Session 1.1 Seasonal Forecasts and Climate Extremes
For a complete list of all of the workshop abstracts click here (PDF).
If you are presenting a poster, please view the Poster Specifications here
Session 1.2: Oral Presentation
Title: Inter-comparison of crop models for simulating canola growth and yield
Authors : Enli Wang1, D. He1,2,
J. Wang2, B. Christy3, M. Hoffmann4, J. Lilley1,
G. O’Leary3, J. Hatfield5, L. Ledda6, P. A. Deligios6, B. Grant7, Q. Jing7, H. Kage8, B. Qian7, E. E. Rezaei9, W. Smith7, W. Weymann8, and The AgMIP-Canola Team
Abstract:
The first comprehensive study on inter-comparison of crop models for simulating growth and grain yield of canola crop has been completed by the AgMIP-Canola team. The team was formed in 2015 and is now finalizing its phase I work and result analysis. The main focus of AgMIP-Canola Phase I was inter-comparison of the major canola models against multiple datasets and analysis of the sensitivity of simulation results to different scenarios of climatic and management changes. Fifteen modelers in 5 countries (Australia, China, Canada, Germany, and Italy) participated with eight crop models (APSIM, CAT-Canola, DSSAT, DayCent, DNDC, HUME, MONICA, SIMPLACE). Experimental data from 6 sites across 5 countries (Australia, China, Germany, Italy and USA) were used to compare model performances. Sensitivity analysis was carried out with a combination of 5 levels of atmospheric CO2 concentrations, 7 temperature changes, 5 precipitation changes, together with 5 N application rates. Preliminary results showed that a partial model calibration (only for phenology) led to poor simulation of biomass and yield, and even the ensemble mean/median differed significantly from measurements across sites. A full calibration with additional data of LAI, biomass and yield from one treatment each site significantly improved model performance for biomass and yield simulations. The multi-model ensemble median yield was better than single-model yield predictions for some but not all models. Results from the sensitivity simulations and the differences in model performances are being analyzed to identify knowledge gaps and new datasets for the model improvement in the next phase.
Session 1.2: Oral Presentation
Title: Lessons Learned from Evaluating APSIM and DSSAT Maize Model Responses to Carbon Dioxide, Temperature, Water, and Nitrogen.
Authors: Kenneth Boote1, C. Porter1, J. W. Jones1, J. Dimes2, J. Hargreaves2, P. Thorburn2, D. MacCarthy3, P. Traore3, W. Durand4, D. Cammarano4, P. Masikati5, S. Homann5, S. Gummadi6, L. Claessen6, D. Murthy7, G. Vellingiri7, S. P. McDermid8, and A. C. Ruane9
1University of Florida, USA, 2 CSIRO, 3 CIWARA, 4 SAAMIP, 5 CLIPS, 6 E. Africa, 7 S. India, 8 New York University, USA, 9 NASA GISS, USA
Abstract:
It is important, in climatic impact assessment, to understand how different crop models respond to CO2, temperature, rainfall, and nitrogen (CTWN). Regional teams in West Africa, East Africa, South Africa, Southeast Africa, and South India obtained farm survey yield data for maize from households in their regions, along with farmer management information,historical weather, soils, and local cultivar calibrations. They selected a representative farm for evaluating DSSAT and APSIM maize model simulations for response to CTWN. Simulated response to N fertilization from 0 to 180 kg N ha-1showed that stable or inert soil C pools for the models had to be set correctly to mimic the yields obtained for zero N fertilizer (or farmer observed yields), and that the yields at high N fertilization represent the genetic potential of the cultivar. The need for correct N response is very important because the teams want to use N fertilization as an option for improving production. The two maize models had a very modest response to CO2 that was less at low N. Increasing temperature reduced yield for both models at most sites, with DSSAT being more sensitive at high temperature. Response to rainfall was less than expected for West and East Africa, because N was so limiting that the low LAI created low transpiration demand. In South and Southeast Africa, where rainfall was lower, the sensitivity to rainfall was stronger, with greater yield reductions for DSSAT than for APSIM. The CTWN exercise was valuable for understanding differential model sensitivity to climatic factors, and guiding model calibration for response to N fertilization for degraded soil conditions.
Session 1.2: Oral Presentation
Title: Who has the ‘best’ crop model?
Authors: Senthold
Asseng, P. Martre, F. Ewert, D. Wallach, B. Liu & AgMIP-Wheat team
Abstract: Who has the ‘best’ crop model? Nobody has! – as no ‘best’ crop model exists so far. Multi-model intercomparisons of many different crop models with field observations have shown that there is no single model ‘best’ suited across different growing conditions. There is also no particular model approach for specific processes (e.g. for photosynthesis) superior across growing conditions. Multi-model intercomparisons showed that while some crop models might reproduce field data very well in one environment, there are different models better in reproducing field data in other environments. Disregarding a crop model based on a few observations might also be premature as such a model might perform very well in other conditions. However, the multi-model ensemble median of crop models has been repeatedly proven as a superior predictor of observed data across different growing conditions (and across crops), compared with any single model. The required number of models in an ensemble for the ensemble median to be a better predictor than any individual model could be as small as 2-5, depending on the growing environment. Comparing simulated impacts from multi-model ensembles with other methods, like statistical methods of impact assessment further supports the validity of a multi-model ensemble approach. Recent progress in AgMIP-Wheat and AgMIP-Wheat linked to other themes in AgMIP will be discussed.
Session 1.2: Oral Presentation
Title: Global Gridded Crop Model evaluation: benchmarking, skills,
deficiencies and implications
Authors: Christoph Müller1, J. Elliott2, and GGCMI phase 1 modeling team
1 PIK, Germany, 2 University of Chicago, USA
Abstract: Large differences between global-scale and also between field-scale crop models have
been recently reported, following a general call to revisit modeling skills and approaches. For this central objective of AgMIP and ISI-MIP 14 global gridded crop models (GGCMs) have contributed simulations of the historic past (1901-2012) in the framework of AgMIP’s GGCM Intercomprison project.
The global scale is especially challenging for model application and evaluation because of the vast climatic and managerial differences between regions but also because of the limited availability of reference data at sufficient detail and a general difficulty in comparing
results of global-scale simulations with site-specific data. Global scale models need to be evaluated at the scale of application, which is typically national or regional aggregates, but site specific data can provide better insights on how well general mechanisms in plant growth and yield formation are
represented in the models.
We here provide a broad model evaluation framework to test performance of GGCMs and also to identify general and individual model deficiencies across different crops and regions. Model skill is evaluated for the four major crops wheat, maize, rice and soy. We find that
GGCMs have substantial skill in reproducing spatial and temporal patterns of yield productivity at the global scale and within individual regions, but that individual models can have complimentary skills, suggesting large potentials for within-ensemble learning. These findings will serve as a basis for further
model development and improvement.
Session 1.2: Oral Presentation
Title: The uncertainty cross-cutting theme
Authors: Daniel Wallach1, L. Mearns2, P. Thorburn3, R. Rotter4, A.C. Ruane5, S. Asseng6, A. Challinor7, J. Jones6, and F. Ewert8
1 INRA, France, 2 NCAR, 3 CSIRO, 4
University of Göttingen, Germany, 5 NASA GISS, USA, 6 University of Florida, USA, 7 University of Leeds, UK, 8
University of Bonn, Germany
Abstract: This
is a report on recent and planned activities of the uncertainty cross-cutting
theme, which aims at developing, analyzing and diffusing effective methods of
estimating the uncertainty in model predictions. A study on the lessons
from the climate modeling community on the design and use of ensembles for crop
modeling was led by a crop modeler and a climate modeler. This is meant to
serve as a roadmap for future methodological studies on ensemble crop modeling.
It covers questions related to the creation of ensembles, model weighting
within ensembles, single model ensembles, super ensembles, uncertainty
evaluation based on ensembles and the use of the ensemble to create better
predictors. A study on model evaluation showed that there are two different
viewpoints about model evaluation, depending on whether the model is treated as
fixed or random. In both cases mean squared error of prediction (MSEP) is a
useful criterion of model prediction uncertainty, but the interpretation and
estimation are very different in the two cases. A study on the effect of
averaging a quantity of interest over space or time on uncertainty showed that
averaging always reduces MSEP, and that the amount of reduction can be
estimated given hindcasts that are relevant to the averaging process. A
new activity, now being organized, concerns crop model calibration. The
objective is to document current methods and propose guidelines for improved
methods, which will be tested by multiple models on common data. This activity
is open to those interested.
Session 1.2: Oral Presentation
Title: Parameterization induced uncertainty of the EPIC model to
estimate climate change impact on global maize yield
Authors:
Wei Xiong, R. Skalsky, C. H. Porter, J. Balkovic, J. W. Jones, and Y. He
1 Institute of Environment and Sustainable Development in
Agriculture, Chinese Academy of Agricultural Sciences; 2 Department
of Agricultural and Biological Engineering, University of Florida; 3
International Institute of Applied Systems Analysis (IIASA), Ecosystem Services
and Management Program; 4 Soil Science and Conservatio Research
Institute; 5 Faculty of Natural Science,Comenius University in
Bratislav
Abstract: Understanding the interactions between agricultural production and climate is necessary for sound decision-making in climate policy. Gridded and high-resolution crop simulation has emerged as a useful tool for building this understanding. Large uncertainty exists in this utilization, obstructing its capacity as a tool to devise adaptation strategies. Increasing focus has been given to sources of uncertainties for climate scenarios, input-data, and model, but uncertainty due to model parameterization are still unknown. Here, we use publicly available geographical datasets as input to the Environmental Policy Integrated Climate model (EPIC) for simulating global gridded maize yield. Impacts of climate change are assessed up to the year 2099 under a climate scenario generated by HadEM2-ES under RCP 8.5. We apply five parameterization strategies by shifting one specific parameter in each simulation to calibrate the model and the effects of parameterization. Regionalizing crop phenology or harvest index appears effective to calibrate the model for the globe, but using various values of phenology generate pronounced difference in estimated climate impact. However, projected impacts of climate change on global maize production are consistently negative regardless of the parameter being adjusted. Model parameterization results in a modest uncertainty at global level, with spread of the global yield change less than 30% by the 2080s. The uncertainty is likely to increase if parameterizing multiple parameters simultaneously that often being practiced in site specific simulations. Parameterization has a larger effect at local scales, implying the possible types and locations for adaptation.
4. Poster Presentation: Session 1.2
Title: Comparative analysis of the simulation of canola phenology from biophysical modules at selected field sites across Canada
Authors: Aston Chipanshi1, Y. Zhang2, G. Bourgeois3, B. Qian2 and B. Bondaruk2
1 Science & Technology Branch, AAFC, Saskatchewan, Canada, 2 Science & Technology Branch, AAFC, Ottawa, Ontario, Canada, 3 Science & Technology Branch, AAFC, Québec, Canada.
Abstract: We evaluated the commonly used cereal based temperature phenology models with the goal of adapting one for canola so that it can be integrated into near real time monitoring of crop conditions and crop yield estimates. Using field data collected from sites across Canada, a comparative analysis of
canola crop stages as calculated from a simple algorithm using growing-degree-day (GDD) accumulation, a module from the Computer Centre for Agricultural Pest Forecasting (CIPRA) and the CROP GRO model were compared with the field observed stages. Field observations were recorded using the BBCH scale and complemented by leaf area index (LAI) and above ground total biomass. Supporting data- the daily maximum and minimum temperature, total precipitation and solar radiation (at some sites) as well soil moisture were measured. Field
nutrient data (N, P, K and S) were also collected.
Between the CIPRA and CROP GROW simulations, there was little difference in the simulated onset of budding, flowering, pod setting and ripening in relation to the observed dates. Even though the CROP GRO simulated dates were within the range of the CIPRA simulated values, it was difficult to ascertain some of the crop and environmental parameters inputs required to arrive at the simulated dates in CROP GROW. Likewise, while the GDD-based simulated stages were broadly comparable to the CIPRA and CROP GROW values, the
canola parameters were first estimated from published sources. Results from the
IPRA method were found suitable for integration into the canola statistical
yield prediction model.
KEYWORDS: Phenology, canola, growing-degree-day, CIPRA, CROP
GROW.
5. Poster Presentation: Session 1.2
Title: Geospatial crop simulation scheme to
deliver amendment measures to alleviate the impact of climate change on crops
Authors: Jonghan Ko, S. Jeong, and J. Choi
Chonnam National University, South Korea
Abstract: Determining effective measures that can
be used to alleviate the impacts of climate change on crops is one of the most
urgent issues facing agriculture. Accordingly, our objectives in this study
were (1) to develop a geospatial crop simulation modeling (GCSM) system to
simulate regional crop production data and (2) to determine the current and
future remedial measures using the developed simulation design. We formulated the
GCSM scheme using the Decision Support System for Agricultural Technology
(DSSAT) crop model package version 4.6, in which crop models can be used to
conduct numerous pixel data runs based on shell scripting in a Linux operating
system. The developed GCSM system was verified by its capability to simulate
barley (Hordeum vulgare) production in South Korea, and the GCSM was then used
to simulate the effects of climate change on barley production for the same
geographical region. Therefore, we will be able to use the GCSM to investigate
and report potential remedial measures that can be used to amend or alleviate
the potential future impacts of climate change on barley production. Although
the current GCSM system requires further development to formulate more concrete
tools for agricultural scientists, farming business managers, and stake
holders, we believe that the system could be effectively used to simulate
geospatial variations of climate change impacts on crops and to search for
potential solutions to the impending food insecurity.
6. Poster Presentation: Session 1.2
Title: Ozone changes the
photosynthesis-conductance relationship for Rice
Authors: Yuji Masutomi1 , T. Yonekura2 , T.
Takimoto3 ,and H. Oue4
1 College of Agriculture, Ibaraki
University, Japan, 2 Center for Environmental Science in
Saitama, Japan, 3 Institute for Global Change Adaptation
Science, Ibaraki University, Japan, 4
Faculty of Agriculture, Ehime University, Japan
Abstract: Ozone (O3) is one of significant factors
reducing crop yields. To date, the reduction of crop yields due to O3 has been
assessed by the dose-response models and, more recently, the fluxed-based
models. These models, however, can’t quantify the combined effect of O3 and
other important factors that affect crop yields, e.g., climate change and the
increase in CO2 concentration. One useful approach for quantifying
the combined effect is to incorporate the O3 impact into a
photosynthesis-conductance model, e.g., the Farquhar/Ball-Woodrow-Berry (BWB)
model. However, fundamental understanding of the O3 impact on the
photosynthesis-conductance models for crops is still missing. For example,
nothing is known about the influence of O3 on the BWB relation for crops. The
objective of this study is to reveal whether O3 has influence on the BWB
relation for rice. To achieve the goal, we grew 4 rice varieties under ambient
and elevated O3 concentration in 2008 in China using FACE system. The
comparison of the observed BWB relations under ambient and elevated O3
conditions revealed that increase in O3 can change the BWB relation for rice.
But there was large difference in the changes of the BWB relation among rice
varieties. These results imply that the change of the BWB relation due to O3
and the difference in the change among rice varieties should be considered when
we assess the risk of ozone for rice.
7. Poster Presentation: Session 1.2
Title: Generation of crop model ensembles by
use of a modelling framework
Authors: Eckart Priesack1 , X. Duan1,2 ,
S. Gayler3 , F. Heinlein1 , C. Klein1 , and
C. Thieme1
1 Institute of Biochemical Plant
Pathology, Helmholtz Center Munich, Germany, 2 Institute of Agricultural Economics and Social Sciences
in the Tropics and Subtropics, University of Hohenheim, Germany, 3 Institute of Soil Science and Land Evaluation,
University of Hohenheim, Germany
Abstract: We present an example for the generation
of model ensembles by use of the model framework Expert-N. Different crop
models are obtained by choosing different sub-models which represent important
processes to determine the dynamics of crop growth. In this way different
sub-models to simulate potential evapotranspiration, actual evaporation, actual
transpiration, soil water flow, soil nitrogen transport, soil carbon and
nitrogen turnover, crop development, canopy photosynthesis, potential and
actual nitrogen uptake and crop growth are combined resulting in different crop
models making up an ensemble of crop models. The sub-models are based on
process descriptions that are included in the crop models CERES, SUCROS, SPASS
and GECROS, but also stem from known soil models such as CENTURY, SOIL, SOILN,
NCSOIL, LEACHM or HYDRUS.
The generated model ensemble is applied to
simulate winter wheat growth at a field site in Southern Germany. Simulation
results are compared to measurements of crop biomass and yields and to soil
water and nitrogen contents. It is concluded that model frameworks as the model
system Expert-N can help to analyse structural uncertainties that lead to
different simulation results between models of a model ensemble.
Biernath, C., Gayler, S., Bittner, S., Klein,
C., Högy, P., Fangmeier, A., Priesack, E.: Evaluating the ability of four crop
models to predict different environmental impacts on spring wheat grown in
open-top chambers. European Journal of Agronomy 35 (2011) 71-82.
Priesack, E., Gayler, S., Hartmann, H.P.: The
impact of crop growth sub-model choice on simulated water and nitrogen
balances. Nutr. Cycl. Agroecosys. 75 (2006) 1-13.
8. Poster Presentation Session 1.2
Title: Multi-wheat-model ensemble responses to
interannual climate variability
Authors: Alex C. Ruane1 , N. I. Hudson2 , S. Asseng3 , D. Camarrano3,4 ,
F. Ewert5,21 , P. Martre6 , K. J. Boote3 , P.
J. Thorburn7 , P.K. Aggarwal8 , C. Angulo5 , B.
Basso9 , P. Bertuzzi6 , C. Biernath10 , N.
Brisson6,11 , A. J. Challinor12,13 , J. Doltra14 , S. Gayler15 , R.
Goldberg2 , R.F.Grant16 , L. Heng17 , J.
Hooker18 , L.A. Hunt19 , J. Ingwersen15 , R.
C. Izaurralde20 , K.C. Kersebaum21 , S. N. Kumar22 , C.
Müller23 , C. Nendel21 , G. O’Leary24 , J.
E. Olesen25 , T. M. Osborne18 , T. Palosuo26 , E.
Priesack10 , D. Ripoche6 , R. P. Rötter26,34 ,
M. A. Semenov27 , I. Scherbak28 , P. Steduto29 , C.
O. Stöckle30 , P. Stratonovich27 , T. Streck15 , I.
Supit31 , F. Tao26,31 , M. Travasso32 , K. Waha23,8 ,
D. Wallach6 , J. W. White33 , and J. Wolf 31
1 NASA Goddard Institute for Space
Studies, USA, 2 Columbia University Center for Climate
Systems Research, USA, 3 University of Florida, USA, 4 James Hutton Institute, UK, 5 Institute of Crop Science and Resource Conservation,
Germany, 6 National Institute for Agricultural
Research (INRA), France, 7 Commonwealth Scientific and Industrial
Research Organization Agriculture, Australia, 8 Consultative Group of International Agricultural
Research, International Water, Management institute, India, 9 Michigan State University, USA, 10 Institute of Biochemical Plant Pathology, Germany, 11 AgroParisTech, France, 12 Institute for Climate and Atmospheric Science,
University of Leeds, UK, 13 CGIAR-ESSP, Colombia, 14 Cantabrian Agricultural Research and Training Centre,
Spain, 15 Institute of Soil Science and Land
Evaluation, Germany, 16 University of Alberta, Canada, 17 International Atomic Energy Agency, Austria, 18 University of Reading, UK, 19 University of Guelph, Canada, 20 University of Maryland, USA, 21 Leibniz Centre for Agricultural Landscape Research, Germany,
22 Indian Agricultural Research Institute,
India, 23 Potsdam Institute for Climate Impact
Research, Germany, 24 Landscape and Water Sciences Department
of Primary Industries, Australia, 25
Aarhus University, Denmark, 26
Natural Resrouces Institute Finland, 27
Rothamsted Research, UK, 28 Queensland University of Technology,
Australia, 29 FAO, Italy, 30 Washington State University, 31 Wageningen University, The Netherlands, 32 Institute for Climate and Water, Argentina, 33 USDA, USA, 34
Georg-August-University Göttingen, Germany
Abstract: We compare 27 wheat models’ yield
responses to interannual climate variability, analyzed at locations in
Argentina, Australia, India, and The Netherlands as part of the Agricultural
Model Intercomparison and Improvement Project (AgMIP) Wheat Pilot. Each model
simulated 1981–2010 grain yield, and we evaluate results against the
interannual variability of growing season temperature, precipitation, and solar
radiation. The amount of information used for calibration has only a minor
effect on most models’ climate response, and even small multi-model ensembles
prove beneficial. Wheat model clusters reveal common characteristics of yield
response to climate; however models rarely share the same cluster at all four
sites indicating substantial independence. Only a weak relationship (R2 ≤ 0.24)
was found between the models’ sensitivities to interannual temperature
variability and their response to long-term warming, suggesting that additional
processes differentiate climate change impacts from observed climate
variability analogs and motivating continuing analysis and model development
efforts.
9. Poster Presentation: Session 1.2
Title: CTWN – Carbon-Temperature-Water-Nitrogen
responses of DSSAT and APSIM models in relation to crop management and initial
soil conditions in wheat
Authors: Nataraja Subash1 , K. J. Boote2 , P.
L. Paulton3 , B. Singh4 , C. Porter2 , S.
P. McDermid5 , H. Singh1 and G. A. Baigorria6
1 ICAR-IIFSR, India, 2 University of Florida, USA, 3 CSIRO, Australia, 4
CIMMYT, India, 5 New York University, USA, 6 University of Nebraska-Lincoln, USA
Abstract: Crop models are extensively used for
assessing the impact of climate change and projecting the food security
scenario of important food crops at local to global level. Even though
these models are basically evolves the crop growth and development around
growing degree concept, however, several subroutines with model specific
processes, depending on the researchers/developers field of specialization and
location of model influences the simulation results. Here under
Agricultural Model Intercomparison and Improvement Project (AgMIP), we
systematically created the CTWN (Carbon, Temperature, Water & Nitrogen)
sensitivity simulation set up in APSIM-wheat and DSSAT-wheat to compare the
responses of these models to changes in these variables. The analysis is done
for 76 farm sites, with the C,T,W &N variables changed one at a time so
that responses can be compared without interactions and so outputs can be
analyzed in detail where model responses differ. The standard CTWN protocol
includes the following 32 simulations: CO2 – 360, 450, 540, 630,
720ppm (run for 30 kg/ha N and 180 kg/ha N) – 10 simulations; Tmax/Tmin – -2,
0, +2, +4, +6, +8 oC – 6 simulations; Rainfall – 25%, 50%, 75%, 100%, 125%,
150%, 175%, 200% – 8 simulations; Fertilizer N – 0, 30, 60, 90, 120, 150, 180, 210
kg/ha – 8 simulations. The sensitivity of APSIM and DSSAT is different for CO2,
temperature and fertilizers, even though both are showing the same trend.
Similarly, it is also found that the initial soil conditions, crop management
conditions influences the model sensitivity to CTWN. Therefore, clear
understanding of the model processes is very much essential for applications of
these models for projecting the impact of climate change on yield as well as to
address food security issues under future projected climate change.