stage,
variety, row spacing, and N rates were conducted.
Collaborative Project with CIMMYT
Variety Selection/Yield Potential
Spring Wheat 1995
CIMMYT
Date
Location
Personnel
Objectives
Feb, 1997
Ciudad Obregon
TEAM-VRT
Jan, 1999
Obregon & Texcoco
Steve Phillips, Joanne LaRuffa,
Wade Thomason, Sherry Britton,
Joe Vadder, Gordon Johnson,
John Solie, Dick Whitney
Discuss potential
collaborative work
IRSP 98, refine INSEY, 2wheel tractor and wheat
bed planter design
Sep, 1999
Texcoco
Erna Lukina
Aug, 2000
Texcoco
Jan-Mar 2001 Ciudad Obregon
Marvin Stone, Kyle Freeman,
Roger Teal, Robert Mullen,
Kathie Wynn, Carly Washmon,
Dwayne Needham
Kyle Freeman
Apr 2001
Ciudad Obregon
Kyle Freeman
July 2001
El Batan
Apr 2002
Ciudad Obregon
Jagadeesh Mosali, Shambel Moges Wheat harvest
Micah Humphreys, Paul Hodgen,
Carly Washmon
Paul Hodgen
NASA Grant
TOTAL
8
26
IRSP 98, use of EY as a
selection tool
IRSP 99, applications of
INSEY, sensor design
for plant breeding
Joint collaboration on
200-03530 NRI Grant
Wheat harvest
In, March, 1996, first variable
rate N applicator demonstrated
to the public
1996
Wet Biomass,
Biomass, kg/ha
kg/ha
Wet
May
May 29
29 1996
1996
9000
9000
Relationship
7000
7000
between total
5000
forage N uptake 5000
3000
3000
and NDVI was used
1000
to apply variable N 1000
rates in turf.
yy == -15770x
-15770x ++ 41909
41909
22
0.73
R
R == 0.73
22
2.1
2.1
2.2
2.2
2.3
2.3
2.4
2.4
2.5
2.5
2.6
2.6
695/405
695/405
Indices developed where we could detect
differences in N and P, independent of one
another. For wheat, numerator wavelengths
between 705 and 735, and denominator
wavelengths between 505 and 545 proved to
be reliable predictors of N and P uptake. In
bermudagrass, the index 695/405 proved to
be reproducible from one season to the next.
Evaluation of management
resolution at 3 locations
1997
In 1997, our precision sensing team
put together two web sites to
communicate TEAM-VRT results.
Since that time, over 20,000 visitors
have been to our sites.
(www.dasnr.okstate.edu/precision_ag)
www.dasnr.okstate.edu/nitrogen_use
The first attempt to combine sensor readings over sites into a single equation for yield
prediction A modification of this index would later become known as INSEY (in-season
estimated yield), but was first called F45D.
6000
6000
Grain Yield
Yield
Grain
5000
5000
4000
4000
Perkins,
Perkins, N*P
N*P
Perkins,
Perkins, S*N
S*N
Tipton,
Tipton, S*N
S*N
3000
3000
2000
2000
yy == 1E+06x2
1E+06x2 -- 12974x
12974x ++ 951.24
951.24
R
R22== 0.89
0.89
1000
1000
00
0.01
0.01
0.02
0.02
0.03
0.03
0.04
0.04
0.05
0.05
NDVI
NDVI F4+NDVI
F4+NDVI F5/days
F5/days from
from F4
F4 to
to F5
F5
0.06
0.06
0.07
0.07
Cooperative research program with CIMMYT.
Kyle Freeman and Paul Hodgen have each spent 4
months in Ciudad Obregon, MX, working with
CIMMYT on the applications of sensors for plant
breeding and nutrient management.
1998
6
N*P Perkins, 1998
5
S*N Perkins, 1998
S*N Tipton, 1998
Cooperative
Research Program
with Virginia Tech
Grain Yield, Mg ha
-1
N*P Perkins, 1999
Experiment 222, 1999
4
Experiment 301, 1999
Efaw AA, 1999
Experiment 801, 1999
3
Experiment 502, 1999
N*P Perkins, 2000
Experiment 222, 2000
2
Experiment 301, 2000
Efaw AA, 2000
Experiment 801, 2000
1
Experiment 502, 2000
Hennessey, AA, 2000
VIRGINIA (7 Loc's)
0
0
0.001
0.002
0.003
0.004
0.005
0.006
INSEY (NDVI Feekes 4-6/days from planting to Feekes 4-6)
0.007
0.008
1999
Applications of indirect measures of electrical conductivity
were evaluated in several field experiments. This work
aims to identify added input variables to refine the inseason prediction of yield.
Covington, 1999
Wheat Grain Yield
Yield
Yield
Method
N Rate
bu/ac
kg/ha
lb/ac
__________________________________________
16.7
1122
0
18.9
1269
Fixed
40
27.5
1846
Fixed
80
35.7
2396
YP-INSEY
68
SED
230
TEAM-VRT entered into
discussions with John
Mayfield concerning the
potential commercialization
of a sensor-based N
fertilizer applicator for cereal
crops.
3.4
__________________________________________
19at lower N rates observed at Covington. Using the inCV, %
Increased yields
season response index (RINDVI), we were able to project responsiveness
to applied N, which changes from location to location based on climatic
conditions specific to each parcel of land, and that changes on the same
land from year to year.
Discovered that the N fertilizer rate
needed to maximize yields varied widely
over years and was unpredictable in
several long-term experiments. This led
to his development of the RESPONSE
INDEX.
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
3
1980
1978
1976
1974
y = 0.65x + 27 (CV = 62)
1971
Fertilizer-N (lb/acre)
Fertilized N required to maximize yield (Lahoma, OK).
90
80
70
60
50
40
30
20
10
0
2000
Year
2.5
y = 1.06x + 0.18
R2 = 0.56
2
RI Harvest
Predicted potential response to
applied N using sensor
measurements collected inseason. Approach allowed us to
predict the magnitude of response
to topdress fertilizer, and in time to
adjust topdress N based on a
projected ‘responsiveness.’
1.5
1
0.5
0
0
0.5
1
1.5
RI NDVI
2
2.5
3
2001
N Fertilizer Optimization Algorithm (NFOA):
1. Predict potential grain yield or YP0 (grain yield achievable with no additional
N fertilization) from the grain yield-INSEY equation, where;
9
INSEY = NDVI (Feekes 4 to 6)/days from planting to sensing (days with
GDD>0)
Biomass (Mg/ha)
8
7
y = 0.0438e6.2862x
R2 = 0.75
6
YP0 = 0.74076 + 0.10210 e 577.66(INSEY)
5
4
3
Feekes 10
2
1
0
0.3
0.5
0.7
0.9
Red NDVI
9
3. Determine the predicted yield that can be attained with added N (YP N)
fertilization based both on the in-season response index (RINDVI) and the
potential yield achievable with no added N fertilization, computed as follows:
8
Biomass (Mg/ha)
2. Predict the magnitude of response to N fertilization (In-Season-ResponseIndex, or RINDVI). RINDVI, computed as; NDVI from Feekes 4 to Feekes 6 in
non-N-limiting fertilized plots divided by NDVI Feekes 4 to Feekes 6 in the
farmer check plots (common fertilization practice employed by the farmer). The
non-N limiting (preplant fertilized) strip will be established in the center of each
farmer field.
7
6
5
YPN = (YP0)/ (1/RINDVI) = YP0 * RINDVI
4
4. Predict %N in the grain (PNG) based on YPN (includes adjusted yield level)
3
2
PNG = -0.1918YPN + 2.7836
Feekes 10
1
5. Calculate grain N uptake (predicted %N in the grain multiplied times YPN)
0
0
2
4
Visual Score
6
8
GNUP = PNG*(YPN/1000)
6. Calculate forage N uptake from NDVI FNUP = 14.76 + 0.7758 e 5.468NDVI
Work with wheat and triticale plant breeders
7. Determine in-season topdress fertilizer N requirement (FNR)= (Predicted
at CIMMYT, demonstrated that NDVIGrain N Uptake - Predicted Forage N Uptake)/0.70
readings could be used for plant selection
FNR = (GNUP – FNUP)/0.70
Engineering, plant, and, soil scientists at OSU release
applicator capable of treating every 4 square feet at 20
mph
6
y=0.4593e246.3x
R2=0.55
Grain yield, Mg/ha
5
Winter Wheat
24 locations in
Oklahoma
1998-2001
4
3
2
1
0
0
0.002
0.004
0.006
0.008
0.01
P erkins N&P, 1998
P erkins S&N, 1998
Tipto n S&N, 1998
P erkins N&P, 1999
Experiment 222, 1999
Experiment 301, 1999
Efaw A A , 1999
Experiment 801, 1999
Experiment 502, 1999
P erkins N&P, 2000
Experiment 222, 2000
Experiment 301, 2000
Efaw A A , 2000
Experiment 801, 2000
Experiment 502, 2000
Hennessey, A A , 2000
P erkins N&P, 2001
Experiment 222, 2001
Experiment 301, 2001
Efaw A A , 2001
Experiment 801, 2001
Experiment 502, 2001
Hennessey, A A , 2001
INSEY=NDVI/Days from planting to sensing GDD>0
12
6.0
Soft White
Winter Wheat
7 locations in
Virginia, 2001
4.0
3.0
y = 0.5643e
R2 = 0.6476
Grain yield, kg/ha
Grain yield, Mg/ha
5.0
Spring Wheat
4 locations in
Ciudad Obregon, MX
2001
10
279.42x
2.0
1.0
8
6
4
y = 649.34e175.42x
R2 = 0.75
2
0.0
0
0.002
0.004
0.006
INSEY=NDVI/days from planting, GDD>0
0.008
0
0
0.005
0.01
0.015
INSEY=NDVI/days from planting to sensing (GDD>0)
0.02
Oklahoma
100 lb N/ac
N uptake, lb/ac
45 bu/ac, 2.5% N in the grain
75 lb N/ac
50 lb N /ac
days with GDD>0?
October
0
February
120
June
240 days
INSEY: Rate of N uptake over 120 days, > ½ of the total growing days
and should be a good predictor of grain yield
In-Season
Estimated
Yield (INSEY)
NDVI at F5
=
days from planting to F5, GDD>0
+Good predictor of final grain yield
+Requires only one sensor reading
+Work over different regions/biotypes
6
y=0.4593e
2
Tipto n S&N, 1998
P erkins N&P , 1999
R =0.55
5
Grain yield, Mg/ha
P erkins N&P , 1998
P erkins S&N, 1998
246.3x
Experiment 222, 1999
Experiment 301, 1999
Efaw A A , 1999
4
Experiment 801, 1999
Experiment 502, 1999
P erkins N&P , 2000
Experiment 222, 2000
Experiment 301, 2000
3
Efaw A A , 2000
Experiment 801, 2000
Experiment 502, 2000
2
Hennessey, A A , 2000
P erkins N&P , 2001
Experiment 222, 2001
Experiment 301, 2001
Efaw A A , 2001
1
Experiment 801, 2001
Experiment 502, 2001
0
0
0.002
0.004
0.006
0.008
INSEY=NDVI/Days from planting to sensing GDD>0
0.01
Hennessey, A A , 2001
Units:
N uptake, kg ha-1
day-1 where GDD>0
•Fertilize whole field with 40 lbs N/ac preplant
•Before exiting the field, apply one strip with 80
(non-N-limiting)
Why is it important to know
the RI for a field?
C.V. = 31
Year
1999
1996
1993
1990
1987
1984
1981
1978
y = -0.0869x + 26.708
1975
60
50
40
30
20
10
0
1971
Yield (bu/acre)
Figure 1a. Wheat yield of plots receiving no
added fertilizer N 1971-2001 (Lahoma, OK)
So, what’s in it for the
farmer?
40
35
30
25
20
15
10
5
0
Yld Loss
Total Loss
Excs N Loss
Time
2000
1998
1996
1994
1992
1990
1988
1986
1984
1982
1980
1978
1976
1974
Ave Loss/ac/yr = $9.77
1971
Loss/acre ($)
Loss from fertilizing with 80 lb N/acre for a 40 bu yield
(Lahoma 502)
RINDVI=NDVI-N-non-limiting/NDVI-farmer check
RINDVI and RIHARVEST
3
2.5
y = 1.06x + 0.18
R2 = 0.56
2
RI Harvest
1998
1999
1.5
2000
2001
1
0.5
0
0
0.5
1
1.5
2
2.5
3
RI NDVI
•
•
•
Strong correlation between RINDVI (vegetative stages) and RI HARVEST
Accurately predict the crop’s ability to respond to N
RINDVI may refine whether or not N should be applied, how much,
and expected NUE
YPMAX
6
y=0.4593e246.3x
R2=0.55
5
Grain yield, Mg/ha
YPN
YP0
4
3
2
1
0
0
0.002
0.004
0.006
0.008
INSEY=NDVI/Days from planting to sensing GDD>0
0.01
P erkins N&P, 1998
P erkins S&N, 1998
Tipto n S&N, 1998
P erkins N&P, 1999
Experiment 222, 1999
Experiment 301, 1999
Efaw A A , 1999
Experiment 801, 1999
Experiment 502, 1999
P erkins N&P, 2000
Experiment 222, 2000
Experiment 301, 2000
Efaw A A , 2000
Experiment 801, 2000
Experiment 502, 2000
Hennessey, A A , 2000
P erkins N&P, 2001
Experiment 222, 2001
Experiment 301, 2001
Efaw A A , 2001
Experiment 801, 2001
Experiment 502, 2001
Hennessey, A A , 2001
YPMAX
YP0
YPN (RI=2.0)
YPN (RI=1.5)
3
1
Predict RI
2
Predict YP0
3
Predict YPN based on RI
4
Fertilizer N = GNUPYPN –GNUPYP0/0.7
2
1
4