The climatic effects of greenhouse gas emissions are primarily increases in temperature and changes in precipitation. The impact on wind is more uncertain: The main lessons from the literature are that future wind conditions in Western Europe are uncertain, and that there are no consensus estimates for change in wind conditions that are significantly different from the no change scenario (Pryor et al. 2005; Haugen and Iversen 2008 and Kjellström et al. 2011). Therefore, inclusion of a change in wind conditions would only add uncertainty to our model, and we decided to disregard changes in wind conditions. Moreover, the uncertainty about cost estimates for wind power is so high that we cannot use LIBEMOD’s output with respect to wind power investments. Hence, in the present study wind power investments are exogenous.
The climatic effects we attempt to model below are:
-
a.
changes in demand for electricity due to changes in the need for heating and cooling,
-
b.
changes in supply of hydropower due to changes in precipitation and temperature, and
-
c.
changes in thermal power supply due to warmer cooling water and therefore lower plant efficiency.
Emissions of greenhouse gases have major damaging effects in the long run, while the short-run effects are minor. This suggests focusing attention on a year several decades ahead, for example, 2100. On the other hand, not much is known about the energy markets in 2100. There is no obvious solution to this trade-off. In order to illustrate the impact of climate change on electricity markets, we use a pedagogical tool, namely to postulate that the average climate in a future time period (2070–2099) materializes in a much earlier year (2030). The year 2030 is far enough into the future to enable optimal investments to change production and transport capacities, but short enough that the economic and political structure can reasonably be expected to continue on the historical trends presently observed. The resulting scenarios and simulations must therefore be carefully interpreted; they are not predictions of actual behavior, but are comparative static simulations of the effects of a climate change on a power system that has had time to adapt.
Below, climate change effects for 2070–2099, for simplicity referred to as 2085, build on the IPCC scenario A1b, see IPCC (2007). This is the most referred emission scenario from IPCC (2007), with a projected global warming of 2.8°C until the end of this century. While IPCC (2007) reports results from global climate model simulations, these must be disaggregated to find climate effects for each Western European country which are needed in our analysis. Downscaling of temperature was performed using an empirical-statistical method based on climate model results, ERA40 re-analysis data (Uppala et al. 2005), and weather station observations, see Benestad (2005, 2008a). The downscaling was based on 20 global climate models described in the IPCC fourth assessment report (Meehl et al. 2007). The estimated multi-model mean temperatures for the period 2071–2100 were used. The complete list of the global circulation models and runs included in this analysis can be found in Table 5 in Benestad (2008b).
Demand for electricity
The demand effect of a warmer climate operates primarily through the need for increased cooling during the summer, and less heating during the winter. These effects are picked up by the annual number of Cooling-Degree-Days (CDD) and Heating-Degree-Days (HDD): Let T
d
be the average daily temperature (°C) on day d. Then T
d
–22 (if positive) is the number of degrees that the average temperature exceeds 22°C on day d. When this (positive) number is summed over all days in a year for which T
d
exceeds 22°C, one obtains CDD; \( \sum\nolimits_{{d = 1}}^{{365}} {{\rm Max} \left( {0,{T_d} - 22} \right)} \). HDD is the corresponding sum of temperatures lower than 18°C; \( \sum\nolimits_{{d = 1}}^{{365}} {{\rm Max} \left( {0,18 - {T_d}} \right)} \).
According to the Mideksa and Kallbekken (2010) review, the sign of the heating and cooling effects on electricity demand seem to be consistent across studies, but with a wide variation on the magnitude of the estimates, see, for example, Baxter and Calandri (1992), Aroonruengsawat and Auffhammer (2009) and Isaac and van Vuuren (2009) and De Cian et al. (2007). Here we build on Eskeland and Mideksa (2009), which used an instrumental variables approach to account for the endogeneity of prices. This is an econometric study of residential demand for electricity using a panel of Western European countries. Using a fixed-effects regression model, this study allows for country differences in the response to temperature changes. As is common in the literature, they include the annual number of CDD and HDD among the independent variables, and find small but significant estimates of the effect of CDD and HDD on per capita electricity consumption.
The Eskeland and Mideksa study has also calculated CDD and HDD separately for each Western European country based on city-specific data in Benestad (2008a). They found that the climate change from 2000 to 2085 increases CDD by 121 days (88%) and decreases HDD by 712 days (28%) for Western Europe as a whole, see Table 1 in Online Resource 1.
Combining the estimated coefficients of CDD and HDD from Eskeland and Mideksa (2009) with the calculated changes in CDD and HDD because of climate change from 2000 to 2085, we find that, cet. par., demand for electricity in Western Europe increases by 3.6% from 2000 to 2085 due to increased cooling needs, but decreases due to lower heating needs by 7.3%. The net direct effect, before the feedback from the model equilibrium changes of supply and prices, is a decrease in annual demand in Western Europe by 3.7%.
Figure 1 shows electricity demand in the base year 2000 and calculated direct changes in demand with 2085 climate because of more cooling and less heating. Demand changes are not uniform across countries. As expected, the Northern European countries decrease their heating demand, but there is almost no increase in their cooling demand since even with a warmer climate there are few days with an average temperature above 22°C. This is mostly reversed in Southern European countries where the increase in cooling demand clearly dominates over the decrease in heating demand. Because heating is needed mainly in winter and cooling mainly in summer, these demand changes are imposed on LIBEMOD in the corresponding season only. The effect of climate change on demand in the model is a shift from northern to southern countries, and a shift from winter to summer.
Inflow of water
Projected changes in runoff have been estimated by the VIC hydrology model (Liang et al. 1994). The model was run at 0.5° spatial resolution for the baseline period (1961–1990), using CRU (Climate Research Unit of the University of East Anglia) meteorological input data (Mitchell and Jones 2005). For the projection period (2070–2099), air temperature and precipitation data from two global circulation models (GCM) run under the A1b emission scenario were used, the Hadley Center Coupled Model (HadCM3, Gordon et al. 2000) and the Max Planck Institute model ECHAM5 (Roeckner et al. 2006). To create the meteorological input data, the ‘delta change approach’ was used: the monthly changes in precipitation and temperature between baseline and projection were calculated, and these changes were imposed on the CRU data. The meteorological data for the baseline and the projection periods were created similarly to the method used by Adam et al. (2009). The results for each run are given as a percentage change in runoff between baseline and future periods for each country and for each season; summer (April 1 – September 30) and winter (October 1 – March 30).
In hydropower production, runoff is only useable to the extent that it reaches run-of-river power plants or the reservoirs; in the latter case the energy content depends on the altitude difference between reservoir and power station. For Norway, a detailed model (“samkjøringsmodellen”) has been used to calculate seasonal inflow measured in energy content of usable water (TWh) that would result from the 2,085 runoffs while keeping the 2000 power system infrastructure. For other countries, we have assumed that changes in inflow are proportional to country-average changes in runoff in each season.
Figure 2 shows inflow by country and season, where the 2,085 numbers reflect direct effects of climate change had the hydropower capacities been kept unchanged from 2000. Annual inflow in Sweden, Finland and Norway increases relative to 2000, whereas there is a decline in Southern European countries. The net effect is a 15% decrease in inflow (and thereby production) for the group of model countries.
In the Nordic and Alpine countries, most of the inflow is received in the summer because winter precipitation falls as snow and is only usable for electricity production when it melts. These countries all expect an increase in winter inflow in 2,085, mostly due to higher temperatures. In Southern Europe, summer inflow decreases as a result of both less precipitation and higher temperatures. The main pattern is a shift in inflow from Southern to Northern Europe, and from summer to winter, exactly mirroring the demand changes.
The Carnot effect
The efficiency of thermal power plants depends on the temperature of the water used for cooling; the hotter the water, the lower is efficiency. Atmospheric warming and changes in river flow are expected to affect river temperatures; see, for example, Mohseni et al. (1999), Mantua et al. (2010) and van Vliet et al. (2011).
Thermal efficiency is in theory linearly dependent on the temperature of the cooling water or cooling air, and thus the average efficiency of thermal plants will be reduced because of climate change – usually referred to as the Carnot effect. Under intense heat waves, nuclear plants may even need to close for safety reasons, this happened in France, Germany and Spain in 2006, and later in Sweden. The Carnot effect means that for the same amount of inputs, output is reduced. However, the Carnot effect will in general lead to a change in the amount of inputs through the market mechanism.
The first step in identifying the Carnot effect is to estimate the temperature changes. Benestad (2009) has estimates for 2,085 quarterly temperatures for the same cities as in Benestad (2008a). We have aggregated these to the seasons and countries in LIBEMOD; we find that the average 2,085 temperature (relative to 2000) increases by 3.0°C in summer and 2.8°C in winter.
The Mideksa and Kalbekken literature review found few studies that quantified the effect of temperature increase on thermal power efficiency. Yet, some studies are available; building on Durmayaz and Sogut (2006), Linnerud et al. (2009, 2011), and taking into account that the effect is not linear in temperature change, we obtain that thermal efficiency in 2085 is reduced by 1.8 (1.7) percent in summer (winter) in fossil plants, and reduced by 2.4 (2.3) percent in nuclear plants. Because plant efficiency differs in our numerical model LIBEMOD, the Carnot effect will be (slightly) different across technologies and countries. Note that the Carnot estimates are uncertain. They have been estimated on small samples (this is in particular the case for fossil fuel plants), and adaptation strategies of power producers, for example, installing water pumps or relocate, have not been taken into account. Still, we believe the Carnot effect should be included in the analysis.
What is the direct impact of the Carnot effect of the 2085 climate, keeping the fuel use constant at year 2000 level? In the winter, the temperature increases are greatest in Northern Europe, for example, 4.9°C in Finland versus 2.4°C in Italy, see Fig. 3, suggesting that the supply reduction is greatest in Northern Europe. Summer temperature differences are smaller, ranging from 2.4°C in Great Britain to 3.7°C in Spain, and hence the shifts in summer supply do not differ that much between countries. Because fossil plants are less common in Northern Europe, there is a small tendency that, cet. par., the reduction as a share of total electricity supply is smallest in the north. Overall, the unambiguous effect is to reduce supply of electricity, while geographical and seasonal patterns are weak.