A process-based modeling approach to forecast and optimize agronomic systems
Designing Optimum Genetic Improvement & Agronomic Systems, Ames, IA, November 29th , 2016
Sotirios ArchontoulisAssistant Professor of Integrated Cropping Systems
Agronomic system: soils + crop-cultivars + climate + management
Planting Harvesting
Grain fill periodflowering
Leaf area index
Soil water
NO3 + NH4
Biomass
Because of the complexity in crop yield, scientists break it down to its components and study only a single component without putting it back together
Very important but usually ignored
Optimizing the agronomic system
What? Yield N-loss Irrigation and nutrients Profits Yield and profits and N-loss
How? Statistical models Process-based models
Type of optimization? Static Dynamic
Optimize or track the system and respond accordingly?
Process-based modeling (APSIM software)
EvaporationRunoff
Drainage Decomposition MineralizationImmobilization
NitrificationDinitrification
Leaching
PhenologyLight interceptionBiomass accumulationBiomass partitioningBiomass retranslocation Leaf area development SenescencePlant death
Water uptakeNutrients uptake
rain temperature radiation
Carry over effects
Grains Stover
INPUTS
INPUTS
OUTPUTS
APSIM has become one of the world’s foremost farming system model > 3,000 licenced users > 100 countries
Holzworth et al. 2014 Environmental Modeling & Software 62: 327-350
Open source codeFree to downloadhttp://www.apsim.info/
ISU work with the APSIM model
Teaching (since 2012) Research (since 2013) Extension (since 2014)
On-going or finished studies Yield – N relationships Manure effects on yields Corn hybrids Soybean varieties Soil water Tile drainage CO2 emissions Rye cover crop Prairies modeling Soil organic carbon Root dynamics Plant growth Soil temperature Biochar effects Erosion Rotation effects N fixation
Databases for Iowa & USA ~ 500 meteorological files (each has 32 years)~ 350 soil profiles (source: web soil survey)~ 60 soybean cultivars (maturity 00 to 9)~ 20 corn hybrids (maturity 85 to 125) ~ 7 workshops
Next workshop: June 2017
The APSIM model and past performance in Iowa Water and Nitrogen limitations included Pest/diseases/phosphorus (not yet)
Puntel , Sawyer, et al., 2016 Front. Plant Science
Continuous corn in central Iowa
Full N
Zero N
Sequential analysis; includes residue, water and nitrate carry overs
How can we optimize the system? Since APSIM has the appropriate algorithms we can optimize the system for NUE
Plant breeders/ Crop physiologists
SoilScientists
Farmers Simulation experiments
Define & test Hypotheses
Optimize the system for each field separately
Statistics (objective functions)Synchronize Demand/Supply
Increase Yield
Decrease N loss
Design field experimentsDefine new hypotheses
Outputs for evaluation
Agronomists
Design and run experiments creation of databases
Farmers (management)
Breeders &physiologists(hybrids)
Crop rotation (previous crop)Cover crops Time of N applicationAmount & type of N applicationManurePlant density & sowing time
PhotosynthesisCanopy stay greenN mobilization and NHIN concentration of tissueRoot architecture (depth x rld) Sink capacity
Soil scientists SOMN inhibitors
Replicates Sum
Factor Level
221010??
??????
??35 years weather data????
Simulation model
Results
Simple exercise
Objective:Increase grain yieldDecrease N leaching
Approach:Explore N management- Rate - Time
Replications:30 historical years
Location:Ames, IA
0 75 150 225 300
0 75 150 225 300
N A M Jun Jul
N A M Jun Jul
May application 180 kg N/ha
Results
Simple exercise
Objective:Increase grain yieldDecrease N leaching
Approach:Explore N management- Rate - Time
Replications:30 historical years
Location:Ames, IA
0 75 150 225 300
0 75 150 225 300
N A M Jun Jul
N A M Jun Jul
May application 180 kg N/ha
optimum
What did we learn?
Yes it works but we need more than
this!
Variability (due to weather) quite high
Risks and logistics
Variability in hybrids was ignored
Example of a static, process-based optimization tool for soybeans
Inputs
http://agron.iastate.edu/CroppingSystemsTools/soybean-decisions.html
Overview of the FACTS platform (dynamic tool)FACTS is a publically available web-tool that deals with production and environmental aspects of cropping systems
Forecast tool Assessment tool
(during the growing season) (before/after the growing season)
perform what-if scenario analysis to identify management practices with the highest profits and lowest environmental impacts
Year 2016 (learn) Multi-year (inform 2017 decisions)
2015 proof-of concept (8 fields) 2016 FACTS publically available (20 fields)
Why we do it Provide quantitative
answers to questions that farmers ask
Improve science behind predictive models
FACTS sites and measurements in 20166 locations (3 with tile drainage)10 corn experiments10 soybean experiments
High resolution measurements; Frequency from 30 minutes up to 1 week
FACTS approach: http://crops.extension.iastate.edu/facts/
Publically availableNo feesNo subscription
Measurements and simulations both presented
Website information 2: soil water and nitrogen
Website information 6: crop yield prediction
Combine measured
Yield Prediction Error (last forecast)
Yield Prediction Error (first forecast)
FACTS web-visits
FACTS assessment tool under development
What-if management scenario analysis – example Ames corn 2016
2016Management
AlternativeManagement
% Yield Change
% N-loss(season)
35,000 plants/acre
5,000 pl/acre more 0.3 -1.8
5,000 pl/acre less -1.3 1.7
May 5th **Planting
12 days earlier planting -5.3 -4.7
12 days later planting 7.7 -7.1
111 RM, 1st week of Sept maturity
Hybrid (7 more days to maturity) 6.9 0
Hybrid (5 less days to maturity) -10.6 -2.4
150 lbs N/acre as UAN At planting
Split N (planting & V6) 0.12 -30.3
PrePlant N application -0.93 23.8
50 lbs/acre more N-rate 0.17 29.1
50 lbs/acre less N-rate -7.3 -15.4
Diagnose / learn from the past optimize the system
Initial soil profile nitrogen and water: Ames corn 2016
2016Values
AlternativeValues
% Yield Change
% N-loss(season)
30 lbs N03-N/acreon January 1
15 lbs less N over the profile -2.1 -12.5
15 lbs more N over the profile 0.2 10.1
3 feet water tabledepth on January 1
4 feet water table depth -8.9 -19.6
2 feet water table depth 0.7 38.6
Diagnose / learn from the past
Alternative management options: Ames soybean 2016 weather
2016Management
AlternativeManagement
% Yield Change
% N-loss(season)
137,000plants/acre
20,000 pl/acre more 1.4 -1.5
20,000 pl/acre less -1.9 1.5
May 6th
Planting 12 days earlier planting -10.0 30.3
12 days later planting -1.6 -10.6
2.7 MG, 1st week of Sept maturity
Variety (10 more days to maturity) 7.8 -4.5
Variety (6 less days to maturity) -15.6 12.1
Summary - % corn yield changes 2016Alternative Management NW NE Central SW SE
More plants 1.1 0.8 0.5 0.1 0.2
Less plants -3.1 -2.2 -1.2 -0.6 -0.6
Earlier Planting -2.7 1.8 -3.3 -3.1 -4.5
Later Planting -4.6 -6.9 2.6 0.7 3.1
Longer Hybrid 6.2 10.7 9.1 10.2 12.1
Shorter Hybrid -12.2 -12.2 -11.4 -12.8 -10.6
Split N 0 0 0 0 0
Pre-Plant N -0.1 0 -0.5 0 -0.1
50 lbs more N 0 0 0.1 0 0
50 lbs less N -9.9 -0.7 -3.7 0 -0.3
Next: “Big runs” to inform 2017 decisions Factor Level
Soil profile water at harvest 5
Soil nitrate at harvest 6
N-rate 20 (every 10 lbs/ac) to estimate EONR
N-time 5 (Fall, Spring, Planting, Split, Late)
N-type 4 (UAN, NH3, NH4NO3, UREA)
Hybrids 20 (maturity x sink capacity)
Planting date 15
Seeding rate 6
Row spacing 2
Tillage type x timing 10 = 5 x 2
Replications (weather years) 40 = 37 (historical) + 3 (forecast )
Sum 17,280,000,000
Model Outputs
Yield
N-loss
Input cost
Profits
Frost risks
FACTS website
X 6 sites Cover cropsResidue managementUpdate the analysis
every 2nd week to decrease uncertainty in weather
Final Reflection
Process-based models are powerful tools, but ….
Optimizing the agronomic system while making lots of assumptions for unknown variables is quite risky
Static optimization tools can aid decision making (select cultivar, decide planting date)
Optimized solutions might not be applicable in practice
Dynamic forecasting and assessment / diagnostic tools (e.g. FACTS) provide the necessary information for the farmers to optimize their cropping system/operation!
Challenges and opportunities: link breeding efforts to cropping systems modeling tools
Acknowledgments
SponsorsDepartment of Agronomy, ISU ANR, Extension ISUIowa Soybean AssociationSoybean Research CenterPlant Science Institute USDA- NIFA (1004346)USDA-NRCSClimate Corp, Monsanto DuPont Pioneer
Main collaborators Mike Castellano Mark Licht
@ArchontoulisLab
Lab-website: http://faculty.agron.iastate.edu/sarchont/FACTS-website: http://crops.extension.iastate.edu/facts/
Email: [email protected]