High ambient temperatures intensify photochemical production of tropospheric ozone, leading to concerns that global warming may exacerbate smog episodes. This widely observed phenomenon has been termed the climate penalty factor (CPF). A variety of meteorological and photochemical processes have been suggested to explain why surface ozone increases on hot days. Here, we quantify an anthropogenic factor previously overlooked: the rise of ozone precursor emissions on hot summer days due to high electricity demand. Between 1997 and 2011, power plant emissions of NOx in the eastern U.S. increased by ~2.5–4.0%/°C, raising surface NOx concentrations by 0.10–0.25 ppb/°C. Given an ozone production efficiency (OPE) of ~8 mol/mol based on the 2011 NASA DISCOVER-AQ campaign, at least one third of the CPF observed in the eastern U.S. can be attributed to the temperature dependence of NOx emissions. This finding suggests that controlling emissions associated with electricity generation on hot summer days can mitigate the CPF.
 Global mean temperature has increased about 0.8°C since the late 19th century, and is projected to increase another 0.4°C over the next two decades [IPCC, 2007]. Temperature is one of the most important factors influencing the abundance of ozone (O3) near the surface [EPA, 2006]. Greater ozone pollution associated with more prevalent heat waves is a clear adverse effect of climate change [EPA, 2006]. Previous studies [e.g., Jacob and Winner2009; EPA, 2009) project an increase of 1 to 10 parts per billion (ppb) in ground level ozone due to climate change over the next several decades.
 The climate penalty factor (CPF), defined as, has been used to quantify the effect of climate change on ozone pollution. Bloomer et al.  quantified CPF in the eastern U.S. to be ~3.2 ppb/°C prior to 2002 and ~2.2 ppb/°C after 2002, based on analysis of ground level ozone and temperature measurements at rural sites. Surface ozone levels in the eastern U.S. have steadily declined during the past decade due to reductions in the emission of NOx from power plants following the EPA NOx State Implementation Plan (SIP) as well as reduced vehicular emissions of NOx [e.g., Bloomer et al., 2009; Cooper et al., 2012]. The photochemical production of tropospheric ozone in the eastern U.S. is controlled by NOx [e.g., Sillman, 1999]. Bloomer et al.  related the decline in the CPF after 2002 to reductions in emissions of NOx from power plants.
 Several possible explanations for the positive value of the CPF (i.e., the rise in surface ozone with increasing temperature) have been suggested. These include enhanced thermal decomposition of peroxyacyl nitrate (PAN) and increased anthropogenic and biogenic emissions of volatile organic compounds (VOCs) as the lower atmosphere warms, as well as the correlation of subsidence, wind stagnation, and UV radiation (increasing photolysis rates) with the meteorological conditions associated with high temperature [Mickley et al., 2004; EPA, 2006]. Anthropogenic emissions of NOx can also vary with temperature. Hot summer days are often associated with increased emissions of NOx from power plants to meet greater electricity demand for air conditioning. Singh and Sloan  reported that vehicular emissions of NOx increase slightly, about 1%/°C, between 20 and 35°C. Auxiliary electricity generators such as backup diesel generators at hospitals running on hot days when energy demand peaks may also add to the burden (private communication, J. McDill, MARAMA and A. Mirzakhalili, DE DNREC). Intensified emissions of NOx will boost the photochemical production of ozone, contributing to the positive value of the CPF.
 We use data from EPA's Continuous Emission Monitoring System (CEMS) to quantify the temperature sensitivity of power plant emissions of NOx; the CEMS monitors emissions from a variety of sources, but the biggest contributors are power generating units. Hereafter we will use the term “power plant emissions” to refer to all point sources in the CEMS inventory. The temperature dependence of emissions in the eastern U.S. during summer () and the response of ground level NOx to temperature () are derived. We then use measurements from the 2011 National Aeronautics and Space Administration (NASA) Deriving Information on Surface Conditions from COlumn and VERtically Resolved Observations Relevant to Air Quality (DISCOVER-AQ) aircraft campaign in the Baltimore/Washington area to calculate the ozone production efficiency (OPE). Finally, these various factors are used to assess the contribution of the temperature sensitivity of power plant emissions of NOx to observed CPF in the eastern U.S.
2 Data and Methods
 Daily CEMS data from power plants in five states of the eastern U.S., Maryland (MD), Ohio (OH), Pennsylvania (PA), Virginia (VA), and West Virginia (WV) were downloaded from the EPA Clean Air Market database (data available at http://ampd.epa.gov/ampd) [He et al., 2013]. We also acquired hourly NOx observations in the Baltimore/Washington area from the EPA Air Quality System (AQS) website (http://www.epa.gov/ttn/airs/airsaqs/detaildata/downloadaqsdata.htm, locations of the selected NOx monitoring sites are shown in Figure S1 of the supporting information). Daily emissions of NOx from power plants and surface concentrations of NOx in summer (June, July, and August) were calculated (see details in [He et al., 2013]).
 Hourly surface temperature data were obtained from the National Oceanic and Atmospheric Administration (NOAA) National Climatic Data Center (http://gis.ncdc.noaa.gov). Observations from five monitoring sites, the Baltimore/Washington International Airport (BWI), the Cleveland Hopkins International Airport (CLE), the Pittsburg International Airport (PIT), the Dulles International Airport (IAD), and the Yeager Airport (CRW), were selected to represent the local ambient temperature in MD, OH, PA, VA, and WV, respectively. Daily mean temperatures were computed for each site.
 The proper quantification of the impact of on ambient ozone requires accurate knowledge of the OPE, defined as the net number of ozone molecules produced per molecule of emitted NOx [e.g., Liu et al., 1987; Kleinman et al., 2002]. We have estimated OPE from the slope of Ox (O3 + NO2) versus NOz (NOy − NOx) [Kleinman et al., 2002] from simultaneous observations of these species. All measurements were obtained by the National Center for Atmospheric Research (NCAR) four-channel chemiluminescence instrument [Walega et al., 1991] during the DISCOVER-AQ campaign over the eastern U.S. in July 2011 (data archive at http://www-air.larc.nasa.gov/cgi-bin/ArcView/discover-aq). Our estimate of OPE is restricted to measurements made in the planetary boundary layer (PBL, pressure > 890 hPa) that show a strong linear correlation (r2 > 0.8) between Ox and NOz for individual vertical spirals.
 For each summer between 1997 and 2011, we have tabulated the total number of days with an ozone exceedance in the Baltimore/Washington region, the summertime mean surface ozone abundance in the region, and number of days when temperature exceeded 32.3°C (90°F) (Figure S2 in the supporting information). The analysis confirms that ozone pollution during the past 15 years is closely related to ambient temperature. Figure 1 shows daily power plant emissions of NOx versus surface temperature in MD, with data grouped into four time periods. For all time periods, power plant emission of NOx rises with increasing temperature. Power plant emissions decreased significantly after 2002 due to the EPA NOx SIP call, with widespread implementation of selective catalytic reduction (SCR) removal of NOx [He et al., 2013]. Plots for other states are similar (Figure S3 in the supporting information).
 We next calculated using linear regression analysis of daily CEMS NOx emissions versus daily mean temperature. Figure 2 presents the temperature response of NOx emissions from power plants (slope) and the correlation coefficient (r). High correlation between NOx emissions and temperature is observed; except for 2003 in WV, which could be caused by partial implementation of NOx controls associated with the NOx SIP call. The value of averaged for the five states decreased from ~15 tons/°C prior to 2002 to ~8 tons/°C after 2002. The summertime power plant emissions of NOx in the five states were ~3000 tons/day in the late 1990s and ~1000 tons/day in the early 2010s [He et al., 2013]. Thus, a 1°C temperature increase could have raised emissions of NOx in the five states considered by ~75 tons prior to 2002 and ~40 tons after 2002, corresponding to increases of ~2.5% and ~4.0%, respectively, with respect to overall power plant emissions. He et al.  reported that ground level NOx is strongly correlated with power plant emissions of NOx in the Baltimore/Washington area; the value of is 4.2 ×10−3 ppb/tons (Figure S4 in the supporting information). Using the slope of the relation between these two quantities, we estimate to be ~0.25 ppb/°C prior 2002 and ~0.10 ppb/°C after 2002.
 Figure 3 shows the relationship between OPE inferred from DISCOVER-AQ measurements and maximum NOx observed during each spiral. The average OPE value was ~8 mol/mol in the Baltimore/Washington area in 2011. Using the value of given above and assuming that the OPE observed in 2011 is applicable for all years, and formulating the temperature sensitivity of surface ozone to power plant emissions as , we estimate to be ~2.0 ± 1.0 ppb/°C prior to 2002 and ~0.8 ± 0.4 ppb/°C after 2002 (uncertainty analysis is described in the supporting information). The decline is due to widespread implementation of SCR technology in power plants after 2002. Bloomer et al.  reported empirical values for the CPF of 3.2 ppb/°C (prior to 2002) and 2.2 ppb/°C (after 2002). Hence, the sensitivity of power plant emissions of NOx to ambient temperature may have contributed approximately two thirds of the observed CPF prior to 2002 and approximately one third of CPF after 2002.
 We have assumed that the OPE inferred from DISCOVER-AQ data during summer 2011 applies for all of the time periods under consideration. Tropospheric ozone production is a nonlinear function of the abundance of NOx and VOCs [e.g., Lin et al., 1988; EPA, 2006]. Observed values of OPE range from 5 mol/mol in urban areas [Kleinman, 2000; Nunnermacker et al., 2000] to ~50 mol/mol in the clean marine atmosphere [Wang et al., 1996]. With decreasing trends of ambient NOx in the eastern U.S. [He et al., 2013], the OPE of ~8 mol/mol may be an overestimate for the late 1990s and early 2000s, because generally OPE falls as NOx rises [Kleinman et al., 2002]. Our estimate that approximately two thirds of CPF can be explained by the temperature dependence of power plant emissions prior to 2002 should therefore be treated as an upper limit, since the OPE in the early 2000s was likely lower than that determined from measurements in 2011. Finally, our analysis only considers the temperature dependence of NOx emissions from power plants equipped with CEMS instrumentation; we do not consider small additional contributions from vehicles and small auxiliary electricity generators situated at locations other than power plants.
4 Concluding Remarks
 Emissions of NOx from vehicles and power plants have declined significantly in the eastern U.S. during the past 15 years. Nonetheless, daily emissions of NOx are tied to electricity demand, which rises on hot days due to increased use of air conditioners. Much of this rise can be attributed to the use of less regulated peaking units in the power plants to meet the excessive demand. It has long been known that surface ozone tends to be higher on hot summer days. We have shown that for the eastern U.S., a considerable portion (approximately one third) of the observed relationship between surface ozone and temperature, termed the CPF, is caused by the increase in power plant emissions of NOx (an important ozone precursor) on hot summer days. This analysis suggests a substantial part of one danger of climate change, deteriorating air quality, can be mitigated by regional or national regulatory actions focused on limiting power plant emissions of NOx on hot summer days.
 We thank Maryland Department of the Environment (MDE) for supporting this study. The 2011 DISCOVER-AQ campaign was supported by NASA. We thank Dr. Andrew J. Weinheimer of NCAR for providing the airborne measurements from which OPE was calculated. Thanks to J. McDill (MARAMA) and A. Mirzakhalili (DE DNREC) for information on auxiliary generators. This research was also supported by NASA grants NNX11AK34G, NNX12AI18G, and NNX12AB10G. The authors thank two anonymous reviewers for helpful suggestions.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.