Climate Change and the Vulnerability of Germany’s Power Sector to Heat and Drought

The effects of extreme weather events and the resilience of the energy sector have become the subject of regulatory initiatives and ongoing research. We demonstrate the vulnerability of the German power sector to climate change and provide a qualitative and quantitative analysis of emerging risks from two types of extreme weather events: droughts and high temperatures. Our analysis is based on datasets covering temperature and drought data for the last 40 years. We present evidence of a higher frequency of power plant outages as a consequence of droughts and high temperatures. To characterize the vulnerability of the power sector we develop a capacity-adjusted drought index. The results are used to assess the monetary loss of power plant outages due to heatwaves and droughts and losses to consumers due to higher wholesale electricity prices and price volatility. An increasing frequency of such extreme weather events will aggravate the observed problem.


INTRODUCTION
The link between climate change and increasingly frequent extreme weather is gaining the international recognition of policymakers and the attention of the scientific community.The IPCC special report defines extreme weather events as "risks/impacts to human health, livelihoods, assets and ecosystems from extreme weather events such as heat waves, heavy rain, drought and associated wildfires, and coastal flooding" (IPCC 2018, p. 11).Climate change is already posing new region-specific challenges for technological and socio-economic systems.The summer of 2018 provided a preview of possible adverse developments.Heat waves were observed in North America, Western Europe, and the Caspian Sea region, while rainfall extremes occurred in Southeast Europe and Japan (Kornhuber et al. 2019).The summer was characterized by an enduring heat wave accompanied by droughts in various regions throughout Europe that lasted until the end of autumn (EC 2019a).During this period, France and Germany reported cuts in nuclear and coal-based electricity generation.Wholesale market prices were highest in Italy and Spain, where weather forecasts predicted temperatures to rise (Platts 2018).
The Energy Journal,Vol. 43,No. 3.This is an open access article under the terms of the Creative Commons Attribution License (CC-BY), which permits use, distribution and reproduction in any medium, provided the original work is properly cited.All rights reserved.
Many studies have focused on long-term temperature-sensitive impacts on the demand side (Rivers and Shaffer 2020), impacts on wind and solar resources, the deterioration of the water supply for cooling, and the lowered potential of run-of-river generation (Cronin, Anandarajah, andDessens 2018, Chandramowli andFelder 2014), or the impacts on transmission capacity (Woerman 2019).Since most of these studies emphasize uncertainty stretched decades into the future, we attract attention to a new aspect of interest for the study of electricity markets: the spatial dimension of short-term planning in the German power market.An assessment of installed generation capacities and their exposure to weather extremes can serve as a starting point for the planning and future design of the electricity sector, including investments, renovation, and phase-outs.Our analysis can be replicated for other countries depending on the availability, granularity, and transparency of weather and power system data.
This study aims to provide a foundation for identifying measures that reduce the electricity system's overall vulnerability, thus supporting energy transition policies that account for current climate change effects.Our results can be applied when assessing the need to renovate or decommission old capacities, considering the costs of renovation.The results of our analysis are essential for the further calibration and estimation of economic damage functions for specific locations of thermal power plants, which are necessary to estimate the future costs from climate change.The paper attempts to provide a new methodological approach to parameterize the effects of extreme droughts and temperatures.The analysis requires meteorological and electricity market data with a high geographical and temporal resolution, therefore, the main limitation of the analysis results from the availability of data.
The paper is organized as follows.In Section 2, we present the key challenges that climate change effects pose to power sector infrastructure.Section 3 describes the main datasets employed in this study.The analysis, methods applied, and results of the analysis are presented in Section 4. Section 5 highlights the main discussion points underlined in the study.We discuss policy in Section 6.

Key power sector challenges posed by climate change effects
Climate change affects all countries to varying degrees.Climate change puts additional stress on developing power systems in regions with more rigorous initial climate conditions.For example, immediate climate change effects are more potent in southern regions with existing water scarcity and hazards that reflect poor long-term land-use practices.This can be observed in the Sudan Sahel region of Nigeria (Umoh and Lugga 2019), Bangladesh (Khan, Alam, and Alam 2013), and Pakistan (Alkon et al. 2019).Adverse extreme events, such as heat waves and droughts, have become particularly intense in Northern and Central Europe over recent years (EDO 2015, 2018a, Magagna et al. 2019).
Another important regional factor that affects electricity systems' resilience is the policy framework that sets targets, regulations, and emission reduction goals for the electricity sector.In this respect, Germany is a good example of a complex system that is on the path to a low-carbon future and is subject to climate change.This section will analyze the key aspects of a wide range of effects associated with climate change in the light of recent developments in policy and research.Our motivation for the analysis is driven by the findings described in the following paragraphs.

Elements of the European policy framework addressing extreme events
The European electricity system is becoming more internationally integrated, which requires higher resilience and cooperation between member states.However, this process also redistributes the risks of cross-border failures in generation and transmission capacities (EC 2016a).At the European level, the threats posed by extreme weather events are addressed in multiple ways.In June 2019, the EU adopted the regulation on risk-preparedness in the electricity sector, an important component of the EU's activities to counteract the risks of extreme events (EC 2019b).Under Article 5 of this regulation, the European Network of Transmission System Operators for Electricity (ENTSO-E) is required to submit a proposal for a methodology to identify "the most relevant regional electricity crisis scenarios" to the European Union Agency for the Cooperation of Energy Regulators (ACER).Concerns about the vulnerability of energy systems are also expressed in the European Commission's program for critical infrastructure protection (EPCIP), which regards both the energy and transport sectors as "European critical infrastructures" (ECIs) (EC 2008).While the program pays specific attention to terrorism, it highlights the vulnerability of network infrastructures in case of extreme natural events that "are not constrained by international borders." The European regulatory framework also plays an important role in national-level member state policies regarding the availability and usage of cooling water for power plants.The Water Framework Directive (EC 2000), which came into force in December 2000, provides a common framework for the management and preservation of European water networks.In 2006, it was replaced by the European Freshwater Directive (EC 2000), which sets limits for the amount and maximum temperature of the mixed water at the discharge point and for the maximum heating range.In addition to electricity generation, inland waterways are also recognized as ECI, which reflects the perceived risk of failures in the bulk delivery of energy carriers and other goods; the disrupted delivery of energy sources such as steam coal poses a risk separate from the scarcity of cooling water.According to the German Coal Importers Association (Verein der Kohlenimporteure e. V.) (Cieslik and Wodopia 2016), up to 50% of imported coal is transported via domestic waterways (see Appendix, Table A 1).The complexity of the interactions between components of the power system was addressed by the "new approach" to making ECI "more secure", introduced by the EC in 2013 (EC 2013), which explicitly addresses the interdependencies between critical infrastructures and industry (Forzieri et al. 2016).
In 2018, the EC established reporting requirements for the member states about their national adaptation actions (particularly the availability of cooling water for power plants) to assess the effects of introduced climate policies and recent climate change impacts (EC 2018).the EC found that measures to create a decarbonized electricity system may even increase the system's vulnerability to extreme weather events, such as carbon capture and storage (CCS) technologies.The EU's emission reduction goal involves increasing the share of variable renewable energies alongside some fossil-based generation that provides balancing power to the electricity system.Carbon capture techniques play a key role under these conditions.The most recent available EU reference scenario projects moderate CCS development until the end of 2050 (EC 2016b).However, while carbon capture deployment allows decarbonizing the electricity system, the water intake of power plants equipped with this technology is considerably higher than that of conventional plants (Byers et al. 2016).This implies an increasing threat of insufficient cooling water availability during droughts and high ambient temperatures, as well as a need for additional investments in cooling technologies.

Key aspects of the impact of heat and drought on the power sector
Table 1 provides an overview of the scientific studies that have elucidated extreme weather events' impacts on power systems and interconnected services, such as the delivery of energy carriers.We focus mainly on heat and drought as the most critical among the other climatic hazards as shown by (Forzieri et al. 2016).In this study, the authors estimated the damage caused to the European energy, transport and industrial sectors by the extreme weather events storms, coastal and river floods, forest fires, droughts, cold and heat waves.They conclude that by the end of the 21st century drought (67%) and heat damage (27%) in Europe will account for 94% of all hazard impacts to the energy sector, which will also experience the highest increase in damage among all sectors (Forzieri et al. 2016).The studies included in Table 1 used different temporal and spatial resolutions and various methods to analyze the effects on the aggregate power sector, including several mathematical modeling approaches.They are based on the assumptions of long-term temperature and  (Hoffmann, Häfele, and Karl 2013, Altvater et al. 2012, van Vliet, Vögele, and Rübbelke 2013, Ecofys 2014, Byers et al. 2016) precipitation trends, as well as structural changes in the electricity demand and future generation mix.In contrast, we assess the increased risk of extreme climate effects at a high spatial resolution in the medium and short term, when they represent the most severe threat to system security, as mentioned in (Boston 2013).
Studies also show that oligopolistic structures in the electricity market can lead to the exercise of market power.(Ghazvini et al. 2013) show, for example, that the planning of maintenance can be influenced by strategic decision-making.The presented descriptive analysis of the relationship between weather conditions and outages does not, however, make it possible to depict strategic decision-making situations, which is why the authors exclude this factor in the further analysis.Based on historic data from 1971 to 2000, van Vliet et al. (2012) ) highlighted tense conditions regarding the availability and high temperatures of cooling water for a future scenario from 2031 to 2060.Similarly, van Vliet, Vögele, and Rübbelke (2013) investigated the effects of water constraints in European electricity markets and their effects on wholesale electricity prices.Hoffmann, Häfele, and Karl (2013) used a control period of 1960 to 1990 for climate data in order to assess the water-electricity nexus for two future periods, 2011-2040 and 2041-2070.Both studies applied a comprehensive approach to estimate the European electricity sector's risks under new climate challenges.They accounted for climate change and its cumulative effects until the end of the 21 st century.

Main data sources
Any analysis of the interrelation between the security of the electricity supply and extreme weather events must account for a diverse set of factors.These include the reduction in available generation capacities, long-term data on temperatures and droughts, and electricity price peaks.In the current study, we aim to evaluate environmental conditions and associated risks by analyzing meteorological and electricity market data at each power plant location.Table 2 presents a summary of the main data sets.Two stress factors for the electricity system, which are caused by extreme weather conditions, are taken into consideration: (i) droughts and (ii) high temperatures (heat waves).Drought data is used to evaluate hydrological droughts, which can affect power plant cooling systems.Precipitation data with a high spatial and temporal resolution is derived from the soil moisture index (SMI) database of the German Drought Monitor (GDM) (Zink et al. 2016).The GDM provides drought data in a 4-km raster grid.Following the SMI specification, the index ranks the severity of droughts from 0 to 1.The index displays daily data and dates back to 1951.A more detailed description of the SMI is provided in Appendix B. The second meteorological dataset, which is used to identify heat waves, is the Climate Data Center of the German Meteorological Service (DWD).In this dataset, temperature data is available at hourly, daily, monthly, and yearly resolutions.
The Transparency Platform of the ENTSO-E lists individual generator and power plant (utility) outages in the European electricity market.The outage data includes both full (shutdown) and partial outages.A more thorough overview of the outage data set can be found in Appendix B. Information on the outages, the associated power plant names, and their Energy Identification Codes (EICs) published by (ENTSO-E 2020) is matched against a database of German power plants, made available by the German Federal Ministry of Economic Affairs and Energy, which contains the geographic location of the individual plants.
Thus, we focus on the impacts of extreme weather events on the German electricity market and quantify the physical effects of climate conditions on thermal power plants' production schedules.This study aims to present relevant results for identifying measures that reduce the overall vulnerability of the electricity system and to provide guidance for policy measures on how to design the energy transition while accounting for the current effects of climate change.

Brief survey of recent drought and high temperature periods from the data
In the short term (approximately 20 years), the growing frequency of extreme events and the persistent increase of the global temperature trend represents additional stress to the electricity generation and transportation systems.Most of the studies in Table 1, following the studies in (Chandramowli andFelder 2014, Cronin, Anandarajah, andDessens 2018) focus on long-term mean surface temperature changes, disregarding the intensity of extreme events.However, the latter tend to increase in time and magnitude.to illustrate the shift in the probability density of the SMI data between two periods shown in Figure 1a.The necessary actions may diverge with consideration of the different time frames of climate change impacts.For example, (Rübbelke and Vögele 2013) analyzed local weather changes that affected nuclear and hydropower plants under the assumption that gas-and coal-fired plants will be used to fill supply gaps due to restricted water availability.Their findings predicted and outlined the current situation on the power market as of summer 2018 (Platts 2018), with heat and wind lull generating temporary profits for coal generation to such an extent that its availability also shrank as temperatures rose.In the last decade, an assessment of adaptation measures for the energy sector highlighted the high annual costs required to deal with climate change impacts in Europe (Altvater et al. 2012).The trade-off between investing in new cooling systems and incurring losses is highly dependent on the location of power plants and regions with high electricity demand.The high granularity of the analysis presented here allows identifying at-risk capacities through multiple criteria, including exposure to restricted cooling water and high ambient temperatures.Focusing on the German power market as a case study is justified because the market has a central geographical location, a high volume of interconnections with neighboring markets, and the highest installed power plant capacity of the European member states.Figure 1b relates the locations of power plant capacity types to regions with higher (green) and lower (red) levels of moisture relative to the long-term trend.The color of the dots represents the power plant type, while the diameter indicates the installed capacity in megawatts of each capacity type.The effects of extreme weather events on thermoelectric generation in Germany can be observed during the heat wave and drought in 2018, which resulted in multiple forced shutdowns of thermal power plants in some areas of Germany (Platts 2018).The power plants, their location and installed capacity are shown in Figure 1b.Hot and dry summers from 2015 to 2018 resulted in average annual moisture levels well below the long-term averages of 1951-2015 for most regions, except the northern seashore (Figure 1b, green).While some regions exhibited only minor changes (Figure 1b, white), most of central Germany experienced extremely low precipitation levels.In light of these alarming observations, it is essential to better understand the effects of climatic changes on the stability of future power systems and the security of the electricity supply, especially as extreme weather conditions intensify.In the following analysis, we assess the risk of future extreme events, in particular droughts and heat waves, on the German power system.

Assessing the risk of droughts in the near future
The distribution of mean SMI in most parts of Germany changed considerably during the period from 1995 to 2018.The investigated sample accounts for 307 geographic locations with thermal power plants that were operational as of 2019.SMI is itself an uncertain variable.For each site, it is known up to a certain probability distribution.The data analysis revealed some important changes in the shape of the distribution.The probability of overall leftward shifts (Figure 1) indicates a reduction of the mean value of SMI over time and a decrease in standard deviation.This tendency is observed for about 80% of all locations (Figure 2).This means that, over time, the power generation fleet has been exposed to increasingly harmful drought impacts when we consider the availability of water for cooling purposes.
The SMI dynamics imply that for the geographic locations represented in Figure 2a, droughts have become more severe within the last 23 years.The risk is significantly higher than that grasped by studies estimating the effects of climate change on water availability and the generation system mentioned in the introduction (van Vliet, Vögele, and Rübbelke 2013, van Vliet et al. 2016, Rübbelke and Vögele 2011).The volatility of the SMI for each location shows a trend toward consistently drier weather conditions in most parts of Germany and an increasing risk of severe droughts for particular regions (Figure 2b).Nearly 300 sites were considered in this analysis, focusing on the locations of thermal power plants.A continuous exacerbation of droughts will expose the power sector to even higher climate related risks in the future.

Cointegration of droughts and temperatures
In addition to this increase in the frequency and duration of droughts, temperatures will also increase due to climate change.Based on the IPCC's projection for the 21 st century, fewer cold temperature extremes and an increase in mean temperatures can be expected (Collins et al. 2013).Furthermore, it is expected that heat waves will occur with a higher frequency and duration with summer temperature extremes over central and southern Europe.Our analysis reveals connections between temperature increase and SMI decline in Germany.
In the following steps, it is necessary to increase the granularity of our analysis.The SMI time series for the chosen locations have a monthly time resolution.In contrast, data on temperatures is available at an hourly scale and is closer to the time resolution of the data on outages.From Figure 3, we can assume that both SMI and temperatures monthly mean time series are negatively correlated-high temperatures occur at low SMI values.It is necessary to prove this relationship formally since it is sensitive to the location, type of vegetation and season as shown by Karnieli et al. (2010).We applied the Johansen test to verify cointegration, thus testing the relationship of two time series: temperatures and droughts.Temperature data is provided by the German Meteorological Service for the period from 1951 to 2018 as monthly mean values of the average soil temperature at a depth of 5 cm.In the framework of the test, the null hypothesis () is that , which assumes no cointegration of the time series.For all the locations considered, the test statistics exceeded the critical test value at the 1% significance level (Table A 2).Thus, we have strong evidence to reject the null hypothesis of no cointegration.We also rejected the hypothesis of the second test, , since the test exceeded the 1% significance level.The results of the Johansen test indicate a significant connection between temperatures and droughts.We estimated the probability distribution and risk of extreme dry events based on data covering the last 70 years for German territory.Referring to the SMI analysis, the results of the cointegration test, and the rising global temperature trend shown in the IPCC report, we conclude that droughts and temperature extremes will become more severe (Figure 2b).The likelihood of low precipitation levels has already almost doubled for particular locations: compare the right-hand side values of the standard deviation curves for SMI 1951SMI -1995SMI and 1951SMI -2018 in Figure 2b.The result of the cointegration test suggests that rising global temperatures will result in a decline of SMI and increase the probability of severe droughts.
Furthermore, temperatures and droughts are unambiguously related to power plants' forced outage rates.Low temperatures (as in the European cold spell in January 2017 and February 2018; (see Schulze et al. 2018)), as well as the combination of high summer temperatures with low precipitation levels during summer and early autumn (Figure 3), coincide with high levels of forced outages of thermal power plants and pump storage sites.The level of exposure to extreme events varies between the regions and depends on their location with some regions, especially exposed to droughts (as will be shown later in Figure 5).In contrast, the cold spells impose a different kind of disruption as discussed by (Bartelet and Mulder 2020) in connection with the European natural gas market.The insufficient temperatures lead to an increase in heating demand and, together with interruptions in natural gas transit, can have a significant impact on day-ahead gas prices, as happened in February and March 2012 and 2013.The impact of the cold snap extends beyond national borders and affects energy demand in exporting and importing countries.

Droughts, high temperatures, and forced outages of thermal power plants
The scatterplot below in Figure 4a maps daily temperatures and the number of forced outages of thermal power plants per unit of time in all locations in Germany from 2015 to 2018.
Nearly 30 outages of thermal capacities (primarily coal plants) in the period from June to August for the years 2016-2018 coincide with the hottest days.The density of outages for each fuel type shows that nuclear capacities tend to have more outages at higher temperatures compared to low and mid-season.Moreover, outages that occur in summer or subsequent dry months of autumn tend to last longer, reaching nearly 20 days (see Figure A 1(a and b) in the Appendix A).The U-shape of density plots against the moderate mid-season temperatures show that hard coal, lignite, and gas capacities also have a seasonal pattern, supporting seasonal outage behavior.The frequency of outages (y-axis), defined here as the number of outages at the same moment of time as reported to ENTSO-e, is distinctively higher at warmer temperatures.The density plot in Figure 5b highlights the difference in outage rates between the years with heat waves (2015,2018,2019) (EDO 2018a) and the years with relatively mild weather conditions in Germany.In 2019, gas, nuclear, and coal capacities experienced multiple outages in the summer months, illustrating a distinctive effect of the 2019 heat wave.The curvature on the left side in Figure 4a is the result of many factors, representing cold, dry spells that appear "when the minimum temperature is below the ten-percentile threshold, and the maximum temperature is below its ten-percentile threshold" (EDO 2018b) combined with reduced solar availability and low levels of water due to hydrological droughts in previous periods.Similar factors explain the curve of the curvature of the density plot Figure 4; the warm periods produced heat waves where "the maximum temperatures are above its ninety-percentile threshold, and the minimum temperature is above its ninety-percentile threshold" (EDO 2018b), accompanied by wind lulls and increasing demand for cooling and air conditioning.As described in the introduction, this study focuses on the implications of climate change for the supply side of power systems.Thus, we disregard effects on the demand side, such as shifts in demand structures or increases in overall electricity demand.

Estimating capacity-and location-specific costs of extreme weather events
The available data does not allow a comprehensive analysis of the cost of climate change for power generation systems.However, we established a robust connection between a shortage of water resources and price volatility in the power market-this section will discuss the method and introduce our findings.Also, we found a notable increase in the frequency and duration of forced outages in response to low SMI and high ambient temperature.These conclusions are drawn through an option pricing model that relates monetary risk to an uncertainty measure called "volatility".Volatility is the standard deviation of the one-period percentage change in the price of the underlying asset.Volatility of the electricity spot market prices creates an additional financial burden on consumers.One way to calculate the economic damage of price volatility is to estimate an implied value of call options on electricity (for the application of options pricing methodology to calculate the financial cost of price volatility, see Cooke and Golub (2020) for the application of real options analysis to calculate the risk-adjusted cost of climate change, see Anda, Golub, and Strukova (2009), Golub and Brody (2017)).To guarantee price stability, consumers should pay an insurance premium that is equal to the value of a call option with a strike price equal to the expected price of electricity.Higher volatility implies higher risk-adjusted electricity costs.In addition to price volatility, there are direct economic costs of forced outages calculated as lost revenue due to a decrease in power production.
Plotting the generation capacity adjusted SMI index against electricity price volatility in Figure 5 reveals another potentially quantifiable element of economic damage due to droughts: for example, risk associated with price volatility using the methodology for risk-adjusted price calculation presented by Cooke and Golub (2020).Figure 5 links low SMI and price volatility.The capacity-adjusted aggregated SMI index is the sum of the products of site-specific SMI indexes multiplied by the share of generation capacity located in the site's proximity.That is, for each type of power generation, the aggregated index is the sum of site-specific capacity-adjusted indexes.The index reflects the share of installed capacities in the German generation mix.A low value of the index indicates that the entire national energy production by a specific technology (i.e., coal, gas, nuclear, etc.) is more exposed to drought.The aggregated index reveals the difference between capacity types most likely to be exposed to the risk of droughts.Hard coal and lignite, accounting for nearly 44 GW (as of 2020) of installed generation capacity, appear to be especially sensitive to cooling problems during droughts.Nuclear, with "only" 9.5 GW installed as of 2020, and gas, with 29.8 GW, behave relatively similarly, justifying rising concerns for the future of gas-fired capacities under current nuclear and coal phase-out policies in Germany.Figure 5 allows comparing price volatility (bars) against the changes in the SMI index (lines).Although only relatively short time series are available, we can see that the lower value of the index corresponds to higher price volatility and therefore to a higher risk-adjusted cost of electricity.Winter 2017-2018 seems to be an outlier in this analysis, with December 2017-January 2018 being warmer than usual in Europe until February arrived with temperatures colder than the 1981-2010 average for that month (Platts 2018).
Besides forced outages, significant changes in the fleet of generation capacities may also affect price volatility.To account for this factor, Table A 3 in the Appendix provides an overview of the commissioning and decommissioning of power plant types in the German power market.Almost 4 GW of thermal power plant capacity was decommissioned from the market in 2017, while nearly 8 GW variable wind capacity was added to the generation mix in preceding years.
So far, we have focused on the occurrence of capacity outages.However, the duration of outages and the amount of capacity that is excluded from the merit order is equally important.To determine the monetary loss of the outages, we can estimate the proxy value of the capacity removed from the market by multiplying the given quantity by the spot market electricity price and subtracting the variable generation costs for each technology type (for detailed cost estimations, see Table A 4).The electricity price is an hourly spot market price from European energy exchange (EEX) available for the period from 2015 to 2018.The following analysis is highly aggregated and serves only to estimate the proxy generator's lost value due to the capacity withdrawn from the market.Table 3 describes the order of values for the estimated indicator.The year 2018 is the most distinctive in this period, reaching nearly 127 million euros, with 37% of these costs being incurred in the summer months.For nuclear generation, the share of value "lost" from June to August is more significant and reaches 63%.Nevertheless, extremely high temperatures do not always cause losses.The gradual advancement of the heatwave and low wind availability in Europe in June 2018 resulted in additional income for German nuclear and coal power plant operators, who ramped up exports to the north (reduced hydroelectric generation) and south (reduced thermal power plant availability) of Europe (Platts 2018).These conditions held only until mid-August, when temperatures rose, and nation-wide thermal plant restrictions came into force in Germany.Whether gains outweigh losses depends on two factors: how German thermal power generators adapt to extreme weather and how heat waves develop, both spatially and temporally, while the share of renewable generation increases.The proxy costs of the forced outages can be compared to the cost of dispatch and feed-in management of renewables, which reached 351.5 million euros in 2018 (see Table 4, BNA 2019b).Re-dispatch will increase in the coming years due to the integration of European electricity markets, the share of the variable renewables, delays in grid expansion, and nuclear and coal phase-out by 2038 (Hirth et al. 2019).Similar reasons affect the "loss" of thermal generation capacities, adding weather extremes to the named factors.Plotting the index vs risk-adjusted price of electricity (see Figure 5) reveals another potentially quantifiable element of economic damage due to droughts.However, a precise analysis requires more detailed data on the wholesale electricity market such as detailed hourly data on power production by individual power plants as well as exports and imports between German regions and within the EU electricity market.This kind of analysis could be a good candidate for the future work.

DISCUSSION
This study provides an in-depth analysis of the risks of climate change, and specifically extreme weather events, for the German electricity system.However, our results can be generalized for other North European countries undergoing an energy transition process.We identify regions with higher hazard levels for generation capacities by introducing a capacity-adjusted drought index.
The effects of climate change negatively affect power systems.Focusing on Germany, we show that the occurrence of heatwaves and droughts will significantly increase in the coming years.We estimated their likelihood and the range of risk at high granularity, assessing the locations of thermal units.Meteorological trends suggest that hydrological droughts and water scarcity, coupled with heatwaves, became more severe within the last decades.Assuming this trend continues, future power systems have to be adjusted to increase their resilience.At the same time, many technological options that are targeted toward decarbonizing future electricity markets (e.g., renewable energy sources, CCS) will be affected by changing climate conditions themselves.
Several studies have assessed the costs of cooling systems for thermal power plants.Due to the nature of such facilities, cost estimations are highly location specific.However, they can be used as a point of reference for estimating necessary investments for the current fleet.In general, once-through water options tend to be the most cost-efficient solutions, while dry-cooling technologies are the most expensive solutions by far.Wet cooling systems with recirculation increase water needs and are restricted by water availability and temperature regulations for intake and discharge.Alternatively, dry cooling is of particular interest since it can be utilized at low water availability.However, as this method uses ambient air to cool exhaust steam, it is exposed to the risk of high temperatures in summer seasons (DOE/NETL 2008, Burillo et al. 2019).Table 4 displays the cost assumptions of several studies, differentiating between power plant technologies and cooling technologies.Adapting existing power plants by investing in more efficient cooling technologies such as dry cooling or alternative configurations of wet cooling towers incurs additional annual generation costs (see Table 4).Moreover, adaptation creates higher operational costs and reduces efficiency, e.g. by 2% (Altvater et al. 2012).In the case of nuclear power plants, every 1°C increase of ambient temperatures above a 21°C threshold causes a 2.2-2.5% reduction in power output (Linnerud, Mideksa, and Eskeland 2011).In the case of Germany, these additional costs occur for capacities that are subject to phase out in the next decade (Table A 5).The necessity of these investments should be evaluated by taking into account the risk of extreme weather events, the growing share of variable renewables, and targets for interconnection capacities between the regions.Long-term investment decisions should rely on a spatially explicit understanding of infrastructure's vulnerabilities to climate change (Burillo et al. 2019, Kennedy andCorfee-Morlot 2013).

CONCLUSION AND POLICY REMARKS
Considering the recent developments of the German power market, coal-fired and nuclear power plants will be decommissioned within the next decade (Rinscheid andWüstenhagen 2019, Jahn andKorolczuk 2012).With nuclear going offline in 2023, arguments for retaining coal-fired capacities remain valid until 2038 (see Table A 5). (Altvater et al. 2012) pointed out that overall, 637.3 million euros per year.will be necessary to adapt European power plants to changing climatic conditions.For Germany, they estimated 8.8 million euros p.a., accounting only for the long-run cooling needs of gas power plants.
According to our study, most of Germany has experienced an exacerbation of droughts over the last 40 years.With ongoing climate change, this strand will likely continue to the future.Our analysis reveals the vulnerability of the German power generation system to extreme weather events like droughts and heatwaves and explores some approaches for the calculation of economic damage due to outages and excessive price volatility on the wholesale electricity market.The post-COVID economic recovery and European Green Deal will shape the transformation of the European economy including rapid decarbonization of power generation.The changing external conditions should be taking into account.This paper provides a methodology for the vulnerability assessment of ongoing transformations.
In the near future, a major share of balancing energy will be provided by gas-fired power plants.Considering the results of this study, additional adaptation measures to strengthen the resilience of the German power system to extreme weather conditions will become ever more impor-tant.Based on the calculation presented in this study, the losses in revenue due to forced outages at extremely high temperatures amounted to 13-17 million euros p.a. and 46 million euro p.a. in 2018, when high temperatures coincided with low wind generation.Given the trend toward more frequent occurrences of such events, the potential losses will also increase.As a result, decisions about adapting the power system to changing climate conditions will not only become a question of system resilience but also of profitability.Interconnector capacities have proven to be an effective asset for increasing the reliability of supply in power systems.Within the next decade, an increase of interconnectors between the European countries of almost 35% is expected (ENTSO-E 2018).An interconnected multi-regional system can reduce overall capacity investments in individual countries up to 31.8% (Hagspiel, Knaut, and Peter 2018).In this context, it becomes more important to take into consideration regional resilience to natural hazards.
Increased interconnection of the European energy system creates a unique opportunity to improve the efficiency and flexibility of the sector and resilience to localized droughts.However, it also increases the vulnerability of the system to widespread droughts when multi-country regions are affected by an extreme event like a heatwave following by drought.
The planning problem of future electricity market designs must consider multiple dimensions, taking into account the spatial and temporal scale of its exposure to weather extremes.While occurrences of weather extremes must be regarded as uncertainties beyond influence, the resilience of the system can be managed.The use of highly granular data for policy assessment planning models can help policymakers and public utilities, generation utility companies, and transmission and distribution system operators.
Electricity market reforms, regulatory initiatives, and incentives for new generation technologies adopted in the face of a changing climate are sources of uncertainty on par with climate change itself.It is essential to account for both sources of uncertainty to ensure the effectiveness of mitigation (including the phase-out of fossil generation capacities and the deployment of bioenergy and CCS technologies) and adaptation (investment in cooling systems and grid expansion) policy options.The methodology proposed in this paper should help to navigate the transformation of the energy system based on a holistic approach to decarbonization and adaptation.Our analysis shows that an exclusive focus on cost efficient decarbonization of power generation might result in an asset's allocation that is in turn less resilient to the effects of climate change.Hence, designers of future low-carbon power systems must consider the effects of climate change on the system, not only the system's effect on climate change.van Vliet, Michelle T.H., John R. Yearsley, Fulco Ludwig, Stefan Vögele, Dennis P. Lettenmaier, and Pavel Kabat (2012). "

Soil moisture index
The GDM provides detailed data and analysis on Germany's soil moisture and drought status.The core of the GDM is the mesoscale Hydrologic Model (mHM) (Samaniego, Kumar, andAttinger 2010, Kumar, Samaniego, andAttinger 2013).The data are collected daily by 2500 weather stations, which are operated by the DWD and interpolated for a grid of 4km by 4km.The processed data are then used as input for the mHM to provide simulated soil moisture.The SMI is formed by transforming the data obtained.The daily moisture values from the past 30 days of mHM are averaged in order to obtain an average monthly value for each cell.Rather than providing absolute values of soil moisture, the SMI provides a range from 0 to 1, with a lower value expressing a higher moisture deficit.Hence, the simulated soil moisture data are normalized with regard to the corresponding saturated water content (Samaniego, Kumar, and Zink 2013).This fraction of monthly measured moisture content to saturated amount in the considered soil layer may show strongly skewed or multimodal distribution, rather than gaussian distributions, where the specific shape is subject to soil properties and general climatic conditions (Koster et al. 2009).Consequently, Samaniego, Kumar, and Zink (2013) use a an non-parametric kernel-based cumulative estimate function to describe the probability density of the monthly moisture fraction, based on a historic reconstruction of soil moisture from 1954 to 2013 (see also Zink et al. 2016). Zink et al. (2016) define the following intervals for the interpretation of the SMI (see Table B 1).

Entso-E Outages
Entso-E's outage data contains both partial reductions and complete shutdowns of production capacity.Outages are reported for both power plants as a whole, and singular power plant units.In addition to information on the capacity reductions and the affected power plant sites, information used in the analysis has different start dates of reporting, this explains the spit in the trend to the separate trends of the mean and maximum temperatures between all reporting station.While the trend in temperatures observed is clear, the SMI data for maximum minimum precipitation needs assessment.D 1. Droughts can be a trigger for power plant outages, but at the same time the reasons for outages can be manifold.For future work an elaborate econometric model, or a probabilistic simulation model could be built to provide a clearer connection.
Important for our analysis are the outliers that are not fully captured by the regression model.The option analysis that allows to include both the change in the distribution of the sample and likelihood of the extreme events is an appropriate tool.
Figure 1: (a) Histogram and probability density (y-axis) of SMI values (x-axis) for two periods, and-(b) Ratio [%] of average SMI index from 2000-2018 to the long-term average 1951-2000.Outages of the power plants are shown for the summer of 2018.

Figure 2 :
Figure 2: Recent developments in the intensity of droughts and increasing risk of extreme droughts in the future.

Figure 3 :
Figure 3: Temperature and drought time series compared to forced power plant outages.

Figure 5 :
Figure 5: Drought risk and wholesale price volatility.
Figure 6 illustrates the maximum capacity mix unavailable per hour among all hours of the month.The highest values (approximately 8 GW) were observed in summer 2018 and 2016.The hourly plot is given in the Appendix, Figure A 2.

Figure A 1 :
Figure A 1: Frequency and duration of outages.

Figure A 2 :
Figure A 2: Frequency and duration of outages.Hourly quantity of capacity in forced outage between 2015-2019.
Figure C 3: Monitoring fluctuation process for the SMI series (from 1951 to 2018).

Figure C 4 :
Figure C 4: BIC and negative log likelihood criteria for number of breakpoints in series.

Figure D 2 :
Figure D 2: Scatter plot of outages number summed by hour per 5°C temperature step.

Table 4 : Overnight investment costs of cooling systems by power plant technology.
Source: Own compilation based on (EPRI 2004) a , (TetraTech 2002) b , and (DOE/NETL 2008) c

Table A 5: Phase out of nuclear and coal capacities.
According to the phase-out strategy described in the "Entwurf eines Gesetzes zur Reduzierung und zur Beendigung der Kohleverstromung und zur Änderung weiterer Gesetze, 29.01.2020". b