We demonstrate a direct connection between large-scale circulation patterns and surface O3 using atmospheric observations obtained during winters 2002 and 2003. Measurements at two rural sites in the northeastern U.S. revealed that median mixing ratios of O3 in winter 2003 were increased by up to 80% compared to 2002, and greatly exceeded previous spring annual maximums. To explain this we propose that strong meridional flows in winter 2003 frequently transported O3-rich mid-tropospheric air masses from high latitudes to the northeastern U.S. while cooling regional climate 4.4°C below normal. Our measurements also show that an exceptionally elevated spring O3 maximum occurred in 2003. The impact from this winter enhancement on the levels of O3 and other species during the following months will be largely driven by actual climatic conditions.
 Air quality and climate are interwoven through complex interactions of atmospheric dynamical, physical, and chemical processes on time scales ranging from seasons to centuries. Ozone (O3), a secondary photochemical product, shapes overall air quality by exerting critical forcings on the oxidizing capacity and radiative balance of the troposphere [Thompson et al., 1990]. Long-term measurements show large seasonal variability in the level of surface O3 at remote sites across the northern hemisphere with monthly mean values peaking at 40–50 parts per billion by volume (ppbv) in springtime (available at http://www.cmdl.noaa.gov/info/ftpdata.html). The level of O3 in these photochemically aged air masses is commonly referred to as its background mixing ratio [Talbot et al., 1994].
 A key factor driving fluctuations in the O3 seasonal cycle is changes in its background mixing ratio. Several numerical studies indicate that the mechanisms responsible for regional shifts in the background level of O3 include stratosphere-troposphere exchange, photochemical processes and intercontinental transport [Janach, 1989; Jacob et al., 1999; Lin et al., 2000; Naja et al., 2003]. Large-scale circulations, an important climate variable, can transport air masses from high latitudes with decreased total column O3 to midlatitudes [Henriksen and Roldugin, 1995; Hadjinicolaou et al., 1997; Callis et al., 1997; Harris et al., 2003]. Consequently, vertical profiles of O3 at midlatitudes can be altered and surface O3 elevated due to subsidence of O3-rich air masses from the Arctic mid-troposphere. This hypothesis is supported by our observations of enhanced surface O3 in the northeastern U.S. during winter 2002–03 compared to 2001–02 (hereafter winters 2003 and 2002).
 Here we examine continuous year-round records of O3 mixing ratios collected during 2001–2003 at two sites operated by the Atmospheric Investigation, Regional Modeling, Analysis and Prediction (AIRMAP) program. These sites are located 125 km apart in New Hampshire at 406 m elevation at Castle Springs (CS, Moultonborough - 43.75°N, 71.35°W) and near sea-level (21 m) at Thompson Farm (TF, Durham - 43.11°N, 70.95°W). Our analysis was centered on a comparison of O3 during winters 2002 and 2003 which exhibited extreme climatic differences in temperature and large-scale transport regimes.
 Ozone was measured at the AIRMAP sites using a Thermo Environmental Instruments model 49C-PS, which is based on UV photometric detection at 254 nm with a limit of detection ∼1.0 ppbv. Observations of nitric oxide (NO) and total reactive nitrogen (NOy = NO + NO2 + HNO3 + PAN + RNO2 + aerosol NO3− + …) were used to compare the photochemical activities of the two winters. Nitric oxide and NOy were measured using a custom modified Thermo Environmental Instruments model 42C. We used a molybdenum converter heated to 350 ± 0.5°C to reduce NOy compounds to NO which was subsequently measured by NO2 chemiluminescence. The conversion efficiency was checked daily by passing ∼25 ppbv of isopropyl nitrate through the converter. An Aadco purifier provided air for zeroing and dilution of ∼0.5 and ∼50 ppbv NO standards (Scott Marrin, Inc., ±2% NIST certified) to ∼1 and 10 ppbv for NO and NOy calibrations. Zeroing and calibration procedures were conducted every 12 hours. One minute averaged data, used in this study, is obtained from each site over the Internet every 15 minutes and displayed in near-real time on the AIRMAP web page (www.airmap.unh.edu).
 Backward air parcel trajectories for the two winters were simulated using the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model [Draxler, 1999] driven by Final Run (FNL) reanalysis wind fields obtained from the National Center for Environmental Prediction [Kalnay et al., 2000]. The trajectories were initiated at TF with starting heights of 500 and 1500 m, which are representative of the planetary boundary layer.
3. Contrasting Climatic Conditions During Two Winters
 Seasonal climate records obtained from the NOAA Northeast Regional Climate Center (available at http://met-www.cit.cornell.edu/climate/Climate_summary.html) show that the mean surface air temperature in New Hampshire during winter 2003 was −7.9°C, which was 5.6°C colder than winter 2002 and 4.4°C below the normal, defined as the temperature averaged regionally over 1961–1990. The northeastern U.S. regional (11 states) mean surface temperature was −5.4°C in winter 2003 as opposed to 0.1°C in winter 2002, the warmest winter on record for this portion of the country.
 These wintertime climate statistics are corroborated by comparison of the seasonal mean geopotential heights at 850 (not shown) and 500 hPa between winters 2002 and 2003 (Figure 1) (available at http://www.cdc.noaa.gov/cgi-bin/Composites/printpage.pl). The trough in winter 2002 was shallow without a ridge-trough pattern at 850 or 500 hPa; the geopotential height contour of 5500 m at 500 hPa only reached southward to central Massachusetts (42.5°N latitude). Consequently, more zonal flow occurred reducing the possibility of a high latitude (>50°N) influence on air masses over the northeast. In 2003 the winter trough over North America was markedly deeper than in 2002. The 500 hPa geopotential height contour of 5500 m formed a ridge along the west coast of Canada and a deep trough on the east coast extending southward to Washington D.C. (39°N latitude). This positioning indicates strong meridional flow in the free troposphere in winter 2003, facilitating southward transport and large-scale subsidence of Arctic air masses over the northeastern U.S.
4. Impact on Surface O3
 The time series of O3 mixing ratios at TF and CS during the fall, winter, and spring seasons of 2001–2003 were examined to identify the salient temporal variations in the northeast over the past two years (Figure 2). Year-to-year differences were the greatest for winter with O3: (1) rarely exceeding 40 ppbv in 2002 but elevated above it 88% of the time in 2003 and (2) at TF only, depleted to values <5ppbv on 42% of the nights in 2002 but exhibiting negligible depletion in 2003. Nocturnal depletion of O3 commonly occurs at low elevation sites (e.g., TF) due to dry deposition in an inversion layer near the surface [Hastie et al., 1993], while it typically does not occur at higher elevation sites such as CS [Zaveri et al., 1995]. The two-fold higher nighttime average wind speeds at TF during winter 2003 (2.2 m s−1) compared to 2002 (1.0 m s−1) often prevented the inversion layer from forming, leading to a constant influx of O3 from above and subsequently a much reduced occurrence of nighttime depletion.
 A compelling rise in O3 mixing ratios began in fall 2002 with the frequency of values >40 ppbv escalating during winter 2003 and manifesting itself with exceptionally high seasonal O3 values of 45 and 53 ppbv at TF and CS, respectively, in spring 2003. To illustrate the trends more clearly we calculated 20-day moving averages that are depicted as smoothed lines on Figure 2. Moving averages for the two years diverged abruptly in late November 2002 when the upward trend in O3 started, and they did not coincide again until early May 2003. The close tracking at the two sites is indicative of a regional signal.
 To summarize the time series of O3 a statistical overview is presented in Table 1. The medians were nearly identical at the two sites, with the 2003 winter and spring values exceeding the previous springtime annual maxima. The median and 10th percentile O3 mixing ratios exhibited maximum values in spring and minimum ones in fall in accordance with other sites across the northern hemisphere [Monks, 2000; Oltmans and Levy, 1994]. While the year-to-year variation in other seasons was insignificant, winter O3 values increased markedly in 2003. The median and 10th percentile values at TF rose from 23 and 6 ppbv in 2002 to 41 and 19 ppbv in 2003, respectively, and they were enhanced similarly at CS. The higher medians for 2003 corresponded to an 80% increase at TF and a 45% increase at CS. This difference is related to the frequency of occurrence of nocturnal O3 depletion, which is dependent on site elevation (Figure 2), and it suggests that the regional enhancement in 2003 was much greater at lower elevation sites.
Table 1. Comparison of O3 Mixing Ratio Statistics (ppbv) at Two AIRMAP Sites for 2001–2003 with Emphasis on Wintertime Values for 2003
10% refers to the 10th percentile. TF means Thompson Farm and CS Castle Springs.
 The enhancements in O3 are more clearly illustrated by the seasonal mean diurnal cycles shown in Figure 3. Comparison of the diurnal cycles between the two years reveals a constant offset of 16 and 12 ppbv throughout the entire day at TF and CS, respectively, during 2003. The lack of a more pronounced mid-day increase suggests minimal in situ photochemical production of O3. Note that the two sites had nearly identical daily peak values while the diurnal amplitude in O3 (i.e., daily maximum–minimum) was much more pronounced at TF than CS due to lower nighttime mixing ratios. Merging of the afternoon maximums at the two sites likely resulted from downward mixing of air masses, which was accentuated in winter 2003 due to large-scale subsidence of Arctic air over the northeastern U.S., as demonstrated later by our trajectory results.
 Diurnal cycles in O3 are commonly attributed to photochemical production and vertical mixing during the daytime and surface deposition and titration in an inversion layer at night. The amplitude of the diurnal cycle remained the same between the two winters, being around 10 ppbv at TF and <3 ppbv at CS. Since CS is an elevated site which has minimal O3 depletion at night, the 8 ppbv maximum difference in nighttime O3 mixing ratios between TF and CS is indicative of the magnitude of regional surface deposition. This assumption is justified since titration of O3 by NO appears to have a minimal effect on the daily cycles (R. Talbot et al., Diurnal characteristics of surface-level O3 and other trace gases in New England, manuscript in preparation, 2004). The remaining 2 ppbv at TF is attributed to daytime vertical mixing and photochemical production, which yields an upper limit of 2 ppbv for a local contribution from photochemistry at the two sites. The AIRMAP observations showed that the levels of NO and NOy in winter 2003 were nearly a factor of 2 smaller than those in 2002. It further confirmed the presence of an aged chemical environment in winter 2003 that was not conducive to enhanced photochemical activity.
 The transport regimes during the two winter seasons were also investigated using daily backward air parcel trajectories. A statistical summary of this analysis is presented in Table 2. There were a total of 10 and 77 days in winters 2002 and 2003 respectively (i.e., 11% versus 88% of the season) with daily maximum O3 > 40 ppbv. Trajectories initiated at 500 m in winter 2003 consistently showed northerly flows, and on 90% of the days with O3 > 40 ppbv the air masses originated near 64°N at an average altitude of 2068 m. On the few days in winter 2002 with O3 > 40 ppbv, trajectories initiated at 500 m originated around 55°N with an average height of 1510 m. Statistics similar to those at 500 m were obtained from the trajectories initiated at 1500 m, but with somewhat higher ending altitudes. Together these cold northerly flows induced regional subsidence in winter 2003 bringing mid-tropospheric (1.5–6 km altitude) air to the surface at midlatitudes.
Table 2. Statistical Summary of Backward Trajectory Analysis for Winters 2002 and 2003
Percentage of days with O3 > 40 ppbv and five-day trajectory ending points >50°N.
Initial starting height of trajectories.
Mean latitude and height of five-day trajectory ending points.
Logan  showed that in January monthly mean O3 levels near the surface at 38°N–48°N are around 20 ppbv or slightly lower, similar to our observations in winter 2002 (Table 1). Furthermore, it was shown that climatological vertical profiles of O3 mixing ratios at latitudes of 67°N–79°N vary from 20 ppbv in the lower troposphere to 60 ppbv in the middle troposphere, as evidenced in recent airborne measurements by Ridley et al. . These O3-rich mid-tropospheric Arctic air masses can be transported by northerly flows to the northeastern U.S., as simulated by our trajectories, and subsequently enhance O3 levels there on a regional scale.
 In addition to advective transport in the free troposphere, an increased frequency of tropopause folding events could provide a source for elevated surface O3. In winter 2003 90% of our backward trajectories originated in the high latitude mid-troposphere, a region potentially impacted greatly by inputs of stratospheric O3. We examined the cross-sectional distribution of potential vorticity (PV) during winters 2002 and 2003 to identify seasonal shifts in PV, a good indicator of changes in stratosphere-troposphere exchange [Danielsen, 1968]. The meridional cross section of PV averaged over the longitudes 90°W–68°W (covering the states east of Mississippi River) showed little variation in PV from winter 2002 to 2003, suggesting fairly constant inputs of stratospheric air over the two years. This implies that enhanced northerly advection of high latitude air was the most important factor influencing O3 levels over the northeast in winter 2003.
 Our findings suggest important implications of greatly increased wintertime O3 for levels of photochemically active species in the atmosphere over the northeastern U.S. during the spring and early summer of 2003. However, photochemical processes are constrained by climate, and therefore subsequent O3 production in the warm season depends on the actual climatic state. Tools such as three-dimensional chemical transport models are needed to predict the impact of such a dramatic elevation in background O3 on atmospheric chemistry and air quality due to the multitude of complex chemistry-climate interactions.
 We gratefully acknowledge the efforts of the AIRMAP team at UNH for operation of the monitoring sites and preparation of final data. This work was supported by NOAA under grant #NA17RP2632.