1 This s a publication by the Joint Research Centre (JRC), the European Commission’s science and knowledge service. It aims to provide evidence-based scientific support to the European policymaking process. The scientific output expressed does not imply a policy position of the European Commission. Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use that might be made of this publication. For information on the methodology and quality underlying the data used in this publication for which the source is neither Eurostat nor other Commission services, users should contact the referenced source. The designations employed and the presentation of material on the maps do not imply the expression of any opinion whatsoever on the part of the European Union concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries.

EU Science Hub https://ec.europa.eu/jrc

Ispra: European Commission, 2019 © European Union, 2019

The reuse policy of the European Commission is implemented by the Commission Decision 2011/833/EU of 12 December 2011 on the reuse of Commission documents (OJ L 330, 14.12.2011, p. 39). Except otherwise noted, the reuse of this document is authorised under the Creative Commons Attribution 4.0 International (CC BY 4.0) licence (https://creativecommons.org/licenses/by/4.0/). This means that reuse is allowed provided appropriate credit is given and any changes are indicated. For any use or reproduction of photos or other material that is not owned by the EU, permission must be sought directly from the copyright holders.

All content © European Union, 2019

The use of country names in this program does not imply any position by the organizers on the official name or legal status of these countries

2 Table of Contents

Plenary session 1 Use of Global models for policy recommendations, M. Wörsdörfer 11

Parallel sessions 1 Model Transparency and sensitivity analysis The challenge of achieving modelling transparency for policy makers, Rosen R.A. 17 It started with a KISS: making complex modelling accessible, transparent and understandable, Caivano A. et al. 18 Knowing unknowns: adapting uncertainty and sensitivity analysis for impact assessment and policy-making, Becker W. et al. 20 What has global sensitivity analysis ever done for us? A systematic review to inform policymaking through earth system modelling, Pianosi F., Wagener T. 22 Modelling of uncertainty in defining safe operating spaces (SOS) in aquatic systems for support of policy development, Gal G., Ofir E. 23 Scenarios, model linkages and data for policy

Assessing the costs of EU truck CO2 targets with DIONE and VECTO models, J. Krause et al. 27 Energy systems modelling: toolbox development for policy decision support, E. Delarue et al. 29 Macroeconomic effects of EU energy efficiency regulations on household dishwashers, washing machines and washer dryers, P. Rocchi et al. 31 A food safety impact modelling system, Bubbico A. et al. 32 The marine modelling framework of the EU Commission. A earth-system model for supporting policy decisions, Macias D. et al. 36 Complex system modelling and multi-criteria decision making An accounting framework for the biophysical profiling of the European agricultural systems, Cadillo Benalcazar J.J., Renner A., et al. 39 Integrative planning for a sustainable, prosperous future, Chan D., Urantowka W. et al 41 Decision support system for maritime spatial planning within blue growth and ecosystem-based management, Abramic A. et al. 42 Methodological reflections on the European Commission's 2013 use of CBA for air pollution policy, Åström S. 44

3 Parallel sessions 2 Model quality and transparency 1 Challenges and opportunities of integrated policy modelling, Hordijk L., Kancs d'A. 49 Modelling for EU policy support: impact assessments, Acs S. et al. 51 From modelling guidelines to model portfolio management, Arnold T. et al. 52 Post-hoc evaluation of model outcomes tailored to policy support, Thulke H. H. et al. 55 The MIDAS touch - Maintaining an overview of model use by the EC through the EC Modelling Inventory and Knowledge Management System MIDAS, Ostlaender N. et al. 56 Integrated assessment modelling Lessons from the use of the GAINS model to inform clean air policies in Europe, Amman M. 61 Science for an evolved Common Agricultural Policy — the Scenar 2030 experience, M'Barek R. et al. 62 Examples of Europe’s marine ecosystem modelling capability for societal benefit Heymans S. JJ. 64 Multi-Objective Local Environmental Simulator (MOLES): model specification, algorithm design and policy applications, Tikoudis I., Oueslati W. 65 Physiologically based kinetic modelling — a bridge between biological science and public health policy, Paini A., et al. 67 Qualitative and quantitative modelling Integrating qualitative scenarios with quantitative modelling: lessons learned from recent EC foresight projects, Ricci A. 71 Towards absolute measure of poverty in the European Union — A pilot study Cseres-Gergely Z., Menyhert B. 72 The Global Conflict Risk Index (GCRI), Halkia M. et al. 74 Modelling stakeholders knowledge and requirements on soil multifunctionality across EU, Bampa F. et al. 75 The politics of models: socio-political discourses in the modelling of energy transition and transnational trade policies, Capari L., Udrea T. et al. 76

4 Parallel sessions 3

Scenarios, model linkages and data for policy 2 Combining micro and macro simulations to assess the distributional impacts of energy transitions. Evidences from the French national low carbon strategy, Ravigné E. et al. 81 Energy security assessment framework to support energy policy decisions, Augutis J. et al. 83 Decarbonising transports in Portugal up to 2050: possible pathways, Tente H et al. 85 Modelling the macroeconomic and employment impacts of future mobility disruption scenarios, Tamba M. 86 Global trends of methane emissions and their impacts on ozone concentrations, Van Dingenen R. et al. 88 Integrated modelling: the case of the food-water-energy-environmental NEXUS management under uncertainty and risks A strategic decision–support system for strategic robust adaptation to climate change and systemic risks in land use systems: Stochastic integrated assessment GLOBIOM model, Ermolieva T. 93 An integrated environmental-economic model for robust pollution control under uncertainty, Wildermeersch M. et al. 94 Policy impact on biofuel or electricity production from biomass in Europe, Leduc S. et al. 95 Robust food-energy-water-environmental security management: linking distributed sectorial and regional models, Ermoliev T. et al. 96 Modelling value chains in global trade Challenges of conducting policy impact analysis using PE models: the case of Brexit, Nti F., Jones K. 101 Shooting oneself in the foot? Trade war and global value chains, Bellora C., Fontagné L. 102 MAGNET — a team-based modular CGE approach for coherent cross-cutting policy assessments, Kuiper M. et al. 103 Environmentally conscious transportation and logistics modelling for agri-food supply chains, De A. et al. 105 Modelling fairness in FVCs: developing quantitative indicators, Gudbrandsdottir I. Y. et al. 109

5 Poster session 1 A global sensitivity analysis of smart grids project cost benefits analysis with correlated inputs, Vitiello S. et al. 115 A participatory process to design regional policy for rural development: the case of the Veneto Region, Trestini S. et al. 116 Defining the applicability of animal-free approaches for skin allergy testing of chemicals, Asturiol D. et al. 117 Employment effect of innovation, Kancs d'A., Siliverstovs B. 119 Forecasting the teaching workforce in Lithuania, Leiputė B et al. 121 Global governance challenges: a case for big modelling, Toth T., Theodoropoulos G. 123 Modelling European economy as an ecosystem of contracts for smart policy making Telesca L et al. 124 Modelling the social impact of open access knowledge repositories, Skulimowski A. M. J. 125 PIRAMID: a new methodology to build baselines for CGE models, Wojtowicz K. et al. 128

Plenary session 2 Assessing social impacts of policies: indicators and methods, Jacob K. 133

Parallel sessions 4

Complex system modelling & multi-criteria decision-making 2 Making sustainability models more robust: dealing with the complexity of the metabolic pattern of social-ecological systems, Giampietro M. 137 Supporting result-based schemes. The case of ex-post assessment of agri-environmental- climatic schemes, Bartolini F. et al. 139 Exploring the (non-)use and influence of models in Belgian climate policy-making: a multiple case study, Fransolet A. 140 Assessing the potential of machine learning algorithms for agent-based models from an innovation policy perspective, Müller M. et al. 142 Machine learning in the service of policy targeting: the case of public credit guarantees, de Blasio G. et al. 143

Engaging stakeholders and policy makers Complex modelling to achieve carbon neutrality in Portugal, Seixas J. 149

6 System dynamics modelling as strategic vision for transforming adult social care in Northern Ireland: from primary research to procurement, Wylie S., Martin J. 150 The use of models to support the design and implementation of the EU Cohesion Policy, Monfort P. 152 Measuring and explaining the EU's Effect on national climate performance Avrami L., Sprinz D. F. 154 The importance of modelling purpose for policy, Edmonds B. 155

Modelling support for EU policy design: the EU's Long Term Strategy for climate Developing a global context for regional 1.5°C scenarios, Keramidas K. et al. 159 Modelling carbon neutral energy systems, Zazias G. et al. 160

Costs and potentials for reducing non-CO2 gases, Höglund-Isaksson L. et al. 162 The land use sector as key to offset residual emissions by 2050, Frank S. et al. 164 Socio-economic aspects of EU decarbonisation, Weitzel M. et al. 166

Parallel sessions 5

Model quality and transparency 2 Map or compass? Improving model contribution to long-term climate strategy. Insights from 4 country studies, Lecocq F. et al. 171 Modelling climate-energy linkages in national energy and climate plans: achievements and challenges, Boromisa A. 172 Analysis of current climate and energy policies: are countries on track to meet their NDC targets?, den Elzen M et al. 174 How to make social-environmental modelling more useful to support policy? Dreßler G. et al. 176 Quantitative pest risk assessment — a modelling methodology supporting EU plant health policy, Bragard C. et al. 178

Scenarios, model linkages and data for policy 3 Strategic sourcing 2.0: improving fiscal efficiency using big data, Borges de Oliveira A., et al. 185 Uses and potential of large-scale, micro level government contracting datasets for policy modelling, Fazekas M. et al. 186 Using synthetic control methods for policy evaluation: modelling challenges, new data, and evidence from EU carbon markets, Bayer P., Aklin M. 187 7 Combining graphical and command line user interfaces for economic analysis: the capriR package for processing, visualizing and analysing large sets of simulation results, Himics M. 189 Wildfire modelling for adaptation policy options in Europe, Krasovskii A. et al. 191

Computational and empirical issues in dynamic economy-wide models Using the Global Multi-Country model in ECFIN’s forecast exercises, Calés L. et al. 195 Stationary rational bubbles in non-linear macroeconomic policy models, Kollmann R. 197 The macroeconomic and sectoral impact of EU competition policy, Cai M. et al. 198 Direct and indirect impacts of European Institute of Innovation and Technology (EIT) investments on regional economic growth, Ivanova O. et al. 199 Modelling the labour market and agricultural sector in an integrated CGE framework: an application to the Brexit case, Angioloni S. et al. 200

Poster session 2 A coupled modelling system for the integrated assessment of continental freshwaters, coastal and marine waters, Friedland R. et al. 205 A platform to manage air quality in Europe with the use of high-resolution modelling tools. Application and validation in Krakow, Vranckx S. et al. 205 Assessing the value of regional cooperation in air pollution control, De Angelis E. et al. 207 Biomass in the European Electricity System: Emission Targets and Investment Preferences, Lehtveer M. et al. 211 Easy-to-use modelling tool for urban air pollution with very high resolution HPC simulations, Horváth Z. et al. 213 MAQ, an integrated assessment model to support air quality policy at regional scale, Turrini E. et al. 215 Modelling the renewable transition: strategies and policies for reducing energetic and raw materials costs, Samsó R. et al. 218 On the sense of applying 'Frankenstein monster' models, Kremers H. 220 Policy failure in the field of electro-mobility, Gómez Vilchez J. 221

The CO2MPAS vehicle simulation model and its role in the European light-duty

Vehicle CO2 certification process, Pavlovic J. et al. 226

8 Plenary session 1 Room 0.A 26 November 10:00 – 11:00

9 10 Use of global models for policy been enhanced for a special regional focus recommendations featured in the freshly released World Energy Outlook 2019. The IEA thereby Mechthild Wörsdörfer, Director of entered unchartered territory in developing Sustainability, Technology and Outlooks, detailed country-level models for the African International Energy Agency continent, which is often neglected in terms The 'EU Conference on Modelling for Policy of data and analysis. The IEA operates support: Experiences, challenges and the numerous detailed technical models which way ahead' in November 2019 in Brussels cover individual sectors in great granularity. is the first time the EU Competence Centre These models form the basis for our on Modelling CC-MOD hosts a large-scale publications in the area of energy technology conference. I appreciate the aim of bringing policies. This year, IEA enhanced its modelling together researchers and policy-makers of hydrogen as well as geospatial analysis involved in modelling activities. A broad and of offshore wind for the release of special diverse range of conference participants are reports. A finer temporal resolution with expected from European and international hourly analysis of storage and demand-side institutions and agencies, Member States, response was a key model improvement for Universities, research institutes, and the World Energy Outlook 2019. Altogether, consultancies. its broad modelling activities put the IEA in a The conference will provide helpful in position as acknowledged and authoritative promoting exchange between modellers agency able to give advice on various and policy-makers and thereby improve energy policy areas covering all fuels and all understanding between professional technologies in the world. With its experience stakeholders. Participants of the conference in using modelling for governmental advise, convene to learn from each other’s models IEA is happy to share its model-specific and to act as sounding board to each other. expertise and knowledge with stakeholders. Exchange on state-of-the-art approaches to Various trainings and enhanced model and modelling is prerequisite for the community data transparency prove the agency’s role to develop common principles, to compare as knowledge provider and its willingness its approaches, to resolve challenges and to to engage into dialogue with the modelling ensure quality standards and progress in the community. In fact, IEA proved its key role advancement of model development. Not in promoting exchange within the modelling the least, it is important that modellers do community when hosting the 2019 edition test their model results in direct exchange of the International Energy Workshop IEW with policy-makers in order to ensure in June 2019. The CC-MOD conference in applicability in the real-world context. Policy Brussels provides for a good opportunity recommendations can be tested against for mutual exchange and sharing of best sanity checks by practitioners and they practices and the IEA is honoured to take part can be tailored to increase their chances of in this endeavour. realization in the political process. Challenges lie ahead. The model community The IEA has long-standing experience in needs to follow numerous developments. the use of large-scale models for policies. First and foremost, digitalisation and Its flagship publication, the ‘World Energy advances in IT landscape set the scene. Outlook’, builds on a comprehensive The pace of technological progress is both a energy model covering global long-term curse and a blessing for policy modelling. On developments of all fuels. The model ensures the one hand, advances in the IT landscape sophisticated techno-economic detail. It increase computational power and speed. allows for concrete policy recommendations New visualisation software allow for to IEA members and beyond based on unprecedented ease in producing graphics bespoke modelling efforts. For example, and interactive tools. Usage and sharing of modelling of the continent of Africa has large-scale databases has become easier

11 than ever before. On the other hand, the Many stakeholders in the modelling speed of change requires constant adaptation community face a similar challenge as the and flexibility. Tools which were state-of-the- IEA. In today’s fast-paced media world, art just years ago may be obsolete by now. it is vital for communication to be timely, Skills of today may be rendered useless in the easy to understand and appealing to a near future, if no constant learning is ensured. broad audience. To get its messages heard The IEA, with its established long-standing by a wide audience, the IEA is constantly modelling culture, copes with these constant embracing innovative visualisation techniques. challenges by continuous adaptation of its New graphing tools and interactive features large-scale models over time. Enhancements provide for powerful ways of convincing policy to its models are made in an evolutionary makers and practitioners. They have become manner, maintaining the good work of standard in conveying policy messages for the past and ensuring high-quality output striving institutions with high ambitions. The while equally benefitting from technological IEA is currently revamping its communicative improvements. Where small models can be tools in order to strengthen its outreach. overhauled easily and re-built from scratch, The new IEA logo is at the start of a longer large-scale modelling tools require a more campaign to secure and expand the IEA’s role careful approach for model development and as a leading authority in energy policy. A new resource planning. website follows suite and will be released soon. New interactive features on the IEA Another major pillar for modelling website allow for better readability and make developments is transparency. The it easier to convey messages to the target European Commission was in the past often audience with a broad range from laymen to reproached for not sharing full details and practitioners and policy experts. assumptions of its modelling efforts on member state level. I noted this during my Model linkages are increasingly becoming time at the European institutions. Similarly, key features in ensuring consistency between IEA receives requests for sharing the entirety models and improving the understanding of underlying model data and assumptions. of the energy system. To which extent Striking the right balance between open- should models and modules be integrated? source and proprietary or confidential As a large agency, IEA operates different information is a question the whole policy independent models. A big challenge is to modelling community faces. Stakeholders ensure consistency across models for data, have interest in understanding models and output and key results. Ideally, consistency they need to understand underlying data. is guaranteed through hard links between The IEA recognises these interests and it is modules, but this is not always technically increasingly sharing its model results and possible. Often, models are aligned manually data in a transparent and open format free through comparing results or use ‘soft links’ of charge. The World Energy Outlook 2019 such as standardized exchange of input data. website offers unprecedented amounts of key Most recently, IEA enhanced its modelling assumptions without paywall. Nevertheless, of hydrogen. As hydrogen is relevant for the core of IEA’s models and some of the many sectors, technologies and regions, its modelling results remain either proprietary modelling is an example of a cross-agency or they can be accessed at reasonable cost. effort of model alignment. Interlinkages are Economic considerations lie behind these omnipresent and they have to be reflected restrictions. Modelling is costly and it needs in modelling efforts when the ambition is to to be financed – most conference participants cover system dynamics. At IEA, we are proud will understand. to operate energy system models which span across all fuels, all technologies, all sectors Communication and visualisation and all countries. The ambition is to reflect techniques are a very decisive area to the energy system in its entirety, without ensure success in the transformation of necessarily giving up on detail. There is no modelling results into real-world policies.

12 blueprint approach to get the optimum of challenges we are facing at IEA are shared models in terms of scope and detail. Surely, by many of conference participants. many conference participants will deal with Therefore, it is good to engage into a fruitful this dilemma. exchange between stakeholder groups at the November 2019 conference in Brussels The IEA is happy to share its experience in and I look forward to sharing my learnings of modelling and policy advisory at the CC/ my time as Director at IEA with previous EU MOD conference in Brussels. The numerous experience.

13 14 Parallel sessions 1

Model transparency and sensitivity analysis

Room 0.A 26 November 11:30 – 13:00

15 16 The challenge of achieving 3. The values of other input assumptions/ modelling transparency for policy coefficients that are not statistically estimated based on historical data makers are only partially supplied in model Rosen R. A., Tellus Instute documentation, even when these input Many types of integrated assessment assumptions are key to understanding models (IAMs) are utilized to facilitate policy the model results. Also, since many input making world-wide, including in the EU. assumptions are often projected for many This is especially true for policy assessment years into the future, it is often not clear research on environmental topics, including what methodologies have been utilized to the mitigation of climate change. In cases make those projections. that are all too common, these models are 4. The results of IAM runs are often overly really just 'black boxes' for policy makers, and aggregated in the published literature so even for other research analysts in the same that detailed aspects of the results that field of inquiry. There are at least four major might help reviewers better understand aspects of the lack of transparency of most the reasonableness and implications of IAMs: the results are not available. Usually, 1. The equations – sometimes a few of there is little evidence as to whether even the key equations are listed in fairly the model developers and users look abstract form, but rarely is a complete at detailed model results to help them list of all the substantive equations determine if the results are reasonable, ever provided in model documentation. counter intuitive, or just plain wrong, Furthermore, model documentation rarely since there is almost never any detailed describes how the equations are solved, discussion of model results presented. even though the solution technique can sometimes matter significantly. Most policy makers are probably aware of at The solution techniques are especially least some of the problems with transparency important to understand since most IAMs listed above. But what can realistically are highly non-linear, and yet need to be be done to solve these problems so that solved over a period of decades which can policy makers can have more confidence in introduce cumulative errors in the solution the modelling results on which they would by the end of the solution time period. like to rely? Model developers, of course, generally do not want to address the above 2. The coefficients within the equations list of problems because it would take many are almost never explained or defined research hours to do so, and would subtract clearly, and the way in which they are from their research budgets for doing the derived from historical data (or theoretical type of work they enjoy most, which is assumptions) is never explained. Even probably writing up their results for peer- their numerical values are usually not reviewed journals. provided. Further, model documentation often does not even describe the time This implies that policy makers and journal period over which historical data used to editors need to re-think the entire process estimate the coefficients. Obviously, for of peer review in light of the four types policy makers to have any confidence that of transparency issues listed above. For the IAM in question is up-to-date, they example, what good is claiming that a need to know the end-point of this time proposed journal article has been peer- period. Finally, the likely accuracy of the reviewed if most aspects of the model itself historical data relied on is usually never cannot be peer-reviewed for the current discussed, nor is the issue of whether submission, and have never been peer these historically-based coefficients reviewed. Clearly, good science, as well as should be changed in the future in model good economics, cannot be done on a 'trust applications. me' basis. At some fairly recent point in the 17 history of the development of an IAM the documentation available for the kinds of model itself must have received a thorough integrated assessment models typically peer review, otherwise what is the value of relied on in IPCC Working Group III and similar just reviewing the write-up of various model reports. The author has also been a peer- results in a proposed journal article? Thus, reviewer for various journals of proposed the EU Competence Centre on Modelling must articles that rely on IAMs, and is a member of work with model developers, model users, The Integrated Assessment Society scientific colleagues in each research field, peer reviewers, journal editors, and report It started with a KISS: making editors to enhance the entire peer-review complex modelling accessible, concept and practice for IAM-based research. transparent and understandable Obviously, however, this first requires that Caivano A., M'Barek R., Ferrari E., European model documentation in the future must Commission, Joint Research Centre include all the four ingredients listed above. 1) Modelling in the policy and research life This implies that research budgets must cycles be explicitly set aside for doing all of the additional work required to achieve this higher Model-based analysis supports the European level of model documentation. Furthermore, Commission in impact assessments and all IAM-based research papers must be analysis of policy options. In the area of accompanied with clear tables of all key agriculture policy analysis and related areas, input assumptions used for each model run/ modelling constitutes a key component for scenario relied on in that paper. Doing this evidence-based policy making. will be a big help to the readers and peer reviewers of those articles for understanding The modelling work belongs to the the significance of each scenario or model 'formulation' phase of the policy cycle, where result. the impact assessments are located. For a transparent policy-making, this process has to For example, it would be relatively easy follow highest standards to allow for tracing when examining mitigation scenarios for back all decision making. climate change to understand why one model or scenario has much nuclear power Therefore, the modelling for policy support for producing electricity in the future than has to include the full step of the research another, if the former scenario assumes life cycle, which does not stop with the much lower nuclear capital costs than the publishing of a report. Instead, it has to latter. Yet, IPCC and other reports rarely ensure open access, dissemination within provide policy makers with these kinds of social media (e.g. making understand metrics key input assumptions for each scenario through visualization) and preservation of presented in their reports. If improved model/ the information in sustainable formats and scenario transparency were achieved, then reliable storage. a policy maker would know which scenario To this end, the Joint Research Centre (JRC) to focus on depending on their view of of the European Commission has developed the likely values of the input assumptions. DataM as a tool which provides interactive Without knowing the key input assumptions dashboards and raw datasets resulting from for each scenario considered for inclusion in the scientific activities of JRC and partners, a policy assessment report, policy makers relating in particular to the economics are incapable of coming to any rational aspects of agriculture and sustainable and scientifically based conclusions as resources. Thus, it is a complement to to appropriate policies to recommend to scientific publications, aiming to improve governments. the usability of traditional scientific reports The analysis in this paper is based on the largely based on big and complex data author’s extensive review of the limited outcomes. DataM provides the readers with 18 on-line tools that enable the self-analysis of download data whereas modelers can also data and allows full accessibility and storage. upload results by managing version, meta- data, reference data and harmonization 2) The challenge of explaining model logics. output ii) a business intelligence engine for Support to policy with (economic) modelling interactive infographics and dashboards. tools has to ensure that the outcome is i) The infographics include some narrative to accessible, ii) transparent, iii) traceable, and resume the results of studies presented in iv) understandable. Without saying, it has to interactive way with a top-down approach provide results and recommendations in a (from generic concepts to details). The timely manner and with high scientific quality. dashboards are straightforward screens based on charts, maps, tables and filters DataM1 (Data portal of agro-economics interrelated among them. research) was built by the JRC as a tool for responding to these challenges. It provides DataM belongs to the acknowledged tools meta-information on the models, including within the section 'EC knowledge, publication, links to documentation in order to allow tools and data platform' of the internet portal a precise understanding of each model’s of the European Commission. specification. Graphical user-interfaces offer flexibility in supporting the appropriate use 4) Example 1: Free trade agreements study of models and their outputs by different user types (e.g. 'viewers', 'users' or 'developers'). A good example is the 2016 study on 'Cumulative economic impact of future trade 3) DataM in details agreements on EU agriculture'. The study, announced by Commissioner Hogan at the The concept at the basis of DataM is to Agriculture Council meeting and published exploit the web plus the recent business (15 November 2016). Models used: MAGNET intelligence technologies to improve the and AGLINK Interactive dashboards, data to usability of our scientific literature. Scientific be downloaded through a query portal or as articles, especially in the case of economic a bulk. studies relying on modelling exercise, are largely based on big and complex data The dashboards are actually short versions of outcomes. the report with interactive visualisations.

Users take great advantage from the usage In 2019 the negotiations on an EU- of on-line tools to self-analyze model MERCOSUR free trade deal gained outcomes from personal perspectives. This momentum and were closed at the Osaka is a paradigm shift as compared to the G20 meeting 29 June 2019. Analysts and traditional scientific articles that show only press did consult the report intensively. some charts and tables in accord to authors' choices. 5) Example 2: BioSAMs and SAMs

This is also a significant improvement as The tool is also supporting the storing compared to the simple provision of complex and visualization of analytical (used for raw data outcomes, not accompanied by any modelling purposes) databases such as tool and guidance to interact efficiently with Social Accounting Matrices (SAMs), which are them. The site is based on: typically employed to calibrate Computable General Equilibrium models. A SAM is a i) a data-warehouse where we manage comprehensive and economy-wide database datasets, often linked by the JRC open data recording data on all transactions between catalogue, the EU open data portal and the economic agents in an economy over a European data portal. Normal users can only certain period of time. SAMs are large

1 https://datam.jrc.ec.europa.eu databases which include national account 19 data plus a series of micro data used to However, even the most advanced models disaggregate the economy at stake to the are approximations to reality, which means level of details needed for the analysis. that the results are inherently uncertain. That DataM is currently storing a EU28 plus all does not mean that models are not useful; Member States' Social Accounting Matrices properly done, a model represents our best with a detailed disaggregation of the bio- understanding of how a system behaves. economy (Mainar Causapé et al., 2018). The Nevertheless, depending on the context, our tool allows researchers to download all the best predictions may still be very uncertain. available SAMs which can be used in any This is particularly the case when forecasting single-country modelling approach or to long-term impacts or modelling complex and visualize the key indicators included within poorly-understood systems, as is often the the database in a friendly manner. case in policy-making.

Based on these sets of SAMS, DataM In this context, it is essential that uncertainty produced an interactive tool to provide the is acknowledged, and quantified to the extent number of jobs that would be generated possible. To do otherwise risks important by an exogenous shock in final demand decisions being made on evidence that may for the selected commodities. This number not be very reliable, which ultimately can lead accounts for direct, indirect and induced to poor policy-making. effects, calculated after (an infinite) feedback effects. The tool shows the variation in job The practice of quantifying the uncertainty creation in each of the sectors shocked and in the results of a model is called uncertainty the aggregate variation (total jobs, jobs in analysis. Uncertainty analysis involves the main productive sector of the commodity, identifying and quantifying the sources of jobs in the other sectors). uncertainty in a model, such as assumptions and input parameters, and propagating their In addition, DataM is also storing and provide effect to the model output. visualization tool for SAMs, developed within the JRC food security projects, for Sub- Sensitivity analysis (SA) is a related and Saharan African countries. So far Kenya and complementary discipline which apportions Senegal SAMs are include while Ethiopia is the uncertainty in the model results to each about to be uploaded. of the sources. Sensitivity analysis is often used to mean both uncertainty and sensitivity Knowing unknowns: adapting analysis, since in most cases an uncertainty uncertainty and sensitivity analysis is a pre-requisite. analysis for impact assessment Sensitivity analysis in impact assessments and policy-making Recognising that these are important Becker W., Rosati R., Albrecht D., European ingredients in model-based studies, European Commission, Joint Research Centre Commission (EC) guidelines on impact Uncertainty in modelling assessments (IAs) have recommended uncertainty and sensitivity analysis since at With the increase of computing power and least 2009. abundance of data, mathematical models have become increasingly prominent tools This work aims to investigate to what extent in policy-making. Models give the possibility sensitivity analysis is practiced in EC impact to quantify the impacts of alternative policy assessments. Using a combination of text options, the results of which can be used to mining and reviewing nearly 500 IAs over the support decision making. Models have also period 2011–2018, we find that the uptake of become increasingly complex, as they aim to sensitivity analysis has been slow, although model economic, physical and social systems recent years show encouraging signs of with ever-higher resolution. improvement. Most impact assessments

20 still do not include a sensitivity analysis, skilled in sensitivity analysis, but 'advanced including many that are based on modelling practitioners' could be trained who can or quantitative analysis. Moreover, sensitivity act as advisors within each DG. A network analyses are often performed in a manner of practitioners could be established via that can greatly underestimate uncertainty, the Modelling Community of Practice, for by varying one assumption at a time. example, centred on the JRC’s dedicated research/support group on sensitivity analysis. This problem is by no means unique to the Sensitivity analyses could also be recorded as EC. A recent review has shown that even in part of the MIDAS model database to provide top-cited academic literature, the occurrence best practice examples and help future IAs. of a statistically-sound sensitivity analysis is relatively uncommon. What causes this gap At the systemic level, addressing all parts of between recommendation and practice? the impact assessment chain could help a shift towards a more 'uncertainty-conscious' Problem identification culture. Policy makers could be given crash courses in awareness and decision- In our opinion, a number of factors contribute making under uncertainty. Quality control to the problem. Among the main ones is procedures (such as the Regulatory Scrutiny a lack of time: policy officers are under Board) could tighten the requirements for considerable time pressure to complete the sensitivity analysis. Models and data should many tasks associated with an IA, and SA also be made transparent, which encourages can be a time-consuming exercise. This may responsible modelling. Reviews could also often be coupled with a lack of the high-level be performed which retrospectively analyse knowledge and statistical/software skills how successful policy modelling was from required to perform a sound SA. Even for previous impact assessments, to learn those with a technical background, sensitivity lessons for future modelling endeavours. This analysis can still be a challenging subject in would also add an element of accountability some circumstances. into the modelling process. The personal difficulties faced by policy Finally, experts in sensitivity analysis officers are complemented by wider systemic should recognise that modelling in impact issues. There seems to be a lack of best assessment occupies a particular niche practice examples, and low incentives. which requires special considerations. This may point to a wider issue, which is Guidelines and training for sensitivity analysis that uncertainty in modelling is under- often tend to be either over-simplified or appreciated, and that there may be unrealistic intimidatingly technical. Like most parts of expectations of what models can reliably impact assessment, sensitivity analysis can predict. This puts pressure on modellers and be applied proportionately to the complexity analysts to produce projections and forecasts and importance of the policy. One route could in cases where it is infeasible, and to be to introduce a bronze/silver/gold system, understate or neglect uncertainties. On top of which gives guidelines to follow depending on this, typical modelling in IAs aims to project the requirements of the impact assessment. the impacts of policies many years into the For example, a low-impact (bronze) policy future, so there is little or no accountability proposal that relies very little on modelling for unreliable forecasts. might only require best-case and worst- Solutions case scenarios, and a candid discussion of uncertainty. In the medium (silver) case, an How can we address these problems? At the impact assessment which uses a cost-benefit personal level, training is an obvious avenue analysis and some basic projections should to improve skills, and reduce the time taken at least investigate a range of scenarios to perform SA. Not all staff involved in corresponding to varying key assumptions impact assessments can be expected to be simultaneously, and include a discussion of

21 wider untestable assumptions. A full (gold) approaches should be used to fully explore sensitivity analysis would require a global the impacts of input uncertainties on model Monte Carlo exploration of uncertainties, and predictions. Global Sensitivity Analysis (GSA) an analysis of the reliability of the modelling: is a set of statistical techniques that provides this ought to be appropriate when large such a structured approach (Saltelli et al., physical/economic models are used for high- 2008). GSA has the potential to massively stakes policy evidence. The idea of giving advance the value of computer models in different 'levels' of sensitivity analysis would earth system sciences, contributing to more aim to direct the effort to where it is needed robust development, evaluation and decision- and to not impose unrealistic expectations in making. However, the application of GSA in every case. many areas of earth system sciences is still relatively limited. We recently published a What has global sensitivity literature review to demonstrate the value analysis ever done for us? A of GSA by showing examples of what its systematic review to inform application has achieved so far for scientists, modellers and policy-makers (Wagener and policymaking through earth Pianosi, 2019). We organised our review into system modelling 10 generic lessons learnt. In this presentation, Pianosi F., Wagener T., Department of Civil we will focus on 3 lessons that are Engineering and Cabot Institute, University of particularly relevant for model-based policy Bristol making, and draw examples from our own Computer models have become essential recent work to demonstrate the point. tools in earth system sciences, improving our 1. Consistency of model behaviour with the understanding of earth system functioning underlying perceptual model of the system and informing policy-making at various is as important as the ability to reproduce spatial and temporal scales. A key challenge observations in the development of computer models is that, even when they represent a relatively The predominant approach to model low number of physical processes, they evaluation still largely relies on the can quickly exhibit complicated behaviours comparison of model predictions to because of the high level of interactions observations. However, even if model between their variables and components. As predictions reasonably fit observations, the level of interactions increases, modellers this might be a relatively fragile result. quickly lose the ability to anticipate and Indeed, the value of the fit-to-data criterion interpret model behaviour and hence to may be undermined by the large (and evaluate that a model achieves 'the right typically poorly known) errors that affect response for the right reason'. The issue is environmental data; the scarcity of the data made more problematic in earth system themselves, which often do not represent modelling where incomplete knowledge of the entire range of system conditions; or the the system, and the scarcity and noisiness of unrepresentativeness of historical data when data, makes it impossible to 'validate' models dealing with changing boundary (e.g. climate) simply based on their fit to observations. Lack or system (e.g. land use) conditions. Here, GSA of transparency about the scope of validity, can be used to understand how the model the limitations and the predictive uncertainty represents system controls, and how such of earth system computer models, can lead controls might change in the future, which is users to overestimate the model’s predictive crucial and sometimes more important than ability or, on the opposite hand, induce them the model’s ability to reproduce historical to rejecting the model completely. observations (e.g. Sarrazin et al., 2019). In order to tackle the above problems, more structured, transparent and comprehensive

22 2. If model predictions are expected to makers can further complement these results support decision-making, then they have with other sources of information to assess to be sensitive to decision-related input how likely those input thresholds are to be factors crossed in the future and hence determine whether actions may be required (e.g. Earth system models are often used as tools Almeida et al., 2017). to support decision-making, by assessing and comparing the effects of different decisions References (which can be related to one or more model's input factors) on an output of interest to the Almeida, S., Holcombe, E. A., Pianosi, F., decision-makers. In this context, an implicit and Wagener, T. (2017), Dealing with deep assumption is that the decisionrelated input uncertainties in landslide modelling for factors exert an influence on the output disaster risk reduction under climate change, that is at least comparable to that of other Nat. Hazards Earth Syst. Sci., 17. uncontrolled factors, such as the model Butler, M., Reed, P., Fisher-Vanden, K., Keller, parameters or forcing inputs. While this K. and Wagener, T. 2014. Inaction and climate influence might be present in the real world, stabilization uncertainties threaten a fiscal one cannot take for granted that it is also cliff. Climate Change Letters, 127. present in the model. Indeed, models built for supporting decision-making typically Saltelli, A., M. Ratto, T. Andres, F. Campolongo, integrate a range of interacting and often J. Cariboni, D. Gatelli, M. Saisana, S. Tarantola, nonlinear components, which means that (2008), Global Sensitivity Analysis. The Primer. their responses to variations across their Wiley many input factors are not immediately obvious. Here GSA can help quantifying the Sarrazin, F., Hartmann, A., Pianosi, F., relative importance of all these factors and Rosolem, R. & Wagener, T. (2018). V2Karst verify whether predictions are sensitive to V1.1: a parsimonious large-scale integrated decision-related ones, which is a necessary vegetation–recharge model to simulate prerequisite for the model to adequately the impact of climate and land cover support decision-making (e.g. Butler et al., change in karst regions. Geoscientific Model 2014). Development, 11.

3. Even in the presence of practically Wagener, T. & Pianosi, F. (2019). What has unbounded uncertainties, learning about Global Sensitivity Analysis ever done for us? the relationship between model controls A systematic review to support scientific and outputs can be relevant for decision- advancement and to inform policy-making making in earth system modelling. Earth-Science Reviews. 194. Another area where GSA has been successfully employed is the investigation of Modelling of uncertainty in so called ‘deep uncertainties’, i.e. input factors defining safe operating spaces whose ranges of variability and probability distributions are poorly known and hence (SOS) in aquatic systems for practically unbounded. The propagation support of policy development of these uncertainties through a model is Gal G., Ofir E., Israel Oceanographic and technically feasible, however the resulting Limnological Research predictions are typically spread over such Efficient policy for the successful wide ranges that they are hardly usable to management of aquatic ecosystems requires directly inform decision makers. Here, GSA balancing between the need to sustain can be used to analyse model simulations the ecosystem over time and the need to and identify thresholds in the input factors provide, often conflicting, ecosystem services. that, if exceeded, would cause the output to Varying environmental conditions, and often cross undesired output thresholds. Decision- 23 large degrees of uncertainty, may hamper ecosystem service. In addition, I account for the development, and implementation, of the two types of uncertainty in defining the successful policies for the management of range of options for policy makers. aquatic. One approach to the merging of the objective of sustaining the ecosystem along The first case study includes the application with the continuous provision of ecosystem of the food-web suite of models, Ecopath services is the definition of safe operating with Ecosim (EwE) and Ecospace, in order spaces (SOS). A SOS in essence defines the to define a commercial fishery’s safe range of management measures that can be operating space (SOS). In order to address applied to the ecosystem in question while the difficulties associated with management maintaining the ecosystem state variables under uncertainty I include uncertainty and within pre-defined ranges thereby sustaining deep uncertainty in the modelling process. the ecosystem over time. The ranges defined I define the SOS based on varying levels of by a SOS, therefore, provides the much- fishing efforts, lake levels, and submerged needed information for policy development. vegetation all based on historical ranges. I do so by running multiple, long term scenarios, There is, however, a large degree of using EwE and Ecospace models developed uncertainty associated with future conditions for the lake. I then incorporate future of aquatic ecosystems and the environmental uncertainty in the abundance of various food drivers acting upon them thereby hindering web components and evaluate the impact successful policy development and on our predictive capabilities and our ability implementation. Furthermore, aquatic to provide effective advice to the decision ecosystem responses to environmental makers and lake managers. drivers are often unpredictable leading to a large degree of uncertainty, in many aspects, The second case study includes the use of over various ranges of temporal scales. coupled hydrodynamical-biogeochemical Hence, the reaction of the ecosystem to models to the lake in order to define changes in the drivers is often hard to predict. allowable nutrient loading into the lake while In addition to the uncertainty associated sustaining acceptable conditions in the lake. with future conditions, an additional type In this case study, a series of scenarios of of uncertainty hampers attempts to model varying lake level and nutrient loading into and manage ecosystems. Known as deep the lake are tested and results are examined uncertainty, this type of uncertainty refers to using a water quality index which allows conditions in which managers, practioners, definition of acceptable conditions which are and stake holders may not agree on the translated into a SOS. I account for the two appropriate models to describe interactions types of uncertainty by introducing parameter among a system’s variables, or the probability uncertainty into the process and a wide range distributions to represent uncertainty about of variability into the forcing conditions. The key parameters in the models (Lempert et results provide robust guidelines for policy al., 2003). Thus, modelling tools utilized to makers required for defining acceptable level assist policy makers in defining, testing and of nutrient loading and lake level changes. applying policy to resource management Both cases highlight methodological must account for both types of uncertainty. approaches to assist policy makers in In this study, I present two case studies defining a policy that, under environmental on a medium-size lake in which I apply uncertainty and deep uncertainty, will ensure various modelling tools to assist in defining continued provision of ecosystem services policies based on SOS that allow, on the one while sustaining the ecosystem over time. hand, sustaining key state variables and, Though the case studies are applied to a on the hand, the on-going provision of key medium-sized lake the approach is general and can be applied to ecosystems in general.

24 Parallel sessions 1

Scenarios, model linkages and data for policy 1

Room 0.B 26 November 11:30 – 13:00

25 26 Assessing the costs of EU truck which vehicle manufacturers (OEM) have to use to certify the fuel consumption and CO CO targets with DIONE and VECTO 2 2 emissions of their vehicles. models Krause J., Tansini A., Fontaras G., European In absence of respective data, a realistic HDV Commission, Joint Research Centre fleet-emissions reference baseline needed to Steininger N., European Commission, be calculated for the regulation . The JRC was Directorate-General for Climate Action tasked to analyze large datasets provided by OEMs, which included VECTO simulation Introduction outputs for their 2016 model-year fleets As a part of its endeavour to limit global and information on the sales numbers. To warming under the Paris agreement, the calculate a representative fleet-wide CO2 EU’s low-emission mobility strategy [1] emissions baseline distribution for the year has reconfirmed the aim to decarbonize 2016, in a stepwise approach the JRC: transport. While Heavy Duty Vehicles (HDVs) 1. Evaluated the data for inconsistencies, are responsible for about a quarter of CO2 emissions from road transport in the EU and 2. Analysed the HDV fleet characteristics for some 6% of total EU CO emissions, their 2 3. Developed a methodology for normalizing emissions were not regulated in the EU until the datasets in order to calculate a robust this year. Therefore, the preparation of the and representative 2016 CO2 baseline and proposal for the first-ever HDV CO2 targets [2], presented by the European Commission in 4. Re-ran VECTO simulations for several 2018 as a basis for the recent EU regulation, thousand different vehicle configurations. required substantial efforts of data collection A key conclusion was that future reporting and rigorous modelling to assess the impacts procedures must be standardised to prevent of alternative regulatory options. In close data inconsistencies. It was found that the cooperation with DG CLIMA, the European initial calculations supplied were in some Commission’s Joint Research Center (JRC) occasions overestimated, possibly leading has provided sound data for policy making to higher CO2 emissions for the 2016 and employed the DIONE and VECTO models fleet. A baseline of lesser precision could to assess the societal and user costs under compromise the robustness of any future diverse regulation scenarios, as well as targets to be included in the regulation. This reconstructed a reference baseline for HDV was an essential observation for defining fleet emissions, as contributions to the impact realistic CO2 limits for the post-2020 period. assessment [3]. More information can be found in [4]. The baseline established was used as a key input Setting up robust HDV Emission Data with for the impact assessment accompanying VECTO the Commission's proposal for setting the respective CO standards, and provided input Prior to 2018, the energy efficiency of 2 new HDV in the EU was not assessed and to the economic modelling described below. monitored officially in the EU. Thus there was Cost Curve Development and Social/User considerable uncertainty with regard to the Cost Calculation with the DIONE model actual situation regarding HDV emissions. As HDVs are highly customizable in order An important step of policy formulation is to fit to their users’ needs, it is challenging to evaluate the costs a policy causes and to design a laboratory certification test like the corresponding impacts on affordability in the case of light duty vehicles. For this for users and OEM competitiveness. With its reason, a simulation approach was chosen for DIONE model, the JRC provides a modelling their emissions certification and monitoring. framework to assess the costs of vehicle CO2 The European Commission has developed the emission standards, first developed within Vehicle Energy Consumption Tool (VECTO), the framework of the 2017 light duty vehicle

27 impact assessment (see [5]). These modules their emission reduction and costs, derived have been adapted and extended to assess from the analysis described above and the costs of HDV emission standards. The provided within the study [7]. An advanced interaction between the modules, as well optimization algorithm (Ants Colony as inputs needed and outputs produced, Optimization) combines technologies into are sketched in Figure 1. The main HDV cost-optimal packages, and continuous cost computational modules are described below. curves are fitted. Some 80 final cost curves More information on the model is available in have been established for HDV. These cover [6]. the four VECTO classes addressed in the regulation, the fuels diesel and liquid natural Firstly, the DIONE HDV cost curve model is gas, two drive profiles, the years 2025 and used to develop curves which describe the 2030, and typical, medium and high-cost technology costs of reducing CO2 emissions estimates. An example of 2025 cost curves is of new HDV, compared to the 2016 baseline. given in Figure 2. It uses input data on available technologies,

Figure 1: Flowchart of DIONE HDV modules

Figure 2: Fitted cost curves for HDV of the sub-groups 4 RD, 5 LH, 9 RD and 10 LH, for Diesel (left) and LNG powertrain (right) in 2025, for cost scenarios typical (blue), medium- cost (green), and high-cost (red)

28 Secondly, the DIONE HDV cross-optimization [3]. COMMISSION STAFF WORKING module runs an optimization to distribute DOCUMENT IMPACT ASSESSMENT emission reduction efforts over the vehicle Accompanying the document Proposal for fleet in a cost-optimal way, building on the a Regulation of the European Parliament cost curves. As inputs, it needs a given and of the Council setting CO2 emission target for CO2 reduction and the HDV fleet performance standards for new heavy composition, the latter taken from an energy duty vehicles, SWD/2018/185 final, https:// systems model scenario. Different options eur-lex.europa.eu/legal-content/EN/ have been explored for HDV, including TXT/?uri=SWD%3A2018%3A185%3AFIN minimizing only technology costs, or total costs (including fuel savings) minimizing [4]. Tansini, A., Zacharof, N., Prado Rujas, costs from different perspectives, e.g. that I., Fontaras, G., Analysis of VECTO data for of a first user (vehicle life years 1–5), a Heavy-Duty Vehicles (HDV) CO₂ emission second user (year 6–15), or from a societal targets, EUR 29283 EN, Publications Office of perspective meeting relative (%) or absolute the European Union, Luxembourg, 2018, ISBN 978-92-79-93163-5, doi:10.2760/551250, (gCO2/km or gCO2/tkm) emission reduction targets, set for the total fleet, or for vehicle JRC112015 classes separately. [5]. Krause, J., Donati, A.V., Thiel, C., Light

Finally, the DIONE HDV total cost module Duty Vehicle CO2 Emission Reduction Cost calculates the fuel savings compared to a Curves and Cost Assessment - the DIONE reference scenario, and the change of total Model, EUR 28821 EN, Publications Office of the European Union, Luxembourg, 2017, ISBN costs of ownership caused by a CO2 standard, from the user and societal perspective. 978-92-79-74136-4, doi:10.2760/87837, JRC108725 To inform policy-making, the above modules have been run for a wealth of settings, [6]. Krause, J., Donati, A.V., Heavy duty vehicle producing several hundred scenario CO2 emission reduction cost curves and cost outcomes. Extensive analysis has been assessment – enhancement of the DIONE carried out to understand the manufacturer model, EUR 29284 EN, ISBN 978-92-79- and user cost impacts of different regulation 88812-0, doi:10.2760/555936, JRC112013 options, and results for selected scenarios are [7]. TNO, TU Graz, CE Delft, ICCT (2018), included in the impact assessment for fuel Support for preparation of impact assessment efficiency standards for HDV [3]. for CO2 emissions standards for Heavy Duty References Vehicles. Final report for Service Request 9 under Framework Contract no CLIMA.C.2./ [1]. European Commission (2016), A FRA/2013/0007 European Strategy for Low-Emission Mobility. Communication from the Commission to Energy systems modelling: toolbox the European Parliament, the Council, the development for policy decision European Economic and Social Committee support and the Committee of the Regions, COM(2016) 501 final Delarue E., Poncelet K., Moncada J.A., University of Leuven, Energy Institute and EnergyVille [2]. European Commission (2018a), Proposal Höschle H., EnergyVille and VITO for a REGULATION OF THE EUROPEAN Kiviluoma J., VTT Technical Research Centre of Finland PARLIAMENT AND OF THE COUNCIL setting CO2 emission performance standards for new Introduction heavy-duty vehicles, COM/2018/284 final, https://eur-lex.europa.eu/legal-content/EN/ A wide variety of models is being developed TXT/?uri=COM:2018:284:FIN and used in the energy sector. Developers and users include academia, industry,

29 policy makers, among others. Models can Model development be found in very different shapes, serving a A flexible and open-source toolbox for wide range of purposes. A model may refer modelling integrated energy systems to a conceptual, physical, or mathematical representation of a physical or economic Due to the increasing complexity of the system. In particular, a mathematical energy system, there is a growing need to model tries to capture the behavior of a integrate or link models covering different system by using mathematical equations or energy sectors or time frames. Currently, relationships. Based on observed/measured the vast majority of energy-system models data, physical relationships, or indeed a have been developed with a specific purpose combination of both, a set of equations in mind (e.g., investment planning, unit can be defined that describe the relations commitment, gas flow simulation). Within the between system variables, and as such the H2020 Spine project, the goal is to develop dynamics of the system.. a flexible energy system model generator that is designed from the outset to perform Energy system models are developed and these different functions and facilitate linking deployed to guide the transition of the energy of models. Key features of Spine model are system and to support energy policy-making. its generically formulated equations and the Specifically, long-term energy-system or interface enabling a direct link between the power-system planning models are frequently data and the source code. In addition, the used. These models typically span a single or Spine toolbox contributes to facilitating model multiple countries and cover multiple decades. linkages and sharing of data by having an API In terms of model input, four categories of to connect to databases of different formats, model input can be distinguished: the demand and tools allowing import and export of data for energy services, fuel prices, technology in different formats. As such, the same data descriptions (including their costs), and can be used to run a variety of models. a policy framework. The output of these models is a description of the transition In addition, transparency is key for models pathway: comprising information about the used to support policy decision making. Using investments in different technologies, how open-source models is a first step towards these technologies are operated and the achieving transparency. In this regard, there is associated costs and emissions. Liberalized a trend to increasingly publish model source energy markets, climate change awareness code in open access repositories. This practice and correspondingly increased levels of not only contributes to transparency in the variable and limitedly predictable renewables, short-run, by making the model source code impose new challenges on the development accessible to all who is interested, but also and use of such models. contributes to increasing transparency (and efficiency) in the long-run, by reducing the In this abstract/presentation, we will focus on need of different institutes to develop their two challenges/new developments in terms own, frequently similar, models. However, of modeling to support policy making, which having access to the models’ source code is we deem crucial in the future role of energy not sufficient to achieve full transparency. systems modeling: A second aspect relates to transparency of data and data handling. Even when data • Providing transparent and flexible tools is made openly available, possible barriers that allow model linkages and sharing of for transparency might arise as, depending data on the license, certain data cannot be republished or shared. Moreover, there are • Enlarging modeling perspective from frequently conversions happening from centralized system optimization to a the raw open data and the data used as market-actor perspective input for the model. The toolbox developed

30 within the H2020 Spine project aims to economic implications of this policy would create a workflow that enables connecting be. Such descriptive scenarios are therefore different data sources in various formats, crucial for translating the visions (where do and helps document the data trail to ensure we want to go?) which can be created using full transparency and replicability of model normative scenarios, to a specific policy results. portfolio (how will we get there?). Equilibrium and agent-based models are identified as From a centralized system perspective to targeted techniques for this descriptive individual market actor behavior perspective. These will be elaborated upon in the presentation based on the work currently To support policy making, long-term planning conducted in the ELDEST12project.. models are widely used to perform scenario analyses. Generating (and comparing) Conclusion different scenarios can inform decision- makers about the environmental, economic, The energy system is changing drastically: the and social implications of certain decisions. levels of variable and limitedly predictable Scenarios on various targets for renewables renewables are steadily increasing, and could, for instance, be run to assess the at the same time, massive infrastructure various implications. investments are required, in a liberalized market environment. In this context, models Depending on the question that needs to are needed to provide policy support on the be addressed, different types of scenario development of proper policy mechanisms exercises can be performed. In this regard, and appropriate market design. In this one can distinguish between normative/ abstract, two key challenges are identified prescriptive scenarios and descriptive that need to be addressed to warrant future scenarios. In normative or prescriptive model use, so as to ensure solid input scenarios, certain boundary conditions of a towards policy decision making: ensuring desired future state of the energy system are flexibility and transparency of models, and imposed upon the model. As such, normative complementing the centralized system scenarios provide information about the ideal perspective with a market-actor oriented transition of an energy system (i.e., where approach. In the presentation, we will do we want to go?). One would typically elaborate further on these aspects. deploy a centralized system perspective, single objective optimization model in this Macroeconomic effects of EU regard. Descriptive scenarios, on the other energy efficiency regulations on hand, do not impose a desired future state, but rather aim to describe a likely evolution household dishwashers, washing of the energy system, i.e., given certain machines and washer dryers assumptions on fuel prices, technology cost Rocchi P., Rueda-Cantuche J.M., Boyano A., evolutions, and policy interventions, how does Villanueva A., European Commission, Joint the energy system evolve. Such scenarios Research Centre can be used to evaluate whether certain policy measures could achieve the desired The JRC B5 unit (Circular Economy and state, and if so, under which conditions. Industrial Leadership) is producing impact For instance, policy-makers could decide to assessments for the Commission's regulations implement a subsidy scheme for solar PV on eco-design requirements for specific panels and wind turbines with the goal of appliances. These impact assessments use reaching the target for the share of RES. both technical and economic information on A descriptive scenario would then allow the appliances production technologies and assessing whether this measure is sufficient costs provided by stakeholders to evaluate to achieve the desired GHG reduction targets the environmental and economic impact of and what the environmental, social and 1 https://www.energyville.be/en/research/eldest-energy-poli- cy-decision-support-toolbox

31 different available policy options using partial required is taken from CEDEFOP, Eurostat microeconomic and technical frameworks. data, OECD data, the POLES model, UNECE The aim of the paper is to complement this data, World Bank data, and data from analysis towards the use of the general National Statistical Institutes of Belgium, equilibrium model FIDELIO, in order to China, Czech Republic, Hungary, India, provide additional information on the macro- Slovakia, Turkey, and the United Kingdom. economic impact of such regulations, taking Moreover, for this specific analysis, we use into account not only the direct impact on additional data provided by stakeholders, the industries producing the appliances, information from the 2010 Household Budget but also all induced effects in the rest of Survey micro-data produced by Eurostat, and the economy of the European countries. In previous analysis such as CLASP (2013)2.3 particular, the paper focuses on the impact of two Commission's Regulations aimed at Considering the aggregate effect in the improving the energy efficiency of household EU, the policies analysed would have a dishwashers, washing machines and washer positive impact in terms of value added dryers. and employment on the industry producing washing and dryer appliances, and some FIDELIO is an econometric recursive related industries such as wholesale and dynamic multi-regional multi-sectoral retail trade sectors and the repairing industry. general equilibrium model with some new- Besides, other industries would benefit due Keynesian features. The model offers a useful to the change in household spending driven instrument to analyse policies influencing by savings in electricity consumption (such household consumption, being the household as the food and beverage industry or the block modelled with relatively high details. accommodation industry). On the other hand, In particular, households consume durable there is a negative impact for the electricity products (housing rents and vehicles) and industry and some related industries. For the non-durable products, such as appliances, economy as a whole, even if the impact on electricity, heating, fuel for private transport, the value added is negative, the variation is public transport, food, clothing, furniture and relatively small compared to a baseline with equipment, health, communication, recreation no regulation (a decrease of 0.01% that is and accommodation, financial services, and around 1.9 billion euro). The total impact on other. For almost all consumption categories, employment is instead positive (around 24 the demand is characterized through thousand jobs), being the industries that are econometric estimations. better off more labour intensive than the industries negatively affected by the policies. To assess the impact of the regulations on household dishwashers, washing A food safety impact modelling machines and washer dryers using FIDELIO, system we introduce two different shocks for households. First, we assume a shock in the Bubbico A., Jones G., Panton S., Rose M., value of appliances consumed, assuming an MacDonald S., Fera Science Ltd Yang J., University of International Business exogenous shock in the price of appliances and Economics (UIBE) caused by the additional cost induced by the new regulation. Second, we assume a shock The emergence of a large-scale increase in the household consumption of electricity in food trade raises the risk of foodborne driven by the efficiency improvement of diseases and requires a status of ongoing appliance induced by the regulation. 2 CLASP (ed.) (2013). Estimating potential additional energy The main bulk of the data used in the model savings from upcoming revisions to existing regulations under the ecodesign and energy labelling directives: a contribution to comes from the international supply and use the evidence base (http://www.clasponline.org). http://www.clasp- tables of the World Input-Output database online.org/en/Resources/Resources/PublicationLibrary/2013/ CLASP-and-eceee-Point-To-Additional-Savings-from-Ecodesign- (WIOD) (2016 release). Additional information and-Energy-Labelling.aspx. last accessed on 08 Apr 2015.

32 awareness and readiness by Government to create the effective vaccine, to deal with authorities (e.g., national health authorities, serious environmental problems of large food standard agencies) and industry (e.g., numbers of infected pig carcasses resulting supply chain stakeholders). Having an early from the incident. insight as to emerging food safety risks is essential to this state of preparedness. FERA Science leads the Work Package Furthermore, increasing global complexity Confidence building and trade facilitation of trade links over recent decades has led within the EU China Food Safety 45 to increased opportunity for economically Partnership with the aim to build confidence motivated food fraud and associated risks for in EU-China trade by improved understanding brand owners. of consumer practices and regulatory frameworks, the latter by developing and Over the last four decades, many food- demonstrating mutual recognition of related scandals (both through unintentional laboratory standards and results. Within incidents and fraud) have contributed to the this project, FERA is developing a modelling global concern around food safety. The global system to evaluate ex ante impacts of integration of the food system increases foodborne outbreaks/food frauds and food supply chain complexity, inevitably increases safety policies on agricultural production, the risks associated with food safety (Ellefson income, markets, trade, and the environment, et al., 2012) and the need for compliance with from global to regional scale economic international quality standards (Lemanzyk impact. et al., 2015). The human cost related to foodborne outbreaks is extremely high: 600 1. The Food Safety Impact Modelling System million – almost 1 in 10 people in the world (FSIMS) uses three modelling tools: – fall ill after eating contaminated food and An Early Warning System (EWS) to predict 420,000 die every year3.4Nevertheless, the the probability of foodborne outbreaks/ impact of food system failures is much higher food frauds; than medical costs and lost productivity. The 2. A Simulation model to estimate the direct inability to meet food safety requirements economic effect on various scenarios may cause a vicious cycle generated by of outbreak/fraud detection periods lower incomes and reduced access to food based on detection standard operation which, in turn, can lead to increased medical procedures, laboratory validation, and costs and decreased worker productivity quality control measures; (Devleesschauwer et al., 2018, p. 84). 3. A Partial equilibrium model to estimate The potential impact of food safety/fraud the ex-ante wider economic impact on the incidents on a business can be devastating; agricultural sector of different scenarios a single event can bring significant economic of the economic shock following the losses due to direct costs (e.g., disruption to outbreak detection and the effect of operations while managing the recall, direct food safety policies, namely the change cost of recalling stock, analytical laboratory in Sanitary and Phytosanitary (SPS) testing) and indirect ones such as the brand measures and agreements. damage and loss of consumer confidence/ change in consumption preferences (Kennedy The EWS implemented by FERA allows for the et al., 2009; Hussain and Dawson, 2013). estimation of the risk of foodborne outbreak/ Moreover, it might cause the huge social costs food fraud. The EWS is a monitoring system comparable to the recent breakout of African concerned with examining a series of data swine fever in China. The Chinese government and identifying changes. These changes needed to take urgent action to detect reflect the mathematical properties of the and control the expansion of the infectious dataset, such as sudden changes relative to disease, to provide high R&D investments recent history. 3 https://www.who.int/en/news-room/fact-sheets/detail/ food-safety 4 http://www.euchinasafe.eu/

33 The simulation model calculates the direct (Figure 1). financial impact of the foodborne incident based on the partial budgeting approach. The EWS allows estimating the export Partial budgeting is a decision-making tool loss due to the economic shock (€80.4 used to compare the costs and benefits. Million) following the decision to recall the It only focuses on the changes in incomes pig meat-related products after the dioxin and expenses that result from the decision contamination in Ireland (Figure 2 left). process within the farm business. Thus, Finally, the analysis of the economic tangible this approach is based on the principle that direct impact of the dioxin incident (i.e. a small change in the supply chain can costs along the supply chain due to controls, eliminate or reduce some costs, eliminate or product recalls and destruction) reveals how reduce some returns, cause additional costs the financial costs increase along the days of to be incurred, and cause additional returns contamination detection (Figure 2 right). to be received (Lascano Alcoser et al., 2011). There is a growing awareness of the Turning now to the partial equilibrium model, challenges around the impact of food-related the assessment of the economic impact outbreaks and food safety policies (i.e., of food-related incidents and food safety non-tariff barriers) to trade. The FSIMS will policies in the agri-food sector requires support decision makers (public authorities economic model-based projections of future and businesses) by providing a standard tool agricultural activity levels. The aim is to use for the assessment of the economic impact well-established models used to support of food-related incidents and for food safety policymakers at the European Commission policy regulation. Moreover, the modeling and in China: The Common Agricultural Policy system will have wider applications in terms Regional Impact Model - CAPRI (Britz and of the ability to assist in the formulation Witzke, 2008) and the China’s Agricultural of EU and Chinese agricultural and food Policy Simulation and Projection Model — safety policies, which have a potential to CAPSiM (Yang et al., 2011; Xiaoyong Zhang, promote the growth of trade with restricting 2003). consequent negative impacts on the economy. Finally, the research outcome will augment In order to set up the modeling system, the analytical capacity of the CAPRI and we use case studies on past foodborne CAPSiM models and enhance their capability outbreaks, namely the Irish pig meat dioxin in terms of modelling of food safety in contamination incident (Kennedy et al, 2009; agriculture. The FSIMS will generate data for Tlustos, 2009) and the Chinese melamine European and Chinese policymakers and will outbreak in 2008 (Zhou and Wang, 2011; Xiu be able to mitigate the risks of ill-founded and Klein, 2010). The EWS can detect the high restrictions on agricultural growth. risk of an outbreak in both the case studies Figure 1: Outbreak detection by the Early Warning System

34 Figure 2: Export Loss (left) and industry direct loss scenarios on the period of contamination detection (right)

Acknowledgements: Kennedy, J., Delaney, L., McGloin, A. and Wall, P.G., 2009. Public perceptions of the dioxin crisis in Irish pork.

Lascano Alcoser, V.H., Velthuis, A.G.J., Hoogenboom, L.A.P. and Van der Fels-Klerx, This project has received funding from the H.J., 2011. 'Financial impact of a dioxin European Union’s Horizon 2020 research incident in the Dutch dairy chain'. Journal of and innovation programme under grant food protection, 74(6), pp. 967–979 agreement No 727864 — the EU-China- Safe project and China R&D Key Programme Lemanzyk, T., Anding, K., Linss, G., Hernández, - intergovernmental S&T innovation J.R. and Theska, R., 2015. 'Food Safety cooperation project 2017YFE0110800. http:// by Using Machine Learning for Automatic www.euchinasafe.eu/ Classification of Seeds of the South-American Incanut Plant'. Journal of Physics: Conference REFERENCES Series (Vol. 588, No. 1, p. 012036). IOP Britz, W. and Witzke, P., 2008. CAPRI model Publishing. documentation 2008: Version 2. Institute for Tlustos, C., 2009. 'The dioxin contamination Food and Resource Economics, University of incident in Ireland 2008'. Organohalogen Bonn, Bonn. Compd, 71, pp. 1155–1159. Devleesschauwer, B., Scharff, R.L., Kowalcyk, Xiu, C. and Klein, K.K., 2010. 'Melamine in milk B.B. and Havelaar, A.H., 2018. 'Burden and products in China: Examining the factors that Risk Assessment of Foodborne Disease'. Food led to deliberate use of the contaminant'. Safety Economics (pp. 83–106). Springer, Food Policy, 35(5), pp. 463–470. Cham. Yang, J., Huang, J., Li, N., Rozelle, S. and Ellefson, W., Zach, L. and Sullivan, D. eds., Martin, W., 2011. 'The impact of the Doha 2012. Improving import food safety (Vol. 85). trade proposals on farmers’ incomes in John Wiley & Sons. China'. Journal of Policy Modeling, 33(3), pp. Hammoudi, A., Grazia, C., Surry, Y. and 439–452. Traversac, J.B. eds., 2015. Food safety, market Zhou, Y. and Wang, E., 2011. 'Urban organization, trade and development. Springer. consumers' attitudes towards the safety Hussain, M.A. and Dawson, C.O., 2013. of milk powder after the melamine scandal Economic impact of food safety outbreaks on in 2008 and the factors influencing the food businesses. Foods, 2(4), pp. 585–589. attitudes'. China Agricultural Economic Review, 3(1), pp. 101–111.

35 The marine modelling framework evaluate different management strategies by of the EU Commission. A earth- creating alternative scenarios. system model for supporting The models within the MMF have been policy decisions already applied to a number of EU regional Macias D., Stips A., Garcia-Gorriz E., Piroddi C., seas, from the Mediterranean Sea to the Miladinova S., Friedland R., Parn O., European Black Sea, Baltic Sea and North Sea. These Commission, Joint Research Centre suit of models could provide information about different criteria addressing a number While the implementation of the Marine of descriptors of the MSFD such as ‘biological Strategy Framework Directive (MFSD) is diversity’ (D1), ‘commercially exploited progressing in the Member States (MS), the species’ (D3), ‘marine food webs’ (D4), European Commission is building up its own ‘eutrophication’ (D5), ‘sea-floor integrity’ analytical capacity in order to improve the (D6), ‘hydrographical alterations’ (D7), understanding of the marine environment ‘contaminants’ (D8) and ‘marine litter’ (D10). from an EU perspective. However, and before the MMF could be used The progress achieved during the last 30 for policy support, it is necessary that the years in marine modelling now allows a tools and their performance is scientifically more realistic simulation of many aspects of evaluated and validated. With this aim, the the marine environment. The use of marine JRC modelling team has produced a number modelling can support the assessment of peer-reviewed articles in the last few process of the marine environment as years using the MMF as scientific analysis foreseen in the MFSD by defining baselines, tool. The engagement of the wider scientific addressing data gaps and allowing for community in designing the MMF has been scenario simulations. To fully exploit this sought with the establishment of a network potential, the Commission is developing of experts in marine modelling (MEME group), a modelling framework for the marine which meet annually and provide inputs and environment (MMF) including all aspects feedbacks on the way JRC is implementing necessary to create a Regional Earth System this work. Modelling tool (i.e., the atmosphere, the hydrology and the ocean living components The level of confidence about the (both biogeochemistry and high trophic performance of the modelling tools to levels)). represent past and present environmental conditions in EU regional seas has allowed the Holistic Ecosystem Assessment is required by generation of future scenarios in the context the MSFD (and other EU pieces of legislation of climate change. Hence, management such as the CFP) in order to assess the options could now be evaluated over the achievement of Good Environmental Status background impacts climate change would (GES). Ecological modelling, combined with provoke in marine ecosystems. high-quality datasets, plays a key role within the implementation of the MSFD by In this presentation we will introduce the supporting the assessment of the ecological different components and the philosophy state of marine ecosystems (for example in behind the MMF providing some specific data-poor regions/periods) and allowing to examples of its application to policy questions in different EU marine basins.

36 Parallel sessions 1

Complex system modelling and multi-criteria decision making 1

Room 0.C 26 November 11:30 – 13:00

37 38 An accounting framework for In the agricultural sector, this situation the biophysical profiling of the is very evident due to the dilemma that exists—in terms of priorities—in terms of European agricultural system its importance and its consequences. The Cadillo Benalcazar J.J., Renner A., Institute of importance of agricultural activities is that Environmental Science and Technology (ICTA), they are responsible for providing nutritious Universitat Autònoma de Barcelona and safe food to socio-economic systems, Giampietro M., Institute of Environmental generating employment and economic Science and Technology (ICTA), Universitat benefits (for farmers, traders, and countries Autònoma de Barcelona and Catalan Institution as wholes). Nevertheless, in causal terms, for Research and Advanced Studies (ICREA) agricultural activities are among the primary Environmental impact of human activity drivers responsible for biodiversity loss and and related dramatic biodiversity loss are environmental impact. Furthermore, global defining characteristics of the contemporary megatrends of continuous population growth sustainability predicament. The concern to as well as changes in dietary patterns reduce environmental impact while preserving suggest that the impacts of agricultural the integrity and the proper functionality of activities will be accentuated in coming years. ecosystems has marked a point of consensus In this respect, the Common Agricultural among citizens, civil organizations, and Policy (CAP) has established nine objectives national and supra-national governmental aimed at covering social, economic, and organizations. However, in practice, equally environmental aspects of sustainability from legitimate but divergent interests of agents a perspective of sustainable development. involved in socio-economic systems have For example, the CAP aims to preserve the made decision making oriented towards socio-economic stability of farmers, protect sustainability ever more complex. the environment, improve food security,

Figure 1. Comparison between the biophysical resources (labor, land, blue water) used in local food production and the biophysical saving of these resources as a consequence of their externalization

39 and provide safe food to the population. system (EU-28 plus Norway) is carried out. The simultaneous achievement of all those This analysis allows for: objectives is a delicate balancing act which requires the taking of informed actions on 1. the quantitative operationalization of complex events occurring across hierarchical the formal relations between system levels—observable only across different components operating both in the scales—that are relevant to a diverse array of technosphere and the biosphere stakeholders e.g. consumers, traders, farmers, 2. to characterize the pattern of internal and institutions. A discussion ignoring the consumption of food (that which can be complex nature of these events and the affected by dietary changes) context in which they take place can lead to misleading results. Therefore, it is imperative 3. to identify for whom and/or for what that decision makers concerned with purpose system components are agricultural option spaces adopt integrated, necessary multi-criteria accounting methods capable of representing the complex relationships 4. to quantify the level of commercial between: (i) resources that are under human openness as a framework for the control (technosphere) and those that are not discussion of food security (biosphere); as well as (ii) resources that are used in a direct way in domestic production 5. to characterize the requirement of and those that are used in an indirect way primary sources (land uses, water, through imported commodities. fertilizers, pesticides, etc.) and primary sinks (emissions, waste, etc.) associated For example, taking coherent actions between with the total consumption of a society the CAP and the Biodiversity Strategy would require more sophisticated analytical tools 6. to differentiate the pressure exerted for studying the trade-offs between the both on the local environment and the productivity of agriculture and protection environments of other societies i.e. via of nature when it comes to implementing externalization. green agricultural policies and establishing The combination of these results provides protected areas. In fact, in this discussion, a system diagnosis and set of indicators there exists an elephant in the room which capable of informing decision makers should be considered: the level of openness concerned with state assessments, including of the agricultural sector in the EU i.e. data referring to processes under human dependency on imports. This consideration control such as population-scale nutrition, is especially important in view of the the technological requirements of food massive increase in the requirement of food production, and the level of commercial production expected in the future at the world openness of the agricultural system, as level and the risk of geopolitical turmoil that well as pressure assessments, including requirement could generate. data referring to processes outside human This work presents a novel accounting control such as the local and externalized framework based on the Multi-Scale demand for a variety of ecological services. Integrated Analysis of Societal and Ecosystem Additionally, the spatial contextualization Metabolism (MuSIASEM) approach and of these pressures allows for an analysis of tailored to the analysis of the water- environmental impact. food-land resource nexus. The versatility The results show that the European of this approach allows the quantitative agricultural norm is found to be one of contextualization of narratives focused substantial externalization of both social and on sustainable development. In order to ecological resource requirements on an order demonstrate the usefulness of the approach, of magnitude at least as large as the internal an analysis of the European agricultural requirements for production factors including 40 labor (human activity), green and blue water The thirty sectors in the base model, which use, and land use (Figure 1 shows examples together capturing dynamics in land use, of the results). Indeed, the externalization of water demand and supply, emissions and agricultural production represents a major waste, soil nutrient, energy, consumption hurdle confronting Europe’s sustainability and generation, material consumption, goals and deserves more attention among biodiversity, population, fertility and policymakers. mortality, health, education, poverty, income distribution, employment, infrastructure, Integrative planning for a finance and balance of payments, agriculture, sustainable, prosperous future industry and service production and government expenditure. These all modeled Chan D., Urantowka W., Pedercini M., Millennium Institute and interact with each other. Each sector has its own mechanisms and is designed to Is there a way to quickly gain a helicopter be comprehensible within itself. We provide view of the impact of certain policies across detailed descriptions of the model through all social, economic and environmental a model documentation available online sectors of society and discuss it with a [8]. Additional indicators and sectors can be diverse of stakeholders? Can we intelligently modeled depending on the context, project develop policies that can be both sustainable and availability of data. and prosperous? What economic, social and environmental consequences could arise if The calibration process of the model is the planet’s temperature were to rise 1.5 essential to the building of confidence in its degrees by 2050? structure and results. Firstly, preliminary simulations are conducted to validate the Models based on Threshold 21 (T21) model with respect to historical data. Further, technology has been continuously developed we take the model structure and conduct over the past 30 years as a tool for policy simulations within it. Then, hypothetical development by the Millennium Institute, a single- or multi-policy scenarios can be not-for-profit organization [1]. Models have introduced and its effect on indicators across been applied in over 40 countries worldwide all sectors can be analyzed. and have been used to develop plans for national development, green economy, Model structures are grounded in scientific sustainable agriculture, renewable energy research specific to each sector [8]. transitions, industrial reform and Sustainable Additionally, system dynamics researchers Development Goal planning in such varied also offers a great wealth of research and contexts as Denmark, Mongolia, Peru, debate into how to model specific structures Senegal, South Africa and the United States and its effects on socioeconomic systems [2], [3]. [9]. Continuous refinement of the base model Additionally, they have been employed to after each subsequent project, constant study economic, social and environmental comparison with data, partial model and full consequences of various climate change model testing methods are used to ensure it scenarios [4], [5]. Models comprise has a good fit with the data and are in line 30 sectorsbroadly categorized as with the research that the model is based off environmental, social and economic and are of.5,10. Each model is calibrated using data built followingthe System Dynamics (SD) specific to each context the model is applied methodology, which excels at deconstructing to. This is typically a country, however the and analyzing complex socio-economic models have been developed to sub-national environments and policy systems [6]. This and supra-national contexts. method of simulation aids in garnering With a fully calibrated model, many insight into complex relationships between hypotheticals can be tested, with projections constructs [7]. of many indicators shown across thirty 41 sectors. Additionally, policy scenarios Millennium Institute’s iSDG model, Expert can be simulated with realistic, costed Perspectives, World Resources Institute. interventions and the effects they have on model dynamics and results can be 5. GRIGGS D., STAFFORD-SMITH M., seen instantaneously. Combinations of GAFFNEY O., ROCKSTROM J., OHMAN M.C., interventions can be simulated within the SHYAMSUNDAR P., … , NOBLE I. (2013), model, and often positive synergies can be 'Sustainable development goals for people found with interventions integrated together. and planet', Nature, 495, 305–307. As one of many examples, typically health 6. STERMAN J. (2000), Business dynamics: and education infrastructure investment is Systems thinking and modeling for a complex more effective if roads are invested into at world, Irwin/McGraw-Hill. the same time, as clinics and schools would become more accessible. T21 encourages 7. DAVIS J.P., EISENHARDT K.M., BINGHAM C.B. collaborative policy planning that is inclusive (2007), 'Developing theory through simulation of what is happening across all sectors. An methods', Academy of Management Review, intuitive interface as well fully transparent 32(2), 480–499. model allows for planning across often siloed government ministries, private sectors 8. MILLENNIUM INSTITUTE (2017), iSDG interests and other stakeholders, which Model Documentation, https://www. will help policy makers to form integrative millennium-institute.org/documentation, solutions to pressing societal problems often accessed 24 February 2019. at lower cost, while taking into account other needs. The model offers a space in 9. GEELS F.W. (2004), 'From sectoral systems which to easily assess the relative impact of innovation to socio-technical systems: (both positive spillovers and negative Insight about dynamics and change from consequences) of differing investment and sociology and institutional theory', Research policy scenarios for more inclusive policy- Policy, 33, 897–920. making. 10. FORD D.N., STERMAN, J. (1998), 'Dynamic References modeling of product development processes', System Dynamics Review, 14, 31–68. 1. PEDERCINI M., ZUELLICH G., DIANATI K., ARQUITT S. (2018), 'Towards achieving Decision support system for Sustainable Development Goals in Ivory maritime spatial planning within Coast: Stimulating pathways to sustainable blue growth and ecosystem-based development', Sustainable Development, 26, 588–595. management Abramic A., Garcia Mendoza A., Phorè S., 2. MILLENNIUM INSTITUTE (2018), Millennium Fernandez- Palacios Y., Haroun Tabraue R., Institute: Projects, https://www.millennium- ECOAQUA Institute, Scientific and Technological institute.org/projects, accessed 24 February Marine Park, University Las Palmas de Gran 2019. Canaria

3. PEDERCINI M., BARNEY G. O. (2009), Introduction Dynamic analysis of interventions designed to The main challenge of PLASMAR Project achieve Millennium Development Goals (MDG): is addressing the implementation of the The case of Ghana, 44, 89–99. Maritime Spatial Planning Directive 2014/89/ 4. ARQUITT S., PEDERCINI M., ONASANYA EU (MSPD) within the present framework of A., HERREN H. (2016), Addressing climate- emerging maritime activities linked to Blue sustainable development linkages in long- Growth development and within the limits term low-carbon strategies: The role of of ecosystem-based management. The main goal is to provide a pilot zoning, for the

42 Exclusive Economic Zone of the Macaronesian Region. Suitability is based on the MSP Region (Azores, Madeira and Canaries data framework parameters (more than archipelagoes), including aquaculture, 80), assessed relation to each maritime offshore wind energy, mineral extraction, sector, and significance given as a weight. To fisheries, maritime tourism and transport, define the parameters’ weights, an analytic based on environmental sustainability. It hierarchical method was used, in conjunction is with this aim that PLASMAR Project has to questionnaires filled by project partners, developed a Decision Support System (DSS) sectorial experts and finally stakeholders, based on available scientific knowledge, during the set of meetings and workshops. collection of scientific data and developed research. As a result, INDIMAR DSS, provides the identification of areas with environmental Data collection sensibility, areas of potential conflicts among maritime sectors, areas with significant lack The first step was data collection, which of data, and, finally, pilot zoning areas for was developed gathering products delivered each maritime sector. by European/global data initiatives: mainly Copernicus (Eye on Earth), European Marine The final version of the DSS INDIMAR will Observation and Data Network (EMODnet), be publicly available as an interactive web European Environment Information and application. Observation Network (EIONET), but also Oceanographic Data and Information Policy support within the DSS INDIMAR Exchange (IODE), national/regional and Developed model and respected results will local data infrastructures. To facilitate the be used by competent authorities (Portugal collection of data and information, a Maritime & Spain) for finalizing Maritime Spatial Plans Spatial Planning (MSP) data framework was for Azores, Madeira and Canaries, needed developed within the project, comprising: for maritime sectors development and a 1. Marine Strategy Framework Directive requirement from the MSPD 2021. Final 2008/56/EC (MSFD): Good Environmental product Decision Support System INDIMAR Status 2017/848/EU (GES) parameters; results includes:

2. Marine Protected Areas information; • A planning scenario based on the preservation of the marine environment, 3. Coastal Land Use data; meeting suitability with respect to the MSFD Good Environmental Status 4. Physical Oceanography data; parameters. This scenario identifies marine areas for the development of 5. Current maritime activities and uses. maritime sectors, where the expected The principles of the INSPIRE Directive impact on the environment will be 2007/2/EC were applied to data collection, minimized; harmonization, and sharing. • Planning scenarios based on the coastal and oceanographic conditions potential Decision Support System for maritime sectors development The developed DSS INDIMAR, is a web (e.g. wind strength, depth… applied for application which is fed with the collected offshore wind parks); data and available scientific knowledge — • A planning scenario based on the gathered by review of scientific and technical avoidance of conflicts among maritime reports. The DSS delivers a multi-criteria sectors. analysis to define environmental suitability • A scenario that integrates all previous of the analyzed maritime sectors within the scenarios. Exclusive Economic Zone of the Macaronesian

43 In June 2019, was approved follow up project the basis for Sustainable Maritime Spatial - PLASMAR+ (2020–2023), where planned to Planning in Macaronesia'), which is supported carry on development of the DSS, including by the European Union and co-financed the social and economic components in the through the European Regional Development model. Fund (ERDF) via the Operational Programme of Territorial Cooperation the INTERREG V-A Acknowledgements Spain-Portugal MAC 2014–2020 (Madeira- Azores-Canarias). This work was developed within the framework of PLASMAR Project ('Setting

Figure 1: INDIMAR DSS snapshot, suitability test for the aquaculture facilities

Methodological reflections on the the help from the advancement of applied European Commission’s 2013 use environmental economics, CBAs have started to consider non-market external effects of CBA for air pollution policy that a contemplated project or policy would Åström S., IVL Swedish Environmental Research have, thereby relaxing the assumption of Institute Ltd. ‘no externalities’ in the idealistic ‘market under perfect competition’. There are Cost-Benefit Analysis (CBA) has been now continuous method and application an important decision support tool for developments, as well as guidance efforts to environmental and air pollution policy for increase transparency and streamlining CBA. decades in Europe and the European Union Thanks to the inclusion of externalities and (EU). Today in many EU countries, some the streamlining of methods and data, one of sort of socio-economic impact assessment the European environmental areas where CBA (IA) of policy proposals is mandated by is used to check socio-economic soundness law, and commonly CBA is utilised. With (that benefits will be higher than costs) of

44 policy proposals is since some 20 years air Pindyck (2013) in his critical climate CBA pollution policy. review proclaimed CBA as fit for purpose for air pollution problems but did for natural On the 18th of December 2013, the European reasons fail to recognise the policy closeness Commission (EC) changed their approach and prescriptive nature of the recent EC to CBA. Instead of using CBA to corroborate AP-CBA proposed after Pindycks’ publication. net socio-economic benefits of a policy Further, even though Gowdy (2007) and ambition level, CBA was used to model the Brennan (2014) presents interesting new 2025 air pollution emission levels where research and perspectives on how to adapt marginal costs of emission reduction would CBA to knowledge developed by experimental equal marginal benefits (i.e. maximize economists, there are yet more perspectives modelled welfare). This CBA model result necessary to complete the picture. Other then served as quantitative basis for the overviews and reviews have been re-iterating proposal of a 2030 policy ambition level, the same critique that has been heard for after adjusting to administrative and political decades now (Frank 2000), without adding considerations. CBA was thereby used to much new substance whilst aiming their prescribe a policy proposal, and consequently message at new audiences (Hwang 2016, used at a high degree of precision: the model Pindyck 2017). Still other texts seem to results specified 2030 emission levels for 5 accept the textbook CBA-concept fully and pollutants in 28 countries. This new use of mainly focus on ensuring that the reader of CBA to prescribe emission targets for policies CBA results is informed on key sensitivities was justified via reference to technological in CBAs (Dudley et al. 2017). But none of infeasibility of reaching the EU long-term the reviews and critique properly addresses air quality objectives formulated in the 6th the practice of basing CBA on prospective and 7th environmental action plans, and via scenarios, nor do they appear to consider adherence to economic rationality. policy environments in which the CBA is made. The purpose with this review is to In parallel development with the new EC discuss and assess the potential reasons use, the academic critique against CBA for for why policy makers are using even more environmental policy has been growing. As exact CBA-results for air pollution policy examples, Ackerman et al. (2009) argues despite a growing body of methodological that CBA, or rather integrated assessment critique against CBA. The guiding questions modelling (IAM) for prescription of climate are: Despite a large and growing body of policy ambition, is difficult to defend from an arguments against the use of CBA, how can it epistemic stand-point. And Pindyck (2013, be that CBA currently is used in a prescriptive 2017) basically dismisses IAM (CBA) as and high-precision version in an air pollution useless. Further, Gowdy (2007) stress the policy context? Is the latest EC approach to need to adjust conventional CBA with respect CBA fit for purpose? How high impact would to lessons learned from the research field any change in approach reasonably have? of experimental economics. The common And what are the alternatives to the current themes in the critique related to uncertainty, approach? discounting, equity, incommensurability, and problems in valuation methods are all Overall the review suggests that the perhaps most attenuated in climate policy but academic support for scepticism towards CBA also applicable to air pollution policy. results has increased during the last years and have gotten more dimensions: system This presentation scrutinizes the new EC analytical, experimental, and psychological. approach to air pollution CBA (AP-CBA) by Further, there are several indications that reviewing recent research developments the reality of air pollution policy is changing. of relevance for CBA and by reviewing the More focus is now on individuals than in policy process in which the CBA was done. the past: if so, decision making by ordinary This review is appropriate since there are citizens will become more important for omissions in earlier overviews and reviews. 45 future air quality and a modelled optimal It can be settled that it is an impossibility emission level assuming economic rationality to require truthful predictions from models will be rendered less reliable. Based on these used for prospective studies of unmeasurable results it could have been argued that the parameters, and democratic policy processes 2013 EC use of AP-CBA to serve as basis for are likely to result in political ambition levels a policy proposal was an example of over- different from modelled ambition levels. confident use of model results. However, the The policy effect of any model improvement air pollution policy process provides input is uncertain, but model developments are that also needs to be considered, where nevertheless desirable. Future AP-CBAs the importance of a rational approach to should compare alternative solutions to reach policy support and the political process policy ambitions as well as compare different must be kept top-of-mind. After all there are types of ambitions. It is also important that indications that the new AP-CBA approach the models used for policy support analysis encouraged higher policy ambitions than complements the economic rationality earlier approaches. presented in AP-CBA with other rationales.

46 Parallel sessions 2

Model quality and transparency 1

Room 0.A 26 November 14:00 – 15:30

47 48 Challenges and opportunities of with conflicting policy perspectives and integrated policy modelling expectations. Thus, when policy questions are complex and more than one stakeholder Hordijk L., Special advisor, European is involved, modellers may be confronted Commission, Joint Research Centre with different interests/stakes. Reaching Kancs d'A., European Commission, Joint Research Centre one policy target may move the society further away from another policy target. An Policy challenges are increasingly complex integrated modelling may reveal trade-offs and go beyond traditional policy or model between targets of different policies and areas. This requires more than ever an inconvenient truth about policy impacts. before combining models for in integrated For example, an integrated modelling assessment framework (e.g. water'energy- framework may reveal trade-offs between agriculture; demography-migration-climate- policy targets of environmental and climate economy, etc.). The benefits of an integrated policies on the one hand and energy and use of models to support policy making are foreign trade policies on the other hand. The more and more acknowledged both by policy scientific integrity requires that modellers makers and modellers. remain neutral, with independency from policy makers being assured1.6Anticipating However, an integrated use of models for potentially conflicting interests of policy policy is associated also with challenges stakeholders behind the different models and pitfalls which are not encountered and conferring them transparently may (encountered less) when using a single model help to mitigate conflicts between different to address a specific question in one policy stakeholders at later stages. area. This paper attempts to raise awareness of key challenges related to an integrated use The presence of multiple stakeholders of models for policy support. Secondly, we requires a careful formulation of policy attempt to identify approaches allowing some questions — preferable this is done jointly by of integrating modelling pitfalls turn into modellers and policy makers. It is generally a opportunities, if addressed in a timely and good practice but in the case of an integrated appropriate manner. modelling it may be a particularly rewarding strategy to formulate policy questions jointly Policy questions. Many policy questions can by policy makers and modellers by accounting be answered by single models. In the same for possibilities to analyse policy questions. time, increasingly, today’s global societal challenges require a joint use of models Multidisciplinary approach. Global societal from different disciplinary areas (social challenges, e.g., energy, water, food security, sciences, environmental sciences, etc.). For global health and urbanisation, involve example, modelling policy questions related the interaction between humans and their to Sustainable Development Goals (SDG) environment. A (mono)disciplinary approach, requires an integrated modelling approach. be it a psychological, economical or technical The first set of challenges that we would one, is too limited to capture any one of these like to point out relates to the formulation challenges. To understand policy impacts in of policy questions as well as potentially presence of interactions between humans conflicting interests of policy stakeholders and environment requires knowledge, ideas related to the different models developed for and research methodology from different different policy areas. disciplines (e.g., ecology or chemistry in the natural sciences, psychology or economy Models that are being linked often have in the social sciences). Hence, collaboration originally been built for different policy between natural and social sciences is questions originally. As result, the integrated needed. system of models might have stakeholders 1 For example, in several Member States the scientific neutrality of scientists is anchored in the national legislation governing scientific institutions. 49 Conceptual challenges. Models for policy Precision and complexity versus robustness support are built on certain theories, each and tractability. Integrated models have a of them contains assumptions, particularly tendency to become more and more complex, models in social sciences disciplines. When more and more detailed and thus more and integrating models from different scientific more difficult to understand and validate. disciplines, likely, their theoretical frameworks An integrated modelling entails a trade-off: will differ. Among these differences, those trying to take the complexity of the world with contradicting theoretical insights into account on the one hand and increased should be paid a particular attention. The uncertainty and difficulty to understand same applies to ensuring consistency in model outcomes on the other hand. In order exogenous assumptions. A further conceptual to approach the trade-off of precision and challenge is related to interfaces/endogenous complexity versus robustness and tractability, variables for linking different models. In a good common understanding of what is a fully integrating modelling system, the important across different disciplines and a same endogenous variables should not be positive attitude toward compromises are computed by different sub-models, as in required. In order to avoid that integrated most cases their results will differ. Instead, an models become 'monsters', key insights iterative feedback loop should be developed from each scientific discipline (an educated for enable to run different models iteratively compromise), as well as trust and respect and converge towards a stable equilibrium. between scientists from different disciplines are required. One way to find the appropriate Empirical challenges. Different models use balance for example between a very detailed different sources of data and depending on natural science component and a very the policy question also different baseline aggregated social science component could assumptions. When integrating different be to undertake a global sensitivity analysis models, there is a fair chance that already of the integrated modelling system. the input data are inconsistent across different sub-models; the more different Transparency and quality assurance. Quality are the scientific disciplines for which the assurance / transparency is not an easy task models have been developed, the larger because in integrated modelling systems will be these differences. To avoid and/or data often come from a large variety of reduce inconsistencies among model data sources, errors percolate through the and baselines, exogenous assumptions and linkages, implying that ensuring a robustness baseline scenarios should be aligned for those of results can be a particularly challenging exogenous variables and parameters that are issue (more challenging than of single model present in several sub-models of the system. results). Indeed, the more complex is the modelling system, the more difficult is to Even when models in different disciplines may ensure robustness. One way of validation of use the same type of data, their spatial and integrated modelling systems is matching temporal resolutions and other dimensions model results with past observations may differ significantly. For example, whereas attempting to replicate historical data. an energy model in economics may suffice Further, local experts from outside modelling with an aggregated country-level energy teams can also contribute with their local supply data with yearly time steps, an energy knowledge to building a credibility of large model in an engineering discipline may work integrated modelling systems. with data at time steps below hours at the grid level. Maintenance, updating and documentation. Large integrated modelling systems Consistent aggregation/disaggregation are challenging to maintain, update and methods are necessary to address differences document. For these tasks, dedicated data across models related to spatial and temporal scientists in the integrated modelling team resolutions and other dimensions. are required that are dedicated to the system

50 maintenance, updating and documentation. phase. In doing so, the EC makes extensive Further, the continuity of staff in integrated use of models. A better understanding of modelling teams is even more important than these models and how they are used can for single models, as such systems tend to be then contribute to a sound use of evidence in more complex than single models. This is also support to EU policies. important for building trust and credibility with the model users and other stakeholders. The Commission's Competence Centre on Modelling (CC-MOD) promotes a responsible, Meta-modelling. A model is an abstraction coherent and transparent use of modelling of phenomena in the real world; a meta- at the EC. As part of its activities, this model is yet another abstraction, highlighting analysis systematically investigates how key properties of the underlying models. In models are used in support to the policy times of increasing policy complexities and formulation phase, by looking at EC IAs which interdependencies, developing/using meta are publicly available. A total of 1063 IAs models to provide integrated modelling carried out in the years 2003 to 2018 have solutions serves a feasible alternative to been investigated to examine the frequency complex large integrated modelling systems. and characteristics of model use, by using Rather than hard or soft coupling, a series text mining techniques complemented by of models or meta modelling might be an manual post processing. The research is alternative. In several policy areas, such as facilitated by and feeds back into MIDAS, the the bio-economy, meta-modelling is well Commission-wide modelling inventory and advanced in the European Commission knowledge management system developed already. For example, within the bio-economy and managed by CC-MOD (Ostlaender et al. domain several sectorial models have been 2019), which directly contributes to enhanced interlinked and integrated in a coherent transparency and traceability of models used modelling framework to support policies.

Modelling for EU policy support: Our results show that models are used in impact assessments 16% of the total IAs (173 out of 1063 IAs), with a positive trend over time, starting with Acs S., Ostlaender N., Listorti G., Hradec J., only two IAs using models in 2004, to around Smits P., European Commission, Joint Research 25–30% from 2015 onwards. Centre Hardy M., UniSystems We identified 123 different models Hordijk L., Special Adivsor, European contributing to IAs. More than half (53%, or Commission, Joint Research Centre 65) of these models were used only once, The objective of this work is to systematically which leads to considerations related to the analyse how models are used in support to efficiency of model use and reuse in the the policy formulation phase of the EU policy EC, as well as on the scope for improved cycle. We focus on European Commission (EC) coordination of models related products and Impact Assessments (IAs). services. On the other hand, some models do dominate: the top 10 models contributed The main framework of the EU regulatory to 10 or more IAs, and were used in 66% (or policy, the Better Regulation Agenda (BR) 114) of the total number of IAs using models. (European Commission 2015), sets a clear commitment to a transparent and sound use Included among the top 10 models are also of evidence for all EU policy making activities. those models used for the development of The Better Regulation Guidelines (European the series of EU Reference Scenarios, led by Commission 2017), which complement DG ENER, DG MOVE and DG CLIMA. Indeed, the Agenda to provide concrete guidance the consistent use of the same series of throughout the policy cycle, recommend to baselines in the areas of energy, transport quantify costs and benefits to the extent and climate goes in the direction of increased possible to support the policy formulation consistency across policy areas and fields 51 of analysis, though there is still room for references or outdated hyperlinks, which strengthening the coherence in baselines made the quantitative evidence untraceable across the whole EC. At the same time, given in several IAs. Since 2017, the BR foresee the dominance of these models to support that, when IA analysis relies on modelling the policy formulation phase, it is essential or the use of analytical methods, the model that they undergo careful quality scrutiny and should be documented in the corporate that maximum transparency and traceability modelling inventory MIDAS (European of results is ensured. Commission 2017). This represents a major step forward in terms of transparency. At the Policy areas with the highest numbers of same time, it is also clear that further action IAs using models are environment (including is needed to use and promote best practices climate), internal market, transport and to ensure transparency and accessibility over energy. However, this could also reflect other time of the evidence base in support to IAs. factors, such as the frequency at which IAs In addition to the BR guidelines, the JRC, as are carried out in the various policy areas. the science and knowledge in house service In addition, it should be remembered that of the Commission, can provide additional models can also be used in other policy assistance and support to the Policy DGs. relevant studies, or for internal notes and analyses which remain unpublished. To conclude, our analysis directly contributes to the implementation of the BR Agenda, Finally, results show that 94% (116) of the by promoting transparency, coherence, total number of models that were used in traceability and accountability in the use of IAs were used for ex ante assessment of evidence for EU policy making. policy options. This is to be expected, since indeed the assessment of policy options is From modelling guidelines to an extremely relevant task for quantitative model portfolio management analysis in IAs. As mentioned, most of the top 10 models were also used for baselines. Arnold T., Grey Bruce Centre for Agroecology Guillaume J.H.A., Australian National University In this respect, transparency is crucial to Lahtinen T., Aalto University understand how models work and to validate Vervoort, R. W., University of Sydney their behaviour, to encourage their sound and Numerical simulation models have moved widespread use in support to policy making from exploratory academic tools to become contributing at the same time to a more mainstream tools in decision making. Models effective and efficient use of resources for are used in a variety of functions such as model development and use within the EC. interpretation of monitoring data, design of The mandatory requirement introduced by monitoring networks, impact assessment, the BR Agenda in 2017 for an annex on and scenario-based planning. Today, model- ‘Analytical models used in the preparation based assessment is prescribed by several of the impact assessment’ contributed to environmental regulations and policies at better model descriptions and an increased multiple scales, such as the Water Framework transparency of the methodology used. Directive in the EU, or the Clean Water Act There is, at the same time, still room for in Ontario, Canada. This abstract takes a improvement in better documenting models high-level management perspective on these as the individual model descriptions in IAs are regulatory processes and offers the concept still quite variable in quality. The information of model portfolio management in order to and reports generated by MIDAS on models govern more effectively. and model use can be used in this respect. Multiple guidelines exist that clearly lay out: At the same time, we also identified some steps involved in a modelling project (e.g. major challenges related to referencing in IAs, Jakeman’s Ten iterative steps in development such as lack of harmonized and adequate and evaluation of environmental models);

52 criteria for how to choose appropriate in our research are lack of awareness and/ model code or modelling methods (e.g. or priority by higher-level management, Vanrolleghem’s Modelling aspects of Water lack of examples to follow, lack of model Framework Directive Implementation); or management guidelines, over-ambition detailed guidance on how to document a of modelling practitioners especially with model application. There is, however, a lack respect to software tools, and an overall of guidance that would help environmental hesitation of senior management to address managers effectively integrate simulation this issue. and decision models into the knowledge infrastructure of an organization, especially Looking for analogous situations in the field if compared with the abundance of guidance of public management, 'project management' for developing and managing geospatial offers guidance for effectively dealing or relational databases. As a consequence, with single projects, and 'project portfolio staff at most agencies that we encountered management' guides high-level managers are managing modelling projects ad-hoc, to orchestrate the efficient acquisition and selecting and applying modelling guidelines execution of multiple projects. In a similar and tools based on personal preferences, way, we define model project management without an explicit strategy of how to re- as the process of effectively rolling out one integrate the knowledge and tools obtained modelling effort, and a model manager as in a modelling project into a larger knowledge someone who manages these efforts. Model strategy. From an agency perspective, portfolio management is an organization’s synergistic strategies for executing modelling governance approach that assures synergies projects could improve cost-effectiveness, and resource efficiency when implementing impact, create more lasting benefits, while multiple modelling projects, assuring reducing dependence on individual staff. that efforts contribute to the building of knowledge infrastructure in ways that align Model management, or management of with an agency’s goals, and the operational numerical models (MNM), is an overarching and administrative support that an agency term that encompasses governance, offers to its model managers. operational support, and administration of modelling: the managerial procedures We provide a systemization of the emerging by which technical modelling projects are discipline of model management around the governed, 2) the technologies that make functions that model applications fulfill for up an organization’s infrastructure for data agencies, the tasks that the agency have and modelling knowledge and IT support, to fulfill, and the leverage points that the and 3) the human resource aspects of how managers of modelling projects have (Figure knowledge is accessed throughout the 1). model’s lifecycle (Arnold, Guillaume, Lahtinen, As one example, the Guide for Actively Vervoort, submitted). Reviewing lessons Managing Watershed-Scale Numerical Models from research and practice highlights that by the Oak Ridges Moraine Coalition (Holysh, management of numerical modelling projects Marchildon, Arnold, Gerber, 2017) summarizes could be greatly improved by moving from lessons and points out low-hanging fruit after ad-hoc and opportunistic decision making reviewing consultant-agency relationships toward intentional, 'explicit' governance of after Ontario designated drinking water modelling projects with defined institutional protection areas and commissioned goals and a long-term strategy for knowledge hundreds of diverse model applications. and software management. Senior Recommendations range from simple rules managers perceive modelling projects as for file naming and file directory structures, complex and requiring a multitude of difficult standard formats for data and metadata, and highly specialized fields of knowledge, over templates for consulting contracts with many of which are beyond their own expertise suggested formulations around intellectual and comfort zone. Major barriers identified property rights, to standard disclaimers. 53 The guide also details potential for the role Benefits include automation of complicated, of a Model Custodian as a shared resource repetitive, technical tasks, reduced duration across multiple agencies. The custodian and cost of modelling cycles, and reduced offers a custodianship plan that supplies a reliance on tacit knowledge by making hidden variety of model management tasks to an knowledge transparent. These organizations agency. This person also supports agencies in drastically reduced the cost of setting up a designing modelling studies, by formulating new model application, and made it easier to goals, requests for proposals and subsequent work in teams or transfer model applications contracts; the custodian reviews proposals from one staff person to another. Yet, many and later project deliverables but also offers rural organizations point out a need for model a platform for archiving and sharing model portfolio management approaches with low applications or parts of these. overhead cost that are appropriate for staff without in-depth modelling experience. The principles of model portfolio management can be taken much further, as demonstrated The presentation will lay out principles for by engineering consultancies, Earth System model and model portfolio management modelling centres, or operational watershed that were identified through our work. It also agencies. Synergies are derived from summarizes leverage points for how model cyberinfrastructure workflow systems that portfolio management can be improved are integrated with data and documentation within public agencies, with the goal of standards, operational procedures and promoting the transparent and cost-effective guidelines, tutorials and other education use of models as management and decision methods, and communication strategies. tools.

Figure 1. Management of numerical models within three embedded modelling cycles. Technical modelling tasks are interdependent with conceptual learning and policy & management activities

54 Post-hoc evaluation of model ecological and/or pathogen ensembles. We outcomes tailored to policy follow the pathway from model predictions to decision and post-hoc final outcome. The support steps are illustrated with the interpretation of Thulke H. H., Lange M., Helmholtz Centre for the decision needs translated into necessary Environmental Research GmbH — UFZ modelling amendments and the quantitative model predictions. We use several historical Expert-knowledge based system models are assessments regarding wild boar (Sus scrofa). flexible tools prepared to support strategy These problems cover natural expansion into identification and policy decisions. Kitching naïve habitat regions as ground for species et al. (2006) note the key question for any management in Denmark (Alban et al. 2005; policy oriented model is whether decisions Moltke-Jordt et al. 2016), the incursion of made with it are more correct than those Foot-and-mouth disease in Bulgarian wildlife made without it. Nevertheless, literature is and the need for EU emergency action (EFSA scarce of scientific evaluations of modelling 2012; Dhollander et al 2016), and the large- initiatives tailored to support policy decisions. scale management following invasion of This presentation addresses the issue using the African swine fever virus into European insights from three practical examples and Union (Lange et al. 2014; EFSA 2017; Lange targets understanding the chance and limits et al. 2018). Further the related managerial when answering the key question for policy implication and implementation is discussed oriented models. before the model outcome is challenged In the EU a prominent area of policy with the real world response to management models addresses health risks to plant and decision. animals due to established (i.e. endemic) or The lessons learnt target three aspects invasive (i.e. foreign) pathogens. The real- of Kitchings’ key question: the purposeful world link between science and policy in application of expert system modelling in the area has several grounds being food crisis context when usually shortage of safety, wellbeing of organisms and nature time argues against; the improved decision conservation perspectives. The multiple recommendation using these models i.e. dimensions of the objectives and the variety why the potential decision implied was more of potential pathogens hazardous to plants correct; the cumulating trust in the model and animals justifies the great spectrum basis for decision support via outcomes of modelling activities in support of policy derived for particular policy questions i.e. decisions. Therefore the field of health risks post-hoc adequacy testing. We present the management created individualised model quantitative predictions spatio-temporally tools of substantial complexity, ad-hoc explicit and compare it with the real world hypotheses and often uncertainty regarding notifications for the example problems. the system-level functioning/interplay of detailed process knowledge. Recent methods References of mathematical computing allow mimicking both the detailed process knowledge driving L. Alban, M.M. Andersen, T. Asferg, A. Boklund, logical conclusions in decision making N. Fernández, S.G. Goldbach, M. Greiner, A. together with the alternative mitigation Højgaard, S. Kramer-Schadt, A. Stockmarr, action or even contradicting interpretations H.-H. Thulke, Å. Uttenthal, B. Ydesen (2005). of observational evidence. The usually short Classical swine fever and wild boar in time horizon between risk assessment, Denmark: A risk analysis. Project report, DFVF, decision taken and real outcome makes the pg. 118. (ISBN: 87-91587-01-8). subject area a good arena for lessons learned. Dhollander S., Belsham G.J., Lange M., We synthesise a list of model applications for Willgert K., Alexandrov T., Chondrokouki E., decision support from past years addressing Depner K., Khomenko S., Özyörük F., Salman independent policy questions and different M., Thulke H.-H., Bøtner A. (2016). 'Assessing

55 the potential spread and maintenance of The MIDAS touch — Maintaining foot-and-mouth disease virus infection an overview of model use by in wild ungulates; general principles and application to a specific scenario in Thrace'. the EC through the EC Modelling Transboundary and Emerging Diseases 63: Inventory and Knowledge 165–174. doi:10.1111/tbed.12240 Management System MIDAS EFSA (2012). 'Scientific Opinion on foot-and- Ostlaender N., Acs S., Listorti G., Hradec J., mouth disease in Thrace'. EFSA Journal 10 (4): Smits P., European Commission, Joint Research 2635. doi:10.2903/j.efsa.2012.2635 Centre Hardy M., Ghirimoldi G., UniSystems EFSA, Depner K., Gortazar C., Guberti V., Masiulis M., More S., Oļševskis E., Thulke The European Commission's (EC) Competence H.-H., Viltrop A., Woźniakowski G., Cortiñas Centre on Modelling (CC-MOD) promotes a Abrahantes J., Gogin A., Verdonck F., responsible, coherent and transparent use of Dhollander S. (2017). 'Scientific Report on modelling to underpin the evidence base for the epidemiological analyses of African EU policies. swine fever in the Baltic States and Poland'. Maintaining an overview of ongoing modelling EFSA Journal 15 (11): 5068. doi:0.2903/j. activities, by documenting the models efsa.2017.5068 and model combinations in use across the Kitching et al. (2006). 'Use and abuse of EC, is an elementary first step for a more mathematical models: an illustration from transparent and coherent use of models in the 2001 foot and mouth disease epidemic the policy cycle. It is, however, also a major in the United Kingdom'. Rev Sci Tech, 25(1), challenge: for example, in the EC more than 293–311. 150 models are currently being used. The list of domains ranges from greenhouse Lange M., Guberti V., Thulke H.-H. (2018). gas emissions, energy consumption and 'Understanding ASF spread and emergency economy, to agriculture and structural control concepts in wild boar populations integrity assessment, to name but a few. In using individual-based modelling and spatio- addition, the majority of these models are temporal surveillance data'. EFSA Supporting run in combination with other models. Thus, Publications 15 (11): EN 1521. doi:10.2903/ they form complex networks of interaction, sp.efsa.2018.EN-1521 together with the related inputs, outputs Lange M., Siemen H., Blome S., Thulke H.-H. and assumptions. Capturing this knowledge, (2014). 'Analysis of spatio-temporal patterns and communicating it in an understandable of African swine fever cases in Russian wild manner to both scientists and policy boar does not reveal an endemic situation'. makers, will not only foster transparency Preventive Veterinary Medicine 117: 317–325. of model use, but enable collaborative and interdisciplinary research that serves cross- Moltke Jordt A., Lange M., Kramer-Schadt S., policy issues. Nielsen L.H., Nielsen S.S., Thulke H.-H., Vejre H., Alban L. (2016). 'Spatio-temporal modeling In order to address these challenges, and to of the invasive potential of wild boar - a grasp the underlying opportunity, CC-MOD conflict-prone species - using multi-source develops and manages MIDAS, the Modelling citizen science data'. Preventive Veterinary Inventory and Knowledge Management Medicine 124: 34–44. doi:10.1016/j. System of the European Commission2.7MIDAS prevetmed.2015.12.017 2 Ostlaender, N. et al. (2019) Modelling Inventory and Knowl- edge Management System of the European Commission (MIDAS), EUR 29729 EN, Publications Office of the European Union, Luxembourg, 2019, ISBN 978-92-76-02852-9 (online), doi:10.2760/900056 (online), JRC116242.

56 contains a description of models in use by the work on the cultural challenge of sharing EC, which directly or indirectly support the a common understanding of cross-domain policy cycle, independently of whether they issues, the continuous involvement of the are developed or run by the EC or by third community of modellers and policy makers parties. MIDAS is situated on the EC Network is fostered through a dedicated and EC wide and accessible to EC staff, offering a platform Community of Practice on Modelling. where data, models, scientific publications To maximise the relevance for the users, and policy actions can be easily correlated, MIDAS puts all models into their scientific and and where these connections can be browsed policy context, by combining the knowledge and better understood. that various modellers and model users in MIDAS was first launched in 2013 as the EC have about a model in a single place, an inventory of models run by the Joint where it can be browsed and visualised. Research Centre of the EC to directly or Using data visualisation techniques helps to indirectly support EU policies3.8After two communicate the resulting complex network successful first years of use, in 2015 the to a non-technical audience, revealing the High Level Reflection Group of Information bigger picture and allowing users to identify Management, whose investigations fed into patterns about model use they might not the COMMUNICATION TO THE COMMISSION have been aware of. In recent years we added on Data, Information and Knowledge tools that allow users to generate their own Management at the European Commission graphs and reports, answering user-specific (C(2016)6626), recommended to extend questions like ‘Which models run by the EC the modelling inventory to the entire EC can be used for Climate Mitigation?’ or ‘Which and to develop it into a corporate tool. In Impact Assessments used the EU Reference the following years, under the lead of a Scenario 2016?’. dedicated Inter-service group of the EC, MIDAS was officially launched as an EC tool the scope and audience of MIDAS was thus in October 20174.fSince 2017 MIDAS is also extended to include models used by any integrated in the workflow for EC impact EC services to support the policy cycle. This assessments, as the revision of the Better now also included models run by external Regulation Toolbox59requests that any model organisations, where the EC used the results used in an impact assessment should be e.g. in the context of an impact assessment. described in MIDAS. In the course of step- When formulating the vision and scope for an wise extension of the MIDAS audience, in EC wide modelling inventory we were aware 2019 parts of the system have been opened that this was not just about designing a to the European Parliament (EP)6.10 technical solution. Instead, great institutional, The information and reports generated by cultural and resource challenges lay ahead MIDAS on models and model use can be of us: scientists and policy makers had to used in practice in documents and reports, invest time and resources to share their such as the compulsory annex introduced by knowledge in an understandable manner and the Better Regulation Agenda711for impact across different domains, with us and with assessments on ‘Analytical models used in their peers, and be ready to maintain the the preparation of the impact assessment’. information updated as long as the models are in use. 4 https://ec.europa.eu/jrc/en/event/conference/laucnh-cc-mod last access: 28th June 2019 The institutional challenges have been 5 European Commission (2017) Better Regulation Toolbox, accessible at: https://ec.europa.eu/info/better-regulation-tool- tackled through a solid governance structure, box_en, last access: 28th June 2019 involving all EC services using modelling 6 MIDAS has been officially opened to the European Parliament results to support the policy process. To during the Science Meets Parliaments event 2019. The opening was done under the umbrella of the Interinstitutional agree- 3 Ostlaender, N., Bailly-Salins, T, Hardy, M., Perego, A., Fri- ment of 13 April 2016 on Better Law-Making (Official Journal is-Christensen, A. dalla Costa, S., (2015) 'Describing models in of the European Union L 123, 12.5.2016, p. 1–14) context – A step towards enhanced transparency of scientific 7 European Commission, accessible at: https://ec.europa.eu/ processes underpinning policy making', International Journal of info/law/law-making-process/planning-and-proposing-law/bet- Spatial Data Infrastructures Research, Vol.10, 27–54. ter-regulation-why-and-how_en , last access: 28th June 2019 57 This could effectively contribute to the This makes MIDAS an important corporate recommendation to enhance transparency and interinstitutional tool to use, reuse and on data, assumptions, methodology and document models in a proper way, directly results and to keep impact assessments contributing to the dissemination and use of understandable and useful for decision- sound methodology underpinning the EC’s makers, as suggested in the academic debate Better Regulation policy. on Better Regulation8. 12

8 Listorti, G. et al. (2019) The debate on the EU Better Regu- lation Agenda: a literature review, EUR 29691 EN, Publications Office of the European Union, Luxembourg, 2019, ISBN 978-92- 76-00840-8, doi:10.2760/46617, JRC116035

58 Parallel sessions 2

Integrated assessment modelling

Room 0.B 26 November 14:00 – 15:30

59 60 Lessons from the use of the GAINS control options on the full range of air model to inform clean air policies pollutants. GAINS can determine least-cost portfolios of these measures that achieve in Europe prescribed targets on health, vegetation and Amann M., International Institute for Applied ambient air quality in the most cost-effective Systems Analysis (IIASA) way and assess the resulting distributions of costs and benefits for different Member Since 1995, the Greenhouse gas — Air States and economic sectors. pollution Interactions and Synergies (GAINS) model developed at the International Institute Inter alia, the GAINS model has been used for Applied Systems Analysis (IIASA) has been for the negotiations on the 2001 and 2013 used to inform Commission proposals and National Emission Ceilings Directives. A subsequent negotiations with the European series of studies conducted with the GAINS Institution on EU clean air policies. model for the European Commission explored the cost-effectiveness, benefits and the The long residence time in the atmosphere distributional impacts of alternative sets of the most relevant air pollutants like fine of national emission ceilings for PM , SO , particulate matter and ground-level ozone 2.5 2 NO , NH and VOC for each Member State. implies that even within cities only a rather x 3 These informed the proposals that were small share of pollutants concentrations presented by the European Commission originates from local sources. Effective to the European Institutions. During the reductions of pollution levels require subsequent negotiations, IIASA conducted coordinated emission cuts in a large region, further studies for the European Parliament often in other countries. Furthermore, fine and the European Council on the implications particulate matter and ground-level ozone of alternative policy proposals. are caused by multiple precursor emissions, including primary PM2.5, SO2, NOx, NH3 and Based on the more than 20 year experience VOC. These are caused by a wide range of with GAINS , the following lessons can be economic sectors, and the potentials for and drawn for model application to support policy costs of further cuts in these emissions vary processes: greatly among the source sectors. • As policy decisions involve choices The GAINS model provides a knowledge that will not necessarily please all base to support informed decision making stakeholders, scientific credibility (through about effective emission reduction strategies a dedicated and transparent peer review) that achieve health and environmental emerges as an absolute prerequisite improvements at least cost and maximize for the acceptance of model outcomes co-benefits on other policy areas, including among stakeholders with diverse greenhouse gas emissions. interests.

GAINS brings together information on the • The same holds for the credibility future evolution of emission generating and acceptance of input data among activities, the current state of applied stakeholders. IIASA held multiple series emission control measures and resulting of bilateral face-to-face consultations emissions for all sources, the potentials for with more than 2000 experts from all and associated costs of further emission Member States. Such efforts are unusual reductions, the chemical transformation and in the academic world and are usually not transport of pollutants in the atmosphere, rewarded by academic systems. and the impacts of ambient pollution on human health, agricultural crops and natural • Open access to model and data over vegetation. Furthermore, the GAINS model the Internet emerged as an important quantifies for each Member State the co- prerequisite for the acceptance of model benefits of more than 400 specific emission results, and the international IIASA

61 supported its role as an impartial broker IFM-CAP and MAGNET (for more details see of scientific knowledge. summary reports 2012 [1] and 2015 [2]).

• While originally the focus of the analyses To serve the policy decision-making was on the identification of cost- process, iMAP has to provide results and effective policy interventions, during the recommendations in a timely manner and negotiations attention shifted towards satisfy high standards of scientific quality the distributional aspects across Member and transparency. Close links with the States, economic sectors and even within current policy agenda have to be maintained. different groups of actors within key Furthermore, harmonised, public databases sectors). should be used whenever possible. In ex ante analysis, baselines, harmonised between • While originally the model was developed models and accepted by clients, provide the to facilitate a common multi-pollutant/ benchmark for counterfactual analysis. multi-effect/multi-country/multi-sector framing of clean air policy, the course At JRC level, iMAP has pioneered the of negotiations added co-benefits on development of modelling platforms and has multiple policy other policy areas and inspired to set up a modelling task force and the multi-level governance aspects serves as an example of a productive policy- (cities/countries/EU level) as additional JRC-academia triangle. dimensions. 2) Scenar 2030— a pre-study for the CAP • In this context, the interactions, synergies reform beyond 2020 and trade-offs with climate policies emerged as an important issue, and On 29 November 2017 the European GAINS is now used both by DG-ENV Commission published the communication and DG-CLIMA for their different policy on 'The future of food and farming' [3] which proposals. aims to modernize the CAP with a tailored rather than ‘one size fits all’ approach, with Science for an evolved Common simpler rules and a more flexible approach. Agricultural Policy — the Scenar In this context, the JRC carried out the 2030 experience Scenar 2030 study [4], which analyses the M'Barek R., Ferrari E., Genovese G., European impact on the agricultural sector of stylised Commission, Joint Research Centre scenarios, reflecting the main drivers of the policy debate and thus providing a framework 1) JRC's support to CAP policy making with for further exploration of the process of iMAP designing the future CAP.

The integrated Modelling Platform for Agro- For the analysis of the social, economic and economic Commodity and Policy Analysis environmental impacts of various options for (iMAP) started in 2005 with the idea of the next CAP, we employed the iMAP platform building up a platform to host agro-economic models MAGNET, CAPRI and IFM-CAP in an modelling tools financed by the European integrated manner covering different spatial Commission, in particular European Union scales (global, European Union, Member State, (EU) Research Framework Programmes. NUTS 2 region and individual farm levels). The Financed mainly by the Joint Research use of three different models and their (soft) Centre (JRC) and the Directorate General linkages added complexity, particularly when for Agriculture and Rural Development (DG- trying to compare results across models (e.g. AGRI), it has developed into a policy support- different commodity categories). oriented platform that disposes of a number of partial equilibrium (PE) and computable The scenarios were co-developed in a general equilibrium (CGE) models, among bottom-up, iterative process with experts those AGLINK-COSIMO, AGMEMOD, CAPRI, from the different directorates in DG AGRI. 62 The scenarios are not representing real impacts of potential efforts to ensure policy options, but underline the potential CAP is fit for today’s world: a policy that is for changes to current agri-food policies to focused on meeting the challenges of a fair address societal challenges and demands. standard of living for farmers, preserving the environment and tackling climate change. The ‘No-CAP’ scenario - removing all budgetary support to farmers - could lead to The continued support to the formal Impact strong decline in farm income by 2030, less Assessment using the same economic jobs in agriculture and bring back the EU as modelling tools of the iMAP modelling net importer of agricultural products. platform to address alternative scenarios contributed to CAP2020+ legal proposal An 'Income and Environment' scenario - and to its associated Impact Assessment [7] maintaining the CAP budget at its current with economic and environmental data and level with stricter environmental rules - could analysis, published in July 2018. result in an overall higher income (with some job losses) while avoiding an increase in A systematic tracing of the impacts of Scenar greenhouse gas (GHG) emissions. 2030 has not yet been performed. However, there are several examples of impacts (apart A 'Liberalisation & Productivity' scenario – from the CAP legal proposal): with a strong reduction in subsidies and a shift to productivity-increasing measures • Court of Auditors 'Opinion No 7/2018: and further trade liberalisation - could lead concerning Commission proposals for to a drop in farming income, job losses regulations relating to the Common and agricultural production. The lower GHG Agricultural Policy for the post-2020 emissions would be mostly offset by higher period' emissions in other regions of the world. • Spanish Ministry for Agriculture: presentation with very positive feedback Whatever policy choices are made, smaller farms are likely to be more heavily impacted • Several quotations in articles and reports by changes to regulations and subsidies. • Presentation and discussion as different academic fora 3) Scenar 2030 — impacts in policy process • Upcoming report by the Committee of the The JRC report 'Scenar 2030 - Pathways for Regions the European agriculture and food sector • JRC Annual Report 2017. beyond 2020' [4] was published, timely after the Communication on 'The future of food 4) Lessons learned and outlook and farming' (29.11.2017) [3], on the very day of the 2017 EU Agricultural Outlook Taking the whole process of the support to conference (18.12.2017), where the JRC the latest CAP reform (including Scenar 2030, Deputy Director General moderated the there are several lessons learned. Session 4 'Enhancing the performance of EU policy and farming' and could also relate to First of all, the good and trustful relationship the report. with DG AGRI played a crucial role for the success of the JRC support (there are several The main report was accompanied by a notes on Commissioner and DG level). Summary report [5] containing the key policy However, important challenges remain, such messages, as well as a comprehensive as time pressure with unexpected requests, interactive version with infographics [6] sometimes insufficient level of coordination prepared within the JRC DataM portal. (both internally and with policy DG), data issues, limits on publication of JRC works, the The JRC’s scientific insight helped complexity of integration of tools. policymakers understand the scope and

63 References Examples of Europe’s marine [1] M'Barek R., Britz W., Burrell A., Delincé J., ecosystem modelling capability for An integrated Modelling Platform for Agro- societal benefit economic Commodity and Policy Analysis Heymans S. JJ, European Marine Board (iMAP) - a look back and the way forward, EUR Skogen M. D., Institute of Marine Research, 25267 EN, Publications Office of the European Norway Union, Luxembourg, 2012, ISBN 978-92-79- Schrum C., Helmholtz Center for Materials and 23554-2 (pdf), doi:10.2791/78721 (online), Coastal Research JRC69667 Solidoro C., National Institute of Oceanographic and Experimental Geophysics, Italy [2] M'Barek R., Delincé J., iMAP, an integrated Serpetti N., Bentley J., Scottish Association or Modelling Platform for Agro-economic Marine Science Commodity and Policy Analysis - New Marine ecosystem models are an important developments and policy support 2012-14, analytical approach to integrate knowledge, EUR 27197 EN, Publications Office of the data, and information; improve understanding European Union, Luxembourg, 2015, ISBN on ecosystem functioning; and complement 978-92-79-47477-4, doi:10.2791/651649, monitoring and observation efforts. They JRC95468 offer the potential to predict the response of marine ecosystems to future scenarios and [3] https://ec.europa.eu/commission/ to support the implementation of ecosystem- presscorner/detail/en/IP_17_4841 based management of our seas and ocean. [4] M'Barek R., Barreiro Hurle J., Boulanger P., In this paper we highlight the current state Caivano A., Ciaian P., Dudu H., Espinosa Goded of the art in European Marine Ecosystem M., Fellmann T., Ferrari E., Gomez Y Paloma modelling through case studies using various S., Gorrin Gonzalez C., Himics M., Elouhichi ecosystem modelling techniques. We build on K., Perni Llorente A., Philippidis G., Salputra the case studies given in the European Marine G., Witzke H.P., Genovese G., Scenar 2030 - Board’s Future Science brief on ecosystem Pathways for the European agriculture and modelling. The case studies include best food sector beyond 2020, EUR 28797 EN, practice in food web modelling, data analyses, Publications Office of the European Union, and uncertainty testing in the marine Luxembourg, 2017, ISBN 978-92-79-73859-3 ecosystem off the west coast of Scotland, (PDF), doi:10.2760/887521, JRC108449 where climate drivers were included to address the impact of climate change on the [5] M'Barek R., Barreiro Hurle J., Boulanger P., future management of fisheries in the area Caivano A., Ciaian P., Dudu H., Espinosa Goded and using an ensemble modelling approach to M., Fellmann T., Ferrari E., Gomez Y Paloma address structural uncertainties in ecosystem S., Gorrin Gonzalez C., Himics M., Elouhichi models of the North Sea. A second case K., Perni Llorente A., Philippidis G., Salputra study in the Irish Sea shows the use of G., Witzke H.P., Genovese G., Scenar 2030 - fishers’ knowledge to address data gaps in Pathways for the European agriculture and ecosystem models, and the implications it food sector beyond 2020 (Summary report), has for policy uptake. A third case study will EUR 28883 EN, Publications Office of the address the socio-economic effects of the European Union, Luxembourg, 2017, ISBN landing obligation in the Adriatic Sea, where 978-92-79-76678-7, doi:10.2760/749027, coupled physical, biogeochemical, food web, JRC109053 and socio-economic models have shown the trade-offs that have to be made in that [6] https://datam.jrc.ec.europa.eu/datam/ ecosystem. mashup/SCENAR2030/index.html Finally, a case study from the Barents Sea [7] https://ec.europa.eu/commission/ shows how the implications of the invasive publications/natural-resources-and- king crabs in Norway has had unforeseen environment 64 consequences, both negative and positive. • Develop a shared knowledge platform All these case studies show the use of for marine models and support the ecosystem models to support policy making development of next generation models and assessment. for sharing information and capability on marine ecosystem models, building The Future Science Brief concludes that that on European initiatives e.g. the pilot there is no single model that can answer all Blue Cloud, and the marine modelling policy questions. In each case the context, framework and associated Network of specific knowledge and scale need to be Experts for ReDeveloping Models of the taken into account to design a model with European Marine Environment; the appropriate level of complexity. It is more practical to assemble several models in order • Enhance trans-disciplinary connections to reach the full End-2-End spectrum. This and training opportunities. Promote requires a transdisciplinary approach and the training that spans fundamental marine inclusion of socio-economic drivers. sciences, modelling and policy and develop an online shared knowledge Key Recommendations and actions proposed training platform to connect marine by the Future Science Brief needed to ecosystem modellers, share opportunities strengthen marine ecosystem modelling and promote inter-disciplinarity. capability include: Multi-Objective Local • Link models to observations and data. Develop models, or coupled models, that Environmental Simulator (MOLES): can incorporate the full spectrum of model specification, algorithm ocean data, including biodiversity from design and policy applications microbes to top predators. Ensure data Tikoudis I., Oueslati W., Organisation for assimilation centers (e.g. the Copernicus Economic Cooperation and Development (OECD) Marine Service, EMODnet, etc.) include all existing and emerging data streams. Use This paper presents the structure, models more actively to design ocean architecture and policy applications of observation networks; the Multi-Objective Local Environmental Simulator (MOLES), i.e. OECD’s new urban • Increase predictability through Computable General Equilibrium model with coordinated experiments and the selected microsimulation features. The model ensemble approach. Design and run is tailored to assess the performance of local coordinated model experiments, e.g. and national policy instruments that target through a common funding scheme, transport, energy consumption and land use to model uncertainty and increase in urban areas. These include, but are not model predictability. Further integrate limited to: urban density restrictions, zoning model predictions, historic data and regulations, public transport subsidies, fuel machine learning to generate sensitive and kilometre taxes, regulations in vehicle adaptive modelling tools that are more ownership and use, congestion pricing, as well representative of the complex interactions as vehicle registration, circulation and parking of evolving marine ecosystems; fees. MOLES assesses these interventions • Make marine ecosystem models more from multiple perspectives, uncovering the relevant to management and policy. potential trade-offs between different policy Increase the credibility of models objectives: environmental performance, by defining and communicating economic efficiency, fiscal and distributional uncertainties. Couple ecosystem- and balance, housing affordability and other physico-chemical- models with socio- wider benefits. The model is also designed economic drivers to include the human to capture possible synergies between urban dimension. Promote co-design between planning and transportation policies. stakeholders and modellers; 65 The policy relevance of MOLES is affected areas being those densely populated underlined by the inextricable links between with high concentrations of PM2.5. urbanisation, economic growth and modern environmental challenges. Urbanisation has The model we present in this paper allows been one the cornerstones of 20th century’s the detailed examination of policies that may vast economic expansion. For decades, the be part of the answer to these challenges. uninterrupted economic growth fuelled — and MOLES is unique in that it combines the was fuelled by – a constant increase in urban internal consistency of a Computable General population and the land uptake of cities. Equilibrium model with the additional useful Slowly, a series of interrelated challenges details of a microsimulation model. As a that exert pressure on the ability of cities CGE model, it represents the value flows to generate prosperity emerged. Local air between various sectors and stakeholders pollution and climate change are two of these in a micro-founded way. However, MOLES challenges. Rough measures by OECD (2010; abstracts from detailed representations 2015a; 2015b) reveal that 60–80% of the of industrial production sectors. Instead, it models with high resolution the real estate global CO2 emissions are generated in cities. A growing share of the urban greenhouse development sector, transport providers gas emissions originate from transport and urban households, as the behaviour of and stationary sources, such as residential these agents determines the locational and heating and cooling. Furthermore, cities span mobility patterns in an urban area. That is, the areas where the largest part of local air- MOLES differentiates housing markets by pollution is generated and dispersed, exposing location and residential type. It represents considerable fractions of population to it. The urban public transport by several modes and OECD (2012) estimates that the worldwide private transport by various vehicle types. mortality due to particulate matter (PM and That resolution facilitates the examination of 2.5 policies affecting population density, urban PM10) is expected to increase from 1 million in 2000 to over 3.5 million in 2050. Therefore, structure and mobility patterns. The CGE to mitigate the overall welfare impact of nature of MOLES facilitates the exploration climate change and air pollution, existing of feedback effects of land-use policies on policies that affect economic activity in urban transport demand and of transport policies on areas have to be streamlined, and new ones the long-run urban development. have to be conceptualized and developed. The microsimulation elements incorporated in The need to examine policies from an urban MOLES introduce behavioural margins of high perspective is expected to intensify as cities environmental and economic importance. expand. The current trends foreshadow In the model, the decision of whether to the total prevail of the city. In 2050, about own a private vehicle from a certain class 85% of the population in OECD countries is of vehicles (e.g. internal combustion engine, expected to live in urban areas. The global electric) is endogenous. Individuals decide urban land cover is projected to increase the optimal number of trips, their type (i.e. steeply: from 603 thousand km2 it was commuting, shopping and leisure) and their in 2000 to over 3 million km2 in 2050. destination locations. They also schedule the Under the business-as-usual scenario, this departure time of these trips across different expansion is highly likely to exacerbate the types of days (working days, bank holidays) problem of air pollution. For instance, OECD and across different periods of one day (e.g. (2016) projects that, in the absence of more on-peak, off-peak). Most important, travel stringent policies, the number of annual demand is modelled separately from the premature deaths due to air pollution will ownership decision. That is, individuals select increase from 3 million people in 2010 to 6–9 which of the scheduled trips will be realised million in 2060. The associated monetized by a private vehicle, public transport, soft cost will increase from USD 3 trillion in 2015 mobility (i.e. walking or bike), or combinations to USD 18–25 trillion in 2060, with the most of them. In each case, mode choice is in

66 line with the constraints imposed by vehicle Physiologically based kinetic ownership, household location and trip modelling — a bridge between destination. The aforementioned elements enable the analysis of policies that aim to biological science and public promote greener forms of urban transport. health policy They also facilitate the assessment of policies Paini A., Whelan M., Asturiol D., Worth A., that alter the within-day travel patterns, by European Commission, Joint Research Centre) moving part of the traffic from the on-peak to less congested time intervals. Mathematical models are an invaluable tool in chemical risk assessment and in medicine, The model has been used by the OECD in providing a means of assessing the safety a comparative analysis of land-use and and efficacy of chemicals in a more efficient transport policy instruments that aim at and effective manner, while respecting ethical decarbonising urban transport in Auckland concerns related to testing on animals and in (New Zealand). A second application of the humans. In particular, modelling can support model, focusing on the comparison of policies research planning and regulatory decision to curb air pollution in Santiago (Chile), is making to (a) protect consumers in several under way. The presentation will focus on the areas, such as cosmetics, food and feed, and future applications of MOLES in the contexts pesticide use; (b) provide better healthcare of Randstad (The Netherlands), Gothenburg through drug development and precision (Sweden) and Istanbul (Turkey). medicine. However, while these approaches are developing rapidly in the research domain, References there is a lack of application in the actual implementation of policies on chemical OECD (2010). Cities and Climate safety and public health. One reason for Change, OECD Publishing. http://dx.doi. this translational barrier is the disconnect org/10.1787/9789264091375-en. between modellers and decision makers, with OECD (2012). OECD Environmental Outlook modellers seeking novel ways of elucidating to 2050, OECD Publishing. http://dx.doi. and reproducing real life phenomena, org/10.1787/9789264122246-en. while decision makers are looking for simplicity, reproducibility, cost-effectiveness, OECD (2015a). Climate Change Mitigation: transparency and credibility in the models Policies and Progress, OECD Publishing, Paris. and their results. http://dx.doi.org/10/1787/9789264238787- en. The Joint Research Centre (JRC) is trying to address this challenge by analysing OECD (2015b). The Economic Consequences of the problem and promoting the use of Climate Change, OECD Publishing, Paris. http:// mathematical modelling in chemical risk dx.doi.org/10.1787/9789264235410-en. assessment and in medicine. Different modelling approaches are being developed, OECD (2016). The economic consequences of evaluated, used and promoted. Among outdoor air pollution, OECD Publishing, Paris. them are the so called Physiologically http://dx.doi.org/10.1787/9789264257474-en. Based Kinetic (PBK) models. PBK models describe the body as a set of interconnected OECD (2019), OECD Employment Outlook compartments, which represent the human 2019: The Future of Work, OECD Publishing, organs and plasma, describing the absorption, Paris, https://doi.org/10.1787/9ee00155-en. distribution, metabolism, and excretion (ADME) characteristics of a compound or a mixture of compounds within the body, with several degrees of complexity (Figure 1). Additionally they can be extended to include a description of the interaction of the chemical

67 or its reactive metabolite with the biological reflecting the need to establish safe target that triggers an adverse effect or exposure levels in the context of food safety efficacious effect. (estragole, a genotoxic substance that is a natural component of certain foods) and the This presentation will illustrate ways in which cosmetics regulation (caffeine). We will also PBK modelling can be used to support the describe a project being led by EURL ECVAM, safety assessment of chemical products in on behalf of the Commission, at the OECD to several regulatory fields (Table 1), as well promote the international acceptance of PBK as in medicine and personalised healthcare. models in chemical risk assessment. We will provide two illustrative examples

Table 1. Regulatory Fields where PBK models can be applied Field Regulation Medical products authorization for human and Regulation 726/2004 veterinary use Precision medicine EU policies are in an exploratory phase Chemical Risk Assessment – REACH Regulation (EC) number 1907/2006 Pesticides – approval of active substances Regulation (EC) No 1107/2009 Biocides – approval of active substances Regulation (EU) No 528/2012 Food & Feed Regulation (EC) No 178/2002 Cosmetics Regulation (EC) No 1223/2009

Figure 1. Schematic representation of a PBK model for estragole, a natural component of certain foods (Paini et al., 2012, TAAP, 245(1)57–66).

68 Parallel sessions 2

Qualitative and quantitative modelling

Room 0.C 26 November 14:00 – 15:30

69 70 Integrating qualitative scenarios This paper discusses lessons learnt from with quantitative modelling: three recent European projects that have explicitly attempted to integrate qualitative lessons learned from recent EC scenarios with modelling. foresight projects Ricci A., ISINNOVA – Institute of Studies for the • PASHMINA [2] built four scenarios Integration of Systems representing possible shifts in socio- economic paradigms and used models The challenges of integrating qualitative to compare their economic and sectoral and quantitative approaches in forward performances looking (FL) exercises have been extensively discussed, which has led to unveiling the • FLAGSHIP [3] developed two long term multifaceted nature of the main barriers visions of the European economy, based arising. As illustrated e.g. in [1] these range on the 3-horizon approach, and used from epistemological consistency all the way models to assess the feasibility and the to practical implementation of interfaces and timing of economic recovery and of the data availability. Cultural issues are of the achievement of the COP21 targets. essence, as FL experts and practitioners often belong to either one or the other of two main • FRESHER [4] devised four storylines communities: foresighters whose background illustrating contrasted, health-focused is mostly in Social Sciences and Humanities scenarios, and used microsimulation (SSH), and modelers whose tools of the modeling to assess their potential trade originate primarily in mathematics impact on the social burden of non and physics. Of course, SSH have over time communicable diseases (NCD). incorporated a good deal of quantitative What these three exercises have in common analysis, but models are often seen as a is (i) the goal to produce results that are mechanistic, somewhat artificial expedient to expressed (also) in quantitative terms (so as feed decision and policymakers with numbers. to speak to policy makers), but reflect deeper Modelers, on the other hand, do not always and more articulate narratives, not (or less) find it easy to recognize the value of purely constrained by the inherent rigidity of models; narrative visions. Economists, in their capacity (ii) an approach that starts by building of social scientists increasingly trained in narrative storylines and then strives to sophisticated quantitative techniques, would 'translate' them into inputs and assumptions appear to be the natural candidates for that can be fed to models; and (iii) the belief bridging the two cultures, and indeed have that the real value of models is not to predict played a major role towards their integration, (which is next to impossible when it comes to although many obstacles still remain. Efforts long term dynamics of complex systems), but have been made by qualitative foresighters to orient, which models can by showing how to include quantitative dimensions in their different dynamics of critical factors (trends, work, through e.g. DELPHI analyses, or by drivers, policies) yield different results. enriching their narratives with indicators in Altogether, the following critical issues an attempt to quantify trends, drivers and emerge. their potential impact, both however falling short of providing a practicable bridge to Discontinuities modeling. Modelers, on the other hand, are keen to feed their tools with inputs elicited Models can perform satisfactorily as long from visions and storylines developed by as they do not have to factor in major qualitative foresighters, but the variables that future discontinuities. This is also the result serve as input to their models are not always of the fact that models are mostly (and immediately traceable in the narratives of FL inevitably) built on the observation and storylines. interpretation of a known past, while socio- technical transitions are likely to modify the

71 way systems work today and generate new implemented in the NEMESIS model to paradigms that shift away from the known account for the differentiated performance past. PASHMINA, which focused on paradigm of R&I spending in ICT, intangibles, and high shifts, clearly showed that these can be education. effectively captured by mathematical models — even when substantial deviations from (Over)simplification past trends are considered – provided the Integrating qualitative and quantitative continuity of pathways and the availability FL approaches can be seen as an attempt of good data, while models are not equipped to formalize implicit (or intuitive) relations to 'automatically' simulate disruptive events between subsystems, or variables: scenarios or, even less, a substantial change in the and storylines can suggest/explore the nature of the interrelation between variables. existence and nature of such relationships, On the other hand, a structural change in while models are tasked with their paradigm entails fundamental modifications formalization in order to check they actually in behaviours, which agent-based models can stand and provide some measure of their represent and simulate. relevance. Along the process, tradeoffs must Scale be found to avoid oversimplification while ensuring that the interactions between the No model can seriously claim to be equally narrative and the equations does not lead performant at different scales of analysis to loss or distortion of information. Both (e.g. macro, meso, micro), whether this refers PASHMINA and FLAGSHIP made extensive to time, space, or/and the granularity of the use of the metamodeling approach [5] as a variables. On the other hand, the ambition middle ground that preserves the rigor of of FL is often to analyse trends observed mathematical formalization while avoiding at e.g. global level in order to understand that the complexity of fully fledged analytics their local or sectoral effects; or, the other hinders the quality of communication way around, to assess the extent that seeds between foresighters, modellers and of change detected at the micro scale can policymakers. be scaled up in the future and with what global consequences. The microsimulation References model developed and adopted in FRESHER is [1] Haegeman K, Marinelli E, Scapolo F, a promising attempt of deriving large scale Ricci A, Sokolov A (2013). 'Quantitative and results by simulating behavioural changes at qualitative approaches in Future-oriented the individual level. Technology Analysis (FTA): From combination Model flexibility and adaptation to integration?'. Technological Forecasting and Social Change. 80. 386–397. doi: 10.1016/j. If the modelling framework is 'imposed' at techfore.2012.10.002 the outset, and narratives built with the sole [2] http://www.isis-it.com/pashmina/ purpose of providing input to the chosen [3] http://www.isinnova.org/portfolio-items/ model, phenomena that are intrinsically flagship/ difficult to quantify are likely to be ignored. [4] https://www.foresight-fresher.eu/ This calls for the willingness and availability of modelers to modify/adapt their equations Towards absolute measure of to account for 'new' phenomena, or for poverty in the European Union – A a novel interpretation of existing ones. pilot study FLAGSHIP has provided several examples of how an open attitude of modelers in this Cseres-Gergely Z., Menyhert B., European respect can increase the overall value of Commission, Joint Research Centre FL. As the FLASGHIP scenarios highlighted Measuring poverty requires a definition of the growing importance of differentiated poverty, a proxy for the level of welfare of R&I investments, new functions were individuals (income-monetary and/or non- 72 monetary, consumption, wealth, etc.), and and adequate social participation. These setting a poverty line under which people poverty thresholds will then be confronted are considered poor. In Europe, monetary with microdata on disposable household poverty is usually measured with the so- income (such as the EU-SILC), to calculate called ‘at-risk-of-poverty’ indicator (AROP), the poverty rate. Importantly, the proposed which defines poverty as the share of people methodology should be applicable to be used with an equivalised disposable income for regular monitoring or EU member states, below the at-risk-of-poverty threshold. The to be scaled to the EU-level and should threshold is set at 60 % of the national enable comparisons across countries and over median equivalised disposable income. As time. argued by many researchers starting from Sen (1983), a relative poverty indicator (such Because the only methodology tested on as the AROP) is a measure of inequality at the EU scale is that of the EU Platform on the bottom of the distribution rather than Reference budgets (EURB, https://www. poverty, as it is unrelated to criteria of need referencebudgets.eu), the ABSPO project and deprivation. Existing measures focusing adopts it as its foundation. In particular, on these latter, like that of severe material it retains the targeted living standards deprivation (SMD), are not monetary. ('adequate social participation') and the mixed-method approach (using expert inputs, This prompts the policy need to improve survey data and focus group discussions) the current measurement of poverty embraced by the EURB, but proposes three from an absolute perspective. While this types of improvements: is challenging, there are several reasons why this is important. First, a monetary 1. Extension towards representativeness indicator of absolute poverty would allow The base is an evolutive, modular extension contextualising EU social indicators on income of the EURB methodology whereby poverty and financial stress, could anchor appropriate 'customized' poverty lines are the poverty threshold, and complete the calculated for the entire population and all picture. Second, this helps monitoring the potential households. Specifically, this means adequacy of income support measures and modelling also households of old age, poor effectiveness of social policies, and better health and rural living environments, as well targeting vulnerable groups and local needs. as using equivalence scales to cover non- Third, in line with the SDGs, an absolute modelled household types. poverty measurement will further help monitor and translate SDG concerning the 2. Introduction of additional elements eradication of extreme poverty for the EU. The additional element in question mostly In line with these objectives, the JRC concerns specific analyses that are only project entitled 'Measuring and Monitoring indirectly related to the core exercise, and Absolute Poverty' (henceforth ABSPO) are aimed at providing valuable context aims to develop a reference-budget based and insights to assess the validity of the indicator of absolute poverty comparable methodology. These include, among others, to existing relative measures, such as the AROP. The project is financed by DG EMPL and • cross-checking the results through a implemented by B01 of the Joint Research survey-based 'fixed-point approach' using Centre and will cover a handful Member EU-SILC; States as a pilot exercise. • collecting survey data on purchasing Specifically, the project will develop habits, store choices, price search efforts; household-specific poverty thresholds based on adequately priced consumption bundles • confronting assumed expenditure associated with minimum living standards patterns with empirical ones around the poverty line. 73 3. Extension with methodological a country, taking into account 25 individual refinements. variables in six risk areas, i.e. social, economic, security, political, demographic and These refinements aim mainly at geographical/environmental areas. All the strengthening the robustness and reliability variables used in the models have extensively of the proposed methodology, either by been used as explanatory or control variables refining certain assumption and using a richer in the conflict research literature. The information base (e.g. price statistics for datasets used are all freely accessible on the costing baskets). internet and have been compiled by diverse international organizations such as the World The delivery and completion of the project, Bank, the United Nations and academic scheduled for June 2021, incorporates inputs institutions. from several different actors: Inspired by the existing quantitative conflict 1. A core team responsible for developing risk models, the GCRI is composed of two the methodology, conflict prediction outputs; the conflict 2. Country teams that construct the baskets probability and the conflict intensity at the local level measured on a scale from zero to ten. Zero represents a no conflict situation while levels 3. National statistical offices that provide one to four cover conflicts that do not include data access on HH expenditure and prices the use of force. Levels five to ten cover violent conflicts, where force is used and 4. An advisory board to provide results in a number of battle-related deaths methodological feedback and support involving at least one state actor, depending on the number of casualties and conflicts 5. An Inter-Service Steering Group that recorded per year. steers the project within the Commission. Using a logistic regression, we measure The Global Conflict Risk Index the probability of conflict outbreak, while (GCRI) the conflict risk intensity is estimated by a Halkia M., Ferri S., Papazoglou M., Van Damme linear regression model. Both estimations M. S., Thomakos D., European Commission, Joint of conflict probability and intensity are Research Centre respectively computed at national (NP) and subnational (SN) levels, depending on the The Global Conflict Risk Index (GCRI) was actors involved and the scope of the conflict. designed by the European Commission’s Indeed, an NP conflict is defined as a civil Joint Research Centre (JRC) in an effort to war over national power with at least one bridge the gap between quantitative conflict of the conflicting parties being a national risk models developed in academic research government whereas a SN conflict can occur institutions and the needs of policy-makers to between a government (one-state actor) and prioritize actions towards conflict prevention. non-state actors, and the conflicting parties typically contest over secession, autonomy, or JRC’s conflict modelling project initiated subnational predominance. four years ago and has been developed in collaboration with an expert panel of The GCRI hereby encompasses four distinct researchers and policy-makers to become the equations, as to assess and describe the full quantitative starting point of the European conflict panorama: Union (EU) Conflict Early Warning System. 1. Probability of violent conflict at the The GCRI uses open-source data from 1989 national level (NP); to 2018 for 191 countries worldwide as 2. Probability of violent conflict at the country-year observations and it considers subnational level (SN); solely the structural conditions characterising

74 3. Intensity of violent conflict at the national Jones A., European Commission, Joint Research level (NP); Centre Sturel S., French Chambers of Agriculture 4. Intensity of violent conflict at the subnational level (SN). Policy making on agri-environmental themes across EU is challenging, especially Developed at the crossroads of science and when it is required to help the local EU policy making for conflict prevention, stakeholders facing daily soil health and policy-relevant choices are made in the land management issues. Simultaneously model development stages to support its society is required to face food security while application. During the modelling stage, this safeguarding natural resources with little combination creates certain consideration time left for policy engagement. For this that are different from a pure scientific- reason researchers’ challenge is to bridge investigation context. The GCRI shows that stakeholder’s knowledge and needs on soil these political-technical considerations can multifunctionality and land management with be and are made within scientific bounds, policy makers. The LANDMARK multi-actor safeguarding scientific rigour as well as project ran 32 workshops across five EU MS objective policy support. countries covering different pedo-climatic conditions and land uses. The purpose was to According to our knowledge, the GCRI is harvest existing knowledge and requirements the only model that calculates both the across levels: from a) local farmers and farm probability and intensity of the forecasted advisors; to b) regional/national stakeholders structural risk of a country to engage in and c) EU policy makers. Information have armed conflict at national or subnational been collected, harmonized and included: level. empirical knowledge and perceptions of This differentiation between an NP and SN soil quality, common actions and valuable model translates the type of conflict to expect tools for land management, indicators and informs policy-makers on the suitable for monitoring soil functions, policies and measures to be undertaken for conflict guidelines that could optimize the supply prevention. and demand of a range of soil functions. But the analysis and comparability of While the GCRI remains firmly rooted by this high amount of qualitative results its conception, construction and modelling in a comprehensible way, represented a methods in the European conflict prevention methodological problem. This has been Agenda, it is validated as a scientifically resolved performing a Multi Criteria Decision robust and rigorous method for a baseline Analysis (MCDA) to generate two qualitative quantitative evaluation of armed conflict risk. multi-attribute decision models (MADM) using DEXi modelling software. Two DEXi models Modelling stakeholders knowledge on knowledge and needs on soil and land and requirements on soil management have been developed and the multifunctionality across EU workshop results used as input data to model. The modelling exercise was able to represent Bampa F., Creamer R., Soil Biology Group, a large qualitative source of information Wageningen University & Research harvested in four different languages in a O’Sullivan L., Johnstown Castle, Teagasc Trajanov A., Debeljak M., Department of condensed, visual and understandable way. Knowledge Technologies, Jozef Stefan Institute The modelling results captured the different Madena K., Chamber of Agriculture Lower EU stakeholders’ perceptions, priorities and Saxony concerns across scales and stakeholders’ Spiegel H., Sanden T., Austrian Agency for level. Concepts and terminology about soil Health and Food Safety quality and soil functions differ regionally. Bugge Henriksen C., Department of Plant Implementation tools and management and Environmental Sciences, University of techniques reflect the different pedo-climatic Copenhagen 75 conditions. Modelling results showed that EU’s Horizon 2020 research and innovation all the 470 stakeholders engaged have programme under grant agreement No a medium to high understanding of soil 63521. multifunctionality, but there are knowledge gaps in terms of soil data, programmes and The politics of models: socio- shared discussions, application of policy political discourses in the instruments. Further, the results showed the modelling of energy transition and contrast between the stakeholder levels: farmer and farm advisors are focussed on transnational trade policies tools for improving local knowledge, while Capari L., Udrea T., Fuchs D., Institute of multi-stakeholders discuss policies and Technology Assessment, Austrian Academy of research solutions at regional and national Sciences level. Focusing on the second model the Bauer A., Department of Science, Technology community identified as high requirements & Society Studies, University Klagenfurt & the following: 1) financial incentives to change Institute of Technology Assessment, Austrian soil and land management and commodity Academy of Sciences risk, 2) clear, independent and validated Energy policy and transnational trade information on soil-friendly management policy are two main areas of ‘modelling techniques, 3) high quality advice for farmers, for policy’. Models in these areas fulfil a 4)farmers’ discussion groups, 5) training variety of functions, from the identification programs especially for farmers and farm and analysis of societal problems to the advisors, 6) funding for applied research and examination of different policy instruments monitoring systems of soil functionality, as and the assessment of the impacts and well as 7) more soil science in education. The costs of planned and implemented policies, results of this modelling exercise provide yet they also serve to justify and legitimize important inputs not only at EU policy agri- public action. In this role, models are not environmental making but also in for bilateral neutral tools, simply providing orientation initiatives such as EIP-AGRI. Furthermore and answers to (exogenously given) societal it can be incorporated in local and regional or political questions but have performative initiatives such as training programmes, effects for socio-political discourses. discussion groups and education courses. Modellers adopt and reframe policy In order to ensure a responsible and questions, they make relevance decisions and transparent use of data collected for policy assumptions regarding parameters, factors support the workshop information, models and scenarios to include, and, in some cases, and modelling results are available in an derive recommendations and options for online platform freely accessible to users societal and political actions. (http://landmark2020.eu/stakeholders- platform/). This decision support modelling In our presentation, we explore and discuss methodology can support evaluation of the (re)production of socio-political discourse alternatives and decision making on soil and and narratives in computational modelling land management issues. The potentiality of and simulation in the two domains, this context specific and novel approach it transnational trade policy and energy is not only to solve problems of integration transition. Methodologically, we build on of opened and closed ended surveys but quantitative and qualitative analysis of also the potential adaptation and use of academic articles. We developed two broad the methodology for other EU policy areas corpora that include scientific articles and dealing with sustainability assessment. reviews on modelling developments and/ or applications and respective results in The work is part of the LANDMARK (LAND relation to the two domains. By the means of Management: Assessment, Research, bibliometric analyses (inter alia, citation, co- Knowledge Base) project, funded by the citation, bibliographic coupling analysis) and

76 text mining approaches (i.e. co-occurrence linked to modelling energy transition in analysis and topic modelling, based on Europe,while the energy security and abstracts), we first explore the main subjects development discourse is linked to less and socio-political semantics of modelling industrialized countries and rural areas. in the two domains separately. Expectedly, Bibliometric mapping (citation, co-citation modelling energy system transitions strongly and bibliographic coupling) further indicates relates to discourses of climate change and that these discourses are nevertheless the decarbonization of the energy system. predominantly driven by Western countries This includes the socio-political narrative (particularly the U.S. and Europe). This on CO2 reduction targets with specific time leadership is reflected in the overall numbers frame (e.g. until 2020/2030/2050) that of publications per country as well as the serves as an important motivation and most relevant and central documents (in legitimization of energy system modelling. terms of citations and positioning in the Yet, energy models are also related to network maps) in the respective corpora. This another policy discourse, revolving around geographic imbalance raises questions about the provision of energy infrastructure to the political implications, particularly when rural, low income countries or geographically considering that less-industrialized countries isolated regions, which have no well- are frequently object of modelling in the established energy system so far. The domains energy transition and trade policy. normative embedding is different in these two examples. One of them is rooted in a We deepen this overview analysis with a globally acknowledged environmental crisis qualitative analysis of selected key articles (climate change) and the second is based from each area. The qualitative analysis on the 'right' of having access to energy allows additional insights in how policy issues (although it is not an acknowledged basic and missions are adopted and transformed human right, discussions exist which address in modelling exercises, how societal visions this). In our second case, i.e. modelling in and ideas find their ways into models and relation to transnational trade policies, the scenarios and whether and how ultimately text mining analyses expectedly reveal a modelling results are used to address specific strong dominance of the free trade discourse, policy actors or decisions. We find quite including references to various free trade distinct patterns across and within the two agreements. Beyond that, a range of further areas. Our analysis shows that socio-political policy objectives and narratives are present aims and discourses frequently serve as and reoccurring in the modelling community, the motivational background of modelling notably welfare, labour as well as poverty exercises, ranging from rather implicit and and inequality. As a time overlay analysis abstract reference to existing social and suggest, more recently trade policies are political discourses to explicit reference linked to environmental and energy issues. to political aims and strategies. Beyond In our presentation, we will discuss these providing the motivational background, policy narratives and their linking to specific concrete policies may feed into modelling modelling approaches in more detail. in different ways, for example as the basis for different scenarios in a simulation or Notably, the different policy discourses end points for which optimized solutions are and narratives feature a strong geographic sought. Such policy input might be defined by component in both modelling domains. the researchers themselves or might base on Thus, we find thematic clusters around the direct involvement of policy-makers (or specific countries or regions (notably Europe, other stakeholders) in the modelling process. China, India) as well as groups of countries Particularly, in the latter case, we analyze or regions (less developed, rural regions, how modelers apply distinct strategies of islands). For example, we observe that boundary management to present their the climate change discourse is strongly results as policy-relevant, while avoiding

77 suspicion of politicization. Overall, with our their development, application and scientific analysis of scientific modelling discourses, we presentation. These analyses may help illustrate how the ‘politics of models’ does experts, policy-makers and the public to not only concern their use at the science- better assess the knowledge claims and policy interface, but is already inscribed in evidence politics of computer modelling.

78 Parallel sessions 3

Scenarios, model linkages and data for policy 2

Room 0.A 26 November 16:00 – 17:30

79 80 Combining micro and macro accounting and energy flows (De Lauretis, simulations to assess the 2017). We calibrate it on an original economy- energy dataset in the form of a 14-sector distributional impacts of energy input-output table produced in collaboration transitions. Evidences from with the French Environment Agency for the French national low carbon 2025, 2030 and 2035 horizons. strategy We combine IMACLIM with microsimulation Ravigné E., Ghersi F., Nadaud F., International as a 3-step procedure. First step is to model Centre for Research on the Environment and households’ expenses trends following the Development, CIRED price and income changes. Second step is accounting for trend-breaking policies Redistributive consequences of environmental targeting the improvement of the thermal policies constitute a major issue for the public efficiency of French homes and the acceptability and hence political feasibility of electrification of private transportation. the transition towards a sustainable economy. We estimate for each household carbon- The direct regressivity of environmental abatement potentials for housing renovation policies, especially carbon taxes — regressive or electric vehicle (EV) purchase. We taxes are inversely proportional to wealth — bound the possible distributions of these is established (Combet, Ghersi, Hourcade, & investments among households by an Théry, 2009; Rausch, Metcalf, & Reilly, 2011). optimistic scenario favouring high-potential Accounting for it requires energy-transition households and a pessimistic scenario where models to consider income and consumption investors are already energy-efficient. Third heterogeneity more than they currently step is micro-accounting through household do (Rao, van Ruijven, Riahi, & Bosetti, reweighting using marginal calibration against 2017; Sánchez, 2018). Moreover, recent control aggregates (De Lauretis, 2017). demonstrations in France have underlined Reweighting ensures consistency between that rising fuel taxes mostly affect the income micro and macro distributions of socio- of workers living in low-density areas. This economics characteristics. is a reminder that the distributive impacts of energy transition policies result not only The convergence of the two numerical from income inequalities but also from other systems lies in an iterative exchange of socio-economic features that influence the variables between the micro- and macro residential and mobility energy consumptions levels. The 3-steps microsimulation procedure of households (Büchs & Schnepf, 2013). computes the reaction of the aggregate budget shares of households’ expenditures Our paper’s research objective is precisely to the income and relative price variations to investigate what socio-economic features computed by IMACLIM. IMACLIM produces its drive the short to mid-term distributive economy-wide outlook under constraint of impacts of the French government’s National these budget shares. Iterating the modelling Low-Carbon Strategy (Stratégie Nationale sequence results in numerical convergence Bas Carbone, SNBC). To that end, we develop towards economic outlooks consistently an original methodology that combines combining the integrative quality of general micro-simulation and computable general equilibrium analysis and the socio-economic equilibrium (CGE) techniques to produce depths of the household database backing economic outlooks both consistent at the the micro-simulation model. economy-wide level and disaggregated across several thousand household types Four simulations, each one removing one characterised by several hundred socio- package of policies concerning carbon tax, EV economic characteristics. and housing renovation, allow us to quantify different policies’ influence on vulnerability — Our CGE model, IMACLIM, is a static hybrid both fuel and monetary poverty. model focused on reconciling national

81 In particular, we assess whether energy- factors and UK households’ home energy, efficiency policies are adequate to bridge transport, indirect and total CO2 emissions'. the gap between the poor and the rich in Ecological Economics, 90, 114–123. https:// the transition. Gini and Atkinson index are doi.org/10.1016/j.ecolecon.2013.03.007 national-level vulnerability indicators to compare distributive performance of scenarii. Combet, E., Ghersi, F., Hourcade, J. C., & We determine horizontal inequalities based Théry, D. (2009). Carbon tax and equity: The on socio-economics features through a importance of policy design. Oxford University principal component analysis regressing Press. fuel vulnerability indicators (Berry, Jouffe, Cronin, J. A., Fullerton, D., & Sexton, S. (2018). Coulombel, & Guivarch, 2016; Hills, 2012) 'Vertical and Horizontal Redistributions from and the share of constrained expenditures a Carbon Tax and Rebate'. Journal of the in household budgets. We then use Association of Environmental and Resource multidimensional analysis to decompose Economists, 6(S1), S169–S208. https://doi. inequalities among most prominent features org/10.1086/701191 (Farrell, 2017). De Lauretis, S. (2017). Modélisation des Our original model is able to test a large impacts énergie/carbone de changements de variety of compensatory measures. Ex-post modes de vie. Une prospective macro-micro policies are mainly recycling of carbon tax fondée sur les emplois du temps. (PhD Thesis). revenues in the form of per capita rebate Université Paris-Saclay. from Social Security, an increase in all social benefits (Cronin, Fullerton, & Sexton, 2018) Douenne, T. (2018). The vertical and horizontal or a simple flat transfer targeting low-income distributive effects of energy taxes. households (Berry, 2017; Douenne, 2018). Potentially popular measures would be to Farrell, N. (2017). 'What Factors Drive lower the marginal tax rate on income, which Inequalities in Carbon Tax Incidence? could trigger Double Dividend (Rausch et al., Decomposing Socioeconomic Inequalities in 2011), or to grant tax carbon rebates based Carbon Tax Incidence in Ireland'. Ecological on the mean damages among a group of Economics, 142, 31–45. https://doi. similar households. org/10.1016/j.ecolecon.2017.04.004

The dataset is still under construction but Hills, J. (2012). Getting the measure of fuel results will be available in due time for the poverty: Final report of the Fuel Poverty conference. Review [Monograph]. Accessed on 22 mai 2019 at http://sticerd.lse.ac.uk/case/ References Rao, N. D., van Ruijven, B. J., Riahi, K., & Berry, A. (2017). Carbon taxation: Designing Bosetti, V. (2017). 'Improving poverty and compensation measures to protect low- inequality modelling in climate research'. income households. Heading Towards Nature Climate Change, 7(12), 857. Sustainable Energy Systems: Evolution or Revolution?, 15th IAEE European Conference, Rausch, S., Metcalf, G. E., & Reilly, J. M. (2011). Sept 3–6, 2017. International Association for 'Distributional impacts of carbon pricing: A Energy Economics. general equilibrium approach with micro-data for households'. Energy Economics, 33, S20– Berry, A., Jouffe, Y., Coulombel, N., & Guivarch, S33. C. (2016). 'Investigating fuel poverty in the transport sector: Toward a composite Sánchez, M. V. (2018). 'Climate indicator of vulnerability'. Energy Research & Impact Assessments With a Lens on Social Science, 18, 7–20. Inequality'. The Journal of Environment & Development, 27(3), 267–298. https://doi. Büchs, M., & Schnepf, S. V. (2013). 'Who emits org/10.1177/1070496518774098 most? Associations between socio-economic 82 Energy security assessment that could do potential damage to the energy framework to support energy system. Since the threats and disruptions are of a stochastic nature, the probabilistic policy decisions model is used for determination of probability Augutis J., Krikštolaitis R., Martišauskas L., distributions for disruption parameters, Lithuanian Energy Institute and Vytautas such as the start, duration and extent of Magnus University disruption, interruption or complete cut-off of energy supply, price increase of energy Security of energy supply plays a sources, availability of technology and other. considerable role in the economic growth This enables to generate a set of disruption and social welfare in any country and is scenarios that is used for modelling of energy an integral part of the Energy Union (EU) system. strategy [1]. Unexpected disruptions in energy systems can have both an economic and In the second step, various scenarios of social cost. As a result, energy security has prospective development of energy system become a key theme not only in the EU, but are modelled using energy system model also worldwide long time ago. The latest implemented in the Open Source Energy strategic European Commission’s (EC) policy Modelling System (OSeMOSYS) [3]. These documents [2] emphasize the importance of scenarios are modelled with a set of diversifying sources of energy and ensuring stochastic disruption scenarios, where values energy security as well. However, to support of disruption parameters are determined energy policy-making, the framework for using the above mentioned probabilistic energy security assessment can be seen as a model. As the main outcome of this model necessary measure. is disruption consequences that include the energy cost increase and possible amounts of The main objective of the presented unserved energy due to disruptions. framework is the assessment of energy security for analysed countries using energy In the third step, ESC is determined as system modelling approach. The framework is an integral characteristic of disruption based on the mathematical model for future consequences: unserved energy and energy perspective of energy security coefficient cost increase. ESC is used to measure energy (ESC) for different development scenarios of security defining its level in the scale from energy systems. It allows to assess energy 0 to 1. It enables to compare various energy security in terms of energy systems resilience development scenarios in terms of energy to disruptions and consists of several steps, system resilience to disruptions and to assess each of which can be described by the the impact of individual energy projects on following models: energy security.

• probabilistic model for the formation of The framework was applied to energy system stochastic disruptions of energy system of Lithuania to measure how ESC relates to and their parameters; different energy development projects in the • mathematical optimisation model for future perspective within three scenarios. modelling of disrupted energy system One of the most important energy security scenarios; assurance requirements is the capacity of the energy system to resist possible disruptions. • energy security metric employed to Fig. 1 demonstrates the average ESC in each measure energy security via energy year for analysed scenarios. security coefficient, which indicates the level of energy system resilience to In 2015, liquefied natural gas (LNG) terminal disruptions. has had a significant impact upon Lithuanian The first step is designed to the identification energy security, as it diversifies the supply of of threats to energy security. Each threat natural gas and has removed the threat of can realise in any energy system disruption total dependence of Lithuania on natural gas 83 supplied from Russia, for which the country the European Continental Network (ECN) is had to pay a monopoly price; besides, the implemented and related to that the second threat of political pressure has also been power connection line with Poland is opened. softened. Disconnection of the Baltic power system from synchronous work with the IPS/UPS and New power connections with Sweden synchronization with the ECN is mandatory and Poland in 2016 have exerted a measure for energy security assurance in positive impact upon ESC mainly due to the Baltic States. This would prevent from improved resilience of power system to a possible total 'black-out' of the power electricity supply and import interruptions network of the Baltic States or unreliable (diversification of electricity import and work of the network and would remove market). possible geopolitical threats from the Eastern countries, which might manifest themselves In 2020, Gas Interconnection Poland- through disruptions in the power system. Lithuania (GIPL) is foreseen as one of The frequency of the electricity systems the projects which can contribute to the of the Baltic States is currently controlled assurance of national energy security. from a central dispatch centre in Russia. Furthermore, in 2016-2021 some old natural Such possible geopolitical threat would be gas and oil technologies end exploitation. For eliminated after the synchronization. Once the this reason, the ESC does not reach higher Baltic States synchronize with the ECN, they level in 2020 as can be expected due to GIPL. will not only operate their systems on that On the one hand, new gas interconnection region’s frequency, but also apply its common diversifies natural gas supply sources and rules. Additional energy security measures routes, integrates gas market of isolated might be increasing the capacity of power Baltic States into the common EU gas market, lines with other countries, RES development ensures natural gas supply security and or other. reliability in Lithuania and may contribute to the rational use and availability support of The developed framework might be seen as LNG terminal. On the other hand, with the a supporting tool for energy policy makers to closure of old power units energy system has see an integrated picture of energy security become more vulnerable due to the provision measures. of proper electricity reserve and technical disruptions. References:

A significant increase of ESC is observed in [1] Building the energy union. European 2025, when the synchronization of Estonian, Commission. 2019. https://ec.europa.eu/ Latvian and Lithuanian power systems with energy/en/topics/energy-strategy-and-energy-

Fig. 1. ESC in the analysed scenarios for energy system of Lithuania

84 union/building-energy-union (accessed 20 The Portuguese Roadmap for Carbon June 2019). Neutrality (RNC2050) was developed to find possible pathways for all Portuguese [2] Energy security. Diverse, affordable, and socioeconomic sectors to achieve zero net reliable energy. European Commission. 2019. carbon emissions by 2050. The RNC2050 https://ec.europa.eu/energy/en/topics/energy- aims to explore feasible technologies, while security (accessed 20 June 2019). pursuing economically viable and social acceptable strategies to attain national [3] M. Howells, H. Rogner, N. Strachan, decarbonisation in the whole sectors. As such C. Heaps, H. Huntington, S. Kypreos, A. its results are transversal to all economic Hughes, S. Silveira, J. DeCarolis, M. Bazillian, sectors from less problematic emissions to 'OSeMOSYS: the open source energy modeling more complex subsectors, like road transport system: an introduction to its ethos, structure decarbonisation. and development', Energy Policy 39 (2011) 5850–5870 In the RCN2050 context, three distinct core socio-economic scenarios have been Decarbonising transports in designed, describing three different visions Portugal up to 2050: possible of the near future country’s development pathways (e.g., production profile, population evolution, spatial planning, energy and climate). These Tente H., Dias L., Monjardino J., Fortes P., visions were produced for the period 2020- Ferreira F., Seixas J., CENSE — Center for 2050. The configuration of these different Environmental and Sustainability Research, trajectories resulted from prospective Universidade NOVA de Lisboa analysis, based on macroeconomic, Portugal publically affirmed its firm demographic and technological scenarios, commitment to be neutral in GHG emissions and also on active participation of several by the end of the first half of the century, stakeholders from different sectors. The during the twenty-second session of the prospective analysis focused on the energy Conference of the Parties (COP 22). The Paris system, buildings, mobility and transport, Agreement (UNFCCC, 2015) opened a new agriculture, land use and forests, and waste, phase in global climate action identifying being considered options and different the need to achieve carbon neutrality by the integration levels of circular economy middle of the 21st century as a condition to (transversal to most economic sectors). ensure that, by the end of this century, the The technological linear optimization model increase in the global average temperature is TIMES_PT was used to evaluate the future held below 2 °C, compared to pre-industrial penetration of different cost-effective levels, and to pursue efforts to limit the technological mobility options in distinct temperature increase up to 1.5 °C. Reaching scenarios. Several updates were added to this goal is particularly challenging for the the model technology database being the Portuguese transport sector which accounts inclusion of shared and/or autonomous for 16.2 million tons of greenhouse gases vehicles one of the most important. The (GHG) emissions, approximately 23% of the TIMES_PT model represents the Portuguese country total, roughly the same emissions energy system and its possible long-term level as power generation. Most emissions developments. The model incorporates (96%) are due to road transport subsector, a high number of modern technologies being the remaining share associated related with the different components of with other mobility forms (i.e. aviation, the energy systems. They are characterized railways and navigation). Transports are in terms of technical (e.g. energy efficiency, also contributing for local Air Quality (AQ) lifetime) and economic (e.g. investment problems in cities, especially due to NO2 and and operation costs) features. The mobility PM10 which tend to exceed AQ limit values in demand for each scenario was defined with some traffic hotspots.

85 a bottom-up approach, considering urban trajectories) highlighting part of the need and peri-urban transport conditions, smart for circular economy approaches also in city expected developments, and impacts the mobility sector. Regarding heavy duty of mobility megatrends. Different scenarios transport, the introduction of new fuels were defined considering distinct levels of (H2) or technologies (electric road systems) future configurations and mobility trends (i.e. depend on the implementation of basic autonomous vehicles and sharing schemes). infrastructure. Major efficiency improvements, A soft link was developed to estimate the in all mobility modes, cause reductions in impacts of carbon neutrality scenarios on air energy intensity between 2005 and 2050 pollutant emissions. (of 74% of energy consumed per pkm and of 83% of energy consumed per tkm). One of the main findings is carbon neutrality technical viability in Portugal by 2050. In 2 of the 3 designed scenarios Portugal Emission reductions imply significant levels reach carbon neutrality by 2050. This goal of renewable sources on final energy brings also strong benefits for air pollutant consumption, reaching 85–90% by 2050, emissions as important co-benefits. Major in particular in the production of electricity, emission reductions in transport sector and consequently on road transport, which are accomplished for NOx (95%) and CO reaches full electrification by 2050. Transport (99%) in 2050. The 2030 air pollutants and mobility is the sector with the greatest targets fulfilment is highly dependent on the energy profile transformation, going from the expected electric vehicle penetration by that most representative sector (37%) to the one horizon. with least expression in terms of final energy consumption (19%). Modelling the macroeconomic and employment Impacts of future Results highlight a fast decarbonisation of transport sector (-98% of GHG emissions in mobility disruption Scenarios 2050, compared to 2005), even with higher Tamba M., Saveyn B., Krause J., Grosso M., demand for mobility in all modes. Traditional Duboz A., Ciuffo B., Fana M., Fernandez-Macias fossil fuels are progressively replaced by E., European Commission, Joint Research electricity, biofuels and H2 (accounting for 93% of the energy consumption in 2050). 1. Introduction Electricity is preponderant in most of the In the context of digitalisation and means of transport (70% of the energy decarbonisation, the transport sector will consumption in 2050). Electric mobility is a undergo radical transformations in the key driver for full transport decarbonisation, decades to come (Alonso Raposo et al., but shared mobility can produce additional 2019). In particular, three key trends hold efficiency gains. a significant disruptive potential for road The passenger mobility demand increase transport: Automation, Connectivity and is ensured both with more public transport Electrification (ACE). In this study, we examine and with the generalization of individual how these new trends in road transport shared and/or autonomous electric transport. may have wider societal implications for Buses have a great potential for electric the European Union, for example in terms mobility, which accounts for about 1/3 of of jobs and CO2 emission reductions. Using demand in the decade 2030–2040 and an energy-environment-focused dynamic 2/3 of demand in the decade 2040–2050, Computable General Equilibrium model, we with the remaining consumption being simulate scenarios with varying deployments assured by biofuels in 2050. Shared mobility of ACE out to 2050 to analyse the potential plays a major role on this energy transition macro-economic, employment and (although uncertainty remains in deployment environmental impacts in a global context.

86 2. Methodology captured through the dynamic adjustment of the vehicle-related capital stock on the The modelling is based on the analysis of production-side of the economy and of the three main scenarios, designed to capture vehicle-related durable good consumption on alternative hypothetical evolutions of road the household side; shifting from conventional transport, and to isolate the impact of manufacturing to the new vehicle type. electrification from those of automation and Moreover, changes induced by the deployment connectivity: of new vehicles, from fuel switching and • The baseline includes limited fuel efficiency, to reduced labour intensity, electrification and no penetration of to varying maintenance requirements, are connected autonomous vehicles (CAVs) in introduced in the model as changes in input- road transport, in line with the modelling output coefficients, and changes in the produced for the EU strategy for long- household consumption matrix, which relates term greenhouse gas emissions reduction industrial outputs to consumption goods (Keramidas et al. 2018). such as transportation. For example, the consumption matrix can be modified to adjust • In an electrification scenario, we assume fuel use and fuel type within the operation a substantially larger electric vehicles of a private vehicle. These operational shifts (EV) penetration, both in freight and are endogenously linked to the deployment passenger transport, in line with the of new vehicles in the stock. A combination EU long-term strategy to limit global of energy system models scenario modelling warming to 2C. results and literature review are used to • An automation-connectivity scenario, quantify and parameterise these changes. where connected and autonomous vehicles (CAV) are assumed to enter the We use the extended model to examine the vehicle fleet as early as 2020 and reach macro-economic and environmental impacts high penetration by 2050. of these mobility disruption scenarios. In addition, a number of sensitivities are • A scenario combining electrification and conducted on key assumptions (e.g. battery automation-connectivity. costs reduction and comparative advantage These scenarios include how the penetration in trade, labour input requirements). Finally, of these technologies could also alter the modelling results are used to further operation of the stock, e.g. in terms of fuel investigate the impacts of ACE on jobs, linking consumption, maintenance costs, etc. We the sectoral employment results to a new analyse these scenarios using the JRC- framework disaggregating the employment GEM-E3 model, a global multi-sectoral structure by occupation, skills and tasks. general equilibrium model, extensively used in EU climate policy research (see for example 3. Results and policy relevance Vandyck et al. 2018). We analyse the general equilibrium effects The JRC-GEM-E3 model is modified in two of the scenarios described above in a global important ways. First, the manufacturing modelling framework with endogenous of vehicles is split between two sectors international trade, explicit supply chain of activity: the existing automotive representation and accounting for CO2 sector and the manufacturing of electric and other GHG emissions. The modelling vehicles. The cost structure of this new work is ongoing and final results will cover vehicle manufacturing sector is derived both aggregate indicators such as GDP, from engineering studies of components exports and emissions but also identify to capture the difference in production sectoral implications (output, investment, processes compared to conventional employment). internal combustion engine vehicles of These results will provide relevant insights today. Second, the deployment of vehicles is for the policy questions currently under 87 investigation in the European Commission Vandyck, T., Keramidas, K., Kitous, A., Spadaro, at the intersection of transport, climate and J., Van Dingenen, R., Holland, M., Saveyn, B, energy issues, for example in the context of 2018. 'Air Quality Co-Benefits For Human the Low-Emission Mobility Strategy (European Health And Agriculture Counterbalance Costs Commission 2016). The study also builds To Meet Paris Agreement Pledges', Nature capabilities to assess policy initiatives to Communications, 9 (4939). doi:10.1038/ foster employment and foresee and mitigate s41467-018-06885-9 any possible negative social impacts. In particular, it will provide a first analysis of Global trends of methane the potential economy-wide impacts of these emissions and their impacts on new transport technologies in the EU, while ozone concentrations the focus on employment will enable a deeper understanding of the transition required in Van Dingenen R., Dentener F., Crippa M., terms of occupation and skills. Guizzardi D., Janssens-Maenhout G., European Commission, Joint Research Centre References Background Alonso Raposo, M. (Ed.), Ciuffo, B. (Ed.), Ardente, F., Aurambout, J-P., Baldini, G., Methane (CH4) is the 2nd most important Braun, R., Christidis, P., Christodoulou, A., anthropogenic greenhouse gas after carbon Duboz, A., Felici, S., Ferragut, J., Georgakaki, dioxide. Since the pre-industrial era, methane A., Gkoumas , K., Grosso, M., Iglesias, M., concentrations have more than doubled, Julea, A., Krause, J., Martens, B., Mathieux, F., and at present sources related to human Menzel, G., Mondello, S., Navajas Cawood, E., activities are about 50% larger than natural Pekár, F., Raileanu, I-C., Scholz, H., Tamba, M., ones. After a period of stagnation, methane Tsakalidis, A., van Balen, M., Vandecasteele, I., concentrations are increasing again since The future of road transport - Implications of the last decade, and by 2020, may reach automated, connected, low-carbon and shared levels that match the most pessimistic mobility, EUR 29748 EN, Publications Office projections used in the IPCC AR5 report. of the European Union, Luxembourg, 2019, It is often forgotten by policymakers that ISBN 978-92-76-03409-4, doi:10.2760/9247, methane is also an important precursor of JRC116644. ozone (O3) in the troposphere. Ozone itself European Commission (2016), A European is a greenhouse gas and short-lived climate Strategy for Low-Emission Mobility. forcer, but it is also an atmospheric pollutant Communication from the Commission to responsible for harmful impacts on human the European Parliament, the Council, the health and damage to crops and vegetation European Economic and Social Committee and for which air quality standards have been and the Committee of the Regions, established. In various parts of the world COM(2016) 501 final environmental policies aim to reduce ground level ozone, but there is a risk that increasing Keramidas, K., Tchung-Ming, S., Diaz-Vazquez, methane emissions will counteract those A. R., Weitzel, M., Vandyck, T., Després, J., regional efforts. Because methane stays Schmitz, A., Rey Los Santos, L., Wojtowicz, about 10 years in the atmosphere, chemical K., Schade, B., Saveyn, B., Soria-Ramirez, A., mechanisms that lead to widespread ozone Global Energy and Climate Outlook 2018: formation involve methane sources from Sectoral mitigation options towards a low- everywhere in the world and mitigation emissions economy — Global context to efforts must become a global goal. Methane the EU strategy for long-term greenhouse mitigation is therefore an issue with gas emissions reduction, EUR 29462 EN, relevance across policy fields, requiring supra- Publications Office of the European Union, national coordination. Here we explore for Luxembourg, 2018, ISBN 978-92-79-97462- an ensemble of future scenarios, the impact 5, doi:10.2760/67475, JRC113446. of projected methane emission trends until

88 2050 on background ozone, and its impacts By contrast, optimistic sustainability on human health and crop yields. scenarios, such as those that target the

2° Paris Agreement goals, projected CH4 Method emission reductions of up to 50%, to 180– 220 Tg CH yr 1 CH by 2050. Such scenarios Understanding the methane contribution 4 4 assume structural changes in the energy, to present and future ozone levels, as well waste and agricultural sectors, together with as the multiple co-benefits of mitigation the implementation of all currently available measures requires a global modelling emission abatement technologies. O may approach. The JRC Fast Scenario Screening 3 decrease by 2 ppb (compared to 2010), Tool (TM5-FASST) is a reduced-form source- saving worldwide 30,000 (-9%) to 40,000 receptor air quality model for the global (-12%) lives. domain (European Commission, 2016; Van Dingenen et al., 2018). The tool uses The maximum CH4-O3 mitigation potential emission-normalized pollutant concentration would be given by a situation without response fields (including all ozone anthropogenic CH4 emissions: global ozone precursors) that were pre-computed with the damage to crops would be reduced by 26%, full chemical transport model TM5 (Krol et and O3 related mortality by 20% compared al., 2005). TM5-FASST linearly scales these to present day ozone impacts. For Europe normalized response with actual CH4 emission we estimate a potential damage reduction changes to obtain global CH4-induced ozone by 40% and 34% for crops and health concentration and ozone exposure metrics respectively. In terms of emission reductions, at a 1°x1° spatial resolution. The exposure the optimistic scenarios reach about 2/3 of metrics are then used to compute impacts this maximum potential. on human health and agricultural crops. We apply the TM5-FASST screening tool to Conclusions evaluate an ensemble of emission scenario families for the years 2030 and 2050, each Impacts of short-lived air pollutant emissions family represented by a low mitigation, a (PM2.5, NO2) are strongly linked to the stringent mitigation and a middle-of-the road emission location and emission controls are pathway. largely driven by countries’ self-interest. In contrast, the transboundary nature of

Results the air quality impacts of CH4 emissions justifies international cooperation to reduce Unabated, global anthropogenic CH4 these emissions. This cooperation may be emissions could increase by 35 to 100% found under the UNFCCC Paris Agreement, (from ca. 330 Tg CH4 yr 1 in 2010 to 450– or regional conventions such as the UNECE 650 Tg CH4 yr 1) by 2050 for a range of Convention Long Range Transport of Air pessimistic scenarios. For these pessimistic Pollution (see Maas and Grennfelt, 2016), or scenarios health-impact weighted O3 could the Arctic Council. Aakre et al. (2018), using rise by 2–4.5 ppb globally, causing 40,000 the TM5-FASST tool also utilised in this work, (+12%) to 90,000 (+26%) more O3 premature argue that collaborations between 3 to 6 key deaths compared to present. regions (‘clubs’), may realize a substantial portion of the global mitigation potential, and Intermediate CH emission reduction 4 overcome some of the difficulties associated scenarios, for instance those compatible with with global agreements. the emission reduction commitments included in the nationally determined contributions In this context the availability of scientific of the signatories of the Paris Agreement, assessment tools, encompassing the full cycle would bring down CH emissions substantially 4 from CH4 emissions to impacts as well as the compared to the pessimistic scenarios, with evaluation of economic costs and benefits the exposure of the global and European is essential in building trust and confidence population to ozone remaining at 2010 levels. between the collaborating partners. 89 CH4 and O3 are both important greenhouse Krol, M., Houweling, S., Bregman, B., van den gases. By 2030, ambitious CH4 emission Broek, M., Segers, A., van Velthoven, P., Peters, reductions, of which many come at zero or W., Dentener, F. and Bergamaschi, P., Atmos negative cost, could close 10 to 20% of the Chem Phys, 5(2), 417–432, 2005. emission gap identified by the United Nations Environment Programme (2019) between the Maas, R. and Grennfelt, P., Eds.: Towards total commitments in the national determined Cleaner Air. https://www.unece.org/fileadmin/ contributions of the signatories to the Paris DAM/env/lrtap/ExecutiveBody/35th_session/ Agreement and the emissions needed to Informal_document_2_Draft_CLRTAP_ reach an end-of-the century 2 °C target. Policymakers_Summary_Report_v05-04-16. pdf 2016. References United Nations Environment Programme: Aakre, S., Kallbekken, S., Dingenen, R. V. and EMISSIONS GAP REPORT 2018, UNEP, S.l., Victor, D. G., Nat. Clim. Change, 8(1), 85, 2018. 2019.

European Commission: TM5-FASST - Fast Van Dingenen, R., Dentener, F., Crippa, M., Scenario Screening Tool, [online] Available Leitao, J., Marmer, E., Rao, S., Solazzo, E. and from: http://tm5-fasst.jrc.ec.europa.eu/ Valentini, L. Atmospheric Chem. Phys., 18(21), (Accessed 3 November 2016), 2016. 16173–16211, 2018

90 Parallel sessions 3

Integrated modeling: the case of the Food-water- energy-environmental NEXUS management under uncertainty and risks

Room 0.B 26 November 16:00 – 17:30

91 92 A strategic decision–support relating to the analysis of uncertainty and system for strategic robust inherent risks, and economic appraisal criteria to inform the robust choice of adaptation adaptation to climate change and actions. The historical performance of land systemic risks in land use systems: use systems (e.g. their impacts on climate Stochastic integrated assessment change, water availability and quality, soil GLOBIOM model quality and erosion, and biodiversity) shows that a proper analysis of interdependencies, Ermolieva T., Havlik P., Boere E., Balkovic J., synergies and trade-offs between Skalský R., Folberth C., Khabarov N., Fritz uncertainties, risks, policy measures and S., Obersteiner M., Ermoliev Y., International systems’ responses within and beyond land Institute for Applied Systems Analysis (IIASA) use systems, is crucial for sustainable land Climate change and variability are expected use systems development. to have significant and highly uncertain impacts on agricultural production and the In the absence of evidences on potential utilization of natural resources, affecting systems’ performance in new uncertain food, water, environmental, and energy conditions (i.e. climate change and increasing (FWEE) security at national and regional weather variability, structural changes and levels, with the potential for world-wide increasing systemic interdependencies), the main challenge is to design new systems spillovers. Different regions will be subject to different types of exposure. To ensure and robust policies, which enable long-term adequate agricultural production and thus mutual stability of the systems irrespective food security at the level of EU countries of the future uncertainty scenario. Therefore, and main economic regions, an analysis in the presence of inherent uncertainties, of local agricultural adaptation strategies the main issue is about designing robust have to be designed and implemented to solutions, which leave systems better-off tackle uncertainties and risks in a robust under all potential scenarios. As the variety way: such strategies include decisions on and the interconnections between LUS governmental regulations, investment in increase, the design of robust solutions has to and reform of water management, land use be based on the analysis of complex systemic practices, agrofood production patterns, and interactions and risk exposures evaluated food trade regulations (European Commission with respect to FWEE security targets. 2009). It has been recognized that climate Our talk discusses the two main issues. First, adaptation is a major concern of the Common we summarize the results of the stochastic Agricultural Policy (CAP) of the European GLOBIOM model studying synergies and Union (EU). In the proposed CAP regulations trade-offs among different measures of for 2014–2020, adaptation has gained great the new Common Agricultural Policy (CAP) prominence, with 'the sustainable use of with the aim to identify the principles that natural resources and climate action' being can ensure the most effective distribution one of the core objectives. of CAP funds to achieve optimal climate In our talk, we discuss the on-going work at change adaptation in the face of inherent IIASA contributing to FP7 Project 'Economics uncertainty and risk. The results demonstrate of climate change adaptation in Europe' numerically that a proper interdependent (ECONADAPT), Horizon 2020 Project 'Co- analysis and appraisal of the EU CAP designing the Assessment of Climate measures under uncertainty and risks can Change costs' (COACCH), and Framework speed-up the mainstreaming of climate Service Contract — Ecosystems Support change uncertainty, risk management for Copernicus and EU Space Policy. The approaches, and robust actions within the projects focus on providing user-orientated CAP implementation plans. The benefits decision-support methodologies and evidence of robust decisions are estimated with the so-called 'Value of Stochastic Solutions'. The

93 recommendations of the stochastic model An integrated environmental- have been summarized in deliverables to the economic model for robust EU Commission. pollution control under uncertainty Second aspect of the discussion focuses Wildemeersch M., Ermolieva T., Ermoliev T., on the improvement of knowledge about Obersteiner M., International Institute for global change processes though advanced Applied Systems Analysis (IIASA) data and information access services such Tang S., The Ohio State University as e.g. COPERNICUS. Any discussion of a policy implementation must be coupled Nitrogen and phosphorus are key elements with the problem of determining the level in agricultural practices for crops growth. of knowledge about uncertainties and At the same time, they are dangerous associated risks, and the appropriate potential polluters of soil, land, and waters, 'security' level reflecting strategic global contributing to poor surface and ground and local food-water-energy-environmental waters quality. In Europe, the Water norms and regulations. Continuous feedback Framework Directive and the Nitrates from the environment is one way to help Directive include direct and indirect measures reduce uncertainty. The potential for to control the use of nutrients in agriculture adaptive improvement of our knowledge and to reduce nutrients leaching from (or learning) about inherent uncertainty in agricultural land to surface and ground combination with strategic robust decisions waters. In our talk, we present an overview can considerably reduce (or increase) the and compare recent studies on developing costs of required adaptation and risks and applying methodologies for nutrients management. Using stochastic GLOBIOM accounting and best management practices model, we illustrate how to provide an (BMPs) to minimize agricultural nutrient integrated assessment of short- and long- losses to the environment in the EU and North term socio-economic and environmental America. Although a wide variety of methods quantifiable and non-quantifiable benefits of exist, research predominantly concentrates improving the knowledge about uncertainty on deterministic frameworks with an through large data services, e.g. as implicit assumption that the ecosystem is COPERNICUS. Quantitative estimates can deterministic and the social planner/farmer reflect Avoided losses or/and damages, Value faces no uncertainties. While convenient, of Information, indicators of knowledge these assumptions may lead to overly spill-overs and potential increasing returns simplistic representations of reality, given from COPERNICUS services and data in that ecosystems are inherently stochastic. various fields (i.e. industries, agriculture, Uncertainty may be particularly important biodiversity protection, air quality, emergency to be captured in coupled human-natural management, crisis prevention, preparedness systems. There is legitimate concern that and response, tourism, etc.), which justify deterministic models calibrated with average the creation and growth of scientific and parameter values may lead to suboptimal business projects using data and information policy recommendations. We therefore verify access services. At the same time, continuous the hypothesis if ignoring the stochastic updating for new and improved datasets nature of nutrient emissions leads to loose poses a challenge for the modelling guidelines for fertilizer application and water and decision support tools provided the and soil pollution control measures. In the requirements for advanced data processing, talk, we focus on phosphorus accounting and assimilation and strategic decision support BMPs modelling. The study examines how analysis. weather uncertainties affect optimal choices

94 of BMPs over adaptation by building on an pollution control policies. We illustrate the integrated economic-agricultural-hydrological framework with two case studies in the UK model of BMPs and combine it with the and US. recent development in decision making under uncertainty relying on methods of nonsmooth Policy impact on biofuel or stochastic optimization. The environmental electricity production from security constraints are introduced in biomass in Europe the form of quantile-based probabilistic constraints, assigning a realistic reliability Leduc S., Patrizio P., Kraxner F., International level to meet the specified environmental Institute for Applied Systems Analysis (IIASA) targets. This stochastic optimization Mesfun S., RISE Research Institutes of Sweden framework for phosphorus management is Staritsky I., Elbersen B., Wageningen University able to provide recommendations regarding & Research Lammens T., BTG Biomass Technology Group robust BMPs under uncertain climate change and stochastic weather events. The model As a low-carbon energy resource and a has been applied in several EU projects, carbon management mechanism, biomass and the results have been negotiated is expected to play essential role in the with representatives from EC DG Agri, transformation of European energy sector Environment, Clima, emphasizing the need for under stringent climate change mitigation risk-adjusted pollution/nutrients accounting accords. In the last two decades, biomass has and BMPs recommendations. attracted a growing interest for developing conversion technologies that generate Comparing the policy guidelines with bioenergy and biofuels using different types a deterministic model using expected of non-food feedstock like agricultural and precipitation and emission levels, we forest residues. The efficient utilization of find that incorporating stochasticity modern bioenergy technologies will be of high results in qualitatively different policy importance for the future development of the recommendations for the BMP adoption European energy supply system, especially path. The results from practical case studies in balancing out fluctuations in energy caution that neglecting stochasticity can generation from other renewable sources like result in insufficient mitigation strategies. We wind and solar. quantify as well how much the stochastic policy guidelines outperform the deterministic The present study investigates the long-term solution in terms of profit, and show that potential of non-food and non-industrial significant gains can be achieved from biomass feedstock for energy purpose using robust solutions. More generally, this in Europe and a strategic model-based work shows that the adoption of stochastic techno-economic feasibility of converting optimization methods in environmental the identified feedstock to energy products. economics provides robust policy A geographically- explicit techno-economic prescriptions under uncertainty and risks, and model, BeWhere has been developed at the that deterministic models based on traditional European scale at a 40km grid size, to assess scenario analysis cannot provide adequate the potential of bioenergy and biofuel from decisions in the presence of uncertainty. In non-food feedstock. The model is a mixed addition, the proposed method allows decision integer linear program, it is based on the makers to select an acceptable risk level minimization of cost and emissions of the along with the corresponding cost of violating full supply chain from feedstock collection to the environmental constraint. By monetizing the final energy product distribution to the risk levels, the framework provides a versatile consumers. The model identifies the optimal tool for decision support with powerful policy bioenergy production plants in terms of implications. The framework demonstrates spatial location, technology and capacity. The urgent need to account for uncertainties feedstocks of interests are woody biomass and risks while deciding on environmental (consisting of eight types from conifers and 95 non-conifers) and five different crop residuals. optimal solutions across sectors and regions. For each type of feedstock, one or multiple Detailed sectorial and regional models have technologies can be applied for either heat, traditionally been used to anticipate and electricity or biofuel production. plan desirable developments of respective sectors and regions. However, solutions The model is run for different policy tools that are optimal for a sub-system may turn such as carbon cost, biofuel production out to be infeasible for the entire system. incentives, or subsidies to name a few. The In this talk, we discuss a new modelling optimal mix of technologies and biomass approach enabling the linkage of detailed needed are optimized to reach a production models of subsystems under joint resource cost competitive against the actual reference constraints. The models act as 'agents' that system which is fossil fuel based. At the communicate with each other via a 'central same time, the impact of the trades on the hub' (a regulator or a planner). In this way, European bioenergy potential is investigated they continue to be separate models, and in three additional scenarios: no trades, different modeling teams do not need to free trades and trades not higher than the exchange full information about their models domestic biomass consumption. — instead, they only need to harmonize the inputs and outputs that are part of the The results show that a maximal amount joint resource constraints. In other words, of emissions substituted (about 60 MtCO2 in this approach the agents operate under a year) would be reached at high carbon asymmetric information. The applicability of price. Allowing trades of feedstock between the developed approach is demonstrated for a the European countries would decrease the case study that focuses on the food-energy- emissions by half compared to a scenario water-environmental nexus of agriculture for which trades are not allowed. The trades and the coal industry that are competing for allow the countries with a high share of fossil limited water and land resources. fuel in their energy mix to increase their bioenergy production and at the same time The approach for linking models is based decrease their emissions. This means that on an iterative process of non-smooth the European countries should collaborate to stochastic optimization converging to the decrease the emissions of the countries which socially optimal solution. It does not require are still heavily depending on fossil fuel. models to exchange full information about their specifications. The 'resource quotas' for Futher reading each sector/region and each resource are BeWhere: www.iiasa.ac.at/bewhere recalculated by sectors/regions independently by shifting their current approximation in Robust food-energy-water- the direction defined by the corresponding environmental security sectorial/regional dual variables from the primal sectorial optimization problem. In this management: linking distributed way, we avoid a 'hard linking' of the models in sectorial and regional models a single code. This also preserves the original models in their initial state for other possible Ermoliev T., Ermolieva T., Havlik P., Rovenskaya E., International Institute for Applied Systems linkages. Using detailed sectorial/regional Analysis (IIASA) models instead of their aggregated simplified versions also allows for taking into account Increasing global-local, as well as sectorial critically important local details, which are interdependencies may significantly affect usually hidden within aggregate data. business-as-usual operations, even under small local disturbances. This calls for the The algorithm is being developed at IIASA development of adequate integrated models in partnership with Institutes from IIASA enabling truly integrative decision support for National Member Organizations for linking

96 national sectorial models with global models systems, which may be subject to shocks (such as e.g. IIASA MESSAGE and GLOBIOM and discontinuities. These methods enable to models). Therefore, linkage is considered as link stochastic models with known marginal not only linking regional and/or sectorial, distributions of sectorial uncertainties, into models but also, more generally, linking cross-sectorial integrated models with models may refer to different local-global joint distributions of collective systemic scales. The proposed computational algorithm risks induced by sectorial uncertainties and is based on generalized gradient methods decisions maximizing a stochastic version of invented for the optimization of non-smooth the function.

97 98 Parallel sessions 3

Modelling value chains in global trade

Room 0.C 26 November 16:00 – 17:30

99 100 Challenges of conducting policy cost of production and the Armington impact analysis using PE models: assumption of product differentiation where consumers view foreign varieties of a product the case of Brexit as imperfect substitutes for the local variety Nti F., Jones K., USDA Foreign Agricultural of that product. This assumption identifies Service the existence of two-way trade (i.e., bilateral exports and imports of goods in the same The objective of this research is to analyze products category) as the result of consumer the export potential of selected agricultural tastes. commodities in the event of a Brexit and highlight the challenges faced using PE The model is parametarized with import models for this analysis. Trade elasticities demand elasticities obtained from Ghodsi et play a central role in estimating the effects al. (2016), which are available by product and of trade agreements as well as estimating country. A semiflexible trans log GDP function the impacts of specific policy changes. approach is used to derive import demand Accurate and up-to-date trade elasticities elasticities and the UN Comtrade system– are important to trade policy makers that based data on import values and quantities rely on these key parameters for assessing across 167 countries and 5,124 products the trade impact of a policy shock. Currently at the six-digit HS level for the period 1996 the trade modeling community relies on through 2014. Export supply elasticities were trade elasticities that are outdated, heavily obtained from different sources because no aggregated, or estimated through poorly one source covers the elasticity estimates specified econometric approaches. The for all the products of interest. Elasticities of elasticities used by most trade modelers were substitutions between domestic and imported estimated from time series data that ended in goods is assumed to be equal and constant the late 1980’s, 1990’s or early 2000’s. Also, across products and countries (constant many of the General Equilibrium (GE) and elasticity of substitution). These elasticities Partial Equilibrium (PE) models are predicated were obtained from different sources on a set of import demand, export supply, because no single source covers the elasticity and substitution elasticities that are often estimates for all the products of interest incomplete and outdated. This would suggest to the Foreign Agricultural Service. These that these elasticities would not adequately sources include Reinhart and Roland-Holst reflect the structural changes that have (1992), Gallaway et al. (2003), and the Global affected the global agricultural production Trade Analysis Project database (Narayanan and food consumption landscape during the and Walmsley, 2008). However, there were past decade. many product–country pairs for which an elasticity estimate was not available. We develop the Foreign Agricultural Service Trade Analysis Model (FASTAM) which is We highlight a series of challenges in the based on the Food and Drug Administration’s development of a PE model, including Trade Impact Model, which was derived from parameterization and data issues. We the Global Simulation Model (GSIM). The GSIM explore approaches to improve the data and is a partial equilibrium model developed by parameters, including filling product–country Francois and Hall (2003) that is scalable and pairs for which an elasticity estimate was not allows for the simultaneous assessment of available. We analyze the potential impact of trade policy changes at the industry level policy changes of Brexit, namely the potential and on a global or national level. The core changes in tariffs on U.S. exports under a assumptions underlie the FAS Trade Model range of scenarios and identify opportunities are perfect competition where competition that could exist for expanding U.S exports from current rivals or the threat of entry by of selected agricultural commodities to the new rivals helps reduce prices to the marginal United Kingdom.

101 Shooting oneself in the foot? Trade to industrial goods (but excluding the war and global value chains automotive sector).

Bellora C., Fontagné L., PSE, University Paris 1 The new tariffs have indeed a direct impact and CEPII on the targeted products and countries, but global value chains, along with general Since early 2018, the United States’ equilibrium effects, trigger consequences administration has taken several measures also on third sectors and countries. These to limit US imports, in particular from are the indirect effects possibly contributing China. This has fuelled retaliation and has to imposing countries doing themselves a escalated in high trade tensions at the disservice. Indeed, global value chain linkages global level. In addition to the measures modify countries' incentives to impose already implemented, the belligerents import protection, as the optimal tariff currently contemplate two alternative routes: depends on the nationality of value-added either open new fronts (particularly in the content embedded in domestic and imported automotive industry, targeted primarily final goods. Tariffs should be decreasing in against the European Union and in particular the domestic content of foreign-produced Germany, but also to Japan), or have a rest to final goods and in the imported content of avoid further damages. domestic production of final goods. However, even before the recent escalation, temporary For the most part, measures currently in trade barriers have moved away from final force increase trade barriers on intermediate goods towards intermediate goods, starting goods, whereas historically goods for final from 2010, following a pattern contrary to consumption were the most protected. the ubiquitous tariff escalation. We show that the current trade policy will be detrimental not only to the targeted CGE modelling in imperfect competition is countries, but also to American value a good candidate to address the effects added. Two mechanisms operate, beyond of this trade war, in particular when the the direct impact of retaliation. First, US dynamic impacts of the trade war have imports subject to higher tariffs inevitably to be characterized. Sectors adjust their contain US value added (e.g. US components intermediate consumption basket to tariff- assembled abroad), notwithstanding the induced price changes, labour force and fine-tuning of the lists of targeted products. capital accumulate, and the overall setting Second, US exports will also suffer a loss of can be linked to a macroeconomic baseline. competitiveness, as production costs increase One drawback of CGE modelling recently in industries that use taxed imported goods fixed is the way GVCs were modelled. We rely as inputs. here on MIRAGE-e2, a version that integrates and importantly differentiates demand of We address the trade and welfare effects goods according to their use, for final or of the current trade tensions by tracing the intermediate consumption, thus properly impact of protection along the value chains, representing GVCs. As for tariff increases, in general equilibrium. We take on board all we rely on the official lists, but our scenarios measures enforced at the time of writing differ from the recent literature (i) in the (including voluntary export restrictions, way we aggregate these information and retaliations and safeguards), based on the (ii) in how we take into account voluntary official lists of additional tariffs, as well as the export restrictions. Tariffs are aggregated current trade agenda. Two measures are on by a simple mean to move from the 8 to the the horizon, namely: (i) the US investigation 6 digit levels and then by a reference group on the automobile industry likely to trigger weighted mean to reach the level used in US trade sanctions; (ii) the European decision our simulations. As far as Voluntary Export of 15 April 2019 to launch negotiations Restraints (VERs) are concerned, we assume with the US on a trade agreement restricted that the negotiated quantities are exported

102 each year; the target in volume is reached which will have widespread repercussions for using an export tax, endogenously computed. the way in which the world economy operates. This solution is appealing in that it generates a rent that accrues to the exporter (contrary SDGs and Paris commitments require to a tariff). policymakers to look at impacts beyond their own domain and decades ahead. According to our estimates, the measures With feedback loops abound, impacts of already implemented would cause significant interventions become theoretically ambiguous value-added losses to China (USD 91 billion requiring ex-ante integrated modelling in the long run), but also to the United States tools to explore expected impacts of policy (62 billion), due to the intertwining of global interventions, trade-offs and synergies across value chains. Because of vertical linkages multiple domains. This calls upon researchers along the value chains, 20 out of our 25 to connect previously separate strands of sectors decrease their value added in the US, research and has resulted in a burgeoning suggesting that with this tariff war the US integrated assessment literature (van Vuuren are shooting themselves in the foot. China et al, 2015, Stehfest et al. 2014). and the United States could experience GDP losses by 0.4% and 0.3% respectively. As in This paper describes how these challenges any war, imposing losses on an enemy comes are met by MAGNET, a global economic at a high cost. model designed to combine generally separate strands of research in a flexible If the tariff war were to escalate, German and coherent manner. MAGNET (Modular industry would pay a heavy toll. The Applied GeNeral Equilibrium Tool) is unique opposite path, a lull through an agreement in covering food security, sustainability and on industrial goods between the United inclusiveness in a single economy-wide States and the European Union, would avoid framework. In contrast to partial agri-food undesirable outcomes, but would bring little models MAGNET is a computable general gain per se to the parties. equilibrium (CGE) model covering income feedback loops and the full (bio)economy MAGNET — a team-based modular thus capturing feedback between primary CGE approach for coherent cross- and industrial sectors (Banse et al. 2011, cutting policy assessments Van Meijl et al. 2018) beyond the grasp of partial models. This wider coverage comes Kuiper M., van Meijl H., Tabeau A., Wageningen at the costs of technological detail, which Economic Research, Wageningen University & is addressed by allowing a link to partial Research models like IMAGE and GLOBIOM, to exploit each other’s comparative advantage (van The policy landscape is becoming increasingly Meijl et al. 2006, Doelman et al. 2018, Frank complex with interrelated global challenges et al. 2019). MAGNET can also be combined stretching across domains previously handled with technical models like TIMER or MARKAL in relative isolation. Prime examples are (Wicke et al. 2015, van Meijl et al. 2018), the sustainable development goals (SDGs) capturing adjustments in cost structures as adopted in 2015 with ambitious goals for well as smoothing changes in technology due both developing and developed countries to to economic feedback loops not accounted end poverty, improve health and education, for in these technology focussed models. reduce inequality and spur economic Compared to other CGE models MAGNET has growth while tackling climate change and more bio-economy detail absorbed through preserve natural resources both on land the cooperation with non-economic models in in the oceans by 2030 (United Nations and combines all its extensions into a single 2015). Alongside the SDGs many countries model instead of parallel developments including the European Union member states, by different teams of researchers. It also committed to halt climate change as part of includes climate specific modules as GHG the 2016 Paris agreement (UNFCCC 2016) 103 emissions and stock, potential and actual We conclude by summarizing lessons learned temperature change and CO2 taxes. MAGNET from last 10 years of MAGNET development has been designed and developed as a relevant for both the modelling and team-based model connecting specialist policymaker communities. expertise from different strands of research in a single model platform. To avoid excessive References complexity in specific applications it has Banse, M., H. van Meijl, A. Tabeau and G. been designed in a modular way allowing Woltjer, 2008. 'Will EU Biofuel Policies affect researchers to combine model extensions Global Agricultural Markets?' European Review tailored to the question at hand. Hence there of Agricultural Economics, 35: 117–141. is no single 'MAGNET model' but the model’s scope can easily be adjusted both in terms of Banse, Martin, John Helming, Peter Nowicki model structure and data preparation. and Hans van Meijl, 2008, 'Future of European Agriculture under Different Policy The core of the paper describes MAGNET is Options'. Agrarwirtschaft, 57: 156–164. it now stands. We go beyond a description of the available modules in also describing the Banse et al. 2011. 'Impact of EU biofuel technical implementation of the modularity policies on world agricultural production and both in setting up the model code and land use'. Biomass and Bioenergy, 35: 2385– additional purpose build software to support 2390. modular development in the GEMPACK software in which MAGNET is coded. Similar Doelman et al. 2018. 'Exploring SSP land-use tools are available for GAMS-based models, dynamics using the IMAGE model: regional hence lessons learned from building and gridded scenarios of land-use change MAGNET are relevant for the wider modelling and land-based climate change mitigation'. community. We also shortly touch upon the Global environmental change: human and team-based development of MAGNET which policy dimensions, 48: 119–135. poses its own opportunities and challenges beyond the technicalities of such a model Francois, J., van Meijl, H. and van Tongeren, platform, and is critical for the long term F. (2005), 'Trade liberalisation in the DOHA success of such an endeavour. Development Round’ Economic Policy , pp. 351–391. MAGNET has been successfully applied to support policymakers across different Frank et al. 2019. 'Agriculture mitigation domains. Examples are in the field of trade wedges for a 1.5 degree world: a multi-model policies, GMOs, and technical change (Meijl assessment'. Nature Climate Change. 9: and Tongeren 1998, 1999, Huang et al 2004, 66–72. Francois et al. 2005, Smeets-Kriskova et al. Hasegawa, T. et al. 2018 'Risk of increased 2017a, 2017b), agricultural and land use food insecurity under stringent global climate policies (Meijl et al. 2006, Nowicki et al. 2009, change mitigation policy', Nature Climate Banse et al. 2008), biobased economy (Banse Change, 8: 699–703 (2018). et al. 2008, Meijl et al. 2018a), food security (Kuiper et al. (fortcoming), Shutes et al. Huang, J. Ruifa Hu, Hans van Meijl, and Frank fortcoming), climate (Nelson et al. 2013, Meijl van Tongeren (2004), 'Biotechnology Boosts et al. 2018b, Hasegawa et al. 2018, Frank to Crop Productivity in China: trade and et al. 2019). These past applications show welfare implications', Journal of Development how MAGNET through its modular team- Economics, 75: 27–54. based development, addresses the need for a flexible but coherent assessment of policies Kuiper, M. et al. 'Labor supply assumptions across multiple and sometimes inherently — a missing link in food security projections' conflicting objectives of SDGs and the Paris (forthcoming). agreement.

104 Nelson, G., et al. 2013, 'Assessing uncertainty spillovers and biased technical change in along the climate-crop-economy modeling agriculture', Economic Systems Research, chain', Proceedings of the National Academy 11(1): 31–48. of Sciences U.S.A. 111(9): 3274–3279. van Meijl et al. 2006. 'The impact of different Nowicki, P., V. Goba, A. Knierim, H. van Meijl, policy environments on land use in Europe', M. Banse, B. Delbaere, J. Helming, P. Hunke, Agriculture, Ecosystems and Environment, 114: K. Jansson, T. Jansson, L. Jones-Walters, V. 21–38. Mikos, C. Sattler, N. Schlaefke, I. Terluin and D. Verhoog (2009) Scenar 2020-II – Update van Meijl et al. 2018a. 'On the macro- of Analysis of Prospects in the Scenar 2020 economic impact of bioenergy and Study – Contract No. 30–CE-0200286/00-21. biochemicals – Introducing advanced European Commission, Directorate-General bioeconomy sectors into an economic Agriculture and Rural Development, Brussels. modelling framework with a case study for the Netherlands'. Biomass and Bioenergy, Shutes et al., 'Factor representation — a 108: 381–397. missing link in household food security projections', (forthcomming(. van Meijl et al., 2018b 'Comparing impacts of climate change and mitigation on global Smeets-Kriskova, Z. et al., 2017, 'The impact agriculture by 2050', Environ. Res. Lett. 13 of R&D on factor-augmenting technical 064021 change – an empirical assessment at the sector level', Economic Systems Research, van Vuuren et al. 2015, Integrated 29(3): 385–417. assessment: Back to the Future. Inaugural lecture, delivered upon acceptance of the Smeets-Kristkova, Z. et al. (2017), 'Assessing Chair in Integrated assessment of global the impact of agricultural R&D investments environmental change, at Utrecht University’s on long-term projections of food security', Faculty of Geosciences of Utrecht University. Frontiers of Economics and Globalization 17: 1–17. Wicke et al. 2015. 'Model collaboration for the improved assessment of biomass Stehfest et al. 2014. Integrated Assessment supply, demand and impacts', GCB Bioenergy, of Global Environmental Change with IMAGE (2015)7: 422–443 3.0. Model description and policy applications, The Hague: PBL Netherlands Environmental Environmentally conscious Assessment Agency transportation and logistics UNFCCC. 2016. The Paris Agreement. 2016. modelling for agri-food supply https://unfccc.int/process-and-meetings/the- chains paris-agreement/d2hhdC1pcy. De A., Aditjandra P., Hubbard C., Gorton M., United Nations. 2015. ‘Sustainable Newcastle University Pang G., University of Birmingham Development Goals: Sustainable Thakur M., SINTEF Development Knowledge Platform’. Samoggia A., University of Bologna Sustainable Development Goals. 2015. https:// sustainabledevelopment.un.org/?menu=1300. Agriculture and fisheries are areas of deep policy integration at the EU level, organised van Meijl, H., F.W. van Tongeren, 1998, 'Trade, through the Common Agricultural Policy technology spillovers and food production in (CAP) and Common Fisheries Policy (CFP) China', Weltwirtschaftliches Archiv, 134(3): respectively. To implement and monitor the 423–449 effects of the CAP and CFP considerable data collection occurs across Member States at the van Meijl, H. , F. van Tongeren, 1999, level of primary production. This allows for 'Endogenous international technology

105 modelling the economic impacts of existing two illustrative cases: salmon in Norway and potential future policy on, for example, and processed tomatoes in Italy, conducted agricultural / fisheries output, incomes, land as part of the EU H2020 VALUMICS project. use / fish stocks through partial and general For both cases, we developed multi-echelon, equilibrium models (European Commission, multi-period, supply chain models, informed 2016a; M'Barek et al., 2017). by the literature (Govindan et al., 2015). The research work carried out in the paper However, the existing infrastructure for pays attention to experiences and problems agri-food policy modelling suffers from two encountered in developing the models, main weaknesses. First, modelling focuses data issues, results and sensitivity analysis. on primary production, rather than the whole Model development, validation and policy supply chain. Hence, existing models typically recommendation occurred in four stages: (i) do not capture the linkages between primary mapping supply chain linkages and product producers and downstream actors despite flows, (ii) designing the mathematical model, the nature of these relationships shaping (iii) data collection for parameters of the profoundly outcomes at the primary level model and (iv) model validation and deriving such as farm and fisher incomes (Falkowski policy recommendation. We concentrate on et al., 2017). Consequently, evaluating the providing key insights pertaining to each impact of new EU policy initiatives that take stages associated with model development, a supply chain perspective (e.g. Directive validation and policy recommendations via 2019/633 on Unfair Trading Practices in the stakeholders’ consultation (Govindan, 2018). agricultural and food supply chain) faces significant challenges within the existing For the first stage, it was necessary to map modelling framework. The second main the supply chain linkages and understand weakness relates to the lack of/limited the nature of flows amongst stakeholders. interactions between policies across sectors This could not be accomplished solely as for example, between agriculture, food from considering the existing literature and transport. The current policy-modelling and available secondary data. Hence, framework is tailored toward a sectoral expert interviews were conducted for each approach (e.g., the Farm Accountancy Data case study to refine the conceptual maps. Network capturing outcomes of the CAP at Without adequate conceptualisation of the the farm level). Yet, the achievement of EU supply chain, it was not possible to build policy objectives increasingly requires an appropriate mathematical models. Figures integrated approach (European Commission, 1 and 2 present the supply chain networks 2016b). For instance, the transport sector for Norwegian salmon and Italian processed accounts for approximately one quarter of tomatoes respectively. overall greenhouse gas (GHG) emissions in the EU (European Parliament, 2019) and the Joint In the second stage, mathematical models Research Centre (2006) estimates that 29% for salmon and processed tomatoes were of all consumption derived GHG emissions are developed based on the framework presented food related. Key indicators such as distance in Figure 3. The objective function within each between supply chain actors, export capacity, mathematical model aims to minimize total fuel efficiency and green tax incentives are cost, comprising of the costs associated with known to influence food prices and output transportation, fuel consumption, inventory (Soysal et al., 2014). However, the integration holding, processing and residuals/waste. of policies in modelling between these two Restrictions associated with carbon emission sectors is limited (Petrov et al., 2017). and wastage are considered for addressing the sustainability aspects. Constraints In response to these gaps, this paper aims related to supply, processing capacity, to provide a robust model of transportation storage capacity, demand, carbon emissions, and logistics for agri-food supply chains for inventory balancing, transportation capacity, policy support. It undertakes this through and different modes of transportation

106 between different types of plants and supply chains, as well as the effects of facilities are taken into consideration. particular policy interventions and market developments (e.g. variation in demand, In the third stage, the primary data collection fuel costs, emission and waste constraints). is performed pertaining to various input They can aid supply chain managers to parameters provided in the framework of the make decisions regarding the amount of mathematical model, supplied by the industry inventory to be kept in different time periods. stakeholders. Information related to the cost The models are developed for a planning components, capacity restrictions, carbon horizon consisting of discrete time periods, emission coefficients, and fuel consumption aiding the possibility of studying demand rates considering various scenarios associated and supply uncertainty and its consequences with choice of transportation mode and the in supply chain decision making. Hence, supply and demand variability in different they help decision makers to identify the time periods. changes in a supply chain network when different transportation routes are adopted The fourth stage aims to resolve the (for example whether maritime routes models and obtain necessary managerial can be adopted or not in place of road/rail recommendations for various scenarios. transportation, to address environmental The models are valuable for policy makers concerns related to fuel consumption and in terms of understanding the costs and carbon emissions). The models generate emissions associated with different food valuable insights for supply chain managers,

Figure 1: Conceptual framework for the Norwegian Salmon Supply Chain Network

Figure 2: Conceptual framework for the Italian Processed Tomato Supply Chain Network

107 Figure 3: Framework for the model for both products

understanding the effects of different European Parliament (2019) Common scenarios associated with demand and transport policy: overview Brussels: European supply uncertainty and the adoption of Parliament. [Online]. Available at: http://www. different transportation routes. Based on the europarl.europa.eu/factsheets/en/sheet/123/ sensitivity analysis, policy implications can be common-transport-policy-overview. drawn. Falkowski, J., Menard, C., Sexton, R.J., Acknowledgements Swinnen, J. and Vandevelde, S. (2017) Unfair trading practices in the food supply This study is part of the VALUMICS project chain: A literature review on methodologies, (www.valumics.eu), and has received funding impacts and regulatory aspects. Seville: Joint from the European Union’s Horizon 2020 Research Centre. [Online]. Available at: http:// research and innovation programme, under publications.jrc.ec.europa.eu/repository/ grant agreement No 72724. handle/JRC108394.

References Govindan, K. (2018) 'Sustainable consumption and production in the food supply chain: A European Commission (2016a) Multiannual conceptual framework', International Journal Union programme for the collection, of Production Economics, 195, pp. 419–431. management and use of data in the fisheries and aquaculture sectors for the period 2017- Govindan, K., Soleimani, H. and Kannan, D. 2019 (C(2016)4329). Brussels: European (2015) 'Reverse logistics and closed-loop Commission. [Online]. Available at: https:// supply chain: A comprehensive review to eur-lex.europa.eu/legal-content/EN/TXT/ explore the future', European Journal of PDF/?uri=CELEX:32016D1251&from=EN. Operational Research, 240, pp. 603–626.

European Commission (2016b) Scoping Joint Research Centre (2006) Environmental study on modelling of EU environment policy. Impact of Products (EIPRO): Analysis of the Brussels: European Commission. [Online]. life cycle environmental impacts related to Available at: http://ec.europa.eu/environment/ the final consumption of the EU-25. Brussels: enveco/studies_modelling/pdf/study_ Joint Research Centre (DG JRC). [Online]. modelling.PDF Available at: http://ec.europa.eu/environment/ ipp/pdf/eipro_report.pdf. 108 M'Barek, R., Barreiro Hurle, J., Boulanger, an economy, they have the potential to be P., Caivano, A., Ciaian, P., Dudu, H., Espinosa especially problematic in food supply chains, Goded, M., Fellman, T., Ferrari, E., Gomez Y as agricultural producers may be placed Paloma, S., Gorrin Gonzalez, C., Himics, M., under undue pressure and have limited Elouhichi, K., Perni Llorente, A., Philippidis, G., bargaining power in negotiations with larger Salputra, G., Witzke, H.P. and Genovese, G. purchasers, such as supermarkets or retailers, (2017) Pathways for the European agriculture given the lack of alternative buyers (Duffy and food sector beyond 2020 — Study et al., 2003, Falkowski et al., 2017). As a Luxembourg: Publications Office of the counter measure, the recent EU Directive European Union. [Online]. Available at: https:// (2019/633) on UTPs aims at protecting ec.europa.eu/jrc/en/publication/eur-scientific- weaker ‘suppliers’, primarily farmers, including and-technical-research-reports/scenar-2030- their organisations (e.g. cooperatives) against pathways-european-agriculture-and-food- their buyers, as well as suppliers of agri- sector-beyond-2020. food products which are further downstream (European Parliament, 2019). Due to the Petrov, L., Annoni, P., Acs, S., Schade, B., Gisolo, topicality and policy relevance of fairness in E., Kulvinskaite, G., Gancheva, M., Christou, FVCs there is value in exploring its dynamics M. and Otto, J. (2017) Analysis of the use of through simulation modeling. models by the European Commission in its Impact Assessments for the period 2009– Aims and objectives 2014. Luxemborg: Publications Office of the European Union. [Online]. Available at: https:// This research forms a part of VALUMICS, an ec.europa.eu/jrc/en/publication/analysis-use- ongoing Horizon 2020 EU funded project on models-european-commission-its-impact- food value chains (FVCs). The overall objective assessments-period-2009-2014. of the project is to develop a comprehensive suite of tools that will enable decision makers Soysal, M., Bloemhof-Ruwaard, J.M. and to evaluate the impact of strategic and van der Vorst, J.G.A.J. (2014) 'Modelling operational policies on the resilience, integrity food logistics networks with emission and sustainability of European FVCs. In considerations:The case of an international particular, the current research concerns the beef supply chain', International Journal of development of a simulation model focused Production Economics, 152, pp. 54–70. on fairness in FVCs and the objective is to develop quantitative indicators of fairness Modelling fairness in FVCs: that are operational in the model. The developing quantitative indicators degree of fairness in inter-firm relations is a perception and therefore it is necessary to Gudbrandsdottir I. Y., Cook D., Olafsdottir G., define quantifiable indicators to be used for Oddsson G., Bogason S., University of Iceland the simulation model. Stefansson H., University of Reykjavík McGarraghy, S., University College Dublin, Modelling fairness National University of Ireland Simulation modeling is well suited for Introduction the development and testing of policy interventions. Food systems have been Unfair trading practices (UTP) within food described as complex adaptive systems supply chains are of increasing concern to (CAS) as they are characterised by a large European Union (EU) and member states’ number of interactions and interdependencies policy makers (DG IPOL, 2015). Findings which leads to nonlinear, emergent system indicate that their negative impact on behaviour that is not easily controlled. SMEs in the EU food sector is affecting Individual agents impact the system with the competitiveness of the industry as a actions resulting from their localised decision- whole (Wijnands et al., 2007). Although making and in turn they are constrained UTPs can arise in any market or sector of by the system structure. In a sense, the 109 system is self-organising from the viewpoint literature from the VALUMICS project (Barling of the individual agent (Choi et al., 2001; and Gresham, 2019). Furthermore, fairness Surana et al., 2005) and its structure and perceptions of agents in the chains will be extended operation are to a large extent studied by surveying actors across the FVCs invisible to them. The same applies to and asking for perceptions on fair compared policy-makers, for whom a lack of a whole- to actual gross margins, as well as analysing chain overview makes it difficult to predict responses to a series of statements on the the effects of policy implementations distributive and procedural components of beforehand (Stave & Kopainsky, 2015). In UTPs according to a conceptual model on addition to their structural complexity, food fairness in sustainable supply chains (IIED/ systems are heavily influenced by social and Oxfam, 2012). environmental factors. Preliminary results Agent-based simulation modeling (ABM) has been successfully used to model CASs. Such The result of the present study will be models are typically built from the bottom quantitative indicators of fairness that can up by identifying agents in the system and be used in the VALUMICS simulation model defining their behaviours, including how to analyse and test policy interventions they interact with other agents and their related to fairness in FVCs. When examining environment. The behaviour of the system as quantitative metrics for distributive fairness a whole emerges out of multiple concurrent we take into account the importance of individual behaviours. In VALUMICS, the aim price, for agents in the FVC in their effort to is to use the ABM to identify the level of maximize their profit or utility. Furthermore, fairness within the system, which emerges the influence of market power with respect via concurrent execution of decision rules on to creating opportunities for misuse of power behalf of multiple independent agents in the in the form of UTPs, is considered to be of FVC. relevance. Price movements threatening the margin of firms being able to exert market Method power are transmitted faster than price movements that improve it (Falkowski et al., As fairness is an intangible concept it is not a 2017). Given this, the VALUMICS project will straightforward task to define it in simulation integrate quantitative economic indicators model operational terms. In order to do so we into its ABM to gain enhanced understanding draw on fairness theory and related literature of distributive fairness in the aquaculture on governance and market power in FVCs and agricultural FVCs. First, the gross profit (Busch & Spiller, 2016; Gereffi et al., 2005; margin obtained by the various actors across Carbone, 2017). To further substantiate the the FVCs will be assessed. Second, the degree establishment of quantitative indicators two of market power will be investigated by using dissimilar European FVCs are explored, an the Lerner Index, which provides an estimate aquaculture chain and an agricultural chain. of market power in an industry, measuring As food supply systems these two chains the price-cost margin through the difference have many things in common. The agents in between the output price of a firm and the the chains are similar (i.e. primary producers, marginal cost divided by the output price producers, retailers and consumers) and the (Elzinga & Mills, 2011). The aim will not be to perishability of the final product affects the determine an absolute measure of fairness workings of both chains. While FVC are similar using these indicators, but rather to ascertain in many aspects, aquaculture and agricultural transitions towards fairer outcomes. This chains vary with respect to governance, approach is in keeping with the European power structure, contractual agreements and Parliament’s depiction, which, rather than pricing practices. The governance structure providing a strict value measure of UTPs, of the FVC is explored in the emerging

110 emphasises the presence of gross deviations Elzinga, K. G., & Mills, D. E. (2011). 'The Lerner away from good commercial conduct. index of monopoly power: origins and uses'. American Economic Review, 101(3), 558–64. Acknowledgements European Parliament (2019). Directive of This study is part of the VALUMICS project the European Parliament and of the Council 'Understanding food value chains and on unfair trading practices in business-to- network dynamics' (www.valumics.eu), and business relationships in the agricultural and has received funding from the European food supply chain. Retrieved from: http://data. Union’s Horizon 2020 research and innovation consilium.europa.eu/doc/document/PE-4- programme, under grant agreement No 2019-INIT/en/pdf 72724. Fałkowski, J., C. Ménard, R.J. Sexton, J. References Swinnen and S. Vandevelde (Authors), Marcantonio, F. Di and P. Ciaian (Editors) Barling D. and Gresham J. (Editors) (2019) (2017), Unfair trading practices in the Governance in European Food Value Chains. food supply chain: A literature review on VALUMICS 'Understanding Food Value Chains methodologies, impacts and regulatory and Network Dynamics', funded by European aspects, European Commission, Joint Union’s Horizon 2020 research and innovation Research Centre. programme, GA No 727243. Deliverable: D5.1, University of Hertfordshire, UK, 237p. IIED/Oxfam (2012). Measuring Fairness in Supply Chain Trading. International Institute Bies, R. J., & Moag, J. S. (1986). 'Interactional for Environment and Development/Oxfam communication criteria of fairness'. Research Publication. in organizational behavior, 9, 289–319. Gereffi, G. Humphrey, J and Sturgeon, Carbone, A. (2017) 'Food supply chains: T (2005), 'The governance of global coordination governance and other shaping value chains'. Review of International forces'. Agricultural and Food Economics,5:3 Political Economy 12:1, 78–104 doi. doi:10.1186/s40100-017-0071-3 org/10.1080/09692290500049805 Choi, T. Y., Dooley, K. J., & Rungtusanatham, Stave, K. A., & Kopainsky, B. (2015). 'A M. (2001). 'Supply networks and complex system dynamics approach for examining adaptive systems: Control versus emergence'. mechanisms and pathways of food supply Journal of Operations Management, 19(3), vulnerability'. Journal of Environmental 351–366. Studies and Sciences, 5(3), 321–336. DG IPOL (2015) Directorate-General for Surana, A., Kumara*, S., Greaves, M., & Internal Policies, Policy Department C, Raghavan, U. N. (2005). 'Supply-chain Citizens’ Rights and Constitutional Affairs networks: A complex adaptive systems (2015) The general principles of EU perspective'. International Journal of administrative procedural law (PE 519.224), Production Research, 43(20), 4235–4265. European Parliament. Wijnands, J. H., van der Meulen, B. M., & Duffy, R., Fearne, A., & Hornibrook, S. (2003). Poppe, K. J. (2007). Competitiveness of the 'Measuring distributive and procedural justice: European food industry: An economic and An exploratory investigation of the fairness legal assessment 2007: Office for Official of retailer-supplier relationships in the UK Publications of the European Communities. food industry'. British Food Journal, 105(10), 682–694.

111 112 Poster session 1

26 November

113 114 A global sensitivity analysis of parameters used. In addition, the current smart grids project cost benefits work also investigates possible dependence relationships among inputs of the CBA. analysis with correlated inputs The following eight input factors were taken Vitiello S., Albrecht D., Rosati R., Mara T. A. into account as causes of output uncertainty: European Commission, Joint Research Centre 1. Social Discount Rate (SDR) In this work, Uncertainty Analysis (UA) and Global Sensitivity Analysis (GSA) are carried 2. Yearly average rate of decrease of out on a Smart Grid project's benefits benefits from software infrastructure and costs in order to identify main input 3. Yearly average rate of decrease of sources of uncertainty. This approach yields benefits from physical infrastructure significant improvements in terms of accuracy 4. Yearly average rate of electricity demand in the estimation of costs and benefits of increase the selected project, namely the Net Product

Value (NPV). 5. Value of 1 ton of CO2-equivalent average price on the ETS market The thorough account of Smart grid projects' costs and benefits is in fact characterized by 6. Emission factor significant uncertainty: each unique, capital In the Rome’s case other two inputs were intensive project tests new technologies added to the study, that is: and site-specific solutions for which a clear 1. Yearly increase in CAPEX costs assessment cannot rely on the evidence from other projects. In this study, for the first time 2. Yearly increase in OPEX costs the Joint Research Centre (JRC) applied to The first step of the analysis was the the Cost Benefit Analysis of a real Smart Grid output variability quantification (NPV). This project a Global Sensitivity Analysis taking variability derives from the input uncertainty into account correlation hypothesis among propagated through the model. Therefore, the variables. ranges of variability of each input and their The analysis is performed around two probability distribution functions must be scenarios: one where the Smart Grid solutions estimated. are implemented only in the pilot project These estimations might depend on several area of ‘Malagrotta’, and another one where elements which should be identified by they are extended to the whole distribution the experts of each Smart Grid project on network of the city of Rome, Italy. the basis of the specific problem they are considered. This means that different UA/SAs Data for both scenarios are gathered from might be necessary when the same model is the Italian Distribution System Operator, ACEA applied if the context of the plan changed. - one of Italy’s biggest one, and a simple Cost Benefit Analysis has been already performed In our study, two different Monte Carlo in Vitiello et al., A Smart Grid for the city simulations of the CBA models have been of Rome - A Cost Benefit Analysis. 2015. conducted, one for the Malagrotta project doi:10.2790/50100. referring to Malagrotta NPV, and one for the Rome’s project to analyse its distribution Based on this latter report, this new network NPV. piece of research aim at showing that the Consequently, two Monte Carlo samples standard practice in CBA of the One-at- of the input of size Nxd (N=256, d=6 for a-time (single-factor) sensitivity analysis Malagrotta and d=8 in the Rome’s case) (SA) might be significantly improved when have been created. These two samples were coupled with Uncertainty Analysis and Global represented by matrices of random input Sensitivity Analysis, that allow capturing the values generated from their probability possible interactions among the different density functions. Each matrix row was a set 115 of the six values (eight inputs for Rome) used inputs,' Reliability Engineering & System to run the CBA model. Also the dependences Safety, vol. 107, pp. 115–121, 2012. among significant variables have been A. Saltelli, M. Ratto, T. Andres, F. Campolongo, shaped, as indicated by the experts. J. Cariboni, D. Gatelli, M. Saisana and S. These correlations among the significant Tarantola, Global Sensitivity Analysis. The inputs were then taken into account when Primer., Chichester: Probability and Statistics, first-order indices were computed. With this John Wiley and Sons, 2008. aim, we made use of the regression technique Q. Shao, A. Younes, M. Fahs and T. A. proposed in Mara and Tarantola (2012). Mara, 'Bayesian sparse polynomial chaos expansion for global sensitivity analysis,' The two projects have undergone the same Computer Methods in Applied Mechanics and sensitivity analysis steps and very similar Engineering, vol. 318, pp. 474–496, 2017. results were obtained. The Malagrotta and Rome’s SDR indices show the highest values S. Vitiello, G. Flego, A. Setti, G. Fulli, S. Liotta, (0.98 and 0.96 respectively). This proves S. Alessandroni, ... & D. Parisse (2015). A once again the high relevance, in terms of smart grid for the city of Rome: a cost benefit uncertainty impact, of this input (see JRC analysis. JRC Science and Policy Report. report). doi:10.2790/50100

The electricity demand increase index A participatory process to remains constant with a percentage of about design regional policy for rural 0.20. Eventually for the Rome’s network, development: the case of the the CAPEX and OPEX indices are significant with an index value equal to 0.42 and 0.28 Veneto Region respectively. Trestini S., Giampietri E., Boatto V., TeSAF Department, University of Padova Note that the sum of the individual main Povellato A., CREA Council for Agricultural Sobol’ indices is higher than one. Under the Research and Economics theoretical assumption of independent input factors, this sum cannot exceed the unity. Recently, the European Union is witnessing We therefore must conclude that some a crucial moment that coincides with the correlation exists among some inputs. conclusion of the multi-annual policy framework programmes 2014–2020 and the Significant improvements over standard necessity to look at the future. Accordingly, sensitivity analysis were obtained by this the European Agricultural Policy (CAP)’s GSA of CBAs for Smart Grids projects. It agenda is undergoing a reform process that allowed knowing more comprehensively the drives the necessity to efficiently analyse and uncertainty affecting the results (NPV), and design the new policy through models that reliably evaluating the contribution of each prefigure the most appropriate strategies parameter in the determination of a positive to achieve future goals. Indeed, agriculture Net Present Value for the project. makes extensive use of models as other relevant policy areas. The CAP takes action The authors strongly recommend the use of with rural development measures with GSA, as a standard tool, to carry out Smart national and regional programmes to address Grids Costs Benefits Analyses. Given the the specific needs and challenges facing wide diffusion of CBA as a tool in impact/risk rural areas: to this purpose, the drafting of assessment, in many fields, this work is a very future rural development strategies (II pillar promising achievement. of the CAP) is experiencing a lot of ferment References in Italy especially at regional level. In this regard, after a long process of exchange and T. A. Mara and S. Tarantola, ' Variance-based debate among the stakeholders (through sensitivity indices for models with dependent structured interviews) the Veneto Region, 116 which is actually the first in Italy to draft a in the European Union in quantities of one strategic document so far, has just concluded tonne or more per annum must be assessed a concrete experience of operational joint for their potential to cause adverse effects collaboration between 31 experts (with the (EC, 2006). One of the endpoints that must direct involvement of both scientists and be assessed is skin sensitisation, which is the policy makers) and a public partnership first step of allergic contact dermatitis (ACD) with a regional conference, thus tracing a in humans. ACD is the clinical manifestation model that could be easily replicated both of a changed responsiveness of the adaptive regionally and nationally for policy support immune system following repeated exposure and implementation. to a sensitising substance.

This abstract describes the participatory The assessment of skin sensitisation has process through which, by means of a traditionally been based on the use of combination of qualitative and quantitative animals, being the less invasive test the local methods, the Veneto regional strategy for lymph node assay (LLNA), which, however, rural development (agriculture, forests, still implies the sacrifice of the animals. The rural development) until 2030 has just been international scenario changed in 2013 when defined, in the context of the perspectives the Cosmetics Regulation (EC, 2009) entered and objectives (e.g. the CAP’s 9 key objectives into force and banned the use of animal + 1 transversal objective) and rules (e.g. the testing for cosmetics in Europe. Moreover, CAP’s New Delivery Model for a more results- since 2016 the REACH regulation demands oriented policy) outlined at the community that testing on vertebrate animals should be and national level for the CAP post-2020 considered only as last resort (EC 2016). as well as in reference to the regional government program. Mechanistically based in vitro methods that capture the key (biological) events (KE) that In particular, starting from the SWOT are considered necessary for the acquisition (strengths, weaknesses, opportunities, and of skin sensitisation (OECD, 2012a,b) threats) analysis and the 32 needs defined represent one of the most prominent at regional level, 43 strategic options have alternatives to animal testing. In fact, three- been evaluated through a multi-criteria animal free methods have been validated by analysis by 31 experts, in terms of efficacy the European Union Reference Laboratory (from low to high) to achieve the EU for Alternatives to Animal Testing (EURL objectives, and selected by more than 100 ECVAM). These methods are: the direct public stakeholders involved in a structured peptide reactivity assay (DPRA) (EURL ECVAM, online survey. Based on these, policy 2013; Gerberick et al., 2004), KeratinoSensTM makers will design their regional priorities, (Emter et al.,2010; EURL ECVAM, 2014; aligned on the EU objectives, and the most Natsch and Emter, 2008), and the human appropriate measures to support farmers at Cell-Line Activation Test (h-CLAT) (Ashikaga regional level, with a view to achieve a more et al., 2006; EURL ECVAM,2015; Sakaguchi performance-based governance system while et al., 2006). Despite these methods being respecting the overall sustainability goals at predictive of LLNA responses with an European level. accuracy of about 80%, the data generated are not considered equivalent to animal test Defining the applicability of data for regulatory purposes. animal-free approaches for skin Instead, the most promising alternative allergy testing of chemicals to animal testing seems to be the use Asturiol D., Casati S., Paini A., Worth A., of these data in combination with other European Commission, Joint Research Centre relevant information (e.g. physicochemical properties, in silico, in chemico, in vitro data) In order to assure the safety of the European in the context of Integrated Approaches population, chemicals produced or marketed to Testing and Assessment and Defined 117 Approaches (DA). DAs consist of a fixed chemical space and applicability domain of data interpretation procedure (DIP) used to DAs, i.e. the interpolation space in which the interpret data generated with a defined set method is expected to provide reliable results. of information sources, that can either be In addition, standard reporting is needed for used alone or together with other information such methods to be used under MAD and sources, to satisfy a specific regulatory need. the JRC has developed standards to describe Several DAs for skin sensitisation have been and report the applicability domain of DAs proposed to combine KE data. These include in a simple and straightforward manner that the use of machine learning algorithms (e.g. can be applied to the different DAs available classification trees, Bayesian networks, and independently of their complexity (can be neural networks) (Asturiol et al. 2016, Hirota applied to a simple classification tree model et al. 2015, and Jaworska et al. 2013). (Asturiol et al., 2016) and to more complex models such as those based on neural The use of these new methodologies based networks (Hirota et al., 2015)). on the integration of data represents a regulatory challenge because so far only It is expected that these DAs will gain OECD individual testing methods have been acceptance by the end of 2019 and that, for translated into OECD test guidelines. Once the first time, computational methods will be adopted, these OECD test guidelines fall included in an OECD test guideline and fall under the mutual acceptance of data (MAD) under MAD. meaning the data generated with one of these methods in any of the OECD member References country with one of these methods should Ashikaga, T., Yoshida, Y., Hirota, M., Yoneyama, be automatically accepted in the other OECD K., Itagaki, H., Sakaguchi, H., & Toyoda, H. countries. (2006). 'Development of an in vitro skin The main difficulties correspond to the fact sensitization test using human cell lines: The that DAs have not undergone any formal human Cell Line Activation Test (h-CLAT): validation to characterise their reproducibility, I. Optimization of the h-CLAT protocol'. transparency, relevance, and accountability; Toxicology in Vitro, 20(5), 767–773. and do not have dedicated OECD test Asturiol, D., Casati, S., & Worth, A. (2016). guidelines. In addition, computational 'Consensus of classification trees for skin methods in general and machine learning sensitisation hazard prediction'. Toxicology in algorithms in particular have traditionally Vitro, 36, 197–209. been treated differently from testing methods and have never been included in a test European Commission. (2006). REACH guideline. Regulation (1907/2006/EC): Authorisation and Restriction of Chemicals (REACH), establishing The JRC is playing a key role in this process a European Chemicals Agency, amending to develop, promote and validate new Directive 1999/. (Official Journal of the methods. In particular, the JRC is leading European Union, Ed.) (Vol. L136). the international efforts to characterise the DAs in the light of the elements above and European Commission. (2009). REGULATION determine the information that needs to (EC) No 1223/2009 OF THE EUROPEAN be reported so that DAs and computational PARLIAMENT AND OF THE COUNCIL of 30 methods are considered at the OECD level November 2009 on cosmetics products. under the MAD and therefore transparent, Official Journal of the European Union, (1223), relevant, accountable, auditable, and 342/59-208. trustable methods. European Commission. (2016). COMMISSION In order to improve the transparency and REGULATION (EU) 2016/1688 Amending trustworthiness of these methods, the JRC Annex VII to Regulation (EC) No 1907/2006 of has proposed the formal definition of the the European Parliament and of the Council 118 on the Registration, Evaluation, Authorisation testing strategy to assess skin sensitization and Restriction of Chemicals (REACH) as potency: from theory to practice'. Journal of regards skin sensitisation. Official Journal of Applied Toxicology, 33(11), 1353–1364. the European Commission, 1688. OECD. (2016a). OECD Guidance document Emter, R., Ellis, G., & Natsch, A. (2010). on the reporting of defined approaches to be 'Performance of a novel keratinocyte-based used within integrated approaches to testing reporter cell line to screen skin sensitizers in and assessment (No. ENV/JM/MONO(2016)28). vitro'. Toxicology and Applied Pharmacology, 245(3), 281–290. OECD. (2016b). OECD Guidance document on the reporting of defined approaches and EURL-ECVAM. (2012). Direct Peptide Reactivity individual information sources to be used Assay ( DPRA ) ECVAM Validation Study within integrated approaches to testing and Report. Retrieved from https://eurl-ecvam.jrc. assessment (IATA) for skin sensitization (No. ec.europa.eu/eurl-ecvam-recommendations/ ENV/JM/MONO(2016)29). eurl-ecvam-recommendation-on-the-direct- peptide-reactivity-assay-dpra Sakaguchi, H., Ashikaga, T., Miyazawa, M., Yoshida, Y., Ito, Y., Yoneyama, K., Suzuki, H. EURL-ECVAM. (2014). KeratinoSensTM (2006). 'Development of an in vitro skin validation study reports. Retrieved from sensitization test using human cell lines; https://eurl-ecvam.jrc.ec.europa.eu/eurl- human Cell Line Activation Test (h-CLAT) ecvam-recommendations/recommendation- II. An inter-laboratory study of the h-CLAT'. keratinosens-skin-sensitisation Toxicology in Vitro, 20(5), 774–784.

EURL-ECVAM. (2015). h-CLAT validation study Employment effect of innovation reports. Retrieved from https://eurl-ecvam.jrc. ec.europa.eu/eurl-ecvam-recommendations/ Kancs d'A., European Commission, Joint eurl-ecvam-recommendation-on-the-human- Research Centre cell-line-activation-test-h-clat-for-skin- Siliverstovs B., Bank of Latvia and KOF Swiss sensitisation-testing Economic Institute, ETH Zurich

Gerberick, G. F., Vassallo, J. D., Bailey, R. E., In setting the Europe 2020 Strategy, the Chaney, J. G., Morrall, S. W., & Lepoittevin, J. P. European Union (EU) has defined five (2004). 'Development of a peptide reactivity ambitious objectives – on employment, assay for screening contact allergens'. innovation, education, social inclusion and Toxicological Sciences, 81(2), 332–343. climate/energy – to be reached by 2020 (European Commission, 2013). Concerning Natsch, A., & Emter, R. (2008). 'Skin the first two targets, the Strategy aims Sensitizers Induce Antioxidant Response at: (i) increasing employment by raising Element Dependent Genes: Application to the the employmentrate of population to at In Vitro Testing of the Sensitization Potential least 75%; and (ii) promoting innovation of Chemicals'. Toxicological Sciences, 102(1), by increasing research and innovation 110–119. expenditures to at least 3% of the GDP.

Hirota, M., Fukui, S., Okamoto, K., Kurotani, S., In the context of these two Europe 2020 Imai, N., Fujishiro, M., Miyazawa, M. (2015). Strategy’s objectives, an important policy 'Evaluation of combinations of in vitro question arises whether innovation sensitization test descriptors for the artificial and employment processes can be neural network-based risk assessment model complementary and hence their EU targets of skin sensitization'. J Appl Toxicol, (August can be achieved at the same time? Further, 2014), 1333–1347. policy makers are interested to know: (i) are there R&D intensity levels when innovation Jaworska, J., Dancik, Y., Kern, P., Gerberick, and employment are positively related to F., & Natsch, A. (2013). 'Bayesian integrated each other and when innovation may have 119 an adverse impact on the firm employment? for our purpose: (i) estimation can be based (ii) what type of innovators create most jobs on a flexible semi-parametric regression and hence provide the highest potential for allowing for a non-linear dependence between policy synergies? Answering these questions the variables of interest without imposing any is the main objective of the present study, as a priori restrictions; and (ii) the elimination of they may help to design policies, which can the selection bias arising from a non-random efficiently contribute to achieving both the assignment of treatment (R&D expenditure) innovation and employment targets of the intensity across firms by conditioning on the Europe 2020 Strategy at the same time. observed firm characteristics.

At a first glance, a simultaneous boosting of In applying the GPS methodology, we attempt both employment and innovation may seem to identify the R&D intensity levels under an easy and most natural task to achieve which innovation can be complementary to as any type of investments (including R&D) employment and under which it may have an increases the labour demand, at least in the adverse impact on employment. To the best short-run. However, the theoretical literature of our knowledge, the application of a flexible suggests that the relationship between semi-parametric counterfactual methods to innovation and employment seems to be the employment-innovation nexus is the first far more complicated than one can naively of this sort in literature and hence constitutes assume initially (Smolny, 1998). Also the our main contribution to literature. econometric results reported in the literature on employment effects of innovation are We base our micro-econometric analysis on rather contradictory both with respect to a large international firm-level panel data their sign and magnitude, suggesting that set for OECD countries and our proxy for increasing the innovation intensity can technology is a measurable and continuous have not only complementary but also variable, while most of previous studies substitutionary effects on employment have relied on either indirect proxies of the (Young, 1993; Piva and Vivarelli, 2005; technological change or dummy variables Antonucci and Pianta, 2002; Van Reenen, (such as the occurrence of product and 1997). process innovation). In particular, we employ the EU Industrial R&D Investment Scoreboard In order to accommodate a wide range of data set, which comprises data on the possibilities in the innovation-employment R&D investment, as well as other financial relationship ranging from highly negative and economic variables for the top 2500 to strongly positive, in the present study innovators worldwide. we propose an alternative methodological approach that has not been employed in the References innovation-employment literature before. In Antonucci, T. and M. Pianta (2002). particular, we relax the linearity assumption 'Employment effects of product and process in the functional relationship between innovation in Europe'. International Review of innovation and employment and hope that it Applied Economics 16(3), 295–307. will contribute towards sorting out the likely reasons for observing such a large range Bogliacino, F., M. Piva, and M. Vivarelli (2012). of estimated employment elasticities with 'R&D and employment: An application of the respect to the firm innovation activity. LSDVC estimator using European microdata'. Economics Letters 116(1), 56–59. We rely on a flexible semi-parametric method – the generalised propensity score Bogliacino, F. and M. Vivarelli (2012). 'The (GPS) estimator – suggested by Hirano and job creation effect of R&D expenditures'. Imbens (2004). Two main features of the GPS Australian Economic Papers 51(2), 96–113. methodology make it particularly attractive

120 European Commission (2013). EUROPE 2020: reforms of the education system in 2017- A strategy for smart, sustainable and inclusive 2018. Project team members from the growth. Brussels: European Commission. Ministry and the Research and Higher Education Monitoring and Analysis Centre Hirano, K. and G. W. Imbens (2004). The (MOSTA) worked together with experts from propensity score with continuous treatments. the Structural Reform Support Programme of In A. Gelman and X.-L. Meng (Eds.), Applied the European Commission, the Department Bayesian Modeling and Causal Inference from for Education in England and University Incomplete-Data Perspectives, pp. 73–84. College Cork in Ireland. Chichester: Wiley. Anticipating shortages in specific teaching Lachenmaier, S. and H. Rottmann (2007). subjects or other areas in the future is 'Employment effects of innovation at the firm crucial for initial teacher training. In order level'. Journal of Economics and Statistics to accommodate demographic and other 227(3), 254–272. challenges, strategic decisions need to be made well in advance. Teachers are key to Pianta, M. (2004). 'Innovation and the learning process, therefore, projecting the employment'. In J. Fagerberg, D. C. Mowery, size of the teaching population that needs and R. R. Nelson (Eds.), Handbook of an investment in preparation, skills and Innovation, Chapter 22. Oxford University continuous support is important. The aim of Press. the teacher workforce planning project is to Piva, M. and M. Vivarelli (2005). 'Innovation set up a pilot forecasting model that provides and employment: Evidence from Italian short term and midterm forecasts on teacher microdata'. Journal of Economics 86(1), demand. Up to now, there has been no 65–83. systematic mechanism to project teacher workforce or to set quotas for publicly funded Smolny, W. (1998). 'Innovations, prices and study placements for initial teacher training. employment: A theoretical model'. Journal of Industrial Economics 46(3), 359–381. The population of Lithuania has been declining in recent years and fell to 2.8m Van Reenen, J. (1997). 'Employment and in 2018. Demographic changes have been Technological Innovation: Evidence from reflected in the pupil population, with a 5 U.K. Manufacturing Firms'. Journal of Labor percent decrease between the school years Economics 15(2), 255–84. 2012/13 and 2016/17. The biggest decrease can be seen in lower and upper secondary Young, A. (1993). 'Substitution and education. This challenge put pressure on the complementarity in endogenous innovation'. school network resulting in some closures Quarterly Journal of Economics 108(3), and consolidation. The pupil-teacher ratio in 775–807 Lithuania is below average OECD figures.

Forecasting the teaching In addition, the teaching workforce is rapidly workforce in Lithuania ageing, nearly half of general education teachers are 50 years and older. If the trend Leiputė B., Padvilikis G., Research and Higher continues, 20 percent of current teaching Education Monitoring and Analysis Centre workforce will be beyond retirement age Hyland A., University College Cork in 5 years’ time. Furthermore, the number of university bachelor degree students This pilot project to develop a teacher graduating from programmes providing workforce forecasting model was teaching qualification decreased almost commissioned by the Office of the threefold. Only a small proportion of Government of Lithuania and the Ministry graduates seek and find employment of Education and Science as part of ongoing at schools. Other challenges such as

121 low salaries, weak learning outcomes of level of education, post type and teaching Lithuanian pupils and significant urban-rural subject was made. In total 83% of teachers disparities can be seen. and pedagogical staff employed during the school year 2017/18 are represented A sophisticated teacher forecasting model in the model. Short-term as well as mid- developed by the Department for Education term projections are provided for each of in England as well as experience of teacher 25 specialisations. Baseline and 2 other forecasting in Ireland was examined and was scenarios are provided. found to be relevant and useful. Considering the findings of the literature review According to the baseline scenario, a shortage undertaken, there is general agreement on of 98 individuals is projected in 2018/19. The the major issues that affect teacher supply highest shortage (177 teachers) can be seen and demand. Common principles relating to in teachers for primary schools. Even if all ITT sources of teacher inflow and outflow are graduates receive employment, a shortage prevalent across the literature. Empirical of over 100 teachers would occur. Therefore, findings by Lithuanian researchers were other type of short-term measures are summarised, however, the use of the results needed to cover the shortage. A surplus for on policy making is not known. Social educators (37) and Pre-school tutors (284) in 2018/19 is expected. However, there The teacher forecasting model is based on might be differences across municipalities labour demand and supply. In the model, that cannot be calculated with the pilot teacher supply is denoted by (a) inflows to the model. teaching pool such as the projected number of ITT students who successfully graduate The highest 4-year (2018/19 — 2021/2022) and receive employment at shortage can be seen among Primary school teachers (700 teachers). Other schools, (b) non-qualified teachers entering significant shortages in Lithuanian language schools via various sector engagement and Mathematics might reach up to 200 programmes or those with unknown individuals in 4 years’ time. The shortage qualification due to short timeseries (c) in Pre-primary teachers will more than inactive qualified teachers who work outside double in size to 123 individuals. Foreign schools and choose to enter or return to language teachers (English and Other) will school. The demand side has two major face a shortage of 171 and 159 respectively. elements: 1. Expansion demand described as Therefore, changes in the ITT student the total teacher numbers each year needed admission might be one of the main keys to to accommodate forecast changes in pupil overcome shortages in midterm. The surplus population 2. Substitution demand described for Pre-school tutors is visible across all three as teachers in stock need to be replaced due scenarios. to death, retirement or resignation. The pilot project results were used to inform The model uses administrative data collected decisions on school year 2018/19 admissions from the national registers on teachers, to initial teacher training programs. The students and pupils. It includes pre-school, model results are open to the public, general and vocational education teachers users can construct their own scenarios by and pedagogical staff members. Different changing assumptions for a selection of levels of detail are used, depending on the model elements. The results not only provide model element. This approach addresses the relevant information for national level policy issue of differences that can occur across makers but also provide information on different levels, for example differences in various teacher labour market prospects that contact hours in rural or urban schools, class are relevant to teacher training providers, mergers, drop-outs and other factors. students and graduates, local authorities and schools. A list of 25 specializations by school type,

122 The project contributes to evidence-based Systems (DDDAS). DDDAS provides an decision-making in educational planning in adaptive feedback loop framework that Lithuania. In addition, it aims to strengthen covers real time collection of data for model the culture of transparency and dialogue adaptation and new initial conditions. among the educational community. To agree on the assumptions, policy reforms and other The second is the fragmentation and siloing methodological elements to be changed or in model development and utilization. added annually, an advisory expert group Such fragmentation reflects the silos as would play a crucial role in further model and defined for the stakeholder organisations planning cycle developments. and entities by their respective mandate as well the different academic disciplines, Global governance challenges: a technologies and standards. However, case for big modelling whole-of-society challenges that involve a whole-of-government response are Toth T., Comprehensive Nuclear-Test-Ban Treaty Systems-of-Systems complex problems. The Organisation PC complexity and scale of such systems calls Theodoropoulos G., Department of Computer for an integrated approach which would Science and Engineering, SUSTech bring all the different elements together to enable a holistic view and analysis. Siloed Global emergencies present immense policy approaches to modelling isolated processes governance challenges to national, political and phenomena at fixed macro-scales are and operational decision-makers. Modelling not sufficient to understand the dynamics and Simulation has been identified as a of such systems. Instead what is needed to crucial force multiplier in the development understand the overall system dynamics, gain and implementation of preparedness and insights and develop the required predictive response measures. Recent years have capacity is the integration of different models witnessed an explosion in modelling and so that the entire set of actors and factors simulation tools for policy and decision whose interplay at finer spatio-temporal support for global challenges while emerging scales creates the emergent dynamics of the technologies such as IoT and remote sensing system can be analysed in context (butterfly enable data collection and analysis at an effect). In addition to contextual analytics, unprecedented scale. the value-added an integrated approach However, despite these significant to modelling and simulation could provide developments, the current state of the art stems from the need to be able to reconfigure of forecasting and predictive modelling modelling & simulation inputs and outputs has overall failed to deliver the effective including those not normally anticipated outcomes expected. This failure may be This paper aspires to conceptualise the need largely attributed to two major factors. for a data-driven, integrated approach to The first is the grounding of models to data modelling and simulation for global challenge which affects the accuracy and reliability governance. It does this by proposing a of the simulations. Models are typically computational framework that link together developed off line, however new rules, initial the different elements of a fragmented conditions and parameters need to be landscape and support a holistic approach constantly provided to dynamically update to decision making. The paper coins the term and calibrate the models to ensure that they 'Big Modelling' to describe such large-scale accurately reflect the real situation on the ecosystems of models, simulations and data. ground as it unfolds. The need for dynamic Big Modelling invokes a new framework since data driven simulations has been recognised the challenges well exceed the capabilities and there have been significant work in this of conventional analytics approaches and direction culminating in the concepts of Info- call for an intermingling of scalable data Symbiotic or Dynamic Data-Driven Application infrastructures and analytics with multi-scale,

123 distributed and agent based simulations measurable and visible the way intangible engines for the creation of digital twins (Big assets and data are should be correctly Model Twins) at a very large scale. accounted in the European open ledger of value. Modelling European economy as an ecosystem of contracts for smart In particular, with the advent of the digital economy, assessing the impact of a change in policy making rules is becoming more and more challenging. Telesca L., Hazard J., Trakti As an example, some regulations impact the relative importance of some actors in 'We are moving from the old ways of networks, because of a variety of reasons measuring and reporting growth based (competition, consumer rights or other). The on making and selling things (i.e., physical changes and impact may be difficult to see capital), rather than today’s growth drivers of using GDP since it does not integrate the developing and creating human, intellectual, wealth of data produced by networks and and network capital… Our research … clearly market structures, that are increasingly indicates a world where networks and digital intangible and digitalized. Networked assets are more valuable than things and analyses and counterparty information, 'access' is more convenient than ownership. In together with all the other data available are the process of creating more efficient, happy, not currently used by regulators and public and technologically supported lives, we may actors for policy making. In contrast, private have to blow up and recreate how we gauge sector actors (large and small companies) are economic prosperity and growth.'1 2. using the power of data (big data) for real time decision-making and consider more than In a world dominated by internet/technology accounting values or GDP data in analysing giants the challenge for the European markets and spot market opportunities. Commission and other policy and regulatory Private sector actors understand that data bodies in the world is mostly related to the and relationships constitute much of the real inconsistency of current economic indicators value of today’s networked economy. in supporting the legislative activities in many different policy areas. Economic measurement As known, the GDP is aggregated vertically approaches, and in particular Gross Domestic from individual transactions. It measures Product (GDP), with the evolution of todays the reward to factors of production over a economies (from industrial to services to given time span, typically a calendar year. information to network), struggles to account In the case of many policies the European for today’s intangible assets—services, Commission is working on, such a distribution insights, and networks. This is not just of the economic wealth created in a calendar because the way of measuring GDP and may not be much impacted at all. All this Generally Accepted Accounting Principles because with the digitalisation of the (GAAP) are not anymore valid methods for society some new elements are becoming measuring economic growth. Although current increasingly important. measures are presenting to the world a 3 difficult trajectory to digest, that tells that In this context, the servification of the we are all headed into negative territory, industries is making more difficult to track we believe those methodologies need to and trace knowledge, data and value based be updated and reflected by new contract on this networked economy paradigm. structures that could facilitate and make This is because economic players adopted new business strategies, totally digital, 1 Libert, Barry, & Beck, Megan, 'GDP Is a Wildly Flawed Measure for the Digital Age', Harvard Business Review, https://hbr. 3 https://books.google.it/books?id=1ZW5BQAAQBAJ&p- org/2016/07/gdp-is-a-wildly-flawed-measure-for-the-digital- g=PA13&lpg=PA13&dq=servification+definition&source=- age bl&ots=iENeQfGhRx&sig=ACfU3U3xYRrqG4ubDZQB- 2 Libert, Barry, & Beck, Megan, The Network Imperative: How to bap2f7eYQz0cHg&hl=it&sa=X&ved=2ahUKEwicmf65sNfiA- Grow Survive and Grow in the Age of Digital Business Models, hUG3aQKHSlEB14Q6AEwBHoECAkQAQ#v=onepage&q=servifi- Harvard Review Business Press cation%20definition&f=false 124 leveraging on new distributions and value such as that of Ronald Coase’s theory of the capturing models (Servification of the firm4 and of Hart and Holmstrom’s theory of industries, Data economy) that exploit the contract incompleteness5. technological advancements like Service Virtualization, Microservices, Platformization The approach is similar to those that of the organizations. Actors like Google, leading-edge enterprises are beginning to Amazon and Facebook drove those trends adopt – a 'graph' of semantically-labelled to create new collaborative models and data, organised into secure 'data lakes,' with business opportunities. This happened since cryptographic access control and assurances they understood the value of data and the of coherence of data across suppliers, possibility to digitalise relationships and customers, managers and regulators, based 6 services creating new opportunities for atomic on standards and open source software . In transactions. Furthermore, with the advent of essence, our approach demonstrates that an blockchainbased applications and services we entire economy can be modelled in the way will see more and more tokenised securities, that leading enterprises are modelling their rights and assets that will disrupt again the own commercial relationships relationship between ownership and transfer of value. Modelling the social impact of open access knowledge The impact of knowledge, software and data repositories are affecting the soundness of GDP as a comprehensive measure to monitor and asses Skulimowski A. M. J., AGH University of Science the wealth of nations. GDP is not invalid, but and Technology, Chair of Automatic Control is less helpful because the data economy and Robotics, Decision Science Laboratory and is more fluid and not measured correctly by International Centre for Decision Sciences and today’s accounting principles. Therefore to Forecasting, Progress & Business Foundation create a better framework for policy making Ex-ante social impact assessment is and monitoring, we need to leverage data increasingly relevant in the design of and enhance current economic indicators advanced cloud-based knowledge provision with network and contracts information that systems (OKPS) involving a large number could give us a better playing field to test of users. OKPS are an important subclass and assess the impact of policies that EU of ‘ICT-enabled social innovations’ [2]. Their institutions would like to implement in the importance is due to the fact that they are market. capable of reaching diverse social policy goals With our work we want to demonstrate how in a timely, relatively low-cost manner, and an innovative approach can be used both according to common citizens’ preferences. for reporting the European economy as an This is due to a rapidly developing ICT aligned ecosystem of contracts and as an actual with the high and ever-growing digital literacy way to describe the connection between in the European societies, a common access the legal and accounting world, basically a to broadband Internet via mobile networks, boundary object between the two worlds. It is and a rich offer of web-based applications. radically more efficient and has the ability to Therefore the expected impact of OKPS provide precise, qualitative and quantitative financed from public funds is a key decision measures of contracting across the entire factor when designing social and digital European economy connecting relationships, policies at regional, national and EU-levels. data, legal and economic frameworks. In our 4 Coase, R. H. (1937), 'The Nature of the Firm'. Economica, 4: research we use the term 'contracting' in a 386–405. doi:10.1111/j.1468-0335.1937.tb00002.x 5 O.Hart, 'Incomplete Contracts and Control', American very large sense to include quasi-contracting, Economic Review 2017, 107(7): 1731–1752 https://doi. permits, approvals, payments and governance org/10.1257/aer.107.7.1731, Hart, Oliver, and Bengt Holmström. documents, roughly coextensive with the 2010. 'A Theory of Firm Scope.' Quarterly Journal of Economics 125 (2): 483–513 notions of 'contract' in economic literature 6 https://www.oreilly.com/ideas/fishing-for-graphs-in-a-hadoop- data-lake 125 An innovative digital knowledge repository combined with scenario planning and developed within a recent project of the EU- anticipatory impact models in multicriteria financed Horizon 2020 research programme decision processes. [3] can serve as an example of such a social innovation. One of the ultimate goals of this At the macro level, where the impact of repository is to provide an efficient training individual OKPS will be aggregated, a hybrid and research support tool for students and stochastic discrete time control model of the young researchers. Its user community user community growth at the learning group building approach presumes a wide use or its partition level can be used. Statistical of existing cooperation networks centred models of global DES influence [4] can be around current research projects, student also used for cases where the dissemination organizations, and exploring the opportunities activities are addressed to large groups offered by social media. of anonymous and mutually independent individuals, such as Internet ad campaigns or Based on the assumption that the social TV broadcasts. impact of the repository will grow with the number of satisfied users, the efficiency Another mechanism contributing to the of disseminating the information about positive impact and user community the content and functionalities offered to building, is the provision of attractive, novel potential users plays a major role in impact and efficient tools, repository services and optimization. Thus arises a three-level model functionalities. The uses of them will provide of user community building and maintenance. users’ feedback indicating the direction in which the services will evolve so that they At the user level, the information about better fulfil the needs and expectations of an OKPS is disseminated spontaneously different user target groups, stakeholders, via social information diffusion and user and policy goals. According to [7], strategic community growth according to the snowball goal attainment can be evaluated by the set principle. The latter will be formally described of quantitative criteria which should fulfil the by learning cellular automata [1]. A simulation following conditions: procedure calculates the impact of individual community-building activities and sums • better values of each criterion correspond them up over a planning period. Moreover, to a higher satisfaction of user or the cooperation results of agents involved stakeholder preferences regarding the in the repository operation on user group project goals (representativeness), development require synergetic models • every change in user or stakeholder capable of aggregating individual actions satisfaction as regards goal attainment and taking into account non-linearity arising can be expressed equivalently as a from their interference. This makes possible change in values of at least one of the a transition to the meso modelling level, criteria (completeness),and if multiple where the interactions inside of different user redundant criteria with stochastic errors groups are modelled as cellular networks of describe the attainment of the same controlled discrete-event systems (CDES). goal, they must be independent random When performing the simulation at the meso variables (weak non-redundancy). level, CDES agent-based models are coupled with anticipatory networks (AN, [6]). The latter The social policy goals of the OKPS defined serve as a long-term impact assessment in the Digital Agenda 2020 and the EC and policy planning tool. Based on the Social Investment Package (SIP) will be experience gained within the project [3], ANs fostered with increased learning and built with Delphi outcomes [5] turned out to rese-arch efficiency and a higher level be particularly suitable for the overall impact of innovativeness. SIP defines the Social modelling and optimization of OKPS. The Investment for Growth and Cohesion as an ANs simultaneously formalize backcasting initiative to delivering, among the others, high

126 levels of employment, productivity and social concerning the efficiency of social educational cohesion. These are among the social goals of and digital policies implemented at different many OKPS, including the repository [3] which levels. primary goal is reaching a 'decisive impact on the innovative capacity of the European References: society' [3]. The 'impact of design decisions [1] H. Beigy and M.R. Meybodi, 'Asynchronous on learning effectiveness' is the secondary cellular learning automata', Automatica, social goal that directly addresses user and vol. 44, pp. 1350–1357, March 2008, DOI: institutional stakeholder satisfaction with the 10.1016/j.automatica.2007.09.018. repository-supported learning. In addition to individual and group users, any public [2] IESI project ('ICT-Enabled Social Innovation organizations such as schools, hospitals, to support the implementation of the Social business and innovation support institutions Investment Package') web site: https:// can be repository’s stakeholders. The social ec.europa.eu/jrc/en/iesi [accessed: June 18, impact builds on the principle that innovation 2019]. leadership knowledge acquired by the users increases their overall productivity and [3] MOVING project ('Training towards creativity. a society of data-savvy information professionals to enable open leadership Furthermore, the OKPS design, innovation') web site: www.moving-project.eu, implementation, and operation should 2019 [accessed: June 18, 2019]. conform to the general objectives of research policies, specifically those included in the [4] L. Sekanina and T. Komenda, 'Global underlying documents of Horizon 2020 and Control in Polymorphic Cellular Automata,' J. Horizon Europe. For example, 'improving of Cellular Automata, vol. 6(4-5), pp. 301–321, the quality of collaborative research April 2011, http://www.oldcitypublishing. and innovation among the EU research com/journals/jca-home/jca-issue-contents/ institutions' is a requirement that should jca-volume-6-number-4-5-2011/jca-6-4- conform to many OKPS project goals. The 5-p-301-321/. above social goals can be quantified and included in the formulation of optimisation [5] A.M.J. Skulimowski, 'Expert Delphi problems, where the portfolio of OKPS Survey as a Cloud-Based Decision Support project, the guidelines to design individual Service', in: IEEE 10th International knowledge repositories and planning conference on Service-Oriented Computing of operational activities of repository and Applications SOCA 2017, 22–25, managements may occur as following specific November 2017, Kanazawa, Japan, IEEE, Piscataway, pp. 190–197, 2017, DOI: 10.1109/ goals Gj: SOCA.2017.33, https://ieeexplore.ieee.org/

G1 - reaching a given number of satisfied document/8241542. users, [6] A.M.J. Skulimowski, 'Anticipatory Network

G2 - reaching the prescribed number and Models of Multicriteria Decision-Making Pro- quality of services offered, cesses', Int. J. Systems Sci., vol. 45(1) 39–59, 2014, https://www.tandfonline.com/doi/full/10 G3 - reaching the content quality and quantity .1080/00207721.2012.670308. in predefined fields that is assessed as satisfactory by the users. [7] A.M.J. Skulimowski and P. Pukocz, 'Enhancing creativity of strategic decision Finally, an analysis of user responses to processes by technological roadmapping various service innovations and community and foresight', in: KICSS 2012: seventh building activities may provide clues international conference on Knowledge,

127 Information and Creativity Support Systems, to calibrate the parameters of the model. Melbourne, Victoria, Australia, 8–10 November Then, production factors, productivities and 2012, IEEE Computer Society. CPS, pp. key parameters of the model are projected 223–230, 2012, https://ieeexplore.ieee.org/ over time to meet exogenously determined document/6405533. targets, such as GDP, GHGs emissions, etc. In our opinion, this approach presents two PIRAMID: a new methodology to drawbacks. First, the baseline is dependent build baselines for CGE models on the base year data and may not be able to reflect structural changes in the economy Wojtowicz K., Rey Los Santos L., Vandyck T., (e.g. new technologies). Second, baselines Tamba M., Weitzel M., Saveyn B., European are created using a model and, therefore, are Commission, Joint Research Centre dependent on the structure of that model. Temursho U., IOpedia This makes it difficult for other modelling Computable General Equilibrium (CGE) teams to replicate the same baseline. models have become one of the most used This paper presents an alternative tools for economic analysis. Originally CGE methodology to build a baseline. We models were developed for short-term policy propose to reverse the order of the two assessment such as fiscal and trade policies. steps in the conventional baseline building. In this context, base year economic structure In our methodology, firstly, base year data was suitable for the comparison of policy structure (Input-Output tables) is projected scenarios. More recently, the need to address over time and then, the parameters of the long-term-issues, such as energy and climate model are calibrated for each period. This policies, has motivated the development of approach requires the projection of all data dynamic CGE models. When analysing long- conventionally used to calibrate a model. On term policies, base year economic structure the other hand, it provides the flexibility to set may not be valid and, therefore, baselines are the values of the main variables of the model required to assess the implication of different and control the economic structure in the policies in the future. A baseline describes the baseline. evolution of the economy without additional Most of the data commonly used to calibrate policies and serves as a benchmark for a CGE model is gathered in the Social assessing economic, energy and climate Accounting Matrix (SAM) and, therefore, policies. is the key element of our projections. The The increasing interest in dynamic models base year SAM is usually built from GTAP has led to an increasing interest in baselines. database, which is also the benchmark for our It is well known that the numeric results of a projections. The methodology used to project model are very dependent on the economic the SAM is the Multi-Regional Generalized structure and, thus, baselines are essential RAS (MRGRAS), and extension of the well- for any comparison exercise. For example, known RAS method. the cost of a climate mitigation policy The projection of the SAM is subject to in 2050 will depend on the preferences, macroeconomic assumptions: GDP (Private productivities, efficiencies and technological and Public Consumption, Investment, Exports options assumed in the reference path. and Imports), tax rates, capital and labour Although the importance of the baselines payments, labour force, unemployment rates, is well known, still there is a wide range of population, etc. These data are obtained different methodologies which try to build a from external sources and, ideally, should be baseline. This lack of consensus hinders the consistent. In addition to the macroeconomic comparative assessment between models. assumptions, we also impose energy data Roughly speaking, conventional baselines are projections from partial equilibrium energy built in two steps. First, base year data is used

128 models such as POLES, PRIMES and/or 1. Flexibility. It allows for a better control of POTENCIA. We fix the use and production of the variables we are interested in and for energy products in the SAM. Energy data is introducing structural changes. balanced not only in value terms but also in 2. Reproducibility. The projected SAMs are quantities and, thus, we can replicate CO 2 not model dependant and can be used by emissions estimated in energy models. other modelling teams. To our knowledge, this is the first attempt to 3. Transparency. The baseline is built create a baseline for CGE models projecting based on transparent macroeconomic base year data. This methodology presents assumptions and energy data. In case of three advantages compared to conventional discrepancy, these assumptions can be baselines: corrected and improved.

129 130 Plenary session 2

Room 0.A 27 November 09:30 – 10:30

131 132 Assessing social impacts of also for this dimension. Firstly, because policies: indicators and methods policies that imply a structural change to society, i.e. create winners and losers, have to Klaus Jacob, Research Director, Environmental be assessed against the benefits to society Policy Research Centre (FFU), Freie Universität as a whole. It should be asked in how far Berlin the gains for one group could compensate the losses for others. Furthermore, impacts The ex ante impact assessment of policies on those groups that are not organized and has become a standard procedure of the thereby cannot articulate their interests policy making process within the European needs special attention. Commission as well as in many countries across the globe. What has begun in many A social impact assessment would ask for countries as an analysis of the economic processes that are triggered or changed by costs of regulation, has now been expanded policies and their impacts on different groups as a comprehensive assessment of all in society. Typical classes of processes are economic, environmental and social impacts of planned policies. The analysis of the • Changes in economic processes expected impacts and the comparison of • Changes in natural systems different policy options based on scientific methods and evidence should inform policy • Geographical changes makers and politicians about the likely • Demographic changes consequences of their actions. It should also inform stakeholders and the public about the • Institutional changes net benefits of a planned policy. • Emancipatory changes

The assessment of economic and budgetary The list shows that social impacts are closely impacts can build on solid databases and interlinked with economic and environmental elaborated modelling tools. For many (not aspects of an impact assessment. all) environmental aspects, similar capacities have been developed. For both dimensions, The impact categories are, however, different. there is an articulated demand from policy Social impacts include: makers and stakeholders to summarize the • Health and wellbeing evidence on likely impacts. • Income For social aspects, the situation is different. The social dimension is in many policy impact • Social environment assessments the least elaborated. There • Impacts on family and community could be several reasons for this: Missing • Institutional and political impacts tools and data, a lack of demand or a poor conceptualisation of social impacts. Unlike Finally, groups can be distinguished along for economic and environmental impacts, for different characteristics, e.g. social aspects it is often not straightforward to attribute social impacts as a desired • Socio-economic and demographic status outcome or not. The acceptable level of • Types of households inequality in society, sources of individual well-being, even the marginal value of income • Position in the economic system is a matter of diverging perspectives. Hence, • Property rights unlike for economic impacts, the distinction between unwanted and desired impacts is in • Geographical aspects many cases not straight forward. • Preferences

There are, however, good arguments to Socio-economic aspects are far more often improve the evidence base for policy making analysed than those aspects that are related 133 to societal participation. Typical indicators the cause to effect relation is most often for socio-economic aspects are income, unclear and often contingent to the contexts. employment or health. Participation, social Therefore, qualitative methods that answer cohesion and well-being in society is, however, such questions should be considered in the not only a matter of income, but also a standard repertoire of Impact Assessments. matter of access to social processes and has a cultural dimension. These aspects are far The talk includes examples for impact less often considered in impact assessments. assessments using microsimulations for They would require a qualitative approach to the analysis of socio-economic impacts, impact assessment. Although comprehensive indicators and indices for well-being and indicators on well-being have been developed, examples for applying qualitative methods.

134 Parallel sessions 4

Complex system modelling and multi-criteria decision making 2

Room 0.A 27 November 11:00 – 12:30

135 136 Making sustainability models indicators: (i) gross inland energy use more robust: dealing with the per capita and (ii) GDP p.c. per year. The following groups of countries each have complexity of the metabolic a same value of EEI: (1) Guatemala, pattern of social-ecological the Netherlands, Germany, Angola, systems Norway, Chile; (2) Sweden, Macedonia, France Azerbaijan, Egypt, Argentina; Giampietro M., Institute of Environmental (3) Australia, Algeria, Finland, Malaysia, Science and Technology (ICTA), Universitat 2 Autònoma de Barcelona and Catalan Institution USA, Turkey . This shows that, at the for Research and Advanced Studies (ICREA) national level, the EEI does not have any discriminatory power to characterize Problem framing societal performance. Why then is the EEI expected to provide useful information The practical consequences of ignoring the for studying the energy performance of implications of complexity in developing countries? quantitative analysis for sustainability policies are illustrated with three blatant cases: 3. EU scenarios of decarbonization. The scenarios provided by the EU predict 1. Used Cooking Oil (UCO). The a monotonic decrease of emissions implementation of the revised Renewable from now to the year 20503. Such a Energy Directive (RED II) is under scrutiny: dramatic reduction in emissions requires UCO has been mixed with palm oil to a radical and quick transformation of take advantage of favourable policies1. the whole economy. The vast majority of This fraud could have been anticipated power capacity for both production and if the relation between primary sources consumption of energy carriers will have and secondary energy carriers had been to be replaced in less than 30 years using considered. (i) Availability of cooking oil fossil energy. However, no emissions for (primary source) in the EU is about 5–6 kg this Olympic effort are considered in the per capita/year. Of this, only 3–4 kg are scenarios. What type of models have theoretically recoverable as UCO. (Austria been used? is the EU country that recovers most UCO with 1 kg p.c. per year). (ii) The production Some systemic epistemological blunders process of secondary carriers (the emerge from these examples: biodiesel) entails that only 75% of the UCO can be transformed in biofuel and 1. Ignoring the forced State-Pressure only 70% of the energy of this biofuel is relation between primary energy sources net (because of the energy consumed in and secondary energy carriers (UCO case). the process itself). Hence, the estimated One cannot study energy systems without production of biodiesel per capita per adopting simultaneously non-equivalent year from UCO in the EU is between 0.5 metrics: For studying external constraints kg (1 x 0.75 x 0.7 using the Austrian 'best – primary sources; for studying internal practice') and 2 kg (assuming an unlikely constraints – the process of production quadrupling of this benchmark). In and use of energy carriers; conclusion, UCO is a waste management 2. Ignoring the multi-scale and open nature problem, not an alternative energy source. of complex adaptive systems organized 2. Use of the Economic Energy Intensity across different levels. Looking only (EEI) to study decoupling. The EEI is a at one level of analysis and ignoring combination of two strongly correlated the openness of the system results in simplistic representations. 1 https://www.euractiv.com/section/agriculture-food/news/ eu-throws-the-ball-to-member-states-to-monitor-red-ii-imple- 2 https://doi.org/10.1016/j.jclepro.2012.12.031 mentation/ 3 https://ec.europa.eu/clima/policies/strategies/2050_en

137 3. Ignoring the biophysical roots of the aspect). Relevant information here is: (i) economic process entails missing the side what are the available primary sources effects of important transformations, and sinks for domestic production? (ii) such as rapid decarbonization. Changing what are the imports of both primary the quantity and quality of flows requires sources and secondary carriers? changing the quantity and quality of the fund elements that produce (power This new generation of quantitative analysis plants), deliver (batteries, distribution is being developed in the Horizon 2020 lines) and use these flows (e.g., means of project ‘Moving towards Adaptive Governance transport, infrastructure and appliances in Complexity: Informing Nexus Security’ in residential, industrial processes). The (MAGIC). Its analytical tool kit combines five production, maintenance and replacement conceptual tools that integrate the use of of fund elements must always be non-equivalent metrics: considered, in terms of both emissions 1. local end-use matrix (how society uses and energy requirements. energy, water, food, labor and land) Can we do better? across its internal components);

A different generation of quantitative models 2. bio-economic pressure matrix (what share is needed to obtain relevant information for of the total consumption of secondary sustainability. These models must establish input goes in boosting the material a series of bridges across non-equivalent standard of living); representations of the performance of energy 3. local environmental pressure matrix systems: (the use of primary sources and sinks 1. Between final energy consumption and affecting the integrity of local ecological material standard of living by looking funds); at internal end-uses across levels. The 4. externalized end-use matrix (the relevant information here is: (i) who uses secondary inputs used by the exporting what energy carriers, (ii) how effective is society embodied in imported products their use (qualitative); (iii) how much of and services); each type of carrier is used (quantitative); (iv) what is the purpose of each end-use 5. externalized environmental pressure (the functions associated with energy matrix (the primary sources and sinks uses). embodied in imported products and services). 2. Between the environmental pressures associated with the energy end-uses in MAGIC’s accounting method uses concepts society and impacts by looking at local from complexity (relational analysis, hierarchy effects of the pressures (use of primary theory, bioeconomics) to characterize the sources and sinks) on the integrity of metabolic pattern of social-ecological ecological funds. The relevant information systems. It is transparent in terms of here is about tracking across levels and assumptions and data used to integrate scales (in GIS) the different pressures information across scales related to: (i) the in relation to resulting environmental energy, water, food, land nexus; (ii) different impacts. types of environmental loadings/impacts; (iii) demographic and socio-economic variables. 3. The level of externalization of the It also integrates the analysis of industrial, energy system by looking at dependency urban and household metabolism in terms on imports (security aspect) and of social practices. These concepts will be externalization of emissions and impacts illustrated using applications developed in the to other social-ecological systems (ethical MAGIC project.

138 Supporting result-based schemes. methodologies. First, a majority of studies The case of ex-post assessment are based on case studies which apply different proxies for environmental quality of agri-environmental-climatic changes, and few studies use composite schemes and overall indicators. Second, existing Bartolini F., Vergamini D., Department of studies apply several methodologies with Agriculture, Food and Environment, University a plethora of alternative assumptions and of Pisa restrictions. The methods differ in terms of Longhitano D., Povellato A., Council for the data used, the measurement proxy for Agriculture Research and Analysis of environmental changes, the possibility of Agricultural Economics, Italy building a counterfactual, the complexity in creating policy scenarios, and, finally, the The current Common Agricultural Policy interpretation of causality. The literature (CAP) architecture includes three measures highlights several causes and barriers which that aim to not only improve positive can reduce the expected benefit of an AECS. environmental externalities but also reduce These motivations can encompass both negative environmental externalities: a) policy failures (in each step of the policy cross-compliance, b) greening, and c) agri- cycle) in private motivation and opportunistic environmental climatic schemes (AECS). The behaviour as well as in asymmetric proposal for the post-2020 CAP indicates a information. more flexible and results-based approach (i.e. measures that focus on payments for Against this background, we estimate a results achieved), giving member states more composite indicator of high nature value at room to manoeuvre to adapt measures to farm level (HNVf) that enables us to track local conditions and environmental priorities. changes at the farm level, and then assess This debate has pushed more focus to the contribution of AECSs to these changes AECS design by highlighting results-based ex post. We use a sample of panel data measures with payments designed based on from the Farm Accountancy Data Network the environmental quality provided. (FADN) for two Italian regions (i.e. Veneto e Tuscany) from 2008 to 2014. In this study, The results-based approach renews the we estimate the effects of different payment role of research in providing reliable ex- levels on a subgroup of farms by applying the post analyses to enable the tracking of generalised propensity score (GPS) approach. changes in environmental quality due to More specifically, we first estimate the GPS agricultural practices and measuring the using a generalised linear model, and we contribution of the CAP measures to these then estimate the dose–response function changes. Measurements of environmental using a flexible parametric form for the performance and the causal effect of regression function of the outcome on the AECSs on environmental performance treatment and the GPS. As AECS payments are largely debated at the academic and are tailored to compensate for foregone political levels. However, the existing costs and income due to participation in evaluation and monitoring system seems environmentally friendly measures, we unsuitable for supporting this new role of assume that the benefit expected from these evaluation. Although the impact of the CAP schemes increases linearly with the payment on sustainability is traditionally addressed received. Thus, we want to observe whether by the agricultural economics literature, increasing payment levels leads to further its contribution to the environment quality environmental benefits or, alternatively, if is still debated, and the literature has not some sub-optimal payment levels do not converged to a consensus. Two factors can maximise environmental benefits because of contribute to the difficulty of comparing the relationship between several sources of the findings of different studies: the use of failure that can arise in designing AECSs. case studies and the application of different

139 Our results demonstrate that several of a different evaluation process for AECSs, agricultural systems and payment levels perhaps by combining advanced modelling affect the provision of HNVf. The results show of the causality of AECSs with reflexive a significant effect of agri-environmental exercises by engaging the relevant actors and schemes on composite indicators of HNV stakeholders. but an uneven distribution across various payment levels. Our results confirm puzzling Exploring the (non-)use and evidence regarding the effect of AECS influence of models in Belgian payments, as alternative payment levels climate policy-making: a multiple have different impacts on HNV. Altogether, the results confirm previous studies showing case study an overall positive effect of AECSs on HNV, Fransolet A., Centre for Studies on Sustainable but a lack of design and targeting (Raggi et Development (DGES-IGEAT), Université Libre de al. 2015) and an agency problem (Bartolini Bruxelles et al. 2012) lead to a lower impact of the measures. Our results can contribute to the Since various public and private actors at the debate on the effectiveness of different international, supranational, national and types of measures, showing that narrow subnational levels started to adopt long- and deep measures might bring higher term targets for reducing greenhouse gas environmental benefits at the farm level. emissions, low-carbon scenario analyses have However, the results justify further additional flourished. Literature reveals an increasing public transaction costs for improving AECSs’ number of model-based analyses envisioning tailoring and targeting. In fact, our results and exploring alternative images of low- show that the instrumental mix results carbon futures, as well as their adjacent in a higher provision of an environmental transition pathways. Scenario approaches or good but requires better coordination and 'foresight' is intended to help policy-makers communication to farmers. to navigate the maelstrom of confusion and conflicts associated with highly complex Using an evidence-based approach to societal challenges such as climate change support policy reform remains a central — i.e. the 'super-wicked' problems. Typical topic in policy discourse and is now legally scenario exercises aim at coping with established within the new delivery model. uncertainty and conflicting values, and hence Consequently, further development in are often claimed as a suitable approach understanding causal mechanisms within for knowing and governing super-wicked the CAP is necessary. In this study, we problems. aimed to contribute to that debate by developing a methodology that enables us When reviewing the scenario literature to account for the multidimensionality of published over the recent years, we observe the provision of public goods and measure significant methodological developments, in the causality of public expenditure. Thus, particular at the level of the calculus or data- although the transition towards more results- sets. These contributions have generated based policy design is desirable to better an increasing technical sophistication of understand the linkage between payments models and scenario building methods, which and environmental performance, the current contrasts with the relative absence of social methodologies and the data collection sciences research on scenarios. Scenario infrastructure are not completely satisfactory analyses have received little academic for ensuring the application of results-based attention from social sciences, whether they payments. Moreover, our study indicates a are political science, sociology, philosophy rather large complexity in detecting policy of science or STS. More specifically, even if failures based on ex-post data owing to foresight products and processes are claimed several difficulties in disentangling the effect as potentially contributing to policy-making, of each cause of failure. This calls for the use relatively little empirical efforts have up to

140 here tried to explore the fundaments of the policy-making and the level of appropriation role of the scenario analyses throughout per actor for each of the four foresight the policy cycle. The present paper aims at studies, I have conducted a cross-case contributing to fill this research gap. analysis in order to highlight the similarities and differences between the four studies. The research questions are formulated as follows: How are low-carbon scenarios The cross-case analysis reveals that the used by political actors? Beyond the purely role in policy-making varies strongly from instrumental use, how do these foresight one foresight study to another. However, exercises influence policy-making? How few examples of purely instrumental use of do the different actors appropriate the studies as orientation and decision support representations of the future developed tools to devise mitigation policies are noted. through these scenario analyses? What are Studies are rather used to justify decisions the factors that affect the use and influence already taken or to improve someone’s of these foresight exercises in policy-making? relative position in the policy systems compared to opponents (i.e. political role). The paper rests on a multiple case study Studies also contribute to expand the analyzing the role in policy-making of four knowledge base and introduce new ideas energy foresight studies carried out for or concept in policy (i.e. conceptual role). Belgium. The chosen studies differ inter Moreover, even if none of the analyzed alia in terms of clients, contractors, year of studies rest on real exercise of scenarios publication, scope, approach (i.e. forecasting co-construction, some studies still have or backcasting), methodology (i.e. bottom-up a process role. The implementation gap modelling, top-down modelling or qualitative between low-carbon scenarios and policy methodology inspired by the work of the is not surprising since the relationship French prospectivists Hugues de Jouvenel between sciences and policy is neither linear, and Michel Godet), level of participation nor mechanic — the vision of a rational and actors involved (i.e. from expert-based knowledge-based policy-making model being analysis to participative exercises), product, rather naïve. The purely instrumental use communication strategy, and follow-up. of scientific knowledge in policy-making is The analysis has been carried out on the generally quite limited, especially in highly basis of an analytical framework developed politicized context like Belgium. Instead, in the knowledge utilization literature (see knowledge usually has a more diffuse and Gudmunson & al., 2009). This framework indirect influence on policy-making — i.e. considers five types of knowledge role conceptual and process roles. That said, in in policy-making, namely instrumental, the present research, such conceptual and conceptual, political, process and distortive process roles were observed among nearly roles. The exploration of the use and influence all the actors, political authorities excepted. of low-carbon scenarios in policy-making The cross-case analysis actually shows is based on documentary analysis and 74 that the level of appropriation of foresight semi-structured interviews with the actors studies varies greatly from one actor to involved in the long-term climate mitigation another, and that the public actors within policy in Belgium. Those actors are political administrations seem to be more permeable authorities, political advisors, administrative to the investigated foreknowledge, whereas, authorities, employers’ federations, workers’ on the contrary, political authorities mark unions, environmental and development little interest in the scenario exercises. Finally, NGOs, advisory councils, energy market the cross-case analysis suggests that a operators, public transport operators, public specific constellation of factors contributes research institutes, consultancy agencies and in explaining the (non-)use and influence of universities. The important empirical material each study. That being said, it seems that the collected was analyzed with the method of particular characteristics of each study affect thematic coding. After exploring the role in

141 less the role of the studies in policy-making provides a natural description of the real than contextual factors. world and that is able to capture emerging phenomena. However, the complexity of The last section of the paper focuses more models created through an ABM approach specifically on the factors that can contribute is limited by our capability to analyse and to explain the very low level of appropriation understand the respective models’ output. of low-carbon scenarios by political If the complexity of the model is reaching authorities. In this discussion, I explain a level where we are no longer able to that the low levels of interest of political understand the processes involved, we cannot authorities in low-carbon scenarios can be understand these artificial complex systems related to the very nature of both foresight any better than we understand the real ones and climate mitigation governance. On the (Gilbert and Terna 2000; Axtell and Epstein one hand, I highlight a mismatch between 1994). At this point, the model decays to foresight and policy-making that makes the a meaningless construct without any real integration of foresight in political practices scientific value, which in the best case can difficult — if not impossible. On the other be used to visualise our little understanding hand, I point out a number of challenges of the matter. Consequently, our ability inherent to climate mitigation governance to analyse the output of complex agent- that lead to political inertia. based models over a large set of different parameters becomes a bottleneck, especially In turn, by providing empirical insights on when applying ABM for innovation policy the (non-)use and influence of model-based purposes. scenario analyses in policy-making, the paper could be instrumental in learning for the One approach that may help us overcoming modelling community. the natural limitations of traditional means of output analysis are recent advances in the Assessing the potential of machine broad field of machine learning and data- learning algorithms for agent- mining tools. These tools may complement based models from an innovation ABMs, especially in the analysis, of model policy perspective outcomes (van der Hoog 2018). Potential applications of machine-learning algorithms Müller M., Bogner K., Pyka A., University of for ABMs are manifolds. They range from Hohenheim variable selection (e.g. Pereda et al. 2017; Kudic M., University of Bremen Edmonds and Lessard-Phillips 2014; Patel et al. 2018) to the use of machine-learning Especially in the field of evolutionary/ tools for validation/verification of ABMs neo-Schumpeterian innovation economics, (e.g. Baqueiro et al. 2009; Remondino and agent-based modelling (ABM) has been Correndo 2006), or even to the automated recognized as one of the most promising analysis of results by emulating the behaviour tools for investigating and capturing the of the whole model (van der Hoog 2017). complex nature and resulting dynamics of processes related to innovation activities In our paper, we aim at identifying the (Morone and Taylor 2010). While using agent- potential of using machine-learning based models without doubt helps improving algorithms for complementing agent-based the understanding of the dynamic relations models in innovation policy research. To be between micro-processes and the emergence more precise, we explicitly try to answer: of macro patterns, researchers generally face 'Which potential role can machine learning a trade-off when applying ABM to complex play in innovation research, in general, and phenomena such as innovation processes. how can ANNs help discovering new patterns ABM allows for creating complex models that from big data sets created by agent-based omit the underlying restrictions of traditional models, in particular?'. To do so we apply a economic modelling approaches. With ABM set of machine-learning tools to an agent- we have a tool at hand that is flexible, that 142 based model designed for the systematic Patel, M. H., Abbasi, M. A., Saeed, M., & Alam, exploration of knowledge exchange/learning S. J. (2018). 'A scheme to analyze agent- and the complex processes affected by based social simulations using exploratory various types of systemic innovation data mining techniques'. Complex Adaptive policy interventions, called VISIBLE model Systems Modeling, 6(1), 1. (Virtual Simulation Lab for the Analysis of Investments in Learning and Education) (Pyka Pereda, M., Santos, J. I., & Galán, J. M. (2017). et al. 2018; Pyka et al. 2019). Based on this 'A brief introduction to the use of machine example, we aim to show how learning techniques in the analysis of agent- based models'. In Advances in Management ABM can be used for improving our Engineering (pp. 179–186). Springer, Cham. understanding for the complex interplay of innovation processes and, therefore, for Pyka, A., Mueller, M., & Kudic, M. (2018). innovation policies. 'Regional Innovation Systems in Policy Laboratories'. Journal of Open Innovation: Preliminary results indicate that the Technology, Market, and Complexity, 4(4), 44. application of machine learning tools to the analysis of ABM outputs offer some Pyka, A., Kudic, M., & Müller, M. (2019). interesting insights and may indeed enhance 'Systemic interventions in regional innovation the interpretation of the results as well as systems: entrepreneurship, knowledge the researchers’ understanding of specific accumulation and regional innovation'. patterns within the model’s output. This, Regional Studies, 1–12. however, only strongly depends on the specific Remondino, M. and Correndo, G. (2006). purpose a researcher pursues and within 'MABS Validation through Repeated Executing narrow boundaries. and Data Mining Analysis'. International References Journal of Simulation Systems, Science & Technology 7(6), 10–21 (2006). Axtell, R.L. and Epstein, J.M. (1994). 'Agent- based modeling: understanding our creations'. van der Hoog, S. (2018). 'Surrogate modelling The Bulletin of the Santa Fe Institute, 9(2), pp. in (and of) agent-based models: A prospectus'. 28–32. In: Computational Economics, pp. 1–19.

Baqueiro, O., Wang, Y. J., McBurney, P., & Machine learning in the service of Coenen, F. (2009, July). 'Integrating data policy targeting: the case of public mining and agent based modeling and credit guarantees simulation'. In Industrial Conference on Data Mining (pp. 220æ231). Springer, Berlin, de Blasio G., Ciani E., D’Ignazio A., Andini M., Heidelberg. Bank of Italy

Edmonds, B., Little, C., Lessard-Phillips, L., & Machine Learning (ML) tools (Hastie et Fieldhouse, E. (2014). 'Analysing a complex al., 2009; Varian, 2014) are increasingly agent-based model using data-mining used to address prediction problems in techniques'. In Social Simulation Conference. applied econometrics. In some instances ML techniques can be used to assist Gilbert, N. and Terna, P. (2000). 'How to decision makers, by providing them with a build and use agent-based models in social decision rule that summarizes the available science'. Mind & Society, 1(1), pp. 57-72. evidence in order to predict which choice is more likely to serve the purpose. This task Morone, P. and Taylor, R. (2010). Knowledge is what Kleinberg et al. (2015) define as diffusion and innovation. Modelling complex 'prediction policy problem'. When the decision entrepreneurial behaviours. Cheltenham: concerns policy targeting, ML methods can Edward Elgar. be employed ex-ante to identify, among

143 the potential beneficiaries, those who will (balance-sheet) dataset, and develop two likely behave in such a way as to ensure the separate ML prediction models, for credit effectiveness of the intervention. constraints and creditworthiness, respectively. All the variables that we use for predicting In this paper, we focus on the 'prediction each status could be potentially available to policy problem' of assigning public credit the GF administration when a firm applies guarantees to firms. Public guarantee for the guarantee. We consider a firm to be schemes aim to support firms’ access to bank credit constrained if the total amount of bank credit by providing publicly funded collateral. loans granted to that firm does not increase The literature has highlighted that these in the six months following a new request for schemes often fail to reach firms that are bank credit from that firm, while we consider actually credit constrained (see, for instance, it credit worthy if it does not have adjusted Zia, 2008). If the guarantee is provided to bad loans in the three-year window following firms that are not credit constrained the the request. We try different ML algorithms to additionality of the program will languish predict each status — LASSO, decision tree, as these firms would have obtained and random forest — and show that the best funding anyway. One of the reasons for this out-of-sample predictive performances are misallocation is that credit rationing is difficult reached with the latter. The predictions for to gauge, while firms’ creditworthiness is financially constrained firms are combined more easily assessed by means of balance with those for creditworthy firms to identify sheet variables. As a result, the eligibility the ML hypothetical beneficiary of the GF. condition usually winds down into naïve rules By comparing the GF assignment with the that pinpoint financially sound borrowers, ML assignment, we show that the GF scoring without considering indicators for credit system is biased against firms that are credit constraints (OECD, 2013). Our exercise aims constrained. at suggesting a benchmark assignment mechanism, based on ML algorithms, that In the second part of the paper, we explicitly accounts for both credit constraints substantiate the validity of our approach and creditworthiness. The nature of this by looking at the ex-post dimension. As task is essentially a forecasting one. As underscored by Athey (2017), ML prediction underscored by Mullainathan and Spiess will not automatically ensure higher program (2017), these prediction policy problems effectiveness because a program might are the ones for which the ML machinery is have heterogeneous effects and ML might extremely well equipped. fail to target those for whom intervention is most beneficial. We provide ex-post We compare our ML-based assignment empirical evidence to test whether that mechanism against the rule originally put ML-based assignment mechanism satisfies in place. The advantages of the ML tool we the aim of increasing the impact of the propose can be shown by comparing its policy. We start by showing the results from performance to that of the current allocation contraction and re-ranking experiments, in rule adopted by the Italian Guarantee Fund the spirit of Kleinberg et al. (2018). Through (GF) to select firms eligible for public support these exercises we estimate the increased when accessing credit. effectiveness that could be attained by excluding some current beneficiaries that are In the first part of the paper, we work as if not ML targets, and (under the assumption we were in the ex-ante situation, in which of selection on observables) by substituting the policymaker must design the allocation them with firms that are not treated under of the guarantee without prior knowledge of the GF rules, but that should have been the intervention effectiveness. We make use eligible for the collateral according to ML. of micro-level data from the credit register Next, to relax the selection-on-observables (CR), kept at the Bank of Italy, and the Cerved assumption, we exploit the threshold for

144 assignment implied under the GF rules and Kleinberg, J., Ludwig, J., Mullainathan, S., run a Regression Discontinuity Design (RDD) and Obermeyer, Z. (2015), 'Prediction policy experiment, separately by ML-targeted and problems'. American Economic Review, 105(5): non ML-targeted groups of firms. We find that 491–495. effectiveness is higher for the firms identified by ML as targets. Kleinberg, J., Lakkaraju, H., Leskovec, J., and Mullainathan, S. (2018), 'Human decisions and We show that around 47 per cent of the machine predictions'. The Quarterly Journal of resources currently allocated by the GF rule Economics, 133(1): 237–293. go to firms that are not a target according to our ML algorithms. By channeling these Mullainathan, S., and Spiess, J. (2017), resources to other firms identified as ML- 'Machine learning: An applied econometric target, the effectiveness of the policy approach'. Journal of Economic Perspectives, improves significantly. 31(2): 87–106.

We discuss the importance, in our case, of OECD (2013), SME and entrepreneurship other issues that are typically related to financing: The role of credit guarantee the use of ML for policy decisions, such as schemes and mutual guarantee societies in transparency and omitted payoffs. supporting finance for small and medium- sized enterprises, Final Report, January 2013. References Varian, H. R. (2014), 'Big data: New tricks Athey, S. (2017), 'Beyond prediction: Using big for econometrics'. Journal of Economic data for policy problems'. Science, 255(6324): Perspectives, 28(2): 3–28. 483–485. Zia, B. H. (2008), 'Export incentives, financial Hastie, T., Tibshirani, R., and Friedman, J. constraints, and the (mis)allocation of credit: (2009), The elements of statistical learning: Micro-level evidence from subsidized export Data mining, inference, and prediction. New loans'. Journal of Financial Economics, 87(2): York: Springer. 498 –527.

145 146 Parallel sessions 4

Engaging stakeholders and policy makers

Room 0.B 27 November 11:00 – 12:30

147 148 Complex modelling to achieve block with the majority of EU countries, carbon neutrality in Portugal becoming more efficient and productive although without significant changes Seixas J., Fortes P., Ferreira F., Tente H., in its economic structure, circularity Monjardino J., Gouveia J.P., Dias L., Palma P., levels increase, resource efficiency is Lopes R., CENSE, NOVA University Lisbon improved, and severe mitigation policies Avillez F., Aires N., Vale G., AGROGES, Sociedade are considered. Each storyline backs a set de Estudos e Projetos, Lda. of socio-economic indicators, validated Martinho S., Barroso J. E., Lasting Values, Lda. by key Portuguese officials, including Barata P., Get2C+ the cabinets of Ministries of Economy, In 2016, one year after the signature of Paris Finance and Environment and the Agreement, the Portuguese prime-minister Portuguese Bank, getting them involved undertook the policy goal of transforming along the envisioning process. Portugal into a carbon-neutral economy 2. Gathering key-stakeholders’ visions for by 2050. A team of 28 national scientists 2050 through seven visioning workshops and experts from different fields (energy, covering: mobility, forest, agri-food agriculture, forest, economy) conducted industry, cities, construction industry, technical studies, including modelling, to waste and energy. More than 160 support the Roadmap for Carbon Neutrality participants from more than 100 entities (RCN2050). This offers the vision and the (private companies, central and local pathways on how the neutrality goal can be government, research institutions and achieved in Portugal, concerning technological NGO) were involved. These collaborative options, economic feasibility, and the need events allowed to define, corroborate for social transformation, while considering and refine assumptions regarding circular economy (CE) as central piece. fundamental activity variables, which represented relevant drivers for the This paper shows how modelling tools, linked modelling exercises, as was the case with a deep stakeholders’ engagement of CE assumptions (e.g., efficiency approaches, have delivered plausible carbon silvicultural practices; recycling rates neutral futures to Portugal. RCN2050 with impact on waste and on energy was constructed through an integrated technologies; industry production framework, embracing five stages. levels, namely cement, glass and paper; 1. Designing socio-economic storylines and households sharing models with effect on quantification of the underlying indicators. household size and equipment’s stock). More than 20 national representative 3. Quantitative assessment of pathways stakeholders have co-created plausible to deliver carbon neutrality by 2050. visions up to 2050 for the Portuguese The greenhouse gas (GHG) emissions society through a structured workshop scenarios of the energy system were and direct interviews. Three distinct generated using the bottom-up, storylines regarding the country’s optimization model TIMES_PT, assuming development were designed: i) Off- the continuation of the 2020 climate Track, Portugal keeps a modest socio- policy in the Off-Track scenario, and a economic growth and is out of the way reduction of 90% of GHG emissions by in the mitigation action; ii) Yellow Jersey, 2050 compared with 2005 values, for the Portugal has a leading position in climate other scenarios. This target corresponds mitigation and in economic development, to the maximum feasible reduction of pushed mostly by creative and knowledge GHG emissions computed by TIMES_PT economy, circularity of economy is without considering negative emissions obtained through the redesign of the or a contraction of energy services productive processes carrying high levels demand. Novel improvements were of efficiency; ii) Pack, Portugal moves in made, particularly the linking between

149 stakeholders’ vision with modelling mobility, causing also a severe decline of NOX inputs and the inclusion of circular emissions (more than 95%) and allowing to economy strategies, such as household’s fulfil air emission targets. Power sector is equipment and car sharing options. For sustained by renewables (mostly solar PV and waste, agriculture and forestry, GHG wind onshore), representing 80% and 100% emissions were estimated based on of electricity production in 2030 and 2050, specific scenarios’ drivers (e.g., evolution respectively. Electrification of economy is a of agriculture productivity, afforestation key decarbonization pillar, increasing from area, recycling rate) and tools (e.g., FOD 26% nowadays to more than 66% by 2050 model for CH4 decay in landfills). (% of electricity in final energy consumption), with important impact in energy dependence 4. Public consultation of the pathways up and the decrease of fuel imports (900 M€/ to 2050, in which stakeholders, including year in average between 2030-2050, a civil society, were able to express quarter of today’s values). CE associated to their views on modelling assumptions, waste management and to precision, organic constraints, and expectations, both online and conservation agriculture, likely declines and during dedicated seminars. More than waste and agriculture emissions up to 70% 80 inputs were received. and 37%, respectively. Likewise, the reduction 5. Assessment of uncertain aspects of livestock is a significant decarbonization raised during public consultation, vector. Forest fires should be kept below 70 through the reformulation of modelling 000 ha/year (vis-à-vis the average 131 000 assumptions and the construction of ha/year of the last 10 years), to preserve the multiple alternative scenarios. Examples sequestration ability of the forest, otherwise include: assumptions on CE (e.g., forest neutrality will be hard to achieve. Although productivity and use of biomaterials; limited, it is estimated a positive impact on shared mobility and freight load factors; economy (0.9% of GDP), stimulated by up to use of secondary materials in industry), 7% growth of investment compared with a the relevance of carbon capture and non-neutrality pathway. storage, hydrogen, electricity imports/ exports, forest fires, animal protein-based These outcomes led to the approval of a diet, among other. For the core Pack Ministries Resolution. Besides reaffirming and Yellow Jersey scenarios, the hybrid the political commitment with the carbon computable general equilibrium GEM-E3_ neutrality goal for Portugal, decarbonization PT was used to access the economic targets for 2030 (around 50% reduction face impacts of carbon neutrality, assuming to 2005), and 2040 (around 70%) were also a soft-link approach with TIMES_PT and settled. sustained by the projected energy system transformation and investment needs. System dynamics modelling as At the end, more than 65 scenarios were strategic vision for transforming generated for Portugal up to 2050. Carbon adult social care in Northern neutrality can be achieved until the middle of Ireland: from primary research to the century, through feasible cost-effective procurement technologies and plausible investments. All sectors of the economy contribute to Wylie S., Strategic Investment Board emissions reductions, although at different Martin J.,Department of Health, Northern pace over time. Energy supply and transports Ireland are the main contributors to decarbonisation, Policy Area & Background contrary to agriculture and industry, the latter due to process emissions. Energy The Department of Health (DoH) in Northern intensity reduces more than 55% up to Ireland is currently taking forward a process 2050 face to today, driven mostly by the to reform adult care and support as part of significant efficiency of electric and shared 150 a former Minister’s ten year vision Health coproduction with service users. The models and Wellbeing 2026: Delivering Together. have been designed so that people with Adult social care describes the suite of a learning disability can understand the activities that enable people with limitations implications and interact with scenarios. to live independently, for example, help with Coproduction is mandated for the design of washing, with dressing or with eating. This all health and social care services in Northern is an issue faced by most European nations Ireland, and SD modelling is providing a because a frail and aging population presents pathway to authentically achieving this. a troublesome demographic in terms of who cares for whom, when, in which homes, Results under what terms, and through which sort System dynamics modelling has created a of economic or social contracts. Unusually, dynamic empirical picture of the stresses in health and social care are integrated in adult social care system in Northern Ireland, Northern Ireland, unlike the rest of the UK many of which are driven by broad stroke where they form separate programs and demographic drivers including an aging budgets. population, reduced working age population, System dynamics (SD) modelling has been longer lifespans with more complicated employed by DoH since 2016 to support the comorbidities (e.g. learning disabilities with translation of various ideas into empirically- dementia), the dynamics of aging carers, and supported policy recommendations. For family pressures on young carers. Through example, in 2017 an expert panel working the modeling, these demographic trends are for DoH published Power to People — 16 seen to be coupled with social trends such proposals for ‘rebooting’ the system. These as changes to the structure of communities 16 proposals for reform were wide-ranging and to mediums of social engagement, as and ambitious, from ideas about valuing well as economic trends, such as how young the workforce to notions of forming a social people reside longer with their parents than concordant, a society-wide agreement on previously (boomerang generation), changes value and risk. System dynamics modelling to the distribution of wealth and cycles has helped to create an operational of indebtedness, consumer choices and understanding of how these proposals might the changing costs of different goods and link up synergistically. services, and the local elasticity of labour.

Design The modelling has shown how demographic, social, and economic trends place stress The suite of system dynamics models on the current system of social care by evaluate the dynamics between (i) prevalence, presenting it with population-level conditions (ii) workforce, and (iii) capital estate for which are different from the conditions in the system as a whole, while also delving which it originally formed. The modelling has into digital twin representations of certain also revealed how these demographics, while services such as domiciliary care services and stressing the old system, hold new energies learning disability services. The modelling and opportunities. effort also includes areas outside of health and social care which drive the dynamics, Moreover, through the SD modelling, some for example, economic inactivity, trends in of the stresses in the system have been technology, trends in home ownership, etc. revealed to be an endogenous product of the system structure itself, for example, In the case of learning disability services a health and social care system unable to — on the back of an institutional crises in help itself from delivering greater intensity various specialist hospitals revealed by of treatment (traumas etc.) at the cost of the news media — people with a learning preventative activities; a system which can’t disability have worked directly to shape keep pace with itself in terms of recruitment the models, setting new best practice for to fill vacancies (endogeneity of attrition rates 151 and churn). A system which, in its response to the leveraging of regulatory and legislative rising medical crises tightens thresholds and drivers. creates even more crises. A blameless ‘race to the bottom’ in procurement and contracting. Format of talk

Outcomes It is likely but not confirmed that policy colleagues would attend and speak to the At a policy level, system dynamics has extent that system dynamics modelling has identified leverage points for change, reasons informed the policy direction for the reform for policy resistance, and potential unintended of adult social care. It is entirely possible for consequences. This is informing a strategic the audience to run simulations via a model plan to be published for public consultation in embedded in a web application on their Autumn 2019. smartphones or computers during the talk.

The digital twin modelling of services — at The use of models to support the both an individual provider as well as at design and implementation of the a market level — has highlighted a need to pilot new procurement structures and EU Cohesion Policy techniques to overcome certain endogenous Monfort P., European Commission, Directorate- problems. To change what decisions are General for Regional and Urban Policy made within a system doesn’t necessarily require anything from procurement except The EU cohesion policy is the EU’s main enacting of past, established processes investment policy. Its objective is to foster around new parameters. To transform a a balanced development in Europe and system — to change the paradigm out of reduce disparities across EU regions and which the system arises — requires changes citizens by co-financing investment projects to procurement decision-rules themselves. in different sectors of the economy. In order A new procurement framework represents to reach these goals and address the diverse a new set of relationships for an industry, a development needs in the EU MS and regions, different pattern of dance emerging between more than €350 billion — almost a third of all actors in the system — administrators, the total EU budget — has been allocated professional staff, businesses, service for cohesion policy for the programming users, regulators, etc. In this instance, the period 2014–2020. This makes cohesion simulation model acts as a layer of insurance policy the second most important EU policy enabling government to take innovation risk. in budgetary terms behind the Common Additionally, there have been indications Agricultural Policy. that a whole-system analysis via a can help to make a compelling case for attracting Assessing impact of cohesion policy is innovation financing. In turn, and analysis of therefore critical and it is one of the the market has revealed leverage points for most evaluated EU policy. However, at policy and regulatory intervention at multiple the macroeconomic level this task is far levels. from trivial. Cohesion policy affects a wide range of macroeconomic variables, such Contribution as GDP, employment, productivity but also consumption, investment, the fiscal balance This work contributes to learning for the and the trade balance. Cohesion policy modelling community by giving a worked interventions have direct and indirect effects. example of embedding system dynamics For instance, projects in the field of transport modelling from primary service design directly boost demand in the short run (e.g. research with service users, to frontline public consumption) and they also improve operational understanding of service delivery, the factors productivity in the longer run (e.g. through to procurement, commissioning, and through an increase in the stock of public finally linked into policy decisions including infrastructure), which should have a positive

152 impact on GDP. At the same time, these Social Fund (ESF) and the Cohesion Fund (CF) interventions increase labour demand which are also geographically distributed according lead to higher wages and hence prices. This to needs that often reflect more the local than indirect inflationary effect negatively affects the national context. This implies that both GDP. the intensity of aid and the policy mix, i.e. the priorities supported by the interventions, In addition, the economic performance of the strongly vary from one region to another, MS and their regions is driven by a wide range even within the same Member State. of other (EU or national) policy actions as well by many different developments affecting the Second, the impact of the policy depends on world economy. These coincide with cohesion the economic and social environment in which policy interventions and the specific impact it is applied. Most Member States present of the latter can therefore not be identified wide regional variations on many aspects by simply looking at the data contained in the that can potentially affect the impact of the national accounts. In order to fully capture interventions and it is therefore relevant to the general equilibrium effects attributable account for regional idiosyncrasies when to the policy and properly assess its impact, analysing the impact of the policy. a counterfactual is needed which generally requires the use of macroeconomic models Third, some mechanisms which need to be at these instruments are among the few to taken into account when assessing the impact encompass all the channels through which of cohesion policy are more likely to play at the policy affects the EU regions' economies. a regional than at a national level. This is Models also offer the possibility to validly for instance the case with spatial spill-overs compare the outcome of different policy through which the programmes implemented scenario which makes them very valuable for in a given region also have an impact in other policy design. regions, especially those that are neighbours by geography. Interventions can also possibly The Commission has always relied on change the balance between agglomeration formal instruments to assess the relevance and dispersion forces, thereby affecting the and effectiveness of the policy, notably spatial distribution of people and economic macroeconomic models. HERMIN and EcoMod activity throughout EU territories. Even were used in the past by the Directorate though cohesion policy could possibly affect General for Regional and Urban Policy of the the spatial equilibrium at the level of the EU European Commission (DG REGIO) to assess Member States, its impact on the location the impact of cohesion policy programmes. of economic activity is more likely to be More recently, the Commission resorted to significant at the regional level. QUEST for conducting impact assessment of the policy. Other institutions have also The analysis of cohesion policy provided used models to analyse the effectiveness by national models can legitimately be of cohesion policy, like the IMF which used complemented by one conducted at the its multi-region GIMF (Global Integrated regional level and this is the main reason Monetary and Fiscal) model to assess the why the Commission decide to enlarge its impact of the EU cohesion transfer in the new analytical arsenal with RHOMOLO, a dynamic Member States from 2004 to 2015. and spatial computable general equilibrium model developed jointly by the DG REGIO and However, these models produce their results the Joint Research Centre of the European at the national level and this makes them Commission. The model simulates the impact blind on several important aspects of of policy interventions on the economies cohesion policy. First, for a large part the EU of 267 EU NUTS-2 regions, taking into cohesion policy is a ‘spatially targeted’ policy, account the spatial spill-overs that are most notably for the interventions supported by the relevant for the policy. The model heavily European Regional Development Fund (ERDF). borrows from New Economic Geography and Interventions supported by the European endogenizes the distribution of economic 153 activity across the regions concerned, emission reduction targets. Our findings show therefore allowing to capture the impact that the EU non-compliance mechanism, on of the policy on the spatial organization of average, does not explain the variation in EU economic activities in the EU. effectiveness, thereby rejecting hypothesis 1. By contrast, our analysis reveals that more The objective of this presentation is to explain ambitious positioning on the use of the Kyoto the value added of the RHOMOLO model for flexible mechanism and higher pre-Kyoto the analysis of the EU cohesion policy and to emission reduction targets reduce the EU’s provide different examples illustrating how effectiveness, hence supporting hypotheses 2 it has been used in the context of impact and 3. assessment or evaluation of the policy and how it complements the information retrieved Elevated use of flexible mechanisms and by national models. early national positioning to reduce emissions allow these countries to escape pressure Measuring and explaining the from the outside. In policy terms, this EU's Effect on national climate implies polarization: those countries that are performance willing to buy emission-reduction services abroad rather than undertake domestic Avrami L., Department of Political Science emission reduction policies, as well as those and Public Administration, National and who embark early on ambitious, unilateral Kapodistrian University of Athens emission reductions, escape the pressure Sprinz D. F., Detlef Sprinz, PIK - Potsdam of the EU climate regime. The EU’s non- Institute for Climate Impact Research & compliance mechanism appears to be an University of Potsdam. ineffective policy instrument. We should, To what extent has the European Union (EU) however, keep in mind that the reported had a benign or retarding effect on what its effect is an average effect across countries member states would have undertaken in and time. the absence of EU climate policies during The findings beg the question whether the 2008–2012? A measurement tool for the EU EU is always a benign force in environmental policy’s effect is developed. The EU’s policy policy. On average, the positive effectiveness effectiveness vis-à-vis its member states score demonstrates that the EU climate is explained by the EU’s non-compliance regime has a benign effect across the EU-14, mechanism, the degree of usage of the Kyoto but the EU cannot rely on the policy lever flexible mechanisms, and national pre-Kyoto under its sole control: the EU non-compliance emission reduction goals. Time-series cross- mechanism. sectional analyses show that the EU’s non- compliance mechanism has no effect, while Instead, our results demonstrate that the ex-ante plans for using Kyoto flexible countries that are willing to use their wealth mechanisms and/or the ambitious pre-Kyoto and wish to advance their climate policies emission reduction targets allow member largely remain unaffected by supranational states to escape constraints imposed by EU institutions. They 'buy' freedom. Conversely, climate policy. those who do not wish to lead, at least among our sample of countries within the We have pursued two aims. First, we derived supranational EU setting, are subject to a measurement method for the effect of the pressures from the EU. Countries remain EU climate regime on the EU-14 members 'interdependent, yet sovereign' (Putnam 1988, during the first compliance period of the p. 434) as the EU institutions can only exert Kyoto Protocol. Second, we explained the pressure on those countries that lack the variation therein by focusing on three core resources to buy greenhouse gas certificated political variables, namely the EU non- abroad and/or are laggards in terms of compliance procedure, the use of the Kyoto the depth of their national commitments. (flexible) mechanisms, and national pre-Kyoto As our results are broadly robust across 154 specifications and methods used, the fine • Prediction is where currently unknown point emerges that it may partially be in aspects are revealed by use of a model. member states’ own hands to define whether This is often claimed or implied by policy and to what extent the EU’s climate change models, but rarely realised. Predictive policy accelerates or retards their national ability should only be claimed when there environmental policy performance. is a track-record of successful prediction before what is predicted is known. Simply While the EU is formally a supranational fitting known data well is not prediction. institution with uniform powers vis-a-vis its member states, our results indicate that • Explanation is when known data is fitted the EU’s institutions systematically affects by the outcomes of a model as a result of member countries in unequal ways. Nudging the structures, processes and setting of the moderates and laggards to increase their that model. This supports an explanation, environmental ambitions might work, but even when the processes are too complex pushing ambitious members states might well to check mentally. Such explanations be beyond the capacity of the EU. inform our understanding of phenomena in an empirical manner. The importance of modelling purpose for policy • Illustration is when a few examples of a process is shown clearly by a model. Edmonds B., Centre for Policy Modelling, It does not show how general this is or Manchester Metropolitan University Business show any strong empirical connection School with what is being modelled. However, it can be informative as to how we think There are many sources of confusion between about things, e.g. as a counter example of policy actors and modellers, but one of the widely held assumptions. most common is when there is a lack of precision about the purpose of a simulation. • Analogy is when a model provides a In particular, this is a problem if the policy way of thinking about some phenomena, actor thinks a model will do one thing, but but no stronger empirical relationship is the analyst intends another. For example, established. This can give insights but can the policy actor might think that a model not be relied upon in any way. might predict the result of a policy, whilst the analysts are only claiming that they will • Theoretical Exploration is when a explore the consequences of some modelling model is used to extensively map out the assumptions. consequences of some assumptions in a general fashion, but it does not claim to Unfortunately, there is a lot of vagueness say anything about anything observed. from analysts concerning what their models can be justified for. A categorisation of some • Risk Analysis is when a model is used to model purposes is suggested into: prediction, reveal and assess processes that might explanation, illustration, analogy, theoretical result from an situation or policy, that exploration, risk analysis and social reflection otherwise might be missed. This does (Edmonds et al. 2019). Each of these is not predict, but does help policy actors described and illustrated with a policy anticipate what might go wrong in the example, along with the key risks that attend application of policy or indicate the each purpose — how each kind of project appropriate monitoring measures. might go wrong — and some measures to mitigate against those risks. The model • Social Reflection is when a model is purpose affects many aspects of model used to reflect the viewpoints of a set of development, checking and use. stakeholders. This can be used to animate the results of these viewpoints and aid

155 discussion and negotiation between Flaminio (2019) 'Different Modelling Purposes' them. It is limited by the knowledge of Journal of Artificial Societies and Social those stakeholders and is not necessarily Simulation 22 (3) 6 . doi:10.18564/jasss.3993

Some examples where confusion about Government Office for Science (2018) model purpose has undermined their policy Computational Modelling: Technological application is discussed. It is proposed that Futures. https://www.gov.uk/government/ this categorisation can be the beginning publications/computational-modelling- of a set of check-lists that could aid the blackett-review communication between policy actors and analysts, forming the basis of more well- Muffy Calder, Claire Craig, Dave Culley, founded policy modelling projects, and hence Richard de Cani, Christl A. Donnelly, inform policy in a more reliable and directed Rowan Douglas, Bruce Edmonds, Jonathon manner (Government Office for Science 2018, Gascoigne, Nigel Gilbert, Caroline Hargrove, Muffy et al. 2018). Derwen Hinds, David C. Lane, Dervilla Mitchell, Giles Pavey, David Robertson, Bridget References Rosewell, Spencer Sherwin, Mark Walport, & Alan Wilson (2018) Computational modelling Edmonds, Bruce, Le Page, Christophe, Bithell, for decision-making: where, why, what, Mike, Chattoe-Brown, Edmund, Grimm, Volker, who and how. Royal Society Open Science, Meyer, Ruth, Montañola-Sales, Cristina, doi:10.1098/rsos.172096. Ormerod, Paul, Root, Hilton and Squazzoni,

156 Parallel sessions 4

Modelling support for EU policy design: the EU’s Long Term Strategy for climate

Room 0.C 27 November 11:00 – 12:30

157 158 Developing a global context for providing the global context within which the regional 1.5°C scenarios EU pathways for the LTS were developed.

Keramidas K., Tchung-Ming S., Diaz-Vazquez A., Reaching the UNFCCC Paris Agreement Després J., Schmitz A., European Commission, goal to limit global warming to well below Joint Research Centre 2°C above pre-industrial levels and aim for 1.5°C requires action from all world The global partial equilibrium simulation countries and in all economic sectors. Global model POLES-JRC (Prospective Outlook net GHG emissions would have to drop to on Long-term Energy System) models has zero by around 2065 to limit temperature been used for more than two decades by increase to 1.5°C above pre-industrial levels the European Commission as an analytical (compared to around 2080 for the 2°C limit). tool for providing energy and greenhouse The analysis shows that this ambitious low- gas (GHG) emissions scenarios to inform carbon transition can be achieved with robust the climate and energy policy trade-offs for economic growth, implying small mitigation climate mitigation and sustainable energy costs. Results furthermore highlight that the development at both world and EU levels. combination of climate and air policies can contribute to improving air quality across The model's main features include a full the globe, thus enabling progress on the UN description of the energy sector, international Sustainable Development Goals for climate energy markets, accounting for energy and action, clean energy and good health. non-energy related emissions of GHGs as well as air pollutants, and a global coverage The global energy system and energy while keeping regional detail for 66 regions. consumption patterns would have to undergo POLES-JRC simulates technology dynamics a profound and immediate transformation to and follows the discrete choice modelling sustain unprecedented levels of global annual paradigm in the decision-making process. decarbonisation rates, between 6.1 and 9.0%/ It determines market shares (portfolio year over 2015–2050, compared to 1.9%/ approach) of competing options (technologies, year over 1990–2016. In particular, a strong fuels) based on their relative cost and climate objective of 1.5°C would require performance while also capturing non-cost massive mitigation efforts in the 2020–2040 elements like preferences or policy choices. decades.

On 28 November 2018, the Commission This transition is based on three main presented the EU’s strategic long-term levers, all of them requiring immediate strategy (LTS) ‘A clean planet for all’ (EC, and strong action: (i) a substantial, cross- 2018) for a climate-neutral economy by sectoral increase in energy efficiency by 2050 in line with the Paris Agreement. decoupling economic growth from energy Following the invitations by the European consumption; (ii) a strong shift of energy Parliament and the European Council, the carriers towards electricity; and (iii) a deep strategy covers nearly all EU policies and decarbonisation of the energy system. Key outlines a vision of the deep economic and mitigation options over 2015–2050 include societal transformations required, engaging the increase of the use of renewable energy all sectors of the economy and society, to sources , reduction of non-CO2 emissions, achieve the transition to zero GHG emissions improved energy efficiency, electrification in the EU by 2050. The quantitative backbone in final energy demand and land use. Total of the LTS is based on an integrated modeling energy-supply-related expenditure needs framework covering in detail all sectors of the would remain similar across scenarios, but economy across several research institutions. the composition changes more towards power The POLES-JRC analytical framework was sector investments. More expenditure would used to analyse global transition pathways be needed for investment in infrastructure, to a low GHG emissions economy, in order to especially in the power sector and for minimise irreversible climate damages, and demand-side energy efficient investments, 159 while operational costs would drop, 2070, with emphasis on the year 2050. The following the declining trend of fossil fuels PRIMES model, developed and operated consumption. by E3MLab, was employed to quantify the alternative pathways for the energy system References: and assess the impacts on energy indicators, investments and costs. The alternative Despres, J., Keramidas, K., Schmitz, A., pathways target energy and process emission Kitous, A., Schade, B., POLES-JRC model reductions between 84% and almost 100% documentation – 2018 update, EUR 29454 by 2050 (compared to 1990); including non- EN, Publications Office of the European Union, CO2 emissions (but excluding LULUCF) this Luxembourg, 2018, ISBN 978-92-79-97300- implies scenarios between -80% and -95% in 0, doi:10.2760/814959, JRC113757 2050. Keramidas, K., Tchung-Ming, S., Diaz-Vazquez, The pathways analysed share the same view A. R., Weitzel, M., Vandyck, T., Després, J., that energy efficiency and renewables are Schmitz, A., Rey Los Santos, L., Wojtowicz, the main pillars of the EU strategy towards K., Schade, B., Saveyn, B., Soria-Ramirez, A., decarbonisation. However, the dominating Global Energy and Climate Outlook 2018: energy carrier in the final energy demand Sectoral mitigation options towards a low- sectors varies across the pathways, with emissions economy – Global context to electricity, biofuels, hydrogen or synthetic the EU strategy for long-term greenhouse carbon-neutral hydrocarbons being gas emissions reduction, EUR 29462 EN, candidates. The conception of the pathways Publications Office of the European Union, emphasises sectorial integration, such as Luxembourg, 2018, ISBN 978-92-79-97462- power and mobility, power and heat, CO 5, doi:10.2760/67475, JRC113446. 2 capture and re-use in industry, biomass EC (2018). A Clean Planet for all A European cultivation and advanced biofuels in mobility, long-term strategic vision for a prosperous, power and production of hydrogen and modern, competitive and climate neutral synthetic hydrocarbons. The dominance of an economy Brussels, European Commission: energy carrier implies high industrial maturity 393. levels and learning-by-doing cost reductions for the technologies associated with its Modelling carbon neutral energy production and use. Data supporting the systems estimation of learning-by-doing potentials for each key technology have been derived after Zazias G., Evangelopoulou S., Fotiou T., a consultation with industry stakeholders. Kannavou M., Siskos P., Moysoglou Y., Statharas S., De Vita A., Capros P., E3MLab, National The PRIMES energy system model represents Technical University of Athens all demand and supply sectors in separate modules. The model formulates a typically The 2015 Paris Agreement has invited all nonlinear and intertemporal optimisation parties to submit, by 2020, mid-century model for each sector, as a typical decision strategies compatible with the goal of problem formulated structurally following containing the rise in average global microeconomic theory. The formulation temperature to well below 2°C above pre- by sector embeds engineering details of industrial levels and pursuing efforts to technologies and technical restrictions in limit it to 1.5oC. The European Commission the economic behavioural problem. The communicated in November 2018 a long- algorithms solve a mathematical optimisation term strategy towards near-zero greenhouse programme for some modules and a mixed gas emissions by 2050 and beyond ‘A clean nonlinear complementarity problem for planet for all’ (EC, 2018). The communication others. Thus, the model derives energy uses several alternative quantitative consumption, energy supply and investment projections of the EU energy system until

160 Figure 1. Global GHG emissions and global average temperature change (with median probability)

Note: The NDC scenario assumes that the global average rate of decarbonisation implied by the NDCs in 2020–2030 is maintained over 2030–2050. Source: POLES-JRC 2018; MAGICC online. depending on prices derived from other • An extension of the industrial energy modules. The pricing modules calculate demand module to include direct use supply costs by sector and determine prices hydrogen directly in high-temperature by sector of product use based on marginal applications, (e.g. iron and steel for direct costs and an allocation of fixed and capital reduction of iron ore), in furnaces and in costs depending on assumptions about the chemical industry as a fuel and as a market conditions. A market balancing routine feedstock to synthesise petrochemicals ensures demand and supply equilibrium in together with captured CO2. all markets simultaneously, and determines market clearing prices, after iterations with all • A new module represents the production modules. The model includes an endogenous of hydrogen, direct air capture of carbon mechanism for EU-ETS carbon prices and (DAC) and production of carbon neutral represents many policy instruments e.g. hydrocarbons, as well as distribution taxes, subsidies, measures removing barriers, of hydrogen either independently or infrastructure investment, technology injected in the gas distribution system. standards, emission or efficiency performing The module calculates feedstock inputs standards, policy targets and others. The to production of synthetic fuels, choice model supports impact assessment of of technologies (among electrolysers and policies by comparing a policy scenario to a production routes for synthetic fuels), baseline. The results provide projections of learning-by-doing, prices of the outputs the energy system, investment, prices, costs and distribution costs. In this way the and emissions until 2070 for each European module determines the prices of the country. synthetic fuels by consumption sector. Blending in the gas distribution is also Several model enhancements were introduced included in the model in order to inform the Long-term strategy in particular in relation to the treatment • The power sector model was extended of new fuels. The model enhancements to add the representation of chemical were introduced in both demand and storage of electricity and the modelling supply modules. An overview of the main of synchronous operation of power enhancements includes: generation, load, renewable resources, 161 storage of the chemical storage inputs organization of energy markets incorporating and the charging and discharging of the very high levels of (variable) RES and various storage systems. Investment competition of resources for energy storage in storage systems including choice of and production of neutral-neutral fuels. volume and the choice of technology mix is endogenous depending on costs of References storage, the prices of the storage inputs EC (2018). A Clean Planet for all A European and, the marginal costs of the power long-term strategic vision for a prosperous, systems that depend on the availability modern, competitive and climate neutral of renewable resources and the demand economy Brussels, European Commission: by end-users for hydrogen and synthetic 393. fuels. The power sector model solves the interconnected system of all European Costs and potentials for reducing countries simultaneously, and captures the sharing of balancing resources. non-CO2 gases

• A module takes care to balance the Höglund-Isaksson L., Winiwarter W., Purohit P., International Institute for Applied Systems capturing of carbon dioxide through Analysis (IIASA) several ways, competing against each other (DAC, biomass, combustion, On 28 November 2018, the Commission industrial processes) and the use of presented the EU’s strategic long-term carbon dioxide as a feedstock for strategy (LTS) ‘A clean planet for all’ (EC, synthetic fuels and its sequestration in 2018) for a climate-neutral economy by 2050 materials (e.g. feedstock for chemical in line with the Paris Agreement. Following substances) and in underground caverns. the invitations by the European Parliament and the European Council, the strategy • Finally, the calculation of overall costs covers nearly all EU policies and outlines a and various policy indicators has been vision of the deep economic and societal extended to include investment and costs transformations required, engaging all sectors related to hydrogen and the synthetic of the economy and society, to achieve the fuels. transition to zero GHG emissions in the EU by The quantitative analysis undertaken via the 2050. The quantitative backbone of the LTS is model confirmed that the decarbonisation based on an integrated modeling framework of the EU economy by mid-century is viable covering in detail all sectors of the economy both technically and economically, regardless across several research institutions. of the ambition level (limit temperature IIASA’s Air Quality and Greenhouse gases increase to well below 2°C or 1.5°C compared (AIR) program develops and uses the GAINS to pre-industrial levels); the pursuit of a model framework to produce detailed and carbon-neutral EU economy by 2050 is a internally consistent analyses of abatement plausible target. Some of the technologies potentials and associated costs for air included in the analysis are at low or medium pollutants and the non-CO2 greenhouse technology readiness levels (TRL), however gases methane (CH ), nitrous oxide (N O) and the number of the pathways assessed 4 2 fluorinated gases (HFCs, PFCs, SF6 and NF3). indicates that the energy system has a Starting in 2008, the program has every 2-3 plethora of innovative options available, none years provided scientific insights on non-CO2 of which can be considered irreplaceable emissions, future abatement potentials and besides some no-regret ones. The analysis costs to various climate policy processes should be complemented in the future with in the EU, including the 2020 climate and further exercises regarding both technological energy package (EC, 2009), the 2030 climate and strategic aspects around key issues, and energy framework (EC, 2013; EC, 2016), such as the origin of hydrogen, the origin of and EU’s recent LTS (EC, 2011; EC, 2018). carbon molecules for feedstock purposes, the 162 GAINS uses a linear model approach with emerging technologies with the latter being emissions estimated by multiplying activity technologies described in literature but with levels with average emission factors for limited current application, and attaching given technology structures applied in a different rates of development to the two given source sector and year. The strength categories. The GAINS model assumptions of the GAINS model lies in its high resolution for estimation of non-CO2 greenhouse gas in terms of emission source sectors and emissions in the EU and technical mitigation technical solutions to reduce emissions. The potentials and costs have been repeatedly GAINS model covers all source sectors for reviewed in bilateral consultations with non-CO2 greenhouse gases, i.e., agriculture, member state experts and in workshops energy production and consumption, industry, with sector and technology experts. All solid waste and wastewater management, assumptions are thoroughly documented in a and fluorinated gases from air conditioning comprehensive methodology report (Höglund- and refrigeration as well as other minor Isaksson et al., 2018) which is publicly sources. The model relies on externally available. Finally, the quality of the scientific produced activity projections for the energy basis of the work is demonstrated by the and agricultural sectors. For contributions to GAINS non-CO2 team’s frequent publications EU’s climate policy processes, macroeconomic in peer reviewed scientific journals (e.g. and energy activity scenarios are imported Höglund-Isaksson, 2012; 2017; Winiwarter from the PRIMES model and agricultural et al., 2018; Purohit and Höglund-Isaksson, activity scenarios from the CAPRI model. In 2017; Höglund-Isaksson et al., 2017). consistency with macroeconomic projections of the PRIMES model, projections for solid References waste generation and consumption of EC (2009). Decision No 406/2009/EC of the F-gases are produced within the GAINS European Parliament and of the Council of model. Feedbacks such as limitations on 23 April 2009 on the effort of Member States supply of organic waste and animal manure to reduce their greenhouse gas emissions as feedstock for production of renewable to meet the Community’s greenhouse gas energy are exchanged between the GAINS emission reduction commitments up to 2020 and PRIMES models. Source and country ('Effort Sharing Decision'). specific emission factors are derived within the model from a wide range of EC (2011). Impact assessment — A Roadmap factors identified as influencing emissions for moving to a competitive low carbon and provided enough country specific economy in 2050. Brussels, European information is available. This ensures a high Commission (EC): 133. level of consistency in emission factors across countries for similar physical and EC (2013). EU Energy, Transport and GHG technological circumstances. For each source Emissions Trends to 2050: Reference Scenario sector a number of emission abatement 2013. Brussels, European Commission options are identified, each with a unique Directorate-General for Energy, DG Climate relative removal efficiency and unit cost. Action and DG Mobility and Transport: 168. The latter is derived from a set of cost parameters that can be both technology EC (2016). EU Reference Scenario 2016 specific (e.g., fixed investment cost, non- Energy, transport and GHG emissions Trends labour operation and maintenance cost) and to 2050 Luxembourg, European Commission country- and year- specific (e.g., labour costs, (EC): 220. electricity and gas prices). Effects on removal EC (2018). A Clean Planet for all A European efficiency and abatement costs from future long-term strategic vision for a prosperous, technological development are captured by modern, competitive and climate neutral dividing abatement technologies into two economy Brussels, European Commission: broad categories; mature technologies and 393.

163 Höglund-Isaksson, L. (2012). 'Global partial equilibrium model of the forest and anthropogenic methane emissions 2005– agricultural sectors that is being developed at 2030: technical mitigation potentials and IIASA’s Ecosystem Services and Management costs'. Atmospheric Chemistry and Physics 12: Program. The model covers most important 9079–9096. crops (>27 crops, four vegetable oil types and related by-products), livestock, and Höglund Isaksson L (2017). 'Bottom-up forest products (different types of stem simulations of methane and ethane emissions wood, primary forestry residues from wood from global oil and gas systems 1980 to harvest, industrial residues) as well as 2012'. Environmental Research Letters 12 (2): biomass feedstocks globally and for the EU. e024007. DOI:10.1088/1748-9326/aa583e. The model is based on a bottom-up approach where the supply side of the model is built- Höglund-Isaksson L, Purohit P, Amann M, up from the bottom (land cover, land use, Bertok I, Rafaj P, Schöpp W, & Borken- management systems) to the top (production/ Kleefeld J (2017). 'Cost estimates of markets). The agricultural and forest the Kigali Amendment to phase-down productivity is represented at the spatially hydrofluorocarbons'. Environmental Science explicit level through link to biophysical & Policy 75: 138–147. doi:10.1016/j. models (crop models, forest growth models envsci.2017.05.006 etc.) for different production systems. The Höglund-Isaksson L, W. Winiwarter, P. Purohit, model computes market equilibrium for A. Gomez-Sanabria, P. Rafaj, W. Schöpp, J. agricultural and forest products by allocating land use among production activities to Borken-Kleefeld (2018). Non-CO2 greenhouse gas emissions in the EU-28 from 2005 to maximize the sum of producer and consumer 2070 with mitigation potentials and costs. surplus, subject to resource, technological, IIASA, 30 October 2018. https://ec.europa. demand, and policy constraints. Demand eu/clima/sites/clima/files/strategies/analysis/ and international trade are represented 58 models/docs/non_co2_methodology_report_ world regions calibrated to 2000. The model en.pdf allows for a full account of all agriculture and forestry GHG sources and is linked to Purohit P & Höglund Isaksson L (2017). the detailed G4M forest sector model. The 'Global emissions of fluorinated greenhouse Global Forest Model (G4M) estimates the gases 2005–2050 with abatement potentials impact of forestry activities (afforestation, and costs'. Atmospheric Chemistry and Physics deforestation, residue harvest, and forest 17: 2795–2816. doi:10.5194/acp-17-2795- management) on biomass and carbon stocks. 2017. By comparing the income of managed forest (difference of wood price and harvesting Winiwarter W, Höglund Isaksson L, Klimont costs, income by storing carbon in forests) Z, Schöpp W, & Amann M (2018). 'Technical with income by alternative land use on the opportunities to reduce global anthropogenic same place, a decision of afforestation or emissions of nitrous oxide'. Environmental deforestation is made. As G4M is spatially Research Letters 13 (1): 014011. explicit (currently on a 0.5° x 0.5° resolution), doi:10.1088/1748-9326/aa9ec9. different levels of deforestation pressure at the forest frontier can be handled. The The land use sector as key to land transitions in G4M are harmonized offset residual emissions by 2050 with GLOBIOM agriculture land demand and includes deforestation driven by the Frank S., Gusti M., Forsell N., Havlík P., expansion of agricultural areas. G4M International Institute for Applied Systems simulates forest management aimed at Analysis (IIASA) sustainable production of wood demanded The Global Biosphere Management Model by GLOBIOM on a regional scale. As outputs, (GLOBIOM) is a global recursive dynamic G4M produces estimates of forest area

164 change, carbon sequestration and emissions for construction and related increase in in forests, impacts of carbon incentives (e.g. the harvested wood carbon sink, as well as avoided deforestation) and supply of biomass lifestyle changes and reduction in animal for bio-energy and timber. The models are product consumption were also identified as regularly used EU’s assessment modelling important strategies that can contribute to framework to provide global and regional mitigation efforts. agricultural and forestry market outlooks, contribute the land use projections for the References EU Reference scenario (EC, 2013; EC, 2016) EC (2011). Impact assessment — A Roadmap and to inform EU climate policies on land use for moving to a competitive low carbon related issues (EC, 2011; EC, 2014; EC, 2016). economy in 2050. Brussels, European GLOBIOM/G4M provided the land use Commission (EC): 133. mitigation pathways and costs for emissions EC (2013). EU Energy, Transport and GHG from the LULUCF sector in the EU’s LTS on Emissions Trends to 2050: Reference Scenario climate change mitigation 'A Clean Planet 2013. Brussels, European Commission for All' (EC, 2018). The models were used to Directorate-General for Energy, DG Climate assess various options as to how the land Action and DG Mobility and Transport: 168. use sector could contribute to achieving the EU’s zero GHG target. Implications of different EC (2014). A policy framework for climate levels of bioenergy demand, feedstock mixes, and energy in the period from 2020 to 2030. and lifestyle changes were assessed across Communication from the Commission to 9 scenarios that differed in their level of the European Parliament, the Council, the ambition for emission reduction. European Economic and Social Committee and the Committee of the Regions. Brussels, Results showed that the LULUCF sector European Commission: 18. is able to maintain and even enhance the current carbon sink of around 320 MtCO2 EC (2016). EU Reference Scenario 2016 until 2050 if sensible management activities, Energy, transport and GHG emissions Trends mitigation policies and lifestyle changes to 2050 Luxembourg, European Commission: were realized despite the need to increase 220. biomass supply for energy production. Hence, the LULUCF sector is key to achieving the EC (2016). Impact Assessment — Proposal for climate neutrality target, as it allows to offset a regulation of the European Parliament and residual emissions from other sectors such of the Council on the inclusion of greenhouse as agriculture. The use of dedicated energy gas emissions and removals from land use, crops instead of stem forest was estimated land use change and forestry into the 2030 to limit negative impact on the forest sink climate and energy framework and amending and therefore helps to maintain the overall Regulation No 525/2013 of the European LULUCF sink across scenarios. Moreover, the Parliament and the Council on a mechanism models showed that even under moderate for monitoring and reporting greenhouse gas economic incentives in particular in the emissions and other information relevant to forest sector targeting mitigation options climate change Brussels, EC: 139. such as increased afforestation (20 MtCO2/ yr at 50 €/tCO2), reduced deforestation (10 EC (2018). A Clean Planet for all A European MtCO2/yr at 50 €/tCO2), and improved forest long-term strategic vision for a prosperous, management (40 MtCO2/yr at 50 €/tCO2) modern, competitive and climate neutral could substantially enhance the carbon sink. economy Brussels, European Commission: In addition, the increased use of biomass 393.

165 Socio-economic aspects of EU For the emission reduction scenarios, different decarbonisation levels of reduction (80% and reduction in line with a global 1.5C pathway) were assessed. Weitzel M., Vandyck T., Rey L., Wojtowicz In addition, it was analysed how the results K., Saveyn B., European Commission, Joint change under different assumptions for Research Centre climate policies in the rest of the world. Under 'fragmented action', it was assumed that only Large scale changes to the energy system the EU would implement climate policy, while and transformation towards a net-zero under 'global action' a coordinated effort emission society will lead to effects for towards a decarbonized global economy was the entire economy. The in-depth analysis followed. The table below shows resulting accompanying the LTS (EC, 2018) therefore impacts on EU GDP. takes a detailed look on how decarbonisation pathways affect macroeconomic indicators Changes to GDP relative to the baseline also like GDP, aggregate investment under altered depend on industry behaviour. GDP impacts investment requirements, international are slightly lower when industries use freely energy trade (e.g., EU net imports for energy), allocated permits to maximize the market as well as sectoral output and employment. share. When the opportunity cost of the free permits is added to the final price ('profit JRC-GEM-E3 is a computable general maximization'), the output of the economy is equilibrium (CGE) model that captures slightly lower. JRC-GEM-E3 was also used to economy wide effects and was used to assess investigate implications for different sectors a range of macro-economic issues stemming of the economy, as the decarbonisation from the analysis of the energy transition. pathways will affect output of various The model captures interactions between sectors differently. Changes in output are different actors (industry, households, also reflected in shifts in the employment government) and captures bilateral trade structure. Different scenarios were run with between 40 regions. The economy is the model with regard to assumptions on the represented by 31 sectors, focussing on labour market. This includes a labour market energy producing and energy intensive in long-term equilibrium where employment industries. is unresponsive to changes in the wage rate In JRC-GEM-E3, both the macro-economic and an alternative formulation where lower baseline as well as the decarbonisation wages can increase employment. Wages can scenarios were constructed using the results be lowered when revenues from auctioning of the PRIMES energy system model for the carbon permits are used to lower labour baseline energy scenario for Europe. For taxes. the rest of the world, the macro-economic References baseline assumes implementation of the Intended Nationally Determined Contributions EC (2018). A Clean Planet for all A European (INDCs) as reported to the UNFCCC and as long-term strategic vision for a prosperous, modelled by POLES-JRC. modern, competitive and climate neutral economy Brussels, European Commission: 393.

166 Table: Impacts on GDP in different scenarios

GDP vs. Baseline, 2050 Fragmented action Global action JRC-GEM-E3 80% 1.5°C 80% 1.5° C reduction reduction

Profit maximisation Perfect labour market -0.13% -0.63% -0.28% -1.30% Lump-sum transfers Market share maximisation Perfect labour market -0.10% -0.59% -0.25% -1.26% Lump-sum transfers

Market share maximisation Imperfect labour market 0.05% -0.29% -0.18% -1.09% revenue recycling

167 168 Parallel sessions 5

Model quality and transparency 2

Room 0.A 27 November 13:30 – 15:00

169 170 Map or compass? Improving model by equivalent removals from natural carbon contribution to long-term climate sinks). strategy. Insights from 4 country Given the complexity of the problem at hand, studies and given the time horizons involved, it is not surprising that countries often rely on Lecocq F., AgroParistech-CIRED techno-economic models as a support tool for Nadaï A., Cassen C., CNRS-CIRED the preparation of their national low-carbon Combet E., Ademe, French Environment and Energy Management Agency strategies (NDCs and mid-century strategies). Models help provide numerically consistent Parties to the Paris Agreement (PA) are pictures of the future, with a degree of asked to prepare and submit by 2020 complexity superior to what the human mind the next round of so-called nationally can achieve, and with a degree of consistency determined contribution (NDC), which define that qualitative forms of prospective cannot their mitigation (and adaptation) objectives reach. typically for the horizon 2030. Parties are also requested to prepare and submit by If using models in support of policy-making 2020 mid-century low greenhouse gas is by no means new, complexity and timing emission development strategies. create a specific challenge for model use in the case of climate mitigation. Specifically, as Preparing these documents constitutes noted above, models must capture an ever a challenge for two reasons. First, the PA expanding range of mitigation options across goal of limiting temperature increase well sectors and across all components of supply below 2°C relative to preindustrial level and demand. This creates a major tension by the end of the century cannot be met between the degree of granularity required by marginal changes in the energy supply for the model to dialogue with stakeholders sector. Major changes in production and in each sector (energy, transport, housing, but consumption patterns across all economic also agriculture, forestry, macroeconomics, sectors is required, at very rapid pace. Finding etc.), and the risk for models to become too technologically, economically, socially and cumbersome or impossible to control. environmentally plausible pathways to successfully complete these transitions thus The objective of this paper is to explore requires to assemble and articulate a very how models currently meet this challenge, large range of expertise that goes beyond and to draw insights on how they might be any individual sector. improved to this regard. The question touches not only the technical structure of individual Second, 2030, let alone 2050, is an unusually models, but also the way different models (for long time horizon for policy-making, where example, different sectoral models) exchange the 'long term' is typically no farther than with each other, and how models and a few years ahead. Given uncertainties modelers interact with stakeholders and with on future parameters (e.g., availability of policymakers. The policy-making process of mitigation technologies, economic conditions, which models and modelers are part is thus citizens’ preferences, etc.) and lack of as important to our study as the technical legitimacy (and possibility) for current policy- structure of these tools. makers to coerce future policy-makers into following their plans make these documents To meet this objective, we rely on four exploratory rather than prescriptive. Yet national case studies in which models have mid-century strategies tend to start from a been used to support the development of low mitigation goal at horizon 2050, such as the carbon strategies: France (Second National factor four (a division by four of emissions Low Carbon Strategy (SNBC), 2018), Sweden relative to, e.g., 2010) or carbon neutrality (Climate Act, 2017), Brazil (NDC, 2015), and (net remaining GHG emissions compensated the USA (Mid-century Strategy for Deep

171 Decarbonization, 2016). These case studies a central, coordinating role, mediating are all based on a variety of models to assess almost all relations with modelers and with national trajectories (e.g. US, Brazil and stakeholders. And in Sweden, the modeling France) but they differ as to their horizon and evaluation is deeply rooted in the political as the ways of framing these trajectories. process (parliament). For example, the US scenarios are technology / innovation centered (e.g., carbon Ultimately, we find that the type of insights sequestration); the Brazil exercise addresses models are tasked to provide to the policy the distributive impact of a set of climate process is critical, especially for the mid- change policies (carbon tax, command and century strategies. There is a tension between control, early or delayed action); and the the objective of exploring a wide range of French SNBC is geared towards 2050 carbon plausible futures (the 'map' function of neutrality, an unprecedented horizon and models) and helping to set the right course, objective for the country, which challenged especially when short-to medium-term methods and ways of interacting with and decisions must be taken (the 'compass' around models. function of models).

Our material is based on a qualitative Modelling climate-energy study combining document analysis and linkages in national energy and semi-directive interviews (with modelers, climate plans: achievements and independent consultants, public agencies and administrations) undertaken in 2019 (April challenges to July). The comparison of the case studies Boromisa A., Institute for Development and (chronology, organization, process, outcomes) International Relations, Croatia allows highlighting a set of valuable lessons for the modeling of NDC and low carbon Implementing Paris Agreement requires a strategies on different scales, horizon, fast and radical transformation of politics, in different settings and with different economy and society. Worldwide emissions objectives. of greenhouse gases need to fall to zero by 2100. Pathways limiting global warming to We find that the way these models are used 1.5 °C show that CO2 emissions need to be is country specific and dependent upon a reduced to net zero globally around 2050 range of factors including the habits and (IPCC Special Report on the impacts of global practices to use these models by public warming of 1.5°C above pre-industrial levels administrations, the existence of academic and related global greenhouse gas emission or private institutions developing such tools, pathways, 2018). dedicated fora gathering stakeholders and modelers etc. Analyzing more in depth the The European Commission calls for climate functioning of these ecosystems allows for neutral Europe by 2050. Implementing such not only capturing the adequacy between the reforms requires energy transition, since use models and the institutional context, but energy sector is globally responsible for more also envisaging improvements in those fields. than half of emissions. There is still vivid debate as to which extent energy transition We find that each country also adopted is justified and feasible. The political will to specific modalities of interactions between make the necessary decisions depends partly modelers and stakeholders in the definition of on improving the analysis and estimates scenarios and their evaluation. For example, of the economics of decarbonisation Brazil has set a rich participatory process compared to reference scenarios. Then the with multiple direct interactions between consequences of unmanaged climate change modelers and stakeholders (representative can be weighed much more transparently from business sectors, NGOs, experts etc.). against the investments and innovations In France the administration set itself with necessary for decarbonisation.

172 According to The Fifth Assessment Report policies and measures to meet the objectives of the Intergovernmental Panel on Climate of energy union and a general assessment Change (IPCC), published in 2013 and 2014 of the planned policies and measures on and Stern (2015) current models tend to competitiveness. underestimate the potential impacts of unmanaged climate change and the benefits The aim of the paper is to map draft NCEPs of a transition to low-carbon growth, both. based on analytical approach and determine Therefore, models that give a more accurate to which extent level of ambitious defined in picture and support formulation of more NCEP depends on the modelling approach. ambitious policies are necessary. This could serve as a basis for outlining common elements that should be considered This paper analyses and maps models used in revisions. in EU member states to formulate policies in integrated national energy and climate The starting hypothesis is that level of plans. Integrated national energy and climate ambitious of NCEP depends on the analytical plans define a pathway to 2030 to ensure base used for policy development. meeting 2030 targets and a transition to In particular, that: a climate neutral economy that is fair and cost-effective for all. In line with Energy Union • First, member states that rely primarily on Governance Regulation(The Regulation (EU) no-regret policy measures and traditional 2018/1999 of the European Parliament and energy planning models (energy supply- of the Council of 11 December 2018 on the demand models, emission reduction Governance of the Energy Union and Climate models, renewable energy models, Action), NCEPs cover the five interlinked forecasting models) define less ambitious dimensions of the energy union: goals and targets.

• Security, solidarity and trust • Second, usage of more sophisticated • A fully integrated internal energy market models (e.g. integrated assessment models, IAMs or third wave models • Energy efficiency — dynamic stochastic computable • Climate action, decarbonising the equilibrium models, DSCEM or agent economy based models, ABM), including models that consider disruptive options (relating • Research, innovation and competitiveness to change of energy mix and the way — supporting breakthroughs in low- energy is used and distributed) enables carbon and clean energy technologies setting more ambitious targets, primarily for the period 2021 to 2030 (and every in the countries with higher innovation subsequent ten year period). index.

Drafts NCEPs had to be submitted by end of Structure of the paper is as follows: 2018. The European Commission published assessment of NCEPs and provided specific First, Energy Union Governance Regulation recommendations addressed to each member and role of NCEPs are briefly presented. Focus state in mid June 2019. The final version of is on the established targets and impact NCEPs should be submitted by the end of assessment included in the draft NCEPs. 2019. Based on available information, data sources used, assumptions and modelling approaches Each NECEP has to include description of the in the EU are categorised. planned policies and measures necessary to achieve goals of the Energy Union, a general This is followed by review of the Commission’s overview of the investment needed to meet assessment and recommendations, focusing the objectives, targets and contributions; an again on the level of ambition and feasibility assessment of the impacts of the planned of achieving goals. The main factors and 173 how member states have weighted them Analysis of current climate and in defining objectives and policies are energy policies: are countries on identified. It is also examined to which extent policy measures and activities are track to meet their NDC targets? coherent with expected impacts, and which den Elzen M., van Soest H., Roelfsema M., PBL impact (financial, societal, impacts on Netherlands Environmental Assessment Agency competitiveness) are considered as most Kuramochi T., Fekete H., Höhne N., NewClimate relevant. This is accompanied with review Institute of interlinkages that are considered (e.g. Forsell N., International Institute for Applied links among affordability, reliability and Systems Analysis (IIASA) sustainability of energy system). Short Summary. Under the Paris Agreement, This analysis, together with results of first countries committed to a variety of climate section, enables mapping NECEPs and testing actions, including post-2020 greenhouse starting hypotheses. gas (GHG) emissions reduction targets. This study [1] compares projected GHG emissions Third section discusses common issues, in in 25 major emitting countries/regions up challenges and best practices identified to 2030, taking into account the emission by mapping relating to elements that trajectories based on current policies and increase quality of analytical basis for policy the implementation of nationally determined development. These include data sources and contributions (NDCs). This study concludes modelling applied. Special attention is given that 16 out of the 25 countries and regions to assumptions related to costing transition, analysed are not on track to achieve the NDC i.e. future evolution of costs and performance targets they have set for themselves. of various technologies, discount rates, comparison with reference scenarios (cost Not all countries on track to achieve of unmitigated climate change and business targets set in NDCs and usual scenario, costs of reference The degree to which the 25 major emitting scenario related to economic growth, cost of countries are likely to achieve their NDC locked in infrastructure and ways to tackle targets under current policies was found disruptive options and costing uncertainties to vary. Of those considered in this study, (the feedback loops in the innovation process China, Colombia, EU28, India, the Russian that interact across the economy, prompting Federation, Saudi Arabia, Turkey, and the institutional and behavioural change, possible Ukraine are likely or roughly on track to discoveries and economies of scale). achieve or even overachieve their self- Finally, based on identified elements and determined unconditional 2025/2030 targets approaches to model interrelations between with currently implemented policies. For the five dimensions of the energy union, Mexico, the achievement of 2030 targets conclusions and recommendations are was found to be uncertain with implemented formulated. policies. The other 16 countries (Argentina, Australia, Brazil, Canada, Chile, Democratic These relate to: Republic of the Congo, Ethiopia, Indonesia, Japan, Kazakhstan, Morocco, Republic of • Models used and identified best practices Korea, South Africa, Thailand, the Philippines, • Interdependence of level of ambition and and the United States), require additional used analytical basis measures to achieve their 2025/2030 • Common issues and challenges related to targets. It should be noted that a country harmonizing approach in future likely to meet its targets not necessarily is undertaking more stringent action on • Range of expected impacts and mitigation than a country that is not on track, interdependencies. as this depends on the ambition level of the nationally determined target, and because

174 countries have different policy-making the Paris Agreement. Previous studies have approaches [2]. shown that, even with full implementation of all the plans countries submitted in this Progress on reducing greenhouse gas agreement, global temperature would rise by emissions also varies 2.6 °C to 3.1 °C, by the end of the century [3].

Currently implemented policies are projected Methodology to influence GHG emissions, but do not prevent emissions from increasing up to NewClimate Institute, IIASA and PBL have 2030 (above 2010 levels). This is the case, estimated the impact of the most effective not only in developing countries (Argentina, current policies on future GHG emissions. Brazil, China, DRC, Ethiopia, India, Indonesia, The calculations by NewClimate Institute Kazakhstan, Morocco, the Philippines, Russia, are largely based on its analyses for, and Saudi Arabia, South Africa, and Thailand) but informed by, the Climate Action Tracker also in OECD countries (Chile, Mexico, Republic project [4], and used existing scenarios from of Korea, and Turkey) up to 2030, compared national and international studies (e.g. IEA's to 2010 levels. Emissions in the remaining World Energy Outlook) as well as their own seven countries are projected to remain calculations of the impact of individual stable, approximately at current levels, or policies in different subsectors. PBL calculated to decrease further, under current policies. the impact of individual policies in different Significant overachievement of NDCs: India, subsectors using the IMAGE integrated Russian Federation, Turkey and Ukraine. For assessment model. The starting point for these countries, the current policies scenario the calculations of the impact of climate projections for 2030 are more than 10% policies is the latest SSP2 (no climate policy) lower than the unconditional NDC target baseline. Current climate and energy policies levels. These countries could revise their in G20 countries, as identified in the CD- NDCs to more ambitious ones under the Paris LINKS project, were added to that baseline by Agreement’s ratcheting mechanism. changing model input parameters to achieve the policy impacts [5]. Both NewClimate and Changes since the Paris Agreement was PBL scenario calculations were supplemented adopted with those on land-use and agricultural policies using IIASA's global land-use model The study also assessed how countries’ GLOBIOM. current policies scenario projections have changed since 2015, when the Paris Future research Agreement was adopted. The comparison with our 2015 study, which covered thirteen The number of studies analysing whether countries, shows only six countries (Australia, countries are on track to achieve their NDC Canada, China, EU, Turkey and the United targets is still limited in literature, whereas States) show lower emissions projections from a policy perspective it becomes for 2030; for the remaining seven countries more and more important to look at the (Brazil, India, Indonesia, Japan, Mexico, implementation of the NDCs [2]. Therefore, Republic of Korea, Russia) the projections it is encouraging that our work is used for were either similar or even higher. the implementation and development of NDC and current policies scenarios for many Still more effort needed to stay well below global and national models in the CD-LINKS 2 °C project [6, 7] and the upcoming European H2020 ENGAGE projects. This has been Even if all the countries’ targets would be done by the development of the modelling fully met, the combined mitigation impact protocol for the implementation of current would fall far short of what is required to policies in all models. That protocol was limit global warming to well below 2 °C and based on the work mentioned above, and an possibly 1.5 °C — the climate targets set in iteration with national experts was added, to 175 ensure inclusion of the most recent and most Social-environmental systems (SES) are important policies (in terms of emissions). characterized by a tight coupling of human and environmental dynamics (Berkes and References Folke, 1998; Folke et al., 2010; Schulze et al., 2017). To support sustainable management 1. Kuramochi, T., et al., Greenhouse gas of these systems, it is crucial to understand mitigation scenarios for major emitting both their social and environmental aspects countries: Analysis of current climate (Carpenter et al., 2009; Ostrom, 2009; Chapin policies and mitigation commitments et al., 2010). Modelling is often proposed as 2018 update, NewClimate Institute an effective tool to address these interlinked (Cologne, Germany), PBL (The Hague, problems and provide sustainable solutions the Netherlands), IIASA (Austria). 2018, to them as it allows disentangling causes https://www.pbl.nl/en. and effects of human behaviour and 2. den Elzen, M., et al., 'Are the G20 environmental constraints. economies making enough progress to meet their NDC targets?' Energy Policy, Whereas social-environmental modelling has 2019. 126: p. 238–250. made many contributions in the scientific realm to provide answers to environmental 3. Rogelj, J., et al., 'Paris Agreement climate issues such as sustainable management of proposals need a boost to keep warming natural resources and ecosystem services well below 2 °C'. Nature, 2016. 534 (Karagiannakos, 1996; Schlüter et al., 2009), (7609): p. 631–639. land-use/land-cover change (Parker et al., 4. CAT, Climate Action Tracker, 2018, 2003; Schulze et al., 2016), or biodiversity Countries (updated 29 August, 2018). conservation (Myers et al., 2000), the impact https://climateactiontracker.org/countries of these studies in real world decision-making [Accessed 29 August 2018]. and policy support has so far been very limited. 5. Roelfsema, M., et al., 'Reducing global GHG emissions by replicating successful In contrast, models from other fields such sector examples: the ‘good practice as transportation planning, epidemiology, policies’ scenario'. Climate Policy, 2018. or pesticide risk assessment have been 18(9): p. 1103–1113. integrated into policy-making processes and 6. McCollum, D.L., et al., 'Energy investment have successfully been applied to real-world needs for fulfilling the Paris Agreement problems. and achieving the Sustainable Development Goals'. Nature Energy, 2018. In this paper, we identify reasons for the 3(7): p. 589–599. limited impact of SES models and discuss steps that need to be taken to increase this 7. Roelfsema, M., et al., 'Takng stock of impact. By learning from fields where models climate policies: evaluation of national have already been successfully involved policies in the context of the Paris in policy making, we aim to facilitate the Agreement climate goals'. Nature Climate integration of social-environmental modelling Change (submitted), 2019. into practice and to increase its relevance for real-world applications. By impact, we How to make social-environmental mean that models have contributed to a modelling more useful to support policy decision in the widest sense, which can policy? range from stimulating a policy process (e.g. raising awareness for so far neglected issues), Dreßler G., Kreuer D., Will M., Thulke H. H., influencing discussions around a decision (e.g. Müller B., UFZ, Department of Ecological laying out certain options or scenarios), up Modelling, Helmholtz Centre for Environmental to policy decisions being directly based on Research - UFZ model results. Impact does not state whether

176 the outcome of the decision was positive or In addition, data collection of social data negative from a given perspective. is associated with more effort and more limitations. However, in specific fields such as To address this question, we investigate coastal fishery in Australia, where a long- seven good-practice examples in detail. We standing relationship between policy and identify key aspects that enable or impede science already exists, the problem of data the application of such models. The selected availability is smaller. Another aspect that studies range from purely ecological problems impedes the use of models in policy making is such as the management of fish populations related to the short-term obligations of policy in rivers (Railsback et al., 2005) or control makers, where long-term perspectives are of of animal diseases (Thulke et al., 2018) to minor importance. coupled SES research such as sustainable fisheries in Australia (Fulton et al., 2014) To summarize these results, we synthesize a and water management in Jordan (Klassert list of key principles of success or failure that et al., 2015). All examples have in common social-ecological modellers should consider that they did not only deliver scientifically when they aim for real-world impact of their innovative insights but were also used to models. In this list, we differentiate between guide actual policy decisions. what lies in the scope of action of modellers, of decision makers, or rather beyond. We For the evaluation of these examples we derive (a) main characteristics needed in use a list of criteria that is inspired by a models to be suitable for policy analysis, framework developed in Gray et al. (2018). (b) necessary steps for interaction between We follow their approach and assess the policy makers and modellers, and (c) ways models with respect to aspects concerning to overcome the drawbacks and challenges the purpose of the modelling endeavour, the associated with the application of models in process of exchange between modellers and policy making. policy makers, details on this partnership, and the product that emerges from this exchange, References i.e. the range and type of application of the model outcome in practice. We interviewed Berkes, F., Folke, C. (Eds.), 1998. Linking the authors of those studies about their social and ecological systems: management experience in the entire modelling process. To practices and social mechanisms for building interpret these achievements, we asked them resilience. Cambridge University Press, to self-assess their studies and evaluate their Cambridge, UK. models against the questions in the criteria Carpenter, S.R., Mooney, H.A., Agard, J., list, without us adding any additional rating. Capistrano, D., DeFries, R.S., Dìaz, S., Dietz, T., Duraiappah, A.K., Oteng-Yeboah, A., Pereira, While the good practice examples that we H.M., Perrings, C., Reid, W.V., Sarukhan, J., examined display a wide range of individual Scholes, R.J., Whyte, A., 2009. 'Science characteristics, several common points for managing ecosystem services: Beyond concerning transdisciplinarity, data availability the Millennium Ecosystem Assessment'. and urgency of the modelling process stand Proceedings of the National Academy of out. Soft factors such as prior experiences Sciences 106, 1305–1312. of the policy makers with modelling, interpersonal trust and communication Chapin, F.S., Carpenter, S.R., Kofinas, G.P., usually determine the chances of success Folke, C., Abel, N., Clark, W.C., Olsson, or failure of a project. Furthermore, we P., Smith, D.M.S., Walker, B., Young, O.R., observe that the confidence of policy makers Berkes, F., Biggs, R., Grove, J.M., Naylor, R.L., in purely ecological models is higher than in Pinkerton, E., Steffen, W., Swanson, F.J., social-environmental ones, where modelling 2010. 'Ecosystem stewardship: sustainability human behaviour is still strongly questioned. strategies for a rapidly changing planet'. Trends in Ecology & Evolution 25, 241–249.

177 Folke, C., Carpenter, S.R., Walker, B., Scheffer, services: a modeling approach'. Ecological M., Chapin, T., Rockström, J., 2010. 'Resilience Research 24, 491–503. Thinking: Integrating Resilience, Adaptability and Transformability'. Ecology and Society 15. Schulze, J., Frank, K., Priess, J.A., Meyer, M.A., 2016. 'Assessing Regional-Scale Impacts Fulton, E.A., Smith, A.D.M., Smith, D.C., of Short Rotation Coppices on Ecosystem Johnson, P., 2014. 'An Integrated Approach Services by Modeling Land-Use Decision'. Is Needed for Ecosystem Based Fisheries PLOS ONE 11. Management: Insights from Ecosystem-Level Management Strategy Evaluation'. PLOS ONE Schulze, J., Müller, B., Groeneveld, J., 9. Grimm, V., 2017. A'gent-Based Modelling of Gray, S., Voinov, A., Paolisso, M., Jordan, R., Social-Ecological Systems: Achievements, BenDor, T., Bommel, P., Glynn, P., Hedelin, B., Challenges, and a Way Forward'. Journal of Hubacek, K., Introne, J., Kolagani, N., Laursen, Artificial Societies and Social Simulation 20, 8. B., Prell, C., Schmitt Olabisi, L., Singer, A., Thulke, H.-H., Lange, M., Tratalos, J.A., Clegg, Sterling, E., Zellner, M., 2018. 'Purpose, T.A., McGrath, G., O’Grady, L., O’Sullivan, P., processes, partnerships, and products: four Ps Doherty, M.L., Graham, D.A., More, S.J., 2018. to advance participatory socio-environmental 'Eradicating BVD, reviewing Irish programme modeling'. Ecological Applications 28, 46–61. data and model predictions to support Karagiannakos, A., 1996. 'Total Allowable prospective decision making'. Preventive Catch (TAC) and quota management system Veterinary Medicine 150, 151–161. in the European Union' Marine Policy 20, 235–248. Quantitative pest risk assessment Klassert, C., Sigel, K., Gawel, E., Klauer, — a modelling methodology B., 2015. 'Modeling Residential Water supporting EU plant health policy Consumption in Amman: The Role of Bragard C., UCLouvain, Earth & Life Institute Intermittency, Storage, and Pricing for Piped Gardi C., Kozelska S., Maiorano A., Mosbach- and Tanker Water'. Water 7, 3643–3670. Schulz O., Pautasso M., Stancanelli Myers, N., Mittermeier, R.A., Mittermeier, G.,Tramontini S., Vos S., European Food Safety C.G., Fonseca, G.A.B. da, Kent, J., 2000. Authority 'Biodiversity hotspots for conservation MacLeod A., Department for Environment, Food priorities'. Nature 403, 853–858. and Rural Affairs, UK Parnell S., School of Environment and Life Ostrom, E., 2009. 'A General Framework for Sciences, University of Salford Vicent Civera A., Centro de Protección Vegetal Analyzing Sustainability of Social-Ecological y Biotecnología, Instituto Valenciano de Systems'. Science 325, 419–422. Investigaciones Agrarias (IVIA) van der Werf W., Centre for Crop Systems Parker, D.C., Manson, S.M., Janssen, M.A., Analysis, Wageningen University Hoffmann, M.J., Deadman, P., 2003. 'Multi- Agent Systems for the Simulation of Land- To support the recent revision of the EU Use and Land-Cover Change: A Review'. plant health legislation, the European Annals of The Association of American Commission requested the European Food Geographers 93, 314–337. Safety Authority (EFSA) to evaluate the status of several plant pests listed in the Annexes Railsback, S.F., Harvey, B.C., Hayse, J.W., of Council Directive 2000/29/EC. Given the LaGory, K.E., 2005. 'Tests of Theory for Diel magnitude of the task, a two-phase approach Variation in Salmonid Feeding Activity and was developed by the EFSA Plant Health Habitat Use'. Ecology 86, 947–959. Panel (EFSA PLH Panel et al., 2018).

Schlüter, M., Leslie, H., Levin, S., 2009. First, a qualitative pest categorisation 'Managing water-use trade-offs in a semi- determines whether the pest fulfils the arid river delta to sustain multiple ecosystem 178 criteria of a quarantine pest or those of a vaccinii. These case studies had sufficient regulated non-quarantine pest. Following the detail to enable the quantification of key outcome of the first phase, for selected pests, processes. However, the assessments were a quantitative risk assessment is performed. kept as parsimonious as possible to ensure Among other elements, this may include transparency and parameterisation with the the evaluation of (i) risk reduction options available data. The modelling framework (RROs) and (ii) the effectiveness of current is thus adaptable to make the assessment EU phytosanitary requirements. In this way, appropriate to the level of detail required. modelling efforts are focused to those pests for which a full risk assessment was felt to be Examples of a conceptual and formal entry needed by risk managers. pathway model are provided in the guidance to illustrate the methodology. The pest Quantitative models are likely to better risk assessment is based on scenarios. A inform pest risk management decisions, as conceptual model describes the system to these are often based on comparisons of be assessed, e.g. an entry pathway leading scenarios, e.g. with or without selected RROs to pest establishment, spread and ultimately in place. Combining uncertainties deriving impact. For each step and scenario of the from the different estimations becomes model, relevant data are crucial, but often straightforward. The approach also makes it not at hand. Models therefore require expert possible to objectively identify which variables knowledge to estimate plausible distributions are more responsible for the uncertainty of for several parameters. These distributions the final outcome. This can inform research reflect the uncertainty of experts in a rigorous needs to reduce uncertainties of future way. It is important to communicate clearly assessments. (e.g. without excessive use of decimals, thus not reflecting the intrinsic uncertainty The new methodology includes a two-tiered of model structure and inputs) and in a approach: depending on risk managers’ consistent manner (e.g. by systematically needs and resources availability, the referring to the variation around mean assessment can be conducted directly, using results) the quantitative results from the weight of evidence and quantitative expert models. Guidance on both verbal and visual judgement (first tier) or applying quantitative communication of model results is provided. models in the various sub-steps of the pest risk assessment(second tier), e.g. pest The results of the pest entry model are prevalence at the origin, trade volumes, reported as the uncertainty distribution climate suitability in the EU for the pest, of the estimated number of founder pest pest dispersal potential, and impacts in the populations. This assessment is made endangered area. separately for each entry (sub)pathway and then combined. Establishment is described We focus here on the second tier, i.e. as the uncertainty distribution of the likely the development of quantitative models number of founder populations establishing for assessing pest entry, establishment, due to entry (e.g. expressed as suitable grid spread and impact. The approach has been cells in a map), in view of climatic and other tested on ten taxa, covering a variety of environmental factors. Spread is depicted as pathogens and invertebrate pests: the an uncertainty distribution for the increase in nematodes Ditylenchus destructor and the spatial occupancy of the pest within the Radopholus similis, the mite Eotetranychus risk assessment area. The consequences of lewisi, the phytoplasma Flavescence pest introduction and spread are conveyed dorée, the bacterium Xylella fastidiosa, as estimated uncertainty distributions of the insect Spodoptera frugiperda, and the changes to crop output, yield or quality fungi Atropellis spp., Ceratocystis platani, under different risk management scenarios. Cryphonectria parasitica and Diaporthe Environmental impacts can also be gauged

179 in terms of estimated changes in ecosystem in the EU at NUTS2 level and carried out services provisioning and biodiversity levels. expert knowledge elicitations to obtain quantitative assessments and uncertainty This modelling framework makes it possible analyses for yield and quality losses, spread to clearly respond to the questions that rate and time to detection. The elicitations motivated the pest risk assessment and were conducted under two scenarios using favours a closer dialogue with risk managers, assumptions common to all pests, making as they are involved in the definition of the the results comparable among the pests scenarios to be developed so as to obtain fit- and translatable by JRC to monetary values. for-purpose outputs. JRC then developed a composite indicator, the Impact Indicator for Priority Pests (I2P2), In addition, the outputs of quantitative in order to rank the 28 species according to pest risk assessments are essential for potential economic, social and environmental designing pest surveillance strategies both impact in the EU and fed it with the values for detection and delimiting surveys. For provided by EFSA. The I2P2 proved its general example, the outputs of models predicting applicability to quarantine plant pests that the likelihood of entry can be used for can potentially affect EU crops and forests. defining relative risks of different locations to survey in a risk-based approach. The Based on the experience gained so far, outputs of spread models could also help in this modelling framework (which is built prioritising the areas to survey. Model outputs upon international principles of pest risk are integrated in the EFSA Pest survey cards assessment) has improved policy support in (see the virtual issue of the EFSA Journal EU plant health. The main advantages are: (i) at https://efsa.onlinelibrary.wiley.com/doi/ flexibility (conceptual and formal models can toc/10.1002/(ISSN)1831-4732.toolkit-plant- be set at appropriate levels of sophistication pest-surveillance), and are thus used for and resolution), (ii) clarity (results can be supporting decision making (e.g. key distances easily compared among scenarios and for designing delimiting surveys of Xylella for different pests), and (iii) transparency fastidiosa). (not only data, but also lack of data, i.e. uncertainties, are built into the model and The advantages of this quantitative presented in the results). approach are also shown by the recent work on EU priority pests (EFSA et al., References 2019). Article 6 of the new EU Plant Health Law (Regulation 2016/2031) sets out the EFSA PLH Panel (EFSA Panel on Plant Health), requirement to establish a list of priority Jeger M, Bragard C, Caffier D, Candresse pests from the list of EU quarantine pests. T, Chatzivassiliou E, Dehnen-Schmutz K, The prioritization needs to be based on an Gregoire J-C, Jaques Miret JA, MacLeod A, assessment of the severity of economic, Navajas Navarro M, Niere B, Parnell S, Potting social and environmental impacts that those R, Rafoss T, Rossi V, Urek G, Van Bruggen A, pests could cause in the EU. The European Van Der Werf W, West J, Winter S, Hart A, Commission asked the Commission’s Joint Schans J, Schrader G, Suffert M, Kertesz V, Research Centre (JRC) to develop a multi- Kozelska S, Mannino MR, Mosbach-Schulz criteria decision analysis to rank pest species O, Pautasso M, Stancanelli G, Tramontini based on their potential impacts and EFSA to S, Vos S and Gilioli G, 2018. 'Guidance on support JRC’s work by collecting the available quantitative pest risk assessment'. EFSA evidence for 28 pest species, identified Journal 2018;16(8):5350, 86 pp. https://doi. as suitable candidate priority pests, on org/10.2903/j.efsa.2018.5350 their potential capacity for establishment, economic and environmental impacts and EFSA (European Food Safety Authority), Baker difficulty of eradication. For each pest, EFSA R, Gilioli G, Behring C, Candiani D, Gogin A, identified the area of potential distribution Kaluski T, Kinkar M, Mosbach-Schulz O, Neri

180 FM, Siligato R, Stancanelli G and Tramontini to rank candidate priority pests as defined S, 2019. 'Scientific report on the methodology by Regulation (EU) 2016/2031'. EFSA applied by EFSA to provide a quantitative Journal 2019;17(6):5731, 61 pp. https://doi. assessment of pest-related criteria required org/10.2903/j.efsa.2019.5731

181 182 Parallel sessions 5

Scenarios, model linkages and data for policy 3

Room 0.B 27 November 13:30 – 15:00

183 184 Strategic sourcing 2.0: improving depth case studies of selected markets. fiscal efficiency using big data In particular, the analytical framework consists of 3 key components which work Borges de Oliveira A., World Bank best together but can also be deployed Fabregas A., Organisation for Economic Co- independently: operation and Development (OECD) Fazekas M., Central European University 1. public procurement market overview, using interactive visualisations; Public procurement amounts up to one third of general government expenditure across 2. regression modelling of unit prices across OECD countries and little over one fifth across government; and Latin American countries (OECD, 2016, 2017). In the face of such an immense spending 3. in-depth case study analysis of selected effort, maintaining and increasing fiscal large markets such as vehicles. efficiency in public procurement remains a key The analytical framework has been challenge for governments across the world. refined through implementation in seven However, so far systematic, real-time, and Latin American countries including Brazil, readily available efficiency measurements Ecuador, Peru, or Uruguay and further sub- as well as explanatory frameworks have national entities such as the state of Santa been largely lacking, making informed policy Catarina and the city of Rio in Brazil. The making challenging. below empirical examples are drawn from With the spread of e-procurement systems the authors’ analysis of Brazil’s federal around the word (OECD, 2016, 2017), government given that it is the largest the ready and real-time availability of country in our sample.1 government-wide, high granularity spending data, typically on the contract or purchased This article makes a contribution both to item level, is increasingly available. Such the academic literature and policy debates. datasets, if of acceptable quality and scope, From an academic perspective, it develops a can potentially provide the much needed carefully tested mixed methods methodology efficiency metrics as well as identify cues as for price modelling rooted in established to which policy-relevant factors can influence theories and methods but applying them them. to a new context characterised by data of exceptional scope and depth. While there Against this background, this paper introduces have been a few specific studies looking at a novel and comprehensive framework isolated factors using similar data, to the for identifying potential savings and best of our knowledge, our study is the first implementing the necessary policy changes in one to bring this literature together and the procurement of standardized goods and propose a comprehensive methodology. From services. a policy perspective, our methodology offers a reliable and specific tool for policy makers The approach proposed is both narrow and to understand public procurement markets comprehensive. It narrowly focuses on unit and to identify factors driving cost savings, prices paid for standardized goods and either directly under policy influence (e.g. services for most of the analysis, but also length of advertising tenders) or indirectly offers a government-wide methodology influencible (e.g. number of bids submitted). proposing policy improvements across the The methodology feeds into day to day policy board. Conceptually, it combines traditional making, leading to recommendations typically methods of strategic sourcing with economic feasible within existing legal frameworks theories of auctions and data science. by tweaking the parameters of policy Empirically, it employs mixed methods combining country-wide descriptive and 1 The empirical analysis extensively draws on, but also goes considerably beyond the authors’ prior analysis Brazil’s federal explanatory quantitative models with in- government published in (World Bank, 2017, pp. 55–66).

185 implementation. As the methodology is fully In pursuance of better understanding transparent and largely automated, real-time and governance of such large portion of policy advice is feasible as well as simulating government activities, we describe a unique, the impact of distinct policy scenarios. micro-level, large-scale database of the EU- wide Tenders Electronic Daily (TED) combined As a demonstration of how the model is with official national public procurement implemented, our analysis of Brazil’s federal datasets. Such a dataset capture public government is taken as an example. This procurement activities across the whole EU- is a suitable choice as it is the largest 28 between 2006-2018 on the tender and country in our sample. Data-driven price contract levels. It is publicly available thanks modelling points at 15 % savings the to Horizon 2020 funding of the DIGIWHIST federal government due to improving public project at opentender.eu. The total dataset procurement processes and decisions, rather contains over 19 million contracts and it is than fundamentally reconfiguring what is continuously updated using advanced web bought or the regulatory framework. Realistic scraping and parsing technologies. policy changes such as increasing the length of advertising tenders or wider use of Among many potential uses of such data, electronic auctions are required to achieving one will be discussed in detail: assessing projected savings with each intervention the quality of institutions at the regional independently priced guiding policy level. In order to enhance prosperity, human prioritisation. We conclude by discussing the well-being and the territorial cohesion of the strengths and weaknesses of the approach EU, the quality of governance (or quality of and what future improvements could be institutions) is a fundamental precondition. made. High-quality institutions are characterised by 'the absence of corruption, a workable Uses and potential of large- approach to competition and procurement scale, micro level government policy, an effective legal environment, contracting datasets for policy and an independent and efficient judicial system', as well as 'strong institutional modelling and administrative capacity, reducing the Fazekas M., Central European University administrative burden and improving the Tóth B., University College London quality of legislation' (European Commission, Czibik A., Government Transparency Institute 2014, p. 161). Such a broad understanding of institutional quality is also underpinned Public procurement plays a crucial role in by influential academic thinking focusing economic development and the quality of on impartial policy implementation rather government across the European Union (EU): than the content of policies or democratic on average, it amounts to about 13 % of GDP decision-making processes (Rothstein & or 29 % of government spending (European Teorell, 2008). Building on this focus on policy Commission, 2016; OECD, 2015). It is a implementation, good governance in public genuinely cross-cutting government function procurement is assessed according to four concerning virtually every public body, and main dimensions: is also one of the principal means by which governments can influence growth rates • Transparency (e.g. amount of and the quality of public services. However, information published in procurement our understanding of the quality of public announcements); procurement processes and outcomes is very • Competition (e.g. average number of much in its infancy, which limits governments’ bidders); capacity to intervene in pursuance of specific public procurement as well as • Administrative efficiency (e.g. length of broader objectives such as public financial decision-making period); and management or local economic development.

186 • Corruption (e.g. the use of non-open, of a policy hence requires to know what opaque procedure types). carbon emissions would have been, had the policy never been introduced, or what causal Each dimension of good governance as well inference scholars call the counterfactual. as a composite score are calculated and their validity tested by comparing them to widely In this paper, we apply the generalized used regional indicators such as GDP/capita, synthetic control method (Xu, 2017) to European Quality of Government Index (EQI), estimate counterfactual emission paths for or public service meritocracy. All tests confirm European countries regulated under the that the indicators proposed, based on prior European Union Emissions Trading System academic and policy literature, are valid. (EU ETS) to assess the effectiveness of EU carbon markets. Importantly, this modelling The new indicators enable a detailed analysis approach offers a statistical method to of the quality of NUTS 3 and NUTS 2 regional 'purge' estimates of the effects of other public procurement governance according (confounding) factors, such as the economic to the four above-mentioned dimensions, crisis or increases in renewable capacity while changes over the last 10 years can over the same time. Drawing on the newly also be explored. We find a mixed picture of created European Union Sectoral Emissions regional convergence between 2006–2015 Data (Bayer, 2019), we show that the EU in the EU. While some Central and Eastern ETS reduced emissions by about -11.5% European regions have converged to the (95% confidence interval: [-16.8%; -4.8%]) EU average, many Mediterranean regions in sectors that it covers compared to the have strongly diverged and, surprisingly, counterfactual in which no EU ETS had been some well-governed Western and Northern implemented. This translates into 1.2 billion European regions have also experienced a tons of CO emissions saved from 2008– strong deterioration in governance quality. 2 2016,or reductions of 3.8% relative to total Overall, governance quality and competition emissions. in particular have deteriorated across the whole EU. Aside from the substantive importance of our findings given constant criticism over Using synthetic control methods the ineffectiveness of carbon markets for policy evaluation: modelling due to persistently low prices, this paper challenges, new data, and introduces a powerful statistical method for evidence from EU carbon markets policy evaluation that travels beyond our application to the EU ETS. In fact, we believe Bayer P., School of Government and Public that the generalized synthetic control method Policy, University of Strathclyde is currently underused to assess policy Aklin M., Department of Political Science, effectiveness and guide political decision- University of Pittsburgh making, so by popularizing this approach, our Motivation and summary work also contributes to the learning of the modelling community as a whole.. Assessing whether a policy works is an enormously difficult task. This assessment Counterfactual modelling and synthetic is complicated by the fact that whenever control a new policy is introduced, changes in the As indicated above, the main challenge for outcome can rarely (if ever) be attributed assessing whether a policy works or not to the policy alone as many other factors is to isolate the effect of the policy, e.g., change at the same time. For example, a the introduction of the EU ETS in our case, country’s carbon emissions may decrease from effects of other factors that change because of new regulatory policy, but the at the same time, e.g., changes to economic reduction may equally well stem from lower growth. A rigorous understanding of policy economic activity. Isolating the causal effect effectiveness therefore requires to compare 187 the observable outcome of a policy against (2008–2016) against the counterfactual. a counterfactual outcome that would To be clear: these results do not mean that have prevailed had the policy never been carbon emissions across Europe reduced by introduced. For us, this means we need to 8.1-–11.5% over the last decade. Reductions compare actual carbon emissions to those were much higher, of course. Instead, these emission levels we had seen in a world reductions are estimates of additional CO2 without the EU ETS. The obvious problem here emission reductions because of the EU is that the latter can never be observed. ETS within the sectors it covered on top of the decline in emissions we have seen One way around this problem is to estimate anyways. In other words, the EU ETS carbon the counterfactual using statistical market policy is associated with substantial techniques. The synthetic control method is decarbonization of about 1.2 billion tons in principle a simple re-weighting approach during 2008–2016. (Abadie, Diamond, and Hainmueller, 2010, 2015). To assess the effect of a policy in Except for emissions from the paper industry, country A on outcome O, we can use a we find similar reduction patterns across re-weighted combination of outcome O other sectors, such as energy, chemicals, in countries B-Z (so called 'donor pool') to minerals, and metals. In a placebo test for approximate as best as possible outcome emissions from the transport sector, which is O before the introduction of the policy in not regulated under the EU ETS, we find no country A — this makes for the synthetic decrease in emissions, which increases our control. The difference between the value confidence in the estimation strategy and the of O after the introduction of the policy in results. country A and the post-policy value of the re-weighted O from the other countries is an Contributions to learning for modelling estimate of the policy’s average treatment community effect on the treated (ATT), i.e. an estimate Aside from the substantive importance of our of its effectiveness. While this approach findings for climate change policy and carbon was originally formulated for a policy in a market regulation in particular, our paper single country, Xu (2017) generalized it to a demonstrates the use of a very powerful multi-country setup where multiple countries statistical method for policy assessment. introduce a policy, or 'get treated' in causal Synthetic control methods remain underused inference parlance. despite a very straightforward application. Main results Importantly, it allows not only to estimate point estimates of policy effectiveness We apply this generalized synthetic control but also measures of uncertainty from method to EU carbon markets (Bayer and bootstrapping. A battery of robustness tests Aklin, 2019). A very practical challenge here also exists, and powerful visualizations allow is that in order to assess the effect of the for an intuitive interpretation of the results.

EU ETS on countries’ CO2 emissions, we need We are confident that popularizing this type emissions data for regulated sectors before of statistical method will be useful for the the introduction of the EU ETS in 2005. For modelling community as a whole, and our this, we construct the novel EUSED data set analysis of the EU ETS is but one application. (available at http://patrickbayer.com/data/) that maps carbon emissions at the sectoral References level from 'National Emissions Reported to Abadie, Alberto, Alexis Diamond, and Jens the UNFCCC' and EUTL data, after accounting Hainmueller. 2010. 'Synthetic Control Methods for differences in reporting standards. for Comparative Case Studies: Estimating Drawing on this data, we estimate that the EU the Effect of California’s Tobacco Control Program.' Journal of the American Statistical ETS reduced CO2 emissions in covered sectors by between -8.1% (2005–2016) and -11.5% Association 105 (490): 493–505. 188 Abadie, Alberto, Alexis Diamond, and Jens several times during the lifetime of a typical Hainmueller. 2015. 'Comparative Politics research project, a clear advantage of user- and the Synthetic Control Method.' American created R scripts emerges: scripts can be Journal of Political Science 59 (2): 495–510. executed repeatedly with a minimum effort for interaction with the software interface, Bayer, Patrick. 2019. European Union Sectoral each time simulation results have been Emissions Data (EUSED). Available at https:// updated. Data visualizations, statistical and dataverse.harvard.edu/dataverse/eused. econometric analyses based on of third- party R packages can be easily replicated and Bayer, Patrick, and Michaël Aklin. 2019. 'The repeated in this manner. European Union Emissions Trading System Reduced CO2 Emissions Despite Low Prices.' We present an R package developed for the Working Paper. Common Agricultural Policy Regionalized Impacts (CAPRI) modelling system, a large- Xu, Yiqing. 2017. 'Generalized Synthetic scale economic model with a particular Control Method: Causal Inference with focus on agriculture and food markets. The Interactive Fixed Effects Models.' Political capriR package includes specific functions Analysis 25 (1): 57–76. for processing, visualizing and analysing both Combining graphical and command the model databases and simulation results. capriR has been designed to complement line user interfaces for economic the specific GUI of CAPRI, which remains the analysis: the capriR package preferred options for quick analysis of model for processing, visualizing and results and for executing the wide range of analysing large sets of simulation modelling tasks from database preparation results until simulation runs. The relative advantages of capriR compared to the GUI include the Himics M., European Commission, Joint (i) dissemination of model-databases and Research Centre simulation results; (ii) automated reporting requiring additional (post-model) calculations Large scale economic models naturally and (iii) creating publication-quality maps and produce large datasets of simulated results. other data visualizations. Analysing large amount of data is often beyond the capacity of the human mind, at As CAPRI covers EU agricultural production least without the help of specific software activities with fine geographical detail tools designed for filtering, transforming (NUTS2 administrative regions), spatial data and visually presenting the datasets. Many and their visualization, analysis is a particular economic models also offer visual aid and challenge. capriR links CAPRI results to easy data access possibilities for their users commonly used spatial data packages thus via graphical user interfaces (GUI). Most GUIs enabling the user to create high-resolution still require the user to do numerous and time intensity maps or even interactive maps. consuming interactions with the software, capriR also includes functions for rapid access slowing down the analysis of simulation of the databases and simulation results of results, and sometimes even hindering the CAPRI modelling system. The modularity model users to find the relevant drivers of the R programming language allows for and other causality chains in model results. directly applying advanced econometric and The R programming language provides statistical techniques from other (open- complementary data exploitation possibilities source) R packages on the data sets retrieved with its command line interface (CUI) and from CAPRI. with a large number of optional packages for analysing and visualizing large datasets. As Although capriR is only directly useful for simulation exercises and the related reporting the relatively small user base of CAPRI, we and data analysis need to be repeated also point out in this paper some general strategies for rapid package development 189 for similar, large scale economic models. inputs and outputs) between different model What makes the capriR approach potentially interesting for a wider economic modelling architectures poses a practical challenge community is that model-specific R packages to modelling groups. Model-specific R can be built with limited efforts, and under packages for data exchange offer a common limited time for the same purposes, including software platform. The large user-base of dissemination of results, visualization and the R programming language in the broader complex post-model data analysis. This scientific community makes such packages is particularly important for modelling efficient in disseminating model results for a approaches where different types of general scientific audience, increasing at the quantitative models are linked. The need for same time the transparency of modelling exchanging large amounts of data (model exercises. Opening the black box of complex

Figure 1: Intensity map of CAPRI simulation results prepared with the capriR package

190 large-scale models by making databases human ignition probability, a gridded and results easily accessible can lead to population density is used. Fuel available huge gains in credibility for the modelling for burning is defined as a combination of community, and is also an essential part in litter and coarse woody debris (CWD) pools, evidence based policy making. excluding stem biomass. We use integrated modeling approach, where biomass dynamics Below is a screenshot (Figure 1) presenting is provided by the IIASA’s global forestry some of the spatial data functionalities of model G4M. The fire suppression efficiency is capriR. implemented in FLAM as the probability of extinguishing a fire on a given day. Wildfire modelling for adaptation policy options in Europe The FLAM’s modeled burned areas for selected test countries in the EU show good Krasovskii A., Khabarov N., Kindermann G., agreement with observed data coming from Kraxner F., International Institute for Applied two different sources (the European Forest Systems Analysis (IIASA) Fire Information System and the Global Fire This study presents a quantitative Emissions Database). We employ climate assessment of adaptation options and projections corresponding to four RCP- related policies in the context of wildfires in scenarios. Our estimation of the potential Europe under projected climate change. The increase in burned areas in Europe under Wildfire Climate Impacts and Adaptation a 'no adaptation' scenario is about 200 % Model (FLAM) is able to capture impacts of by 2090 (compared with 2000–2008). The climate, population, and fuel availability on application of prescribed burnings has the burned areas. FLAM uses a process-based potential to keep that increase below 50 fire parameterization algorithm that was %. Improvements in fire suppression might originally developed to link a fire model with reduce this impact even further, e.g. boosting dynamic global vegetation models. The key the probability of putting out a fire within features implemented in FLAM include fuel a day by 10 % would result in about a 30 moisture computation based on the Fine Fuel % decrease in annual burned areas. By Moisture Code (FFMC) of the Canadian Forest identifying policy options that emphasize on Fire Weather Index (FWI), and a procedure to adaptation options such as using agricultural calibrate spatial fire suppression efficiency. fields as fire breaks, behavioral changes, and long-term options, burned areas can be Currently FLAM operates with a daily time- potentially reduced further than projected in step at 0.25-arc degree spatial resolution. our analysis. All inputs in FLAM are adjusted to fit this resolution. FLAM uses daily climate data In the talk we will discuss policy for temperature, precipitation, wind, and recommendations, as well as demonstrate relative humidity. When calculating the some visualization tools.

191 192 Parallel sessions 5

Computational and empirical issues in dynamic economy-wide models

Room 0.C 27 November 13:30 – 15:00

193 194 Using the Global Multi-Country Bokan et al., 2018; Karadi et al., 2018; Erceg model in ECFIN’s forecast et al., 2006). exercises The GM model within the EC forecast Calés L., Cardani R., Croitorov O., Di Dio F., rounds Frattarolo L., Giovannini M., Hohberger S., Since autumn 2015, the GM model is Pataracchia B., Ratto M., European Commission, Joint Research Centre, regularly employed in the EC’s institutional Pfeiffer P., Roeger W., Vogel L., European forecast to understand the drivers of the Commission, Directorate General for Economic evolution of euro area macro-economic and Financial Affairs variables, providing input to the Thematic Boxes included in the EC Spring and Autumn Introduction Forecast reports. More specifically, by means of a structural model, EC policy analysts can We present the application of the Global get a sound interpretation of macroeconomic Multi-country model (GM) in the context of data, by decomposing the dynamics of GDP, the European Commission (EC)’s institutional inflation, consumption, investment, trade, forecasts. GM is a structural macroeconomic employment, etc. into key drivers, such as the model, jointly developed by the Joint Research evolution in domestic and foreign demand, Centre and DG ECFIN to perform forecasting, commodity prices, and productivity, as well as medium term projections and spillover fiscal and monetary policy. analysis. To do so, the dataset of macro-economic GM is a fully estimated model that is flexible variables is extended up to the forecast to allow for different country configurations. horizon using the European Commission The two-region version of GM (GM2: Euro forecasts. This extended dataset is analysed Area and Rest of the World) is the main to recover the drivers (shocks) mainly version applied for the EC forecast rounds. responsible for the evolution of the macro- The three-region version of GM (GM3: Euro variables, which can then be represented in a Area, US and Rest of the World) has been decomposed form, where all the drivers add- used to analyse different patterns of the up to replicate the data. post-crisis slump in the Euro Area and the US (Kollmann et al., 2016). The EMU-countries An example: the EC Autumn Forecast 2018 version (GM3-EMU)1 is estimated for the four largest European economies (Germany, Graphs 1-4 show the decompositions of EA France, Italy and Spain), and has been used GDP growth, output gap, consumer price for the cross-country comparison of the inflation and trade-balance to GDP ratio post-crisis evolution of Euro Area economies (European Commission, 2018). These graphs (Albonico et al., 2019a, b). show how the estimation of a structural The GM model builds on the estimated model allows the identification of the factors version of the QUEST III model (Ratto et that drive the short- and medium-term al., 2009), from which it inherits most of deviations of key macro-economic variables its structure. In GM2, the global economy from their long-run trends (including GDP, consists of two regions, the domestic Euro inflation, domestic demand, and trade Area (EA) and the Rest of the World (RoW), balance). which are connected by trade and financial In short, these model-based decompositions linkages. The EA region is fully specified and attribute above-trend euro area real GDP features households, firms, and fiscal and growth in 2019 to continued domestic monetary policy authorities’ blocks, while demand growth and to an accommodative the RoW has a simpler structure. The GM monetary policy stance (Graph 1). Domestic model falls into the class of large-scale DSGE demand levels remain below the historical models used for policy analysis (see, e.g. average, however, implying a still negative (although diminishing) contribution to the 1 EMU stands for Economic and Monetary Union.

195 Note: The solid black line shows the historical data, and the dashed line represents the European Commission’s forecast for 2018 and 2019. The coloured bars show the contribution of the groups of driving forces to deviations of GDP growth from its long- run trend of approximately 1.4%, i.e. the value at which the (solid) horizontal axis intersects the vertical axis. Bars above (below) this horizontal axis indicate positive (negative) contributions to GDP growth in a given year. The sum of positive and negative contributions matches the data (black solid line) for any point in time.

level of economic activity (Graph 2). Below- the Global Multi-country model,' Economic average levels of domestic demand also Modelling, 81: 242–273. explain low levels of inflation in the euro area (Graph 3) and a significant part of the trade ALBONICO, A., L. CALÈS, R. CARDANI, balance surplus (Graph 4) incorporated in the O. CROITOROV, F. DI DIO, F. FERRONI, M. European Commission’s forecast for 2019. GIOVANNINI, S. HOHBERGER, B. PATARACCHIA, F. PERICOLI, P. PFEIFFER, R. RACIBORSKI, M. References RATTO, W. ROEGER AND L. VOGEL (2019b): 'The Global Multi-Country Model (GM): an ALBONICO, A., L. CALÈS, R. CARDANI, O. Estimated DSGE Model for the Euro Area CROITOROV, F. FERRONI, M. GIOVANNINI, S. Countries,' ECFIN Discussion Paper No. 102 , HOHBERGER, B. PATARACCHIA, F. PERICOLI, European Commission, 2019. R. RACIBORSKI, M. RATTO, W. ROEGER AND L. VOGEL (2019a): 'Comparing post-crisis BOKAN, N., A. GERALI, S. GOMES, P. dynamics across Euro Area countries with JACQUINOT, AND M. PISANI (2018): 'EAGLE-

196 FLI: A macroeconomic model of banking and for non-linear models. In the sunspot financial interdependence in the euro area,' equilibria considered here, the economy may Economic Modelling, 69, 249–280. temporarily diverge from the no-sunspots trajectory, before abruptly reverting towards ERCEG, C. J., L. GUERRIERI, AND C. GUST that trajectory. In contrast to rational bubbles (2006): 'SIGMA: A New Open Economy Model in linear models (Blanchard (1979)), the for Policy Analysis,' International Journal of bubbles considered here are stationary- Central Banking, 2. -their expected path does not explode to infinity. Numerical simulations suggest that EUROPEAN COMMISSION (2018): 'European non-linear DSGE models driven by stationary Economic Forecast – Autumn 2018,' European bubbles can generate persistent fluctuations Economy Institutional Paper, November 2018. of real activity and capture key business cycle KARADI, P., S. SCHMIDT, AND A. WARNE stylized facts. Applications to both closed and (2018): 'The New Area-Wide Model II: an open economies are analyzed. extended version of the ECB micro-founded This paper seems highly relevant for model for forecasting and policy analysis with the Brussels conference, because it a financial sector,' Working Paper Series 2200, challenges linearized macroeconomic European Central Bank. models that are routinely developed and KOLLMANN, R., B. PATARACCHIA, R. used for quantitative policy analysis RACIBORSKI, M. RATTO, W. ROEGER, AND by the European Commission and other L. VOGEL (2016): 'The post-crisis slump in policy institutions. See, e.g., Giovannini the Euro Area and the US: Evidence from et al. (2019); Kollmann et al. (2016, 2015, an estimated three region DSGE model,' 2014, 2012, 2013). The results here show European Economic Review, 88, 21–41. that the predictions of standard dynamic macroeconomic policy models can change RATTO, M., W. ROEGER, AND J. IN ’T VELD radically when non-linearity is taken into (2009): 'QUEST III: An estimated open- consideration. In particular, standard policy economy DSGE model of the euro area prescriptions need not be valid anymore, with fiscal and monetary policy,' Economic in a nonlinear environment. This project Modelling, 26, 222–233. contributes, thus, to the construction of more reliable macroeconomic policy models. Stationary rational bubbles in non-linear macroeconomic policy Further reading: models http://www.robertkollmann.com/KOLLMANN_ Bubbles_NonLin_DSGE_updated.pdf Kollmann R., Université Libre de Bruxelles and Centre for Economic Policy Research (CEPR) http://www.robertkollmann.com/KOLLMANN_ Linearized dynamic stochastic general SLIDES_Bubbles_NonLin_DSGE_updated.pdf equilibrium (DSGE) models with a unique References stable solution are the workhorses of modern quantitative macroeconomics and of Blanchard, Olivier, 1979. 'Speculative quantitative macroeconomic policy analysis. Bubbles, Crashes and Rational Expectations'. This paper shows that stationary sunspot Economics Letters 3, 387–398. equilibria exist in completely standard non- linear macroeconomic models, even when Blanchard, Olivier and Charles Kahn, 1980. the linearized versions of those models have 'The Solution of Linear Difference Models a unique solution. Thus, those models have under Rational Expectations'. Econometrica additional stationary solutions, if non-linearity 48, 1305–1311. is considered. The classic Blanchard and Kahn (1980) conditions are, hence, irrelevant Giovannini, Massimo; Stefan Hohberger, Robert Kollmann, Marco Ratto, Werner Roeger 197 and Lukas Vogel, 2019. 'Euro Area and U.S. effects. Intertemporal deterrence effects External Adjustment: The Role of Commodity arise from companies’ expectations that Prices and Emerging Market Shocks', Journal the European Commission will continue of International Money and Finance 94, its competition policy interventions at the pp.183–205. same pace into the foreseeable future. Sectoral deterrence effects come from Kollmann, Robert; Beatrice Pataracchia, Rafal the transmission of the direct effects of Raciborski, Marco Ratto, Werner Roeger and competition policy interventions in a given Lukas Vogel, 2016. 'The Post-Crisis Slump market to the remainder of the sector to in the Euro Area and the US: Evidence from which this market belongs. The within-sector an Estimated Three-Region DSGE Mode', diffusion of the direct effects is modelled by European Economic Review 88, 21–41. way of a logistic function, which has been used in the literature to model the diffusion Kollmann, Robert; Jan in’t Veld, Marco Ratto, of innovation or the use of new technologies Werner Roeger and Lukas Vogel, 2015. 'What but is used here to model deterrence. Model Drives the German Current Account? And simulations show that the total effects How Does it Affect Other EU Member States?' (including the deterrent effects) of the Economic Policy 40, 47–93. European Commission’s competition policy Kollmann, Robert; Jan in’t Veld, Beatrice interventions on GDP are sizeable, but slightly Pataracchia, Marco Ratto and Werner Roeger, lower than the estimated impact of the 2014. 'International Capital Flows and implementation of the EU Services Directive. the Boom-Bust Cycle in Spain', Journal of Under the baseline scenario, GDP increases International Money and Finance 48, by 0.29% after five years and by 0.56% in 314–335. the long term. The effects on employment rise from 0.20% after five years to 0.26% Kollmann, Robert; Werner Roeger and Jan in’t in the long term. The employment effects Veld, 2012. 'Fiscal Policy in a Financial Crisis: are smaller than the GDP effects due to the Standard Policy vs. Bank Rescue Measures', increase in labour productivity associated American Economic Review 102, 77–81. with the increased competitive pressures.

Kollmann, Robert; Marco Ratto, Werner Roeger The current paper also exploits the available and Jan in't Veld, 2013. 'Fiscal Policy, Banks information on the sector distribution and the Financial Crisis', Journal of Economic of the European Commission's merger Dynamics and Control 37, 387–403. interventions and cartel prohibitions over the recent period. The version of the QUEST III The macroeconomic and sectoral macro-model used is a single-sector model impact of EU competition policy unsuited to explore the industry spill-over effects of competition policy interventions. Cai M., Cardani R., Pericoli F., European This is the reason why the macro-model Commission, Joint Research Centre simulations have been complemented by an Dierx A., Ilzkovitz F., European Commission, input-output model analysis, which allows Directorate-General for Competition tracking the interlinkages between industries This paper provides an assessment of the and identifying the differential effects of macroeconomic and sectoral impact of competition policy interventions affecting merger interventions and cartel prohibitions different industries of the economy. by the European Commission over the period This input-output model is used to exploit 2012-2018. available information on the distribution of The macroeconomic simulations conducted merger interventions and cartel prohibitions using the QUEST III macro-model consider across industries and to explore how the both the direct effects of competition policy avoided price increase resulting from interventions and two types of deterrent a competition policy intervention in a

198 given industry is transmitted to the price European Structural and Investment Funds levels in other industries. The total price and European Fund for Strategic Investment. reduction associated with competition policy interventions over the period 2012- The present study focuses on evaluating 2018 is significant (-0.2%). Price reductions the impacts of the European Institute of occur in industries that are important Innovation and Technology (EIT) investments for the purchasing power of consumers in the period 2020-2050 using a spatially (motor vehicles, financial services) and the explicit macroeconomic model for Europe with competitiveness of the European economy a regional and sectoral detail that captures (telecommunications, electronics). The spillovers from investment in the knowledge industry spill-over price effects of competition and human capital. policy interventions may be as important as In order to undertake a comparative scenario their within-industry price effects. Industry analysis and assess impacts of selected EIT spill-overs were particularly high in 2013, investment support policies first a baseline when two important cartels in financial scenario is constructed and simulated. There markets were prohibited. Spill-overs tend is no EIT supported investment implemented to be higher in industries having strong in the baseline scenario; baseline indicators, interlinkages with the rest of the economy, such as an additional investment leverage such as financial services, network industries or impact on GDP, are used as benchmark or intermediate goods industries. against which to compare EIT policy scenario Direct and indirect impacts of outcomes. European Institute of Innovation Second, alternative EIT investment support and Technology (EIT) investments (counterfactual) scenarios are constructed on regional economic growth and simulated. The policy scenario construction requires data on private co- Ivanova O., Thissen M., PBL Netherlands funding rates for each year in the EIT Environmental Assessment Agency scenario. The rest of the co-funding is coming Kancs d'A., European Commission, Joint from the EIT funds. The Horizon Europe Research Centre proposal sets out the budget for the EIT (EUR 2 The present study evaluates regional 3 billion for the period 2021–2027) . Further, economic impacts of the European we assume that the spatial investment Institute of Innovation and Technology (EIT) pattern of the EIT will remain the same investments in the period 2020-2050 using as in the base year EIT expenditure data spatially disaggregated Computable General meaning that only a subset of EU28 regions Equilibrium (CGE) model with endogenous will receive the funding and that the regional growth engines. The model used for the pattern of investments follows the historical analysis allows us to take into account pattern. both direct and indirect impacts of the EIT The EIT investment support affects investments via inter-regional trade linkages economy, society and environment in many and endogenously determined global different ways, posing challenges to the knowledge frontier. methodological framework for capturing Innovation and human capital have been all the impacts, as they are diverse and widely recognised as key drivers of a complex due to various inter-sectoral, inter- sustainable economic growth in the long- regional and inter-temporal linkages and run. The European Union is implementing a interdependencies. In the context of EIT number of policy instruments to promote activities in the areas of research/innovation, the innovation activity in Europe including education and business creation/support in among others the Framework Programme, 2 Article 9, COM(2018) 435 final.

199 improving the innovative performance of the investments on the productivity of the Member States and the Union, three types economic sectors. The total EU direct effect of effects of EIT-supported investments are related to EIT investments impact on sectoral of particular interest: (i) demand effects productivity in the period 2020-2050 amount (e.g. hiring of workers, machinery), structural to 2 170 million of Euros whereas the overall effects (e.g. productivity and human capital effects in the same period amount to 18 520 growth) and macroeconomic effects (e.g. on million of Euros. The EU-wide ratio of the GDP and employment). Generally, there are total to direct effects of EIT investments in many more economic impacts, as well as the period 20202–2050 is around 6. Overall societal and environmental effects which, about 1/3 of EU NUTS2 regions experience however, are beyond the scope of the present positive effects of EIT investments and 2/3 of analysis. EU NUTS2 regions experience negative effects of EIT investments. These negative effects In order to calculate the direct impacts of are experiences not only by the regions that the EIT investments in the period 2020- do not receive EIT investments but also by 2050, we have calculated the impacts the regions that receive relatively small EIT of EIT investment on changes in sectoral investments. productivity in each of EU NUTS2 regions based on the estimated TFP regressions Modelling the labour market and regional data. The relative changes in and agricultural sector in an productivity of each of the sectors have been translates into the changes in sectoral value integrated CGE framework: an added by sector and by region and the new application to the Brexit case regional GDP has been calculated as the sum Angioloni S., Wu Z., Department of Agri-Food of the increased value added of the sectors. Economics, Agri-Food Biosciences Institute Among the regions that have the largest Berrittella M., Dipartimento di Scienze direct benefits from EIT investments are the Economiche Aziendali e Statistiche, Università Provincia Autonoma di Trento in Italy, Noord- degli studi di Palermo Brabant in the Netherlands, Ile-de-France in Aguiar A. H., Center for Global Trade Analysis France, Oberbayern and Berlin in Germany in Project, Purdue University the order of their direct benefits. The direct benefits for these regions range between Free cross-border movements of goods, 130 and 400 million of Euros in the period services, investment and people are the 2020–2050. principal pillars of the European single market. The UK is preparing to leave the EU Total regional effects of EIT investments with departure date temporarily scheduled can be both positive and negative meaning for the end of October 2019 plus a possible that economic growth of the regions that transition period (House of Commons are directly affected by EIT investments Library, 2019). Whenever this happens, the can results in economic decline in some UK will need to put in place new trade and other regions. This can be due to increased immigration policies necessary to deliver competitiveness of the regions with EIT the novel relationship with the EU. Ideally, investments and the respective relocation the UK would like to decide immigration of economic activities to these regions. The policy and trade policy separately such that regions with the largest negative indirect it can maximize the outcome from each effects include Massa-Carrara and Veneto one of them. More realistically, the UK will in Italy, Nord-Pas de Calais and Bretagne in be required to face trade-offs between France as well as Arnsberg and Weser-Ems in trade and immigration policies especially Germany. in sectors such as the agri-food industry. A complete understanding of the effect of these The largest total effects of EIT investments combinations is therefore impossible without are associated with the same regions that understanding how trade and immigration had the largest direct effects of these policies interact within the UK economy. 200 This paper investigates the effects of the Immigration scenarios were based on the interactions between trade and labour forecasts of the EEA net inflow of workers immigration policies. We apply a computable in the UK until 2030, the prediction window general equilibrium (CGE) model based on employed in this study. The first immigration the GTAP 10 database specifically designed shock is the most extreme and assumes zero to count for imperfect native-migrant net EEA inflow in the UK (-2.1% reduction of substitution across sectors and to allow for labour force). The second scenario assumes imperfect labour mobility between agriculture that there are no modifications to the current and the rest of the economy. The CGE model immigration policy. This would imply a modest is employed to analyse the effect of different reduction of labour force (-0.2%) to reflect trade and labour immigration restrictions on the trend started after the Brexit referendum the UK economy within the Brexit debate. that has already witnessed a reduction of the number of EEA migrant workers in the UK. The process of relocating labour can be slow Finally, the third scenario imposes zero net because skills are not perfectly transferable inflow of unskilled EEA workers (-1.1% labour across sectors and retraining takes time and force). This option was included to reflect money (Campo, Forte, and Portes, 2018). In some immigration options considered in the other words, workers specialize in occupations UK to limit only the number of this type of where they have a comparative advantage. workers (UK Government, 2018; Migration Similarly, natives and migrants have different Advisory Committee, 2018). Immigration skill sets and immigration encourages shocks were integrated with three baseline workers to specialize where they have a trade shocks: a no deal scenario, an average lower opportunity cost (Peri and Sparber, free trade agreement, and a version of soft 2009). Following Ottaviano and Peri (2012), Brexit as described in UK Government (2018). we estimated the native-migrant constant elasticity of substitution (CES) by industry Our findings show that in the UK severe to for the UK to differentiate the production of moderate immigration shocks can generate value added, and thus the demand for labour, a reduction in GDP and welfare larger by skill type and migrant/native status. than a free trade agreement and a soft Brexit scenario. Moreover, UK agriculture Besides, although farm and non-farm wages is particularly reliant on the EEA unskilled have moved together, there is a substantial workers. A reduction of EEA unskilled labour evidence that wage differential persists would reduce output and increase prices in in developed economies (Kilkenny, 1993). agriculture more than in other non-agriculture In the UK, low unemployment rates and a sectors. This suggests the implementation high number of vacancies are persistently of ad-hoc immigration policies such as the observed in many sectors of the agri-food seasonal agricultural workers scheme and the industry (Office of National Statistics, 2018). extension of the shortage occupation lists to Following the GTAP-AGR model developed agricultural labourers would be needed if the by Keeney and Hertel (2005), we have UK will give up to the freedom of movement integrated a module for segmenting the with EU nationals. These results can be market for mobile factors in the agricultural employed to improve the decision-making and non-agricultural sectors according to the process in matter of immigration, trade, and change in relative prices and the elasticity of agricultural policies outside the Brexit debate. transformation. References We also included two further specifications. First, livestock sectors have their own Campo, F., G. Forte, and J. Portes. 2018. demand system for feedstuff as to reflect 'The Impact of Migration on Productivity their imperfect substitutability. Second, we and Native-born Workers’ Training, External allowed the agricultural sectors to have their Report'. Migration Advisory Committee; own production system that employs other London, United Kingdom. agricultural inputs as imperfect substitutes. 201 House of Commons Library. 2019. 'Brexit Office of National Statistics. 2018. 'Population Delayed Again: Until 31 October 2019?' of the UK by Country of Birth and Nationality: Research Briefing. London, United Kingdom. 2017.' Office for National Statistics, Accessed April 2018. Keeney R. and Hertel T.W. 2005. 'GTAP-AGR: A Framework for Assessing the Implications of Ottaviano, G. I. P., and G. Peri. 2012. Multilateral Changes in Agricultural Policies'. 'Rethinking the Effect of Immigration on GTAP Technical Paper No. 24, Center for Wages.' Journal of the European Economic Global Trade Analysis, Purdue University, Association 10 (1):152–197. doi:10.1111/ Indiana, USA. j.1542-4774.2011.01052.x.

Kilkenny, M. 1993. 'Rural/Urban Effects Peri, G., and C. Sparber. 2009. 'Task of Terminating Farm Subsidies,' American Specialization, Immigration, and Wages.' Journal of Agricultural Economics 75: 968– American Economic Journal: Applied 980. Economics 1 (3):135–169. doi:10.1257/ app.1.3.135. Migration Advisory Committee. 2018a. 'EEA Migration in the UK: Final Report.' UK Government. 2018. 'EU Exit: Long-Term Accessed October 2018. https://www.gov.uk/ Economic Analysis, Technical Reference government/publications/migration-advisory- Paper'. Accessed January 2019. committee-mac-report-eea-migration.

202 Poster session 2

27 November

203 204 A coupled modelling system for publically available on a dedicated webserver, the integrated assessment of as well as the option for users to develop own scenarios and assess the costs. continental freshwaters, coastal and marine waters Following the regional particularities, the outcome of the scenarios with respect to Friedland R., Bisselink B.,Bouraoui F., de Roo A., freshwater runoff and nutrient loads differ Gelati E., Garcia-Gorriz E., Grizetti B., Guenther strongly between the regional seas. Having S., Macias D., Miladinova S., Parn O., Piroddi C., Stips A., Vigiak O., European Commission, Joint developed the needed interfaces within the Research Centre project, the nutrient loads scenarios will be applied for the regional seas and analysed While the implementation of the Marine for impacts on the marine environment, e.g. Strategy Framework Directive (MSFD) and to assess if Good Ecological Status (GES) the Water Framework Directive (WFD) are thresholds (defined within the MSFD) might progressing in the Member States (MS), the get fulfilled. European Commission is building up its own analytical capacity in order to improve the A platform to manage air quality understanding of the coupled freshwater in Europe with the use of high- and marine environment from an EU resolution modelling tools perspective. Managing as well as modelling the marine environment in a sustainable Application and validation in manner are only possible as long as the Krakow full continuum of freshwater systems (like Vranckx S., Hooyberghs H., Vanhulsel M., Blyth rivers), coastal waters and the open sea is L., Smeets N., Veldeman N., Maiheu B., Lefebvre considered. The environmental state of all W., Janssen S., VITO, Vlaamse Instelling voor seas is largely determined by the supply of Technologisch Onderzoek freshwater, nutrients, contaminants, litter, Bielas E., Zaleski W., Transport System Division, etc. from land. Hence, measures developed Municipality of Krakow within the implementation cycles of EU’s MSFD and the Water Framework Directive ATMO-Plan is a user friendly web based have – wherever possible – to be assessed decision support tool, designed to facilitate for the full continuum. A step forward to the assessment of the impact of emission support this integrated understanding is reduction scenarios on air quality in Europe the BLUE2 project, where a freshwater on an urban scale at a high spatial resolution. resources model (LISFLOOD) and a nutrient The tool is pre-configured with EU generic load model (GREEN) were coupled to the data for the EU-28, users can upload their marine biogeochemical models. This allows own data to further improve the quality of an integrated assessment of pan-European the air quality information obtained. water resources from the sources on land to The web-tool applies an operational the sinks in the regional seas. modelling chain starting from regional A database was developed including background concentrations, meteorology, envisaged measures to reduce pollution fleet data and the road network with and water abstraction, originating from traffic intensities. Regional background the implementation of various directives. concentrations for the whole interregional Combining climate change projections and hotspot area can been modelled using RIO[1]. measures enabled us to develop joint future This model is based on a residual kriging scenarios, assuming different progresses in interpolation scheme starting from hourly the implementation of the various measures, pollutant concentrations as measured by the ranging from 'Business As Usual' to 'Maximal official monitoring stations and using land use Technical Feasible'. The outcome of the (CORINE) and population density as spatial spatially explicit scenarios will be made information. A leaving-one-out validation of hourly background concentrations shows 205 the strength of this modelling approach. challenges. To validate the quality of the Alternatively, CAMS reanalysis data have been model chain, the city of Krakow has organized pre-configured for the EU-28 as background two validation campaigns in summer and concentrations. winter 2017. Each campaign lasted 28

days applying NO2 passive samplers at As road traffic emission model, VITO’s 115 locations throughout the city. The RIO- FASTRACE model is applied, relying on the FASTRACE-IFDM modelling chain has been COPERT methodology and starting from the operated for 2017 and successfully validated local road network with traffic intensities and against the results of these campaigns fleet information. An EU-wide traffic database highlighting the capabilities of the modelling for the EU-28 has been developed based on chain. COPERT mileage per country and downscaling using satellite-based proxy data. If available, ATMO-Plan is offered as a user-friendly local traffic data can be applied. The city of web-based air quality decision support tool, Krakow operated the VISUM traffic model offering an online interface for configuring operated and the Polish fleet information has the modelling of air quality scenarios and been enriched with local traffic counts. analysing the results. Users can import and export data for further processing offline and In the operational chain, these traffic creation of more detailed input data. The emissions are used in the bi-gaussian application can be used for many applications model IFDM model [1]. IFDM combines on such as screening of the effect of a low- an hourly resolution these contributions emission-zone, assessment of urban air from local traffic with the hourly background quality in high-resolution, impact of traffic concentrations taking into account a method reduction scenarios, environmental impact to avoid double counting. Optionally, point assessments and to identify which measures source emissions can be added as well. The sufficiently improve the air quality to improve tool has been pre-configured with ECMWF the local environmental quality and to meet Era5 meteo data for the EU-28, which can be EU air quality limits. replaced with more detailed local meteo if available. The model output is presented as Acknowledgements time series at points of interest and annual average concentration maps for NO , PM The validation study has been accomplished 2 10 as part of the LIFE Integrated Project and PM2.5. The tool can be applied for both a reference year and scenarios to assess 'Implementation of Air Quality Plan for the impact of possible traffic measures. Małopolska Region – Małopolska in a Additionally, the functionality is available to healthy atmosphere' (LIFE14IPE PL 021). The draw the boundaries of a low-emission-zone development of the web based air quality and adjust the fleet for exclusion of banned decision support tool ATMO-Plan is partly vehicles. funded by AirQast, a project that has received funding from the European Union’s Horizon As part of the LIFE IP project Małopolska 2020 research and innovation programme in a healthy atmosphere, the tool has been under grant agreement No 776361. configured, applied and validated for the city of Krakow in Poland. Krakow is located in the References Malopolska province, which forms together [1]: 'Evaluation of the RIO-IFDM-street canyon with the bordering province Silesia and the model chain'; W. Lefebvre, M. Van Poppel, cross-border regions in Czech Republic and B. Maiheu, S. Janssen, E. Dons; Atmospheric Slovakia an area with serious air quality Environment 2013 (77) 325–337

206 Figure 1: Screen shot of ATMO-Plan Krakow with NO2 annual average concentrations.

Assessing the value of regional In this work, an integrated assessment cooperation in air pollution control system composed of two different tools has been used (Figure 1): SHERPA (Thunis De Angelis E., Carnevale C., Turrini E., Volta M., et al., 2016) and RIAT+ (Carnevale et University of Brescia al., 2012). SHERPA, Screening for High Ferrari F., Maffeis G., TerrAria s.r.l. Emission Reduction Potential on Air tool, Guariso G., Politecnico di Milano provides different modules to support Pisoni E., European Commission, Joint Research policymakers in air quality management Centre (source apportionment, scenario assessment, In the past decades, major efforts have been governance); furthermore, it provides the made to tackle air pollution, but despite the input data for optimizing air quality plans improvements of air quality, most of the in any European region using RIAT+. RIAT+ (Regional Integrated Assessment Tool Plus), urban population is still exposed to PM2.5 concentrations that exceed the EU limit in turn, allows the definition of efficient air values and the stricter WHO guideline values. quality policies through a multi-objective According to the European Environmental or a cost-effectiveness approach including Agency (EEA), in 2015 premature deaths due the measure implementation costs in the decision process. In the current study, SHERPA to PM2.5 concentrations were 422000 in EU, 60600 in Italy. In Northern Italy, particularly is used to create for a specified domain: the in the Po Valley, high concentrations are a database of emission abatement measures common problem both in urban and rural (based on GAINS model database), a mapping areas. This area is characterized by high between emission classification CORINAIR population and emission density. Furthermore, SNAP3-fuel emission classification and GAINS the orography and the local meteorology specific sector-fuel classification, the emission (low wind speed, temperature inversion) inventory on the whole Europe, and the worsen the air stagnation. In the past source-receptor models needed to describe years, interregional cooperation plans were the relation between precursors emissions implemented to identify concerted actions to and various air quality indicators (AQIs), e.g., abate air pollution. the yearly average PM2.5 concentration.

207 For each geographical domain considered The integrated assessment methodology and for each value of the implementation briefly described here is used to analyze the costs, RIAT+ determines the best degree impact of cooperation between Northern of adoption of the end-of-pipe abatement Italian regions in comparison with individual measures, meaning the technologies that regional plans. Using as AQI the average reduce emission before being released yearly PM2.5 concentration, the optimization in the atmosphere without any energy problem was first solved on the Po Valley consumption modification, and the associated domain shown in Figure 2. implementation cost.

Figure 1. Integrated Assessment System composed by SHERPA (providing input data) and RIAT+ (implementing a multi-objective or cost-effectiveness optimization).

Figure 2. Case Study domain

208 The multi-objective optimization results are with the cooperative policy. In this case, the represented in the objective space, where investment does not change appreciably for the vertical axis refers to the AQI values Piedmont, Emilia Romagna, and Veneto, but is and the horizontal axis to the minimum higher in Lombardy where 2.2 M€/yr more are implementation cost to obtain such an necessary, as shown in Table 1. The impacts AQI. Costs are computed with respect to of a coordinated policy are clear in Lombardy, the reference scenario CLE2020 (Current especially because of its central position Legislation scenario), representing the in the Po Valley. In Figure 4, corresponding application of local, regional, national and emission reductions are presented for each European policy already in force in 2020. precursor and each CORINAIR macrosector Selecting for instance the optimal policies for the Po Valley policy (coordinated policy, in the point of maximum curvature of CP) and for the regional policy (RP) where the Pareto front, shown in Figure 3, the the emission reduction is the sum of each implementation cost is 67.7 M€/yr and an regional variation. Efficient measures should AQI reduction of 1.1 μg/m3 (7.8%) with be implemented mainly in agriculture, respect to CLE 2020 conditions. RIAT+ can road transport and domestic heating (non- now be used over the four separate regional industrial combustion plants) to reach the domains to understand what is the cost for expected impact. each region to autonomously work towards the PM2.5 concentration reduction obtained

Figure 3. Pareto front of the multi-objective optimization computed on the Po Valley optimization domain

209 Table 1. Cost over CLE2020 in M€/yr to autonomously obtain the same impacts on Air Quality resulting from a coordinated policy between regions

Cost of regional Cost of coordinated Optimization domain ΔAQI policy (RP) policy (CP)

Lombardy 8.8% 25.7 23.5 Emilia Romagna 6.9% 13.5 13.5

Piedmont 7.5% 15.2 15.2 Veneto 7.9% 15.5 15.5 Po Valley 7.8% 69.9 67.7

Figure 4. Emission reduction for each precursor and each macrosector for the coordinated policy (CP) and the regional policy (RP)

In this work, a Decision Support System effect of end-of-pipe measures common to that aims to help policy maker in the all regions. Even if, for political reasons, the implementation of efficient Air Quality adoption of different measures in different Plan is used in a critical European area to regions within the same concerted action study how convenient is an interregional seems difficult to accept, these results coordinated plan in comparison with actions emphasize the need of also examining the autonomously defined by each regional effect of non-technical and energy efficiency authority. SHERPA-RIAT+ have proved to be measures, which can make a further helpful to understand the priority areas, contribution to the decrease of pollutant where there is room for increased adoption concentration. of emission control measures, at different spatial scales. However, the range of possible References improvement of air quality turns out to be Carnevale, C. et al. (2012) ‘An integrated quite limited because many technology-based assessment tool to define effective air quality actions should be already in place by 2020 policies at regional scale’, Environmental and, in this study, we have only examined the Modelling & Software, 38, pp. 306–315. 210 Thunis, P. et al. (2016) ‘On the design and long-term, dynamic optimization model that assessment of regional air quality plans: The optimizes the investments in the power sector SHERPA approach’, Journal of Environmental and has partial coverage of heat sector via Management. Academic Press, 183, pp. combined heat and power plants (CHPs) and 952–958. heat pumps. The composition of electricity system is used as input to the Electric POwer Biomass in the European Dispatch (EPOD) model [2, 3]. This model Electricity System: Emission minimizes the operating cost on an hourly Targets and Investment basis for a selected period (usually 1 year), thus being able to investigate variations from Preferences wind and solar resources. The system models Lehtveer M., Department of Space, Earth use a comprehensive database as input and Environment, Chalmers University of to represent the existing electricity supply Technology infrastructure (power plants) [4] and hourly Fridahl M., The Centre for Climate Science wind and solar resources. The ELINEPOD and Policy Research (CSPR), Department of modelling package covers 27 EU member Thematic Studies, Environmental Change, states (EU-27), i.e. all but Croatia, as well Linköping University as Norway and Switzerland. For this study, Introduction the island states of Cyprus and Malta are excluded from the geographical scope, i.e. in Biomass is an important resource that can total, this study covers 27 countries. help decarbonisation in many sectors. If used in electricity generation, biomass can Since the amount and the cost of biomass complement variable renewables as well as that can be supplied sustainably and would provide negative emissions if coupled with be available to electricity system is highly carbon capture and storage technologies. uncertain, as demonstrated by Kluts et al. Yet, how these system services compete [5], we refrain from assuming the cost-supply or complement each other is not clearly curve of biomass for our model runs and understood. Furthermore, adoption of instead allow for unlimited biomass use at biomass in the electricity sector is also varied prices to illustrated the cost-effective pending policy incentives and acceptance use of biomass for the electricity system. levels. In this paper we investigate the cost- Based on previous research [6], we test three effectiveness of biomass in the European different price levels for biomass: 20, 50 and electricity system based on costs of 100 Euro/MWhth (megawatt hours of heat). biomass and emission requirements posed To estimate the effect of negative emissions on the electricity system as policy targets on the cost optimal allocation of biomass we and compare the results with investment run the model with three different emission preferences for biomass technologies in scenarios: reaching zero emissions by 2050, selected countries obtained via International meaning that no emissions from any part of Negotiation Survey (INS). the electricity system are allowed; reaching net-zero emissions by 2050, meaning that Method emissions in part of the electricity system can be offset by negative emissions in another, To evaluate the need for biomass in and; reaching net-negative (–10%) emissions European electricity system we use the compared to 1990 levels by 2050. Combining ELINEPOD modelling package. The Electricity these scenarios with biomass costs gives us Systems Investment Model (ELIN) originally thus nine different cases for the model. constructed by Odenberger and Unger [1], has previously been used to study the The INS survey data on investment transformation of the European electricity preferences in the energy supply sector were obtained through questionnaires system to meet the policy targets on CO2 emissions. The ELIN model is a bottom up, distributed at five negotiating sessions of

211 the UN Framework Convention on Climate respondents’ views of whether or not to Change (UNFCCC): the 42nd Subsidiary direct investments towards BECCS in order to Bodies meeting in Bonn, June 2015 (n = transform the electricity generation system 134); the 21st Conference of the Parties towards low-carbon configuration in their (COP) in Paris, December 2015 (n = 577); country of residence. For most countries, the COP22 in Marrakech, November 2016 (n = respondents neither agree nor disagree that 892); COP23 in Bonn, November 2017 (n such should be done, with a slight tendency = 944); and COP24 in Katowice, December to lean towards disagreeing. The same test 2018 (n = 996). In total, 3,543 responses provides evidence that country of origin has been collected of which 1,115 are influences respondents’ views of whether or from UNFCCC delegates residing in the 27 not to direct investments towards bioenergy European countries focused in this article. without CCS (p = .001). A Mann-Whitney The data used in this article builds on and U test also reveals that governmental extends previously used data on investment actors are generally more in favour of both preferences [7]. The extended number of bioenergy without CCS (p = .021) and BECCS responses allows for a more finely granulated (p = .000) compared to nongovernmental analysis, moving from global regional analysis actors across all the selected countries to look at European domestic levels. The represented in Figure 1. questionnaire was designed using a Likert- style response option format. References

Results 1. Odenberger, M., Pathways for the European electricity supply system to

The preliminary model results show a clear 2050 — Implications of stringent CO2 difference in biomass use among the cases. reductions., in Energy and Environment. In net-zero and negative emission runs 2009, Chalmers University of Technology. biomass is mainly used in biomass-fuelled steam power plants with carbon capture and 2. Goop, J., M. Odenberger, and F. Johnsson, storage (BECCS). Negative emissions are used Congestion patterns in the european (aside for meeting the emission requirement electricity transmission grid with high in –10% emissions case) to enable the use levels of solar generation. Submitted for of fossil power plants to utilise the existing publication, 2016. capacity and provide the flexibility to the 3. Göransson, L., et al., 'Linkages between system. BECCS plants are concentrated in few demand-side management and countries, providing negative emissions also congestion in the European electricity for other member states. In zero emission transmission system'. Energy, 2014. 69: p. case biomass is used in biogas fuelled power 860–872. plants and in combined heat and power plants (CHPs) that are more evenly distributed 4. Kjärstad, J. and F. Johnsson, 'The across Europe. One can note also that even European power plant infrastructure — when biomass cost is extremely high (100 Presentation of the Chalmers energy Euro/MWhth), it is still costeffective for the infrastructure database with applications'. system to use some amount of it instead of Energy Policy, 2007. 35(7): p. 3643–3664. other variation management options such as increased storage. Biomass is also competing 5. Kluts, I., et al., 'Sustainability constraints with nuclear power; high biomass price means in determining European bioenergy more competitive nuclear power plants and potential: A review of existing studies vice versa. and steps forward'. Renewable and Sustainable Energy Reviews, 2017. 69: p. Survey results show varying preferences for 719–734. bioenergy, but the general level of preference is rather low. A Kruskal-Wallis test provides 6. Johansson, V., M. Lehtveer, and L. no evidence that country of origin influences Göransson, 'Biomass in the electricity 212 Figure 1. Attitudes of UN climate change conference delegates, from the selected countries, towards directing investments in a long-term transition to low-carbon electricity generation to bioenergy without carbon capture and storage (BE w/o CCS) and BECCS.

system: A complement to variable for ensuring citizens with clean air ([3]). renewables or a source of negative Since the implementation of the Directive emissions?' Energy, 2019. 168: p. 532– 2008/50/EC, modelling can be used for some 541. assessments ([4]).

7. Fridahl, M., 'Socio-political prioritization Very high spatial resolution would be of bioenergy with carbon capture and necessary for accurate epidemiology storage'. Energy Policy, 2017. 104: p. exposure computations since there may 89–99. be substantial differences between the concentrations of air pollutants over the city: Easy-to-use modelling tool for hot spots may arise and stay at certain places urban air pollution with very high regularly resulting high exposure while the resolution HPC simulations overall (average) air quality indicators used in policies, are below policy thresholds. However, Horváth Z., Liszkai B., Kovács A., Budai T., Tóth there were obstacles to implementation of C., Széchenyi István University operational models of these features for policy and decision makers, in particular huge Motivation and aims computing power and difficult operation of Bad air quality in many cities results in 3 applications of the required supercomputers million premature deaths worldwide per year were mentioned as main obstacles. according to WHO reports ([1]) and more than Using the novel research results of high 70.000 deaths across the EU-28 countries performance computing (HPC), artificial ([2]). EC introduced assessment methods of intelligence (AI), cloud computing and exposure measurements and set up policies 213 mathematical technologies, the latter one for Traffic simulations of the demonstration city, constructing efficient algorithms - these all let Győr, Hungary are based upon the traffic us develop and implement a suitable solution monitoring sensor network. In the latter case to be presented below. the sensor network, to be implemented during the HiDALGO-project consists of a plate HiDALGO goals recognition camera system that monitors the whole traffic at the main junctions of HiDALGO is a Centre of Excellence for HPC the demonstration city, thus providing the and Big Data for global systems funded traffic simulation with traffic information by EC from Horizon 2020 ([5]). HiDALGO including origin-destination and trip data in advances HPC and AI technologies as tools on real time. The traffic monitoring system will supercomputers in order to improve data- be completed with affordable cost weather centric computation in general. HiDALGO and air quality sensors as well for validation runs the urban air pollution pilot as well purposes. Chemistry between the main which provides a user-friendly application as components of pollution, in particular NO , service that accurately and quickly forecasts 2 NO, O is taken into account. Emissions of the air pollution of cities with very high spatial 3 city infrastructure, in particular from heating resolution using HPC, supercomputers and of buildings and long range emissions are tools of the CoE. This pilot will be running modelled by CAMS data. operationally. Further, a traffic control system will be developed and tested to minimize air Post-processing computation of features of pollution while keeping traffic flow constraints pollutions, in particular hot spots, e.g. area in being satisfied at prescribed level. a cross section above certain concentration value, or computed local exposure values, Methods of HiDALGO Urban Air Pollution which might be components for new policy Pilot methods are provided. Visualization of 2D To achieve these goals, the pilot has (e.g. concentration maps in cross sections) developed an HPC-framework for simulating and of 3D (e.g. by virtual reality tools of the air flow in cities by taking into account HiDALGO) will be also presented. real 3D geographical information data of the Novel mathematical technologies for city, applying highly accurate computational modelling fluid dynamics (CFD) simulation on a highly resolved mesh (cca. 2 meter resolution at To reduce the huge amount of computing street level) and using weather forecast or resources for each analysis, novel re-analysis data from ECMWF as boundary mathematical technologies, namely model conditions. For CFD solvers both ANSYS order reduction for CFD is applied in HIDALGO. Fluent, leading commercial software and This is based on, e.g. proper orthogonal an open source framework have been used. decomposition and its variants which involve Dispersion of the emitted pollutants in the several offline HPC-simulations, snap-shot wind field is computed via strong coupling collection, data analysis of the snap-shots to the air flow computation. For uncertainty (e.g. clustering similar states into groups and quantification an ensemble model has been dimensional reduction in groups) and then the developed for the urban wind flow and the reduced model composition for the original pollutant information. variables. The reduced models are less computationally demanding while losing only The emission of the traffic, which is the most small part of the accuracy from the original significant producer of NO , one of the most 2 full complexity model. dangerous pollutant, is computed via using either the Copernicus Atmosphere Monitoring Web portal for user interface Service (CAMS) inventory data for emissions or via the Copert-4 model applied to traffic Users are served via a web based portal volumes got by traffic simulations with SUMO. where they specify the area and time period

214 of the assessment, source of data, select Matter (PM), is still a major problem, from the supported HPC infrastructure and especially in urban densely populated areas. choosing the parameters of solvers for In such areas, end-of-pipe measures are simulations. Then they launch the assessment usually not enough to reduce atmospheric by pushing a button. Status of the running pollutant concentrations to acceptable levels. is monitored from the dashboard. Post- Since secondary atmospheric pollutants, such processing is also run simply from the portal. as PM10, are generated through complex The prototype of the portal was developed in and non-linear processes of production, MSO4SC, another H2020 project and tailored accumulation and transport, environmental further by the subproject for industrial authorities need tools for building and mathematics of the University-Industry implementing air quality plans. Cooperation Centre of the Széchenyi István University. MAQ model [1] is an integrated assessment model designed to support decision makers, Demonstrations often in the need to select air quality control policies with economic constraints. The Full set of features of the assessment, methodology implemented can be interpreted in particular the traffic sensor network starting from the DPSIR scheme (Drivers- supported traffic simulation are ready for Pressures-State-Impacts-Responses), adopted Győr, the demonstration city of the project. by the EU [2]. In particular, MAQ solves an Further cities, e.g. Stuttgart are ready for the optimization problem by iteratively changing hands-on demonstration of the developed a set of abatement measures that directly service during the presentation or the reduces emissions (PRESSURES) or alters the conference. human activities (DRIVERS). This modifies air quality (STATE) resulting in the variation References of IMPACTS. The impacts are evaluated, and [1] WHO, air pollution. https://www.who.int/ RESPONSES are accordingly changed until airpollution/en/ efficient solutions are reached.

[2] EEA, Air quality in Europe - 2015 report MAQ solves a multi-objective decision (EEA Report No 5/2015). https://www.actu- problem that at the same time minimizes environnement.com/media/pdf/news-25756- n air quality indices (e.g. the yearly average rapport-air-aee.pdf PM10 or NO2 concentrations) and the total cost to reach such indices. The decision variables [3] Clean Air Policy Package. http://ec.europa. of the problem are the application rate of the eu/environment/air/clean_air/index.htm emission reduction measures. Two classes of measures are considered: end-of-pipe [4] Models for analysing climate change and measures, reducing the pollutant emissions air pollution policies. http://ec.europa.eu/ without changing the energy consumption of environment/air/publications/models.htm the anthropic activity and energy efficiency measures, reducing the level of fuel [5] HiDALGO – HPC and Big Data for Global consumption. Such measures also include Systems. https://hidalgo-project.eu/ energy-switch ones (the replacement of an activity by another one that is more effective MAQ, an integrated assessment from an energy consumption point of view), model to support air quality policy as well as behavioural measures (e.g. active at regional scale mobility strategies).

Turrini E., Carnevale C., De Angelis E., Volta M., The methodology has been applied over DIMI, University of Brescia Lombardy region, a densely populated and industrialized area, located in the Po basin Despite the growing political and scientific and subject to high PM pollution levels. concern arisen in recent years, Particulate

215 The baseline scenario includes the Current agriculture dominates the other ones; as this Legislation requirements for 2020 (CLE2020). sector is relatively unregulated with respect to the other two, there are a number of Four decision problems have been solved effective and low-cost measures that can be adopting the MAQ optimization approach: adopted.

• the identification of efficient abatement In terms of maximum potential PM10 mean measures considering, one by one, concentration (Table 1), Non-Industrial three emission sectors identified as combustion (c) could have the highest impact the main contributors to the regional but at very high costs (1046 M€/year). PM10 concentrations: (a) road traffic, (b) agriculture and (c) non-industrial The efficient solutions, when all the sectors combustion (mainly domestic and are considered (d), are represented in Figure commercial heating systems); 2. As expected, multi-sectorial policies have a higher impact with respect to the policies • the identification of efficient abatement identified considering specific emission measures considering all the emission sectors. Policy A is the solution of maximum sectors (d). curvature. This policy allows a reduction in population weighted PM10 yearly mean Figure 1 shows the Pareto curve representing, concentrations of 5.6 µg/m3 at a cost of 525 in the objective space (Costs Vs. PM10 M€/year. population weighted mean concentrations), the optimal solutions for decision problems Figure 3 shows the estimated costs and (a), (b) and (c). The solution considering emission reductions for Policy A. Costs are

Figure 1. Pareto curves for the three single macro-sector decision problems (a), (b) and (c)

Table 1. Maximum potential PM10 mean concentration reduction and costs of measures in one by one emission sectors

maximum potential PM10 emission sectors mean concentration reduction Cost [M€/year)] [ µg/m3] Road Traffic (a) 0.9 603 Agriculture (b) 1.6 102 Non-Industrial combustion (c) 2.3 1046

216 Figure 2. Pareto curve for the decision problem all the macro-sectors (d)

distributed in different sectors, but highest activities in Lombardy region (Italy). The investments are computed for 'Commercial analysis estimates the maximum potential and residential combustion plans' (2), reduction of different sectors and, assessing Agriculture (10) and 'Solvent use' (6). costs, proves and quantifies the efficiency of a plan that includes measures for all sectors, Premature mortality due to PM10 exposure emphasizing the substantial role of energy is estimated in terms of average per-capita measures. months of life lost [3]. Maps in Figure 4 show the health impact reduction of one by one References scenario, highlighting the effectiveness of Policy A. [1] Turrini, E. et al. (2018) 'A non-linear optimization programming model for air In conclusion, MAQ model is implemented to quality planning including co-benefits for GHG assess the relative effectiveness of air quality emissions', Science of The Total Environment. policies that consider one by one emission Elsevier, 621, pp. 980–989. sectors and a plan affecting all anthropogenic

Figure 3. End-of-pipe and energy measure costs (left); PM10 precursors emission reduction (right) aggregated for CORINAIR – SNAP1 macro-sectors, related to Policy A

217 Figure 4. Maps showing per-capita months of life lost over the study domain due to long- term PM10 exposure, related to the baseline scenario CLE2020, maximum potential of measures in one by one sector and Policy A

[2] EEA (1999) Technical report No 25/1999. of all economic sectors and households to use renewable energy, will entail an enormous [3] Bickel, P, Friedrich, R. (2005) ExternE. monetary investment, and a significant Externalities of Energy, Methodology 2005 increase in energy and raw materials demand. Update. Luxemburg-IER. Not adequately managed, those resources risk becoming bottlenecks to complete the Modelling the renewable transition. transition: strategies and policies for reducing energetic and raw The aim of this work is to evaluate the potential energy and raw materials costs materials costs of the transition, and to derive policy Samsó R., Solé J., Madurell T., García Olivares recommendations in order to minimise them. A., Instituto de Ciencias del Mar, Consejo To do so, we used the MEDEAS model at EU Superior de Investigaciones Científicas, Spain scale.

The European Union must decisively engage The MEDEAS models are open-source System in the renewable transition in order to Dynamics energy-economy-environment minimize its contribution to global climate Integrated Assessment models (IAM) that change. However, the structural changes are currently being developed within the required to replace fossil fuels by renewable framework of the MEDEAS project ('Modelling energy are not free of costs and impacts. the Renewable Energy Transition in Europe'). They are available at two geographical Indeed, building new infrastructure for the scales, EU and World, and are structured generation, transmission and storage of in 7 submodules: Economy, Energy, renewable energy, and the adaptation process Infrastructures, Materials, Land Use, Social

218 and Environmental Impacts Indicators and the model. In terms of energy, results from Climate Change. Being a child of the World the simulations with the TRANS scenario model (nested approach), the EU model indicate that the initial cost of the transition inherits the same characteristics as its will inevitably have to be covered with fossil parent, with the only exception that the EU fuels (mostly imported), with a subsequent version includes imports from the rest of the increase in greenhouse gas emissions across world. Europe. Despite an initial period of energy scarcity, if the investment in renewable The models bring innovative features energy sources is maintained over time, the such as the integration of Input-Output progressively larger installed capacity will be matrices in a System Dynamics model, the able to cover even larger proportions of the inclusion of supply-demand closures, the energy demand, reaching above 90% of the dynamic estimation of the Energy Return on share of the energy supply by 2050. Investment (EROI) of renewable energy (RE) technologies, the impact of climate change in Good planning and an adequate geographical the energy consumption and the estimation distribution of the different technologies will of potential scarcity of certain raw materials. be key to optimize the potential of each RE Most importantly, the MEDEAS model uses a technology, and to avoid the net energy cliff hybrid bottom-up/top-down approach, that that might occur if the EROI of the system allows for testing hypothesis from which becomes too small. policy recommendations can be derived. In terms of materials costs, only Indium First, a scenario named TRANS (for transition) and Tellurium might become a threat to the was designed. Scenarios in the MEDEAS completion of the transition at European level, context are a set of hypothesis regarding and substitutes of some of these materials the future evolution of the system that are will need to be found as soon as 2035. fed to the model at the beginning of every simulation. A minimum of 90% substitution Regarding the three hypotheses evaluated we of fossil fuels by renewable energies by 2050 conclude that: was the main prerequisite for the design • Reducing the size of the aviation sector of the TRANS scenario, and the selected by half results in savings of 4% of the hypotheses to achieve it were: stabilisation total final energy consumption of the EU. of the economy; large implementation rates of RES technologies; large increase in storage • To compensate for the scarcity of Indium capacity; enforcing electrification of all and Tellurium, recycling rates of such economic sectors combined with efficiency elements would have to increase between improvements; phase-out of oil for electricity 40 to 45% annually. and heat generation by 2060; and an afforestation program to capture Carbon from • Delaying the phase-out of oil for the atmosphere. electricity and heat generation does not prevent an episode of energy scarcity Based on that scenario, 3 simulation that arises around 2030, causing a minor experiments were performed in order to economic decline. evaluate the overall effects of a) cutting the aviation sector size by half, b) increasing the Based on these results, a set of policy recycling rates of materials, and c) delaying recommendations aimed at reducing energy the phase-out of oil for electricity and heat and materials costs are proposed. These generation. policies target the stabilisation of the economy, guaranteeing supplies of non- The energy and materials costs associated renewable fuels to make the transition, with such transition scenario, and to each implementing RE technologies with the of the three individual hypothesis, were highest EROIs, the electrification of all evaluated by looking at relevant outputs of economic sectors, downsizing the aviation 219 sector, the electrification and downsizing of constructed by Dr Frankenstein from body the household transport sector, improving parts of a variety of dead human-beings. It recycling rates and the construction of the turned out to be an ugly monster since, due high-voltage European interconnection. to this large variety, these parts did not fit together well. If Dr Frankenstein would have Finally, the assumptions made in this work been an economist, his monster would have allow to meet the EU target of reducing GHG looked much different (even uglier?) than emissions by 80% with respect to those if he were a climate scientist. Every model of 1990. Therefore, we conclude that the often uses similar concepts, like ‘equilibrium’, adoption of the TRANS scenario will allow but it soon turned out that we all had a the transition to be almost complete by different understanding. Every sub model 2050 without severely affecting people’s life originated from a different scientific culture standards nor depleting material resources, which caused Frankenstein's monster model and leaving a clear path to reaching zero to become rather ‘culture shocking’, hence emissions before the end of the century. heavily rejecting the other body parts. Every one of us was an expert on a sub model and On the sense of applying understood their own models into the last 'Frankenstein monster' models detail. There was nobody who understood the complete model. So, how could you use such Kremers H., ModlEcon a model for policy impact advise? My first encounter with quantitative models Computable General Equilibrium modeling for policy impact analysis was a three year can be seen as ‘theory with numbers’. In the contract in a project called ‘ECOBICE’ financed eyes of many ‘practitioners’ as policy makers by the ‘Volkswagen Stiftung’. The project and engineers often consider themselves, this intended to bring together various models often makes it suspect. Nevertheless, applying dealing with the impact of climate change Computable General Equilibrium modeling to into one big integrated assessment ‘economy- policy advise offers large benefits, if applied climatebiosphere’ model. The ECOBICE conscientiously. First of all, it functions as a integrated assessment model combines an common language between us theorists or economic model, a vegetation model, and a modellers, and the policy makers who are climate model, with a land use model, a water used to communicate in numbers. Secondly, model, a tourism model, and a population using a consistent theory to communicate model with their interlinkages into one your findings provides an elaborate set of ‘monster’ model. assumptions under which these findings are My first impression (as an economist) was valid. Thirdly, in order to compute the cost that such a ‘monster’ would not make any of inefficiencies in the economy, one needs sense. I nevertheless accepted the huge a benchmark which can be proven to be challenge this project offered. What better efficient. The Computable General Equilibrium way to quickly and efficiently obtain the model can be proven to fulfil the two Welfare necessary knowledge on all aspects of Theorems on efficiency introduced by Cournot, climate change impact? I started with the Walras and others. Hence, this model offers a responsibility for the economic sub model perfect benchmark. and, at the end of the project, integrating all The ‘ECOBICE’ experience provided me with a the modeling contributions into the overall good overview of what could go wrong with ‘ECOBICE’ model, see Kemfert et al. (2006). quantitative modeling and its application During the project, I often compared to policy impact analysis. The paper to this the ‘ECOBICE’ model to the monster of presentation presents this overview, and Frankenstein. According to the novel by provides my ideas on how we could possibly Mary Shelley (1869), this monster was solve these problems. I also address some

220 misconceptions among the various scientific Hawellek, J., C. Kemfert, and J.A.W.M. Kremers cultures involved. The paper can be seen (2003), 'Uncertainties of the costs of the as a qualitative introduction to a follow-up Kyoto Protocol − A meta-analysis', Working quantitative study on a meta-analysis of the Paper, University of Oldenburg, Oldenburg, influence of such modeling characteristics Germany. on policy impact. Hawellek and Kremers (2003) for example already performed Kemfert, C., W. Knorr, L. Criscuolo, K. such an analysis with respect to modeling Hasselmann, G. Hooss, H. Kremers, K. experiments of the US Stanford University Ronneberger, N. Kim-Phat, K. Tanaka, R. Tol, based Energy Modeling Forum (EMF) M. Zandersen (January, 2006), 'Bewertung regarding the impact of implementing the des Klimawandels anhand von gekoppelten Kyoto Protocol, see Weyant and Hill (1999). global vernetzten Ökonomie-Biosphäre-Klima Modellen ECOBICE (Economy-Biosphere- There is a theoretical and an applied Climate)', Endbericht Forschungsprojekt (quantitative policy analysis) side to gefördert von der Volkswagen Stiftung, integrated assessment modeling. The applied Programm Nachwuchsforderung in der side has many characteristics of engineering, fachübergreifenden Umweltforschung. and the theoretical side is completely left out. What is left looks like a sort of ‘engineering Rodrik, D. (2015). Economics Rules: The Rights tool’. But, the real challenge is to find the and Wrongs of the Dismal Science. W. W. correct equilibrium between theory and Norton & Company. engineering practices in order to come to a Weyant, J. and J. Hill (1999). The Costs of the conscientious policy impact analysis. We will Kyoto Protocol. The Energy Journal. Stanford be looking for this equilibrium. Within this Energy Modelling Forum, International context, I in particularly refer to Rodrik (2016), Association for Energy Economics. who tries to defend the social sciences and in particular the economic sciences against the Policy failure in the field of criticism following the financial crisis of 2009. electro-mobility During the presentation, I hope to bring this overview and my suggestions to the audience Gómez Vilchez J., European Commission, Joint Research Centren of policy makers and modellers, open for discussion and alternative proposals, and The global stock of electric cars exceeded 5 possibly for cooperation. My presentation million units in 2018, led by China with 2.3 intends to provoke the conference attendants million electric cars (EVI, 2019). In the same to think further what they are applying and year, there were over 1.2 million electric cars why. The ECOBICE experience was a good in use in the European Union (EU) (EAFO, one to learn about all aspects of climate 2019). In the passenger car market, electro- change, but was it worth the massive costs? mobility is speeding up. The key component Couldn't we have done the same with existing of electric cars continues to be the battery. small models? Is this really the best way to Despite declining lithium-ion battery costs in communicate results to the policy makers? recent years (BNEF, 2018), the purchase price Can we improve upon current modeling of an electric car remains higher than the practices? At the end of the conference, I hope price of a conventional car. For this reason, to bury the monster of Frankenstein! financial incentives are available in most EU Member States (ACEA, 2019). Policies References at the EU level such as the CO2 emission Shelley, M. (1869), Frankenstein, or the targets for new cars sold (EU, 2014b) and the modern Prometheus. Boston. deployment of alternative fuel infrastructure (EU, 2014a) are expected to promote electric

221 car market growth, thereby contributing PTTMAM models the decisions of four main towards the Paris Agreement (EU, 2017). agent groups involved in the EU electro- Notwithstanding this, the future market mobility system as well as their interactions: uptake of this powertrain technology remains users, vehicle manufacturers, infrastructure uncertain. providers and authorities. The model is documented by Harrison et al. (2016a) and To understand the factors influencing car has been applied by Pasaoglu et al. (2016), powertrain choice, including electric cars, the Harrison and Thiel (2017a), Harrison and Thiel JRC conducted a stated preference survey in (2017b) and Harrison et al. (2018). mid-2017. Over 1,200 car owners in France, Germany, Italy, Poland, Spain and the United The TE3 model examines energy demand Kingdom answered the questionnaire (for and greenhouse gas emissions from nine details, see Gómez Vilchez et al. (2017)). As a car powertrains, including electric cars, in result, a discrete choice model was estimated, four major non-EU car markets: China, India, where driving range and recharging time were Japan and the United States (US). The model identified, in addition to the purchase price, is documented by Gómez Vilchez (2019) as statistically significant factors influencing and has so far been applied by e.g. Haasz choice (Rohr et al., 2019). et al. (2018). For the purpose of this work, a sub-model that represents the Chinese bus The need to consider explicitly the market was added. interactions of multiple factors in a complex system such as this calls for the adoption Figure 1 shows the default (i.e. ‘no linkage’) of a modelling strategy that facilitates the evolution of the battery price, as simulated representation of the ‘endogenous point in each model. The following assumptions of view’ (see Richardson (2011)). This is underlie this figure: the battery capacity best captured by a simulation model that assumed for PTTMAM is 30 kWh, the uses feedback structures (Forrester, 1961) exchange rate is 1.2 dollars per euro and a (Richardson, 1999). Two classic examples of 10% mark-up is assumed to translate cost feedback processes present in the field of into price. As can be seen, the simulated electro-mobility are the ‘chicken-and-egg’ battery prices differ significantly between problems related to high battery prices and 2013 and 2022. Whereas the PTTMAM insufficient infrastructure availability. The simulation has a relatively good fit with the system’s behaviour over time that arises historical data over the period 2012-2014, the from such feedback structures is relevant empirical evidence suggests a better match for model-based policy analysis. Of interest with the TE3 simulation in 2015 and 2016. to the policy analyst is the identification of leverage points (see Forrester (1971)). In addition, Figure 1 shows the results after linking the modules (i.e. ‘modal linkage’ The objective of this paper is to illustrate a curve). In this case, the simulated battery potential policy failure in the field of electro- price exhibits a more plausible behaviour. mobility by means of a modelling example, Finally, future battery prices are shown with a focus on the electric car battery price. in the figure. The 2020, 2025 and 2030 For this purpose, a model linkage between trajectory is taken from the ‘middle scenario’ the JRC in-house Powertrain Technology reported by JRC (2018). The simulation after Transition Market Agent Model (PTTMAM)1 model linkage seems to be in line with that and the Transport, Energy, Economics, trajectory. Environment (TE3)2 model was implemented. Both simulation models are grounded on Policy failure may be defined in this paper system dynamics. as the ineffectiveness of policy measures for meeting on time a certain target that has been set to overcome a given problem. 1 Available under the EUPL at: https://ec.europa.eu/jrc/en/ptt- mam As an example, the case of Germany can be 2 Available at: http://www.te3modelling.eu/

222 Figure 1. Evolution of the electric car battery price Sources: data from BNEF (2017), scenario by JRC (2018) and own simulations

mentioned: it set the target of deploying one negative feedback processes. In the particular million electric vehicles (EV) by 2020 and has field of electro-mobility modelling, we offered purchase incentives since 2016 (BMU, suggest the complementary use of simulation 2019). Despite this, the one million EV target tools that model the actions of key non-EU is unlikely to be met on time (see e.g. Thiel et car markets such as China or the synergetic al. (2019)). effects of policy measures, so that policy failure in the field of electro-mobility can be Model-based policy analysis based on the avoided. isolated output of any of the two models would have been misleading. Once major This paper was limited to a single example markets influencing cumulative battery and other potential sources of policy failure manufacturing experience (e.g. bus market, remained unexplored. Relevant to the context China and the US) are explicitly taken into of the European Battery Alliance (EBA, 2019), account, a more realistic behaviour of a supply-side constraints may jeopardise crucial model variable could be determined. electric car market uptake (Gómez Vilchez, Thus, this may provide a sounder basis for 2018). Further research on the representation choosing a level of purchase subsidy for a of this and additional feedback processes new electric car. in the existing models is needed. Finally, these models can be coupled with the JRC We conclude that models that do not in-house DIONE fleet impact model (Krause explicitly model feedback loops are likely to et al., 2017) to more accurately estimate miss important mechanisms that are part of vehicle energy demand and emissions, thus the electro-mobility system. As a result, the building on the previous work by Harrison et side effects or unanticipated consequences al. (2016b). of well-intentioned policies may take policy- makers by surprise (see Sterman (2000) for References a discussion). Therefore, our general policy recommendation is to use nonlinear models ACEA. (2019). Electric vehicles: Tax benefits that deal with the interplay of positive and & incentives in the EU. European Automobile 223 Manufacturers Association (ACEA). Retrieved THE COMMITTEE OF THE REGIONS Delivering from https://www.acea.be/uploads/ on low-emission mobility. A European Union publications/Electric_vehicles-Tax_benefits_ that protects the planet, empowers its incentives_in_the_EU-2019.pdf consumers a. Retrieved December 4, 2017, from http://eur-lex.europa.eu/legal-content/ BMU. (2019). Maßnahmenpaket der EN/TXT/?uri=CELEX:52017DC0675 Bundesregierung. Regierungsprogramm Elektromobilität. Bundes-Ministerium für EVI. (2019). EV Global Outlook 2019. Electric Umwelt, Natur-Schutz und nukleare Sicherheit Vehicles Initiative (EVI). OECD/IEA. (BMU). Retrieved from https://www.bmu. de/themen/luft-laerm-verkehr/verkehr/ Forrester, J. W. (1961). Industrial Dynamics. elektromobilitaet/bmu-foerderprogramm/ Massachusetts Institute of Technology massnahmenpaket-der-bundesregierung/ Press. Retrieved from https://books.google.it/ books?id=4CgzAAAAMAAJ BNEF. (2017). Lithium-ion Battery Costs and Market. Bloomberg New Energy Finance Forrester, J. W. (1971). 'Counterintuitive (BNEF). behavior of social systems'. Technological Forecasting and Social Change, 3, 1–22. BNEF. (2018). Electric Vehicle Outlook 2018. https://doi.org/https://doi.org/10.1016/S0040- Bloomberg New Energy Finance (BNEF). 1625(71)80001-X

EAFO. (2019). European Alternative Fuels Gómez Vilchez, J. (2019). The Impact of Observatory (EAFO). European Commission Electric Cars on Oil Demand and Greenhouse (EC). Retrieved from https://www.eafo.eu/ Gas Emissions in Key Markets. Ph.D. Thesis. KIT Scientific Publishing, Karlsruhe. EBA. (2019). Building a European battery Retrieved from https://www.ksp.kit. industry. European Battery Alliance (EBA). edu/9783731509141 Retrieved from https://www.eba250.com/ Gómez Vilchez, J. (2018). 'Exploring EU. (2014a). Directive 2014/94/EU of the the battery market for electric cars'. In European Parliament and of the Council Proceedings of the 36th International of 22 October 2014 on the deployment of Conference of the System Dynamics Society. alternative fuels infrastructure. European Union Law. Retrieved May 22, 2017, from Gómez Vilchez, J., Harrison, G., Kelleher, L., http://eur-lex.europa.eu/legal-content/EN/ Smyth, A., & Thiel, C. (2017). Quantifying TXT/?uri=celex%3A32014L0094 the factors influencing people’s car type choices in Europe: Results of a stated EU. (2014b). Regulation (EU) No 333/2014 of preference survey. Retrieved from http:// the European Parliament and of the Council publications.jrc.ec.europa.eu/repository/ of 11 March 2014 amending Regulation handle/111111111/50291 (EC) No 443/2009 to define the modalities for reaching the 2020 target to reduce Haasz, T., Vilchez, J. J. G., Kunze, R., Deane, P.,

CO2 emissions from new passenger cars. Fraboulet, D., Fahl, U., & Mulholland, E. (2018). European Union Law. Retrieved May 15, 2017, 'Perspectives on decarbonizing the transport fromhttps://eur-lex.europa.eu/legal-content/ sector in the EU-28'. Energy Strategy Reviews. EN/ https://doi.org/10.1016/j.esr.2017.12.007

EU. (2017). COMMUNICATION FROM Harrison, G., & Thiel, C. (2017a). 'An THE COMMISSION TO THE EUROPEAN exploratory policy analysis of electric PARLIAMENT, THE COUNCIL, THE EUROPEAN vehicle sales competition and sensitivity ECONOMIC AND SOCIAL COMMITTEE AND to infrastructure in Europe'. Technological

224 Forecasting and Social Change, 114, 165–178. Pasaoglu, G., Harrison, G., Jones, L., Hill, A., https://doi.org/10.1016/j.techfore.2016.08.007 Beaudet, A., & Thiel, C. (2016). 'A system dynamics based market agent model Harrison, G., & Thiel, C. (2017b). 'Policy simulating future powertrain technology insights and modelling challenges: The case transition: Scenarios in the EU light duty of passenger car powertrain technology vehicle road transport sector'. Technological transition in the European Union'. European Forecasting and Social Change, 104, 133–146. Transport Research Review, 9(3), 37. https:// https://doi.org/10.1016/j.techfore.2015.11.028 doi.org/10.1007/s12544-017-0252-x Richardson, G. P. (1999). Feedback Thought in Harrison, G., Thiel, C., & Jones, L. (2016a). Social Science and Systems Theory. Pegasus Powertrain Technology Transition Market Agent Communications. Retrieved from https:// Model (PTTMAM): An Introduction. Retrieved books.google.it/books?id=nyJAPgAACAAJ from http://publications.jrc.ec.europa.eu/ repository/handle/111111111/40434 Richardson, G. P. (2011). 'Reflections on the foundations of system dynamics'. System Harrison, G., Krause, J., & Thiel, C. (2016b). Dynamics Review, 27(3), 219–243. https://doi. 'Transitions and Impacts of Passenger Car org/10.1002/sdr.462 Powertrain Technologies in European Member States'. Transportation Research Procedia, Rohr, C., Lu, H., Smyth, A., Kelleher, L., Gómez 14(Supplement C), 2620–2629. https://doi. Vilchez, J.J., Thiel, C., Harrison, G. (2019). org/10.1016/j.trpro.2016.05.418 'Using Stated choice experiments to quantify the impact of vehicle characteristics that Harrison, G., Vilchez, J. J. G., & Thiel, C. influence European’s propensity to purchase (2018). 'Industry strategies for the promotion electric vehicles'. In TRB Annual Meeting of E-mobility under alternative policy and Online. Transportation Research Board economic scenarios'. European Transport (TRB) 98th Annual Meeting, January 13–17, Research Review. https://doi.org/10.1186/ Washington D.C. Retrieved from http:// s12544-018-0296-6 amonline.trb.org/68387-trb-1.4353651/ t0015-1.4368127/1694-1.4368253/19- I. Tsiropoulos, D. Tarvydas, N. L. (2018). 01325-1.4364644/19-01325- Li-ion batteries for mobility and stationary 1.4368274?qr=1 storage applications - Scenarios for costs and market growth. EUR - Scientific and Technical Sterman, J. (2000). Business Dynamics: Research Reports. Joint Research Centre Systems Thinking and Modeling for (JRC). Retrieved from http://publications. a Complex World. Irwin/McGraw-Hill. jrc.ec.europa.eu/repository/bitstream/ Retrieved from https://books.google.it/ JRC113360/kjna29440enn.pdf books?id=CCKCQgAACAAJ

Krause, J., Donati, A. V., & Thiel, C. (2017). Thiel, C., Julea, A., Acosta Iborra, B., De Light Duty Vehicle CO2 Emission Reduction Miguel Echevarria, N., Peduzzi, E., Pisoni, Cost Curves and Cost Assessment - the DIONE E., Gómez Vilchez, J.J., Krause, J. (2019). Model. JRC Science for Policy Report. Joint 'Assessing the Impacts of Electric Vehicle Research Centre (JRC), European Commission. Recharging Infrastructure Deployment Efforts Retrieved from http://publications.jrc. in the European Union'. Energies . https://doi. ec.europa.eu/repository/bitstream/ org/10.3390/en12122409 JRC108725/kjna28821enn.pdf

225 Sustainable Transport Unit, following the The CO2MPAS vehicle simulation model and its role in the European request of DG Climate Action, and is currently used for the type-approval of new vehicles in light-duty vehicle CO certification 2 Europe. process

The development specifications of CO2MPAS Pavlovic J., Arcidiacono V., Maineri L., were challenging, as it had to be highly Anagnostopoulos K., Komnos D., Saporiti F., accurate, exhibit fast operation, and function Corsi S., Valverde V., Ciuffo B., Fontaras G., European Commission, Joint Research Centre with a limited number of input data. The Lindvall S., Directorate-General for Climate core of CO2MPAS is a backward-looking, Action longitudinal dynamics physical model simulating energy flow and losses at various Light-duty vehicles (passenger cars and light components. Investigations indicated that commercial vehicles) produce a significant the four most important factors affecting share of the European Union's (EU) total CO2 emissions for WLTP and NEDC are, carbon dioxide emissions (CO2), the main energy demand at the driveline, gear- Greenhouse Gas (GHG). The EU has so far shifting strategy for automatic transmission used a series of policy instruments to curb vehicles, hot and cold start engine fuel

CO2, such as voluntary agreements and consumption, and the operation of specific fleet-wide CO2 targets. Current fleet-wide, fuel-saving technologies. In order to maximize sales weighted, CO2 targets (95 g/km for cars accuracy and in lack of detailed input data in 2021 and 147 g/km for light commercial CO2MPAS has an integrated self-calibration vehicles in 2020) are based on the New functionality. European Driving Cycle (NEDC) originally established in the early seventies. The JRC CO2MPAS achieves low errors in the and other stakeholders have signalled already prediction of the NEDC cycle that in the since the last decade the need for a new controlled sample used for its development up-to-date test protocol. As of July 2017, were of the order of 1% with a standard the emissions type-approval of light-duty deviation of 3%. In the period from 09/2017 vehicles in Europe is based on the Worldwide to 03/2019 CO2MPAS has been used for Harmonized Light-duty vehicles Test official type-approval of 2882 vehicle Procedure (WLTP). WLTP was introduced to families (2882 vehicle 'high' and 2442 replace the old and outdated NEDC. However vehicle 'low' configurations as defined by the the introduction of the new test protocol regulation). The average CO2MPAS error is would not be possible if the entire regulatory 2.7% for vehicle high and 2.6% for vehicle framework governing vehicle emissions would low including possible biases introduced by have to change, and most prominently the the manufacturers effort to optimize their emissions towards lower values (increasing pre-existing NEDC-based CO2 targets. In order to allow sufficient lead time to vehicle the error). More importantly, the percentage manufacturers and national authorities to of vehicles with an error lower than 4%, adapt to the new procedure, and also to the acceptance limit set in the regulation, is save them from the burden of double testing 69% for the vehicle configuration high and the vehicles over the two protocols (NEDC 68% for the low configuration (Figure 1). & WLTP) in parallel, a simulation-based In other words, for about 2/3 of the newly certified vehicles the manufacturer’s declared approach was chosen to calculate the CO2 emissions and fuel consumption of vehicles value was validated with CO2MPAS, avoiding over the NEDC in the period 2017–2021, additional experimental testing. This is an based on the WLTP test results. A dedicated important achievement as it reduces the costs and time necessary to certify light-duty vehicle simulation model (CO2MPAS) was developed for the purpose by the JRC’s vehicle CO2 emissions during the transitional period.

226 Further to the above, together with the Regarding future activity and developments,

CO2MPAS tool the JRC introduced the DICE3 given the expected shift towards hybrid and server, a new channel that is used for electric powertrains in the years to come, communicating type approval information the development team is working on the to the JRC, in order to monitor both the necessary adaptations for covering also performance of the CO2MPAS tool and the electric powertrains and conventional vehicle implementation of the respective regulation. operation over broader operating conditions.

The data produced by CO2MPAS and received The authors are investigating the possibility by the DICE3 have been used to support of calibrating CO2MPAS using real-world DG Climate Action for various monitoring measurement data in order to extend the activities and for ensuring the integrity of the tool’s applicability to fuel consumption implementation of the new regulation. prediction over real-world operation for consumer information and research purposes.

Figure 1. Distribution of in-use CO2MPAS error

227 Notes

228 Notes

229 Notes

230 Notes

231 © European Union, 2019

232