Active-Crop Sensor Calibration Using the Virtual-Reference Concept K. H. Holland (Holland Scientific) J. S. Schepers (USDA-ARS, retired) 8 th ECPA Conference.

Active-Crop Sensor Calibration Using the

Virtual-Reference Concept

K. H. Holland (Holland Scientific)

J. S. Schepers (USDA-ARS, retired)

8th ECPA Conference 2011

“N-Rich” Reference

• Postage-stamp calibration

• Ramped calibration strip

• Randomized calibration block(field strips)

check

N-richother N rates

Modified Postage Stamp

• Combine-width plots

• Randomized except check plot

• Each block of treatments should have minimal soil variability

• Repeated replications

check

Modified Calibration Ramp

• Standard ramp of plots

• No randomization

• Each N rate in the ramp has a nearby check and adequate N reference

• Could randomize N rates in the ramp strip

check

N-Rich

Commercial applicators and large producers - - TELL US :

• N-rich strips are problematic– May be hard to locate (legal problems)– Need to move each year

• Can not expect operators to understand how the algorithm and sensor calibration work

• Need a “turn-key” approach that does not require N-rich strip or highly-skilled operator

• Algorithm needs to be simple, versatile, and easy to adapt for local conditions

Algorithms

1) based on predicted yield potential(Oklahoma State University) (GreenSeeker)

2) based on producer-set minimum and maximum N rates (Europe & Missouri)

3) based on extension of crop N-response function (Holland and Schepers)

Note: All algorithms use sensor data that are normalized to “healthy crops” .

In-Season N Management

Crop vigor during the growing season

is proportional to

yield at harvest

How to Characterize Healthy Crops ?

• N-Rich Strip (or Ramp Calibration Strip)average (as with plot studies)

programmed (highest 3 consecutive seconds)

• Normal Field Transectsidentify healthy plants from frequency distribution of all plants (histogram) (MS Excel)

3-second Running Average

0

1

2

3

4

5

6

1 459 917 1375 1833 2291 2749 3207 3665 4123 4581 5039 5497 5955 6413

Number of Readings

CI

red

-ed

ge

Mexico - White Corn, 2010 Crop Circle

600 m @ ~6 kmph

0.6

07

0.8

15

81

.02

46

1.2

33

41

.44

22

1.6

51

1.8

59

82

.06

86

2.2

77

42

.48

62

2.6

95

2.9

03

83

.11

26

3.3

21

43

.53

02

3.7

39

3.9

47

84

.15

66

4.3

65

44

.57

42

4.7

83

4.9

91

85

.20

06

5.4

09

45

.61

82

5.8

27

6.0

35

8

0

50

100

150

200

250

0%

20%

40%

60%

80%

100%

120%

Histogram

Bin (vegetation index)

Fre

qu

en

cy

Mexico, 2010

0.6

07

0.8

15

81

.02

46

1.2

33

41

.44

22

1.6

51

1.8

59

82

.06

86

2.2

77

42

.48

62

2.6

95

2.9

03

83

.11

26

3.3

21

43

.53

02

3.7

39

3.9

47

84

.15

66

4.3

65

44

.57

42

4.7

83

4.9

91

85

.20

06

5.4

09

45

.61

82

5.8

27

6.0

35

8

0

50

100

150

200

250

0%

20%

40%

60%

80%

100%

120%

Histogram

Bin (vegetation index)

Fre

qu

en

cy

Mexico, 2010

95%“Happy Corn”

Gustavo Field

0

1

2

3

4

5

6

7

1 453 905 1357 1809 2261 2713 3165 3617 4069 4521 4973 5425 5877 6329

Number of Readings

CI

red

-ed

ge

Mexico - White Corn, 2010

600 m @ ~6 kmph

Crop Circle

95 Percentile

3-second Running Average = 5.057

95 Percentile = 5.206

3% lower

0.6

07

0.8

15

81

.02

46

1.2

33

41

.44

22

1.6

51

1.8

59

82

.06

86

2.2

77

42

.48

62

2.6

95

2.9

03

83

.11

26

3.3

21

43

.53

02

3.7

39

3.9

47

84

.15

66

4.3

65

44

.57

42

4.7

83

4.9

91

85

.20

06

5.4

09

45

.61

82

5.8

27

6.0

35

8

0

50

100

150

200

250

0%

20%

40%

60%

80%

100%

120%Histogram

Bin (vegetation index)

Fre

qu

en

cy

Mexico, 2010

95%50%Field Average

0.6

07

0.8

85

41

.16

38

1.4

42

21

.72

06

1.9

99

2.2

77

42

.55

58

2.8

34

23

.11

26

3.3

91

3.6

69

43

.94

78

4.2

26

24

.50

46

4.7

83

5.0

61

45

.33

98

5.6

18

25

.89

66

Mo

re

0

50

100

150

200

250

0%

20%

40%

60%

80%

100%

120%Histogram

Bin (vegetation index)

Fre

qu

en

cy

95%50%

4.47

5.21SI = = 0.85

N CreditsPreplant N

EONRProducer Optimum

N Accumulation(based on growth stage)

Sufficiency Index

Back-Off Strategy SI to start cutback SI to cut-off

Algorithm

SpatialSoil / Topography

Adjustment0

2

4

6

8

10

12

14

0 50 100 150 200 250 300

Fertilizer N Rate, kg/ha

Yie

ld, M

g/h

a

FieldReference

Holland K.H. and J.S. Schepers. 2010. Derivation of a variable rate nitrogen application model for in-season fertilization of corn. Agronomy Journal 102:1415-1424.

S e e

Nappl = ( Nopt – Ncred )

√

√ (1 – SI)

∆ SI

Farmer Rateor NEONR

6 3310

20

40

60

80

100

120

140

160

180

200

Distance or Time

N ra

te A

pplie

d, k

g/ha

Mexico, 2010

6 3310

20

40

60

80

100

120

140

160

180

200

Distance or Time

N ra

te A

pplie

d, k

g/ha

Mexico, 2010

Uniform Rate

N Rates (0, 50, 100, 150, 200 kg/ha)

OptRx

Check Plot

SoybeanPrevious

Year

Histogram

0

50

100

150

200

250

0.27

1

0.38

43142

86

0.49

76285

71

0.61

09428

57

0.72

42571

43

0.83

75714

29

0.95

08857

14

1.06

42

1.17

75142

86

1.29

08285

71

1.40

41428

57

1.51

74571

43

1.63

07714

29

1.74

40857

14

1.85

74

1.97

07142

86

2.08

40285

71

2.19

73428

57

CI red-edge

Fre

qu

ency

0%

20%

40%

60%

80%

100%

120%Frequency

Cumulative %

95 Percentile

3-Second Running Average

0.000

0.200

0.400

0.600

0.800

1.0001.200

1.400

1.600

1.800

2.000

1 372 743 1114 1485 1856 2227 2598 2969 3340 3711 4082 4453 4824

Sample Number

CI

red

-ed

ge

CI red-edge values : 95 percentile 1.9853-second average 1.889

5% lower

GreenSeeker

0

20

40

60

80

100

120

140

160

180

200

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31

200 N

150 N

100 N

50 N

0 N

Chlorophyll Index Bins (0.7 to 2.2)

Oc

cu

rre

nc

eIrrigated Corn - 2009

V9 Growth Stage

95 percentile

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 240

10

20

30

40

50

60

70

80

90

100

Drive First

Drive & Apply

Plot Order (400 m)

N R

ate

(kg

/ha)

V12

Virtual Reference Strip (0-200 kg N/ha preplant)

check

0 100 200 300 400 500 6000

20

40

60

80

100

120

140

160

180

200

VRS firstD&A (H to L)D&A (L to H)

Distance (m)

N R

ate

(kg/

ha)

Mexico, 2010 Drive and Apply

There’s Probably a Lot More Information in a Histogram than We Realize !

• Where’s it at ?

• How to get it out ?

Mexico 2010 - Irrigated Corn

V5 Growth Stage

SI = 0.7

Mexico - White Corn, 2010

600 m @ ~6 kmph

Crop Circle

95%

Cut-back level

Conclusions• The virtual reference concept offers producers a convenient

approach to quantify the vigor and chlorophyll status of crops for in-season N applications.

• Histograms of active sensor data and related analyses offer a quick glimpse of where to focus management efforts.

New sensors and tools will be

needed to help fine tune

management decisions.

Jim Schepers

402-310-6150

[email protected]

Historic Perspective

• N-Rich treatment was initially used to normalize data from plot studies and allow leaf N concentration comparisons across time, fields, cultivars, etc. (1988)

• Extended to normalization concept to SPAD meters. (1990)

• Adapted to field situations and N-Rich strips to accommodate crop canopy sensors. (~2000)

Historic Perspective

• N-Rich plot concept extended to postage stamp arrangement with multiple N rates. (2002)

• Ramped calibration strip with multiple N rates introduced. (2005)

• Need for active sensor calibration technique to accommodate commercial applications. (2007)