Transcript March 29, 2016 Crop Modeling Webinar

AgMIP Regional Team Crop-Climate
Modeling Webinar
Ken Boote, Cheryl Porter, Alex Ruane,
Wei Xiong, Sonali McDermid, John Dimes, etc.
March 29, 2016
Phase 2 – Reviewing CM1 and CM2 (Climate Scenarios)
Crop/Climate Modeling
Webinar Goals
 Evaluate Phase 2 crop model simulations of climate scenario
impacts.
 Review Crop Model Simulations of Climate Impacts
 Discuss deadlines, Uploads to ftp site
 April 3: Final CM1 and CM2 simulations submitted
 Review simulated climate impacts (CM1 and CM2)
 Tools for creating box-and-whisker plots?
 What do the economists need?
Identify areas where help of climate scientists and crop modelers
is needed.
3 teams out of 7 have submitted CM1 and CM2 simulations
Actions: 1) Send updated ppt with correct change ratio, 2) Make
r-scripts for box-and-whisker plots, 3) Send timeline to all
2
March 25 (new) Deadline for
submitting CM1 and CM2
March 25 - Deadline to submit CM1 analysis (current management,
current climate, 30 years) and CM2 analyses (5 future climate
scenarios with current management, 30 years) to the ftp site.
CM1, CM2 files:
• Harmonized data (aceb, dome, alnk, acmo files)
• Include box and whisker diagrams comparing outputs from each
analysis. (These can be done using FACE-IT.)
Visualization:
• Box and whisker diagrams of absolute yields using “per farm” mean
of 30 years, where the variation in box-and-whisker is from farms:
Baseline, plus 5 climates, done for two crop models
• Box and whisker plot of climate change ratio (CCR), where CCR =
per-farm 30-year mean(future) / per-farm 30-year mean(base):
Baseline, plus 5 climates, done for two crop models.
• Some analysis of the 6 climates (base, hot-dry, cool-wet, etc.)
3
Box-and-Whisker Plot Chickpea
Example: Absolute Yield
Ignore that the two models are
different. Calibration not done.
Beware when many farms are
at zero yield
Concern for probability of exceedance
with many “zero” yields
Beware when many farms are at zero yield. Economists need
climate change ratio. Not possible if you have zero yield. Here
caused by APSIM failure at low N. Solved by higher initial N.
Box-and-Whisker Plot Chickpea
Example: Climate Change Ratio
Need to understand why a crop
model goes up or down on CCR
Box-and-Whisker Plot Chickpea
Example: Climate Change Ratio
Need to understand
why a crop model
goes up or down on
CCR
South India Selection of 5 GCMs for
Impact Assessment: Mid century RCP 8.5
Choice of Models under different climate caterory for Impact assessment
Cool/Wet: MIROC 5 (O)
Hot/Dry : IPSL-CM5A-MR (N)
Cool/Dry : inmcm4 (L)
Hot/Wet : CanESM2 (D)
Middle : HadGEM2-AO (Y)
South India: Base and future maize and rice
productivity, 2 models & 5 climates
I would put base
climate on plot to right
Rice farms - Base and future rice productivity
South India: Base and future maize and rice
climate change ratio, 2 models & 5 climates
Maize farms - Impact of climate change
on maize productivity
Rice farms - Impact of climate change on
rice productivity
IGB Example of Table: Absolute
Yield and Change Ratio
Comparison of wheat yield under baseline and future
climate scenario RCP4.5 – mid term
Simulated yield (kg/ha) Yield Change (%)
APSIM
DSSAT
APSIM DSSAT
Baseline
4156
4674
GLXF
Cool/Dry
4608
4722
10.89
1.0
GMXF
Hot/Wet
3258
4827
-21.61
3.3
GQXF
Hot/Dry
3516
4726
-15.39
1.1
GSXF
Cool/Wet
3854
5160
-7.26
10.4
GYXF
Middle
3969
5011
-4.50
7.2
Table works, but does not
give concept of variability.
11
IGB Example of Table: Absolute
Yield and Change Ratio
Comparison of wheat yield under baseline and future
climate scenario RCP8.5 – mid term
Baseline
IAXF
ICXF
IPXF
IQXF
ITXF
Simulated yield
(kg/ha)
Yield Change (%)
APSIM
DSSAT
APSIM
DSSAT
4156
4674
Middle
3985
4936
-4.1
5.6
Hot/Wet
3232
4949
-22.2
5.9
Cool/Dry
3515
5199
-15.4
11.2
Hot/Dry
3415
4819
-17.8
3.1
Cool/Wet
3803
5188
-8.5
11.0
Table works, but does not
give concept of variability.
Review: January 22 Deadline
Requested (not complete?): Upload CM0 (historical) and CTWN
analyses to ftp site for review by DSSAT and APSIM modelers.
CM0 files:
• Survey data, domes, linkage files, model inputs, model outputs,
harmonized data (aceb, dome, alnk, acmo files).
• Use the directory structure described in the Handbook Appendix 3
• Include a probability of exceedance graph with both models and
observed data for the CM0 analyses.
• All modifications made to adjust simulated outputs must be made
in the data or DOMEs such that simulated outputs can be
replicated from the harmonized data inputs.
CTWN files:
• Survey data, domes, linkage files, acmos, model inputs, model
outputs.
• Include box and whisker diagrams for the standard CTWN
sensitivity analyses.
13
Climate Team Planning for
Dakar
• Request from Alex and DFID
– Do you know of any longer-term 10-plus years of yield
data (network of trials, district level, national level)
available for your regions?
– Alex and DFID wish to document the extent to which
crop models are able to simulate seasonal weather
variation, including extremes (drought, heat, etc.).
– This is not possible in single year yield survey data.
• Long-term historical yield data provided by SAAMIP, IGB,
and South India.
The Most Important Issues
Discovered from CTWN
• Issues learned from CTWN and survey plots
– # 1 Problem: Setting of SLSC=SOM3 and SLFIC=Finert.
These set yield at zero N. Must have both models
predicting same “low” yield levels at zero N. Experience so
far: needed high SLSC=SOM3 of 0.96 to 0.98 to mimic
degraded soil. SLFIC=? If this is off, models will differ on
the other plots (CO2, temperature, rainfall).
– # 2 Problem: High initial N (in survey file) can also cause
lack of response to fertilizer N as well as differences among
models.
– # 3 Problem: Cultivar yield potential set too high, can give
very high yield at 180 kg N/ha. Sets asymptote.
– Model differences in temperature response, gives opposite
Last
(negative vs. positive) response to scenarios.
Slide