Frédérick Garcia

An open platform to build, evaluate and simulate integrated models of farming and agro-ecosystems

Due to significant changes in agro-ecological contexts, farmers need new solutions to produce goods. Modelling complements field experiments in the design of new farming systems. French researchers involved in such design issues developed a specific modelling platform to help model, simulate and evaluate cropping systems. After testing several existing environments, the RECORD platform was developed under the VLE environment, allowing the design of atomic and coupled models. It integrates different time steps and spatial scales and proposes some standard formalisms used to model agro-ecosystems (e.g. difference equations, differential equations, state charts...). A graphic user interface was designed to simplify coding tasks. A variety of research projects already use this platform. Examples are given showing the ability to recode simple models, encapsulate more complex models, link with GIS and databases, and use the R statistical package to run models and analyse simulation outputs. The option to use web interfaces enables application by non-scientist end-users. As the models follow a given standard, they can be placed in a repository and used by other researchers. Linking RECORD to other international platforms is now a compelling issue.

Interactive simulation modeling in farm decision-making

Of late, a series of methods and tools have evolved in order to assist farmers in making decisions, based on the development of computers. In this paper, an attempt has been made to focus on simulation tools as a means to expand interactivity. The first part deals with the understanding of interactivity. Then the use of simulation as a support for interactivity is developed. In this endeavor, the evolution of understanding of the role played by simulation tools for farm management decision-making has been considered crucial, based on experiments carried out in an interactive manner with both farmers and consultants. Finally, we introduce and discuss numerous potential opportunities provided by the new techniques through automatic machine learning and multi-agent modeling.

MELODIE: a whole-farm model to study the dynamics of nutrients in dairy and pig farms with crops.

In regions of intensive pig and dairy farming, nutrient losses to the environment at farm level are a source of concern for water and air quality. Dynamic models are useful tools to evaluate the effects of production strategies on nutrient flows and losses to the environment. This paper presents the development of a new whole-farm model upscaling dynamic models developed at the field or animal scale. The model, called MELODIE, is based on an original structure with interacting biotechnical and decisional modules. Indeed, it is supported by an ontology of production systems and the associated programming platform DIESE. The biotechnical module simulates the nutrient flows in the different animal, soil and crops and manure sub-models. The decision module relies on an annual optimization of cropping and spreading allocation plans, and on the flexible execution of activity plans for each simulated year. These plans are examined every day by an operational management sub-model and their application is context dependent. As a result, MELODIE dynamically simulates the flows of carbon, nitrogen, phosphorus, copper, zinc and water within the whole farm over the short and long-term considering both the farming system and its adaptation to climatic conditions. Therefore, it is possible to study both the spatial and temporal heterogeneity of the environmental risks, and to test changes of practices and innovative scenarios. This is illustrated with one example of simulation plan on dairy farms to interpret the Nitrogen farm-gate budget indicator. It shows that this indicator is able to reflect small differences in Nitrogen losses between different systems, but it can only be interpreted using a mobile average, not on a yearly basis. This example illustrates how MELODIE could be used to study the dynamic behaviour of the system and the dynamic of nutrient flows. Finally, MELODIE can also be used for comprehensive multi-criterion assessments, and it also constitutes a generic and evolving framework for virtual experimentation on animal farming systems.

Simulating rotational grazing management

Dairy systems predominantly based on rotational grazing are notoriously hard to manage. In order to ensure profitability, this type of production requires quite good organisation, planning, and operating capability on the part of the farmer. A simulation-based decision support system, called SEPATOU, has been developed for this purpose. At the core of the decision support approach lies an explicit and rigorous modelling of the management strategy that underlies a dairy farmers decision-making behaviour (real or hypothetical). The SEPATOU system is a discrete-event simulator that reproduces the day-to-day dynamics of the farmers decision process and the response of the controlled biophysical system for which models of grass growth, animal consumption, and milk production are used. SEPATOU provides the means to evaluate and compare tentative strategies by simulating their application throughout the production season under different hypothetical weather conditions. The relative worth of a strategy can be assessed by analysing the effects on the biophysical system and their variability across the representative range of possible conditions that is considered. The activities to be managed concern the type and amount of conserved feed, where to fertilise and how much, the choice of fields to harvest, and most importantly, which field to graze next. Typically, SEPATOU is designed to be used by extension services and farming system scientists. It is implemented in C++ and is currently undergoing a validation process with the intended users.

Planning with graded nondeterministic actions : A possibilistic approach

This article proposes a framework for planning under uncertainty given a partially known initial state and a set of actions having nondeterministic (disjunctive) effects, some being more possible (normal) than the others. The problem, henceforth called possibilistic planning problem, is represented in an extension of the STRIPS formalism in which the initial state of the world and the graded nondeterministic effects of actions are described by possibility distributions. Two notions of solution plans are introduced: γ‐acceptable plans that lead to a goal state with a certainty greater than a given threshold γ, and optimally safe plans that lead to a goal state with maximal certainty. It is shown that the search of a γ‐acceptable plan amounts to solve a derived planning problem that has only pure (nongraded) nondeterministic actions. A sound and complete partial order planning algorithm, called NDP, has been developed for such classical nondeterministic planning problems. The generation of γ‐acceptable and optimally safe plans is achieved by two sound and complete planning algorithms: POSPLAN that relies on NDP, and POSPLAN* that can be seen as a hierarchical version of POSPLAN. The possibilistic planning framework is illustrated throughout the article by an example in the agronomic domain.

Implications of growth uncertainties associated with climate change for stand management

Climate change is expected to have substantial effects on many aspects of forest ecosystems, including timber production. Temperatures in northern Europe are expected to increase considerably, although there is substantial uncertainty about both the seasonal and average changes that will occur. In Scandinavia, production is predicted to increase across most of the area covered by boreal forest, since the growth of trees in the region is currently limited by temperature. Therefore, we have analyzed the importance of adapting management practices to future climate changes and considered possible ways to address associated stand management problems. For this purpose, we simulated climate scenarios with temperature increases ranging from 2.5 to 6.0°C over a 100-year period, and effects on typical Swedish stands with several species, then optimized their management with simulated annealing. The results indicate that the maximum considered temperature trend would raise the economic value of the stands by almost 5% more than the minimum trend. However, the importance of optimizing management plans in accordance with the correct temperature scenario appears to be limited. The plan optimized for the minimum temperature trend was only marginally inferior to the plan optimized for the maximum temperature trend in the maximum trend scenario, and vice versa. It also seemed adequate to use a deterministic formulation of the problem, and in cases where a stochastic climate change model generated more robust plans, the advantage could be attributed to model artifacts rather than climate change per se.

Possibilistic Planning: Representation and Complexity

A possibilistic approach of planning under uncertainty has been developed recently. It applies to problems in which the initial state is partially known and the actions have graded nondeterministic effects, some being more possible (normal) than the others. The uncertainty on states and effects of actions is represented by possibility distributions. The paper first recalls the essence of possibilitic planning concerning the representational aspects and the plan generation algorithms used to search either plans that lead to a goal state with a certainty greater than a given threshold or optimally safe plans that have maximal certainty to succeed. The computational complexity of possibilistic planning is then studied, showing quite favorable results compared to probabilistic planning.

TiMDPpoly: An Improved Method for Solving Time-Dependent MDPs

We introduce TiMDPpoly , an algorithm designed to solve planning problems with durative actions, under probabilistic uncertainty, in a non-stationary, continuous-time context. Mission planning for autonomous agents such as planetary rovers or unmanned aircrafts often correspond to such time-dependent planning problems. Modeling these problems can be cast through the framework of Time-dependent Markov Decision Processes (TiMDPs). We analyze the TiMDP optimality equations in order to exploit their properties. Then, we focus on the class of piecewise polynomial models in order to approximate TiMDPs, and introduce several algorithmic contributions which lead to the TiMDPpoly algorithm for TiMDPs. Finally, our approach is evaluated on an unmanned aircraft mission planning problem and on an adapted version of the well-known Mars rover domain.

A novel method for the sustainable management of wheat crops: exploration by simulation

Designing crop management strategies to meet both environmental and economic objectives is a growing preoccupation for advisers and researchers, in the move towards sustainable development. We describe here a simple and original method designed to facilitate this task and appropriate field experimentation. This method involves the use of a multi-stage procedure in which a simulation tool is used to explore a decisional space defined by an expert. As a case study, we designed crop management strategies for winter wheat in the Paris Basin. We aimed to maintain a high wheat yield of more than 9 t.ha−1 and grain quality whilst limiting the risks of nitrogen pollution by maintaining post-harvest soil nitrogen content below 30 kg.ha−1. The method was applied on two representative field situations with contrasting soil nitrogen supplies. We used the “DéciBlé” simulation tool to evaluate management strategies, expressed as sets of decision rules, for the possible technical choices such as cultivar, sowing date, sowing density and nitrogen supply. Our step-by-step approach involved progressive limitation of the decisional domain to be explored, and some of the solutions obtained were not intuitive. This method has two main novel features: (1) a simple design dealing with a complex problem, without reduction to a single judgement criterion and (2) results expressed as action plans similar to those implemented by farmers.

Optimising irrigation management at the plot scale to participate at the regional scale water resource management

According to the FAO, a great challenge for the coming decades will be the task of increasing food production to ensure food security for the steadily growing world population. Most of that increase will have to come from intensified agriculture, supported by irrigation. Where irrigated agriculture is developed, water used for irrigation can represent more than 90% of water consumption. In an increasing number of countries, existing resources are fully exploited. Drinking-water supplies and the maintenance of a continuous, minimum water flow in rivers are often given priority over irrigation. Planning and management of water resources have become a very important issue everywhere in the world. In particular, accurate estimation of water demand by agriculture is a key need for water management. Water management includes planning (with decisions on a multiyear time scale, such as building dams), strategic management (seasonal decisions with possible adjustment during the season, such as determining the total water volume used for irrigation), and tactical management (daily decisions, such as releasing water from particular dams). Developing tools by integrating knowledge in models may be of some help for planners. However, modeling water management is complex because it concerns different scales (scale of decision, scale of action, scale of planning, scale of management) and different actors (water users, including farmers, factory managers, and general public, policy makers, and water manager). Crop models that simulate the dynamic of plant growth and water demand of one or several crops can provide quantitative contributions to the environmental impact assessment and be very useful for water management