This analysis is part of the Climate Change Impacts and Risk Analysis (CIRA) project (Waldhoff et al. 2014, this issue) which aims to quantify the physical and economic impacts of climate change in the United States. One of the central features of the CIRA project is the use of consistent socio-economic and climatic projections. To that end, this analysis uses temperature pathways developed by the MIT IGSM-CAM model as described in Monier et al. (2014, this issue). The socioeconomic and emissions projections underlying these pathways may be found in Paltsev et al. (2013, this issue).
This study compares the electricity demand and supply results from three electricity models (GCAM, IPM, and ReEDS) across six scenarios that differ by temperature pathway and climate policy (see Online Resource Table 1). The scenarios are summarized below.
Control scenario - No temperature change, no policy. This scenario holds ambient air temperatures constant over time. The Control scenario reflects a typical baseline simulation of each model in which electricity demand is unaffected by temperature change.
REF CS3 - Reference temperature change with climate sensitivity of 3°. This scenario incorporates the effects of rising temperatures under reference (i.e., no policy) emission levels using the same global reference GHG emission pathways at equilibrium climate sensitivities (CS) of 3° Celsius.
REF CS6 - Reference temperature change with climate sensitivity of 6°. The higher climate sensitivity represents a low probability, higher temperature scenario.
POL4.5 CS3 - Emission reduction policy and temperature pathway consistent with a radiative forcing target of 4.5 W/m2. Cumulative power sector emissions from 2015 to 2050 are reduced by 8.9 %.
POL3.7 CS3 - Emission reduction policy and temperature pathway consistent with a radiative forcing target of 3.7 W/m2. Cumulative power sector emissions from 2015 to 2050 are reduced by 21.3 %.
TEMP 3.7 CS3 - Temperature pathway consistent with a radiative forcing target of 3.7 W/m2, but without the emission reduction policy. This scenario isolates the effect of a small temperature change under a low emission scenario from the combined policy and temperature effects in POL3.7 CS3.
The two climate policy scenarios (POL4.5 CS3 and POL3.7 CS3) represent the emissions reductions required in the U.S. electric power sector consistent with global emissions pathways required to achieve equilibrium levels of radiative forcing of 4.5 and 3.7 W/m2 in 2100. The power sector emissions pathways for the two policy scenarios, POL4.5 CS3 and POL3.7 CS3, are based on the change in power sector emissions from the global GCAM simulations of these scenarios (see Calvin et al. 2014, this issue). We apply the percent reduction in cumulative power sector emissions from 2015 to 2050 of the GCAM model to the three models in this analysis. The percentage change in emissions is appropriate to use instead of an absolute value because GCAM-USA, IPM, and ReEDS have different emissions pathways in the control scenario. Cumulative emissions versus the Control scenarios fall by 8.9 and 21.3 % in the POL4.5 CS3 and POL3.7 CS3 scenarios, respectively. To meet this cumulative target, the 2050 annual emissions fall by roughly 23 and 55 % below reference emissions in 2050 for the respective POL4.5 CS3 and the POL3.7 CS3 scenarios.
This section provides a description of the models used in the analysis including an overview of the model, the translation of temperature change into electricity demand, and implementation of the policy scenarios. See Table 1 for a summary of the models’ attributes. To aid in comparing the results across models with different native geographic resolution and make the analysis more tractable, results are aggregated—using population weighting where appropriate (e.g., heating and cooling degree days)—to six national regions as shown in Table 2.
The GCAM-USA model is a detailed, service-based building energy model for the 50 U.S. states (Kyle et al. 2010; Zhou et al. 2014). Nested within the global GCAM model (Kim et al. 2006), it allows for greater spatial representation of U.S. buildings sector while maintaining the full interaction with other U.S. sectors and other global regions. GCAM is a recursive dynamic model that projects greenhouse gas emissions and energy trends to the end of the century, and it includes partial equilibrium economic models of the global energy system and global land use.
The heating and cooling demands come from the buildings sector in each state, which is based on two representative building types: residential and commercial. Each building type demands six service categories: heating, cooling, lighting, hot water, appliances (residential) or office equipment (commercial), and others. These services are provided by end use technologies, the number of which depends on the service. These technologies use four types of fuels including electricity, natural gas, fuel oil and biomass. Other electricity demands (e.g., industrial and transportation) are modeled at the state level.
The electricity demand from buildings is a function of floor space, building shell efficiencies, end use technologies, state economic and population growth, population-weighted heating and cooling degree-days (HDD/CDDs), and other technical and calibration parameters (see Zhou et al. 2014 and the Online Resource). The model is calibrated to a base year of 2005. The change in energy demand for the non-controls is based on the changes in HDD/CDD over time from the CIRA scenarios’ temperature data.
The Regional Energy Deployment System model (ReEDS) is a deterministic, myopic, optimization model of the deployment of electric power generation technologies and transmission infrastructure for the contiguous United States. It is designed to analyze critical energy issues in the electric sector, especially power sector emissions constraints and clean energy standards. ReEDS provides a detailed treatment of electricity-generating and electrical storage technologies and specifically addresses a variety of issues related to renewable energy technologies, including accessibility and cost of transmission, regional quality of renewable resources, seasonal and diurnal generation profiles, variability of wind and solar power, and the influence of variability on the reliability of the electrical grid. ReEDS addresses these issues through a highly discretized regional structure (e.g., 134 balancing areas), explicit statistical treatment of the variability in wind and solar output over time, and consideration of ancillary service requirements and costs.
To translate temperature change to change in power demand, a temperature-sensitive econometric demand model was developed for each of ReEDS’ regions. The model estimates the change in electricity demand load as a function of HDD/CDD over a reference level of demand for each of the power control areas. The model parameters are based on detailed empirical utility load data for over 300 transmission zones over a 2 year period (see Sullivan et al. 2015 and Denholm et al. 2012). Parameter estimates are obtained for four seasons, which captures both heating and cooling seasons, and four daily time slices. Unlike the structural equations used in GCAM and IPM that specify residential and commercial heating and cooling, the ReEDS demand model aggregates all temperature-sensitive demand changes including industrial. The model assumes a fixed ratio of temperature-sensitive demand to total demand. A limitation of this approach is that the model does not capture changes in consumer preferences, shifts in population, and technological change (e.g., end-use efficiency improvements).
ReEDS, an electricity-only model, requires additional information relating a global, economy-wide GHG reduction pathway to U.S. electricity-sector CO2 emissions limits. As an estimate for the electricity sector’s share, ReEDS uses as input the electric-sector CO2 emissions from the relevant GCAM scenarios, rescaled to match ReEDS’ 2010 emissions levels. In contrast to GCAM and IPM, ReEDS assumes that carbon credits are given away to emitting sources rather than auctioned off. This policy assumption has the effect of reducing the cost of a GHG policy to utilities, thereby damping the price change seen by consumers and the demand response.
The Integrated Planning Model (IPM®) is a well-established electric sector dispatch and capacity planning model used by both the public and private sectors to inform business and regulatory policy decisions. The implementation of IPM used for this study (EPA Base Case v4.10) represents the power system of the contiguous United States and Canada in 32 model regions (see Online Resource Figure 1). The model is a fully forward-looking linear programming model that determines the least-cost method of meeting energy and peak demand requirements over the period of 2012 to 2050.Footnote 2 It provides integration of wholesale power, system reliability, environmental constraints, fuel choice, transmission, capacity expansion, and all key operational elements of generators on the power grid.
For the present work, IPM treats electricity demand as an input from a separate electricity demand model. The demand model uses structural equations that represent electricity demand as a function of activity (e.g., population or employment), structure (e.g., square footage, climate), and intensity (i.e., electricity used per unit of activity) as described in Jaglom et al. (2014). The population, employment, and square footage factors change over time, driven by assumptions about regional population growth contained in the Annual Energy Outlook (U.S. DOE 2010). Intensity can be thought of as a semi-empirical measure of the technology, efficiency, and consumer preference that drive demand in a particular region. The intensity factors are estimated based on the base-year and projected electricity consumption from the Annual Energy Outlook and the observed HDD and CDD from 2000 to 2009. The intensity factors are constant across the scenario, but change over time to capture shifts in exogenous variables such as consumer preferences and end-use efficiencies.
For the non-control scenarios in which temperature changes over time, the demand model uses HDD/CDD data to calculate the changes in temperature sensitive demand (i.e., residential and commercial heating and cooling). A 30-year centered average of temperatures is used in the HDD/CDD calculations. These changes in demand are applied to the exogenous control scenario demand that feeds into IPM. The single policy scenario run by IPM, POL3.7 CS3, is implemented as a cap-and-trade system in which banking is allowed; borrowing is not.
Comparison of heating and cooling degree-day estimates
All of the models used the same method for calculating HDD/CDD from the temperature data using a base temperature set-point of 65 °F, a common convention (see Isaac and van Vuuren (2009) and equations in the Online Resource). This convention was chosen because of its widespread use in the literature and simplicity. In doing so, we acknowledge the work of Hekkenberg et al. (2009) that suggests this method may lead to a conservative energy demand estimate.
The national, population-weighted HDD/CDD values for each model are shown in Online Resource Figure 2. The models exhibit closely aligned trends of falling HDD and rising CDD over time and the values fall within a range of 1.6 to 6 % of the median. From 2005 to 2050, HDD declines by 780 to 990 degree-days (−19 to −24 % from 2005) while CDD rises by 540 to 670 degree-days (32 to 43 % from 2005).
Snapshots of HDD and CDD broken-out by region, model, and scenario for 2005 and 2050 may be found in Online Resource Figures 3 and 4. Across the regions, the absolute decrease in HDD is greatest in the northern regions (Northeast, Midwest, and Northwest). As expected the decline is more pronounced with the higher temperature scenarios (REF CS6,3) than the policy scenario (POL3.7). The absolute increase in CDD over time is greatest in the southern regions (Southeast, South Central, and Southwest). The increase in CDD is greatly mitigated in the policy scenario.
Though the absolute differences in CDD are largest in the southern regions, the percentage change over time is greatest in the northern regions as shown in Fig. 1. The average increase in Northeast and Northwest is 70 % (averaged across the models for REF CS3) followed by 35 % in the Midwest and Southwest. The percent changes within a region are consistent across most regions with the exception of the Northeast and Northwest in which IPM is higher than the other models by over 20 percentage points. The variation in the changes between the models is attributed to differences in population assumptions, spatial resolution, and smoothing method applied to noisy temperature or HDD/CDD time series data. The percent change in heating degree day shows more uniformity across regions and models at around −20 % (REF CS3) with the Southwest slightly greater at −29 %. Under the POL3.7 scenario, the average change in CDD drops to 19 % and the change in HDD moves to −15 %.