Integrated assessment of agricultural systems - Seamless

Available online at www.sciencedirect.com

Agricultural Systems 96 (2008) 150–165 www.elsevier.com/locate/agsy

Integrated assessment of agricultural systems – A component-based framework for the European Union (SEAMLESS) Martin K. van Ittersum a,*, Frank Ewert a, Thomas Heckelei b, Jacques Wery c, Johanna Alkan Olsson d, Erling Andersen e, Irina Bezlepkina f, Floor Brouwer g, Marcello Donatelli h, Guillermo Flichman i, Lennart Olsson d, Andrea E. Rizzoli j, Tamme van der Wal k, Jan Erik Wien k, Joost Wolf a a

e

Plant Production Systems, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands b Food and Resource Economics, University of Bonn, Nussallee 21, D-53115 Bonn, Germany c UMR System #1123, SupAgro Cirad Inra, 2 Place Viala 34060 Montpellier, France d Lund University Centre for Sustainability Studies, Lund University Box 117, 221 00 Lund, Sweden FLD, Royal Veterinary and Agricultural University (KVL), Hørsholm Kongevej 11, Hørsholm DK-2970, Denmark f Business Economics, Wageningen University, Hollandseweg 1, 6706 KL Wageningen, The Netherlands g LEI, Wageningen UR, P.O. Box 29703, 2502 LS The Hague, The Netherlands h CRA-ISCI, Via di Corticella 133, 40128 Bologna, Italy i IAMM-CIHEAM, 3191 Route de Mende, 34093 Cedex 5, Montpellier, France j IDSIA-SUPSI, Via Cantonale, Galleria 2, 6928 Manno, Lugano, Switzerland k Alterra, Wageningen UR, P.O. Box 47, 6700 AA Wageningen, The Netherlands Received 10 January 2007; received in revised form 16 July 2007; accepted 17 July 2007 Available online 1 October 2007

Abstract Agricultural systems continuously evolve and are forced to change as a result of a range of global and local driving forces. Agricultural technologies and agricultural, environmental and rural development policies are increasingly designed to contribute to the sustainability of agricultural systems and to enhance contributions of agricultural systems to sustainable development at large. The effectiveness and efficiency of such policies and technological developments in realizing desired contributions could be greatly enhanced if the quality of their ex-ante assessments were improved. Four key challenges and requirements to make research tools more useful for integrated assessment in the European Union were defined in interactions between scientists and the European Commission (EC), i.e., overcoming the gap between micro–macro level analysis, the bias in integrated assessments towards either economic or environmental issues, the poor re-use of models and hindrances in technical linkage of models. Tools for integrated assessment must have multi-scale capabilities and preferably be generic and flexible such that they can deal with a broad variety of policy questions. At the same time, to be useful for scientists, the framework must facilitate state-of-the-art science both on aspects of the agricultural systems and on integration. This paper presents the rationale, design and illustration of a component-based framework for agricultural systems (SEAMLESS Integrated Framework) to assess, ex-ante, agricultural and agri-environmental policies and technologies across a range of scales, from field–farm to region and European Union, as well as some global interactions. We have opted for a framework to link individual model and data components and a software infrastructure that allows a flexible (re-)use and linkage of components. The paper outlines the software infrastructure, indicators and model and data components. The illustrative example assesses effects of a trade liberalisation proposal on EU’s agriculture and indicates how SEAMLESS addresses the four identified challenges for integrated assessment tools, i.e., linking micro and macro analysis, assessing economic, environmental, social and institutional indicators, (re-)using standalone model components for field, farm and market analysis and their conceptual and technical linkage. Ó 2007 Elsevier Ltd. All rights reserved.

*

Corresponding author. Tel.: +31 317 482382; fax: +31 317 484892. E-mail address: [email protected] (M.K. van Ittersum).

0308-521X/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.agsy.2007.07.009

M.K. van Ittersum et al. / Agricultural Systems 96 (2008) 150–165

151

Keywords: Bio-economic farm model; Cropping system model; Impact assessment; Indicators; Market model; Sustainable development

1. Introduction Agricultural systems around the globe continuously change as a result of enlarging trade blocks, globalisation and liberalisation, introduction of novel agro-technologies, changing societal demands and climate change. Parallel to liberalisation of markets, the European Union (EU) has engaged in a political ambition to devise policies that aim to improve sustainability of agricultural systems, i.e., their economic viability, environmental soundness and social acceptability, and to enhance the contribution of agricultural systems to sustainable development of society and ecosystems at large (EC, 2001, 2005b). Agricultural, environmental and rural development policies must contribute to these aims, but in a cost-effective and efficient manner to make them politically acceptable. Strong interactions between policies and adoption of agro-technologies exist. Assessing the strengths and weaknesses of new policies and interactions with agro-technologies, prior to their introduction, i.e., ex-ante integrated assessment, is vital to devise policies that promote sustainable development. The European Commission (EC), for instance, has introduced Impact Assessment of its policies as an essential step in the development and introduction of new policies since 2003 (EC, 2005a). Integrated Assessment has been defined as ‘‘an interdisciplinary and participatory process combining, interpreting and communicating knowledge from diverse scientific disciplines to allow a better understanding of complex phenomena’’ (Rotmans and van Asselt, 1996). Integrated assessment and modelling (IAM) has been proposed by research as a means of enhancing the management of complex systems and to improve integrated assessment (Parson, 1995; Harris, 2002; Parker et al., 2002). It is based on systems analysis as a way to consider, in a balanced integration, the biophysical, economic, social and institutional aspects of a system under study. The assumption underlying IAM is that computerized tools from science contribute to better informed ex-ante integrated assessments of new policies and technologies. They certainly do not replace a participatory process in which many other factors and knowledge sources play a determining role, but allow safe and relatively cheap experimentation, and quantification of effectiveness and efficiency of different policy alternatives. Agricultural science has a history in using systems analysis and what may be characterized as integrative modelling approaches for analyzing bio-economic problems (Heckelei et al., 2001; Kropff et al., 2001; Van Ittersum and Donatelli, 2003; Arfini, 2005; Verburg et al., 2006). However, based on our interactions with potential users of research tools for integrated assessment and studying the literature we argue there are several major challenges to overcome to make research tools more useful for integrated assessment.

First, existing tools, methods and data each cover only some of the hierarchical levels needed within an integrated assessment and in particular do not link the micro (field– farm-small region) and macro (market or sector) levels. This is partly a matter of not bridging scales and partly a matter of lack of interdisciplinarity. Policy questions to be addressed cannot be solved at micro or macro level only, but need cross-scale consideration. Dalgaard et al. (2003) recognized scaling from one hierarchical level to another as a key issue in agro-ecology. Hansen and Jones (2000) describe different methods for upscaling, and Ewert et al. (2006a) address the issue of bridging different hierarchies of scales in natural, economic and social disciplines. Initial attempts to bridge micro and macro analyses focused on developing countries for relatively local markets (Sissoko, 1998; Kruseman, 2000). A crucial challenge is to develop multi-scale methods in general and, more specifically, methodologies that allow bridging analyses at micro and macro scales. Second, the existing IAM approaches are heavily biased towards either the biophysical, economic or social disciplines, and imbalanced in their degree of quantification. Social aspects pertain to employment, income distribution, quality of life of farmers, gender in farming, etc., and are generally not well represented in modelling tools. Furthermore, institutional constraints are often entirely lacking in present integrated research tools. Institutions are defined as the formal and informal rules of a society or of organisations (Spangenberg et al., 2002), that can either facilitate or hamper the decision making, subsequent implementation of policies or use of new technologies, and thus, influence the resulting behaviour of targeted actors (e.g., compliance with regulations, intended behavioural changes). In short, current integrated assessment tools are still restricted in the range of issues they tackle. Third, many of the existing models and databases are currently case specific, restricting their re-use in new problems and their timely availability when new issues arise. Also, their limited re-use is not cost-effective. Although there is an inherent tension between being generic in model formulation and sufficiently meaningful in applications, we think we can largely benefit from the concept of component-based modelling which breaks up larger models in discrete and re-usable components (Szyperski et al., 2002; Argent, 2004). This author advocates the use of component-based modelling (modelling frameworks) in overcoming the tension between the need for good science and the need to be relevant in terms of applications. The fourth obstacle and challenge is related to all three previous challenges and strongly to the third one. Integration of research and the cross-fertilisation of ideas from different disciplines are hindered by the variety of formalisms, which is also reflected in the software tools implementing


the research results. Models are rarely re-usable outside the environments for which they were developed. Over the past two decades various research groups have been trying to exploit integrated modelling in supporting inter-disciplinary research (Rizzoli et al., 1998). This has resulted in the successful examples of model linkage and modelling frameworks, in water- and environmental-resource management modelling (Leavesley et al., 1996), cropping system modelling (ModCom; Hillyer et al., 2003), hydrological modelling (TIME; Rahman et al., 2004 and OpenMI; Gijsbers and Gregersen, 2005). This paper introduces the design and functionality of the SEAMLESS Integrated Framework (SEAMLESS-IF) for an ex-ante, integrated assessment of agro-environmental policies and agro-technological innovations in the European Union (EU). SEAMLESS stands for System for Environmental and Agricultural Modelling; Linking European Science and Society, and brings together over 100 scientists from a broad range of disciplines and 15 countries. The definition of SEAMLESS-IF was driven by the four challenges described above. SEAMLESS aims at delivering an integrated framework to underpin integrated assessment of agricultural systems at multiple scales (from field, farm, region to EU and global), to provide analytical capabilities for environmental, economic, social and institutional aspects of agricultural systems and to develop a component-based system that allows re-use for new problems, while using a software infrastructure that facilitates the re-use and linkage of the components. A vision of the final version of SEAMLESS-IF, due early in 2009, is presented in this paper and illustrated with an example from its first and second working prototypes which are available on www.seamless-ip.org (Ewert et al., 2006b). We realize that the contents and design of research tools are just one determining factor in the usefulness and uptake of such tools in integrated assessment; the process of user engagement and participatory development is probably equally important (McIntosh et al., in preparation). Although the four issues mentioned above are a result of an interactive process of user interaction in relation to impact assessment in the EC, the process of user engagement and participatory development is not a specific subject of this paper.

cessing and packing) and other land uses (e.g., nature). In SEAMLESS, we have conceptualized the system by distinguishing actors (e.g., farmers, policy makers) taking actions which have an effect on the environment (in the broadest sense, i.e., biophysical, economic or social), which results in certain conditions that in turn may affect the actors. Impacts on the institutional environment are not considered but the constraints that institutions may pose on actors and decisions are included in the assessment (see Section 4.2). As such SEAMLESS embodies interdisciplinary, integrated assessment and modelling. SEAMLESS-IF facilitates assessing proposed policy options by comparing a baseline scenario capturing the autonomous trends and already accepted policies with policy scenarios which differ from the baseline in the proposed policies. The scenarios are assessed through a set of indicators that capture the key economic, environmental, social and institutional issues of the questions at stake (Fig. 1). The indicators in turn are assessed using outputs from quantitative model components, typologies and databases. In the final steps (Fig. 1) the indicator values are visualized, aggregated or weighted in an interactive process with users and stakeholders. The framework is designed to compare policy alternatives, in interaction with agro-technological options, for a defined time horizon; these time horizons are defined by the policy questions at stake. The models used within SEAMLESS-IF (Section 3) have some flexibility in terms of time horizons, although the economic market model is currently designed to handle questions with a time horizon of up to 10–15 years from now. Spatially, the framework allows analysis at EU-25 level (and wherever data are available also for the two new member states Bulgaria and Romania, and for Norway, Switzerland, Croatia, Bosnia-Herzegovina, FYR Macedonia, Turkey and Albania), at member state level and administrative units (so-called NUTS-2 regions which, approximately, match provinces within the member states; NUTS stands for Nomenclature of Territorial Units for Statistics in the European Union). Within the NUTS-2 Pre Pre-modelling

2. SEAMLESS integrated framework: domain, approach and users

SEAMLESS focuses on the land-bound agricultural activities (arable cropping, grasslands, livestock, perennials, including orchards, agro-forestry and vineyards) and their interactions with the environment, economy and rural development. SEAMLESS-IF aims at providing analytical capacity to assess sustainability of agricultural systems in the European Union and contributions of the EU’s agricultural systems to sustainable development at large, including some effects on the entire production chain (transport, pro-

Users/stakeholders

2.1. System definition

Problem definition

Scenario description

Indicator development selection

Modelling Definition of simulation experiment

Model selection and composition

Parameterization and simulation

Data and knowledge base

152

Post-modelling Post-model Post analysis

Visualization of results

Documentation/ communication communication

Fig. 1. Integrated assessment procedure using SEAMLESS-IF, with premodelling, modelling and post-modelling phase.


regions agri-environmental zones and farm types are distinguished. At the other end of the spectrum SEAMLESS-IF aims at the assessment of some interactions between the EU and the rest of the world in terms of (agricultural) economy and trade flows. Applicability of SEAMLESS-IF is evaluated and improved in two real-world Policy Cases. Policy Case 1 is driven by global economic policy changes, analysing the impact of further trade liberalisation as currently negotiated in the Doha round of the World Trade Organization. Policy Case 2 analyses what would happen if the EU countries, regions and farmers implement instruments to comply with the EU directives on water and nitrate. The impacts will be assessed with the economic, social and environmental indicators at the various levels represented in SEAMLESS-IF. Specific attention will be paid to the interaction between the various agro-technologies and land uses (such as integrated crop management, conservation agriculture and agro-forestry) and the existence and degree of specific policy incentives to use these technologies. 2.2. Component-based modelling in SEAMLESS-IF Inspired by the approach of modular modelling by Zeigler (1987) and by recent works in component-oriented software engineering (Szyperski et al., 2002) we have opted for an Integrated framework supporting integrated assessment in which Individual (stand-alone) knowledge components can be linked through a software Infrastructure, allowing the use of selected and linked components to underpin integrated assessments. We have named this the Triple I concept. The individual components can be either models representing different processes at specific hierarchical levels, databases, or indicators derived from model outputs and/or data. The software infrastructure of SEAMLESS-IF allows the seamless linkage of selected components into model chains that assess certain indicators. These components have value in their own right and can be either, existing or newly developed models and databases, focusing, for instance, on crop growth and externalities, farm responses or market simulation. The models have been designed to simulate aspects and processes of agricultural systems at specific levels of organisation, i.e., point or field scale, farm, region, EU and world. Linking models designed for different scales and pertaining to different domains is the essential trait of integrated modelling. Such a task requires cross-disciplinary expertise, which is rarely embodied in a single researcher, but it is in the union of the experiences and knowledge of teams of researchers. The software architecture of SEAMLESS-IF allows the re-use of model components for scale and disciplinary integration and their inclusion in model chains for the computation of indicators. 2.3. Users of SEAMLESS-IF SEAMLESS-IF is being designed and developed as a response to a research call from the Directorate General

153

(DG) Research of the European Commission. Frequent interactions with DG Research shaped the requirements and design of the framework (see Section 1 of this paper). Since the start of the project, interactions with various DGs of the EC have been initiated to engage with foreseen users of the framework, in particular DG Agriculture, DG Environment and DG Economics and Finances. User Forum meetings are organized twice a year and smaller meetings in between upon request. We originally distinguished six classes of users in the analysis of requirements for SEAMLESS-IF, i.e., coders, linkers, runners, providers, viewers and players, with distinct requirements. Each of these classes still determine development of the framework but for the development of graphical user interfaces they have been merged into three groups of users, i.e., the policy expert, the integrative modeller and the domain specific modeller. The policy experts are engaged in the pre-modelling and post-modelling phase (Fig. 1) to define scenarios and indicators and are mainly interested in the results of the impact assessment. The integrative modellers are those foreseen to set-up integrated assessments with SEAMLESS-IF, implementing and running model chains which involve model components pertaining to different domains and spanning different scales. They work in tight interaction with the policy expert, to prepare and assess the scenarios which will be studied during an impact assessment procedure. The domain specific modellers are the experts of the individual components and their code or data; they are not being served through a specific graphical user interface. 3. Detailed description of SEAMLESS-IF This section presents the major components of SEAMLESS-IF, the quantitative models, the database and the indicators and the software infrastructure that allows the (re-)use and linkage of these components in applications pertaining to different impact assessments. 3.1. SEAMLESS-IF software infrastructure The SEAMLESS-IF software infrastructure is a result of analyses of technical and user requirements within the project and review of software frameworks which are aimed at supporting a component-oriented approach to modelling (Szyperski et al., 2002; Argent, 2004). A number of environmental modelling frameworks allowing improved model development and deployment and model component linking have inspired the software architecture and design of SEAMLESS-IF (Argent, 2004; Argent and Rizzoli, 2004; Argent et al., 2006). The main philosophy is to allow for (re-)using and linking a variety of available knowledge components, such as models, databases, expert rules and analysis tools while facilitating linking, transparency and documentation through using semantically rich meta-information (Van der Wal et al., 2005). These

154


semantically annotated components might further evolve, while they can be used in the integrated framework at the same time. The components follow a strict separation of data, model and graphical user interfaces. The main component in the SEAMLESS-IF software infrastructure is SeamFrame, the core component that runs on a server and provides the services that can be used by the several SEAMLESS client components/applications. The main components of SeamFrame are: the modelling environment, project manager, processing environment and the domain manager (Fig. 2). The SeamFrame server interacts with the SEAMLESS database and knowledge base. The SEAMLESS ontology plays a central role in SEAMLESS-IF to harmonize and relate the different concepts from models, indicators, source data, etc. SeamFrame uses ontology to structure domain knowledge and semantic meta-information about components in order to facilitate organisation, retrieval and linkage of knowledge in components. To guarantee consistency between the database and the ontology, the domain manager generates the relational database schemas from the ontology. The ontologies and their content are stored in a so-called knowledge base (Rizzoli et al., 2007; Villa and Athanasiadis, 2006; Villa et al., 2006). The use of ontologies to semantically annotate the component models allows, among other things, for checking the match between sources in terms of linking the proper output variables of a component to the input variables of a second component. Modelling environments assist users (domain specific modellers) to develop and edit their executable models and datasets with as current choices MODCOM (Hillyer et al., 2003) for the biophysical models and GAMS (Brooke et al., 2006) for farm economic and market models (Section 3.2.2). Modellers can use a model development

tool to deliver models which will be wrapped up as components thanks to specific wrappers in SeamFrame. All model components implement an interface based on the standard Open Modelling Interface (OpenMI; www.openmi.org; Gijsbers and Gregersen, 2005). OpenMI provides a standardized interface to define, describe and transfer data between software components that run simultaneously or subsequently. The standard is extended to meet the requirements of SEAMLESS (Gijsbers et al., 2006). The model components can then be arranged in model chains by domain specific modellers and integrative modellers. The project manager assist the user in the configuration of the integrated assessment problems: the user is guided in the definition of the problem description, the selection of the indicators and the model chains used to compute them. The project manager also allows to set the model parameters, define which data sets are to be used as inputs and to choose among alternative policy options to be tested and evaluated. The processing environment orchestrates the execution of the experiments associated with an integrated assessment problem: it launches simulations and optimisations within the model components. It is envisaged that in future releases the processing environment will allow to perform sensitivity analyses covering the whole chain of models. On the client side we find a set of rich internet applications (Tidwell, 2006) that have been designed to facilitate the interaction with the user. At present, there are the graphical user interface of the Project Manager (SEAMLESS-IF GUI) that assists the user in the formulation of a project to perform the integrated assessment of alternative policy options, and the graphical user interface of the result browser (Seam:PRES) which enables the user to easily access, display, and compare the results of the policy assessments.

SeamFrame server

SEAMLESS client

Seam:PRES

Domain manager

EXPAMOD

SEAMLESS Ontology

SEAMCAP