The Role of Natural Gas in a Low-Carbon Europe: Infrastructure and Supply Security

In this paper, we analyse infrastructure needs of the European natural gas market in response to decarbonisation of the European energy system. To this end, we use numerical modelling and apply the Global Gas Model. We investigate three pathways of future natural gas consumption: i) a decreasing natural gas consumption, following the scenarios of the EU Energy Roadmap 2050; ii) a moderate increase of natural gas consumption, along the lines of the IEA’s New Policies Scenario; and iii) a temporary increase of natural gas use as a “bridge” technology, followed by a strong decrease after 2030. Our results show that current import infrastructure and intra-European transit capacity are sufﬁcient to accommodate future import needs in all scenarios. This is despite a pronounced reduction of domestic production and a strong increase in import dependency. However, due to strong demand in Asia, Europe must increasingly rely on exports from Africa and the Caspian region, leading to new infrastructure capacity from these regions. When natural gas serves as a “bridge” technology, short-term utilisation rates of LNG import capacity temporarily increase instead of instigating large scale pipeline expansions.


INTRODUCTION
The role that natural gas will play in the transition to a decarbonised European energy system is unclear.There is a broad range of perspectives on natural gas in the future European energy mix: natural gas could play the role of a "bridge fuel" during a transition phase, or serve as the main "backup fuel" for intermittent renewable power generation.However, natural gas could also be steadily phased out and substituted for by non-fossil fuel alternatives, which will quickly become economic under stringent climate policies. 1While the European Energy Roadmap to 2050 proposes development in the latter direction with natural gas consumption declining over the next few decades (EC, 2011a), the International Energy Agency sees a consistently large role for natural gas in Europe in the coming decades in its New Policies Scenario (IEA, 2015).
In the first part of this paper, we base our analysis on two scenarios of the EU Energy Roadmap 2050 (EC, 2011a).This Roadmap triggered a vivid academic debate and motivated the Energy Modeling Forum (EMF) to devote a round of model comparisons of European decarbonisation scenarios: EMF 28 on "the effect of technology choices on EU climate policy" (cf.Knopf  et al., 2013). 2 We use data from both the EU Energy Roadmap and EMF 28 as a starting point for our sectoral analysis, where we focus on the impact of climate policy on the European natural gas infrastructure.
To this end, we construct four scenarios.Our first scenario (Reference) complies with the EU 2020 targets. 3It is, hence, a moderate climate scenario.The second scenario (HRes) is a more stringent climate scenario with high shares of renewables in line with the EU 2030 targets. 4Both scenarios foresee a decreasing importance for natural gas in the European energy system (pathway i above).We define two alternative scenarios to investigate other possible developments of the European natural gas sector (pathways ii and iii): The first alternative scenario (Backup) allows us to investigate infrastructure needs in an environment of increasing natural gas consumption.The second alternative scenario (Bridge) focuses on natural gas as a transitional fuel towards a decarbonised European energy system.
Our main results suggest that the pipeline and LNG capacities already in place or currently under construction are sufficient to accommodate future European demand for natural gas in all scenarios.This holds particularly for scenarios with declining natural gas consumption.However, allowing for a more diverse natural gas supply, and taking into account competition with Asia for Russian natural gas, new connections are advisable.In particular, pipeline connections from Africa and the Caspian region towards Central Europe could be significantly expanded.Moreover, within Europe, there is need for small but important infrastructure investments for improved interconnection between regions (e.g., between the Iberian Peninsula and the rest of Western Europe) and for reverse flows (West-East direction).These small additional capacities do not only serve to import additional volumes, they also considerably improve supply security by diversifying trade flows.An increasing natural gas consumption (Back-Up Scenario) is characterised by the most significant pipeline expansions.In contrast, the Bridge scenario suggests lower investments in pipelines but higher expansions of LNG import facilities to accommodate the temporary increase in consumption.
The remainder of this paper is organised as follows.In Section 2, the GGM and its dataset are presented.In Section 3, we discuss results for decreasing European natural gas consumption (pathway i).The two alternative scenarios, pathways ii and iii, are discussed in Section 4 and Section 5 concludes.5.The mathematical formulation of the GGM can be found in Appendix A. 6.The region "EU" consists of all EU-27 member states except for Cyprus and Malta.Note that Croatia only acceded to the European Union after our base year 2010 and is excluded from the EU, when we discuss aggregate results in the following sections.Furthermore, our dataset separately includes ten East European countries as well as Russia, Turkey, Norway, and Switzerland.All other countries are combined in the six regions Africa (AFR), Asia-Pacific (ASP), the Caspian Region (CAS), the Middle East (MEA), North America (NAM), and South America (SAM) to represent all production, consumption, and trade in the global natural gas market.See Appendix B for a list of country and regional abbreviations.

THE GLOBAL GAS MODEL
The Global Gas Model (GGM) is a partial equilibrium model of the natural gas market.It numerically simulates regional natural gas production, consumption, and patterns of international trade. 5The model is set up with a high level of detail featuring demand seasonality, market power exertion vis-a `-vis final consumers, as well as endogenous investments in storage and transport capacity, both of pipelines and along the LNG supply chain.While Egging (2013) presents a stochastic version of the model, in this paper, we use a deterministic version with a particular focus on, and a more detailed representation of, Europe.Twenty-five of the EU member states are incorporated individually among the global total of 45 regional nodes. 6

Model Description
The GGM represents all important market agents along the natural gas value chain.These comprise producers, traders, storage system operators (SSO), and transmission system operators (TSO), while final consumption is represented by aggregate inverse demand functions.All agents operate rationally under complete information and maximise the sum of discounted profits over the entire model horizon under operational constraints (such as production capacity limits) and technical and infrastructure restrictions (such as pipeline capacities and loss rates).All agents are price takers, except for selected traders who can exert market power vis-a `-vis final consumers by taking into account the effect of their supplied quantities on market prices.This accounts for the oligopolistic market structure of the European natural gas sector. Figure 1 illustrates the supply chain structure incorporated in the GGM for two trading countries, highlighting the interaction between the different market agents.In this stylised example, Country 1 produces, consumes, and exports natural gas to Country 2, which, by contrast, does not produce natural gas domestically.
Producers maximise discounted profits of selling natural gas to assigned traders, bearing the costs of extraction.These costs are defined as proposed by Golombek et al. (1995) by means of logarithmic functions that consist of a constant per unit term, a linearly increasing term, and a third term inducing a steep cost increase whenever production approaches the maximum capacity.Huppmann (2013) provides a comprehensive discussion of the properties of a Golombek production function.Production cannot exceed producer-and time-specific capacities.
Traders, in turn, generate revenues by selling natural gas to final consumers, while bearing the costs of purchasing natural gas from the assigned producer, as well as of renting storage and transportation services.There is a one-to-one mapping between traders and producers at the country (or region) level.A trader can, hence, be interpreted as the sales arm, and a producer, in turn, as the extraction arm of a supplying company.The trader, in our setup, is the agent that can exert market power on final consumers.
The TSOs manage the transportation network and rent out capacity to traders.This can be capacity of pipelines, liquefaction, shipping, or regasification infrastructure.Along a particular LNG supply chain, one TSO is responsible for the liquefaction terminal, one TSO manages the shipping route to the destined import country, and a third TSO is in charge of the regasification terminal.The TSOs maximise profits from congestion rents on transportation capacities minus incurred investment costs. 7Similarly, SSOs manage the storage facilities that traders can rent to arbitrage between seasonal price variations.To relieve the pressure of infrastructure bottlenecks, the TSO and SSO can endogenously invest in additional transportation and storage capacities respectively, and will do so whenever profitable.
Furthermore, marketers at each consumption node serve to balance natural gas supply with the combined demand of three different sectors (residential/commercial, industrial, and power generation).Consumption from the different sectors is represented by one aggregate inverse demand function for each country node. 8his "multi-agent economic game on an underlying transportation network" (Egging,  2013) is formulated as a mixed complementarity problem (cf.Facchinei and Pang, 2003), programmed in GAMS (Brooke et al., 2008) and solved using the PATH solver (Ferris and Munson,  2000).The model is fully parameterised regarding production capacities and costs, reference prices and consumption levels, transportation and storage capacities, and costs and losses.The base year is 2010 and the reporting horizon is 2050, including every fifth year in between. 9

Data
For each node, reference demand, production, and price levels, as well as production costs and capacities are based on assumptions and results of various energy system models, namely BP 10.See Knopf et al. (2013) for a comparison of EPPA results to other models in the model comparison EMF28 "Technology scenarios for transforming the European energy system".EPPA is the only top-down model in EMF28 that provided detailed data for the natural gas sector.
(2011), the World Energy Outlook (IEA, 2012), and EMF 28 modelling results obtained from the EPPA model (Paltsev et al., 2005). 10For EU member states, we vary reference consumption and production levels across scenarios as discussed in Sections 3 and 4. For instance, in our Reference scenario we use results of the energy system model PRIMES that are the basis for the EU Energy Roadmap 2050 and were provided for EMF 28 (cf.Knopf et al., 2013).PRIMES results reflect the optimal choice between different technologies, given an EU-wide GHG emission ceiling.
Node-and time-specific inverse demand curves are based on reference consumption levels and data on sector shares in the year 2010.In order to construct the aggregate demand curve, we combine sector-specific consumption and price elasticities.The sector-specific elasticities are assumed as follows: for the residential sector -0.25, the industrial sector -0.4, and for the power sector -0.75 (c.f., van Oostvoorn et al., 2003).Hence, the price elasticity of the aggregate inverse demand function at the reference price-quantity equilibrium depends on the sector shares of natural gas consumption.Consequently, a consumption node with high natural gas usage in power generation is characterised by a higher price elasticity at the reference point.However, because we approximate demand by linear functions, aggregate price elasticities are not fixed.
Production capacities, in turn, are set exogenously at a level that is higher by 1.5% to 15% than projected production levels, to account for a certain degree of slack capacity.Initial pipeline capacities within Europe are to a large extent based on GTE (2011).Initial liquefaction and regasification capacities are taken from GIIGNL (2011).We account for infrastructure projects that are currently under construction or planned and include these expansions in the model periods 2015 and 2020 exogenously.We limit endogenous investments in infrastructure-both in pipelines and along the LNG chain-in the first two periods of the model run.After 2020, however, investments in transportation capacity are unrestricted in order to allow for a future network configuration predetermined as little as possible by modellers' restrictions.Transportation losses and costs for the pipeline and the LNG technology are linearly increasing in the distance between the origin and destination of the natural gas flows.Similarly, investment costs for a particular pipeline depend on the length of the onshore and offshore segments of that pipeline.All costs are inflated by 2.75% annually, while the discount rate for future cash flows is 10%.
Our analyses of the European natural gas sector in response to a European decarbonisation takes place in the context of global consumption and production of natural gas growing by 67% between 2010 and 2050.In regions outside the EU, consumption is projected to rise in all world regions, particularly in the Asia-Pacific region, projected to see the largest growth of natural gas consumption with a doubling by 2025 and an almost three-fold increase until 2050.In most regions, natural gas production is projected to develop roughly along the same trajectories as consumption.This holds for North and South America, the Middle East, and Africa.However, for Russia and especially the Caspian region, the increase in production is projected to be significantly larger than the rise in consumption.Russian production increases by about two thirds until 2035, while production in the Caspian region doubles in the following decades until 2050.The exports allowed by these increased production levels are used to meet the import needs of Europe and the Asia Pacific (Holz et al., 2015).

NATURAL GAS IN THE EU ENERGY ROADMAP 2050
This section takes a closer look at two scenarios of the EU Energy Roadmap 2050 (EC, 2011a) with a focus on natural gas trade flows and infrastructure expansions.For both scenarios, reference demand and production levels for European countries are based on PRIMES (see Section 2).
Our first scenario, the Reference scenario, is defined by a moderate climate policy.It incorporates the two binding EU 2020 targets of a 20% GHG emissions reduction relative to 1990levels and a 20% share of renewable energy in final consumption.GHG emissions in 2050 are 40% below the level of 1990.Both CCS and nuclear power are options to decarbonize the future energy mix.Energy efficiency and renewable energy supplies increase according to current learning-curves.
Our second scenario, the HRes scenario, is defined by a more ambitious EU climate policy that incorporates the EU 2030 targets of a 40% GHG emissions reduction relative to 1990 and a 30% share of renewable energy in final consumption.Furthermore, GHG emissions are reduced by 80% in 2050.While the set of available technologies is not restricted, renewable energies play a larger role than in the Reference scenario.This second scenario is based on the "High RES" scenario of the EU Energy Roadmap 2050 (EC, 2011a).

Reference Scenario: 40% GHG Emissions Reduction until 2050
The Reference scenario projects a steady decline in EU natural gas consumption.Figure 2 shows that EU natural gas consumption decreases to below 400 bcm/y as of 2040, and decreases further to less than 80% of the 2010-level in 2050. 11The reduction in the EU is spread unevenly across its member states.The strongest decrease in natural gas consumption between 2010 and 2050 takes place in the Netherlands ( -41%), the UK ( -35%), France ( -35%), and Germany ( -27%).By contrast, some countries increase their natural gas consumption in a shift away from coal, e.g., Greece ( + 126%), Spain ( + 30%), and Bulgaria ( + 26%).Accordingly, Spain becomes a more important natural gas consumer with a higher consumption level by 2025 than France and the Netherlands.12.Note that these EU shares are import dependencies for the 25 EU Member States that are included in the dataset.This holds for all results shown for the aggregate EU in the paper.As discussed above, Croatia is assigned to "South East Europe"; Cyprus and Malta are not included.
Production of natural gas in the EU decreases even more strongly than consumption.By 2030, only 44% of the 2010 production level is reached, and only 20% by 2050.Except for the Netherlands and Romania, all producing countries stop extracting natural gas by 2040 at the latest.With domestic production decreasing faster than consumption, import dependency increases.While in 2010, 64% of EU natural gas consumption was covered by imports, this goes up to 82% by 2030 and above 90% by 2050 (see Figure 2). 12Net imports increase over time, by 10% until 2030 and by 13% to about 350 bcm/y until 2050 compared to 2010 levels.
Underlying these increasing EU imports, large heterogeneity can be observed among countries (see Figure 3).In particular, those countries with increasing demand (e.g., Spain, Greece, Belgium, and Bulgaria) experience a significantly higher than average increase in net imports.Furthermore, countries such as the UK with initially large domestic production become more and more dependent on imports ( + 36% net imports between 2010 and 2030).Eventually the Netherlands produce at a level just above self-sufficiency-a shift from being a significant net exporter in 2010 to a transit country by 2050.Italy, and particularly Austria, become more and more important in the intra-European transit of natural gas.Some large natural gas consumers such as Germany and France experience a significant reduction of net imports over time, in line with their declining consumption.
Most imports to the EU are delivered by pipelines, today and in the coming decades.LNG imports stay below a 30% share of total net imports (fluctuating around 80 bcm/y over time as depicted in Figure 2) with the highest share in 2020.Most of the imported LNG continues to originate from the Middle East and Africa (jointly more than 70% in 2050).However, African LNG exports to the EU decline after 2020, whereas LNG supply from Russia and South America rises to almost 30% of EU LNG imports in 2050.
13.Even in the absence of additional import needs, a costly infrastructure expansion can be rational because we assume spatial price differentiation between markets.Profit-maximising traders with the ability to exert market power then have an incentive to supply to high-price regions with little competition.
14.In the past, such a project between Georgia and Romania was discussed under the denomination "White Stream." European natural gas imports-in the form of LNG or transported through pipelines-are mainly provided by African, Russian, and Norwegian supplies, and to a lesser extent from the Caspian region and the Middle East.In particular, Africa's share increases significantly (by pipeline exports) while Norwegian exports to Europe become smaller over time.A notable divergence from current trade flows are Russian exports; throughout the entire model horizon Russian exports to Europe and Ukraine are considerably lower than in 2010.This is due to increased competition for Russian gas within the domestic Russian market as well as strong demand in Asia-Pacific, where only a limited number of suppliers is available.Consequently, Russia does not completely use all of its large export capacity towards Europe.
Stagnating European LNG imports must also be seen in the context of a globally increasing demand for natural gas, which triggers a considerable increase of LNG imports in regions other than Europe, particularly in the Asia-Pacific region with a projected boom in natural gas consumption in China and India, and the ever-strong demand in Japan.A large share of the threefold increase of natural gas consumption in Asia between 2010 and 2050 is satisfied by LNG from other world regions.
More than half of European LNG imports are shipped to the UK and Spain.In particular, the UK imports more LNG over time due to decreasing domestic production levels and little interconnection with the continental pipeline network.Spain's LNG imports steadily decline over time as African pipeline gas becomes more important.Similarly, in Italy, additional pipeline gas imports from Africa and the Caspian replace imports of LNG.Moreover, France's LNG imports from Africa are phased-out over time due to unfavourable costs.In contrast, Ireland, Poland, Germany, and the Netherlands start importing LNG.
These findings explain, and are jointly determined by, infrastructure expansions.We identify two main drivers: first, an increase in total imports due to strongly decreasing EU production levels, and second, the profit-seeking motive of new suppliers to the European market. 13igure 4 depicts all major interconnecting pipelines towards Europe as well as the regasification infrastructure, comparing capacities in 2030 with those already existing in 2010.The largest cross-border pipelines in 2030 are projected to run from Russia to the Ukraine (119 bcm/y via the already existing Brotherhood system) and to Germany (59 bcm/y via Nord Stream), between Africa and Italy (64 bcm/y via TransMed, Greenstream, and GALSI) and Spain (32 bcm/y via MEG and Medgaz), as well as from Norway to the UK (47 bcm/y via Langeled) and to Germany (43 bcm/y via Europipe and Norpipe).
There are five major pipeline expansions that can be identified (see Figure 5).First, the (exogenously included) Nord Stream pipeline from Russia to Germany (59 bcm/y) is built until 2015, but will not be further expanded.Second, a pipeline across the Black Sea from the Caspian region to Central Europe is endogenously added to the European pipeline system to be available in 2020. 14This translates into major expansions between the Caspian region and Romania (by 34 bcm/ y), from Romania to Hungary (by 31 bcm/y), and further to Austria (by 28 bcm/y).From there, additional pipeline capacity to Germany (plus 33 bcm/y) is needed while the existing one from Austria towards Italy is sufficient.Third, capacity is added endogenously from Africa to Italy (GALSI pipeline with 29 bcm/y) to satisfy Italian demand and to further transport natural gas to 15. Construction of the offshore section of South Stream started in December 2012, but the project was abandoned during the Russian-Ukrainian crisis in December 2014.In spring 2016, Russian and Central European activities to revive the project were reported.It is unclear whether the project is still a viable option for a future export route of Russian natural gas to Europe.
16.For Poland and the Netherlands, these expansions have already been scheduled and are incorporated exogeneously into the model.For Germany, the endogeneouly determined expansion is below 1bcm/y and, hence, not depicted in Figure 4.
Western and Central Europe via Austria.This leads to an expansion from Italy to Austria (by seven bcm/y) and explains the significant expansion between Austria and Germany.The fourth major expansion towards Europe takes place between Africa and Spain (Medgaz pipeline with 25 bcm/ y).This endogenous expansion is partly explained by increased African exports to France, which are facilitated by additional capacity between Spain and France (11 bcm/y).Finally, the first section of South Stream is included with exogenous expansions between Russia and Bulgaria (16 bcm/y) and from Bulgaria towards Serbia (5 bcm/y). 15The Serbia to Hungary link is used sparingly, only a small expansion is needed.
Most of the described expansions take place before 2030.Exceptions include later capacity expansions from Africa to Italy and to Spain, as well as from Italy to Austria, and from Austria to Germany (see Figure 5).
Moreover, in Ireland, Poland, Germany, and the Netherlands new LNG regasification terminals are built (together 26 bcm/y until 2050). 16In other countries like Spain, France, Italy, or 17.This figure includes the total costs of investments in regasification facilities and of all inter-country pipeline expansions within the EU.For simplicity, we account for only half of the expansion costs of those pipelines that either start or end in the EU.For instance, half of the investment costs of the interconnection between Turkey and Greece is added to the EU's investment figure .Portugal the model suggests that it is not economically sensible to build additional LNG import capacities.Currently planned-and therefore included-expansions in these countries are not utilised.Given our assumptions on pipeline length and expansion costs, and taking into account investments into additional LNG infrastructure, total investment costs for the EU can be estimated at around € 25 bn.until 2050. 17More than 65% of these costs are due before 2020, and more than 94% before 2025.
Notably, some major projects under discussion are not endogenously built in the Reference scenario.Nabucco, TAP (Trans-Adria Pipeline), and ITGI (Interconnector Turkey-Greece-Italy) are now developed to bring Caspian natural gas to Central Europe.However, new infrastructure capacities contribute to an increased diversification of supplies as new import paths and transit routes open up.Accordingly, infrastructure investments improve supply security and lead to increased competition.

HRes Scenario: 80% GHG Emissions Reduction until 2050
Aggregate natural gas consumption is significantly lower in the HRes scenario than in the Reference scenario (see Figure 6).This has important implications for EU imports of natural gas and infrastructure expansion needs.While in the Reference scenario, EU net imports increase over the entire time horizon; in the HRres scenario, they only slightly increase in the first decades, before starting to decrease strongly after 2030.In line with the development of consumption and imports, infrastructure expansions are generally smaller than in the Reference scenario.Figure 7 depicts the absolute deviations of cumulative infrastructure expansions until 2050 relative to the Reference scenario.In particular, the connections from Africa to Italy and Spain are expanded by much smaller amounts.Likewise, the pipeline from the Caspian region to CEE with all subsequent pipeline sections in Europe is built with a significantly lower capacity.This translates into fewer expansions between Italy and Austria, and from Austria to Germany.There are only two minor exceptions: small amounts of additional capacity are built between Germany and Denmark to satisfy a considerably higher Danish demand, and between Greece and Italy to compensate for the faster decline in Italian production.Regasification facilities are built endogenously only in Ireland, but with smaller capacity than in the Reference scenario (less than to two bcm/y of new capacity in 2030).

AN ALTERNATIVE PERSPECTIVE: GROWING IMPORTANCE OF NATURAL GAS IN A LOW-CARBON EUROPE
The two EU Energy Roadmap 2050 scenarios discussed in the previous section are characterised by a decrease in EU natural gas consumption.In the political and public debate, however, natural gas is often characterised as an important energy carrier on the way to a low carbon economy, i.e., as a "bridge technology". 18Natural gas is the fossil fuel with the lowest carbon content per energy unit and natural gas power plants have the flexibility to be used as backup for intermittent renewable power generation.The discrepancy between the advantages of natural gas and its declining consumption in the EU Energy Roadmap has led us to investigate projections from other model frameworks that are also in line with EU climate targets.To this end, we construct two alternative scenarios in which natural gas plays-at least temporarily-a vital role in the transition to a low-carbon Europe.
Our first alternative scenario, the Back-Up scenario, is based on the New Policy Scenario (NPS) of the World Energy Outlook (IEA, 2012), which provides projections for the global energy mix until 2035.In terms of stringency of EU climate policy, the NPS is comparable to our Reference scenario as it includes the EU 2020 targets of 20% GHG reduction relative to 1990 and a 20% share of renewables in the energy demand, and its emissions path is in line with a 40% reduction until 2050.The NPS shows how EU climate policy could lead to an overall reduction in the use of fossil fuels with a shift from coal and oil to renewables and natural gas.The projected emissions reduction in Europe in the NPS can be decomposed into two effects.On one hand, the joint consumption of the three main fossil fuels oil, coal, and natural gas is projected to steadily decline (a scale effect).On the other hand, the relatively carbon-intensive fossil fuels, oil and particularly coal, are substituted for by natural gas, whose combustion generates less CO 2 per energy unit (substitution effect).About two thirds of the overall emissions reductions until 2035 can be attributed to the scale effect, and one third to the substitution effect.According to IEA (2012), EU natural gas consumption will increase by 14.7% in 2035 relative to 2010.This stands in sharp contrast to the reduction in natural gas consumption of 14.9% in our Reference scenario-however, leading to the same emissions pathway.
Our second alternative scenario, the Bridge scenario, is based on results of the PET model presented within the EMF 28 (see Labriet et al., 2012, for the model description and Knopf et al., 2013, for a comparison with other models' results in the EMF 28 group).We follow a scenario characterised by a reduction of EU GHG emissions of 80% until 2050 that is similar to the HRes scenario discussed in the previous section.In this Bridge scenario, natural gas consumption increases slowly until 2030 and it decreases sharply therafter.This scenario comes closest to the often-cited "bridge into a low carbon future".   in the Back-Up scenario, EU consumption of natural gas is well above levels in the Reference scenario.In 2030, about 526 bcm/y is consumed (27% more than in the Reference scenario) increasing to 579 bcm/y in 2050 (52% more).Consequently, falling production and steadily increasing consumption levels lead to an even higher import dependency, reaching 94% in 2050.
In every period, the EU's share of global consumption is higher in the Back-Up scenario than in the Reference scenario since the Rest of World assumptions are the same in both scenarios.In the Reference scenario, the EU share in global consumption decreases from 15% in 2010 to 7% in 2050; it decreases much less, to 11% in the Back-Up scenario.The increased consumption levels in 2030 relative to the Reference scenario are spread across all EU member states with significantly higher levels in Germany, Italy, the UK, and Poland due to a significant substitution of natural gas for coal in power generation.
In 2030, total net imports into the EU are 33% higher than in the Reference scenario, rising to a 57% higher level (about 550 bcm/y) in 2050.All major trade flows from the Reference scenario show up in larger magnitude in the Back-Up scenario (see Figure 9 for a 2030 comparison).The role of Austria, Hungary, Romania, and Slovakia as transit countries is more pronounced in the Back-Up scenario.
LNG imports oscillate around a 25% share of total EU net imports.Accordingly, almost 132 bcm/y of LNG is projected to be imported by the EU in 2050.Although LNG imports from the Middle East increase over time, this region's market share declines.Both South America and Russia are projected to gain a larger share in the European LNG market.North American LNG will be exported to the EU by mid-century.Higher LNG imports relative to the Reference scenario lead to a small increase in the expansion of regasification facilities in the EU.Until 2050, an additional capacity of 50 bcm/y is built up compared to only 45 bcm/y in the Reference scenario (of which around 39 bcm/y is exogenously given in both scenarios).The higher expansion is divided between Ireland and Germany (about + 3 bcm/y until 2050 each).In Poland, there is no additional investment to the scheduled (exogenous) expansion in either scenario.
A larger difference between the two scenarios can be observed in the expansion of the pipeline network directed towards Europe.Figure 10 contrasts cumulative pipeline expansions for  20.Note that we do not include the possibility for shale gas production in Europe in our data set, due to the lack of reliable data.However, given the high costs that shale gas production in Europe would likely have and the overall reduction of natural gas consumption in most scenarios, shale gas hardly seems to have a bright prospect in Europe.
21. See Richter and Holz (2015) for a specific analysis of such disruption crises of Russian natural gas supplies to Europe.
22. Pipelines originating in Germany that carry supply towards the Netherlands, Denmark, and Belgium are expanded in the Back-Up scenario between 7 and 13 bcm/y each.
of the EU with the Asia-Pacific and Russian domestic consumption.Total EU infrastructure investments in the Back-Up scenario are projected to be almost twice as high as in the Reference scenario.They reach almost € 43 bn.until 2050, with investments before 2025 being significantly higher than total expenditures in the Reference scenario.
Central and Eastern Europe (CEE) currently suffers from a strong dependency on Russian natural gas exports.The only direction of pipeline flow is from the East (Russia and subsequent transit countries) westwards.Some countries currently have modest domestic production, such as Romania, Poland, Czech Republic, and Hungary, but this will phase out by 2040 (except in Romania). 20While natural gas is usually not the dominant fuel in the energy systems of the CEE countries, it is often the input fuel for peak power generation (e.g., on high-demand winter days).This makes these countries vulnerable to unilateral disruptions by Russia, which occured in the winters of 2008/2009 (gas dispute between Russia and Ukraine) and 2011/2012 (strong winter and gas dispute), and particularly, during the 2014 Russia-Ukraine-EU crisis. 21CEE countries would benefit from a diversification of supplies, even if it affects only small shares of total imports.Proposals include reverse flow capacity, LNG terminals in coastal countries, and increasing storage capacity in order to improve the supply security by increasing the number of potential exporters to CEE.
A few major pipeline expansions in the Back-Up scenario are relevant for CEE.First, Caspian natural gas finds its way to central Europe via a significant pipeline expansion through Bulgaria, Romania, and Hungary to Austria.This leads to a finer meshed pipeline network towards and within CEE than is currently the case, resulting in a potentially higher diversification and a reduced dependence on Russian natural gas.This goes hand in hand with a relatively small expansion from Russia towards Bulgaria ("South Stream").Second, the offshore project between the Caspian region and Romania ("White Stream") is preferred endogenously to other projects in the Southern corridor.In particular, there are no investments along the original route of the Nabucco project via Turkey and Romania towards central Europe.However, a similar pipeline route from Turkey to Bulgaria and further on via Serbia to Hungary is expanded by a small amount.The model outcome of the high-demand scenario (Back-Up) suggests a more attractive interconnection from Turkey via Greece to Italy (the TAP).Third, both expanded pipeline routes to Austria-from the Caspian Region via CEE as well as the connection between Italy and Austria-lead to a further capacity expansion towards Germany.In the Back-Up scenario, this expansion accounts for almost 66 bcm/y. 22Fourth, and importantly, West-East natural gas transfers (reverse flow) become possible due to new pipelines.Poland can import more natural gas via Denmark and the Czech Republic.The pipeline from Austria to Slovakia is built and the pipelines from Austria to Hungary and further on to Romania and from Italy to Slovenia are also built.Finally, there will be no LNG regasification capacity construction in CEE other than the small Polish terminal currently under construction.Pipeline supplies, including reverse flows, will remain the preferred method of import, even for coastal countries.In this section, we take a closer look at the Bridge scenario.In particular, comparison to the "Back-Up" scenario reveals some interesting insights concerning infrastructure expansions.Until 2030, one can observe similar paths between these two scenarios concerning production, consumption, and trade patterns for the EU aggregate.After 2030, however, the scenarios substantially diverge (see Figure 11).In 2030, the EU aggregate consumption levels are almost the same for both cases at about 10% above 2010 levels.However, while consumption increases steadily in the Back-Up scenario, it decreases fast in the Bridge scenario.In 2035, consumption in the latter is 23%, in 2050 it is 74% below the Back-Up consumption level.The 2050 consumption level in the Bridge scenario is 69% below the 2010 consumption level.Both absolute imports (LNG and via pipelines) and the import dependency are close to but higher than the Bridge scenario until 2030.LNG imports in this scenario reach a level of more than 130 bcm/y in 2020-a share of more than 32% of net imports.
When contrasting scenarios, there is spare LNG import capacity in both the Energy Roadmap scenarios discussed in Section 3 and in the Back-Up scenario.However, all LNG import capacity is used in the Bridge scenario until 2030.LNG imports can flexibly satisfy additional demand for a limited time period.As demand is not sustained in the long-term, many infrastructure expansions in the pipeline network are not economically justifiable and LNG regasification is utilised as the alternative short-term option to serve the high demand.
Pipeline expansions in the Bridge scenario are close to the Reference scenario but generally smaller (with a few exceptions) than under the other 80% scenario, HRes (see Section 3).In contrast, expansions of regasification facilities are highest in the Bridge scenario (58 bcm/y compared to 52 bcm/y in the Back-Up scenario).The decline in demand after 2030 in the Bridge scenario reduces total infrastructure investments and replaces pipeline expansions with short-term economical and more flexible LNG import facility expansions (see Figure 12).Until 2025, the two alternative scenarios follow a similar expansion path, slightly higher than expansions of the Reference scenario.After 2025, only in the Back-Up scenario do cumulative expansions continue to increase gradually up to a significantly higher level.
Total EU infrastructure investments in the Bridge scenario are about € 26 bn., close to the expenditure figures in the reference scenario but well below the investment costs in the Back-Up scenario.Among all three scenarios, the Bridge scenario is characterised by the highest investment levels between 2010 and 2020.Moreover, in this scenario, 99% of all investments are made before 2025 compared to only 69% in the Back-Up scenario.

CONCLUSIONS
In this paper, we take a closer look at how the envisaged decarbonisation of the European energy sector will affect the natural gas sector.We focus on infrastructure needs to accommodate the transition to a low-carbon economy.To this end, we apply the Global Gas Model, a multi-period complementarity model of the world natural gas market.We analyse three potential and quite opposite pathways for the role of natural gas in Europe: first, a continuously decreasing consumption of natural gas in the EU; second, a slightly increasing consumption path; and third, the role of natural gas as a "bridge fuel" to a low-carbon Europe.
First, we construct two climate scenarios based on the EU Energy Roadmap 2050.Both scenarios are characterised by decreasing natural gas consumption-the more stringent the climate policy the lower the level of consumption.Consequently, in the 80% GHG reduction scenario (HRes), there is no need for large-scale pipeline expansions.In the moderate climate policy scenario, however, the decline in European domestic production and the increasing reliance on African and Caspian exports lead to some expansions of import capacity.This includes pipeline expansions from Africa to Spain and Italy, and from the Caspian region to Central and East Europe.European LNG imports, in turn, stagnate and even fall after a peak in 2020 because of strong demand in the Asia-Pacific region.The availability of shale gas, both as LNG exports from North America as well as with increased production capacity in some major demand regions of the world (e.g., China, Poland) could change this picture somewhat by reducing competition for natural gas with Asia-Pacific.
Second, we construct a set of alternative scenarios that are comparable in underlying GHG emissions reductions but that differ by continuously increasing natural gas consumption for part or all of the time horizon.Expansions in the moderate climate scenario with rising EU natural gas consumption (Back-Up) are significantly higher than in the Reference scenario.This holds, particulary for the connections from Africa and the Caspian region to central Europe, while a new pipeline from the Middle East towards Turkey, Greece, and Italy is constructed.The results of the Bridge scenario, with only temporary increases in natural gas consumption, show that long-term trade relations are needed to economically justify pipeline infrastructure construction.Instead of large scale pipeline expansion, the existing idle LNG import capacities are used during years of high demand, supported by some additional expansions in regasification facilities.
All scenario results show improved import diversification due to the build-up of West-East (reverse flow) capacity that is still largely lacking in today's market.This is the result of an economic cost minimisation mechanism in a capacity-constrained market with market power despite no explicit consideration of supply security considerations.In other words, supply security would benefit from relaxing the (institutional, political, and technical) constraints on investments as we assume for the period after 2020.
The next decade will show if the tendency towards lowering natural gas consumption in Europe, as indicated by the Energy Roadmap, will come to fruition.In contrast, a stronger reliance on natural gas may be a probable energy future and our alternative scenarios indicate the economic feasibility of such a pathway.
Future research should include a more detailed look at the developments in other world regions outside of Europe to capture all global dynamics.In particular, in the Asia-Pacific region, one can expect strongly increasing demand in the emerging natural gas markets of China, India, Thailand, and others, booming natural gas production and LNG exports in Australia, and a sustained high demand in Japan and Korea.Climate and energy policies in this region would impact the trade flows in the entire global natural gas market and merit a more detailed modelling analysis.Moreover, there is uncertainty on a number of factors in natural gas markets, and stochastic modelling may be an alternative to the deterministic scenario analysis presented in this paper.Fodstad et al. (in  this issue) present results using a stochastic modelling approach.Finally, our modelling exercise shows that there is a need for a clearer picture of the role that natural gas can play in Europe in the upcoming decades.This is particularly necessary for today's decisions on the future natural gas infrastructure.The stochastic version of the Global Gas Model was introduced and used in Egging (2010,  2013).Notation and explanations have been modified and shortened here.For details, we refer to these earlier publications.In the following, we give the complete optimization problems of the representative agents that are modeled in the GGM.The optimization problems of the various agent types are linked by market-clearing conditions and solved simultaneously by using their Karush-Kuhn-Tucker (KKT) conditions.The KKTs are the first-order (optimality) conditions of the optimization problems under constraints, under certain conditions-that are met in our problem setupthey give the optimal solution to the problems.The KKTs constituting the mixed complementarity problem (MCP) are collected at the end of this appendix.

Trader
Traders are the central agents in the model.They act as the trading arm of a supply company with one production region.Traders may act perfectly competitively or exert market power with respect to end users.We apply a conjectural variation approach: parameter is assigned values cv tny between 0 and 1 varying by trader, region and/or year.A value of 0 implies perfectly competitive behaviour, a value of 1 Nash-Cournot oligopolistic behaviour, and values in between indicate moderate levels of market power exertion.Consequently, sales revenues are expressed as: , where the conjectural variation parameter is

Producer
A producer supplies natural gas to its trader at marginal cost.He maximises discounted profits Eq. (A.4): revenues minus costs subject to capacity constraints Eq. (A.5).
p P P P p q c (q ) ndy ndy ndy ndy His decision variables are production levels in all seasons of all years in the time horizon: .The following two sections describe the two types of infrastructure services, transmission and storage, both with a Third Party Access (TPA) regime and regulated service fees.If capacity is restrictive, market-determined congestion fees are added to the regulated fees, such that for the resulting aggregate fee, the capacity services demanded by the traders equal the available capacity.

Transmission System Operator (TSO)
The network of transmission arcs includes pipelines as well as liquefaction, shipping, and regasification activities in the LNG value chain.Arcs are directed.A pair of nodes may have two arcs connecting them, at most one in each direction.LNG liquefaction and regasification is represented using auxiliary geographical nodes.The TSO maximises discounted congestion revenues minus investment costs, subject to capacity constraints.He allocates pipeline capacities , and is A f ady responsible for network expansions .

Figure 1 :
Figure 1: Representation of the Natural Gas Market and Supply Chain in the Global Gas Model (GGM)

Figure 2 :
Figure 2: EU Consumption, Production, and LNG Trade, as well as Import Dependency (right axis) in the Reference Scenario (in bcm/y and percent)

Figure 3 :
Figure 3: Imports and Exports by Country and Type (pipeline natural gas or LNG) in 2010, 2030, and 2050 in the Reference Scenario (in bcm/y)

Figure 4 :
Figure 4: Pipeline and Regasification Capacities to Europe in the Reference Scenario in 2010 and 2030 (in bcm/y)

Figure 5 :
Figure 5: Pipeline Expansions in the Reference Scenario with a Destination within the EU Member States (lower part of horizontal axis is the pipeline's origin; in bcm/y)

Figure 6 :
Figure 6: EU Consumption Levels and Net Imports by Scenario (in bcm/y) 19

4. 1
Back-Up Scenario: An Increasing Role for Natural Gas in a 40% Reduction Scenario EU aggregate differences between the Reference and Back-Up scenarios are summarised by Figure8.While EU production levels are only marginally higher than in the Reference scenario,

Figure 8 :
Figure 8: Comparison of EU Consumption, Production, LNG Trade, and Import Dependency between Back-Up and Reference Scenarios (in bcm/y and percent)

Figure 9 :
Figure 9: Imports and Exports through Pipelines and in the Form of LNG in 2030; Back-Up vs. Reference Scenario (in bcm /y)

Figure 11 :Figure 12 :
Figure 11: Comparison of EU Consumption, Production, LNG Trade, and Import Dependency between Back-Up and Bridge Scenarios (in bcm/y and percent)

Table 4 : Dual Variables and Prices
A e q ay Dual to Arc expansion limit S e q ny Dual to Storage expansion limit A s ady Congestion rate for arc S s ndy Congestion rate storage injection used to weigh the inverse demand curve that can be affected, and the market price (exogenous to a perfectly competitive trader).The trader maximises discounted season-length weighted dr d y dprofits, resulting from sales revenues to end users in different regions, minus gas purchase costs and fees for using transmission arcs and storages.His decision variables, in all seasons of all years in the time horizon, are sales Mass balance must be maintained at each node Eq. (A.2), n must equal the extractions: Eq. (A.3), the storage cycle constraint.