Summertime weekly cycles of observed and modeled NO x and O 3 concentrations as a function of satellite-derived ozone production sensitivity and land use types over the Continental United States

To show how remote-sensing products can be used to classify the entire CONUS domain into “geographical regions” and “chemical regimes”, we analyzed the results of simulation from the Community Multiscale Air Quality (CMAQ) model version 4.7.1 over the Conterminous United States (CONUS) for August 2009. In addition, we observe how these classifications capture the weekly cycles of ground-level nitrogen oxide (NO x) and ozone (O3) at US EPA Air Quality System (AQS) sites. We use the Advanced Very High Resolution Radiometer (AVHRR) land use dominant categories and the Global Ozone Monitoring Experiment-2 (GOME-2) HCHO/NO 2 column density ratios to allocate geographical regions (i.e., “urban”, “forest”, and “other” regions) and chemical regimes (i.e., “NOx-saturated”, “NOx-sensitive”, and “mixed” regimes). We also show that CMAQ simulations using GOME-2 satellite-adjusted NO x emissions mitigate the discrepancy between the weekly cycles of NO x from AQS observations and that from CMAQ simulation results. We found geographical regions and chemical regimes do not show a one-to-one correspondence: the averaged HCHO / NO 2 ratios for AVHRR “urban” and “forest” regions are 2.1 and 4.0, which correspond to GOME-2 “mixed” and “NO xsensitive” regimes, respectively. Both AQS-observed and CMAQ-simulated weekly cycles of NO x show high concentrations on weekdays and low concentrations on weekends, but with oneor two-day shifts of weekly high peaks in the simulated results, which eventually introduces the shifts in simulated weekly-low O 3 concentration. In addition, whereas the high weekend O 3 anomaly is clearly observable at sites over the GOME-2 NO x-saturated regime in both AQS and CMAQ, the weekend effect is not captured at sites over the AVHRR urban region because of the chemical characteristics of the urban sites ( ≈ GOME-2 mixed regime). In addition, the weekend effect from AQS is more clearly discernible at sites above the GOME-2 NO x-saturated regime than at other sites above the CMAQ NO x-saturated regime, suggesting that the GOME-2-based chemical regime classification is more accurate than CMAQ-based chemical classification. Furthermore, the CMAQ simulations using the GOME-2-derived NOx emissions adjustment (decreasing from 462 Gg N to 426 Gg N over the US for August 2009) show large reductions of simulated NO x concentrations (particularly over the urban, or NO x-saturated, regime), and mitigates the large discrepancies between the absolute amount and the weekly pattern of NO x concentrations of the EPA AQS and those of the baseline CMAQ.


Introduction
Photochemical ozone (O 3 ) production near the earth's surface depends on the chemical environment, which is heavily influenced by the ratio of volatile organic compounds (VOCs) to nitrogen oxide (NO x = NO + NO 2 ) emissions.While biogenic sources (vegetation) are primarily responsible for VOC emissions in many parts of the country, manmade sources contribute the majority of NO x emissions in the United States (US).According to the US Environmental Protection Agency (US EPA), anthropogenic NO x emissions in Published by Copernicus Publications on behalf of the European Geosciences Union.
the US are estimated to be 21.2Tg per year (as in 2002): 38 % from on-road vehicles, 22 % from electric generation power plants, 22 % from off-road equipment, 11 % from commercial fuel combustion, and the rest from industrial processes and miscellaneous sources.Because of the dominance of mobile sources, NO x emissions demonstrate a clear daily and weekly pattern related to temporal variations in the human use of fossil fuels.Therefore, an investigation of the daily or weekly variations of surface O 3 concentrations could illuminate the controlling effects of the key precursors of O 3 concentrations.Previous studies have focused on daily or weekly O 3 variations over several metropolitan areas, including New Jersey, Southern California, Los Angeles, Atlanta, Chicago, Denver, New York City, Dallas, Houston, Phoenix, Washington, DC, Baltimore, and their neighboring regions (e.g., Lebron, 1967;Cleveland et al., 1974;Elkus and Wilson, 1977;Vukovich, 2000;Marr and Harley, 2002;Fujita et al., 2003;Qin et al., 2004;Blanchard and Tanenbaum, 2006;Shutters and Balling Jr., 2006;Blanchard et al., 2008;Yarwood et al., 2008).These studies have highlighted the weekend effect on urban regions, where higher ground-level O 3 concentrations occur during the weekends rather than weekdays.However, the peaks of their precursors show an opposite trend: Higher urban region O 3 concentrations during the weekends are attributed to lower surface NO x emissions in the NO xsaturated environment.
The relationship between the weekly cycles of NO x concentrations and emissions has been investigated utilizing remote sensing NO 2 column density products.For example, Beirle et al. (2003) and Kaynak et al. (2009) examined the weekly cycle of the NO 2 column density using retrieval products from the Global Ozone Monitoring Experiment (GOME) and the SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY (SCIAMACHY), respectively, and found that temporal variations in the NO x column density are proportional to emissions.In particular, Beirle et al. (2003) revealed a lower GOME NO 2 column density on weekends and a higher column density on weekdays over the US, European countries, Japan, and South Korea.Similarly, Kaynak et al. (2009) found relatively large SCIAMACHY NO 2 columns on weekdays and a significantly lower column density on weekends over polluted regions.However, no such large reduction has been found in rural areas.A strong correlation was also observed between the Ozone Monitoring Instrument (OMI)-derived surface NO 2 measurements and ground-based NO 2 measurements at the US Air Quality System (AQS) and Environment Canada's National Air Pollution Surveillance (NAPS) networks in polluted areas (Lamsal et al., 2008).
In addition to characterizing the weekly cycles of NO x emissions, estimating a photochemical indicator, which is the ratio of VOCs to NO x emissions, is crucial to a more thorough understanding of photochemical O 3 production because the photochemical environment strongly influences production.Sillman et al. (1990) and Sillman (1999) intro-duced a photochemical indicator that uses the ratios of certain chemical species to represent the O 3 -NO x -VOC sensitivity of a particular geographical area.Recently, Martin et al. (2004) and Duncan et al. (2010) utilized the ratio of satellite HCHO to the NO 2 column density from GOME (spatial resolution, 40 km × 320 km) and OMI (spatial resolution, 13 km × 24 km) as a photochemical indicator consistent with the ratio of VOCs to NO x emissions over the surface.Martin et al. (2004) used the GOME-derived indicator to characterize geographic regions (e.g., North America, Europe, and East Asia) as chemical regimes (NO x -saturated and NO xsensitive regimes).Using increased OMI-derived indicator, Duncan et al. (2010) found that most US cities had become more NO x -sensitive regimes from 2005 to 2007.
Many previous studies (e.g., Lebron, 1967;Cleveland et al., 1974;Elkus and Wilson, 1977;Vukovich, 2000;Marr and Harley, 2002;Fujita et al., 2003;Qin et al., 2004;Blanchard and Tanenbaum, 2006;Shutters and Balling Jr., 2006;Blanchard et al., 2008;Yarwood et al., 2008) showed the weekly pattern of surface NO x and/or O 3 over the AVHRRderived US Geological Survey Land Use Land Cover (USGS LULC) regions (referred to as "AVHRR regions", which consist of urban, forest, and other regions) (Loveland et al., 2000) or GOME-2-derived chemical regimes (referred to as "GOME-2 chemical regimes", which are NO x -saturated, NO x -sensitive, and mixed regimes).Several other studies (e.g., Martin et al., 2004;Duncan et al., 2010) showed how remote sensing products can be used to determine the chemical environment.In this study, despite the uncertainty of remote sensing, we show that remote-sensing-derived chemical environments can be used for determining chemical regimes by directly showing the weekly cycles of surface O 3 from in-situ surface measurements over remote sensing-derived chemical regimes.No previous studies have addressed how the AVHRR-derived geographical regions can be chemically classified.This study investigates the chemical characteristics of AVHRR geographical regions by estimating the column ratios of GOME-2 formaldehyde (HCHO) to nitrogen dioxide (NO 2 ).In addition, two other previous studies (Kim et al., 2009;Russell et al., 2010) presented evidence of a large reduction in mobile source emissions over the western US or California since 1999 and since 2005.The reduction in point source emissions (e.g., from power plants) was accounted for during the preparation of the standard EPA National Emission Inventory 2005 (NEI 2005), but changes in mobile and other source emissions were not because of the lack of explicit data pertaining to changes over the entire CONUS domain.In this study, we perform two different CMAQ simulations -CMAQ simulation with NEI 2005 and CMAQ with GOME-2-derived NO x emissions -to examine the impact of emissions changes on the weekly cycles of surface NO x concentration.
To investigate the weekly cycles of surface O 3 concentrations at EPA AQS stations over different geographical regions and chemical regimes, we divide the 12 km CMAQ model grids covering the CONUS into two different types of satellite-derived classifications: AVHRR-derived geographical regions and GOME-2-derived chemical regimes.We will also introduce CMAQ-derived chemical regimes (referred to as "CMAQ chemical regimes") to investigate the differences between the weekly cycles of surface O 3 concentrations at AQS stations of the GOME-2 and those of the CMAQ chemical regimes.Section 2 describes which measurement data are used for this study and then provides a description of the CMAQ 4.7.1 model.Section 3 briefly describes the methods used to define geographical regions and the GOME-2-and CMAQ-derived chemical regimes, and investigates variations in both the AQS-observed and CMAQsimulated weekly cycles of daytime (01:00-05:00 p.m., local time) surface NO x and O 3 concentrations at corresponding sites over AVHRR geographical regions and GOME-2 chemical regimes for August 2009 and the weekly cycles of NO x and O 3 over GOME-2-and CMAQ-derived chemical regimes.In addition, Sect. 3 shows how CMAQ with GOME-2-derived emissions mitigate the discrepancies of the pattern of NO x weekly cycles.Section 4 concludes and discusses the findings of this study.

EPA air quality system (AQS) O 3 and NO x
Hourly surface O 3 and NO x concentrations are obtained from the EPA AQS measurement network (http://www.epa.gov/ttn/airs/airsaqs/detaileddata/downloadaqsdata.htm).Hourlyarchived O 3 data from about 1100 measurement sites are utilized and mapped onto 12 km CMAQ model grid cells.The total number of CMAQ grids, including AQS O 3 measurement sites, is 874 (AVHRR urban, forest, and others), 897 (GOME-2 HCHO/NO 2 < 1: GOME-2 NO xsaturated regime, 1 < GOME-2 HCHO/NO 2 < 2: mixed regime, GOME-2 HCHO/NO 2 > 2: NO x -sensitive regime) and 875 (CMAQ HCHO/NO 2 < 1: CMAQ NO x -saturated regime, 1 < CMAQ HCHO/NO 2 < 2: mixed regime, CMAQ HCHO/NO 2 > 2: NO x -sensitive regime) (Table 1).Hourlyarchived NO x data from 265 measurement sites are utilized and mapped onto 227 model grids (AVHRR urban, forest, and other), 240 model grids (GOME-2 NO x -saturated regime, mixed regime, NO x -sensitive regime), and 234 model grids (CMAQ NO x -saturated regime, mixed regime, NO x -sensitive regime) (Table 1).We filtered out some data for the study of the weekly cycles of surface NO x and O 3 because the weather conditions during the period of the remnant low of Tropical Storm Ana affecting the US (17-19 August 2009) produced large uncertainties regarding its meteorological impact on O 3 chemistry.Martin et al. (2004), mainly because consistent dynamical random errors have appeared since January 2009 in the OMI product (http://www.knmi.nl/omi/research/science/).Daily GOME-2 NO 2 and each orbit datum point of the HCHO column retrieval products are from http://www.temis.nl/airpollution.GOME-2 vertical NO 2 and HCHO column density data were prepared for a cell size of 0.25 • , but some of the data were filtered out with a cloud fraction of > 40 %.Details pertaining to the NO 2 retrieval algorithm using the DOAS approach and error analysis are provided by Boersma et al. (2004).The retrieval data used in this study are obtained from the European Space Agency (ESA) Tropospheric Emission Monitoring Internet Service (TEMIS) (http://www.temis.nl/airpollution/no2.htm).A detailed description of the HCHO column product can be found in De Smedt et al. (2008).TM4NO2A version 2.1 is used for the GOME-2 NO 2 column density and TEMIS version 1.2 is for the GOME-2 HCHO column density.(Houyoux et al., 2000).The emission sectors that vary according to meteorological conditions are simulated with various emission models in the PRE-processor to CMAQ (PRE-MAQ).In this study, monthly mean lateral boundary conditions derived from the GEOS-CHEM (Bey et al., 2001) global model simulation results are used for August 2009, in light of the findings of earlier CMAQ evaluations that emphasized the drawback of using the climatologically-averaged static boundary condition (e.g., Tong and Mauzerall, 2006;Tang et al., 2008).

AVHRR geographical regions and GOME-2 and CMAQ chemical regimes
Previously, Kaynak et al. (2009) derived three geographical sites -urban, rural, and rural-point (rural areas, including large EGUs) -for their NO x weekly cycle study.They chose seven urban sites based on Census 2000, eleven rural sites far from both urban regions and large EGUs, and more than 100 rural-point sites.In this study, instead of selecting specific sites, we divide the entire CONUS domain into three AVHRR-based geographical regions -urban, forest, and others (Fig. 1) -to examine how surface O 3 concentrations vary at all the available EPA AQS measurement sites over these geographical regions.Such a classification  (Anderson et al., 1976;Loverland et al., 2000).
differs from that used in previous studies to characterize the weekly cycles of O 3 , including the weekend effect in several urban regions (e.g., Cleveland et al., 1974;Elkus and Wilson, 1977;Vukovich, 2000;Marr and Harley, 2002;Fujita et al., 2003;Lebron, 2004;Qin et al., 2004;Blanchard and Tanenbaum, 2006;Shutters and Balling Jr., 2006;Blanchard et al., 2008;Yarwood et al., 2008).This study categories all of the model grid cells and EPA AQS measurement stations using a dominant AVHRR land use type.The AVHRR urban regions, which represent cities and their surrounding areas, comprise the smallest portion of the three geographical regions.
Previous remote sensing studies (e.g., Martin et al., 2004;Duncan et al., 2010) characterized two or three different regimes, including the NO x -saturated/NO x -sensitive and NO x -saturated/mixed/NO x -sensitive regimes.In this study, similar to Martin et al. (2004) and Duncan et al. (2010), we first utilize the ratio of GOME-2 HCHO to the NO 2 column density to divide the CONUS into three chemical regimes by using O 3 sensitivity derived from the GOME-2 ratio.(in red) and between CMAQ and CMAQ with a 30 % reduction in VOC emissions (in blue) in accordance with the ratio of GOME-2 or CMAQ HCHO to the NO 2 column density.No clear transitions take place among the NO x -sensitive, mixed, and NO x -saturated regimes, but changes in daytime surface O 3 are proportional to those in VOC emissions over the NO x -saturated regime (for low HCHO / NO 2 ratio < 1), but surface O 3 is highly sensitive to changes in NO x emissions over the NO x -sensitive regime (for a high HCHO / NO 2 ratio of > 2).Over the mixed regime, some changes in surface O 3 are affected by those in VOC emissions, but more changes are affected by NO x emissions changes (for ratios between 1 and 2).A high O 3 sensitivity to changes in NO x emissions was similarly shown as the ratio of HCHO to the NO 2 column density increased, as was also shown in previous studies by Martin et al. (2004) and Duncan et al. (2010).The CONUS domain is divided into three chemical regimes using two transitions (ratios are 1 and 2, see Fig. 2).The left panel of Fig. 3 identifies three chemical regimes -NO xsaturated, mixed, and NO x -sensitive -using the transition.This study further categorizes other chemical regimes using the ratios of CMAQ HCHO to NO 2 column density (see the right panel of Fig. 3) to investigate how weekly cycles of NO x and O 3 vary among the measurement stations of the GOME-2 and CMAQ chemical regimes of the CONUS.The CMAQ HCHO / NO 2 ratios are estimated at the GOME-2 overpass time.
Using one month-averaged GOME-2 NO 2 columns, some HCHO / NO 2 ratio data from CMAQ modeling grids are not considered when the NO 2 column density is less than 1 × 10 15 molecules cm −2 , typical of regions remote to anthropogenic sources (e.g., Martin et al., 2006;Russell et al., 2010) (Fig. 3).The distribution of two different chemical regimes derived by GOME-2 and CMAQ is generally consistent over the CONUS, but with clear differences over some metropolitan areas (e.g., Houston, New Orleans, and Tampa, in Fig. 3).Thus, over such regions, another classification could introduce different surface O 3 sensitivity in model simulation.A zoomed-in regionalized study over the Southeastern US may highlight differences among these chemical regime definitions.In addition, NO x -sensitive regimes (in red) from CMAQ are larger than those from GOME-2, implying higher surface O 3 sensitivity to changes in NO x emissions in the model simulation.
We also analyze how the AVHRR geographical regions (e.g., urban, other and forest regions) represent the ratio of the GOME-2 HCHO / NO 2 columns (see Fig. 4).The GOME-2 HCHO / NO 2 mean values (standard deviations) of the AVHRR urban, other and forest regions are 2.1 (1.3), 3.8 (1.8), and 4.0 (1.9), respectively.A large variability implies that each geographical region could be classified as two or more chemical regimes.The GOME-2 HCHO/NO 2 mean values (standard deviations) of the GOME-2 NO x -saturated, mixed, and NO x -sensitive regimes are 0.6 (0.2), 1.6 (0.2), and 4.4 (1.6), respectively.The GOME-2 chemical ratios for the AVHRR urban and forest regions are 2.1 and 4.0, respectively, which likely correspond to those for the GOME-2 mixed (1.6) and NO x -sensitive regimes (4.4), respectively.Interestingly, the chemical characteristic of the AVHRR urban regions is similar to that of the GOME-2 mixed regimes, which might affect the weekly cycle of the surface O 3 concentrations over the region.This issue will be described in detail in Sect.3.4.

Weekly variation of NO x emissions
To understand the weekly cycle of ground-level O 3 concentrations, we must evaluate weekly variations in the emissions of the O 3 precursor, NO x .For the sake of consistency with the evaluation of the weekly O 3 cycle, only the daytime (01:00-05:00 p.m., local time) emissions from CMAQ is used.Large weekday/weekend variations in NO x emissions are shown at EPA AQS stations over both the AVHRR Fig. 3. Three GOME-2-derived and CMAQ-derived chemical regimes using two transitions (1 and 2, see Fig. 2) using the ratio of the GOME-2 HCHO to NO 2 (left panel) and the CMAQ HCHO to NO 2 column density (right panel).The CMAQ ratios are estimated at the overpass time.Black represents the category 1 region (HCHO / NO 2 < 1, an NO x -saturated regime), green the category 2 region (1 < HCHO / NO 2 < 2, a mixed regime), and red the category 3 region (HCHO / NO 2 > 2, an NO x -sensitive regime).The cell size of the GOME-2 HCHO and NO 2 column density data (from http://www.temis.nl/airpollution/)differs from that of CMAQ (12 km), and thus, both GOME-2 and CMAQ column density data are interpolated into 36 km for this comparison.Ratios are estimated only when the GOME-2 NO 2 column density is larger than 1.0 × 10 15 molecules cm −2 .regions and the GOME-2 regimes (Fig. 5).Over the urban regions and the NO x -saturated regimes, changes in weekday/weekend NO x emissions are larger than those in other regions or regimes.From Friday to Saturday, NO x emissions decrease by 29.4 %, 26.7 %, and 25.3 % over the urban, other, and forest regions, respectively.The reduction in NO x emissions over the urban regions in this study is smaller than that in NO x emissions (about 34 %) in the Los Angeles basin, estimated by Yarwood et al. (2008).Reductions are similarly estimated by 29.4 %, 26.6 %, and 26.4 % over NO x -saturated, mixed, and NO x -sensitive regimes, respectively.The greatest NO x emissions are also shown on Friday over three AVHRR regions (7.76 mol s −1 , 1.94 mol s −1 , and 0.97 mol s −1 ) and the GOME-2 regimes (5.25 mol s −1 , 2.57 mol s −1 , and 1.75 mol s −1 ).Thus, changes in the absolute amounts of NO x emissions over the urban regions from weekdays to weekends are larger than those over the NO xsaturated regimes, but changes in NO x emissions over the other and forest regions are smaller than those over the mixed and NO x -sensitive regimes.

Weekly variation of AQS and CMAQ ground-level NO x
We investigate weekly anomalies of ground-level AQS and CMAQ NO x concentrations at EPA AQS stations over AVHRR geographical regions and GOME-2 chemical regimes, using only daytime (01:00-05:00 p.m., local time) observed and simulated NO x concentrations and estimate weekly anomalies by subtracting averaged NO x concentrations for each day from the average of all the available NO x concentrations during the month over the urban, forest, and mixed regions (see the two left columns of Fig. 6) or the NO x -saturated, mixed, and NO x -sensitive regimes (see the two right columns of Fig. 6).For each day (from Sunday to Saturday), the number of data, the standard deviation, and the standard error of the mean (SEM) of AQS NO x observations and corresponding CMAQ simulation results are also estimated (Table 2).The SEM is estimated the standard devi-ation dividing by the square root of the sample size (http: //en.wikipedia.org/wiki/Standarderror), ranging from 0.1-0.9.The low weekend and high weekday NO x is clearly observable at the stations located in all the AVHRR regions and GOME-2 regimes (Fig. 6).Over the urban regions and the NO x -saturated regimes, AQS observations show the smallest NO x concentrations during weekends (on Sunday) and the largest during weekdays (on Thursday and Friday).The weekly pattern of NO x concentrations over urban regions is similar to that of the NO 2 column density and the NO 2 mixing ratio in previous studies (e.g., Beirle et al., 2003;Shutters and Balling Jr., 2006;Kaynak et al., 2009).In particular, Kaynak et al. (2009) showed the smallest NO 2 column density over the urban regions on Sunday and the largest on Wednesday, Thursday, and Friday (e.g., Wednesday for Chicago; Thursday for Houston, Atlanta, New York, and Phoenix; Friday for Los Angeles and Seattle).The CMAQ simulation shows a similar pattern for the smallest NO x concentrations on Sunday and the largest on Thursday (over urban regions) and Wednesday (over NO x -saturated regimes) instead of Friday, as in AQS.In general, the model-simulated NO x peaks occur one or two days earlier than the AQSobserved NO x peaks.Interestingly, the weekly pattern of the NO x emissions inventory is not consistent with this simulated NO x pattern.Note that the NO x emissions for Friday are slightly greater or similar to those for Wednesday over the regime or region (Fig. 5).However, the simulated NO x concentrations significantly decrease from Wednesday to Friday over the regime or region, which introduces a large discrepancy between the weekday NO x pattern of the model and that of the observation.The possible reason for this difference is described in Sect.3.6.This shift in the highest NO x day (Fig. 6) may contribute to the shift in the lowest O 3 day (Fig. 7), particularly at stations over urban regions or NO xsaturated regimes because of the relatively greater amount of NO x emissions than over other regions or regimes (Fig. 5).Details pertaining to the shifts in low O 3 peak days are described in the next section.
The weekly cycles of NO x concentrations at AQS measurements sites over the AVHRR other region and GOME-2 mixed regimes in AQS are more or less similar to those in the AVHRR urban region and the GOME-2 NO x -saturated regimes, respectively.CMAQ shows that the weekly cycle of NO x concentrations at AQS sites over the AVHRR other region are similar to those over the AVHRR urban region.The simulated high-peak NO x days (Tuesday-Thursday) over the NO x -saturated regime shifted to late weekdays (Wednesday-Friday) over the mixed regime.In other words, although simulated weekly cycles of NO x concentrations of urban and other regions are similar, simulated peak NO x occurs two days later over mixed regimes than it does over NO xsaturated regimes.Interestingly, the simulated NO x cycle at AQS stations over mixed regimes is similar to the observed cycle at stations over NO x -saturated regimes.It is most likely the result of the closeness of the mixed regime to the NO xsaturated regime.
The AQS shows the largest NO x concentration at stations over forest regions and NO x -sensitive regimes on Friday, but CMAQ shows the largest NO x concentrations in these areas on Thursday (Fig. 6).A study by Kaynak et al. (2009) also showed similar patterns of NO 2 column densities over their rural areas (large NO 2 column densities on Thursday and Friday and small column densities on Saturday and Sunday).The pattern of NO x cycles at stations over the AVHRR forest regions in AQS and CMAQ is similar to that over the AVHRR other regions in AQS and CMAQ, respectively.Over forest regions and NO x -sensitive regimes, the highest NO x peak day occurs one day earlier in CMAQ (on Thursday) than in AQS (on Friday), which might be closely related to the one-or two-day shifts of low O 3 peak days during weekdays in CMAQ compared to those in AQS.

Weekly anomalies of AQS and CMAQ ground-level O 3
Weekly anomalies of ground-level O 3 concentrations from AQS and CMAQ are compared at EPA AQS measurement stations over the AVHRR geographical regions and the GOME-2 chemical regimes.Weekly O 3 anomalies and weekly NO x anomalies are estimated in the same way.For each day, the number of data, the standard deviation and the SEM of AQS O 3 observations and corresponding CMAQ modeling results are also estimated (Table 3).Over urban     Blanchard et al. (2008) showed more diverse highest and lowest peak days than our study, likely resulting from the local characteristics of the measurement sites.
Previous studies clearly showed the weekend effect in several polluted areas (e.g., Cleveland et al., 1974;Elkus and Wilson, 1977;Vukovich, 2000;Marr and Harley, 2002;Fujita et al., 2003;Lebron, 2004;Qin et al., 2004;Shutter and Balling Jr., 2006;Yarwood et al., 2008), but this study finds no weekend effect at AQS stations over the AVHRR urban regions in AQS nor in CMAQ because the AVHRR urban region has a similar chemical environment to the GOME-2 mixed regime, instead of the GOME-2 NO x -saturated regime.However, at other AQS sites over NO x -saturated regimes, the weekend effect is clearly shown in both AQS and CMAQ, indicating a peak ground-level O 3 concentration on Sunday both in AQS and CMAQ.Both AQS and CMAQ also show the second highest peak on Monday.At AQS stations over NO x -saturated regimes, AQS observations indicate the smallest O 3 on Friday, but the CMAQ simulation shows the smallest on Thursday (see Fig. 7).The difference between the observed and simulated lowest O 3 days is most likely the result of temporal variations of NO x concentrations over the regime (see Fig. 6).
At AQS stations over the AVHRR other regions, the largest and smallest O 3 concentrations are shown on Wednesday and Sunday, respectively.The CMAQ model simulation results show the same highest and lowest O 3 days.Both AQS and CMAQ show negative anomalies on Saturday (Fig. 7), which are consistent with negative anomalies in NO x concentrations (Fig. 6).At AQS sites over the mixed regime, both the AQS observation and the CMAQ simulation show the largest O 3 concentrations on Tuesday.The AQS observation shows the smallest O 3 concentrations on Sunday, but the CMAQ simulation shows the smallest on Thursday.The reason for negative anomalies on Thursday and Friday in CMAQ is not clear, but the AQS observation also shows similar negative anomalies on Friday and Saturday.As we explained earlier, the differences between the lowest O 3 days in AQS and those in CMAQ are similarly shown over the AVHRR urban region.Similarly, the low O 3 peak day occurs one day earlier in CMAQ (Thursday) than in AQS (Friday) during the weekdays.The transport of small O 3 concentrations over NO xsaturated regimes might introduce small O 3 concentrations over its neighboring mixed regime.At AQS stations over forest regions, both AQS and CMAQ show the largest O 3 concentrations on Tuesday, but they show the smallest on Sunday or Thursday, respectively.At AQS stations over NO xsensitive regimes, both AQS and CMAQ show the largest O 3 concentrations on Tuesday, the second largest on Wednesday, and the smallest on Sunday.The highest O 3 peak day over the NO x -sensitive regime is similar to that of the AVHRR forest region, likely due to the similar chemical environment of the regime to that of the region (Fig. 4).

Weekly anomalies of NO x and O 3 over the CMAQ chemical regimes
In this section, we further investigate the effects of the difference between the GOME-2-and CMAQ-derived chemical regimes (see Fig. 3) on weekly cycles of surface NO x and O 3 concentrations at AQS measurement station sites.This study compares the weekly NO x and O 3 cycles at the AQS stations over the two different NO x -saturated regimes (from GOME-2 and CMAQ).The weekly cycles of NO x at corresponding stations in AQS and CMAQ over the two different chemical regimes (see the two right columns of Fig. 6 and the two left columns of Fig. 8) are generally similar.Table 4 represents the number of data, the standard deviation, and the SEM of AQS NO x or O 3 observations and corresponding CMAQ simulation results at AQS stations over the CMAQ-derived chemical regimes.As discussed in the previous section, the O 3 weekend effect in AQS and CMAQ (see the two right columns of Fig. 7) are clearly observable at AQS stations over the GOME-2 NO x saturated regime, but high O 3 anomalies on Sunday and Monday are not obvious at AQS stations over the CMAQ NO x saturated regime in AQS (see the third column of Fig. 8).Interestingly, the pattern of weekly O 3 cycles over the CMAQ NO x -saturated regime is similar to that over the AVHRR urban region (see the first column of Fig. 7).This finding indicates that CMAQderived NO x -saturated regime stations might be characterized as AVHRR urban region stations that, in reality, include some AVHRR urban region stations.This finding further suggests that utilizing the GOME-2-derived photochemical indicator might define NO x -saturated regime stations better than the CMAQ-derived photochemical indicator.

The impact of the GOME-2-derived emissions inventory on the weekly NO x pattern
As we described in Sect.2.2, GOME-2 NO 2 retrievals were obtained for August 2009 and used to calculate monthlyaveraged NO 2 column density for the CONUS domain, and an equivalent monthly mean column-integrated value of NO 2 concentrations was calculated from the CMAQ with NEI 2005.The ratios of the CMAQ NO 2 column density to the GOME-2 NO 2 column density are also estimated (Fig. 9).A comparison of modeled and satellite-observed NO 2 columns exhibited general overestimates in urban areas close to concentrated population areas over the CONUS.
In particular, the CMAQ model over-predicted the NO 2 column density in the urban areas over Houston, Texas, New Orleans, Louisiana, Tampa and Jacksonville, Florida, Portland, Oregon, Minneapolis, Minnesota, Tulsa, Oklahoma, Kansas City, Kansas, Charlotte and Raleigh, North Carolina, and Los Angeles, California.Even though other regions exhibited various prediction trends, Martin et al. (2006) and Choi et al. (2008Choi et al. ( , 2009) also showed similar over-prediction trends of the GEOS-CHEM model (using NEI1999) and the Regional chEmical trAnsport Model (REAM) model (using NEI 1999 or NEI 1999 with a 50 % reduction in EGU and non-EGU point sources, respectively) in urban areas over the southern US.Unlike in the results discussed in the previous paragraph, the baseline model generally underpredicts NO 2 column density in areas near San Francisco, California, Springfield, Illinois, Bloomington, Indiana, Binghamton, New York, and Scranton, Pennsylvania, and large portions of rural area over the CONUS (Fig. 9).Napelenok et al. (2008) also showed that NO 2 column densities are generally underpredicted in rural areas over the eastern US in the baseline CMAQ model.Interestingly, over San Francisco and its neighboring region, the GEOS-Chem (Martin et al., 2006) and the REAM (Choi et al., 2008(Choi et al., , 2009) ) CTMs also showed underprediction of the NO 2 columns.
To analyze the effect of the GOME-2-derived emissions inventory on the weekly cycles of surface NO x over the AVHRR geographical regions or the GOME-2 chemical regimes, we performed additional simulations, including the GOME-2-derived NO x fossil-fuel emissions.The GOME-2derived NO x fossil-fuel emissions inventory (E t ) is first estimated following Martin et al. (2003) and Choi et al. (2008) by fitting E t to a priori bottom-up emission E a from NEI 2005 with the ratio of the retrieved NO 2 column ( r ) to the corresponding simulated column at the GOME-2 overpass time ( s ): For a new emissions inventory, we estimated the ratios of the CMAQ NO 2 column density to the GOME-2 NO 2 column density.For this application, we estimated both the ratios of the CMAQ NO 2 columns to GOME-2 NO 2 columns   in a 36 km spatial resolution in order to effectively avoid the adverse effect of the transport of NO 2 tracers in a fine resolution and the GOME-2-derived NO x emissions inventory by applying the ratios to the standard NEI 2005 (Fig. 9).The GOME-2-derived NO x emissions (426 Gg N) are 7.8 % less than the NEI 2005 (462 Gg N) emissions over the US (Fig. 9). Figure 10a and b show a comparison of the simulated surface NO x concentrations from the CMAQ with NEI 2005 and the second CMAQ with the GOME-2-derived emissions inventory.The simulated NO x concentrations from the second CMAQ are significantly less at EPA stations, particularly over the AVHRR urban region and the GOME-2 NO xsaturated regime (Fig. 10a and b).The number of available AQS data, the standard deviations and the SEM of corresponding CMAQ simulation results with GOME-2-derived emissions are estimated (Table 5).Interestingly, the standard deviation and the SEM values from the 2nd CMAQ simulations become similar to those from the AQS observations.We also found that the large reduction in NO x emissions affects the weekly pattern of surface NO x at stations over the urban region or the NO x -saturated regime (Fig. 10a and b).As we described earlier, the CMAQ-simulated NO x peak days occur in one-or two-day shifts compared with those from the EPA AQS observations (Fig. 6).The trend in weekly NO x emissions (see Fig. 5) reveals similar emissions from Monday to Friday, but the standard CMAQ shows a peak NO x day on Wednesday.The simulated high peak occurrence on Wednesday is probably related to the longest NO x lifetime in the middle of the week, which could be caused by the overestimated surface NO x concentrations or favorable meteorological condition for NO x tracers on Wednesday in CMAQ.The details need to be further investigated.Note that large reductions in NO x emissions mitigate rapid changes in simulated NO x concentration (an increase from Monday to Wednesday and a decrease from Wednesday to Friday) during the week-days at stations over the urban region or the NO x -saturated regime (see the second and third columns of Fig. 10a and b).

Conclusion and discussion
This study compares the weekly cycles of surface ozone and its precursors simulated by the CMAQ model to that observed at the EPA AQS stations over satellite derived geographical regions and chemical regimes over the contiguous United States.We found that the CMAQ model generally captures the weekly cycles of ground-level NO x and O 3 as observed at AQS sites, except occasional shifts between positive and negative anomalies during weekdays.Over AVHRR other and forest regions and GOME-2 mixed and NO xsensitive regimes, both AQS and CMAQ show positive NO x anomalies during weekdays and negative NO x anomalies during weekends.Similarly, AQS and CMAQ show negative O 3 anomalies during weekends and positive O 3 anomalies during weekdays.However, AQS observations do not reveal any weekend effects (high O 3 weekly anomalies during weekends) at AQS stations over the AVHRR urban region, but they clearly show weekend high O 3 anomalies at other AQS stations over the GOME-2 NO x -saturated regime, suggesting that characterizing the CONUS as GOME-2 chemical regimes could benefit the analysis of weekly cycles (including the weekend effect).Over the AVHRR urban region and the GOME-2 NO x -saturated regime, the greatest negative O 3 day differs slightly in AQS (on Friday) and CMAQ (on Thursday).A shift in the largest negative O 3 anomaly day in CMAQ is likely to result from a shift in the positive NO x day in the model, unlike shifts in AQS.comparison of CMAQ and GOME-2 NO 2 columns proves to reduce the large discrepancy in the absolute amounts of surface NO x concentrations and the weekly pattern of surface NO x concentrations between CMAQ and EPA AQS, in particular over the AVHRR urban region and the GMOE-2 NO x saturated regime.This study also analyzes and compares the weekly cycles of NO x and O 3 over the two different chemical regimes from GOME-2 and CMAQ.A weekend effect is clear at AQS stations over the GOME-2-derived NO x -saturated regime, but it is not at other AQS stations over the CMAQ-derived NO x -saturated regime.After selecting a specific region of interest as a chemical regime using a satellite-derived photochemical indicator, we are able to compare observed O 3 weekly cycles with corresponding simulation results in order to evaluate model performance.Through the comparison, we are able to understand and even forecast the highest and lowest O 3 anomaly days over the region.
The results of this research warrant future research that addresses several remaining issues.First, our definition of a "chemical regime" is loosely associated with satellite or land use-based characteristics.Besides the inherent uncertainties from satellite retrievals and land use data, the categorization approach used to derive chemical regimes may not collocate with the actual chemical environment in the lower troposphere, where surface ozone is photo-chemically produced.
The satellite HCHO and NO 2 columns represent the vertical accumulation of corresponding species from the ground to the top of the troposphere.The use of column data can be justified by the fact that the majority of emission sources of NO x and VOCs originate at (mobile, area, and biogenic) or near (power plant sources, typically with chimneys in the lower km) the earth's surface.Nevertheless, the potential impact of the vertical distribution of these species, particularly that caused by differences in the chemical lifetime and emission source distribution, on the determination of a chemical regime must also be taken into account.Second, our study is also limited to one month (August 2009).Using a single month for analysis could be problematic in terms of statistics for looking at weekday-weekend differences.Longer-time scale simulations need to average out the impact of synoptic variability on the weekly variations of O 3 and its precursors.As the number of the sampling data increases, the standard error of the mean (SEM) of EPA AQS observations and corresponding simulation results for each day decreases.Therefore, further investigation of changes in the chemical regime over space during other seasons/years could expand our understanding of this phenomenon.Such future investigations would need to determine the chemical environment (e.g., satellite-derived HCHO / NO 2 ratios) for each month in order to distinguish the chemical regime stations.In addition, the anthropogenic and biogenic emissions are expected to change from season to season and from year to year in various directions, resulting in a so-called "seasonal transition" of the chemical regime (e.g., Jacob et al., 1995;Martin et al., 2004).While the static land use-based indicator may not capture such seasonal changes, a satellite-based dynamic indicator could more realistically reflect the temporal evolution of the chemical environment.
Figure 2 represents the differences between the daytime (01:00-05:00 p.m., local time) surface O 3 of the baseline CMAQ and that of the CMAQ with a 30 % reduction in NO x emissions

Fig. 2 .
Fig. 2. Differences between the surface O 3 in the baseline CMAQ and that in CMAQ with 30 % NO x reduction (in red, baselinesensitivity) or a reduction in VOC emissions (in blue, baselinesensitivity) in accordance with the ratios of the GOME-2 HCHO to NO 2 column density (upper panel) and the CMAQ HCHO to NO 2 column density (lower panel).The differences are estimated by averaging the data for the daytime (01:00-05:00 p.m., local time) during August 2009 only when GOME-2 NO 2 or the CMAQ NO 2 column density is larger than 1.0 × 10 15 molecules cm −2 .So that the trend can be more clearly observable, all five O 3 differences are averaged according to the GOME-2 or CMAQ ratios.

Fig. 4 .
Fig. 4. The ratio of GOME-2 HCHO columns to NO 2 columns over AVHRR-derived geographical regions (i.e., urban, others, and forest regions, left panel) and GOME-2 derived chemical regimes (i.e., NO x -saturated, mixed, and NO x -sensitive regimes, right panel) for August 2009.The bar represents the standard deviation of GOME-2 HCHO / NO 2 ratio.

Fig. 5 .
Fig. 5. Weekly variation in NO x emissions at EPA AQS stations over three AVHRR-derived geographical regions (i.e., urban, other, and forest regions, left column) and GOME-2-derived chemical regimes (i.e., NO x -saturated, mixed, and NO x -sensitive regimes, right column) in CMAQ for August 2009.For the sake of consistency with the daytime O 3 comparison in Fig. 7, only daytime (01:00-05:00 p.m., local time) NO x emissions are used; some data are filtered out from 17-19 August 2009, when Tropical Storm Ana strongly affected air quality over the eastern US.

Fig. 6 .
Fig. 6.Weekly anomalies of AQS-observed and CMAQ-simulated ground-level NO x concentrations at EPA AQS stations over AVHRRderived geographical regions (i.e., urban, other, and forest regions, left two columns) and GOME-2-derived chemical regimes (i.e., NO xsaturated, mixed, NO x -sensitive regimes, right two columns) for August 2009.For the sake of consistency with the daytime O 3 comparison in Fig. 7, only daytime (01:00-05:00 p.m., local time) NO x concentrations are used; some data are filtered out from 17-19 August 2009, when Tropical Storm Ana strongly affected air quality over the eastern US.

Fig. 8 .
Fig. 8. Weekly anomalies of AQS-observed and CMAQ-simulated ground-level NO x at EPA AQS NO x stations (left two columns) and O 3 concentrations at EPA AQS O 3 stations (right two columns) over CMAQ-derived chemical regimes (i.e., NO x -saturated, mixed, and NO xsensitive regimes, see the right panel of Fig. 3) for August 2009.Only daytime (01:00-05:00 p.m., local time) NO x and O 3 concentrations are used; some data are filtered out from 17-19 August 2009, when Tropical Storm Ana strongly affected air quality over the eastern US.

Fig. 9 .
Fig. 9.The ratio of the monthly-averaged CMAQ NO 2 columns versus GOME-2 NO 2 columns (left panel, both GOME-2 and CMAQ column data are interpolated for 36 km for this comparison), the monthly-averaged NO x emissions from EPA National Emissions Inventory 2005 (NEI 2005, middle panel, 462 Gg N over the US) and the monthly-averaged NO x emissions from the GOME-2-derived emissions inventory (right panel, 426 Gg N over the US) for August 2009.
Fig. 10.(a) Weekly mean variations of AQS-observed, CMAQ-simulated, and the 2nd CMAQ-simulated (with GOME-2-derived emissions inventory) ground-level NO x concentrations at EPA AQS stations over AVHRR-derived geographical regions (i.e., urban, other, and forest regions) for August 2009.For the sake of consistency with the daytime O 3 comparison in Fig. 7, only daytime (01:00-05:00 p.m., local time) NO x concentrations are used; some data are filtered out from 17-19 August 2009, when Tropical Storm Ana strongly affected air quality over the eastern US.Note that the index of the vertical y-axis for the urban region stations (0-25 ppbv) differs from the indices (0-10 ppbv) of other and forest region stations.(b) Same as in (a), but over GOME-2-derived chemical regimes (i.e., NO x -saturated, mixed, NO x -sensitive regimes).Note that the index of the vertical y-axis for NO x -saturated regime stations (0-25 ppbv) differs from the indices (0-10 ppbv) of mixed and NO x -sensitive regime stations.

Table 2 .
The number of data (weekly anomaly), the standard deviation, and the standard error of the mean (SEM) of EPA AQS observations and CMAQ simulations for NO x measurement sites over the AVHRR-derived geographical regions and the GOME-2-derived chemical regimes.

Table 3 .
The number of data (weekly anomaly), the standard deviation, and the standard error of the mean (SEM) of EPA AQS observations and CMAQ simulations for O 3 measurement sites over the AVHRR-derived geographical regions and the GOME-2-derived chemical regimes. 2

www.atmos-chem-phys.net/12/6291/2012/ Atmos. Chem. Phys., 12, 6291-6307, 2012 6302 Y. Choi et al.: Summertime weekly cycles of observed and modeled NO x and O 3Table 4 .
The number of data (weekly anomaly), the standard deviation, and the standard error of the mean (SEM) of EPA AQS observations and CMAQ simulations for NO x and O 3 measurement sites over the CMAQ-derived chemical regimes.

www.atmos-chem-phys.net/12/6291/2012/ Atmos
Compared to GOME-2 NO 2 column in August 2009, the CMAQ NO 2 column is generally larger, particularly over the AVHRR urban region and NO x -saturated regime.Adjusting fossil-fuel NO x emissions (from 462 to 426 Gg N over the US) based on a . Chem.Phys., 12, 6291-6307, 2012

Table 5 .
The number of data (NO x concentration), the standard deviation, and the standard error of the mean (SEM) of CMAQ simulations with GOME-2 adjusted emissions for NO x measurement sites over the AVHRR-derived geographical regions and GOME-2-derived chemical regimes.