Final Report Summary - MODEXTREME (MODelling vegetation response to EXTREMe Events)
Executive Summary:
MODEXTREME has as its strategic goal to help the European (and non-European) agriculture to face extreme weather events by improving the capability of biophysical models simulating plant responses to integrate weather extremes. The two key targets are the production of novel (reusable) libraries of models and the improvement of yield monitoring and forecasting systems. For this purpose, the project has improved the capability of biophysical models simulating crop/grassland/tree responses to represent the impact of extreme weather events (such as heatwaves, cold shocks, drought and frost) via development of generically usable software units implementing libraries of models. In so doing, it has extended the multi-model platform BioMA (Biophysical Model Applications) for the simulation of plant growth and development, thus providing direct support to the European Commission Joint Research Centre (MARS: Monitoring Agricultural ResourceS – Agri4Cast action), while also reaching out to local stakeholders by organizing dissemination workshops and training sessions addressed to scientists, technicians and decision makers.
Expressed mathematically, fundamental concepts were implemented into dedicated modules and coupled to existing modelling solutions (EMS) to illustrate that the crop/grassland/tree models have the elements needed to reproduce experimentally established system performances under extreme conditions. In particular, dedicated components modelling the impact of extreme events were developed and coupled to simulation models of arable crops (i.e. generic models such as WOFOST, CropSyst, AquaCrop and Syrius-Quality, and the rice model WARM) but also to the grassland model PaSim, the olive model OLIVECAN and the vineyard model GRAPE, but they are potentially applicable for use with other models as well. Existing and modified (for the impact of extremes) solutions of these models were assessed in a comparative fashion (EMS versus MMS) against field data from a variety of agricultural sites worldwide. This investigation was based on common protocols for model calibration and sensitivity analysis, which allowed identifying the most influential crop parameters in different pedo-climatic (including with scenarios of climate change) and management contexts. Downscaled (bias-corrected) projected storylines of climate change were generated for the study-areas from a set of GCM-RCM simulations to assess the sensitivity of both EMS and MMS to alternative weather scenarios including more frequent and severe extreme events (based on radiative forcing for the high RCP8.5 which is the most extreme of the emissions pathway scenarios developed for the IPCC’s Fifth Assessment Report).
Results with simulation models highlighted limitations in the response to extreme weather events of currently used modelling solutions. The improvement introduced by MMS was the basis for a discussion with local stakeholders and the FAO around the employment of these modelling tools to support food security issues in Europe and worldwide. For the generic crop model WOFOST and the rice-specific model WARM, both EMS and MMS were employed within the Agri4Cast forecasting system of the European Commission (based on modelling tools and agro-climatic indicators) to quantify their reliability to estimate yield for the main cereal/non-cereal crops cultivated in Europe. The reliability of forecasting results was improved in most cases, sometimes greatly increasing the percentage of variance explained (e.g. up to 44% for spring barley in Poland). The addition of reasoned agro-climatic indicators (also developed within the project) to model outputs further enhanced forecast reliability (e.g. for rice in Spain).
To date, the project has led to the publication of 27 peer-reviewed articles and 29 other papers acknowledging MODEXTREME support, the organization of more than 20 dissemination and training events, and the production of 31 software units and packages corresponding to entirely new or refactored applications, model components and modelling solutions.
Project Context and Objectives:
CONTEXT
Model-based estimates of agricultural production are key elements for food security in the context of climate change. More important than mean temperature rise, projected climate change calculations indicate that frequency and severity of atmospheric extremes - which include floods, droughts, severe heat and cold, and storms - are altered by changing climate means. Changes in precipitation distribution, along with increased frequency and intensity of extreme weather events likely determine an augmented variability of agricultural productions and price volatility as well as changes in commercial exchanges. Also, the impacts of extreme events on crop/grassland/tree productivity are of increasing concern for food security policies. Changes in climate drivers could alter the frequency and magnitude of hazards, and changes in socio-economic drivers can lower the capacity of agricultural systems to cope with hazards. Extreme events are often combined and different plant stress factors (e.g. high temperature, evaporative demand and light) may simultaneously occur with different correlations among geographical areas. This generates an apparent complexity in the climate forcing-plant response relationships across a wide range of temporal and spatial scales and a poor description by simulation models. For instance, many biological processes may undergo sudden shifts at particular thresholds for temperature or precipitation. However, the mechanistic links between plant processes, and the relative roles in these processes of extreme events, have been only fragmentarily documented. Therefore, the extent of the inherent response to extreme weather events that plant species and varieties may be able to undergo remains an open field of research.
VISION OF THE PROJECT
A main concept of the project is that novel but still relatively simple responses to each climatic variable may generate complex responses. Most models have been established and calibrated on genotypes released in the years 1990s and on climate patterns to which plants were well adapted. Consequently, such modelling solutions had to be evaluated and possibly new capabilities could be required to predict plant responses under altered climate patterns. Translating basic knowledge into system models (and evaluating them) requires an interdisciplinary approach combining plant eco-physiology, software engineering and climatology. The MODEXTREME consortium has been built along this interdisciplinary approach and the project aimed to extend and strengthen European research in cooperation with non-EU partners and international organizations. MODEXTREME has brought together on-going efforts made by European and international teams on the modelling of extreme events impacts (including groups involved with knowledge of the development and use of MARS infrastructure, and involvement in international modelling initiatives such as AgMIP and MACSUR).
The project has three main objectives:
1. SCIENTIFIC CONTENT: To formalise and integrate into statistical and process-based simulation models the responses of main crop/grassland/tree systems to extreme weather events, assess them against existing datasets and use them to generate possible trajectories of agricultural productivity in short-medium time horizons.
2. INFRASTRUCTURE: To develop generically reusable software units implementing libraries of models that extend the modelling capabilities of the BioMA (Biophysical Model Applications) platform towards weather extremes, within a framework to support the monitoring and forecasting system of the European Commission Joint Research Centre (MARS: Monitoring Agricultural ResourceS).
3. SOCIO-INSTITUTIONAL: To transfer the knowledge gained and the modelling solutions developed to MARS analysts and model users, for dedicated implementation with local stakeholders to improve yield forecasts and increase food security at the EU level and worldwide.
Farmers, consultants, experimental scientists and decision-makers in Europe, Latin America, People’s Republic of China, and Sub-Saharan Africa are key beneficiaries of the project. MODEXTREME has, for instance:
- Tested modelling solutions for the benefit of model users and agricultural advisors
- Delivered guidelines to build, select, post-process and use climate scenarios into crop/grassland/tree simulation modelling for the benefit of model users and agricultural advisors, and to support food security policies
- Extended modelling tools and platforms beyond the impact of weather extremes for the benefit of model users, policy makers and agricultural advisors
Several dissemination and training courses have been organized to improve the understanding of simulation tools, dedicated libraries and model development environments for the benefit of scientists, technicians, advisors and policy makers, building model capacity regarding both use and development of models.
Project Results:
Key result no 1: Development of indices apt to gauge impacts of critical temperatures and water shortage on crop performance (Task 1.1)
Workpackage: WP1
Research aims and background
The existent modelling solutions for the simulation of arable crops, grasslands and trees are developed and tested using the average patterns of weather drivers and are considered satisfactory even if their predicting capacity fails when extreme weather events appear. This situation hinders the correct yield prediction under the specific situations of some extreme meteorological years (considered “outliers”), although the overall performance of the model is actually good. For some model applications, like the prediction of yield for a given year, the overestimation of yield after an extreme weather event is a serious issue for institutions and agencies devoted to agricultural production forecasting.
Results and applications
Weather extremes (namely low and high temperature extremes and drought) were identified with thresholds as functions of their effects on crop yield. A set of eight core indices was defined and tested on the basis of daily data for 1975-2013 as obtained from JRC for France and Spain. Statistical analysis of the indices allowed to identify critical years both at the local as well as at the regional scale. A web-based application to analyse existing datasets and identify potentially yield-reducing events was developed (WEBEXTREME: http://www.modextreme.org/webextreme).
Significance and benefits
The procedures for the identification of extreme events as function of their effect on crop production are new and allow a univocal identification of the years that bring potential production effects for several crops out of long time series. The procedure was automated and tested; the potential benefits are open to all the Europeans countries.
Successful applications
The identification of extreme events was implemented by means of a web-based application open to the users worldwide.
References:
MODEXTREME Deliverable D1.1. Identification of extreme events, statistical adjustments and modelling of extreme yield anomalies.
Tommy Klein, Pierluigi Calanca, Tamara Ben Ari, David Makowski, Gianni Bellocchi, 2015, Spatial extension of heat and drought stress in European wheat producing areas, Poster at 5th AgMIP Global Workshop, February 25-28, Gainesville, FL, USA. http://modextreme.org/wp-content/uploads/sites/4/2015/02/Poster_KleinEtAl_v3.pdf
Tommy Klein, Argyrios Samourkasidis, Ioannis N. Athanasiadis, Gianni Bellocchi, Pierluigi Calanca, 2016, webXTREME: a simple web tool for calculating agroclimatic indicators of extreme events In: Sauvage, S., Sáchez-Pérez, J.-M. Rizzoli, A.E. (Eds.), Proceedings of the 8th International Congress on Environmental Modelling and Software, July 10-14, Toulouse, France, Vol. 2, 471. ISBN: 978-88-9035-745-9.
Tommy Klein, Argyrios Samourkasidis, Ioannis N. Athanasiadis, Gianni Bellocchi, Pierluigi Calanca, webXTREME: R-based web tool for calculating agroclimatic indices of extreme events. Submitted to Computers and Electronics in Agriculture.
Key result no 2: Development of a general modelling framework to account for the yield reduction in many crops by means of effects on harvest index (Task 1.2)
Workpackage: WP1
Research aims and background
The existent crop modelling solutions are developed and tested using the average patterns of weather drivers and are considered satisfactory even if their predicting capacity fails when extreme weather events appear. This situation hinders the correct yield prediction under the specific situations of some extreme meteorological years (considered “outliers”), although the overall performance of the model is actually good. For some model applications, like the prediction of yield for a given year, the overestimation of yield after an extreme weather event is a serious issue for institutions and agencies devoted to agricultural production forecasting.
Results and applications
The modelling framework developed in MODEXTREME assumes that yield variations due to an extreme event (cold temperature, high temperature or water deficit) are due to flower death and failure in pollination (high or low temperatures) while water stress reduces seed set. All these effects are mediated by a change in Harvest Index (HI), while the main effect of weather on crop performance is already captured by existing crop models. In crops whose yield is a vegetative organ (e.g. potato) the response is calculated over the Leaf Area Index. The response functions of HI depression vs. environmental critical conditions were developed and formalized. The framework allows also to account for the combinations of multiple events during and after the flowering, and of the joint effect of high temperature and water stress. Although tables with the critical thresholds for the main crops have been provided, the modelling framework is theoretically applicable to any crops even beyond those mentioned in the DOW after specific calibration.
Significance and benefits
The modelling framework demonstrated to be easily applicable to many crops and to increase the yield prediction capability of already existing models on extreme years. The general HI approach has been found to be simple enough to be applicable to many crops after calibration of the response functions parameters, while being robust and performant. The HI-based modelling framework for the extreme weather impact on yields may also be applied as an independent module over the output of the crop models, allowing to deal with the extreme-event effect without need of modifying or re-coding the existing modelling solutions.
Successful applications
The approaches developed to model the impact of extreme events were implemented in a dedicated software component for use within crop modelling frameworks.
References:
MODEXTREME Deliverable D1.2. Report on modelling approaches for simulating the impact of extreme events on agricultural production.
Gianni Bellocchi, Francisco J. Villalobos, Marcello Donatelli, Ole B. Christensen, Oscar Rojas, Roberto Confalonieri, Ioannis N. Athanasiadis, Irina Carpusca, Claudio O. Stöckle, 2014, Extending existing models to capture vegetation response to extreme weather events: the MODEXTREME project In: Ames, D.P. Quinn, N.W.T. Rizzoli, A.E. (Eds.), Proceedings of the 7th International Congress on Environmental Modelling and Software, June 15-19, San Diego, California, USA. ISBN: 978-88-9035-744-2. http://www.iemss.org/sites/iemss2014/papers/iemss2014_submission_215.pdf
Francisco J. Villalobos, José Paulo de Melo e Abreu, Luca Testi, Gianni Bellocchi, 2015, Impact of extreme meteorological events on crop yield: a common framework approach, Poster at 5th AgMIP Global Workshop, February 25-28, Gainesville, FL, USA. http://modextreme.org/wp-content/uploads/sites/4/2015/02/Poster-Agmip_T1.2.pdf
Key result no 3: Micrometeorological calculation methods for the estimation of canopy temperature from air temperature and disaggregation of daily maximum and minimum temperature into hourly temperatures
Workpackage: WP1
Research aims and background
Both the available historical datasets and the future climate projections of meteorological data have a time resolution of day (maximum and minimum air temperatures). While this resolution is usually adequate for the majority of crop models, at the time of calculating the effect of extreme temperatures over the plant is not the air but the canopy temperature what matters, and the exact amount of time during which the canopy remains in a dangerous temperature state. There was no practical and generalized way to account for these key issues in the existing modelling solutions.
Results and applications
Advanced calculation methods based on crop micrometeorology were presented, formalized and tested with previous high-resolution canopy temperature datasets. These methods allow the disaggregation of daily maximum and minimum air temperatures into a diurnal time series of air temperature at an hourly or shorter time step, using empirical polynomial or sine disaggregation during the day and a modified application of Brunt law during the night. From the air temperature pattern, the canopy temperature is calculated by means of advanced applications of methods based on crop energy balance and aerodynamic features.
Significance and benefits
These calculation methods are new and original. They allow to calculate the canopy temperature in any time of the day or the night of virtually any crop, whenever its precise definition is required, from standard meteorological data.
Successful applications
These methods were integrated in the module for the calculation of yield response to extreme events.
References:
MODEXTREME Deliverable D1.2. Report on modelling approaches for simulating the impact of extreme events on agricultural production.
Jose Paulo Melo-Abreu, Francisco J. Villalobos, Luciano Mateos, 2016. Frost protection. In: Villalobos F.J. Fereres E. (Eds.), Principles of Agronomy for Sustainable Agriculture. Springer. ISBN 978-3-319-46115-1.
Francisco J. Villalobos, Luca Testi, Luciano Mateos, Elías Fereres, 2016, Manipulating the crop environment. In: Villalobos, F.J. Fereres, E. (Eds.), Principles of Agronomy for Sustainable Agriculture. Springer. ISBN 978-3-319-46115-1.
Villalobos FJ, Lopez-Bernal A. In press. El clima. In: Barranco D, Fernández-Escobar R, Rallo L (Eds.). El cultivo del olivo. Mundi-Prensa, 213-249. (in Spanish)
Francisco J. Villalobos, José Paulo de Melo e Abreu, Luca Testi, Gianni Bellocchi, 2015, Impact of extreme meteorological events on crop yield: a common framework approach, Poster at 5th AgMIP Global Workshop, February 25-28, Gainesville, FL, USA. http://modextreme.org/wp-content/uploads/sites/4/2015/02/Poster-Agmip_T1.2.pdf
Key result no 4: Development of statistical models for ex-post (static) adjustment of estimated yields, suitable for application at the “statistical level” of the MARS monitoring system (Task 1.4)
Workpackage: WP1
Research aims and background
The existent crop modelling solutions are developed and tested using the average patterns of weather drivers and are considered satisfactory even if their predicting capacity fails when extreme weather events appear. This situation hinders the correct yield prediction under the specific situations of some extreme meteorological years (considered “outliers”), although the overall performance of the model is actually good. For some model applications, like the prediction of yield for a given year, the overestimation of yield after an extreme weather event is a serious issue for institutions and agencies devoted to agricultural production forecasting.
Results and applications
Yield statistics over the past period were used to derive statistical indicators of climate impact on plant production. To this end, climate variables aggregated at the national scale and/or climate indicators developed in task T1.1 were used.
For statistical adjustment, models relying on single climate variables perform equally well as models relying on complex indices. Monthly indices were found useful to target key time periods. Variables/indices with the highest predictive capacity were identified. Prediction of yield extremes is improved by combining several climate indices using logistic regression.
Significance and benefits
The statistical indices developed and tested are now available for cereals yield forecasting in Europe.
Successful applications
Scripts for the calculation of statistical indices are available for integration to the statistical layer of the MARS crop monitoring and forecasting system.
References:
MODEXTREME Deliverable D1.1. Identification of extreme events, statistical adjustments and modelling of extreme yield anomalies
Tamara Ben-Ari, Juliette Adrian, Tommy Klein, Pierluigi Calanca, Marjin Van der Velde, Stefan Niemeyer, Gianni Bellocchi, David Makowski, 2015, Identifying accurate climate indicators of extreme yield loss in Europe, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/Poster_CSA_8.pdf
Tamara Ben-Ari, Juliette Adrian, Tommy Klein, Pierluigi Calanca, Marijn Van der Velde, David Makowski, 2016, Identifying indicators for extreme wheat and maize yield losses, Agricultural and Forest Meteorology 220, 130-140.
Key result no 5: Development of OLIVECAN, a model to simulate the water and carbon exchange of olive groves. (Task 1.3)
Workpackage: WP1
Research aims and background
Models of tree crops are rare and those existent often are focused to cover only some specific processes of interest. The development of a complete model for the olive tree, the main tree crop in Europe, with the capacity of predicting net the carbon assimilation, water use and yield formation is extremely useful for the olive crop cultivation and oil production industry.
Results and applications
OLIVECAN is a model aimed at the simulation of water and carbon exchange between olive orchards and the atmosphere. It is composed of sub-models that deal with the more important processes like interception of PAR radiation, soil water uptake, potential and actual transpiration, photosynthesis and assimilation of carbon, respiration rates of the different organs and yield formations. The model was conceived to work under localised irrigation conditions (i.e. calculates the water balance over two different soil compartments with different profiles of water content). The model includes the harvest index responses functions to extreme temperatures (heat and frost) and drought developed in task 1.2.
Significance and benefits
OLIVECAN has a strong predictive capacity for water use, biomass accumulation and yield. It can be used at different level of complexity to simulate the physiological response of olive to different cropping management (i.e. irrigation, pruning, planting strategies) and is able to respond correctly to the CO2 increase of future conditions in terms of transpiration, photosynthesis and respiration processes, including the effects of extreme meteorological events on yields.
Successful applications
A novel version of OLIVECAN was successfully tested in Spanish conditions.
References:
MODEXTREME Deliverable D1.2. Report on modelling approaches for simulating the impact of extreme events on agricultural production.
Omar García-Tejera, Alvaro Lopez-Bernal, Luca Testi, Francisco J. Villalobos, 2017, A soil-plant-atmosphere continuum (SPAC) model for simulating tree transpiration with a soil multi-compartment solution, Plant and Soil, doi:10.1007/s11104-016-3049-0
Omar Garcia-Tejera, Álvaro Lopez-Bernal, Francisco J. Villalobos, Francisco Orgaz, Luca Testi, 2016, Effect of soil temperature on root resistance: implications for different trees under Mediterranean conditions. Tree Physiology 36, 469-478.
Alejandro Morales, Peter A Leffelaar, Luca Testi, Francisco Orgaz, Francisco J. Villalobos, 2016, A dynamic model of potential growth of olive (Olea europaea L.) orchards. European Journal of Agronomy 74, 93-102.
Key result no 6: Improvement of BioMA framework
Workpackage: WP2
Research aims and background
BioMA is a modelling framework designed to develop reusable software components for composing modelling solutions for the integrated analysis of agro-ecological systems (http://www.biomamodelling.org). It consists of a set of core components, a suite of coding assistant tools plus a set of Window applications for configuring simulations, running the models, optimizing models against reference data, exploring the outputs and statistically analyse the model estimates. The technology provided by BioMA appears particularly suitable to the objectives of MODEXTREME. The project activities require handy tools to run multiple models, which must be easily extendible to increase their explanatory power in order to analyse the crop response to multiple biotic and abiotic stresses.
The BioMA framework can satisfy all these needs; first, by making available flexible Modelling Solutions (MS) which can be easily extended by integrating additional components to include more functionalities. By adding the MODEXTREME.WeatherExtremesImpact component (see D2.1 Movedi et al., 2016) to MS based on existing crop models (e.g. CropSyst), new MS can be obtained with the additional capacity of simulating crop response to extreme weather events. Both conventional and the extended version can be compared. A set of ancillary applications, embedded in the BioMA-Site and BioMA-Spatial as plugins, assisted MODEXTREME users across the whole chain of model testing activity, from model optimization to analysis of output estimates. BioMA was completely refactored and optimized around MODEXTREME needs, but it is a mature technology which can be fruitfully applied to other research areas involving agroecosystem analysis. For instance, current ongoing applications are the impact analysis of climate change, and the optimization of agricultural management.
Results and applications
This software has enabled MODEXTREME partners to perform all the simulation workflow steps: simulation configuration, model parameterisation, model optimization and statistical analysis of output, data view and exploration.
The products which were made available are the following ones:
i. BioMA-Site and BioMA-Spatial: MS Windows applications to deploy and run modelling solutions. They are also the interfaces which expose all the other functionalities as plugin tools.
ii. Code generator tools: they facilitate model components development, limiting the coding effort requested to modellers to the algorithmic representation of models.
iii. Plugin tools for simulation configurations, parameter editing, component structure exploring and data view and analysis: MPE Model Parameter Editor, MSE Modelling Solution Explorer, ACG AgroManagement Configurator Generator, MCE Model Component Explorer, GDD Graphical Data Display.
iv. Applications for optimizing models and analysing results: LUISA (sensitivity analysis), Optimizer (model optimization against reference data), IMMA (statistical analysis of estimates).
All these tools can be downloaded from the BioMA portal (http://www.biomamodelling.org) along with modelling solutions. Documentation, tutorials, and sample simulations are provided. D2.3 (Donatelli et al., 2016) and Donatelli et al. (in preparation) give an overview of the available tools.
Significance and benefits
The BioMA platform represents a significant advancement in the field of agroecosystem modelling targeting flexibility in modelling solution development and re-usability of discrete model units. A large set of resources are made publicly available, which allow a vast set of applications. The BioMA architecture was designed to allow the extensibility of components, so resources can be easily extended and adapted virtually to most user needs.
Successful applications
WP4 partners extensively tested released modelling solutions, and BioMA-Site and Spatial applications. WP4 produced a large amount of simulation files covering the whole South-East Europe, Ukraine and portions of China, Argentina, Brazil, and South Africa. WP4 partners made available country-specific soil parameter sets, crop parameters files as resulting from their calibration, peculiar agro-management files, and tested BioMA-Site/Spatial configurations (available at: http://modextreme.org).
References:
MODEXTREME deliverable D2.1. Abiotic stress model component: v1.
MODEXTREME deliverable D2.3. The BioMA development.
Marcello Donatelli, Davide Fumagalli, Iacopo Cerrani, Davide Fanchini, Simone Bregaglio, Ioannis Athanasiadis, Andrea E. Rizzoli, Enhancing biophysical models reuse: shifting the focus from frameworks to components in the BioMA framework. In preparation for submission to Agricultural Systems.
Key result no 7: Development of Modelling Solutions
Workpackage: WP2
Research aims and background
The interest on simulation models for the integrated analysis of agroecosystems has increased over the last decades. Furthermore, new functionalities are continuously required and models are expected to be always to date with scientific advancements, to support emergent scientific and practical applications.
In principle, models with such requirements could be built by assembling the most advanced modelling resources available, but the lack of standardization and the heterogeneity of languages, approaches and programming paradigms, has prevented so far this possibility.
With the advent of innovative programming concepts targeting software reusability, the possibility of composing simulation tools from reusable components has become realistic. These concepts are at the very foundation of the BioMA framework, where fully working modelling solutions can be assembled by aggregation of smaller software components. By the same way, it can be possible to produce extended or modified versions of an existing model by just adding new components to take care of additional tasks, or by integrating alternative versions of a component already present.
In practice, component reusability solves the “reinventing the wheel” problem. If a large part of a new simulation system can be built using ready-made reusable components, the developers’ efforts can be concentrated on the new functionalities and on the related science, thus accelerating the development process, and improving the scientific content of software.
It therefore becomes easier to explore multiple versions of models, thus helping their improvement, and to use multiple model for more robust analysis. This has been the case in MODEXTREME, where modelling solution were built also reusing existing model libraries previously implemented in software components.
Results and applications
The team of WP2 released a total of seven fully working modelling solutions (MS), notably: i) CropSyst; ii) CropSyst+Extreme; iii) WOFOST; iv) WOFOST + Extreme; v) WARM; vi) WARM + Extreme; vii) AquaCrop (+ an “Extreme” version still under development).
These MS allows the simulation of cropping systems for the potential and water-limited production. The “extreme” versions are extensions of basic models, accounting for the effects of stress deriving from extreme weather events, such as frost, heat and drought.
They can be applied to various scientific and practical applications, especially where water limitation is relevant to the study. The individual components can be utilized as well in new MS or new components can be added to the given MS to produce extended version for new applications.
The MS can be downloaded (http://www.biomamodelling.org) and deployed into either BioMA-Site (for one site simulation) or into BioMA-Spatial (for multiple locations). D2.4 (Donatelli et al., 2016) and Donatelli et al. (in preparation) outline concepts at the basis of MS composition.
Significance and benefits
For model developers, the availability of a set of software components implementing most common models for basic process in biophysical modelling, allow easily building MS around specific needs, and extending already existing ones by just adding the new features desired to fully describe the system of interest. For model users, the availability of a large set of MS allow for many applications in various domains, with the possibility of adopting multi-model approaches thanks to the availability of powerful graphical user interfaces to run MS and analyse the outputs. Respect to previously existing modelling resources, those released during the MODEXTREME project present other additional advantages, such as higher transparency about algorithms, variables and parameters, parameter editing and optimization facilities, an easily configurable and extendible management simulation model, incorporated in each MS.
Successful applications
The MS are being used in other research projects. Ongoing current researches are employing them in scenario analysis to assess the impact of climate change on annual crops in Europe. Another ongoing application uses MS to optimize management strategies in view of upcoming climate change.
References:
MODEXTREME deliverable D2.4. Modelling solutions.
Marcello Donatelli, Davide Fumagalli, Iacopo Cerrani, Davide Fanchini, Simone Bregaglio, Ioannis Athanasiadis, Andrea E. Rizzoli, Enhancing biophysical models reuse: shifting the focus from frameworks to components in the BioMA framework. In preparation for submission to Agricultural Systems.
Key result no 8: Platform interoperability demonstrative actions
Workpackage: WP2
Research aims and background
The BioMA platform was designed to maximise component reusability, which is the capability of easily replace software units within a given software framework. In BioMA the architecture chosen has explicitly included features to make BioMA-components and modelling solutions able to run/be composed even in other frameworks with a reasonable ease, to maximise reusability to the benefit of model development and knowledge sharing.
Model reuse can have different meanings; within biophysical modelling, a “model” can be:
• A process-based representation of either a single process or of a compartment;
• A whole modelling solution allowing the simulation of the system of interest.
However, the point in reuse is not the granularity of the model unit, as in model development. The key difference is if either the discrete unit is to be reused in workflows, or is composed to iterate with other model units over the execution of time-steps. If the reuse is of modelling solutions, adaptors are built following the requirements of a specific software framework either to enable links in workflows, or for ensemble runs. In this case the “model” units run asynchronously and potential incompatibilities of binaries can be overcome. In the case of iteration across model units over time, the requirements for an effective reuse increase. Key features in the architecture of model units are required and binaries incompatibility across models units becomes an insurmountable obstacle in practical terms. Model development requires being able to modify and link model units working synchronously at run time.
Results and applications
Some demonstrative actions were undertaken, consisting in setting up fully working modelling solutions under the following conditions:
1. BioMA components incorporated into models developed under external frameworks;
2. Whole BioMA modelling solutions running into external frameworks;
3. Models designed for external frameworks running into BioMA-compliant modelling solutions;
For all these case studies, the key to the successful implementation is the correct clock handling, which in turns impact on integration mechanism. In the first case study, the BioMA Disease components were incorporated into the APSIM-Wheat model. In the second case study, the BioMA CropSyst-Extreme modelling solution was implemented into an APSIM model container, becoming a modelling solution runnable in that software framework. As for the preceding case, clock handling, in BioMA performed at the Composition Layer level, was handled by APSIM. These experiences allowed refining the procedures for aggregating components as implemented in CLIC (Composition Layer Interactive Coder), the coding-assistant tool to perform composition of modelling solutions. A new feature has been introduced, to generate APSIM-compliant Modelling Solutions in which clock handling is excluded from the model runner and demanded to APSIM.
In the third case study, the converse operation was attempted, that is, to run a non-BioMA compliant software within a BioMA modelling solution. The PaSim model developed by INRA was chosen for this study. PaSim core structure was written in C++, according to different programming paradigms. The interoperability problem was approached by developing a non-BioMA C# wrapper, which implemented homologous methods mapping individually to each original PaSim method. This wrapper has to be called from within a BioMA strategy, to run the whole process workflow. A prototype wrapper with limited functionality has been realized, which proved to function within a BioMA strategy, which in turn could be composed into a demo-MS. Case studies are documented in D2.5 (Donatelli, 2016) and in Holzworth et al. (in preparation).
Significance and benefits
Re-using software units in different modelling platforms substantially means making available to researchers, so far anchored just to their own platform, the opportunity to share an enormous stock of knowledge. The actions that have been undertaken in the MODEXTREME project represent just the proof that the platform interoperability can be achieved. At first, however, technical specifications on how to reach interoperability need to be discussed. Burning questions to answer regard the type of discrete units to share among platforms (strategies implementing simple processes, specific domain component, whole modelling solution), the programming language to use, and a minimum set of requirements for the API - application programming interface.
Successful applications
The BioMA CropSyst-Extreme modelling solution was successfully linked to the APSIM framework.
References:
MODEXTREME deliverable D2.5. Modelling solutions with adapters to other platforms. MODEXTREME project deliverable.
Dean P. Holzworth, Simone Bregaglio, Jeremy P.M. Whish, Marcello Donatelli, Modelling frameworks interoperability: The case BioMA-APSIM. In preparation for submission to Environmental Modelling & Software.
Key result no 9: Provision of climate model based extremes indices for agricultural model changes
Workpackage: WP3
Research aims and background
One important use of agricultural models is to provide information about possible changes in yields caused by anthropogenic climate change in the future. In the MODEXTREME project several important model improvements have been developed and implemented, and it is relevant to study how these improvements affect projections of climate change impacts. As part of this work, a novel set of indicators of extreme weather events appropriate for crop/grassland/tree modelling was identified. These indicators had the purpose to help with decisions about important changes in agricultural circumstances due to climate change.
Results and applications
Bias correction was applied to CORDEX RCM model data before calculation of indices.
In total the indices are (full details can be found in deliverable D3.4):
• Seven standard temperature indices (FD, SU, TR, GSL, WSDI, WSDImax, CSDI)
• Seven standard precipitation indices (RX1day, RX5day, SDII, R10mm, R20mm, CDD, CWD)
• Seven cold temperature phenological indices (S.EM.CRIT.8 S.EM.CRIT.3 EM.AN.CRIT.2 LASTFROST.0 LASTFROST.8 LASTFROST.3 LASTFROST.2 – NB the first three indices require information about sowing, emergence and flowering dates which was only available for SW Europe, hence these indices are only provided for this region)
• Four extreme heat phenological indices (TMAX.40 TMAX.45 GSL.GRASS VHOT.DAYS)
For each of the available 12km Euro-CORDEX simulations and all MODEXTREME study areas, these 25 different indices based on average and extreme values of temperature and precipitation have been made available on the same server and can be accessed at http://ensemblesrt3.dmi.dk/data/MODEXTREME/extremes. These indices include a number of phenological indices relating to both extreme heat and extreme cold. Seven of these indices, all precipitation based, were used in the model selection process.
Significance and benefits
Changes in these indices can be studied and will lead to indications about the most important meteorological changes associated with anthropogenic climate change. This can be used to put calculated results with the MODEXTREME agricultural models into perspective.
Successful applications
Adding this layer of climate detail to the discussion about the future of the agricultural sector in the context of climate change has led to new conclusions about the impact of extreme weather events.
References:
MODEXTREME deliverable D3.1. Historical and perturbed baseline time series for scenario periods.
MODEXTREME deliverable D3.2. Processed RCM/GCM output.
MODEXTREME deliverable D3.3. Report on methods, guidance and recommendations.
MODEXTREME deliverable D3.4. Indices of extremes, simulated and observed.
MODEXTREME deliverable D3.5. Report on projected changes in indices/extremes.
Cathrine Fox Maule, Ole B. Christensen, Peter Thejll, 2014, Selection of a best subset of GCM-RCMs from an ensemble for impact studies, Poster at 3rd Lund Regional-scale Climate Modelling Workshop, June 16-19, Lund, Sweden. http://modextreme.org/wp-content/uploads/sites/4/2014/09/RCM2014_Lund_MODEXTREME_4.rar
Cathrine Fox Maule, Marianne Sloth Madsen, Wilhelm May, Jens Hesselbjerg Christensen, Shuting Yang, Ole Bøssing Christensen, 2015, On procedures for model selection in providing climate scenario data for impact studies – A challenge to both communities, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-9244, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-9244.pdf
Cathrine Fox Maule, Peter Thejll, Richard Cornes, Clare Goodess, Ole B. Christensen, 2015, Effects of bias correction on the climate change signal of extreme indices of precipitation, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-4489, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-4489.pdf
Ole B. Christensen, Cathrine Fox Maule, Richard Cornes, Clare Goodess, Gianni Bellocchi, 2015, Use of regional climate model output for modelling the effects of future weather extremes on agriculture, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/150316_MODEXTREMEWP3_RCC_CMG_OBC.pdf
Fabrizio Ginaldi, Gianni Fila, Davide Fumagalli, Antonio Zucchini, Marcello Donatelli, Exploring the impact of climate variability estimates on crop models predictions in CC impact assessment studies. In preparation for submission to European Journal of Agronomy.
Key result no 10: Bias correction of climate model data for sensitivity studies of agricultural model changes
Workpackage: WP3
Research aims and background
One important use of agricultural models is to provide information about possible changes in yields caused by anthropogenic climate change in the future. In the MODEXTREME project several important model improvements have been developed and implemented, and it is relevant to study how these improvements affect projections of climate change impacts. In WP3 it was the aim to provide processed output from climate models for this purpose to the other work packages in the project. This necessitates performing bias correction of relevant fields for the use in agricultural models, which are quite sensitive to biases in meteorological variables.
Results and applications
Bias correction methods have been developed and implemented for regional climate model (RCM) simulations from the CORDEX project for selected periods, resulting in sets of bias corrected gridded daily data sets of daily maximum and minimum surface air temperature as well as daily total precipitation. Other meteorological variables, in particular solar irradiation, have been calculated with the help of weather generators due to a lack of consistent measurements, which prevents bias correction of daily data. For south-western Europe the observational dataset underlying the bias correction is the gridded E-OBS dataset, which provides all three variables on a 25 km-grid. RCM output data have been interpolated to this grid, and quantile-quantile based bias correction has been applied. For the other MODEXTREME study areas, station observations have been used, either from public databases or provided by local project partners.
The quantile-based method is constructed by sorting all values for the relevant point and period in the model and correspondingly sorting observation values. Then, each fractile (the highest, second-highest, etc.) of the model is corrected to the corresponding observational value. This does not affect the chronology of the modelled events, and we will therefore hardly ever have unphysical situations like heavy precipitation coinciding with very high temperature due to bias correction. A piecewise linear function is then adapted and employed also for values outside the period used for fitting. For values entirely outside the range of model values in the fitting period, constant correction is used taking from the nearest fractile.
Significance and benefits
Plants are extremely sensitive to small variations in temperature and precipitation, and therefore agricultural models have a very high bar for the minimum quality of input data from numerical climate models. For this reason bias correction is a necessity for obtaining information about agriculture in possible changed future climates with any hope of realism. In MODEXTREME the focus has been on indices of extremes most relevant for impact on crop/grassland/tree species, including a number of new phenological indices which are not normally used by the climate research community. Several different European projects have constructed bias corrected climate simulation data using a variety of methods and underlying observations. At the start of MODEXTREME this was not the case, and therefore the work performed in WP3 was necessary for being able to investigate the effects of climate extremes on agricultural outputs using the models developed in MODEXTREME. Even though alternatives do exist now, the MODEXTREME climate datasets are still deemed useful and applicable to not just agricultural modelling but also to other forms of impacts models.
Successful applications
Adding this layer of climate detail to the discussion about the future of the agricultural sector in the context of climate change has led to new conclusions about the impact of extreme weather events.
References:
MODEXTREME deliverable D3.1. Historical and perturbed baseline time series for scenario periods.
MODEXTREME deliverable D3.2. Processed RCM/GCM output.
MODEXTREME deliverable D3.3. Report on methods, guidance and recommendations.
MODEXTREME deliverable D3.4. Indices of extremes, simulated and observed.
MODEXTREME deliverable D3.5. Report on projected changes in indices/extremes.
Cathrine Fox Maule, Ole B. Christensen, Peter Thejll, 2014, Selection of a best subset of GCM-RCMs from an ensemble for impact studies, Poster at 3rd Lund Regional-scale Climate Modelling Workshop, June 16-19, Lund, Sweden. http://modextreme.org/wp-content/uploads/sites/4/2014/09/RCM2014_Lund_MODEXTREME_4.rar
Cathrine Fox Maule, Marianne Sloth Madsen, Wilhelm May, Jens Hesselbjerg Christensen, Shuting Yang, Ole Bøssing Christensen, 2015, On procedures for model selection in providing climate scenario data for impact studies – A challenge to both communities, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-9244, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-9244.pdf
Cathrine Fox Maule, Peter Thejll, Richard Cornes, Clare Goodess, Ole B. Christensen, 2015, Effects of bias correction on the climate change signal of extreme indices of precipitation, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-4489, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-4489.pdf
Ole B. Christensen, Cathrine Fox Maule, Richard Cornes, Clare Goodess, Gianni Bellocchi, 2015, Use of regional climate model output for modelling the effects of future weather extremes on agriculture, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/150316_MODEXTREMEWP3_RCC_CMG_OBC.pdf
Key result no 11: Selection of climate model data sets for sensitivity studies of agricultural model changes
Workpackage: WP3
Research aims and background
One important use of agricultural models is to provide information about possible changes in yields caused by anthropogenic climate change in the future. In the MODEXTREME project several important model improvements have been developed and implemented, and it is relevant to study how these improvements affect projections of climate change impacts. As part of this work, a novel set of indicators of extreme weather events appropriate for crop modelling was identified. In WP3 it was the aim to provide processed output from climate models for this purpose to the other work packages in the project. This processing consists of two major aspects: Bias correction of relevant fields for the use in agricultural models, which are quite sensitive to biases in meteorological variables; and an implemented method for selection of representative climate model simulations for cases where the amount of computations involved in agricultural modelling makes it necessary to not study all available climate model simulations. It was decided to use at most four regional model simulations selected as one central and 3 sampling the set of projected changes in a set of seven precipitation-based indices. This selection is done for each area under investigation: for south-west Europe as a whole, for Ukraine, China, South Africa, and South America individually.
Results and applications
For determination of model differences, the following precipitation based indices were used:
• PR Average precipitation
• CC Climate Change signal of annual mean
• RX1day Highest precipitation amount in a one-day period
• RX5day Highest precipitation amount in a five-day period.
• SDII Simple Daily Intensity Index
• R10mm Heavy precipitation days
• R20mm Very heavy precipitation days
• CDD Consecutive Dry Days
• CWD Consecutive Wet Days
These indices were calculated for the present-time as well as the far-future bias-corrected time slice of 30 years for all models. The relative change of each index for the area in question then form a nine-element vector for each model, and ordinary PCA analysis can be applied in order to choose the most different modelled climate change. Employing a principal-component analysis method for model selection, four simulations were selected from a set of 10 simulations for south-west Europe and for Ukraine and from a set of 18 simulations for South Africa. For the remaining areas no selection was necessary, as less than four simulations were available. The resulting sets of at most four simulations, bias corrected in order to have distribution functions identical to the available observational data for the present-day simulation period, have been available for most of the project duration at a server with access for project partners. For each simulation and study area, 25 alternative indices based on average and extreme values of temperature and precipitation have been made available on the same server. These indices include a number of phenological indices relating to both extreme heat and extreme cold.
Significance and benefits
Selection of ensemble members from a larger ensemble is necessary for the purpose of many different kinds of impacts modelling with computer-heavy impacts models. This is the case for the agricultural models in MODEXTREME, but the concept is generally useful.
Successful applications
Adding this layer of climate detail to the discussion about the future of the agricultural sector in the context of climate change has led to new conclusions about the impact of extreme weather events.
References:
MODEXTREME deliverable D3.1. Historical and perturbed baseline time series for scenario periods.
MODEXTREME deliverable D3.2. Processed RCM/GCM output.
MODEXTREME deliverable D3.3. Report on methods, guidance and recommendations.
MODEXTREME deliverable D3.4. Indices of extremes, simulated and observed.
MODEXTREME deliverable D3.5. Report on projected changes in indices/extremes.
Cathrine Fox Maule, Ole B. Christensen, Peter Thejll, 2014, Selection of a best subset of GCM-RCMs from an ensemble for impact studies, Poster at 3rd Lund Regional-scale Climate Modelling Workshop, June 16-19, Lund, Sweden. http://modextreme.org/wp-content/uploads/sites/4/2014/09/RCM2014_Lund_MODEXTREME_4.rar
Cathrine Fox Maule, Marianne Sloth Madsen, Wilhelm May, Jens Hesselbjerg Christensen, Shuting Yang, Ole Bøssing Christensen, 2015, On procedures for model selection in providing climate scenario data for impact studies – A challenge to both communities, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-9244, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-9244.pdf
Cathrine Fox Maule, Peter Thejll, Richard Cornes, Clare Goodess, Ole B. Christensen, 2015, Effects of bias correction on the climate change signal of extreme indices of precipitation, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-4489, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-4489.pdf
Ole B. Christensen, Cathrine Fox Maule, Richard Cornes, Clare Goodess, Gianni Bellocchi, 2015, Use of regional climate model output for modelling the effects of future weather extremes on agriculture, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/150316_MODEXTREMEWP3_RCC_CMG_OBC.pdf
Key result no 12: Model evaluation
Workpackage: WP4
Research aims and background
The agricultural system is increasingly threatened by climate change with an increase in the frequency/intensity of extreme weather events. Simulation models of crop, grassland and tree systems have been often used in environmental conditions markedly different from the ones they were developed. This has been achieved by reproducing time trends of observations by fitting the values of the parameters included in core equations, degrading the process-based nature of models to purely statistical tools, hence unsuitable to explore conditions where model performance cannot be corroborated by model testing, as in scenarios of climate change. A model accounting for the impact of extreme weather events must be able to perform under both ordinary and anomalous weather conditions without changing its parameterization. Available crop, grassland and tree models were developed based on datasets collected – in most cases – under conditions for which plants were well adapted, thus without explicitly accounting for extreme weather events. This is one of the main reasons for the poor capability of models to capture the yearly variability in official yields at regional and country level and poor coherence between the mathematical description of crop/grassland/tree responses to the environment and the values of the parameters used. In WP4 Agricultural system models’ ability to simulate plant growth was evaluated by comparing existing (EMS) and modified (MMS) modelling solutions, the latter accounting for the impact of extreme weather events. The evaluations were done for the historical period against observed data, and for the future period with downscaled climate projections. The results were presented to a variety of stakeholders in each country to receive feedback. The WOFOST model’s sensitivity to different parameters was also tested.
Results and applications
Regarding evaluation with the observed data, the results highlight some limitations in the response to extreme weather events of state-of-the-art modelling solutions, as emerged from the assessment performed at a variety of sites and systems. It is difficult to draw a general conclusion from the comparison between modified and existing modelling solutions as the agro-ecosystems assessed are diverse, and the exercise was often based on weather and crop data of relatively short length. In the evaluation with future climate data, clearer differences between EMS and MMS emerged, in comparison to evaluation with historical data. For WOFOST, in general MMS was found to simulate lower yield and AGB than EMS due to the additional “extreme” component of MMS. For CropSyst, as with WOFOST, MMS tends to simulate lower yield and AGB than EMS by accounting for negative impacts of extreme weather events on crop growth. Similar results were obtained for PaSim (grassland model), OLIVECAN (olive tree model) and WARM (rice model).
The sensitivity analysis of WOFOST identified the most influential WOFOST parameters to simulate crop yield. The influential parameters vary depending on the combination crop x site. With a few exceptions, the parameters identified by the Morris method obtained the highest eFAST total sensitivity index, which underlines that the two sensitivity analysis methods are closely related. These results suggest the plasticity of the WOFOST model. The WP4 results indicate that it is important to consider the effect of extreme weather events in agriculture system models. It is likely that most simulations of plant growth without explicit consideration of “extreme” component overestimate crop productivity and underestimate its variability. The improvement introduced by MMS can be the basis for a discussion around the employment of these modelling tools to support food security issues in Europe and worldwide.
Significance and benefits
The results indicate the importance of considering the effect of extreme weather events in agriculture system models. Better modelling tools that perform well even under anomalous weather conditions can be built for operational use, for example, in-season yield forecasting and for climate change studies. Stakeholders suggest that the WP4 results are useful for elaboration of long-term agriculture development projects and policy planning for agriculture and food security, e.g. agricultural insurance, credits, investments, resource allocation/mobilization, irrigation systems. MODEXTREME models can also guide further research on agricultural vulnerabilities, variety trials, or plant breeding.
Successful applications
It is recommended to improve simulation of the effects of other extreme weather events such as heavy rainfalls or extremely cold temperatures. Integration of pests and diseases model is another suggestion. There was a high interest in the use of models for support to farm management decisions at much smaller scales to identify and implement appropriate adaptation strategies on the farm. To draw implication for the wider society, food security at the national and international levels needs to be discussed in a much broader context, and by accounting for other aspects of food security beyond food quantity (or productivity).
References:
MODEXTREME deliverable D4.1. Results on model testing against experimental/observational and gridded data.
MODEXTREME deliverable D4.2. Results on model evaluation in future climate scenarios.
MODEXTREME deliverable D4.3. Interaction with stakeholders in the frame of food security.
MODEXTREME deliverable D4.4. Report on the assessment of new modelling solutions.
Robert Mangani, Eyob Tesfamariam, Gianni Bellocchi, Abubeker Hassen, Comparison of existing and modified CropSyst crop model to assess the impact of extreme weather events on maize yield in South Africa. Submitted to Climate Research.
Renáta Sándor, Catherine Picon-Cochard, Raphael Martin, Frédérique Louault, Katja Klumpp, David Borras, Gianni Bellocchi, Plant acclimation to temperature: developments in the Pasture Simulation model. Submitted to Field Crops Research.
Carlo Gilardelli, Roberto Confalonieri, Gianni Bellocchi, Sensitivity analysis of the effect of climate changes on crop production in Europe: modelling solutions of the crop model WOFOST. In preparation.
Key result no 13: Evaluation of forecast reliability of agro-climatic indicators
Workpackage: WP5
Research aims and background
In the last decades, a variety of forecasting systems were developed. The first methods were based on surveys or crop scouting; these approaches were replaced since the 1990s by more objective and sound techniques, based on the single or integrated use of agro-climatic indicators, remote-sensing information, and crop models. In specific contexts where crop production fluctuations are driven by few main factors, statistical models based on relationships between a few, relevant agro-climatic indicators and crop yields can be able to accurately explain the inter-annual crop yield variability, and thus to reliably forecast crop yields. In recent years, many climatic and agro-climatic indicators were developed and related to the impact of extreme weather events (e.g. drought, extreme temperatures) on agricultural productions. These metrics can be very simple, like those based on counts, i.e. the number of times a phenomenon occurs, sums (e.g. thermal and rainfall sums), or more complex, like those considering plant susceptibility to an extreme event in different moments during the crop cycle.
The aims of this research were:
• to evaluate the potential for forecasting crop yields based on spatially aggregated temperature and drought indicators for the main winter and summer crops and for mown grasslands in all European countries where the specific crop is cultivated;
• to identify and analyse the combinations crop × country where forecasting systems based on agro-climatic indicators present sufficient reliability in case of extreme events.
Results and applications
The indicators ARIDcr (number of days with ARID higher than a fixed threshold, with ARID being 1 – actual to potential transpiration ratio) and Tmaxcr (number of days with Tmax higher than a fixed threshold) were those with the largest forecasting capability, whereas the indicator involved with the impact of low temperatures were rarely selected by the forecasting system. The best performances were achieved for wheat, barley and rye in Spain, with 82%, 85% and 76% of yield variability captured by the forecasting system, and with EF (Nash and Sutcliffe modelling efficiency) ranging from 0.76 to 0.8. For these cases, the trend of yields during the time window was well reproduce by the system, without marked differences between official and forecasted yields. Indeed, Spain is mainly characterized by a Mediterranean climate and winter crop production is mostly driven by rainfall volumes and distribution during the growing season. Satisfactory results were achieved also for sunflower in Bulgaria and sugar beet in Croatia, with R2 values of 0.81 and 0.74 respectively. However, a great part of yield variability was explained by the introduction of technological innovations (a significand technological trend was found), especially in Bulgaria, where the regression model explained only 20% of the year-to-year yield fluctuations.
Significance and benefits
The applied methodology led to satisfactory forecasting performance in countries where crop production is mainly driven by a specific meteorological variable and where this variable often assumes extreme values (e.g. drought, temperatures higher or lower than critical thresholds).
Successful applications
The above represents an important benefit for the community, since – under the conditions depicted in the previous section – simple and robust crop yield forecasting systems can be successfully adopted, without the need of complex approaches based on dynamic simulation models.
References:
MODEXTREME deliverable D5.1. Evaluation of forecasts based on agro-climatic indicators.
Key result no 14: Evaluation of forecast reliability of improved modelling solutions
Workpackage: WP5
Research aims and background
The most sophisticated forecasting systems are based on crop simulation models. An example is represented by the European Commission Crop Growth and Monitoring System (CGMS), used within the MARS project for short-term yield forecasts for the main food crops at European level. The MARS system is based on the WOFOST model for all crops but rice, for which the WARM model is used. The uncertainty in simulations is reduced by post-processing model outputs with historical series of official yields. The need for a statistical layer is due to a portion of variability affecting actual yields which is not reproduced by the monitoring and forecasting system. Among the sources of uncertainty, a key role is played by the absence – in many process-based crop/grassland models – of algorithms for reproducing the effect of biophysical processes actually affecting yields. Concerning the general project goal, existing forecasting systems are based on models developed for conditions of good adaptation and often designed for temperate environments. The effects of unusual weather events on crop productivity, including crop failure in case of extreme conditions, are thus often overlooked or unsatisfactorily simulated by the available crop models. However, raising temperature and a projected increase in the frequency and intensity of extreme weather events such as droughts, heat waves, frost events, are increasingly threatening the usefulness of forecasting services. The aim of this work was to evaluate – within the European Commission MARS workflow – the accuracy of forecasting systems based on standard (CGMS) and improved (including MODEXTREME impact models) modelling solutions for the main crops in Europe.
Results and applications
The standard CGMS system applied to maize showed satisfactory performances in most of the countries, with model reliability increasing after the flowering stage. As an example, in Hungary (the third European maize producer) the system, without the presence of a significant technological trend, explained 46%, 71% and 77% of yields variability while approaching maturity for the three forecasting moments (20th, 24th and 29th ten-day periods, respectively). In general, the best results for the standard solution were achieved in Germany at maturity for potato and grain maize, and in Spain for soft wheat, with 86%, 88% and 86% of explained variability, respectively. The situation was less satisfactory for spring barley, with the exception of Spain, where the comparison of official and forecasted yields led to R2 and EF values of 0.81 and 0.79 respectively. The inclusion of MODEXTREME impact models within the CGMS standard solution led to improve the forecasting reliability in 57 out of 75 combinations crop × country × moment when the forecasting event was triggered. In particular, improvements were achieved in almost all the countries for durum wheat, soft wheat and grain maize, although satisfactory performance were achieved for the latter also with the standard system. The best forecasting performances were achieved for grain maize in Germany, where 92% of inter-annual variability in yields was explained by the MODEXTREME solutions. In most of the combinations crop × country × ten-day period, response functions to extreme events and minimum/maximum canopy temperatures improved the forecasting capability more than state variables. In particular, the response function for cold stress was one of the most relevant indicator.
Significance and benefits
The inclusion of MODEXTREME impact models within the CGMS standard solution led to improve the forecasting reliability in 57 out of 75 combinations crop × country × moment when the forecasting event was triggered. In particular, improvements were achieved in almost all the countries for durum wheat, soft wheat and grain maize, although satisfactory performance were achieved for the latter also with the standard system. The best forecasting performances were achieved for grain maize in Germany, where 92% of inter-annual variability of yields was explained by the MODEXTREME solutions. The improved solution increased up to 44% the amount of variance explained by the forecasting system for spring barley in Poland. In general, the most satisfactory forecasting results were achieved after flowering period.
Successful applications
Achieved results suggest adopting the modelling solutions improved during the MODEXTREME project (those including the impact models for extreme weather events) for forecasting purposes in Europe. A dedicated workshop was organized with the EC-JRC MARS team to define strategies to transfer MODEXTREME results within the MARS operational workflow.
References:
MODEXTREME deliverable D5.2. Report on the evaluation of improved process based modelling solutions and of the standard CGMS ones.
Roberto Confalonieri, 2014, CoSMo: A simple approach for reproducing plant community dynamics using a single instance of generic crop simulators, Ecological Modelling 286, 1-10.
Giovanni Cappelli, Sevim Seda Yamaç, Tommaso Stella, Caterina Francone, Livia Paleari, Marco Negri, Roberto Confalonieri, 2015, Are advantages from the partial replacement of corn with second-generation energy crops undermined by climate change? A case study for giant reed in northern Italy, Biomass and Bioenergy 80, 85-93.
Livia Paleari, Giovanni Cappelli, Simone Bregaglio, Marco Acutis, Marcello Donatelli, Gian Attilio Sacchi, Elisabetta Lupotto, Mirco Boschetti, Giacinto Manfron, Roberto Confalonieri, 2015, District specific, in silico evaluation of rice ideotypes improved for resistance/tolerance traits to biotic and abiotic stressors under climate change scenarios, Climatic Change 132, 661-675.
Simone Bregaglio, Francesca Orlando, Emanuela Forni, Tommaso De Gregorio, Simone Falzoi, Chiara Boni, Michele Pisetta, Roberto Confalonieri, 2016, Development and evaluation of new modelling solutions to simulate hazelnut (Corylus avellana L.) growth and development, Ecological Modelling 329, 86-89.
Roberto Confalonieri, Simone Bregaglio, Myriam Adam, Françoise Ruget, T. Li, T. Hasegawa, X. Yin, Y. Zhu, K. Boote, S. Buis, T. Fumoto, D. Gaydon, T. Lafarge, M. Marcaida, H. Nakagawa, Alex C. Ruane, B. Singh, U. Singh, L. Tang, F. Tao, J. Fugice, H. Yoshida, Z. Zhang, L.T. Wilson, J. Baker, Y. Yang, Y. Masutomi, Daniel Wallach, Marco Acutis, B. Bouman, 2016, A taxonomy-based approach to shed light on the babel of mathematical models for rice simulations, Environmental Modelling & Software 85, 332-341.
Carlo Gilardelli, Tommaso Stella, Nicolò Frasso, Giovanni Cappelli, Simone Bregaglio, Marcello E. Chiodini, Barbara Scaglia, Roberto Confalonieri, 2016. WOFOST-GTC: a new model for the simulation of winter rapeseed production and oil quality, Field Crops Research 197, 125-132.
Livia Paleari, Roberto Confalonieri, 2016. Sensitivity analysis of a sensitivity analysis: we are likely overlooking the impact of distributional assumptions, Ecological Modelling 340, 57-63.
Key result no 15: Identification of the most reliable workflows for yield forecasts
Workpackage: WP5
Research aims and background
A variety of systems for crop yield forecasts was developed in the last decades to face the increasing demand of timely and reliable information to be used for properly managing agricultural markets and policies. The most widespread systems are based on agro-climatic indicators, remote-sensing and crop models, with those techniques used alone or in an integrated fashion. In general, simple relationships between specific agro-climatic indicators and crop yields can be sufficiently accurate when applied to context where crop yield fluctuations are driven by few key meteorological drivers. However, the most sophisticated methods are based on crop simulation models (e.g. the European Commission Crop Growth and Monitoring System). Most of available systems were developed targeting temperate environments. However, the projected increase in the frequency and intensity of extreme weather events (e.g. droughts, heat waves and frost events) is threatening the usefulness of current forecasting services.
The aims of this work were:
• to analyse and compare – at European scale – the reliability of forecasting systems based on (i) agro-meteorological indicators; (ii) outputs of the CGMS standard modelling solution; (iii) outputs of the CGMS improved (including MODEXTREME impact models) modelling solution; (iv) the combination of agro-climatic indicators and outputs of both the standard and improved modelling solutions (hybrid approach);
• identify the most reliable workflow for each combination crop × country × forecasting event.
Results and applications
In general, the improved modelling solutions (including MODEXTREME impact models) showed the most satisfactory forecasting performances, especially for durum wheat, soft wheat and maize. The highest reliability was obtained for maize in Germany at the 24th 10-day period, with 92% of explained yield variability. Model outputs accounting for extreme weather events allowed increasing the forecasting capability in most cases. The output that often explained a large part of inter-annual variability was the harvest index response function to cold. However, in the northern states, simulated damages to leaf area due to frost played a key role. The system based on agro-meteorological indicators showed the best performances only for potato in Poland and in Germany at the 16th 10-day period. The integrated use of the different forecasting systems (hybrid) allowed to improve CGMS performances in 87% of the combinations crop × country × forecasting moment. However, most of the improvement in the forecasting reliability was due to the MODEXTREME impact models (improved dynamic solutions). The addition of agro-climatic indicators to those provided by the improved solutions led to a slight improvement of forecasting performances in 24% of the cases.
Significance and benefits
This part of the study allowed identifying the most reliable workflow for each combination crop × country × forecasting moment, providing a crucial support for operational forecasting services in Europe. Pros and cons of different approaches (agro-climatic indicators, standard and improved solutions, hybrid systems) were highlighted for different agro-climatic contexts.
Successful applications
The hybrid forecasting system developed and tested within the project can be a powerful tool within operational forecasting systems. Like for the system based on improved modelling solutions, a dedicated workshop was organized with the EC-JRC MARS team to define strategies to transfer MODEXTREME results within the MARS operational workflow.
References:
MODEXTREME deliverable D5.3. Report on the identification of the most reliable workflows for yield forecasts.
Roberto Confalonieri, Simone Bregaglio, Marco Acutis, 2016, Quantifying uncertainty in crop model predictions due to the uncertainty in the observations used for calibration, Ecological Modelling 328, 72-77.
Roberto Confalonieri, Francesca Orlando, Livia Paleari, Tommaso Stella, Carlo Gilardelli, Ermes Movedi, Valentina Pagani, Giovanni Cappelli, Andrea Vertemara, Luigi Alberti, Paolo Alberti, Samuel Atanassiu, Matteo Bonaiti, Giovanni Cappelletti, Matteo Ceruti, Andrea Confalonieri, Gabriele Corgatelli, Paolo Corti, Michela dell’Oro, Alessandro Ghidoni, Angelo Lamarta, Alberto Maghini, Martino Mambretti, Agnese Manchia, Gianluca Massoni, Pierangelo Mutti, Stefano Pariani, 2016, Uncertainty in crop model predictions: What is the role of users?, Environmental Modelling & Software 81, 165-173.
Valentina Pagani, Tommaso Guarneri, Davide Fumagalli, Ermes Movedi, Luca Testi, Tommy Klein, Pierluigi Calanca, Francisco Villalobos, Stefan Niemeyer, Gianni Bellocchi, Roberto Confalonieri, Accounting for the effect of weather extremes on cereal yield forecasting in Europe. In preparation.
Potential Impact:
POTENTIAL IMPACTS
The European and international added value of the project lies in increasing the capability of yield monitoring and forecasting system to support agricultural and food security policies in the frame of the European Commission and beyond. For that, MODEXTREME has:
- formalised and integrated into statistical and process-based simulation models the responses of main crop/grassland/tree systems to extreme weather events;
- assessed the above approaches against experimental/observational datasets;
- used the above approaches to generate trajectories of agricultural productivity in short-medium time horizons;
- developed generically reusable software units implementing libraries of models;
- extended the modelling capabilities of the BioMA platform towards weather extremes;
- supported the monitoring and forecasting system of the European Commission JRC-MARS;
- transferred the knowledge gained and the modelling solutions developed to JRC-MARS analysts and other beneficiaries, for dedicated implementation with local stakeholders to improve yield forecasts and increase food security at the EU level and worldwide.
By providing alternate modelling solutions to simulate a variety of agricultural systems worldwide, yet under scenarios of climate change, MODEXTREME provided evidence of the operational usability of modelling approaches in forecasting systems, by:
- showing the potential of novel modelling tools to account for the impact of weather extremes on agricultural systems;
- reducing uncertainties in the estimates of agricultural production under weather extremes;
- strengthening the role of the EU forecasting system in agriculture.
SCIENTIFIC, TECHNOLOGICAL AND POLICY OUTCOMES
MODEXTREME has developed novel approaches of salient plant processes to represent mathematically the impact of extreme weather events (heatwaves, aridity, cold shocks and frost) on leaf development and grain accumulation as well as other physiological pathways (e.g. thermal acclimation of grassland species). They provided scientific ground for the implementation of dedicated software code for use into system models, which increased the capability of existing simulation tools to reproduce experimentally established system performances. The simulation models used in MODEXTREME have benefited of the improvements introduced, which can be extended to virtually any other simulation model of plant growth and development. Agro-climatic indicators have also been developed and used to identify extreme weather events along plant growing seasons and assess their impact on agricultural production. To facilitate the calculation of indicators of climatic extremes having impact on agricultural production, the web application webXTREME was developed and made freely accessible online as a way to promote collaboration within the scientific community and between scientists and stakeholders.
Novel software components for the impact of extreme events, new modelling solutions for crop/grassland/tree systems and several supporting tools have been developed within MODEXTREME, mostly for and through the platform BioMA (Biophysical Model Applications). BioMA provides an advanced technological support (open and flexible to develop modelling solutions upon reusable software components) to the European Commission for scenario analyses in agriculture and climate change in response to the need of the work program of JRC MARS - Agri4Cast. The latter includes impact assessment and the development of adaptation strategies for agriculture under climate change scenarios, as a key element in the EU agriculture monitoring capabilities. The framework BioMA is currently being used as production tool in the EU forecast run by the JRC of the European Commission and it holds potential for further developments and extension to new functionalities in the future (for instance to deal with challenges to food security associated with the exacerbated impact of pests and diseases).
Newly developed datasets of climate scenarios in project regions opens up to the possibility to view projected weather data in unprecedented detail, that is downscaled, bias-corrected, post-processed and put in a consistent framework for direct use for simulation modelling in agriculture. The dataset has been shared with the JRC and several scientists worldwide which see it as the missing piece in a wide range of studies they planned to do. Modelling teams outside of the project requested access to these datasets. The quantified scenario drivers based on RCP8.5 benefited strongly from the most recent IPCC scenario development process (through the CORDEX initiative). Used as inputs to force models simulating agricultural production, they provide new insights about the substantial potential for food security assessment. Adding this new layer of detail to the discussion about the future of the agricultural sector in the context of climate change may lead to new conclusions about the impact of extreme weather events.
The experimental/observational results, the modelling approaches and the simulation outputs that have been produced in MODEXTREME offer benefits to a variety of end-users. These include both sector specialists and farmers, as well as local, regional, national and EU policy makers working directly or indirectly with agriculture and also scientists involved in food security assessment. The agro-environmental modelling community would also benefit from both the developments in modelling approaches and improved parameterization for simulations of crops, grasslands and tree crops (olive and grape) at a variety of contrasting sites. The uncertainty quantification of model estimates of agricultural production and other outputs from agricultural systems holds potential to direct further work across a larger variety of conditions than those examined in MODEXTREME. For instance, modelling approaches could be expanded to assess the impact of extreme weather events on subsistence farmers, which account a large portion of the rural society.
MODEXTREME provided detailed results showing the best workflow in the estimate of agriculture production in Europe for combinations of crop-country-forecasting events, based on alternate modelling solutions and agro-climatic indicators. They show the increased forecasting reliability, obtained with novel modelling solutions and indicators developed in MODEXTREME, in a spatially-explicit way by indicating for each European country the best forecasting workflow for a broad set of crops. The information provided by these results is readily exploitable, and the tools and approaches developed are of direct use by the European Commission as they represent improved solutions of the Agri4Cast MARS Crop Yield forecasting system. This important outcome has been achieved thanks to close collaboration with JRC MARS - Agri4Cast.
COMMUNICATION AND OUTREACH
A project brochure outlining the aims of the project and its structure was developed and handed out at presentations, exhibitions and conferences, and a project website (which will be maintained beyond the funding period) was deployed with open access to all project outcomes. Particular emphasis has been placed on dissemination events to enhance sharing of project results across broad communities of scientists and non-scientific actors. In particular, four dissemination workshops were organized as part of the dissemination plan:
1. “Modelling the impact of extreme climatic events in agriculture” in the frame of 5th International Symposium for Farming Systems Design (Montpellier, France, September 10th, 2015);
2. “Modelling the impacts of extreme climatic events on agrifood systems” at FAO headquarter (Rome, Italy, November 3rd, 2015);
3. “Agricultural Modelling Hackathon” in the frame of the 8th Biennial Congress of the International Environmental Modelling and Software Society (Toulouse, France, July 12th, 2016);
4. “Improving agricultural modelling for the impact of extreme weather events” at European Commission DG AGRI headquarter (Brussels, Belgium, September 20th, 2016).
Young scientists were supported by the project to illustrate results in these workshops. To give wider access several e-learning materials in a variety of formats (presentations, training documents, science briefs, software tools) were made available through the public website, and two public newsletters were distributed electronically, which ensured effective dissemination among broad communities of scientists and policy makers worldwide.
A Stakeholder Platform was established to create a participatory framework linking key end-users to the project team. It comprised representatives of the European Commission (JRC, DG AGRI and other DGs) and local actors and experimental scientists identified by partners (and put under FAO coordination) such chambers of agriculture, agricultural authorities, governmental bodies, etc.
EC delegates attended all the project meetings (kick-off and three annual meetings) and subsequently produced reports of their recommendations in relation to their role, their expectations of the project outcomes and on the process of consultation between the project team and the stakeholder platform. Dissemination and training plans were produced based on the description of work, and updated after each annual meeting to incorporate the above recommendations and those from the Scientific Advisory Board (three distinguished scientists) about key actions to implement the project plan. Interactions with the main stakeholder (JRC MARS Unit) have become more intense in the final period (e.g. a meeting was organized in Ispra in October 2015 followed by several exchanges and contacts), resulting in a joint action plan that responded to the needs and opportunities identified in the project (such as exploitation of data of the MARS Crop Yield Forecasting System and creation of modelling solutions compatible with that same system). The project also held two sessions at JRC (Ispra, Italy) as part of the training activities addressed to MARS agents.
Dedicated workshops addressed to local stakeholders were organised in Argentina, Brazil, China, France, Italy, South Africa, Spain, Switzerland and Ukraine. Views from local workshops supported the compilation by FAO of questionnaire-based feedback on the modelling support to food security issues and led to a definition of key outputs from the project that are of relevance to stakeholders, their likely impact, and strategies for post-project dissemination and implementation of the outputs.
The dissemination activity of the project was completed by illustrating relevant research outcomes in international events such as the 5th AgMIP global workshop (Gainesville FL, USA, February 25th-28th, 2015), the 3rd Global Science Conference on Climate - Smart Agriculture (Montpellier, France, March 16th-18th 2015) and the 8th Biennial Congress of the International Environmental Modelling and Software Society (Toulouse, France, July 10th-14th, 2016).
An internal collaborative platform has been designed so as to ensure a regular exchange among the project partners, the stakeholder platform and the scientific advisory board (also assisted by periodic circulation of internal newsletters). Supported by the management team, the internal platform will be maintained beyond the funding period to ensure access to consolidated training material, datasets and scientific outcomes.
List of Websites:
Project public website
The website address is the following: http://modextreme.org
To support dissemination and to host web-based training courses, the open public website informs on the aims and structure of the project, project outputs (including access to software), presentations of project workshops and training events, and other news of the project. Public deliverables are also published on the website. Two public newsletters, also available on the website, have been produced to outline the main outcomes of the project and likely impact.
Relevant contact details
MODEXTREME Coordinator: Gianni Bellocchi
INRA
5 chemin de Beaulieu, 63039 Clermont-Ferrand (France)
Tel.: +33 (0) 4 43 76 16 01
Email: gianni.bellocchi@inra.fr
MODEXTREME Project Manager: Léa Tourneur
INRA Transfert
3 rue de Pondichéry, 75015 Paris (France)
Tel.: + 33 (0) 1 76 21 62 06
Email: lea.tourneur@inra.fr
MODEXTREME has as its strategic goal to help the European (and non-European) agriculture to face extreme weather events by improving the capability of biophysical models simulating plant responses to integrate weather extremes. The two key targets are the production of novel (reusable) libraries of models and the improvement of yield monitoring and forecasting systems. For this purpose, the project has improved the capability of biophysical models simulating crop/grassland/tree responses to represent the impact of extreme weather events (such as heatwaves, cold shocks, drought and frost) via development of generically usable software units implementing libraries of models. In so doing, it has extended the multi-model platform BioMA (Biophysical Model Applications) for the simulation of plant growth and development, thus providing direct support to the European Commission Joint Research Centre (MARS: Monitoring Agricultural ResourceS – Agri4Cast action), while also reaching out to local stakeholders by organizing dissemination workshops and training sessions addressed to scientists, technicians and decision makers.
Expressed mathematically, fundamental concepts were implemented into dedicated modules and coupled to existing modelling solutions (EMS) to illustrate that the crop/grassland/tree models have the elements needed to reproduce experimentally established system performances under extreme conditions. In particular, dedicated components modelling the impact of extreme events were developed and coupled to simulation models of arable crops (i.e. generic models such as WOFOST, CropSyst, AquaCrop and Syrius-Quality, and the rice model WARM) but also to the grassland model PaSim, the olive model OLIVECAN and the vineyard model GRAPE, but they are potentially applicable for use with other models as well. Existing and modified (for the impact of extremes) solutions of these models were assessed in a comparative fashion (EMS versus MMS) against field data from a variety of agricultural sites worldwide. This investigation was based on common protocols for model calibration and sensitivity analysis, which allowed identifying the most influential crop parameters in different pedo-climatic (including with scenarios of climate change) and management contexts. Downscaled (bias-corrected) projected storylines of climate change were generated for the study-areas from a set of GCM-RCM simulations to assess the sensitivity of both EMS and MMS to alternative weather scenarios including more frequent and severe extreme events (based on radiative forcing for the high RCP8.5 which is the most extreme of the emissions pathway scenarios developed for the IPCC’s Fifth Assessment Report).
Results with simulation models highlighted limitations in the response to extreme weather events of currently used modelling solutions. The improvement introduced by MMS was the basis for a discussion with local stakeholders and the FAO around the employment of these modelling tools to support food security issues in Europe and worldwide. For the generic crop model WOFOST and the rice-specific model WARM, both EMS and MMS were employed within the Agri4Cast forecasting system of the European Commission (based on modelling tools and agro-climatic indicators) to quantify their reliability to estimate yield for the main cereal/non-cereal crops cultivated in Europe. The reliability of forecasting results was improved in most cases, sometimes greatly increasing the percentage of variance explained (e.g. up to 44% for spring barley in Poland). The addition of reasoned agro-climatic indicators (also developed within the project) to model outputs further enhanced forecast reliability (e.g. for rice in Spain).
To date, the project has led to the publication of 27 peer-reviewed articles and 29 other papers acknowledging MODEXTREME support, the organization of more than 20 dissemination and training events, and the production of 31 software units and packages corresponding to entirely new or refactored applications, model components and modelling solutions.
Project Context and Objectives:
CONTEXT
Model-based estimates of agricultural production are key elements for food security in the context of climate change. More important than mean temperature rise, projected climate change calculations indicate that frequency and severity of atmospheric extremes - which include floods, droughts, severe heat and cold, and storms - are altered by changing climate means. Changes in precipitation distribution, along with increased frequency and intensity of extreme weather events likely determine an augmented variability of agricultural productions and price volatility as well as changes in commercial exchanges. Also, the impacts of extreme events on crop/grassland/tree productivity are of increasing concern for food security policies. Changes in climate drivers could alter the frequency and magnitude of hazards, and changes in socio-economic drivers can lower the capacity of agricultural systems to cope with hazards. Extreme events are often combined and different plant stress factors (e.g. high temperature, evaporative demand and light) may simultaneously occur with different correlations among geographical areas. This generates an apparent complexity in the climate forcing-plant response relationships across a wide range of temporal and spatial scales and a poor description by simulation models. For instance, many biological processes may undergo sudden shifts at particular thresholds for temperature or precipitation. However, the mechanistic links between plant processes, and the relative roles in these processes of extreme events, have been only fragmentarily documented. Therefore, the extent of the inherent response to extreme weather events that plant species and varieties may be able to undergo remains an open field of research.
VISION OF THE PROJECT
A main concept of the project is that novel but still relatively simple responses to each climatic variable may generate complex responses. Most models have been established and calibrated on genotypes released in the years 1990s and on climate patterns to which plants were well adapted. Consequently, such modelling solutions had to be evaluated and possibly new capabilities could be required to predict plant responses under altered climate patterns. Translating basic knowledge into system models (and evaluating them) requires an interdisciplinary approach combining plant eco-physiology, software engineering and climatology. The MODEXTREME consortium has been built along this interdisciplinary approach and the project aimed to extend and strengthen European research in cooperation with non-EU partners and international organizations. MODEXTREME has brought together on-going efforts made by European and international teams on the modelling of extreme events impacts (including groups involved with knowledge of the development and use of MARS infrastructure, and involvement in international modelling initiatives such as AgMIP and MACSUR).
The project has three main objectives:
1. SCIENTIFIC CONTENT: To formalise and integrate into statistical and process-based simulation models the responses of main crop/grassland/tree systems to extreme weather events, assess them against existing datasets and use them to generate possible trajectories of agricultural productivity in short-medium time horizons.
2. INFRASTRUCTURE: To develop generically reusable software units implementing libraries of models that extend the modelling capabilities of the BioMA (Biophysical Model Applications) platform towards weather extremes, within a framework to support the monitoring and forecasting system of the European Commission Joint Research Centre (MARS: Monitoring Agricultural ResourceS).
3. SOCIO-INSTITUTIONAL: To transfer the knowledge gained and the modelling solutions developed to MARS analysts and model users, for dedicated implementation with local stakeholders to improve yield forecasts and increase food security at the EU level and worldwide.
Farmers, consultants, experimental scientists and decision-makers in Europe, Latin America, People’s Republic of China, and Sub-Saharan Africa are key beneficiaries of the project. MODEXTREME has, for instance:
- Tested modelling solutions for the benefit of model users and agricultural advisors
- Delivered guidelines to build, select, post-process and use climate scenarios into crop/grassland/tree simulation modelling for the benefit of model users and agricultural advisors, and to support food security policies
- Extended modelling tools and platforms beyond the impact of weather extremes for the benefit of model users, policy makers and agricultural advisors
Several dissemination and training courses have been organized to improve the understanding of simulation tools, dedicated libraries and model development environments for the benefit of scientists, technicians, advisors and policy makers, building model capacity regarding both use and development of models.
Project Results:
Key result no 1: Development of indices apt to gauge impacts of critical temperatures and water shortage on crop performance (Task 1.1)
Workpackage: WP1
Research aims and background
The existent modelling solutions for the simulation of arable crops, grasslands and trees are developed and tested using the average patterns of weather drivers and are considered satisfactory even if their predicting capacity fails when extreme weather events appear. This situation hinders the correct yield prediction under the specific situations of some extreme meteorological years (considered “outliers”), although the overall performance of the model is actually good. For some model applications, like the prediction of yield for a given year, the overestimation of yield after an extreme weather event is a serious issue for institutions and agencies devoted to agricultural production forecasting.
Results and applications
Weather extremes (namely low and high temperature extremes and drought) were identified with thresholds as functions of their effects on crop yield. A set of eight core indices was defined and tested on the basis of daily data for 1975-2013 as obtained from JRC for France and Spain. Statistical analysis of the indices allowed to identify critical years both at the local as well as at the regional scale. A web-based application to analyse existing datasets and identify potentially yield-reducing events was developed (WEBEXTREME: http://www.modextreme.org/webextreme).
Significance and benefits
The procedures for the identification of extreme events as function of their effect on crop production are new and allow a univocal identification of the years that bring potential production effects for several crops out of long time series. The procedure was automated and tested; the potential benefits are open to all the Europeans countries.
Successful applications
The identification of extreme events was implemented by means of a web-based application open to the users worldwide.
References:
MODEXTREME Deliverable D1.1. Identification of extreme events, statistical adjustments and modelling of extreme yield anomalies.
Tommy Klein, Pierluigi Calanca, Tamara Ben Ari, David Makowski, Gianni Bellocchi, 2015, Spatial extension of heat and drought stress in European wheat producing areas, Poster at 5th AgMIP Global Workshop, February 25-28, Gainesville, FL, USA. http://modextreme.org/wp-content/uploads/sites/4/2015/02/Poster_KleinEtAl_v3.pdf
Tommy Klein, Argyrios Samourkasidis, Ioannis N. Athanasiadis, Gianni Bellocchi, Pierluigi Calanca, 2016, webXTREME: a simple web tool for calculating agroclimatic indicators of extreme events In: Sauvage, S., Sáchez-Pérez, J.-M. Rizzoli, A.E. (Eds.), Proceedings of the 8th International Congress on Environmental Modelling and Software, July 10-14, Toulouse, France, Vol. 2, 471. ISBN: 978-88-9035-745-9.
Tommy Klein, Argyrios Samourkasidis, Ioannis N. Athanasiadis, Gianni Bellocchi, Pierluigi Calanca, webXTREME: R-based web tool for calculating agroclimatic indices of extreme events. Submitted to Computers and Electronics in Agriculture.
Key result no 2: Development of a general modelling framework to account for the yield reduction in many crops by means of effects on harvest index (Task 1.2)
Workpackage: WP1
Research aims and background
The existent crop modelling solutions are developed and tested using the average patterns of weather drivers and are considered satisfactory even if their predicting capacity fails when extreme weather events appear. This situation hinders the correct yield prediction under the specific situations of some extreme meteorological years (considered “outliers”), although the overall performance of the model is actually good. For some model applications, like the prediction of yield for a given year, the overestimation of yield after an extreme weather event is a serious issue for institutions and agencies devoted to agricultural production forecasting.
Results and applications
The modelling framework developed in MODEXTREME assumes that yield variations due to an extreme event (cold temperature, high temperature or water deficit) are due to flower death and failure in pollination (high or low temperatures) while water stress reduces seed set. All these effects are mediated by a change in Harvest Index (HI), while the main effect of weather on crop performance is already captured by existing crop models. In crops whose yield is a vegetative organ (e.g. potato) the response is calculated over the Leaf Area Index. The response functions of HI depression vs. environmental critical conditions were developed and formalized. The framework allows also to account for the combinations of multiple events during and after the flowering, and of the joint effect of high temperature and water stress. Although tables with the critical thresholds for the main crops have been provided, the modelling framework is theoretically applicable to any crops even beyond those mentioned in the DOW after specific calibration.
Significance and benefits
The modelling framework demonstrated to be easily applicable to many crops and to increase the yield prediction capability of already existing models on extreme years. The general HI approach has been found to be simple enough to be applicable to many crops after calibration of the response functions parameters, while being robust and performant. The HI-based modelling framework for the extreme weather impact on yields may also be applied as an independent module over the output of the crop models, allowing to deal with the extreme-event effect without need of modifying or re-coding the existing modelling solutions.
Successful applications
The approaches developed to model the impact of extreme events were implemented in a dedicated software component for use within crop modelling frameworks.
References:
MODEXTREME Deliverable D1.2. Report on modelling approaches for simulating the impact of extreme events on agricultural production.
Gianni Bellocchi, Francisco J. Villalobos, Marcello Donatelli, Ole B. Christensen, Oscar Rojas, Roberto Confalonieri, Ioannis N. Athanasiadis, Irina Carpusca, Claudio O. Stöckle, 2014, Extending existing models to capture vegetation response to extreme weather events: the MODEXTREME project In: Ames, D.P. Quinn, N.W.T. Rizzoli, A.E. (Eds.), Proceedings of the 7th International Congress on Environmental Modelling and Software, June 15-19, San Diego, California, USA. ISBN: 978-88-9035-744-2. http://www.iemss.org/sites/iemss2014/papers/iemss2014_submission_215.pdf
Francisco J. Villalobos, José Paulo de Melo e Abreu, Luca Testi, Gianni Bellocchi, 2015, Impact of extreme meteorological events on crop yield: a common framework approach, Poster at 5th AgMIP Global Workshop, February 25-28, Gainesville, FL, USA. http://modextreme.org/wp-content/uploads/sites/4/2015/02/Poster-Agmip_T1.2.pdf
Key result no 3: Micrometeorological calculation methods for the estimation of canopy temperature from air temperature and disaggregation of daily maximum and minimum temperature into hourly temperatures
Workpackage: WP1
Research aims and background
Both the available historical datasets and the future climate projections of meteorological data have a time resolution of day (maximum and minimum air temperatures). While this resolution is usually adequate for the majority of crop models, at the time of calculating the effect of extreme temperatures over the plant is not the air but the canopy temperature what matters, and the exact amount of time during which the canopy remains in a dangerous temperature state. There was no practical and generalized way to account for these key issues in the existing modelling solutions.
Results and applications
Advanced calculation methods based on crop micrometeorology were presented, formalized and tested with previous high-resolution canopy temperature datasets. These methods allow the disaggregation of daily maximum and minimum air temperatures into a diurnal time series of air temperature at an hourly or shorter time step, using empirical polynomial or sine disaggregation during the day and a modified application of Brunt law during the night. From the air temperature pattern, the canopy temperature is calculated by means of advanced applications of methods based on crop energy balance and aerodynamic features.
Significance and benefits
These calculation methods are new and original. They allow to calculate the canopy temperature in any time of the day or the night of virtually any crop, whenever its precise definition is required, from standard meteorological data.
Successful applications
These methods were integrated in the module for the calculation of yield response to extreme events.
References:
MODEXTREME Deliverable D1.2. Report on modelling approaches for simulating the impact of extreme events on agricultural production.
Jose Paulo Melo-Abreu, Francisco J. Villalobos, Luciano Mateos, 2016. Frost protection. In: Villalobos F.J. Fereres E. (Eds.), Principles of Agronomy for Sustainable Agriculture. Springer. ISBN 978-3-319-46115-1.
Francisco J. Villalobos, Luca Testi, Luciano Mateos, Elías Fereres, 2016, Manipulating the crop environment. In: Villalobos, F.J. Fereres, E. (Eds.), Principles of Agronomy for Sustainable Agriculture. Springer. ISBN 978-3-319-46115-1.
Villalobos FJ, Lopez-Bernal A. In press. El clima. In: Barranco D, Fernández-Escobar R, Rallo L (Eds.). El cultivo del olivo. Mundi-Prensa, 213-249. (in Spanish)
Francisco J. Villalobos, José Paulo de Melo e Abreu, Luca Testi, Gianni Bellocchi, 2015, Impact of extreme meteorological events on crop yield: a common framework approach, Poster at 5th AgMIP Global Workshop, February 25-28, Gainesville, FL, USA. http://modextreme.org/wp-content/uploads/sites/4/2015/02/Poster-Agmip_T1.2.pdf
Key result no 4: Development of statistical models for ex-post (static) adjustment of estimated yields, suitable for application at the “statistical level” of the MARS monitoring system (Task 1.4)
Workpackage: WP1
Research aims and background
The existent crop modelling solutions are developed and tested using the average patterns of weather drivers and are considered satisfactory even if their predicting capacity fails when extreme weather events appear. This situation hinders the correct yield prediction under the specific situations of some extreme meteorological years (considered “outliers”), although the overall performance of the model is actually good. For some model applications, like the prediction of yield for a given year, the overestimation of yield after an extreme weather event is a serious issue for institutions and agencies devoted to agricultural production forecasting.
Results and applications
Yield statistics over the past period were used to derive statistical indicators of climate impact on plant production. To this end, climate variables aggregated at the national scale and/or climate indicators developed in task T1.1 were used.
For statistical adjustment, models relying on single climate variables perform equally well as models relying on complex indices. Monthly indices were found useful to target key time periods. Variables/indices with the highest predictive capacity were identified. Prediction of yield extremes is improved by combining several climate indices using logistic regression.
Significance and benefits
The statistical indices developed and tested are now available for cereals yield forecasting in Europe.
Successful applications
Scripts for the calculation of statistical indices are available for integration to the statistical layer of the MARS crop monitoring and forecasting system.
References:
MODEXTREME Deliverable D1.1. Identification of extreme events, statistical adjustments and modelling of extreme yield anomalies
Tamara Ben-Ari, Juliette Adrian, Tommy Klein, Pierluigi Calanca, Marjin Van der Velde, Stefan Niemeyer, Gianni Bellocchi, David Makowski, 2015, Identifying accurate climate indicators of extreme yield loss in Europe, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/Poster_CSA_8.pdf
Tamara Ben-Ari, Juliette Adrian, Tommy Klein, Pierluigi Calanca, Marijn Van der Velde, David Makowski, 2016, Identifying indicators for extreme wheat and maize yield losses, Agricultural and Forest Meteorology 220, 130-140.
Key result no 5: Development of OLIVECAN, a model to simulate the water and carbon exchange of olive groves. (Task 1.3)
Workpackage: WP1
Research aims and background
Models of tree crops are rare and those existent often are focused to cover only some specific processes of interest. The development of a complete model for the olive tree, the main tree crop in Europe, with the capacity of predicting net the carbon assimilation, water use and yield formation is extremely useful for the olive crop cultivation and oil production industry.
Results and applications
OLIVECAN is a model aimed at the simulation of water and carbon exchange between olive orchards and the atmosphere. It is composed of sub-models that deal with the more important processes like interception of PAR radiation, soil water uptake, potential and actual transpiration, photosynthesis and assimilation of carbon, respiration rates of the different organs and yield formations. The model was conceived to work under localised irrigation conditions (i.e. calculates the water balance over two different soil compartments with different profiles of water content). The model includes the harvest index responses functions to extreme temperatures (heat and frost) and drought developed in task 1.2.
Significance and benefits
OLIVECAN has a strong predictive capacity for water use, biomass accumulation and yield. It can be used at different level of complexity to simulate the physiological response of olive to different cropping management (i.e. irrigation, pruning, planting strategies) and is able to respond correctly to the CO2 increase of future conditions in terms of transpiration, photosynthesis and respiration processes, including the effects of extreme meteorological events on yields.
Successful applications
A novel version of OLIVECAN was successfully tested in Spanish conditions.
References:
MODEXTREME Deliverable D1.2. Report on modelling approaches for simulating the impact of extreme events on agricultural production.
Omar García-Tejera, Alvaro Lopez-Bernal, Luca Testi, Francisco J. Villalobos, 2017, A soil-plant-atmosphere continuum (SPAC) model for simulating tree transpiration with a soil multi-compartment solution, Plant and Soil, doi:10.1007/s11104-016-3049-0
Omar Garcia-Tejera, Álvaro Lopez-Bernal, Francisco J. Villalobos, Francisco Orgaz, Luca Testi, 2016, Effect of soil temperature on root resistance: implications for different trees under Mediterranean conditions. Tree Physiology 36, 469-478.
Alejandro Morales, Peter A Leffelaar, Luca Testi, Francisco Orgaz, Francisco J. Villalobos, 2016, A dynamic model of potential growth of olive (Olea europaea L.) orchards. European Journal of Agronomy 74, 93-102.
Key result no 6: Improvement of BioMA framework
Workpackage: WP2
Research aims and background
BioMA is a modelling framework designed to develop reusable software components for composing modelling solutions for the integrated analysis of agro-ecological systems (http://www.biomamodelling.org). It consists of a set of core components, a suite of coding assistant tools plus a set of Window applications for configuring simulations, running the models, optimizing models against reference data, exploring the outputs and statistically analyse the model estimates. The technology provided by BioMA appears particularly suitable to the objectives of MODEXTREME. The project activities require handy tools to run multiple models, which must be easily extendible to increase their explanatory power in order to analyse the crop response to multiple biotic and abiotic stresses.
The BioMA framework can satisfy all these needs; first, by making available flexible Modelling Solutions (MS) which can be easily extended by integrating additional components to include more functionalities. By adding the MODEXTREME.WeatherExtremesImpact component (see D2.1 Movedi et al., 2016) to MS based on existing crop models (e.g. CropSyst), new MS can be obtained with the additional capacity of simulating crop response to extreme weather events. Both conventional and the extended version can be compared. A set of ancillary applications, embedded in the BioMA-Site and BioMA-Spatial as plugins, assisted MODEXTREME users across the whole chain of model testing activity, from model optimization to analysis of output estimates. BioMA was completely refactored and optimized around MODEXTREME needs, but it is a mature technology which can be fruitfully applied to other research areas involving agroecosystem analysis. For instance, current ongoing applications are the impact analysis of climate change, and the optimization of agricultural management.
Results and applications
This software has enabled MODEXTREME partners to perform all the simulation workflow steps: simulation configuration, model parameterisation, model optimization and statistical analysis of output, data view and exploration.
The products which were made available are the following ones:
i. BioMA-Site and BioMA-Spatial: MS Windows applications to deploy and run modelling solutions. They are also the interfaces which expose all the other functionalities as plugin tools.
ii. Code generator tools: they facilitate model components development, limiting the coding effort requested to modellers to the algorithmic representation of models.
iii. Plugin tools for simulation configurations, parameter editing, component structure exploring and data view and analysis: MPE Model Parameter Editor, MSE Modelling Solution Explorer, ACG AgroManagement Configurator Generator, MCE Model Component Explorer, GDD Graphical Data Display.
iv. Applications for optimizing models and analysing results: LUISA (sensitivity analysis), Optimizer (model optimization against reference data), IMMA (statistical analysis of estimates).
All these tools can be downloaded from the BioMA portal (http://www.biomamodelling.org) along with modelling solutions. Documentation, tutorials, and sample simulations are provided. D2.3 (Donatelli et al., 2016) and Donatelli et al. (in preparation) give an overview of the available tools.
Significance and benefits
The BioMA platform represents a significant advancement in the field of agroecosystem modelling targeting flexibility in modelling solution development and re-usability of discrete model units. A large set of resources are made publicly available, which allow a vast set of applications. The BioMA architecture was designed to allow the extensibility of components, so resources can be easily extended and adapted virtually to most user needs.
Successful applications
WP4 partners extensively tested released modelling solutions, and BioMA-Site and Spatial applications. WP4 produced a large amount of simulation files covering the whole South-East Europe, Ukraine and portions of China, Argentina, Brazil, and South Africa. WP4 partners made available country-specific soil parameter sets, crop parameters files as resulting from their calibration, peculiar agro-management files, and tested BioMA-Site/Spatial configurations (available at: http://modextreme.org).
References:
MODEXTREME deliverable D2.1. Abiotic stress model component: v1.
MODEXTREME deliverable D2.3. The BioMA development.
Marcello Donatelli, Davide Fumagalli, Iacopo Cerrani, Davide Fanchini, Simone Bregaglio, Ioannis Athanasiadis, Andrea E. Rizzoli, Enhancing biophysical models reuse: shifting the focus from frameworks to components in the BioMA framework. In preparation for submission to Agricultural Systems.
Key result no 7: Development of Modelling Solutions
Workpackage: WP2
Research aims and background
The interest on simulation models for the integrated analysis of agroecosystems has increased over the last decades. Furthermore, new functionalities are continuously required and models are expected to be always to date with scientific advancements, to support emergent scientific and practical applications.
In principle, models with such requirements could be built by assembling the most advanced modelling resources available, but the lack of standardization and the heterogeneity of languages, approaches and programming paradigms, has prevented so far this possibility.
With the advent of innovative programming concepts targeting software reusability, the possibility of composing simulation tools from reusable components has become realistic. These concepts are at the very foundation of the BioMA framework, where fully working modelling solutions can be assembled by aggregation of smaller software components. By the same way, it can be possible to produce extended or modified versions of an existing model by just adding new components to take care of additional tasks, or by integrating alternative versions of a component already present.
In practice, component reusability solves the “reinventing the wheel” problem. If a large part of a new simulation system can be built using ready-made reusable components, the developers’ efforts can be concentrated on the new functionalities and on the related science, thus accelerating the development process, and improving the scientific content of software.
It therefore becomes easier to explore multiple versions of models, thus helping their improvement, and to use multiple model for more robust analysis. This has been the case in MODEXTREME, where modelling solution were built also reusing existing model libraries previously implemented in software components.
Results and applications
The team of WP2 released a total of seven fully working modelling solutions (MS), notably: i) CropSyst; ii) CropSyst+Extreme; iii) WOFOST; iv) WOFOST + Extreme; v) WARM; vi) WARM + Extreme; vii) AquaCrop (+ an “Extreme” version still under development).
These MS allows the simulation of cropping systems for the potential and water-limited production. The “extreme” versions are extensions of basic models, accounting for the effects of stress deriving from extreme weather events, such as frost, heat and drought.
They can be applied to various scientific and practical applications, especially where water limitation is relevant to the study. The individual components can be utilized as well in new MS or new components can be added to the given MS to produce extended version for new applications.
The MS can be downloaded (http://www.biomamodelling.org) and deployed into either BioMA-Site (for one site simulation) or into BioMA-Spatial (for multiple locations). D2.4 (Donatelli et al., 2016) and Donatelli et al. (in preparation) outline concepts at the basis of MS composition.
Significance and benefits
For model developers, the availability of a set of software components implementing most common models for basic process in biophysical modelling, allow easily building MS around specific needs, and extending already existing ones by just adding the new features desired to fully describe the system of interest. For model users, the availability of a large set of MS allow for many applications in various domains, with the possibility of adopting multi-model approaches thanks to the availability of powerful graphical user interfaces to run MS and analyse the outputs. Respect to previously existing modelling resources, those released during the MODEXTREME project present other additional advantages, such as higher transparency about algorithms, variables and parameters, parameter editing and optimization facilities, an easily configurable and extendible management simulation model, incorporated in each MS.
Successful applications
The MS are being used in other research projects. Ongoing current researches are employing them in scenario analysis to assess the impact of climate change on annual crops in Europe. Another ongoing application uses MS to optimize management strategies in view of upcoming climate change.
References:
MODEXTREME deliverable D2.4. Modelling solutions.
Marcello Donatelli, Davide Fumagalli, Iacopo Cerrani, Davide Fanchini, Simone Bregaglio, Ioannis Athanasiadis, Andrea E. Rizzoli, Enhancing biophysical models reuse: shifting the focus from frameworks to components in the BioMA framework. In preparation for submission to Agricultural Systems.
Key result no 8: Platform interoperability demonstrative actions
Workpackage: WP2
Research aims and background
The BioMA platform was designed to maximise component reusability, which is the capability of easily replace software units within a given software framework. In BioMA the architecture chosen has explicitly included features to make BioMA-components and modelling solutions able to run/be composed even in other frameworks with a reasonable ease, to maximise reusability to the benefit of model development and knowledge sharing.
Model reuse can have different meanings; within biophysical modelling, a “model” can be:
• A process-based representation of either a single process or of a compartment;
• A whole modelling solution allowing the simulation of the system of interest.
However, the point in reuse is not the granularity of the model unit, as in model development. The key difference is if either the discrete unit is to be reused in workflows, or is composed to iterate with other model units over the execution of time-steps. If the reuse is of modelling solutions, adaptors are built following the requirements of a specific software framework either to enable links in workflows, or for ensemble runs. In this case the “model” units run asynchronously and potential incompatibilities of binaries can be overcome. In the case of iteration across model units over time, the requirements for an effective reuse increase. Key features in the architecture of model units are required and binaries incompatibility across models units becomes an insurmountable obstacle in practical terms. Model development requires being able to modify and link model units working synchronously at run time.
Results and applications
Some demonstrative actions were undertaken, consisting in setting up fully working modelling solutions under the following conditions:
1. BioMA components incorporated into models developed under external frameworks;
2. Whole BioMA modelling solutions running into external frameworks;
3. Models designed for external frameworks running into BioMA-compliant modelling solutions;
For all these case studies, the key to the successful implementation is the correct clock handling, which in turns impact on integration mechanism. In the first case study, the BioMA Disease components were incorporated into the APSIM-Wheat model. In the second case study, the BioMA CropSyst-Extreme modelling solution was implemented into an APSIM model container, becoming a modelling solution runnable in that software framework. As for the preceding case, clock handling, in BioMA performed at the Composition Layer level, was handled by APSIM. These experiences allowed refining the procedures for aggregating components as implemented in CLIC (Composition Layer Interactive Coder), the coding-assistant tool to perform composition of modelling solutions. A new feature has been introduced, to generate APSIM-compliant Modelling Solutions in which clock handling is excluded from the model runner and demanded to APSIM.
In the third case study, the converse operation was attempted, that is, to run a non-BioMA compliant software within a BioMA modelling solution. The PaSim model developed by INRA was chosen for this study. PaSim core structure was written in C++, according to different programming paradigms. The interoperability problem was approached by developing a non-BioMA C# wrapper, which implemented homologous methods mapping individually to each original PaSim method. This wrapper has to be called from within a BioMA strategy, to run the whole process workflow. A prototype wrapper with limited functionality has been realized, which proved to function within a BioMA strategy, which in turn could be composed into a demo-MS. Case studies are documented in D2.5 (Donatelli, 2016) and in Holzworth et al. (in preparation).
Significance and benefits
Re-using software units in different modelling platforms substantially means making available to researchers, so far anchored just to their own platform, the opportunity to share an enormous stock of knowledge. The actions that have been undertaken in the MODEXTREME project represent just the proof that the platform interoperability can be achieved. At first, however, technical specifications on how to reach interoperability need to be discussed. Burning questions to answer regard the type of discrete units to share among platforms (strategies implementing simple processes, specific domain component, whole modelling solution), the programming language to use, and a minimum set of requirements for the API - application programming interface.
Successful applications
The BioMA CropSyst-Extreme modelling solution was successfully linked to the APSIM framework.
References:
MODEXTREME deliverable D2.5. Modelling solutions with adapters to other platforms. MODEXTREME project deliverable.
Dean P. Holzworth, Simone Bregaglio, Jeremy P.M. Whish, Marcello Donatelli, Modelling frameworks interoperability: The case BioMA-APSIM. In preparation for submission to Environmental Modelling & Software.
Key result no 9: Provision of climate model based extremes indices for agricultural model changes
Workpackage: WP3
Research aims and background
One important use of agricultural models is to provide information about possible changes in yields caused by anthropogenic climate change in the future. In the MODEXTREME project several important model improvements have been developed and implemented, and it is relevant to study how these improvements affect projections of climate change impacts. As part of this work, a novel set of indicators of extreme weather events appropriate for crop/grassland/tree modelling was identified. These indicators had the purpose to help with decisions about important changes in agricultural circumstances due to climate change.
Results and applications
Bias correction was applied to CORDEX RCM model data before calculation of indices.
In total the indices are (full details can be found in deliverable D3.4):
• Seven standard temperature indices (FD, SU, TR, GSL, WSDI, WSDImax, CSDI)
• Seven standard precipitation indices (RX1day, RX5day, SDII, R10mm, R20mm, CDD, CWD)
• Seven cold temperature phenological indices (S.EM.CRIT.8 S.EM.CRIT.3 EM.AN.CRIT.2 LASTFROST.0 LASTFROST.8 LASTFROST.3 LASTFROST.2 – NB the first three indices require information about sowing, emergence and flowering dates which was only available for SW Europe, hence these indices are only provided for this region)
• Four extreme heat phenological indices (TMAX.40 TMAX.45 GSL.GRASS VHOT.DAYS)
For each of the available 12km Euro-CORDEX simulations and all MODEXTREME study areas, these 25 different indices based on average and extreme values of temperature and precipitation have been made available on the same server and can be accessed at http://ensemblesrt3.dmi.dk/data/MODEXTREME/extremes. These indices include a number of phenological indices relating to both extreme heat and extreme cold. Seven of these indices, all precipitation based, were used in the model selection process.
Significance and benefits
Changes in these indices can be studied and will lead to indications about the most important meteorological changes associated with anthropogenic climate change. This can be used to put calculated results with the MODEXTREME agricultural models into perspective.
Successful applications
Adding this layer of climate detail to the discussion about the future of the agricultural sector in the context of climate change has led to new conclusions about the impact of extreme weather events.
References:
MODEXTREME deliverable D3.1. Historical and perturbed baseline time series for scenario periods.
MODEXTREME deliverable D3.2. Processed RCM/GCM output.
MODEXTREME deliverable D3.3. Report on methods, guidance and recommendations.
MODEXTREME deliverable D3.4. Indices of extremes, simulated and observed.
MODEXTREME deliverable D3.5. Report on projected changes in indices/extremes.
Cathrine Fox Maule, Ole B. Christensen, Peter Thejll, 2014, Selection of a best subset of GCM-RCMs from an ensemble for impact studies, Poster at 3rd Lund Regional-scale Climate Modelling Workshop, June 16-19, Lund, Sweden. http://modextreme.org/wp-content/uploads/sites/4/2014/09/RCM2014_Lund_MODEXTREME_4.rar
Cathrine Fox Maule, Marianne Sloth Madsen, Wilhelm May, Jens Hesselbjerg Christensen, Shuting Yang, Ole Bøssing Christensen, 2015, On procedures for model selection in providing climate scenario data for impact studies – A challenge to both communities, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-9244, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-9244.pdf
Cathrine Fox Maule, Peter Thejll, Richard Cornes, Clare Goodess, Ole B. Christensen, 2015, Effects of bias correction on the climate change signal of extreme indices of precipitation, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-4489, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-4489.pdf
Ole B. Christensen, Cathrine Fox Maule, Richard Cornes, Clare Goodess, Gianni Bellocchi, 2015, Use of regional climate model output for modelling the effects of future weather extremes on agriculture, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/150316_MODEXTREMEWP3_RCC_CMG_OBC.pdf
Fabrizio Ginaldi, Gianni Fila, Davide Fumagalli, Antonio Zucchini, Marcello Donatelli, Exploring the impact of climate variability estimates on crop models predictions in CC impact assessment studies. In preparation for submission to European Journal of Agronomy.
Key result no 10: Bias correction of climate model data for sensitivity studies of agricultural model changes
Workpackage: WP3
Research aims and background
One important use of agricultural models is to provide information about possible changes in yields caused by anthropogenic climate change in the future. In the MODEXTREME project several important model improvements have been developed and implemented, and it is relevant to study how these improvements affect projections of climate change impacts. In WP3 it was the aim to provide processed output from climate models for this purpose to the other work packages in the project. This necessitates performing bias correction of relevant fields for the use in agricultural models, which are quite sensitive to biases in meteorological variables.
Results and applications
Bias correction methods have been developed and implemented for regional climate model (RCM) simulations from the CORDEX project for selected periods, resulting in sets of bias corrected gridded daily data sets of daily maximum and minimum surface air temperature as well as daily total precipitation. Other meteorological variables, in particular solar irradiation, have been calculated with the help of weather generators due to a lack of consistent measurements, which prevents bias correction of daily data. For south-western Europe the observational dataset underlying the bias correction is the gridded E-OBS dataset, which provides all three variables on a 25 km-grid. RCM output data have been interpolated to this grid, and quantile-quantile based bias correction has been applied. For the other MODEXTREME study areas, station observations have been used, either from public databases or provided by local project partners.
The quantile-based method is constructed by sorting all values for the relevant point and period in the model and correspondingly sorting observation values. Then, each fractile (the highest, second-highest, etc.) of the model is corrected to the corresponding observational value. This does not affect the chronology of the modelled events, and we will therefore hardly ever have unphysical situations like heavy precipitation coinciding with very high temperature due to bias correction. A piecewise linear function is then adapted and employed also for values outside the period used for fitting. For values entirely outside the range of model values in the fitting period, constant correction is used taking from the nearest fractile.
Significance and benefits
Plants are extremely sensitive to small variations in temperature and precipitation, and therefore agricultural models have a very high bar for the minimum quality of input data from numerical climate models. For this reason bias correction is a necessity for obtaining information about agriculture in possible changed future climates with any hope of realism. In MODEXTREME the focus has been on indices of extremes most relevant for impact on crop/grassland/tree species, including a number of new phenological indices which are not normally used by the climate research community. Several different European projects have constructed bias corrected climate simulation data using a variety of methods and underlying observations. At the start of MODEXTREME this was not the case, and therefore the work performed in WP3 was necessary for being able to investigate the effects of climate extremes on agricultural outputs using the models developed in MODEXTREME. Even though alternatives do exist now, the MODEXTREME climate datasets are still deemed useful and applicable to not just agricultural modelling but also to other forms of impacts models.
Successful applications
Adding this layer of climate detail to the discussion about the future of the agricultural sector in the context of climate change has led to new conclusions about the impact of extreme weather events.
References:
MODEXTREME deliverable D3.1. Historical and perturbed baseline time series for scenario periods.
MODEXTREME deliverable D3.2. Processed RCM/GCM output.
MODEXTREME deliverable D3.3. Report on methods, guidance and recommendations.
MODEXTREME deliverable D3.4. Indices of extremes, simulated and observed.
MODEXTREME deliverable D3.5. Report on projected changes in indices/extremes.
Cathrine Fox Maule, Ole B. Christensen, Peter Thejll, 2014, Selection of a best subset of GCM-RCMs from an ensemble for impact studies, Poster at 3rd Lund Regional-scale Climate Modelling Workshop, June 16-19, Lund, Sweden. http://modextreme.org/wp-content/uploads/sites/4/2014/09/RCM2014_Lund_MODEXTREME_4.rar
Cathrine Fox Maule, Marianne Sloth Madsen, Wilhelm May, Jens Hesselbjerg Christensen, Shuting Yang, Ole Bøssing Christensen, 2015, On procedures for model selection in providing climate scenario data for impact studies – A challenge to both communities, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-9244, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-9244.pdf
Cathrine Fox Maule, Peter Thejll, Richard Cornes, Clare Goodess, Ole B. Christensen, 2015, Effects of bias correction on the climate change signal of extreme indices of precipitation, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-4489, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-4489.pdf
Ole B. Christensen, Cathrine Fox Maule, Richard Cornes, Clare Goodess, Gianni Bellocchi, 2015, Use of regional climate model output for modelling the effects of future weather extremes on agriculture, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/150316_MODEXTREMEWP3_RCC_CMG_OBC.pdf
Key result no 11: Selection of climate model data sets for sensitivity studies of agricultural model changes
Workpackage: WP3
Research aims and background
One important use of agricultural models is to provide information about possible changes in yields caused by anthropogenic climate change in the future. In the MODEXTREME project several important model improvements have been developed and implemented, and it is relevant to study how these improvements affect projections of climate change impacts. As part of this work, a novel set of indicators of extreme weather events appropriate for crop modelling was identified. In WP3 it was the aim to provide processed output from climate models for this purpose to the other work packages in the project. This processing consists of two major aspects: Bias correction of relevant fields for the use in agricultural models, which are quite sensitive to biases in meteorological variables; and an implemented method for selection of representative climate model simulations for cases where the amount of computations involved in agricultural modelling makes it necessary to not study all available climate model simulations. It was decided to use at most four regional model simulations selected as one central and 3 sampling the set of projected changes in a set of seven precipitation-based indices. This selection is done for each area under investigation: for south-west Europe as a whole, for Ukraine, China, South Africa, and South America individually.
Results and applications
For determination of model differences, the following precipitation based indices were used:
• PR Average precipitation
• CC Climate Change signal of annual mean
• RX1day Highest precipitation amount in a one-day period
• RX5day Highest precipitation amount in a five-day period.
• SDII Simple Daily Intensity Index
• R10mm Heavy precipitation days
• R20mm Very heavy precipitation days
• CDD Consecutive Dry Days
• CWD Consecutive Wet Days
These indices were calculated for the present-time as well as the far-future bias-corrected time slice of 30 years for all models. The relative change of each index for the area in question then form a nine-element vector for each model, and ordinary PCA analysis can be applied in order to choose the most different modelled climate change. Employing a principal-component analysis method for model selection, four simulations were selected from a set of 10 simulations for south-west Europe and for Ukraine and from a set of 18 simulations for South Africa. For the remaining areas no selection was necessary, as less than four simulations were available. The resulting sets of at most four simulations, bias corrected in order to have distribution functions identical to the available observational data for the present-day simulation period, have been available for most of the project duration at a server with access for project partners. For each simulation and study area, 25 alternative indices based on average and extreme values of temperature and precipitation have been made available on the same server. These indices include a number of phenological indices relating to both extreme heat and extreme cold.
Significance and benefits
Selection of ensemble members from a larger ensemble is necessary for the purpose of many different kinds of impacts modelling with computer-heavy impacts models. This is the case for the agricultural models in MODEXTREME, but the concept is generally useful.
Successful applications
Adding this layer of climate detail to the discussion about the future of the agricultural sector in the context of climate change has led to new conclusions about the impact of extreme weather events.
References:
MODEXTREME deliverable D3.1. Historical and perturbed baseline time series for scenario periods.
MODEXTREME deliverable D3.2. Processed RCM/GCM output.
MODEXTREME deliverable D3.3. Report on methods, guidance and recommendations.
MODEXTREME deliverable D3.4. Indices of extremes, simulated and observed.
MODEXTREME deliverable D3.5. Report on projected changes in indices/extremes.
Cathrine Fox Maule, Ole B. Christensen, Peter Thejll, 2014, Selection of a best subset of GCM-RCMs from an ensemble for impact studies, Poster at 3rd Lund Regional-scale Climate Modelling Workshop, June 16-19, Lund, Sweden. http://modextreme.org/wp-content/uploads/sites/4/2014/09/RCM2014_Lund_MODEXTREME_4.rar
Cathrine Fox Maule, Marianne Sloth Madsen, Wilhelm May, Jens Hesselbjerg Christensen, Shuting Yang, Ole Bøssing Christensen, 2015, On procedures for model selection in providing climate scenario data for impact studies – A challenge to both communities, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-9244, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-9244.pdf
Cathrine Fox Maule, Peter Thejll, Richard Cornes, Clare Goodess, Ole B. Christensen, 2015, Effects of bias correction on the climate change signal of extreme indices of precipitation, EGU General Assembly 2015, Geophysical Research Abstracts, Vol. 17, EGU2015-4489, April 12–17, Wien, Austria. http://meetingorganizer.copernicus.org/EGU2015/EGU2015-4489.pdf
Ole B. Christensen, Cathrine Fox Maule, Richard Cornes, Clare Goodess, Gianni Bellocchi, 2015, Use of regional climate model output for modelling the effects of future weather extremes on agriculture, Poster at Climate-Smart Agriculture 2015 Global Science Conference, March 16-18, Montpellier, France. http://modextreme.org/wp-content/uploads/sites/4/2015/03/150316_MODEXTREMEWP3_RCC_CMG_OBC.pdf
Key result no 12: Model evaluation
Workpackage: WP4
Research aims and background
The agricultural system is increasingly threatened by climate change with an increase in the frequency/intensity of extreme weather events. Simulation models of crop, grassland and tree systems have been often used in environmental conditions markedly different from the ones they were developed. This has been achieved by reproducing time trends of observations by fitting the values of the parameters included in core equations, degrading the process-based nature of models to purely statistical tools, hence unsuitable to explore conditions where model performance cannot be corroborated by model testing, as in scenarios of climate change. A model accounting for the impact of extreme weather events must be able to perform under both ordinary and anomalous weather conditions without changing its parameterization. Available crop, grassland and tree models were developed based on datasets collected – in most cases – under conditions for which plants were well adapted, thus without explicitly accounting for extreme weather events. This is one of the main reasons for the poor capability of models to capture the yearly variability in official yields at regional and country level and poor coherence between the mathematical description of crop/grassland/tree responses to the environment and the values of the parameters used. In WP4 Agricultural system models’ ability to simulate plant growth was evaluated by comparing existing (EMS) and modified (MMS) modelling solutions, the latter accounting for the impact of extreme weather events. The evaluations were done for the historical period against observed data, and for the future period with downscaled climate projections. The results were presented to a variety of stakeholders in each country to receive feedback. The WOFOST model’s sensitivity to different parameters was also tested.
Results and applications
Regarding evaluation with the observed data, the results highlight some limitations in the response to extreme weather events of state-of-the-art modelling solutions, as emerged from the assessment performed at a variety of sites and systems. It is difficult to draw a general conclusion from the comparison between modified and existing modelling solutions as the agro-ecosystems assessed are diverse, and the exercise was often based on weather and crop data of relatively short length. In the evaluation with future climate data, clearer differences between EMS and MMS emerged, in comparison to evaluation with historical data. For WOFOST, in general MMS was found to simulate lower yield and AGB than EMS due to the additional “extreme” component of MMS. For CropSyst, as with WOFOST, MMS tends to simulate lower yield and AGB than EMS by accounting for negative impacts of extreme weather events on crop growth. Similar results were obtained for PaSim (grassland model), OLIVECAN (olive tree model) and WARM (rice model).
The sensitivity analysis of WOFOST identified the most influential WOFOST parameters to simulate crop yield. The influential parameters vary depending on the combination crop x site. With a few exceptions, the parameters identified by the Morris method obtained the highest eFAST total sensitivity index, which underlines that the two sensitivity analysis methods are closely related. These results suggest the plasticity of the WOFOST model. The WP4 results indicate that it is important to consider the effect of extreme weather events in agriculture system models. It is likely that most simulations of plant growth without explicit consideration of “extreme” component overestimate crop productivity and underestimate its variability. The improvement introduced by MMS can be the basis for a discussion around the employment of these modelling tools to support food security issues in Europe and worldwide.
Significance and benefits
The results indicate the importance of considering the effect of extreme weather events in agriculture system models. Better modelling tools that perform well even under anomalous weather conditions can be built for operational use, for example, in-season yield forecasting and for climate change studies. Stakeholders suggest that the WP4 results are useful for elaboration of long-term agriculture development projects and policy planning for agriculture and food security, e.g. agricultural insurance, credits, investments, resource allocation/mobilization, irrigation systems. MODEXTREME models can also guide further research on agricultural vulnerabilities, variety trials, or plant breeding.
Successful applications
It is recommended to improve simulation of the effects of other extreme weather events such as heavy rainfalls or extremely cold temperatures. Integration of pests and diseases model is another suggestion. There was a high interest in the use of models for support to farm management decisions at much smaller scales to identify and implement appropriate adaptation strategies on the farm. To draw implication for the wider society, food security at the national and international levels needs to be discussed in a much broader context, and by accounting for other aspects of food security beyond food quantity (or productivity).
References:
MODEXTREME deliverable D4.1. Results on model testing against experimental/observational and gridded data.
MODEXTREME deliverable D4.2. Results on model evaluation in future climate scenarios.
MODEXTREME deliverable D4.3. Interaction with stakeholders in the frame of food security.
MODEXTREME deliverable D4.4. Report on the assessment of new modelling solutions.
Robert Mangani, Eyob Tesfamariam, Gianni Bellocchi, Abubeker Hassen, Comparison of existing and modified CropSyst crop model to assess the impact of extreme weather events on maize yield in South Africa. Submitted to Climate Research.
Renáta Sándor, Catherine Picon-Cochard, Raphael Martin, Frédérique Louault, Katja Klumpp, David Borras, Gianni Bellocchi, Plant acclimation to temperature: developments in the Pasture Simulation model. Submitted to Field Crops Research.
Carlo Gilardelli, Roberto Confalonieri, Gianni Bellocchi, Sensitivity analysis of the effect of climate changes on crop production in Europe: modelling solutions of the crop model WOFOST. In preparation.
Key result no 13: Evaluation of forecast reliability of agro-climatic indicators
Workpackage: WP5
Research aims and background
In the last decades, a variety of forecasting systems were developed. The first methods were based on surveys or crop scouting; these approaches were replaced since the 1990s by more objective and sound techniques, based on the single or integrated use of agro-climatic indicators, remote-sensing information, and crop models. In specific contexts where crop production fluctuations are driven by few main factors, statistical models based on relationships between a few, relevant agro-climatic indicators and crop yields can be able to accurately explain the inter-annual crop yield variability, and thus to reliably forecast crop yields. In recent years, many climatic and agro-climatic indicators were developed and related to the impact of extreme weather events (e.g. drought, extreme temperatures) on agricultural productions. These metrics can be very simple, like those based on counts, i.e. the number of times a phenomenon occurs, sums (e.g. thermal and rainfall sums), or more complex, like those considering plant susceptibility to an extreme event in different moments during the crop cycle.
The aims of this research were:
• to evaluate the potential for forecasting crop yields based on spatially aggregated temperature and drought indicators for the main winter and summer crops and for mown grasslands in all European countries where the specific crop is cultivated;
• to identify and analyse the combinations crop × country where forecasting systems based on agro-climatic indicators present sufficient reliability in case of extreme events.
Results and applications
The indicators ARIDcr (number of days with ARID higher than a fixed threshold, with ARID being 1 – actual to potential transpiration ratio) and Tmaxcr (number of days with Tmax higher than a fixed threshold) were those with the largest forecasting capability, whereas the indicator involved with the impact of low temperatures were rarely selected by the forecasting system. The best performances were achieved for wheat, barley and rye in Spain, with 82%, 85% and 76% of yield variability captured by the forecasting system, and with EF (Nash and Sutcliffe modelling efficiency) ranging from 0.76 to 0.8. For these cases, the trend of yields during the time window was well reproduce by the system, without marked differences between official and forecasted yields. Indeed, Spain is mainly characterized by a Mediterranean climate and winter crop production is mostly driven by rainfall volumes and distribution during the growing season. Satisfactory results were achieved also for sunflower in Bulgaria and sugar beet in Croatia, with R2 values of 0.81 and 0.74 respectively. However, a great part of yield variability was explained by the introduction of technological innovations (a significand technological trend was found), especially in Bulgaria, where the regression model explained only 20% of the year-to-year yield fluctuations.
Significance and benefits
The applied methodology led to satisfactory forecasting performance in countries where crop production is mainly driven by a specific meteorological variable and where this variable often assumes extreme values (e.g. drought, temperatures higher or lower than critical thresholds).
Successful applications
The above represents an important benefit for the community, since – under the conditions depicted in the previous section – simple and robust crop yield forecasting systems can be successfully adopted, without the need of complex approaches based on dynamic simulation models.
References:
MODEXTREME deliverable D5.1. Evaluation of forecasts based on agro-climatic indicators.
Key result no 14: Evaluation of forecast reliability of improved modelling solutions
Workpackage: WP5
Research aims and background
The most sophisticated forecasting systems are based on crop simulation models. An example is represented by the European Commission Crop Growth and Monitoring System (CGMS), used within the MARS project for short-term yield forecasts for the main food crops at European level. The MARS system is based on the WOFOST model for all crops but rice, for which the WARM model is used. The uncertainty in simulations is reduced by post-processing model outputs with historical series of official yields. The need for a statistical layer is due to a portion of variability affecting actual yields which is not reproduced by the monitoring and forecasting system. Among the sources of uncertainty, a key role is played by the absence – in many process-based crop/grassland models – of algorithms for reproducing the effect of biophysical processes actually affecting yields. Concerning the general project goal, existing forecasting systems are based on models developed for conditions of good adaptation and often designed for temperate environments. The effects of unusual weather events on crop productivity, including crop failure in case of extreme conditions, are thus often overlooked or unsatisfactorily simulated by the available crop models. However, raising temperature and a projected increase in the frequency and intensity of extreme weather events such as droughts, heat waves, frost events, are increasingly threatening the usefulness of forecasting services. The aim of this work was to evaluate – within the European Commission MARS workflow – the accuracy of forecasting systems based on standard (CGMS) and improved (including MODEXTREME impact models) modelling solutions for the main crops in Europe.
Results and applications
The standard CGMS system applied to maize showed satisfactory performances in most of the countries, with model reliability increasing after the flowering stage. As an example, in Hungary (the third European maize producer) the system, without the presence of a significant technological trend, explained 46%, 71% and 77% of yields variability while approaching maturity for the three forecasting moments (20th, 24th and 29th ten-day periods, respectively). In general, the best results for the standard solution were achieved in Germany at maturity for potato and grain maize, and in Spain for soft wheat, with 86%, 88% and 86% of explained variability, respectively. The situation was less satisfactory for spring barley, with the exception of Spain, where the comparison of official and forecasted yields led to R2 and EF values of 0.81 and 0.79 respectively. The inclusion of MODEXTREME impact models within the CGMS standard solution led to improve the forecasting reliability in 57 out of 75 combinations crop × country × moment when the forecasting event was triggered. In particular, improvements were achieved in almost all the countries for durum wheat, soft wheat and grain maize, although satisfactory performance were achieved for the latter also with the standard system. The best forecasting performances were achieved for grain maize in Germany, where 92% of inter-annual variability in yields was explained by the MODEXTREME solutions. In most of the combinations crop × country × ten-day period, response functions to extreme events and minimum/maximum canopy temperatures improved the forecasting capability more than state variables. In particular, the response function for cold stress was one of the most relevant indicator.
Significance and benefits
The inclusion of MODEXTREME impact models within the CGMS standard solution led to improve the forecasting reliability in 57 out of 75 combinations crop × country × moment when the forecasting event was triggered. In particular, improvements were achieved in almost all the countries for durum wheat, soft wheat and grain maize, although satisfactory performance were achieved for the latter also with the standard system. The best forecasting performances were achieved for grain maize in Germany, where 92% of inter-annual variability of yields was explained by the MODEXTREME solutions. The improved solution increased up to 44% the amount of variance explained by the forecasting system for spring barley in Poland. In general, the most satisfactory forecasting results were achieved after flowering period.
Successful applications
Achieved results suggest adopting the modelling solutions improved during the MODEXTREME project (those including the impact models for extreme weather events) for forecasting purposes in Europe. A dedicated workshop was organized with the EC-JRC MARS team to define strategies to transfer MODEXTREME results within the MARS operational workflow.
References:
MODEXTREME deliverable D5.2. Report on the evaluation of improved process based modelling solutions and of the standard CGMS ones.
Roberto Confalonieri, 2014, CoSMo: A simple approach for reproducing plant community dynamics using a single instance of generic crop simulators, Ecological Modelling 286, 1-10.
Giovanni Cappelli, Sevim Seda Yamaç, Tommaso Stella, Caterina Francone, Livia Paleari, Marco Negri, Roberto Confalonieri, 2015, Are advantages from the partial replacement of corn with second-generation energy crops undermined by climate change? A case study for giant reed in northern Italy, Biomass and Bioenergy 80, 85-93.
Livia Paleari, Giovanni Cappelli, Simone Bregaglio, Marco Acutis, Marcello Donatelli, Gian Attilio Sacchi, Elisabetta Lupotto, Mirco Boschetti, Giacinto Manfron, Roberto Confalonieri, 2015, District specific, in silico evaluation of rice ideotypes improved for resistance/tolerance traits to biotic and abiotic stressors under climate change scenarios, Climatic Change 132, 661-675.
Simone Bregaglio, Francesca Orlando, Emanuela Forni, Tommaso De Gregorio, Simone Falzoi, Chiara Boni, Michele Pisetta, Roberto Confalonieri, 2016, Development and evaluation of new modelling solutions to simulate hazelnut (Corylus avellana L.) growth and development, Ecological Modelling 329, 86-89.
Roberto Confalonieri, Simone Bregaglio, Myriam Adam, Françoise Ruget, T. Li, T. Hasegawa, X. Yin, Y. Zhu, K. Boote, S. Buis, T. Fumoto, D. Gaydon, T. Lafarge, M. Marcaida, H. Nakagawa, Alex C. Ruane, B. Singh, U. Singh, L. Tang, F. Tao, J. Fugice, H. Yoshida, Z. Zhang, L.T. Wilson, J. Baker, Y. Yang, Y. Masutomi, Daniel Wallach, Marco Acutis, B. Bouman, 2016, A taxonomy-based approach to shed light on the babel of mathematical models for rice simulations, Environmental Modelling & Software 85, 332-341.
Carlo Gilardelli, Tommaso Stella, Nicolò Frasso, Giovanni Cappelli, Simone Bregaglio, Marcello E. Chiodini, Barbara Scaglia, Roberto Confalonieri, 2016. WOFOST-GTC: a new model for the simulation of winter rapeseed production and oil quality, Field Crops Research 197, 125-132.
Livia Paleari, Roberto Confalonieri, 2016. Sensitivity analysis of a sensitivity analysis: we are likely overlooking the impact of distributional assumptions, Ecological Modelling 340, 57-63.
Key result no 15: Identification of the most reliable workflows for yield forecasts
Workpackage: WP5
Research aims and background
A variety of systems for crop yield forecasts was developed in the last decades to face the increasing demand of timely and reliable information to be used for properly managing agricultural markets and policies. The most widespread systems are based on agro-climatic indicators, remote-sensing and crop models, with those techniques used alone or in an integrated fashion. In general, simple relationships between specific agro-climatic indicators and crop yields can be sufficiently accurate when applied to context where crop yield fluctuations are driven by few key meteorological drivers. However, the most sophisticated methods are based on crop simulation models (e.g. the European Commission Crop Growth and Monitoring System). Most of available systems were developed targeting temperate environments. However, the projected increase in the frequency and intensity of extreme weather events (e.g. droughts, heat waves and frost events) is threatening the usefulness of current forecasting services.
The aims of this work were:
• to analyse and compare – at European scale – the reliability of forecasting systems based on (i) agro-meteorological indicators; (ii) outputs of the CGMS standard modelling solution; (iii) outputs of the CGMS improved (including MODEXTREME impact models) modelling solution; (iv) the combination of agro-climatic indicators and outputs of both the standard and improved modelling solutions (hybrid approach);
• identify the most reliable workflow for each combination crop × country × forecasting event.
Results and applications
In general, the improved modelling solutions (including MODEXTREME impact models) showed the most satisfactory forecasting performances, especially for durum wheat, soft wheat and maize. The highest reliability was obtained for maize in Germany at the 24th 10-day period, with 92% of explained yield variability. Model outputs accounting for extreme weather events allowed increasing the forecasting capability in most cases. The output that often explained a large part of inter-annual variability was the harvest index response function to cold. However, in the northern states, simulated damages to leaf area due to frost played a key role. The system based on agro-meteorological indicators showed the best performances only for potato in Poland and in Germany at the 16th 10-day period. The integrated use of the different forecasting systems (hybrid) allowed to improve CGMS performances in 87% of the combinations crop × country × forecasting moment. However, most of the improvement in the forecasting reliability was due to the MODEXTREME impact models (improved dynamic solutions). The addition of agro-climatic indicators to those provided by the improved solutions led to a slight improvement of forecasting performances in 24% of the cases.
Significance and benefits
This part of the study allowed identifying the most reliable workflow for each combination crop × country × forecasting moment, providing a crucial support for operational forecasting services in Europe. Pros and cons of different approaches (agro-climatic indicators, standard and improved solutions, hybrid systems) were highlighted for different agro-climatic contexts.
Successful applications
The hybrid forecasting system developed and tested within the project can be a powerful tool within operational forecasting systems. Like for the system based on improved modelling solutions, a dedicated workshop was organized with the EC-JRC MARS team to define strategies to transfer MODEXTREME results within the MARS operational workflow.
References:
MODEXTREME deliverable D5.3. Report on the identification of the most reliable workflows for yield forecasts.
Roberto Confalonieri, Simone Bregaglio, Marco Acutis, 2016, Quantifying uncertainty in crop model predictions due to the uncertainty in the observations used for calibration, Ecological Modelling 328, 72-77.
Roberto Confalonieri, Francesca Orlando, Livia Paleari, Tommaso Stella, Carlo Gilardelli, Ermes Movedi, Valentina Pagani, Giovanni Cappelli, Andrea Vertemara, Luigi Alberti, Paolo Alberti, Samuel Atanassiu, Matteo Bonaiti, Giovanni Cappelletti, Matteo Ceruti, Andrea Confalonieri, Gabriele Corgatelli, Paolo Corti, Michela dell’Oro, Alessandro Ghidoni, Angelo Lamarta, Alberto Maghini, Martino Mambretti, Agnese Manchia, Gianluca Massoni, Pierangelo Mutti, Stefano Pariani, 2016, Uncertainty in crop model predictions: What is the role of users?, Environmental Modelling & Software 81, 165-173.
Valentina Pagani, Tommaso Guarneri, Davide Fumagalli, Ermes Movedi, Luca Testi, Tommy Klein, Pierluigi Calanca, Francisco Villalobos, Stefan Niemeyer, Gianni Bellocchi, Roberto Confalonieri, Accounting for the effect of weather extremes on cereal yield forecasting in Europe. In preparation.
Potential Impact:
POTENTIAL IMPACTS
The European and international added value of the project lies in increasing the capability of yield monitoring and forecasting system to support agricultural and food security policies in the frame of the European Commission and beyond. For that, MODEXTREME has:
- formalised and integrated into statistical and process-based simulation models the responses of main crop/grassland/tree systems to extreme weather events;
- assessed the above approaches against experimental/observational datasets;
- used the above approaches to generate trajectories of agricultural productivity in short-medium time horizons;
- developed generically reusable software units implementing libraries of models;
- extended the modelling capabilities of the BioMA platform towards weather extremes;
- supported the monitoring and forecasting system of the European Commission JRC-MARS;
- transferred the knowledge gained and the modelling solutions developed to JRC-MARS analysts and other beneficiaries, for dedicated implementation with local stakeholders to improve yield forecasts and increase food security at the EU level and worldwide.
By providing alternate modelling solutions to simulate a variety of agricultural systems worldwide, yet under scenarios of climate change, MODEXTREME provided evidence of the operational usability of modelling approaches in forecasting systems, by:
- showing the potential of novel modelling tools to account for the impact of weather extremes on agricultural systems;
- reducing uncertainties in the estimates of agricultural production under weather extremes;
- strengthening the role of the EU forecasting system in agriculture.
SCIENTIFIC, TECHNOLOGICAL AND POLICY OUTCOMES
MODEXTREME has developed novel approaches of salient plant processes to represent mathematically the impact of extreme weather events (heatwaves, aridity, cold shocks and frost) on leaf development and grain accumulation as well as other physiological pathways (e.g. thermal acclimation of grassland species). They provided scientific ground for the implementation of dedicated software code for use into system models, which increased the capability of existing simulation tools to reproduce experimentally established system performances. The simulation models used in MODEXTREME have benefited of the improvements introduced, which can be extended to virtually any other simulation model of plant growth and development. Agro-climatic indicators have also been developed and used to identify extreme weather events along plant growing seasons and assess their impact on agricultural production. To facilitate the calculation of indicators of climatic extremes having impact on agricultural production, the web application webXTREME was developed and made freely accessible online as a way to promote collaboration within the scientific community and between scientists and stakeholders.
Novel software components for the impact of extreme events, new modelling solutions for crop/grassland/tree systems and several supporting tools have been developed within MODEXTREME, mostly for and through the platform BioMA (Biophysical Model Applications). BioMA provides an advanced technological support (open and flexible to develop modelling solutions upon reusable software components) to the European Commission for scenario analyses in agriculture and climate change in response to the need of the work program of JRC MARS - Agri4Cast. The latter includes impact assessment and the development of adaptation strategies for agriculture under climate change scenarios, as a key element in the EU agriculture monitoring capabilities. The framework BioMA is currently being used as production tool in the EU forecast run by the JRC of the European Commission and it holds potential for further developments and extension to new functionalities in the future (for instance to deal with challenges to food security associated with the exacerbated impact of pests and diseases).
Newly developed datasets of climate scenarios in project regions opens up to the possibility to view projected weather data in unprecedented detail, that is downscaled, bias-corrected, post-processed and put in a consistent framework for direct use for simulation modelling in agriculture. The dataset has been shared with the JRC and several scientists worldwide which see it as the missing piece in a wide range of studies they planned to do. Modelling teams outside of the project requested access to these datasets. The quantified scenario drivers based on RCP8.5 benefited strongly from the most recent IPCC scenario development process (through the CORDEX initiative). Used as inputs to force models simulating agricultural production, they provide new insights about the substantial potential for food security assessment. Adding this new layer of detail to the discussion about the future of the agricultural sector in the context of climate change may lead to new conclusions about the impact of extreme weather events.
The experimental/observational results, the modelling approaches and the simulation outputs that have been produced in MODEXTREME offer benefits to a variety of end-users. These include both sector specialists and farmers, as well as local, regional, national and EU policy makers working directly or indirectly with agriculture and also scientists involved in food security assessment. The agro-environmental modelling community would also benefit from both the developments in modelling approaches and improved parameterization for simulations of crops, grasslands and tree crops (olive and grape) at a variety of contrasting sites. The uncertainty quantification of model estimates of agricultural production and other outputs from agricultural systems holds potential to direct further work across a larger variety of conditions than those examined in MODEXTREME. For instance, modelling approaches could be expanded to assess the impact of extreme weather events on subsistence farmers, which account a large portion of the rural society.
MODEXTREME provided detailed results showing the best workflow in the estimate of agriculture production in Europe for combinations of crop-country-forecasting events, based on alternate modelling solutions and agro-climatic indicators. They show the increased forecasting reliability, obtained with novel modelling solutions and indicators developed in MODEXTREME, in a spatially-explicit way by indicating for each European country the best forecasting workflow for a broad set of crops. The information provided by these results is readily exploitable, and the tools and approaches developed are of direct use by the European Commission as they represent improved solutions of the Agri4Cast MARS Crop Yield forecasting system. This important outcome has been achieved thanks to close collaboration with JRC MARS - Agri4Cast.
COMMUNICATION AND OUTREACH
A project brochure outlining the aims of the project and its structure was developed and handed out at presentations, exhibitions and conferences, and a project website (which will be maintained beyond the funding period) was deployed with open access to all project outcomes. Particular emphasis has been placed on dissemination events to enhance sharing of project results across broad communities of scientists and non-scientific actors. In particular, four dissemination workshops were organized as part of the dissemination plan:
1. “Modelling the impact of extreme climatic events in agriculture” in the frame of 5th International Symposium for Farming Systems Design (Montpellier, France, September 10th, 2015);
2. “Modelling the impacts of extreme climatic events on agrifood systems” at FAO headquarter (Rome, Italy, November 3rd, 2015);
3. “Agricultural Modelling Hackathon” in the frame of the 8th Biennial Congress of the International Environmental Modelling and Software Society (Toulouse, France, July 12th, 2016);
4. “Improving agricultural modelling for the impact of extreme weather events” at European Commission DG AGRI headquarter (Brussels, Belgium, September 20th, 2016).
Young scientists were supported by the project to illustrate results in these workshops. To give wider access several e-learning materials in a variety of formats (presentations, training documents, science briefs, software tools) were made available through the public website, and two public newsletters were distributed electronically, which ensured effective dissemination among broad communities of scientists and policy makers worldwide.
A Stakeholder Platform was established to create a participatory framework linking key end-users to the project team. It comprised representatives of the European Commission (JRC, DG AGRI and other DGs) and local actors and experimental scientists identified by partners (and put under FAO coordination) such chambers of agriculture, agricultural authorities, governmental bodies, etc.
EC delegates attended all the project meetings (kick-off and three annual meetings) and subsequently produced reports of their recommendations in relation to their role, their expectations of the project outcomes and on the process of consultation between the project team and the stakeholder platform. Dissemination and training plans were produced based on the description of work, and updated after each annual meeting to incorporate the above recommendations and those from the Scientific Advisory Board (three distinguished scientists) about key actions to implement the project plan. Interactions with the main stakeholder (JRC MARS Unit) have become more intense in the final period (e.g. a meeting was organized in Ispra in October 2015 followed by several exchanges and contacts), resulting in a joint action plan that responded to the needs and opportunities identified in the project (such as exploitation of data of the MARS Crop Yield Forecasting System and creation of modelling solutions compatible with that same system). The project also held two sessions at JRC (Ispra, Italy) as part of the training activities addressed to MARS agents.
Dedicated workshops addressed to local stakeholders were organised in Argentina, Brazil, China, France, Italy, South Africa, Spain, Switzerland and Ukraine. Views from local workshops supported the compilation by FAO of questionnaire-based feedback on the modelling support to food security issues and led to a definition of key outputs from the project that are of relevance to stakeholders, their likely impact, and strategies for post-project dissemination and implementation of the outputs.
The dissemination activity of the project was completed by illustrating relevant research outcomes in international events such as the 5th AgMIP global workshop (Gainesville FL, USA, February 25th-28th, 2015), the 3rd Global Science Conference on Climate - Smart Agriculture (Montpellier, France, March 16th-18th 2015) and the 8th Biennial Congress of the International Environmental Modelling and Software Society (Toulouse, France, July 10th-14th, 2016).
An internal collaborative platform has been designed so as to ensure a regular exchange among the project partners, the stakeholder platform and the scientific advisory board (also assisted by periodic circulation of internal newsletters). Supported by the management team, the internal platform will be maintained beyond the funding period to ensure access to consolidated training material, datasets and scientific outcomes.
List of Websites:
Project public website
The website address is the following: http://modextreme.org
To support dissemination and to host web-based training courses, the open public website informs on the aims and structure of the project, project outputs (including access to software), presentations of project workshops and training events, and other news of the project. Public deliverables are also published on the website. Two public newsletters, also available on the website, have been produced to outline the main outcomes of the project and likely impact.
Relevant contact details
MODEXTREME Coordinator: Gianni Bellocchi
INRA
5 chemin de Beaulieu, 63039 Clermont-Ferrand (France)
Tel.: +33 (0) 4 43 76 16 01
Email: gianni.bellocchi@inra.fr
MODEXTREME Project Manager: Léa Tourneur
INRA Transfert
3 rue de Pondichéry, 75015 Paris (France)
Tel.: + 33 (0) 1 76 21 62 06
Email: lea.tourneur@inra.fr
