SiteSpecific and Ecologically Based Weed Management - PowerPoint PPT Presentation

Loading...

PPT – SiteSpecific and Ecologically Based Weed Management PowerPoint presentation | free to view - id: 12ab9f-OTc0N



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

SiteSpecific and Ecologically Based Weed Management

Description:

Lisa Rew, Land Resources and Environmental Science ... Migration (immigration and emigration) Biotic and abiotic impacts on the above ... – PowerPoint PPT presentation

Number of Views:130
Avg rating:3.0/5.0
Slides: 99
Provided by: brucem76
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: SiteSpecific and Ecologically Based Weed Management


1
Site-Specific and Ecologically Based Weed
Management
  • Bruce Maxwell
  • Cooperating Personnel
  • Lisa Rew, Land Resources and Environmental
    Science
  • Nicole Wagner, Land Resources and Environmental
    Science
  • Ed Luschei, University of Wisconsin
  • Perry Miller, Land Resources and Environmental
    Science
  • Daniel Goodman, Department of Ecology
  • David Buschena, Agricultural Economic/Economics
    Department

2
Outline
  • The net return equation
  • Cutting inputs How do you do it without
    increasing risk?
  • Shifting to a more ecologically based form of
    management
  • Adaptive management and the role of precision
    agriculture

3
Economic return to farmer
Net return Gross returns - Costs /ha
yield (kg/ha) seed (/ha) price
(/kg) fertilizer (/ha)
insecticides (/ha) herbicides
(/ha) equipment (/ha)
insurance (/ha) technology
(/ha) Environment
4
Net Return Yield (Price) (CHCTCSCFCTechCE
nv) /ha kg/ha /kg
/ha
Yield f(Nw, etc.)
Temporal Dynamics
Applied Research
  • Next year yield
  • Next year yield

Spatial Dynamics
Basic Research
NR
Nw
NR
Y
5
(No Transcript)
6
Extrapolation from Small Plot Competition
Experiments (SPCE)
Ed Luschei
Other site where we want to predict weed impact
Farm
SPCE
7
Process Model Description
x
x
x
x weed density z growing season precipitation
8
Generation of SPCE Data
Bozeman 1993
Yield (T ha-1)
0
Do this 8 times before fitting a curve
THROUGH ALL 8
Weed density (plants m-2)
9
Calculation of the Experimental and Farm
Economic Threshold (x)
Net Return with weed control
Net Return ( ha-1)
Net Return Without weed control (experimental)
xexp.
Weed density (plants m-2)
10
Predict Farm Threshold at Different Location From
Experiments
Randomly choose precipitation (cm) for farm
location
Precipitation (cm)
Calculate farm weed impact and threshold
11
Each SPCE uses a Randomly Drawn Precipitation
Record
12
Combine Sets of 8 SPCE and Fit Predictive
Equation to Predict Farm Weed Impact
y
xWeed density
zGSP
13
Experimental vs. Farm Economic Threshold
150
125
Farm Threshold
100
75
50
25
0
0
25
50
75
100
125
150
Experimental Threshold
14
Experimental vs. Farm Economic Threshold
150
125
Farm Threshold
100
75
50
25
0
0
25
50
75
100
125
150
Experimental Threshold
15
How to parameterize models at the field scale???
16
Field Scale Research
AJ Bussan Ed Luschei Lee Van Wychen
17
Experimental assessment of cost-effectiveness of
precision wild oat control in spring wheat
  • E. Luschei, L.Van Wychen,
  • B. Maxwell, A.J. Bussan, D. Buschena, D. Goodman

Funding Montana Noxious Weed Trust Fund
18
Questions
  • Does precision weed control make sense?
  • Improved net return?
  • reduced inputs?
  • Can we use this technology for on-farm
    experimentation?

19
Work from1998.
  • Provided observational evidence that
  • patch spray might economically outperform
    broadcast application while using less herbicides
  • Wild oat control might, however, be worse with
    patch spraying

20
1998 Use data prediction and simulation
to examine what if we used this strategy...
Consultant patches
No Spray
Broadcast 1X
21
Blue number above box is predicted mean wild oat
density at harvest
22
(No Transcript)
23
1999 Experiment
  • Consultant maps weeds, 4 sites
  • Create prescription map which has treatment
    replicates (12 reps/site, 3 treatments)
  • Spray according to prescription map
  • Harvest yield with yield monitor equipped combine
  • Find mean yields in replicates
  • Calculate mean field values for treatments from
    replicates

24
From consultant wild oat map to experiment...
Broadcast spray
Consultant patch map
Control (no spray)
Patch spray
Computer tells controller to spray in purple
and green areas
25
Lee Van Wychen
26
(No Transcript)
27
Data processing -- yield monitor data
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
(No Transcript)
34
Conclusion
  • Patch spraying can increase net returns by 10/a
    for field that are 20-50 infested
  • Wild oat control may be more variable with patch
    spraying (2 of 4 cases had no difference in wild
    oat rating, the other 2 had approximately twice
    the rating)
  • Field-scale experiments work with current
    technology!

35
Reduced Inputs Increased Risk
36
Adaptive Management
Experiments
Management
37
Adaptive Management
A. Bussan, L. Rew, D. Goodman, D. Buschena
  • Adaptive management is based upon the premise
    that managed ecosystems are complex and
    inherently unpredictable.
  • The adaptive approach embraces the uncertainties
    of system responses and attempts to structure
    management actions as "weak" experiments from
    which learning is a critical product.

38
Goal Maximize Net Returns To Farmers Through
Adaptive Management of Inputs
Whole Field Input Prescription
Previous Information
Data Synthesis
Weed Control
Yield
Weeds
Seeding
Seeding
Fert.
Fertilizer
Experiment
Building the database Increases the predictive
ability
Bayesian estimation
39
Proof of Concept Experiment
  • Create a reasonable model to predict yield and
    net return
  • Use the model to compare net return outcomes for
    site-specific, conventional high input
    full-field, and conventional full-field
    low-input approaches to management.

40
Predicting Crop Response to theEnvironment and
Inputs
Decision Support Model Structure
  • Weed damage function including the impact of crop
    seeding rate.
  • Nitrogen and water, crop response function
  • Weed response to management function (herbicide
    dose response)

41
Crop Yield is a function of
Decision Support Model Structure
  • Crop seeding rate (SR)
  • Weed density (Nw)
  • f(management intensity or herbicide rate)
  • Nitrogen available in the soil (N)
  • Plant available water (stored soil moisture
    growing season precipitation) (PAW)

42
Decision Support Model Structure
Sensitivity Analysis
Ymax Ywf most sensitive parameter
43
(No Transcript)
44
(No Transcript)
45
Herbicide Rate Response
46
Decision Support Model Structure
NR Y(P-D) - (SC FC WC O)
NR Net Return (/acre) Y Yield P Price
received (/bu) D Price dockage (/bu)
SC Seed cost (/acre) FC N-Fertilizer cost
(/acre) WC Weed control cost (/acre) O All
other costs (/acre)
47
Weed demographic response to predict future
population levels and subsequent crop impacts
  • Seed survival in the seed bank
  • Germination and emergence
  • Seedling survival
  • Seed production
  • Migration (immigration and emigration)
  • Biotic and abiotic impacts on the above
  • Spatial and temporal variability in biotic and
    abiotic environment

48
Spatial Analysis Input
Optimization MSU Weed Ecology Group
version 0.5 This computer
program has the minimum requirements of 1) a
georeferenced map of wild oat abundance
(density), 2) crop yield map from a yield
monitor, 3) and a map of where herbicides were
applied to the field. Other georeferenced
information can greatly improve the
predictions of the model e.g. 4) crop density
map 5) soil moisture map ...etc. ltEntergt to
continue...
49
Enter field information and amount of inputs
used in the field that was mapped. ---------------
--------------------------------------------------
-------------- For estimating stored soil water
using the Paul Brown Soil Moisture Probe 1.
Predominant soil texture a) Coarse - sand
b) Coarse - loamy fine sand c) Mod. coarse -
sandy loam d) Mod. fine - clay loam, sandy
clay loam e) Fine - sandy clay, silty clay,
clay ------------------------------------------
---- Enter a letter for the soil type...?
d 2. Enter the moist soil depth determined with
the Moisture Probe in inches? 14 3. Using the
70 probability of precipitation map, enter the
expected precip. for the growing season in
inches...? 5
50
Enter Economic parameter values -------------
--------------------------------------------------
------ Price expected to received for grain /bu
? 3.00 Cost of weed control at label rate
including application cost in /acre ? 18 Cost
of 60lbs of seed for planting (/acre) ? 1.20
Cost of 80lbs of N fertilizer (/acre) ? 2.50
All other crop husbandry costs in /acre ?
60 -----------------------------------------------
---------------------- Enter to
continue...
51
(No Transcript)
52
(No Transcript)
53
(No Transcript)
54
(No Transcript)
55
Patchy Wild Oat Distribution
100
90
80
70
Probability of Higher ANR
60
50
40
30
20
10
0
SS better than CONV
SS better than LI
CONV better than LI
56
Continuous Wild Oat Distribution
100
90
80
70
Probability of Higher ANR
60
50
40
30
SS better than CONV
CONV better than LI
SS better than LI
SSbtConv
SSbtLI
57
Summary from Proof of Concept Experiment
  • Logical model with reasonable estimate of
    variation can produce reasonable results.
  • With inclusion of variation (Monte Carlo
    Simulation) results could be useful to producers.
  • Site-specific management is likely to improve
    input management in small grain systems

58
Objective
Constructing the black box
Select and refine empirical small grain yield
models to include multiple factors including
nitrogen fertilizer rate, herbicide rate and
precipitation as well as wild oat density, and
develop methods that allow on-farm
parameterization of the models for weed
management decision support. Nicole Wagner
(USDA-FSA Argentina crop predictions)
59
Crop Yield as a Function of 5 Variables
Weed Density(Nw)
Crop Density(NC)
Crop Yield
Precipitation (gsp)
Available Nitrogen (N)
Herbicide Rate (r)
Nicole Wagner
60
  • Imagine the factorial experiment required to
    assess 5 continuous variables producing nonlinear
    responses.
  • 1024 plots with no replication

61
Sub-models from Disjoint Data Sets
Data Sets
Sub-models
PAW Crop Yield
PAW crop SR
N rate PAW Crop Yield
N rate PAW
herbicide rate
Herbicide rate Crop Yield
Global Model (all 5 variables)
herbicide crop SR
N rate Herbicide rate Crop Yield
Jackknife Method
62
Sub-Model Selection Method
  • Sum of Squares lowest SS best model
  • Minimal assumptions about uncertainty
  • Ymax 0.7 0.01 gsp 0.001gsp2 0.002N
    0.0006 gspN
  • Likelihood Method highest likelihood best
    model
  • Establishes the prior
  • Characterizes error which can be incorporated
    into Monte Carlo simulation
  • Use probability distributions to fit parameters

63
Monte Carlo Simulation Sensitivity of Yield
64
Sensitivity of Net Return
65
Model Development
  • Review of historical experiments and models
  • Independent field data (36 site-years from around
    the world)
  • Greenhouse data (full factorial experiment)
  • Virtual field simulation (demonstration of
    practical application of the empirical model)

66
Model Development
  • Review of experiments and models
  • Independent field data
  • 1. gathering data
  • 2. investigating data
  • 3. fitting candidate models to data
  • 4. assessing model performance
  • Greenhouse data
  • 1. conducting experiment
  • 2. investigating data
  • 3. fitting candidate models to data
  • 4. assessing model performance
  • Virtual field simulation

67
Investigating individual data sets
wheat yield (kg/ha)
wild oat density (plants/m2)
Van Wychen 2004
68
Results of scatter plots, standardized regression
69
Results of scatter plots, standardized regression
yield
crop density
70
Results of scatter plots, standardized regression
yield
yield
crop density
weed density
71
Results of scatter plots, standardized regression
yield
yield
crop density
weed density
yield
soil moisture
72
Results of scatter plots, standardized regression
yield
yield
crop density
weed density
yield
yield
?
fertilizer rate
soil moisture
73
Candidate models
1.
Cousens 1985
rw wild oat density ywf weed-free yield
(i.e. maximum yield) i, a parameters
74
Candidate models
1.
no weeds
2.
yield
high weed level
rw wild oat density rc wheat density r
intrinsic growth rate b, f parameters of
competition
crop density
Baeumer and deWit 1968, Wright 1981, Weiner 1982,
Jollife et al. 1984
75
Candidate models
1.
2.
3.
Kim et al 2002
rw wild oat density ywf wheat density b0
weed competition at 0 herbicide R herbicide
rate B response rate of the herbicide
76
Candidate models
1.
ymax
yield
2.
j
crop density (rc )
3.
4.
rw wild oat density rc wheat
density ymaxmaximium yield
Jasieniuk et al 2000
77
Candidate models
4.
5.
Streibig et al 1993
78
Candidate models
4.
5.
6.
79
Candidate models
4.
5.
6.
7.
80
historical models
observed yield (kg/ha)
predicted yield (kg/ha)
81
Assessing model performance
Limitations of the best models
  • best-fitting model does not include the history
    of agronomic modeling
  • 2 does not include influence of managed inputs
  • 3 does not include fertilizers influence on
    wild oat nor herbicides influence on wheat (i.e.
    crop injury)

82
Why a greenhouse experiment ?
  • seeking ecological 1st principles
  • easier to control (e.g. water treatments, time of
    emergence, individual plant distances)
  • relatively quick to repeat
  • generation of hypotheses that can be further
    tested in the field (Freckleton Watkinson 2001)

83
5-variable greenhouse experiment
  • 9 combinations of wheat/wild oat densities
  • 4 herbicide rates
  • 4 nitrogen rates
  • 3 water treatment levels

84
Assessing model performance
best-fitting models
Model 6
Model 5a
Model 5b
85
observed yield (kg/ha)
predicted yield (kg/ha)
86
Model Development
  • Review of experiments and models
  • Independent field data
  • 1. gathering data
  • 2. investigating data
  • 3. fitting candidate models to data
  • 4. assessing model performance
  • Greenhouse data
  • 1. conducting experiment
  • 2. investigating data
  • 3. fitting candidate models to data
  • 4. assessing model performance
  • Virtual field simulation

87



wheat density
wild oat density
soil water
Virtual Field Maps
Van Wychen, unpublished data
88
(No Transcript)
89
maximizing yield
kg N/ha 0 22.5 45 90 112.5 135 157.5 169
herbicide rate 0 0.25 0.5 0.75 1
90
Predicted Yield
Herbicide rate
Nitrogen rate
kg N/ha 157.5
full herbicide rate (1x)
High rate Inputs
kg N/ha 45
quarter herbicide rate
Low rate Inputs
91
Low rate scenario
Variable rate scenario
High rate scenario
kg/ha
lt 800 801-1000 1001-1200 1201-1400 gt 1400
Predicted Yield
92
Localized variable rates
High input rate
Low input rate
93
Localized variable rates
High input rate
Low input rate
probability 86
100
94
Spring data
crop density
Model Yield Equation
weed density
available water
refit model-- update parameters
Machinery for variable-rate application data
collection
herbicide rate
nitrogen rate
yield map
95
Summary
  • Overall, previously collected field data sets
    revealed a large amount of variation,
    illustrating lack of knowledge of ecological
    mechanisms upon which farmers currently make
    management decisions.
  • Field data supported the development of a
    5-variable linear regression model, but a
    5-variable data set (i.e. greenhouse experiment)
    was necessary to build a 5-variable nonlinear
    model.
  • Application of the 5-variable nonlinear model to
    prescribe localized variable-rate fertilizer and
    herbicide management on farms will promote
    site-specific parameter estimation and increased
    parameter stability.

96
Decision Support System
Previous years data
Possible management strategies
crop density
Global Model
weed density
Yield Equation
nitrogen rate
herbicide rate
available water
Predicted Returns
On-farm data collection
Net return
Strategy
97
Risk Distribution based on simulations for
predicted yield
Strategy 1 SSM
Net Return
Strategy 2 Low input
98
(No Transcript)
99
Acknowledgements
Dr. Ed Luschei
Dr. Sharlene Sing
Perry Hofferber
Lee Van Wychen
Dr. Lisa Rew
Dr. Marie Jasieniuk
Nicole Wagner
About PowerShow.com