Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
8/1/2017
Prof. R. Khosla, Colorado State University 1
Nitrogen Management
Raj KhoslaColorado State University
https://www.euractiv.com/wp‐content/uploads/sites/2/2016/10/Digital‐farming.jpg
Nitrogen management
?I
Adapted from: Lassaletta et al. 2014 Environmental Research Letters
Nitrogen Use Efficiency
(<50%)
8/1/2017
Prof. R. Khosla, Colorado State University 2
Ammoniumfertilizer
NH3
NH4+
NH3
N2
NO2- NO3
-
N2O,NO, NO2
OM
Urea
N2O
RunoffLeaching
NH3
Forms Of Nitrogen
Only two are plant available: NO3- and NH4
+
NO2‐
NO2
HNO2
HNO3
NO‐
N2
N2O
NH2OH
N2H4
NH3NO3
‐
NH4+
How do we manage nitrogen for crop production?
CSUAgricultural Research,
Development & Education Center
Eastern Colorado Research Center
Arkansas Valley Research Center
San Luis Valley Research Center
Plainsman Research Center
Southwestern Research Center
Western Research Center
200 lbs
150 lbs0 lbs
50 lbs
200 lbs150 lbs
0 lbs50 lbs
8/1/2017
Prof. R. Khosla, Colorado State University 3
Calculating the Optimal N Rate
Nitrogen Rate (lbs/Acre)
Gra
in Y
ield
(B
u/A
cre)
Optimal Range
N rate = 35+ (1.2 X EY (bu/ac))
N ManagementState N Rate Recommendation
CO 35+ (1.2 X EY (bu/ac)) – (8 X Average ppm NO3 N in Soil) – (.14 X EY (bu/ac) X %OM)‐ Other N
Credits
KS (1.6X YG (bu/ac))‐(%OM X 20) ‐ Profile N ‐ Legume N‐ other N Credit
OH ‐27 + (1.36 X Yield Potential (bu/ac) ‐100) – N credit (lb/ac) or 110 + [1.36 X (Yield potential
(bu/ac) ‐100)] – N credit (lb/ac)
IN ‐27 + (1.36 X Yield Potential (bu/ac) ‐100) – N credit (lb/ac) or 110 + [1.36 X (Yield potential
(bu/ac) ‐100)] – N credit (lb/ac)
MI ‐27 + (1.36 X Yield Potential (bu/ac) ‐100) – N credit (lb/ac) or 110 + [1.36 X (Yield potential
(bu/ac) ‐100)] – N credit (lb/ac)
MO Fertilizer N Recommendation (lbs/ac) – Pre‐plant N Test Credits (lbs/ac)
MT N Fertilizer YG Recommendation (lbs/ac) ‐ PSNT NO3‐ (lbs/ac) *Wheat
ND Fertilizer N recommendation (lbs/ac)‐ Soil Nitrate Concentration (lbs/ac)‐ N Credits (lbs/ac)
NE 35+ [1.2 X EY (bu/ac)] – (8 X Average ppm NO3 N in Soil) – (.14 X EY (bu/ac) X %OM)‐ other N
credits
OR YG (bu/ac) X Required N Protein Goal (lb/ac) – Residual Soil N (lb/ac) *Wheat
PA EY (bu/ac) – ( (Manure since last harvest (lb/ac) + Previous Crop Factor (lb/ac) + Three year
Manure History Factor (lb/ac)) X Soil Nitrate (lb/ac))
SD YG bu/ac X 1.2 – Soil NO3 (lbs/ac) – Manure N (lb/ac) + no‐till Adjustment
VA EY (bu/ac) – ((Applied Manure Factor Last Year (lb/ac) + Leguminous Crop Factor (lb/ac) +
Manure History Factor (lb/ac)) * (PSNT (ppm))
IA N Rate Web Application
WI N Rate Web Application
MN N Rate Web Application
IL N Rate Web Application
ND N Rate Web Application
Common VariablesState N Rate Recommendation
CO 35+ (1.2 X EY (bu/ac)) – (8 X Average ppm NO3 N in soil) – (.14 X EY (bu/ac) X %OM)‐ other N Credits
Estimated Yield (EY)
Soil N Test
N Credits
Web Application
Max Economic Return To Nitrogen
8/1/2017
Prof. R. Khosla, Colorado State University 4
+/- 2 bu/A from the mean
%+/- 10 bu/A from the mean
Only 36%
Mean: 182.5 bu/A
>192.5 bu/A
40%Under-fertilized
<172.5 bu/A
24%Over-fertilized
Yield MapPixels = Average?
8%
high
med
med
low
low
Management Zones are delineated on farm fields by classifying the field into different sections or zones.
* CSU, USDA-ARS, Centennial Ag Inc.
Based on the research conducted in Colorado*
N rate = 35+ (1.2 X EY (bu/ac))
Average
In 9 out of 10 site years we can separate low from high zone but NOT low from medium or medium from high zones based on grain yield
Mean grain yield across MZs
16
12
8
4
0
a a b
Low Medium High
Management zones
Gra
in y
ield
(M
g ha
-1)
12
9
6
3
0
ab b
Low Medium High
Management zones
Gra
in y
ield
(Mg
ha -1
)
a
20
15
10
5
0
b b
Low Medium High
Management zones
Gra
in y
ield
(M
g ha
-1)
a
Source: Koch et al. 2004
Low Productivity (Zone 3)
MediumProductivity (Zone 2)
High Productivity (Zone 1)
The three data layers
Aerial Imagery
Topography
Farmer’s experience
are stacked as GIS layersto delineate the zone
Traits such as dark color, low-lying topography, and historic high yields were designated as a zone of potentially high productivity or high zone
Delineating management zones…
8/1/2017
Prof. R. Khosla, Colorado State University 5
Macro-variability
Micro-variability
Low Productivity (Zone 3)
MediumProductivity (Zone 2)
High Productivity (Zone 1)
Landsat 8
Worldview‐2 Natural Color display (Bands 5‐3‐2)
Captured 9/13/2013Spatial Resolution: 2m
Boulder Creek Flood Plain
Landsat 8Natural Color display (Bands 4‐3‐2)
Captured 9/17/2013Spatial Resolution: 30mBoulder Creek Flood Plain
500 ft
NDVI = NIR- Red / NIR + Red
How to translate NDVI readings into N rate recommendations?
Nitrogen Algorithm(s)
One of the first modern applications of remote sensing and it’s use…
to determine N rates by estimating yield using NDVI
NDVI provides an estimate of above ground biomass
First Nitrogen Application Algorithms were derived from yield estimates using remote sensing
Big turning point in the history of data‐driven N management
Raun et al 2001. In-Season Prediction of Potential Grain Yield in Winter Wheat Using Canopy reflectance
Cumulative Growing Degree Days (GDD)
Abo
ve G
roun
d B
iom
ass
NDVI Time 1
NDVI Time 2
Expected Yield (EY) = (NDVI T1 + NDVI T2) / GDD(INSEY)
• In 2002, Raun et al., developed the Nitrogen Fertilization Optimization Algorithm (NFOA)
• a multi-step process:
1. Generate Yield Prediction Equation (YP0) from the INSEY and previous year’s yield data
2. Field data collection for N response
Nitrogen Algorithm(s)
Raun et al 2002 Improving NUE in Cereal Grain Production with Optical Sensing and Variable Rate Application
Generate Yield Prediction Equation (YP0)
INSEY
Gra
in Y
ield
(kg
/ha)
YPo= a X e(b X INSEY)
8/1/2017
Prof. R. Khosla, Colorado State University 6
N-Rate Field Experiment
Collect sensor readings
Collect temperature data for GDD
Nitrogen Algorithm(s)
NDVI 0.85
NDVI 0.59
NDVI 0.73
Response Index (RI) = NDVIRich / NDVIReferencePotential for yield increase ~44%
with additional NRI= .85 / .59 = 1.44
RI= .85 / .73 = 1.16 Potential for yield increase ~16% with additional N
How much additional N?
3. Calculate Yield Potential with added N fertilizer (YPN)
YPN = YP0 * RI
4. Compute Grain N uptake at YP0 & YPN
GNUP_YP0 = YP0 x % N Grain
GNUP_YPN = YPN x % N Grain
5. Final N Rate = (GNUP_YPN – GNUP_YP0) / NUE
Nitrogen Algorithm(s)
Raun et al 2002 Improving NUE in Cereal Grain Production with Optical Sensing and Variable Rate Application
Limitations:
I. NDVI saturates at high LAI values
II. This algorithm does not account for location of plant in the field
VEGETATION INDICES EQUATION
Normalized Green Index (GRI)
Normalized Red Edge Index (NREI)
Normalized Difference Red Edge Index (NDREI)
Green Chlorophyll Index (GRI)
Red Edge Chlorophyll Index (RECI)
Green Soil Adjusted Vegetation Index (GSAVI)
Green Optimal Soil Adjusted Vegetation Index (GOSAVI)
Modified Chlorophyll Absorption in Reflectance Index
G/(NIR+RE+G)
G/NIR+RE+G
RE/(NIR + RE + G)
RE/(NIR + RE + G)
(NIR – RE)/(NIR + RE)
(NIR – RE)/(NIR + RE)
NIR/G‐1
NIR/RE‐1
1.5 * [(NIR – RE)/(NIR + RE +.5)]
(1 + .16)(NIR – G)(NIR + G + .16)
[(NIR – RE) ‐ .2 *(NIR – G)]/(NIR/RE)
Cao et al 2014: Active Canopy Sensing of Winter Wheat Nitrogen Status: An evaluation of two Sensor Systems
New Vegetation Indices to Detect N StatusVEGETATION INDICES Equation
Normalized Green Index G/(NIR + RE +G)
Normalized Red Edge Index RE/(NIR + RE +G)
Normalized NIR Index NIR/(NIR + RE +G)
Red Edge Ratio Vegetation Index NIR/RE
Green Ratio Vegetation Index NIR/G
Red Edge Green Ratio Vegetation Index RE/G
Green Difference Vegetation Index NIR‐G
Red Edge Difference Vegetation Index RE‐G
Normalized Difference Red Edge (NIR‐RE)/(NIR+RE)
Green Normalized Difference Vegetation Index (NIR‐G)/(NIR+G)
Red Edge GNDVI (RE‐G)/(RE+G)
Green Wide Dynamic Range Vegetation Index (a*NIR‐G)/(a*NIR+G)(a‐.12)
Red Edge Wide Dynamic Range Vegetation Index (a*NIR‐RE)/(a*NIR + RE)(a‐.12)Optimized Vegetation Index 1 100*(lnNIR‐lnRE)
Modified Double Difference Index (NIR‐RE)‐(RE‐G)
Modified Normalized Difference Index (NIR‐RE)/(NIR‐G)
Green Chlorophyll Index NIR/G‐1
Red Edge Chlorophyll Index NIR/RE‐1
Modified Red Edge Simple Ratio (NIR/RE‐1/SQRT(NIR/RE+1)
Modified Green Simple Ratio (NIR/G‐1)/SQRT(NIR/RE+1)
Modified Enhanced Vegetation Index 2.5* (NIR‐RE/(NIR+6*RE‐.75*G+1)
Modified Normalized Difference Red Edge [NIR‐(RE‐2*G)]/[NIR+(RE‐2*G)]
Modified Chlorophyll Absorption in Redlectance Index [(NIR‐RE)‐.2*(NIR‐G)](NIR/RE)
Modified Transformed CARI 3*[(NIR‐RE)‐.2*(NIR‐G)(NIR/RE)]
Green Soil Adjusted Vegetation Index 1.5*[(NIR‐G)/(NIR+G+.5)]
Red Edge Soil Adjusted Vegetation Index 1.5*[(NIR‐RE/(NIR+RE+.5)]
Green Optimal Soil Adjusted Vegetation Index (1+.16)(NIR‐G)/(NIR+G+.16)
Red Edge Optimal Soil Adjusted Vegetation Index (1+.16)(NIR‐RE)/(NIR+RE+.16)
Red Edge Transformed Vegetation Index .5[120*(NIR‐G)‐200*(RE‐G)]
Grenn re‐Normalized Difference Vegetation Index (NIRE‐RE)/SQRT(NIR+RE)
Limitations:
NDVI saturates at high LAI values
II. Accounting for location of plant in field
8/1/2017
Prof. R. Khosla, Colorado State University 7
N Rate (kg ha-1) = (135.3 x (NDVIRef. / NDVITarget)2) – (134.8 x (NDVIRef. / NDVITarget)) + 1
~96 lbs/a
~96 lbs/a
~96 lbs/a
NDVI
0.41
NDVI
0.41
NDVI
0.41
Coupling site-specific management zones with active proximal sensors
N Rate (kg ha-1) = (135.3 x (NDVIRef. / NDVITarget)2) – (134.8 x (NDVIRef. / NDVITarget)) + 1
~92 lbs/a
~144 lbs/a
~37 lbs/a
High
Medium
N Rate (kg ha-1) = Crop properties + Soil Properties
NDVI
0.41
NDVI
0.41
NDVI
0.41
Low
Crop Based Management
Micro-variabilityMacro-variability
High MZ Medium MZ Low MZ
N management strategies
High Medium Low
UniformRemote Sensing
0 kg
/ha
112
kg/h
a
224
kg/h
a
0 kg
/ha
112
kg/h
a
224
kg/h
a
0 kg
/ha
112
kg/h
a
224
kg/h
a
Management Zones
224 kg/ha168 kg/ha
112 kg/ha
Remote sensing within Management Zones
112
kg/h
a
168
kg/h
a
224
kg/h
a
56 k
g/ha
112
kg/h
a
168
kg/h
a
0 kg
/ha
56 k
g/ha
112
kg/h
a
N management strategies
224 kg/ha
High Medium Low
8/1/2017
Prof. R. Khosla, Colorado State University 8
150
120
90
60
30
0
d
c
b a
d
c
ba
NU
Ea
(kg
Gra
in /
kg N
)
d
c
ba
NUEa
2010 2011 2012
Uniform MZ RS RS + MZ
224
168
112
56
0
N a
pplie
d (k
g/ha
)
Improvement in NUE and reductions in N loadings in the biosphere.
Uniform MZ RS RS + MZ0
28
Dif
fere
nce
in
N a
pp
lied
(kg
/ha)
N loadings Uniform MZ RS RS + MZ
15
10
5
0
a a aa
a a
ba
Yie
ld (
Mg/
ha)
Yield
a a aa
Uniform MZ RS RS + MZ
2010 2011 2012
Dif
fere
nce
in N
2O e
mis
sion
(k
g/h
a/y)
0
0 75
Uniform MZ RS RS + MZ
2010 2011 2012
6.0
4.5
2.0
1.5
0
N2O
em
issi
on (
kg/h
a/y)
N2O emissions
-54%
-55%-50%
Reductions in N2O linked to fertilizer
8/1/2017
Prof. R. Khosla, Colorado State University 9
There will be even more complex soil and crop models that encompass many other sensitive parameters
Machine learning
Model Time ScaleDaily time-step; historical weather data to predict N flux
Weather InputsReal-time; solar radiation; temperature; precipitation
Soil InputsNRCS SSURGO; root depth; slope; SOM; drainage
Cultivar; maturity class; population; yieldCrop Inputs
Management Inputs Tillage; manure; previous crop characteristics
N Fertilizer Inputs Type; rate; timing; pricing
N Rate OutputMass balance; deterministic and stochastic; price risk factors
Graphical OutputN loss and uptake; N dynamics; crop development; fertilizer maps
Method Approach
Sella et al 2016 Adapt-N Outperforms Grower-Selected Nitrogen Rates in Northeast and Midwestern United States Strip Trials,
N Fertilizer Inputs
Management Inputs
Graphical Output
N Rate Output
Nrec=Nexp_yld - Ncrop_now – Nsoil_now
- Nrot_credit – Nfut_gain – loss - Nprofit_risk
8/1/2017
Prof. R. Khosla, Colorado State University 10
Increasing NUE with advanced decision making process
N2O
N
Yield
Thank [email protected]