MENA Compared to Europe: The Influence of Land Use, Nuclear Power, and Transmission Expansion on Renewable Electricity System Costs

Most studies of near-zero-carbon power systems consider Europe and the United States. In this paper, we focus on the Middle East and North Africa (MENA), where weather conditions, especially for solar, differ substantially from those in Europe. We use a green-field linear capacity expansion model with over-night investment to assess the effect on system cost of (i) limiting/expanding the amount of land available for wind and solar farms, (ii) allowing for nuclear power and (iii) disallowing for international transmission. This is done under three different cost regimes for solar PV and battery storage. We find that: - The amount of available land for wind and solar farms can have a great impact on the system cost. We found a cost increase of 0-50% as a result of reduced available land. In MENA, the impact on system cost is greatly influenced by the PV and battery cost regime, which is not the case in Europe. - Allowing for nuclear has nearly no effect in MENA, while it can decrease system costs in Europe by up to 23%. In Europe, the effect on system cost of whether nuclear power is allowed is highly dependent on the PV and battery cost regime, which is not the case in MENA. - Disallowing for international transmission increases costs by up to around 25% in both Europe and MENA. The cost increase depends on cost regime for PV and batteries. The impact on system cost off these three controversial parts of a decarbonized power system thus plays out differently, depending on (i) the region and (ii) uncertain future costs for solar PV and storage. We conclude that a renewable power system in MENA, is less costly than in Europe irrespective of the cost regime. In MENA, the system costs vary between 37 and 83 euro/MWh. In Europe, the system costs vary between 43 and 89 euro/MWh.


Introduction
of renewable power systems in any region of the world. We use MENA and Europe as test cases, since they differ in terms of resources. In addition, applying the model to Europe allows us to benchmark our results against those in the literature, e.g., [4,8,20]. The over-arching research questions in this paper are: • What is the cost of a CO2 neutral future power system in MENA/Europe?
• What is the impact of weather conditions and demand density on the cost of carbonneutral power systems?
• What is the impact of (i)-(iii) on system cost?
The paper is organized as follows: Section 2 describes the scenarios, model, and data input, and provides resource availability in the two regions in the form of supply curves for wind and solar. Section 3 outlines the results for the four scenarios (base, land availability, nuclear, and transmission expansion). In Section 4 we discuss the results relative to the literature as well as policy, and Section 5 provides concluding remarks and indicates a direction for future research.

Method
We use an energy-system model based on a model developed at the division of Physical Resource Theory at Chalmers University of Technology, which is available online [36]. We apply the model to MENA and Europe. By evaluating both regions with the same model, the difference in results between the two regions may be attributed to differences in demand and weather, rather than different model formulations and cost assumptions. Four different scenarios are evaluated: one base scenario and three scenarios where the premises for transmission expansion, nuclear power and land availability for wind-and solar farms are changed, respectively, see Table 1. Table 1: Scenarios. Nuclear power and international/inter-subregional transmission are either fully available as investment options or totally excluded. All scenarios restrict land available for onshore wind-and solar exploitation. All land-based water-areas, natural reserves, and areas with a population density higher than 75 [capita/km 2 ] are removed. We then assume that a percentage of the remaining land is available for onshore wind-and utility solar PV (the percentage indicated in the table applies to each technology, e.g., 10% corresponds to 10% for wind power and 10% for solar power, or 20% in total). All scenarios are evaluated for high-, mid-and low PV-and battery costs, see Table 2. The PV costs are retrieved as the low, mid, and high cost-scenario projections by NREL [37]. In addition to the technologies listed in Table 2, there is the option of residential PV, PV rooftop. The cost for PV rooftop is assumed to be 50% higher than the cost for PV Utility, see Section 2.2.2 below. The evaluated costs for batteries are retrieved from utility-scale lithiumion storage projections made by W. J Cole [38], as the highest, midrange and lowest projected costs.

Model
We use a bottom-up linear investment energy-system network model with hourly resolution for a full chronological year, to minimize total system cost for a power system that meets the demand at all times. Our model is based on an early version of the Supergrid capacity expansion model [36]. Since the focus is to evaluate the cost-efficiency of a future system with inter-continental grid connections, rather than the pathway to reach such a system, we employ overnight investment in a green-field optimization approach. The exception is hydropower, which is assumed to be installed at its present capacity, as reported by the World Energy Council [39]. Technology costs and electricity demand are projections for 2040. Demandand weather data, as well as costs and technology performances, are exogenous to the model. The model is implemented in Julia using JuMP, a domain-specific modeling language for mathematical optimization embedded in Julia.
Variables subject to optimization are capacity investment, electricity generation, storage and transmission. These variables are functions of the subregions R = {r 1 , ..r n }, the technologies possible to invest in K = {k 1 , ..k n }, different classes of solar-and wind power C = {c 1 , ..c n } (depending on capacity factor) and the hours over one year H = {h 1 , ..h n }. Parameters given to the model include technology costs, technology efficiencies, demand, distance between subregions and capacity factors. The model represents wind and solar power using five resource classes for each region, see details in the supplementary material. We use the GlobalEnergyGIS package [40] to generate the maximum potential capacity (in GW ) and hourly capacity factors for each technology, resource class and model region.
The objective function to be minimized is the total system cost.
By convention we use uppercase for variables and lowercase for parameters.
The transmission cost is divided by two since the model is investing in transmission lines between subregions r 1 and r 2 and between r 2 and r 1 , even though only one line is needed. The transmission cost (tc, [€/GW]) is a function of transmission line cost (tlc, [e/GW/km]), distance between subregions (di r 1 ,r 2 , [km]), transmission substation cost (tsc, [e/GW ]) and a fixed transmission cost (tf c, [%of ic]). Two substations are assumed to be needed for each transmission line.
The implemented constraints follow. First, we need to ensure load balance, i.e., make sure that demand is met at all times.
The electricity generation per hour (G, [GW h/h]) must be less then or equal to the installed capacity (C, [GW ]) multiplied by the capacity factor (cf ) for each technology, hour and subregion.
The storage level (SL r,k,h , [GW h/h]) cannot be negative: The maximum storage level depends on the installed capacity (C r,k,c , [GW ]) and the discharge time (dt r,k , [h]) for the storage technology. For batteries, the discharge time is set to 8 hours; for hydro dams, it depends on the dam size.
If h=1, the first term after the inequality sign is instead the storage level in the last hour of the previous year. Note that the first term is less than or equal to and not simply equal to. This is due to spillage when the water inflow is greater than the amount of water that the dam can handle.
Charging batteries requires batteries: Transmission constraints assure that the transmitted electricity (T G r 1 ,r 2 , [GW h/h]) does not exceed the installed transmission capacity ([T C r 1 ,r 1 , GW ]) and that the installed transmission between subregion r 1 and r 2 is the same as between r 2 and r 1 .
In order to partially mimic realistic constraints on nuclear power plants, ramping constraints and a minimum generation level in percentage of installed capacity are imposed.
We assume a limited stock of biogas. No more than 5% of the total annual electricity generation can be produced by biogas turbines.

Data
Input data include: definition of subregions; estimates of transmission distances between subregions; cost-and performance data for technologies and fuels; hourly subregional demand for a full chronological year; capacity factors; and capacity limits for solar power, wind power, and hydropower. This section contains information on how these input data were retrieved and implemented in the model. All investment costs are annualized using a discount rate of 5%. Cost inputs are in dollars but the results are given in euros, with a conversion rate of 0.87 €/$.   The assumed transmission costs are presented in Table 3 and are retrieved from ETSAP [41], except for the fixed cost which is taken from NREL [42]. The lifetime of HVDC lines is assumed to be 35 years [43].

Technology-and Fuel Costs
The power generating technology options are wind power (on-and offshore), PV (utility and rooftop), concentrated solar power (CSP) and biogas turbines (GT). The assumed costs and efficiencies for each technology are shown in Table 4. The costs are retrieved from the National Renewable Energy Laboratory's (NREL) Annual Technology Baseline (ATB) Database 2018 [37]. This database contains technology cost projections for 2040 for a low-, mid-, and highcost scenario; the assumed costs used in this study are the mid-cost scenario projections in the NREL database [37]. For PV rooftop, the cost is assumed to be 50% higher than PV utility due to higher installation costs for smaller systems, which is in line with the cost projections in NREL database 2018 [37]. The investment cost for batteries is retrieved from utility-scale lithium-ion storage projections by W. J Cole [38]. Lifetime and round-trip efficiency for the batteries are also retrieved from W. J Cole [38]. The modeled batteries are assumed to be lithium-ion battery packs with a discharge time of 8 hours and the cost can therefore be converted from $/kW to $/kWh with a factor 8. We assume the cost of biogas to be the average biogas cost in [13], USD 60 per MWh.

Wind-and Solar Data
Installation limits for wind and solar power capacity are based on assumptions on typical wind and PV farm densities (W/m2) and available land (m2). Several auxiliary datasets were used to exclude areas where solar-and wind power cannot be placed and to estimate solar-and wind potentials for each region. The datasets include population (GPWv4, [44]), land cover (MODIS, [45]), protected areas (WDPA, [46]), and topography and bathymetry (ETOPO1, [47]) [40]. After masking out unsuitable locations, a certain fraction of the remaining area is considered available for solar and wind farms, see available land in Table 5. Assumptions on typical wind and PV farm densities (W/m2) are then used to convert the resulting available area to potential capacity (GW).
Hourly time series with capacity factors for PV and wind power are calculated using solar irradiation and wind speed from the ECMWF ERA5 database and the Global Wind Atlas [40,48,49]. The modeled subregions are divided into pixels (0.01°x0.01°), capturing the different solar and wind conditions with an hourly time resolution. Solar irradiation is used to calculate the annual PV capacity factor profiles assuming fixed-latitude-tilt; wind speed is translated into capacity factors based on a power curve for a typical wind park with Vestas 112 3.075 MW wind turbines [40]. To reduce the computational demand, the pixels in each subregion are aggregated into five classes, depending on yearly average capacity factors for solar-and wind power. The pixels within a resource class, in each subregion, are then assumed to have the same capacity factor time series (the average of all capacity factor time series in those pixels), see Supplementary Material and [40]. Supply curves for PV and wind power built on input data to the model are shown in Figure  3.

Figure 3:
Supply curves for PV and wind power as percentages of net demand, assuming midrange costs for PV and wind power and 10% available land. Net demand corresponds to the total demand less production from hydro power. These supply curves corresponds to the assumptions in the base scenario. Supply curves for other costs and assumptions on available land are presented in the Supplementary Material.

Hydropower
Installed hydropower capacities and annual production in each subregion are assumed to be as in 2016 according to the World Energy Council [39] and can be found in the Supplementary Material. Monthly hydro inflow profiles for each region were retrieved from the GRanD database [50,51]. The inflow profiles are converted to hourly inflow assuming an even flow within each month and the dam size is assumed to be equal to the annual production divided by 12, i.e., the dam can roughly hold a months's worth of energy before spillage occurs, depending on the inflow profile.

Hourly Demand Profiles
Due to a lack of comprehensive real-world demand data for each of our 23 subregions, we generate synthetic hourly electricity demand based on a machine learning approach. We model the profile of how future hourly electricity demand per capita varies over the year as a function of: (i) consumption-adjusted regional GDP [52], (ii) hourly temperature time series for 2015 from NASA MERRA-2 [53], (iii) country-level annual electricity generation for 2017 from [54] scaled to match regional final electricity demand in the SSP2-34 scenario for 2050 [52], and (iv) gridded global population of data [55]. We fit the model based on electricity demand for 44 countries from [56] (from 2015, matched with data from (i), (ii), (iv)) and fit a gradient boosting regression model [57] to the normalized demand time series for the different countries. Our model thus fits the profile over the year, and this is scaled up by the level given by (iii). For model validation, we use cross-validation to withhold the data for each of the 44 different countries (once per country) as test data and train the model on the remaining data (countries). Our model predicts quantitatively and qualitatively similar time series for most countries, lending credit to the view that we have a model with generalization across larger geographical regions. The resulting demand series are treated as inelastic in the optimization model (for more details, see [58]).

Results
The results are structured as a comparison between MENA and Europe, for three different scenarios, one with variable land availability (Section 3.1), one with a nuclear power option (Section 3.2) and one without the option of inter-subregional transmission (Section 3.3). Each scenario was investigated with a range of cost projections for solar PV and battery storage, see Table 2.
We begin by presenting a summary of the cost range obtained by varying scenario assumptions as well as PV and battery costs. System LCOE varies substantially between 43 -89 e/MWh (Europe) and 37 -83 e/MWh (MENA), based on the different scenarios and PV and battery costs, see Figure 4. The system LCOE assuming midrange costs for PV and batteries for the base scenario, a system with inter-subregional transmission and without nuclear power, is 63 €/MWh for Europe and 53 €/MWh for MENA (denoted with a black line in Figure 4).

Figure 4:
Range of system LCOE in Europe and MENA for the cases studied. Low, mid, and high values for the PV and battery investment costs are investigated for each of the three scenarios: one scenario with variable land availability for wind and solar farms; one scenario that allows for nuclear power; and one that does not allow for inter-subregional power transmission. The black line corresponds to the cost assuming midrange costs for PV and batteries for the base scenario, which does not allow for nuclear power but does allow for power transmission between subregions.
Depending on the scenario, and the cost level for PV and battery investments, the system cost is 6-35% lower in MENA compared to Europe. Thus, for any given assumption on land availability and investment costs, the system cost is lower in MENA than in Europe.
The main focus of this paper is how socio-political realities (regarding land availability, nuclear power, and transmission infrastructure) and cost developments for solar PV and battery capacity impact the system cost for low-carbon power. However, we include the energy mix in order to provide some information on system design. Figure 5 shows the optimal generation mixes in the base scenario for low-, mid-and high cost PV and batteries for both MENA and Europe. For both regions, the mix is dominated by wind power for high and mid-level investment costs, with transmission dominating storage. The mix is instead dominated by solar power when investment costs for PV and batteries are assumed to be low, for both regions, with storage winning out over transmission in Europe and completely dominating transmission for MENA, see Figure 5. (For the optimal generation mix for the other scenarios, see the Supplementary Material.) Figure 5: Optimal generation mix for low-, mid-, and high-cost PV and batteries in both MENA and Europe, as a share of total generation. The categories Transmission and Batteries represent the shares of total power generation that pass through transmission lines and battery storage, respectively, which is why the total generation exceeds 100% of demand.

Available Land for PV and Wind Power
Decreasing the land available for wind and solar farms means that the supply curves ( Figure  3) become steeper, and it is necessary to exploit sites with lower output in order to cover demand, thus increasing system cost. Assuming less land is available for wind and solar farms (2% instead of 10% as in the base scenario) increases the system cost by up to 47% in MENA and 25% in Europe, see Figure 6. In Europe, the increase in system cost due to less land being available is up to 23-25% regardless of PV and battery costs. In MENA, the system cost increase due to less available land for VRE exploitation is smaller when investment costs for PV and storage are low, see Figure 6, with land availability having almost no effect for low investment costs. This has to do with wind power production being significantly more constrained by available land compared to in Figure 3. This is also shown in the generation mix: When there is less available land, deployment of wind decreases, while solar generation increases, an example of which may be seen in Figure 7. The reason the system cost does not increase significantly as land becomes more scarce in the Low case may be explained by the generation mix already being dominated by solar in that case, and thus the impact of the land constraint on wind power is less relevant. Solar figures prominently in the mix for Europe, too, in the Low case, but wind is still important, see Figure 5.

Allowing for Nuclear Power in the System
Allowing for nuclear means expanding the available technology options, thus always inducing a system cost decrease (or keeping the system cost at the same level as the base scenario). However, the resulting system-cost reduction is contingent on the costs of the alternative generation technologies, see Figure 8. Two things stand out from the results: First, if the investment costs for PV and batteries are assumed to be low, allowing for nuclear power does not yield any system cost reduction. Secondly, the cost reduction is smaller in MENA than it is in Europe. While the cost reduction range in Europe is 10% and 23% for Mid and High investment costs for PV/batteries, respectively, the cost reduction in MENA is only a few percent, see Figure 8. The difference between MENA and Europe in the cost-reducing effect of allowing for nuclear may be explained by more favorable solar and wind resources in MENA. The abundance of solar and wind resources in MENA entails that a renewable system, including the necessary flexibility capacity (batteries and trans-mission), will out-compete nuclear in most subregions. In comparison, in Europe, there is less low-cost wind and solar resource in relation to its demand (see Figure 3), which makes nuclear power relatively more competitive.

Excluding Transmission
We find an increase in system cost in both MENA and Europe when transmission is excluded, i.e., when subregions are isolated, see Figure 9. The cost increase is roughly the same in MENA and Europe: 11-25% in Europe and 4-23% in MENA, depending on cost assumptions for PV and batteries.
. The cost increase is significant (23% and 25%, respectively for MENA and Europe, when PVand battery costs are assumed 'High', roughly corresponding to today's cost. Conversely, the effect of excluding transmission is less significant when costs for PV and batteries are low. The explanation is twofold: First, allowing for transmission is more important when wind power is a large part of the mix, hence the smaller effect from excluding transmission when PV and batteries are low-cost and thus form a larger part of the mix, see Figure 5. At low costs for PV and batteries, the PV dominated system in MENA suffers only a small increase in system cost from excluding transmission (4% system cost increase). Secondly, if batteries are cheap, especially in combination with cheap solar, the effect on cost from allowing for trade of variations (through transmission expansion) is smaller.
Excluding inter-subregional transmission leads to significantly higher system costs (23% and 25%, respectively, for MENA and Europe), for high PV and battery investment costs, roughly corresponding to today's cost. Conversely, the effect of excluding transmission is less significant when costs for PV and batteries are low. The explanation is two-fold: First, allowing for transmission is more important when wind power is a large part of the mix, hence the smaller effect of excluding transmission when PV and batteries are low-cost and thus form a larger part of the mix, see Figure 5. At low costs for PV and batteries, the PV-dominated system in MENA suffers only a small increase in system cost from excluding transmission (4% system cost increase). Second, if batteries are cheap, especially in combination with cheap solar, allowing for trade (through transmission expansions) has a smaller effect.

Discussion
This paper investigates the system cost for a renewable power system in the MENA region, focusing on three influential factors: the availability of land for wind and solar farms; the option of including nuclear power; the possibility of transmission expansion. By comparing MENA to Europe, we trace the connection between resource quality (wind and solar conditions) and the cost of a renewable power system.

Comparison between Europe and MENA: System Cost and Technology Mix
Our resource quality assessment ( Figure 3) shows that MENA can satisfy its demand with either solar or wind power at a lower LCOE than Europe. We also find that for all investigated scenarios, MENA has a lower system cost than Europe. Our results thus show a correlation between resource quality and system cost. However, there are many additional factors that determine the cost and capacity mix of optimal renewable power systems, including the available variation management strategies, e.g., the abundance of reservoir hydropower, as well as the nature of spatial and temporal variations in VRE resources. Solar and wind power display different spatial and temporal variations, with solar power having a diurnal pattern, while wind power production displays variations on both shorter and longer time scales [59,60]. Complexities associated with how technologies complement each other in time and space influence cost and other features of a renewable power system, including the technology mix. Without prior knowledge of the system properties of wind and solar, respectively, it would, for instance, be difficult to infer that the system mix in the base scenario for midrange costs on PV and batteries is dominated by wind ( Figure 5, middle bar), since solar sites with lower LCOE than wind are abundant in both Europe and MENA (Figure 3, left column). However, our results still indicate that some tangible information on system cost may be inferred simply from considering the supply curves. In order to understand that connection better, more research is needed, both on other regions and with other technology scenarios. In the current study, the availability of nuclear resulted in the difference in system cost between MENA and Europe all but disappearing in the Mid and High investment cost cases, i.e., assuming midrange or high investment costs for PV and batteries, see Figure 8.

Transmission Expansion
We find that system costs increase by between 4 and 25% when inter-subregional transmission is excluded. Isolating subregions leads to over-investment in VRE capacity and more investment in storage and. in thermal generation capacity with potentially low full-load hours. This cost increase is consistent with the literature [4, 6-11, 13, 61]. However, unlike the majority of these papers, we investigate how the benefit of optimal transmission expansion depend on the cost of solar PV and battery storage. The benefit of adding inter-subregional transmission is greater when investment costs for PV and batteries are high compared to when they are low. A similar result was found by Schlachtberger et al. [4]. The underlying mechanism is that increased transmission mainly benefits systems with a high share of wind power, and low-cost solar PV and batteries lead to a smaller share of wind power in the optimal generation mix. Low-cost solar PV and batteries systems instead rely on more battery storage. For example, the low investment cost case (Low) results in 5.4 T W h of battery storage in MENA, equivalent to 63 million Tesla car batteries. This quantity of batteries may have consequences in terms of the use of materials. Thus, decarbonized power systems may entail hard-to-swallow features, such as large-scale transmission or large amounts of batteries or nuclear power.

Nuclear Power Option
By modeling the power systems in MENA and Europe with and without the possibility of investing in nuclear power, we show that allowing nuclear power may either have almost no impact on system cost or decrease system costs significantly, by about 20%. The effect of allowing for nuclear power is contingent on the supply of low-cost VRE resources (Figure 3) as well as low-cost variation management resources (here battery storage). Thus, the higherquality VRE resources in MENA, compared to Europe, entail that allowing nuclear power in the system has very little effect on system cost (the reduction in system cost is less than 4%), regardless of the cost assumptions for solar PV and batteries). In Europe, the benefit of including nuclear power is highly dependent on the cost assumptions, varying between 0% and 23% for the different cost assumptions for solar PV and battery investments. These system cost reductions are in line with results in the literature, where the effect of including nuclear power is moderate [62,63]. However, the literature also includes claims that decarbonizing without nuclear power (or other carbon-neutral thermal technologies such as coal with CCS) is substantially more expensive sepulvedaJenkins2018, hong2018economic. That said, those studies use narrower system boundaries, where regions are isolated and may not benefit from the flexibility provided by trade with other areas on a continental scale. Thus, we argue that power systems based primarily on renewable resources, and comparisons with such systems (as in [13,64]), are best investigated using models that have the flexibility of cross-border trade.
Our results, which show that the reduction in system cost of allowing for nuclear power in MENA is very small, support strategies that decarbonize power systems without investing in nuclear power and avoid associated concerns about safety and proliferation.

Amount of Land Available for Wind and Solar Farms
Both wind and solar power face social opposition in some regions in the world [34,35]. We show that the assumptions on available land are indeed an important determinant for the system cost of a fully renewable system. In fact, if the available land for VRE exploitation is cut in half (from 8% to 4%), the system cost may increase by up to 25%, depending on the cost for solar PV and battery storage, see Figure 6. Schlachtberger et al. [4] found an increase in system cost by 10% when the land available for onshore wind was reduced to zero, but our study shows that the effect in MENA (but not in Europe) is highly dependent on solar and battery costs. If investment costs for PV and batteries are assumed to be midrange or high, power systems in MENA and Europe are typically dominated by wind power, which means that system costs are more sensitive to land availability than for systems dominated by solar power. The impact on system costs of the assumptions on available land are of the same magnitude as the impact of allowing for inter-subregional transmission or nuclear power. This suggests that land-availability assumptions should feature more prominently than currently in policy discussions. The difference in system cost incurred by assumptions on available land could, for instance, be interpreted as an opportunity to give financial incentives to the people negatively affected by the construction of wind and solar power.

Limitations
The main results of this paper are about how circumstances due to policy and public opinion (land for VRE, nuclear power, and the availability of transmission expansion) impact the cost of a renewable power system, and how this impact may differ depending on regional resource endowment. The limitations of the model framework with potential consequences for this set of questions are: 1. Limitation: The system boundary in this study is set around the electricity sector and does not include other sectors in the energy sector such as heat, transportation, and industry. In addition, only one storage technology (lithium-ion batteries) is considered. Bearing on cost difference between Europe and MENA: Sector coupling is likely to entail an increased demand for electricity, both through electrification and use of electro-fuels, thus increasing scarcity of land for VRE farms. Thus, in the future, differences regarding the resource-to-demand relationship (Figure 3) may be greater and have a larger effect on cost, thus increasing the cost in Europe compared to MENA. This would also impact the relative benefit of using nuclear for electricity generation, especially in Europe. Sector coupling increases the temporal flexibility of the system. In this sense, it resembles the effect of low-cost storage. Thus, if there is sector coupling, it is likely that the availability of other variation management strategies, such as transmission, becomes less consequential for system cost. Similarly, it would be comparatively less costly to integrate renewables, thus rendering the nuclear option less important for cost reduction. However, since sector coupling increases the demand for electricity, the land-availability issue becomes more pressing.
2. Limitation: Political realities are not considered when modeling international transmission expansion, and only two cases are considered: no (international, i.e., intersubregional) transmission and optimal transmission. These two extreme points of transmission expansion are both unlikely. In fact, transmission between subregions already exists in both Europe [65] and MENA [66]. Bearing on impact on cost of inter-subregional transmission/nuclear option/NIMBY: The impact on system cost from extending transmission is in fact smaller than estimated in this paper, because the min-imum amount of transmission is already greater than zero, and the maximum feasible transmission grid is likely smaller than the optimal grid.

Limitation:
We have not allowed for expansion of hydropower. Bearing on cost difference between MENA and Europe: MENA exhibits a greater potential, compared to Europe, for expansion of hydropower. Allowing for expansion may increase the difference in system cost between the regions.

Limitation:
We model every subregion as a copper plate, i.e., electricity transmission within each subregion is assumed to be unlimited. Due to this, internal transmission requirements are not considered.

Bearing on results:
This assumption means that the cost for renewable power systems is underestimated in general, and likely especially so in cases with large volumes of power traded between subregions. This model artefact could potentially have a greater effect on the cost estimate for MENA, since its model subregions are generally larger. This issue was addressed in reference [61], where it was seen that cost and capacity mix did not change significantly as the spatial resolution was gradually coarsened 1 . The authors speculate that this is dues to that, as the spatial becomes coarser, there are two mechanisms that counteract each other: the transmission needs are underestimated, but at the same time the VRE resource is underestimated. This is due to that the method to estimate the VRE availability used in [61], entails that VRE resources are averaged, so that the best sites are no longer visible for larger regions. This latter is a trait which is less likely to interfere in the present study, since we employ wind-and solar classes in our model, thus capturing more of the resource heterogeneity compared with the method used in reference [61]. However, the lack of literature on the subject entails that we cannot be confident about the extent to which the large regions, and, especially, the unequal region size between Europe and MENA, impact the results. We believe that this topic merits more research in the future.

5.
Limitation: There are no ramping or start-and-stop costs for the thermal power plants.
Bearing on results: Cebulla and Fichter [67] showed that including such constraints are of little consequence for predominantly renewable power systems.
We deem the first of these limitations as likely having the largest effect on results, since it provides an alternative variation management strategy, which impacts the cost-effectiveness of nuclear as well as transmission expansion. Sector coupling also effectively increases the demand for electricity, thus putting more strain on land for production, which potentially increases the effect of less available land.

Conclusions
This paper investigates the effects of three socio-political factors on renewable power system costs, the availability of: (i) nuclear power, (ii) international transmission, and (iii) land for wind and solar deployment. The analysis is applied to MENA and Europe separately, which allows for a comparison regarding how a priori conditions (such as population density, available land for RE and weather conditions) may be used to predict the cost and capacity mix of renewable power systems. We find that: • For any combination of assumptions on investment costs for solar PV and batteries, as well as transmission/nuclear/land availability, the system cost is lower in MENA than in Europe. This suggests that the lower system cost is linked to the better wind and solar resource quality.
• The cost for a decarbonized power system ranges between 37 and 89 e/MWh in this study.
• Allowing for nuclear power reduces the system cost by 0% to roughly 20%. The magnitude depends on investment costs for solar PV and batteries, resource quality, and the availability of land for wind and solar. Because these factors are more favorable in MENA, the availability of nuclear power has a greater impact on system cost in Europe than in MENA.
• Allowing for optimal transmission expansion decreases the system cost by between 5 and 25%. The highest cost decrease (25%) is found when PV and batteries are high-cost, due to a corresponding higher share of wind power which is favored by transmission, since it smooths out wind variations. The cost impact from optimal transmission is similar in Europe and MENA.
• Public acceptance of wind and solar farms may have a large impact on the cost of a renewable power system. Our results indicate that a decrease of available land (from 10 to 2%) can increase system cost by about 50% for the case without nuclear but with the option of transmission expansion.
In summary, socio-political factors, here exemplified by whether nuclear power is included as a technology option, whether international transmission is possible, and the extent of land available for wind and solar farms, have markedly different impacts on results depending on the region (weather and demand conditions) and the cost of solar PV and storage capacity. Any judgment on the necessity of a specific socio-political factor for the realization of a decarbonized power system is contingent on assumptions regarding, for instance, investment costs and region. We also conclude that while the land available for wind and solar exploitation, which is affected by public acceptance issues, seems important for the system cost, this issue has not been investigated as thoroughly as other factors in model-based research. Future research could explore its importance in greater detail, with more realistic assumptions on restrictions for wind and solar expansion.

Supplementary Material Wind and solar resource classes
For solar technologies, the five classes are determined by the annual average capacity factor, where classes 1 to 5 are given by the ranges 0.1-0.15, 0.15-0.2, 0.2-0.24, 0.24-0.28, 0.28-1. For on-and offshore wind power, the classes are based on annual average wind speed, where the classes 1 to 5 for onshore wind are 4-5, 5-6, 6-7, 7-8, 8-99 m/s, and for offshore wind, 5-6, 6-7, 7-8, 8-9, 9-99 m/s. When assuming that the land available for solar and wind power is a fraction of the remaining land after maskings have been applied (see Section2.2.3), we use the same distribution of classes as for the total land, i.e., the model cannot only invest in the best classes for wind and solar power, but in the assumed fraction for every class of land.

System LCOE
LCOE for the total electricity system (system LCOE) is calculated as the system cost divided by the total demand. Since hydropower is modeled as no-cost in this study, the LCOE is calculated as the system cost divided by the difference between the demand and the annual hydropower production.
where SC is the total electricity system cost, D is the total demand and H is the annual hydropower production. Figure S1: Supply curves for PV and wind power in percentage of demand, assuming low-costs on PV and wind power and 10% available land. Net demand corresponds to the total demand subtracting the hydro power production. These supply curves corresponds to the assumptions in the base scenario.

Supply curves
V Figure S2: Supply curves for PV and wind power in percentage of demand, assuming mid-costs on PV and wind power and 10% available land. Net demand corresponds to the total demand subtracting the hydro power production. These supply curves corresponds to the assumptions in the base scenario.
VI Figure S3: Supply curves for PV and wind power in percentage of demand, assuming high-costs on PV and wind power and 10% available land. Net demand corresponds to the total demand subtracting the hydro power production. These supply curves corresponds to the assumptions in the base scenario.
VII Figure S4: Supply curves for PV and wind power in percentage of demand, assuming mid-costs on PV and wind power and 2% available land. Net demand corresponds to the total demand subtracting the hydro power production. These supply curves corresponds to the assumptions in the base scenario.
VIII Figure S5: Supply curves for PV and wind power in percentage of demand, assuming low-costs on PV and wind power and 20% available land. Net demand corresponds to the total demand subtracting the hydro power production. These supply curves corresponds to the assumptions in the base scenario.

Results
Optimal generation mix Figure S6: Optimal generation mix for low-, mid-and high-cost PV and batteries in MENA and Europe for the base scenario (10% available land). Generation is given as a percentage of total power production. Batteries can be interpreted as the share of total power production that passes through battery storage. Figure S7: Optimal generation mix for low-, mid-and high-cost PV and batteries in MENA and Europe for the no transmission scenario (10% available land). Generation is given as a percentage of total power production. Batteries can be interpreted as the share of total power production that passes through battery storage.
X Figure S8: Optimal generation mix for low-, mid-and high-cost PV and batteries in MENA and Europe for the nuclear scenario (10% available land). Generation is given as a percentage of total power production. Batteries can be interpreted as the share of total power production that passes through battery storage. Figure S9: Optimal generation mix for low-, mid-and high-cost PV and batteries in MENA and Europe for the base scenario but with 2% available land. Generation is given as a percentage of total power production. The categories Transmission and Batteries can be interpreted as the shares of total power production that pass through transmission lines and battery storage, respectively.
XI Figure S10: Optimal generation mix for low-, mid-and high-cost PV and batteries in MENA and Europe for the base scenario but with 10% available land. Generation is given as a percentage of total power production. Transmission and batteries can be interpreted as the shares of total power production that pass through transmission lines and battery storage, respectively.
Figure S11: Optimal generation mix for low-, mid-and high-cost PV and batteries in MENA and Europe for the base scenario and 20% available land. Generation is given as a percentage of total power production. Transmission and batteries can be interpreted as the shares of total power production that pass through transmission lines and battery storage, respectively.
XII Solar and wind power production Figure S12: The annual PV and wind power production as a function of available land, MENA to the left and Europe to the right.