Powerful Google developer tools for immediate impact! (2023-24 C)
Western crop science society of america conference oregon, 2013 - greenseeker-pocket sensor
1. Western Crop Science of America
Conference
June 10-13, 2013
Pendleton, OR
Olga Walsh
Assistant Professor, Soil Nutrient Management
Western Triangle Agricultural Research Center
Montana State University
Evaluation of Sensor-Based Technologies and N
Sources for Spring Wheat
2. Cooperators
Mr. Robin Christiaens, Research Asossiate,
WTARC, Conrad, MT
Dr. Mal Westcott, Professor and Supt., WARC,
Corvallis, MT
Ms. Martha Knox, Research Specialist,
WARC, Corvallis, MT
Lindsey Martin, Producer, Pendroy, Teton
County, MT
3. Objective
To evaluate two sensors (GreenSeeker, and Pocket
Sensor) for developing NDVI-based topdress fertilizer N
recommendations in spring wheat in Montana
To determine whether sensor-based recommendations
have to be adjusted depending on what N fertilizer
source (liquid UAN, or granular urea) is used
4. Materials and Methods
3 experimental sites: 2 dryland (WTARC, and on-farm
study (Lindsey Martin, Pendroy, Teton County), and 1
irrigated (WARC)
Choteau spring wheat variety
Comprehensive soil sampling - preplant fertilizer
application rates for all nutrients except N
Plot size: 5’x 25’.
5. Materials and Methods
4 replications
4 preplant N rates (20, 40, 60, and 80 lbs N ac)
2 topdress N fertilizer sources (granual – urea, 46-0-0,
and liquid – urea ammonium nitrate (UAN), 28-0-0)
Topdress N fertilizer rate determined using NDVI
obtained using GreenSeeker and Pocket Sensor at
Feekes 5 growth stage
7. GreenSeeker
Real-time active light source sensor
Emits light at 670nm (red) and 780nm (NIR)
Measures crop canopy reflectance at 200 readings /sec
Outputs Normalized Difference Vegetative Index
(NDVI)
Equivalent to a plant physical examination
NDVI is correlated with:
Plant biomass
Plant chlorophyll
Crop yield
Water stress
Plant diseases, and
Insect damage
8. Pocket Sensor
Real-time active light source sensor
Pre-calibrated to GreenSeeker
Can be calibrated to any NDVI sensor
NDVI can be directly compared independent of what
sensor is used to sense the crop
Cost ~ 25% of GreenSeeker cost
No storage capability
2010
2012
9. Concept Summary
1. How much
biomass is
produced ?
2. What Yield is
attainable without
addition of N?
3. How
responsive is
the crop to N?
4. What Yield is
attainable with
addition of N?
YPN = INSEY*RI
NDVI = (NIR-red)/(NIR+red)
INSEY = NDVI/GDD>0
RI = NDVI (NRS) /NDVI (FP)
10. Yield
Trt
Preplant
N
Fertilizer
Rate, lb
N ac-1
Topdress
N
Fertilizer
Source
Mean spring wheat grain yield, lb ac-1
2011 2012
WTARC WARC WTARC WARC
Martin
1 0 - 829 (f) 1822 (e) 4229 (d) 3512 (f) 1698 (c)
2 220 urea 2378 (a) 3335 (abc) 4433 (d) 4981 (e) 1837 (bc)
3 20 urea 1369 (e) 2488 (d) 4797 (c) 5121 (de) 1995 (ab)
4 40 urea 1388 (e) 3061 (bc) 5178 (a) 5299 (bcde) 1996 (ab)
5 60 urea 1662 (cd) 3453 (ab) 5140 (abc) 5746 (abc) 2072 (ab)
6 80 urea 1925 (b) 3558 (a) 5262 (a) 5273 (cde) 2115 (a)
7 20 UAN 1298 (e) 2907 (cd) 4824 (bc) 5563 (abcd) 1997 (ab)
8 40 UAN 1465 (de) 3136 (abc) 4958 (abc) 5674 (abcd) 2065 (ab)
9 60 UAN 1771 (bc) 3004 (bc) 4951 (abc) 5862 (ab) 1980 (ab)
10 80 UAN 1935 (b) 3210 (abc) 5160 (ab) 5871 (a) 2027 (ab)
•Preplant fertilizer N was applied as urea.
• ** Topdress fertilizer N rates for Treatments 3-10 were determined based on the NDVI values obtained using GreenSeeker.
•The means in the same column followed by the same letter are not significantly different, p<0.05.
Consistently, there were no substantial differences in grain yields associated with topdress
fertilizer N source (urea vs UAN) at any of 5 site-years. This indicated that topdress N
fertilizer rates do not need to be adjusted based of fertilizer sources used, i.e. the same N
rates should be prescribed whether urea or UAN is applied.
11. Prescribed N rates vs Yield
Site-
year
Trt
Preplant
N rate,
lb N ac-1
GS
NDVI
Prescribed
topdress N
rate, lb N
ac-1
Total N
rate,
lb N
ac-1
N rate
difference
,
lb N ac-1
Grain
yield,
bu ac-1
Yield
gain,
bu ac-1
1
WTARC,
2012
2 220 0.3 62 282
-178
74 (d)
+14
6 80 0.5 24 104 88 (a)
2
Martin,
2012
5 60 0.3 0 60
+37
35
=6 80 0.4 17 97 35
3
WTARC,
2011
6 80 0.4 9 89
- 42
32 (b)
+10
7 20 0.3 27 47 22 (e)
4
WTARC,
2012
3 20 0.5 13 33
+91
80 (c)
+8
6 80 0.5 24 124 88 (a)
• As in the first growing season, in 2012, Spring Wheat (Canada)
Algorithm and Generalized Algorithm did not prescribe any topdress N
fertilizer to be applied at any of the 3 experimental locations.
• The recommended application rates generated by the Sensor-Based
Nitrogen Optimization Algorithm (USA/Canada/Mexico) ranged from of 0
lb N ac-1 at Martin in 2012 to 99 lb N ac-1 at WARC at 2012, depending
on the NDVI values
• Sensor-based generated topdress N rates did not always optimize grain
yields.
• Some rates were excessive (1), (2); Some rates did not make sense (3);
Some rates made sense (4)
12. Results: GS vs PS
y = 2.0825x2 - 1.0582x + 0.4613
R² = 0.50
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.10 0.20 0.30 0.40 0.50 0.60 0.70
PocketSensorNDVI
GreenSeeker NDVI
Relationship between
GreenSeeker NDVI and
Pocket Sensor
NDVI, WTARC and
WARC, 2011, and
WTARC, WARC, and
Martin, 2012.
y = 1.2049x - 0.1196
R² = 0.91
0.25
0.30
0.35
0.40
0.45
0.50
0.37 0.39 0.41 0.43 0.45 0.47 0.49
PocketSensorNDVI
GreenSeeker NDVI
Relationship between
GreenSeeker NDVI and
Pocket Sensor
NDVI, WTARC and
WARC, 2011, and
WTARC, WARC, and
Martin, 2012. NDVI
values are averaged by
treatment over all five
site-years.
There was a strong relationship observed between NDVI values obtained with GreenSeeker and with Pocket
Sensor. Understandably, the relationship was improved dramatically when mean NDVI values averaged by
treatment were used (R2 = 91 vs R2 = 50) importance of replication when taking the canopy reflectance
readings because it helps to account for spatial variability present within a field
13. Results: GS vs Yield
y = 3455.9x + 640.46
R² = 0.97
y = -2E+06x2 + 2E+06x - 361440
R² = 0.80
y = 24288x2 - 17481x + 7558
R² = 0.39
y = 7452.3x - 1134.3
R² = 0.96
y = 7408.9x - 917.65
R² = 0.75
0
20
40
60
80
100
120
0.2 0.3 0.3 0.4 0.4 0.5 0.5 0.6 0.6 0.7
Springwheatgrainyield,buac-1
GreenSeeker NDVI
martin-12 warc-12 wtarc-12 wtarc-11 warc-11
•Strong linear relationship was observed between GS NDVI and yields at 4 of 5
site-years - GS NDVI values predicted 75 to 97 % of variation in grain yields.
• At WARC in 2012, 80% of variation in yields was explained by variation in NDVI;
however, unexpectedly, the observed trend was: the higher NDVI, the lower the
yield – Negative slope.
•Labus et al. (2002): “early season NDVI were not consistent indicators of wheat
yields”. Extensive study in Montana, - the strength of NDVI-yield relationships was
highly dependent on site-specific and region-specific characteristics.
• Crop reflectance measurements aim not to predict yield, but estimate yield
potential.
14. Results: GS vs Yield
y = -157329x2 + 150526x - 32458
R² = 0.91
30
35
40
45
50
55
60
65
0.37 0.39 0.41 0.43 0.45 0.47 0.49
Springwheatgrainyield,buac-1
GreenSeeker NDVI
• Relationship between mean GreenSeeker NDVI values and mean
spring wheat grain yields (averaged over site-years) at WTARC and
WARC, 2011, and at WTARC, WARC, and Martin, 2012.
• GreenSeeker NDVI was able to predict 91 % of variation in spring
wheat grain yields across site-years (R2 = 0.91)
15. PS NDVI vs Yield
y = -79220x2 + 74697x - 14020
R² = 0.96
1500
2000
2500
3000
3500
4000
0.32 0.37 0.42 0.47
Springwheatgrainyield,lbac-1
Pocket Sensor NDVI
Relationship between mean Pocket Sensor NDVI values and mean spring wheat grain yields (averaged over site-years)
at WTARC and WARC, 2011, and at WTARC, WARC, and Martin, 2012.
• Robust linear relationship was also evident between PS NDVI and yields at
3 of 5 site-years in 2011 and 2012, where spring wheat grain yield was
predicted midseason with 83 to 92 % accuracy.
• When averaged across site-years, PS NDVI values collected at Feekes 5
growth stage were able to predict 96 % of variation in spring wheat grain
yields.
16. Total N applied vs Yield
Strong polynomial relationships between the total amounts on N applied
(preplant plus topdress) was observed at all 5 site-years.
The highest topdress N rates prescribed did not result in increase in grain
yield, but in most cases, caused yield reduction.
y = -0.0007x2 + 0.264x + 15.397
R² = 0.97
y = -0.0014x2 + 0.4024x + 35.272
R² = 0.81
y = -0.0012x2 + 0.2704x + 73.691
R² = 0.86
y = -0.0021x2 + 0.5048x + 71.991
R² = 0.60
y = -0.0004x2 + 0.0972x + 30.071
R² = 0.66
0
20
40
60
80
100
120
0 50 100 150 200 250
Grainyield,buac-1
Total N rate applied, lb N ac-1
wtarc-11
warc-11
wtarc-12
warc-12
martin-12
17. Protein
Trt Preplant N
Fertilizer
Rate, lb N
ac-1*
Topdress
N Fertilizer
Source**
Mean grain protein content, %
2011 2012
WTARC WARC WTARC WARC Martin
1 0 - 9.5 (bc) 14.1 (bcd) 9.6 (a) 11.5 (e) 14.3 (b)
2 220 urea 9.7 (ab) 16.4 (a) 15.4 (a) 14.9 (a) 16.7 (a)
3 20 urea 9.5 (bc) 14.6 (bc) 10.5 (a) 14.0 (abc) 15.3 (a)
4 40 urea 9.7 (a) 13.1 (d) 11.1 (a) 14.1 (abc) 15.7(a)
5 60 urea 9.4 (c) 15.2 (ab) 11.6 (a) 13.9 (bc) 15.9(a)
6 80 urea 9.6 (ab) 15.1 (ab) 12.8 (a) 14.4 (ab) 16.2 (a)
7 20 UAN 9.6 (abc) 14.5 (bc) 10.6 (a) 12.9 (d) 15.6 (a)
8 40 UAN 9.5 (abc) 13.5 (cd) 11.0 (a) 13.4 (cd) 15.8 (a)
9 60 UAN 9.5 (bc) 14.3 (bcd) 12.0 (a) 14.0 (abc) 16.0 (a)
10 80 UAN 9.5 (bc) 15.1 (ab) 13.0 (a) 14.3 (abc) 16.4 (a)
The highest preplant N rate was associated with the best protein content.
Topdress N rates did not optimize grain protein at any of the 5 site-years.
At irrigated site, WARC, urea resulted in significantly higher grain protein content,
compared to UAN (15.4 vs 14.3)
At dryland sites, WTARC and Martin– no significant differences in grain protein
content associated with N source.
18. Lessons Learned
In both growing seasons, the rates generated by the
USA/Canada/Mexico Algorithm were not appropriate
for grain yield optimization
Much higher rates were prescribed for the irrigated
site (WARC) compared to those for dryland sites
WTARC and Martin. This makes sense since the
expected yield potential at the irrigated site was
much greater
However, grain yields obtained at WTARC were just
as high as at WARC yield potential was either
underestimated at WTARC or overestimated at
WARC
Separate algorithms developed for dryland spring
wheat and for irrigated spring wheat production
systems
19. Both sensors perform well and are useful in
predicting spring wheat grain yield potential mid-
season
Algorithms developed in other regions do not provide
the topdress N rates appropriate for Montana spring
wheat varieties and growing conditions
It is expected that this study will continue for one
more growing season at 3 experimental locations to
expand database and to summarize results
Future studies are needed to pinpoint the rate of N
loss due to volatilization and immobilization and
other pathways in Montana wheat production
systems for improved N recommendations.
Lessons Learned
20. Protein Yield concept
Spring wheat is produced for its quality,
represented by high grain protein content.
Evaluating NUE in spring wheat should
take into an account both grain yield and
protein content.
Combining yield and protein into protein
yield, as proposed by Jackson (1998)
makes sense because N is vital to both
yield and protein production.
Protein Yield = grain protein content (%) * grain yield (lb ac-1)