The Cost of Undisturbed Landscapes

By 2030 Austria aims to meet 100% of its electricity demand from domestic renewable sources, with most of the additional generation coming from wind and solar energy. Apart from the benefit of reducing CO2 emissions and, potentially, system cost, wind power is associated with negative impacts at the local level, such as interference with landscape aesthetics. Some of these impacts might be avoided by using alternative renewable energy technologies. Thus, we quantify the opportunity cost of wind power versus its best feasible alternative solar photovoltaics, using the power system model medea. Our findings suggest that the cost of undisturbed landscapes is considerable, particularly when solar PV is mainly realized on roof-tops. Under a wide range of assumptions, the opportunity cost of wind power is high enough to allow for significant compensation of the ones affected by local, negative wind turbine externalities.


Introduction
"Holding the increase in the global average temperature [. . . ] well below 2°C above preindustrial levels and pursuing efforts to limit the temperature increase to 1.5°C above pre-industrial levels" (United Nations, 2015) requires rapid and substantial reductions in greenhouse gas emissions worldwide (Masson-Delmotte et al., 2018). Increasing the share of electricity generated from renewable sources consequently is amongst the United Nations' sustainable development goals (indicator 7.2.1) and a prominent goal of many climate and energy strategies. The European Union, for example, aims to meet at least 32 % of its final energy consumption from renewable sources by 2030 (European Parliament and European Council, 2018). 1 Against this background, Austria aims to generate 100 % of its electricity consumption from domestic renewable sources on annual balance by 2030 2 . The required expansion of renewable energy generation shall be incentivized by technology-specific subsidies rather than CO 2 pricing.
According to government estimates, an additional 27 TW h of electricity needs to be generated from renewable sources annually to meet this policy goal. Out of the targeted total increase in renewable electricity generation, 1 TW h is thought to be sourced from biomass, and 5 TW h from new hydropower generation, according to the government's climate and energy strategy (BMNT and BMVIT, 2018). The comparatively small projected increase in electricity generation from these sources results from the perception of biomass being ecologically 3 (Erb et al., 2018) and economically unsustainable 4 , while further development of Austria's hydropower potential was considered 'challenging' (Wagner et al., 2015) even before the projected 5 TW h increase in hy-1 In its "European Green Deal" the EU Commission proposes to increase ambitions and reduce CO 2 emissions by capacities (including seasonal hydro storages) in place, which should limit PV integration cost.
Thus, our findings should be qualitatively generalizable to power systems with high renewable shares but lower storage capacities with climates comparable to Austria. Methodologically, we rely on the power system model medea, which is described in section 2.
Our analysis contributes to a more comprehensive understanding of the trade-offs faced in the design of highly renewable electricity systems, thereby helping to inform policies about the socially optimal expansion of RET. Any such policy necessarily needs to account for the full social cost, including all externalities, of RET. Further, we contribute to the analysis of near-optimal power system configurations (see e.g. Neumann and Brown, 2019;Schlachtberger et al., 2017).
Corresponding results are presented in section 3. Finally, our investigation into the cost of undisturbed landscapes generates insight into the possibilities and limitations of funding potential compensating measures for the benefit of the ones negatively affected by wind turbines. We illustrate these potentials by an approximation of the valuation of undisturbed landscapes that is implicit in Austria's climate and energy strategy. Section 4 summarizes policy implications and concludes.

Data and Methods
To investigate the cost of undisturbed landscapes (i.e. the opportunity cost of wind power) we use the power system model medea, as summarized in subsection 2.1. We instantiate the model with scenario assumptions and data as observed in 2016, which was a low-wind year in Austria, while solar and hydro resources were close to long-term averages (see subsection 2.2). Based on this, we approximate the opportunity cost of wind power through the following procedure: 1. We derive the unrestricted system cost-minimizing deployment of wind and solar power, given the scenario assumptions 2. We restrict deployment of wind power by a small margin (so that PV, the next best RET, substitutes for wind power) and observe net system cost c net for Austria, calculated as total system cost including air pollution cost net of the balance of trade.
3. We repeat step 2 till no wind power can be deployed.
4. Finally, we approximate the opportunity cost of wind power (at given wind power capacity w) OC w by the change in c net in response to a change in wind power capacity w deployed, i.e.
where ∆ is the difference operator and opportunity cost OC w are expressed in A C per MW wind power foregone.

Power system model medea
We make use of the power system model medea to simulate (dis)investment in, and hourly operation of, the prospective Austrian and German power systems in the year 2030. The model is cast as a linear optimization seeking to minimize total system cost, which consists of fuel and emission costs, quasi-fixed and variable operation and maintenance (O&M) costs, the costs of investment in energy generation, storage, and transmission assets, and the (potential) cost of nonserved load. From an economic perspective, the model reflects a perfectly competitive energy-only market with fully price-inelastic demand and perfect foresight of all actors.
The modeled system is required to meet exogenous and inelastic demand for electricity and heat at any hour of the year. Energy supply, in turn, is constrained by available installed capacities of energy conversion, storage, and transmission units. Co-generation units convert fuel to heat and power subject to a feasible operating region defined by the unit's electrical efficiency, the electricity loss per unit of heat production, and the backpressure coefficient. Electricity generation from intermittent sources (wind, run-of-river hydro, solar) is subject to exogenous hourly generation profiles, which are scaled according to total installed capacities. Electricity from these sources can be curtailed at no additional cost (free disposal).
Electricity can be stored in reservoir and pumped hydro storages, batteries, or hydrogen storage (more precisely: water electrolysis, hydrogen storage, and reconversion to electricity in fuel cells).

Data
We set up our model with the goal to resemble Austria's prospective electricity and district heating systems in the year 2030. We also include Austria's largest electricity trading partner, Germany, to account for potential effects from electricity trade. For Germany, our scenario reflects current and announced future electricity sector policies up to 2030, so that we set generation capacities to levels consistent with these policies. These assumptions are laid out in subsections 2.2.1 and 2.2.2, and summarized in Table 1.

Scenario Assumptions for Austria
The Austrian government has set itself the goal of generating 100 % of electricity consumption from domestic renewable sources on annual balance by 2030 (BKA, 2020). However, industry own consumption and system services, which currently account for about 10 % of annual electricity consumption, are exempt. The government plans to achieve this goal by generating an additional 27 TW h of electricity annually from renewable sources 5 , for a total of 78.1 TW h of 5 Please note that this policy goal does not imply that the generated electricity must actually be consumed. Hence, we count curtailed electricity as contributing to the policy goal. 6 renewable electricity generation in 2030. New hydropower plants are thought to contribute 5 TW h annually, while additional electricity generation from biomass is envisaged to account for 1 TW h annually. In our scenarios, we add generation capacities sufficient to reach these targets. Further, the government projects the remainder to come from solar PV and onshore wind turbines with an annual contribution of 11 TW h and 10 TW h (BKA, 2020), respectively.
As we are interested in determining the opportunity cost of wind power versus its best alternative, we allow for endogenous investment in wind and solar power without enforcing announced technology-specific targets. Other low-carbon energy technologies are not feasible at a large scale in the Austrian context, as we have laid out in the introduction. Initial generation capacities for Austria are summarized in Table 1.

Scenario Assumptions for Germany
Germany has announced specific capacity targets for several power generation technologies.
Following these announcements, we anticipate an end to nuclear power generation, a (partial) is incentivised through technology-specific government subsidies, so that we assume targeted capacities will be in place by 2030. The corresponding capacity assumptions are displayed in Table 1. In addition, we allow for endogenous (dis-) investment in energy conversion capacities.

Energy supply
We represent 21 energy conversion and storage technologies that are expected to be operated in 2030. In addition to given, initially installed capacities (see Table 1), the model can endogenously add further generation capacities that are compatible with stated policy objectives. We calculate the annualized investment cost of each technology based on an assumed weighted average cost of capital of 5 % over the plant's lifetime. All technologies can also be decommissioned so that we adopt a long-run perspective on the power system.  continued on next page 9  Power plants running on fossil fuels emit CO 2 into the atmosphere. We approximate the amount of CO 2 released through the carbon intensity of fossil fuels, as displayed in Table 3. These estimates are based on an analysis by the German environmental agency (Juhrich, 2016).
In the baseline scenario, we assume capital cost of solar PV of A C625/kWp installed. These Generation from non-dispatchable technologies solar PV, wind turbines, and run-of-river hydro is assumed to follow hourly generation profiles 6 as observed in 2016. German solar and wind power profiles are sourced from Open Power Systen Data (2019). For Austria, solar and wind 6 A generation profile tracks the share of installed intermittent capacity that is generating electricity over time.  (2019), which we scaled to match annual generation as reported in the national energy balance (Statistik Austria, 2020). Subsequently, the scaled generation time series

10
were divided by installed capacities as published in Biermayr et al. (2019). Similarly, generation profiles for run-of-river hydropower were derived from generation data reported by ENTSO-E Transparency Platform (2020a), which we scaled to match annual run-of-river electricity generation as reported by Austria's regulatory body E-Control (2020).
Data on inflows into Austrian hydro reservoirs are not publicly available. However, the ENTSO-E Transparency Platform (2020b) reports weekly data on hydro storage filling levels along with hourly electricity generation from hydro reservoirs and pumping by pumped hydro storages (ENTSO-E, 2020b). Based on this data, we have approximated inflows as the part of the change in (weekly) hydro storage fill levels that isn't explained by upsampled hourly pumping and generation. To arrive at hourly inflows, we have interpolated our weekly estimates by piecewise cubic Hermite interpolation. Although inflows to water reservoirs are given exogenously, all storage technologies (including hydro reservoir and pumped storage plants) are charged and dispatched endogenously. Annual heat consumption is derived from energy balances for Austria and Germany, respectively (AG Energiebilanzen, 2018; Statistik Austria, 2020). Subsequently, we break annual heat consumption down into hourly heat consumption, based on standard natural gas load profiles for space heating in the residential and commercial sectors (Almbauer and Eichsleder, 2008). These load profiles make use of daily average temperatures to calculate daily heat demand. We extract spatially resolved temperature data from ERA-5 climate data sets (C3S, 2017) and compute a capacity-weighted average of temperatures at locations of combined heat and power (CHP) generation units. Daily heat demand based on these weighted average temperatures is then broken down to hourly consumption based on standardized factors accounting for weekday and time-of-day effects. Descriptive statistics of electricity and heat consumption are provided in Table 4.

Electricity Transmission
Electricity transmission between the modeled market areas is limited to 4.9 GW, in line with the introduction of a congestion management scheme by German and Austrian authorities in 2018 (Bundesnetzagentur, 2017). Section 3.4.2 explores the effects of increasing transmission capacity to 10 GW.

Prices
Monthly prices for exchange-traded fuels (hard coal, crude oil (Brent), natural gas) are retrieved from the International Monetary Fund's Commodity Data Portal. We convert these prices to A C/MW h based on the fuel's energy content and market exchange rates obtained from the European Central Bank (2020). Finally, we resample prices to hourly frequency using piecewise cubic Hermite interpolation. Due to its low energy density, lignite is not transported over large distances and consequently also not traded on markets. Instead, lignite-fired power plants are situated in proximity to lignite mines. According to estimates from Öko Institut (2017), the price of lignite in Germany is close to A C1.50/MW h. Biomass-fired power plants run on a wide variety of solid and gaseous fuels, some of which are marketed. However, the continued operation of biomass-fired plants typically relies on sufficient subsidies. As a first-order approximation to more complex prices impact renewables deployment (Brown and Reichenberg, 2020;Kirchner et al., 2019) and affect the results of our analysis. As the future efficient price of CO 2 is highly uncertain, we have conducted our analysis for CO 2 prices in the range of A C0 to A C100 per tonne. Implicitly, we assume efficient pricing of CO 2 emissions, i.e. all otherwise external cost of CO 2 emissions are internalized through the prevailing CO 2 price.  (2017). We have, however, converted these output-related estimates to input-related values, so that variable cost refer to the energy content of fuel used to generate electricity. The air pollution cost of RET mainly reflect the cost of air pollution during manufacturing of the plants. To avoid distortions from different underlying assumptions on RET utilisation, we have converted these cost estimates to reflect the specific annual cost per unit of capacity installed. The corresponding estimates for the external cost of air pollution from fossil fuel combustion are summarized in Table 6. Please note that these cost are considered external and consequently do not enter system cost minimization.
The data retrieval and processing scripts are published on GitHub https://github.com/ inwe-boku/medea under the MIT license.

Results and Discussion
To derive the social opportunity cost of not using wind power (but solar photovoltaics instead) in 2030, we start by determining the long-run power market equilibrium without constraints on renewable capacity addition. This gives us our unconstrained baseline scenario with minimal 14 system costs. Subsequently, we report the system effects of a gradual restriction of wind power and the corresponding estimates of the cost of undisturbed landscapes (in terms of the opportunity cost of not installing wind turbines).

Unconstrained Baseline
By 2030, Austria generates the bulk of its electricity from intermittent sources. Given our baseline assumptions and carbon price of A C25/t CO 2 e, this comes at a total system cost of A C2.56 billion. Table 7 summarizes the breakdown of the total system cost.
Due to the large share of electricity generation from renewable sources, operation and maintenance cost of A C1 billion exceeds spending on fuels and emission certificates required by thermal power plants. Burning fossil fuels gives rise to 7.1 Mt of CO 2 emissions, which are priced at A C177 million.
Thermal generation capacity, running on natural gas, waste, and biomass, is not expanded in the baseline scenario. Hence, the annualized investment cost is entirely spent on the expansion of renewable electricity generation capacity. In our baseline scenario, 10.6 GW of onshore wind turbines are added to the power plant stock, while no investment in solar PV or storage technologies, such as batteries and hydrogen storage, occurs. In total, 59.7 TW h electricity is generated is a necessary consequence of the government's policy target. While (almost) as much electricity is generated from domestic renewable sources as is being consumed, fossil thermal generation units are still required to generate heat and to substitute for renewable generation in times of low electricity generation from intermittent renewable sources.
In the unconstrained baseline scenario with all additional renewable electricity generation coming from wind power, Austria is a net exporter of electricity throughout most of the year. Minor net imports occur in January, as is shown in Figure 1 a. Net exports are relatively stable over the year, as the seasonality of onshore wind power generation is favorably complementary to generation from pre-existing run-of-river hydropower plants (see Figure 2), which historically dominate Austria's power generation mix. Austria tends to export electricity to Germany when domestic wind power generation is high (Pearson correlation coefficient r(X AT , wind AT ) = 0.46, significant at the 1 % level for hourly data), and to import electricity from Germany when German electricity (excluding the cost of CO 2 emissions) amount to A C226 million.

Restricting wind power
With the gradual restriction of wind turbine additions, electricity is increasingly sourced from solar PV. This induces changes in the composition and the operation of the electricity system. As long as the policy constraint of generating a given amount of electricity from renewable sources is binding, any decline in wind power generation must be offset by a corresponding increase in power generation from an alternative renewable power generation technology. Thus, the installed capacity of solar PV increases by 1.98 GW for each GW reduction in installed wind power capacity. The corresponding change in investment cost is indicated in Figure 3. eration mix reduces the variance of total intermittent electricity generation somewhat, while the impact on minimal intermittent generation is minor. In effect, resource adequacy can be achieved with slightly lower thermal capacity in our baseline scenario. More thermal capacity can be decommissioned so that O&M cost decreases slightly. This reverses as wind power is phased out completely. The variance of intermittent electricity generation is being reduced only marginally, but minimum intermittent generation declines stronger. Thus, an electricity system dominated by solar PV requires slightly larger thermal generation capacities. In effect, fewer thermal units can be decommissioned so that system-wide O&M costs increase.
Changes in the cost of fuel and CO 2 emissions are closely linked to trade flows. As wind power deployment is gradually limited to 6 GW from the initial unconstrained optimum, net ex- balance with Germany by A C95 million. The move towards electricity generation from solar PV also leads to a shift in the seasonal pattern of electricity generation, imports, and exports. Figure 1 shows the monthly electricity generation and electricity exchange for the two extreme scenarios in which additional adjustable renewable electricity generation is entirely sourced from wind power Given these estimates, we can calculate the valuation of undisturbed landscapes implied by announced government policies. The government intends to source an additional 10 TW h annually from wind power, which corresponds to an additional 5.05 GW wind turbines being installed.
At this level of wind power deployment, the opportunity cost of wind power averaged over all considered CO 2 price scenarios is approximately A C22 500/MW and year. The present value of the cost of undisturbed landscapes implied by government policies amounts to approximately A C1.3 million over the lifetime of a standard 3.5 MW wind turbine at a 5 % discount rate.

Sensitivity Analysis
To assess the robustness of our results, we analyze the sensitivity of model outcomes to changes in parameters that can be expected to have a particular influence.  prices of A C50/t CO 2 e or lower, this decline is particularly pronounced, with a reduction of the optimally deployed wind power capacity by approximately 0.6 GW for each A C10/kWp decline in the capital cost of solar PV. At capital costs of A C400/kWp or lower, solar PV covers all additionally required renewable energy generation if the price of CO 2 is at A C25/t or lower. For a carbon price of A C50/t CO 2 e, about 2 GW of wind power is deployed in optimum at this capital cost. As capital costs fall further, deployed wind power capacity descends slowly, approaching zero as PV capital cost reach A C275/kWp. With CO 2 price at A C75/t or higher, when government policy is not binding anymore, deployed wind power capacity remains sturdy. Even at a capital cost of PV as low as A C275/kWp, added wind power capacity tops 6 GW. Moreover, it is worth noting that the impact of the lower PV capital cost on CO 2 emissions largely depends on the prevailing CO 2 price level.
With emission prices of A C25/t CO 2 e or lower, a halving of PV capital cost 0 /t CO 2 25 /t CO 2 50 /t CO 2 75 /t CO 2 100 /t CO 2 In line with less wind power capacities deployed, the opportunity cost of wind power declines.
Given PV's capital cost of A C560/kWp, a 10 % reduction from our baseline, we estimate the opportunity cost of wind power at 5.05 GW added at A C17 500/MW wind power avoided, a 21.9 % reduction compared to baseline. However, the reduction in opportunity cost is smaller when restrictions on wind power deployment are tighter. The last megawatt of wind power foregone comes at an opportunity cost of approximately A C27 200, as long as emitting a tonne of CO 2 costs A C75 or less.

Sensitivity to transmission constraints
Following a decision by ACER, electricity transmission capacity between Austria and Germany was lowered from an initially (virtually) unlimited capacity to 4.9 GW in 2018. In consequence, the previously unified Austro-German electricity market zone was split up, allowing Given that current energy policy goals turn Austria from a net importer into a net exporter of electricity, we have analyzed the effects of a potential lifting of ACER's decision to cap electricity transmission by setting transmission capacity to 10 GW. Under this assumption, the opportunity cost of wind power is higher compared to baseline at low penetrations of solar PV. However, the increase in opportunity cost as the PV penetration increases is less pronounced compared to baseline.
Higher transmission capacity allows (fossil) power plants to generate and export electricity at times when interconnectors with lower transmission capacity would be congested. In consequence, use of fossil fuels for thermal power generation and CO 2 emissions are higher (by 390 kt to 670 kt at a carbon price of A C25/t CO 2 e) compared to our baseline with lower interconnection capacity.
Moreover, thermal capacity is not being shut down (as in the baseline scenario), so that changes in O&M cost are very small.
Higher interconnector capacity allows for higher quantities of imports and exports. Indeed, Austrian imports are up by 3.1 TW h/a, while exports increase by about 2.6 TW h/a, compared to the unconstrained baseline scenario.
Even though net exports decline, Austria's (net) export surplus increases, mostly on the back of a lower average import price for electricity from Germany. Moreover, increased interconnector capacity leads to a trade balance that is less responsive to changes in installed RET capacities.
Thus, the opportunity cost of wind power are higher at at high wind power penetration, but lower at low wind power penetration (see Figure 7).

Limitations of the Analysis
It is worth noting that our analysis is not spatially resolved, which has important consequences.
Our analysis of the opportunity cost of wind power does not depend on a specific spatial allocation of RET or a specific grid topology. Hence, our results are generalizable to the extent that any assessment of specific RET expansion plans must consider the effects of spatially distributed RET on resource quality and grid cost. However, there are good reasons to believe these factors will not change the results of our analysis substantially. First, we implicitly assume that wind Lastly, in our analysis we focus on the time-horizon up to 2030. Thus, our findings are not necessarily transferable to the (almost) fully decarbonized electricity systems we might establish in the more distant future. Yet, fundamental properties of renewable resources, such as seasonal and diurnal patterns in wind speeds or solar irradiation, will persist over these time frames. Thus, we expect to find positive opportunity costs of wind power also in a decarbonized Austrian electricity system, particularly as future options to facilitate the system integration of variable renewables that we did not model, are unlikely to come at significantly lower cost than the included options.

Conclusions and Policy Implication
Not disturbing landscapes can come at considerable opportunity cost, as our analysis reveals.
How high this cost is, depends on the valuation of CO 2 and the cost of the best alternative to wind power. If the cost of CO 2 is low and the best alternative solar PV is realized at utility-scale in open space, wind power's system cost advantage vanishes in Austria. If, on the other hand, the cost of CO 2 is high, if open-space PV itself causes negative external effects at the local level, or if there is a preference for rooftop solar PV, the opportunity cost of wind power is worthy of consideration.
Given the current policy target of adding capacities for generating 10 TW h wind energy, the present value of the cost of undisturbed landscapes amounts to at least A C995 000 over the lifetime of each 3.5 MW wind turbine not erected. 9 While these numbers indicate that in the absence of local wind turbine externalities, the government targets for RET deployment aren't economically efficient, a complete policy evaluation must weigh the cost of undisturbed landscapes against its benefits arising from the preservation of current landscapes aesthetics, for example. Yet, any significant expansion of renewable electricity generation technologies will interfere with landscapes.
Relying completely on open-space solar PV would require about 315 km 2 land 10 for replacing the approximately 3 030 wind turbines of the 3.5 MW class that would be needed to meet policy goals by wind power alone.
For a comprehensive policy evaluation, our findings need to be complemented by a spatially highly resolved analysis of the local cost of disturbed landscapes. Together, such findings would allow determining the socially optimal deployment of renewable energy generation technologies.
Here, we see a fruitful field for further research.
Attention should also be paid to the distributive consequences of wind power expansion in Austria. As we have shown, a large-scale expansion of wind turbines comes with an increase in electricity exports, i.e. at high penetration some wind power capacity is mostly used to generate electricity for export. Revenues from these export are accruing to the owners of power generation assets, while negative external effects of wind turbines have to be borne by local residents. This imbalance could be narrowed down by policies requiring wind turbine owners to share wind turbine income with affected residents, either directly or through transfers. Fair sharing of benefits and burdens has the potential to foster acceptance of wind power (Scherhaufer et al., 2017). higher the price of CO 2 , the stronger the incentive for investment in renewable energy technologies. Hence, there is a CO 2 price that is equivalent to reaching the same emission reduction. In our baseline scenario, the policy reduces CO 2 emissions by 1.2 Mt versus the same scenario without the policy constraint. An emission reduction of similar size could also be reached through raising the CO 2 price to approximately A C40/t. However, the resulting system is not equivalent in terms of installed capacities of electricity generated from domestic renewable sources. Reaching this policy objective would require a carbon price of approximately A C63/t CO 2 e. For emission prices above this level, the government policy is not binding anymore. A major difference between a binding government policy objective and a hypothetical CO 2 price lies in the measures' effect on electricity trade. Under the renewables target, Austria necessarily turns into a net exporter of electricity and benefits from the associated revenues. Under CO 2 pricing, Austria would remain a net importer of electricity and maintain a negative balance of payments in the electricity sector up to a CO 2 price of approximately A C45/t.
Appendix C. Description of the power system model medea