Long Term Trends in New York : PM

Measurements of PM2.5 mass and particle components over a 14–15 year period are used to explore trends at urban and rural locations across New York State. Such data are used to determine compliance with national ambient air quality standards (NAAQS), as well as track the effectiveness of reductions in source pollutant emissions. Since 2000 annual mean PM2.5 mass shows a downward trend with decreases of 4–7 μg m in urban areas versus 3–4 μg m for rural background sites. Much of this change can be attributed to particle sulfate (SO4) and nitrate (NO3) which showed annual decreases of 2–3 μg m and 0.5–1.0 μg m, respectively. Determining accurate trends for the carbonaceous particle material is challenging due to the changes in the carbon sampling and analysis methods which impact the measurements. Analysis of the data indicate some decreases in particle elemental carbon (EC) but little change in the organic carbon fraction (OC). Significant improvements in air quality are observed since 2000, when New York City metropolitan area exceeded the 1997 annual PM2.5 standard of 15 μg m, to the present conditions where it meets the current standard of 12 μg m. Further improvements in air quality may benefit from controls on carbonaceous material which is currently the major component of PM2.5.


INTRODUCTION
As a consequence of the various studies (for example Dockery et al., 1993;Pope et al., 1995) linking outdoor exposure to particulate matter (PM) with adverse health effects, the Environmental Protection Agency (EPA) promulgated National Ambient Air Quality Standards (NAAQS) for PM mass in 1997 (http://www3.epa.gov/ttn/naaqs/criteria.html).An annual average standard of 15 µg m -3 and a 24 h standard (based on the 98th percentile of observed 24 h values) of 65 µg m -3 were established for PM 2.5 (particles with an aerodynamic diameter at or below 2.5 microns).Both standards were applied to a 3 consecutive year period.Subsequent reviews led to a lowering of the 24h PM 2.5 NAAQS to 35 µg m -3 in 2006 and the annual primary to 12 µg m -3 in 2012.In the late 1990's a national PM 2.5 monitoring network was established to monitor spatial and temporal variability in PM 2.5 mass for compliance with the standard.Subsequently, the PM 2.5 Chemical Speciation Network (CSN) was designed to monitor spatial and temporal patterns in the particle chemical components useful for emission control implementation and provide data in support of health based research.
Generally, the major components of PM 2.5 particles in the US include SO 4 , NO 3 , ammonium (NH 4 ) and carbonaceous material (USEPA, 2012).The crustal or soil material proportion is small although it can be important in arid locations.In the northeastern US, particle SO 4 and associated NH 4 as well as carbonaceous material are the dominant components with particle NO 3 being important in urban areas particularly during colder periods (Tolocka et al., 2001;Schwab et al., 2004).
Over the past few decades large reductions in air pollutants have occurred throughout the US (USEPA, 2012) as a result of various emission controls (USEPA, 2014).These include the Clean Air Act (CAA) and Clean Air Act Amendments (CAAA), NO x State Implementation Plan (SIP) Call, Clean Air Interstate Rule (CAIR) and Tier 2 light duty vehicle standards and lower fuel S rule as well as improved vehicle technology.The recent downturn in the economy around 2008-2009 and consequent reduced energy consumption may also have resulted in lower emissions.A recent review (Lurmann et al., 2015) reported reductions in ambient air pollutants over a 20 yr period in Southern California.Blanchard et al. (2013) found major improvements in air quality since 1999 in the southeastern US.Reductions in PM 2.5 mass and major components have also been observed at locations in Europe (Cusack et al., 2012).Decreases in PM 2.5 mass have been noted in selected New York urban centers over the period 1999-2007 (Buckley and Mitchell, 2011).With a view to gaining insight into major source categories Schwab et al. (2004) analyzed PM chemical components measurements from various New York sites over a 2 year period.In this study, we evaluate trends in PM 2.5 mass and major components using an extensive dataset covering an approximate 14-15 year period.This dataset is also used to explore urban-rural differences as well as seasonal patterns.Such datasets are important in evaluating the effectiveness of existing emission controls and for future regulations.

Site Description
A list of the sites used for this analysis along with their locations are shown in Table 1 and Fig. S1.In the New York City (NYC) area PM 2.5 mass is measured across the 5 boroughs; Manhattan, Bronx, Brooklyn, Queens, and Staten Island.Data from a total of 11 sites in NYC, listed in Table 1, are included in the analysis of PM 2.5 mass.Outside of NYC measurements from 3 urban sites; Buffalo (BUFF), Rochester (ROCH) and Albany (ALB) and two rural sites; Pinnacle State Park (PSP) and Whiteface Mountain (WFM) are also used.PSP is a rural site located near the village of Addison (line of sight approx.310 km northwest of NYC) within 12 km of the border with Pennsylvania (Schwab et al., 2009).The site is at an elevation of 500 m and impacted by pollutants transported from the southwest across the region.WFM is located approximately 410 km north of NYC in the village of Wilmington, at an elevation of 600 m on the side of Whiteface Mountain.This configuration of sites provides a representation of measurements across urban as well as urban-rural contrasts.PM 2.5 components measured at seven of the sites MAN, IS52, QUEENS, ROCH, BUFF, PSP and WFM are also included in the analysis.The MAN site operated from two nearby (line of sight distance under 1 km) locations, Canal Street up to March 2007 and Division Street thereafter.Since species measurements at the New York Botanical Gardens (NYBG) ceased in December 2005 and measurements at ALB only commenced in August 2007 particle species data from these two sites were deemed not suitable for this long term analysis.

Sampling Instruments
PM 2.5 mass samples were collected based on the Federal Reference Method (FRM) from midnight to midnight generally on a 1 in 3 day schedule with the following exceptions.At QUEENS the 1 in 3 day schedule switched to daily sample collection in May, 2001.Sample collection at the IS52 and PSP sites was on a daily schedule up to 2010 but switched to a 1 in 3 day thereafter.WFM sampling was also setup on a daily schedule but switched to a 1 in 6 day collection in 2007.Both ALB and BUFF sites switched to daily sampling from 2007 to 2011 but reverted back to a 1 in 3 day schedule in 2012.
The FRM employed the Partisol Plus Sequential Air Sampler, Model 2025, Thermo Scientific, Franklin, MA for PM 2.5 mass sampling.A summary of sampler details is provided in Table 2.The sampler employs a PM 10 sizeselective inlet followed by a Well Impactor Ninety-Six (WINS) impactor with a 2.5 µm cut-point.Air flow is volumetrically controlled at 16.7 LPM and particles are collected on a 47 mm Teflon filter housed in a cassette.A supply chamber which stores filters prior to sampling is maintained to within 5°C of ambient using a continuous air ventilation system.Unused filters automatically move across under pneumatic pressure to the sample chamber for sampling   Peters et al. (2001).
while the already sampled filter moves into a storage chamber.This facilitates unattended sample collection over holiday and weekend periods.Blank filters were collected by installing a filter into the chamber with no flow through the filter over the 24 h period.Ambient meteorological and filter parameter data are recorded for each sample.Sampled filters are shipped to RTI International, Durham, North Carolina for analysis.Mean 24h PM 2.5 mass measurements for blank filters in 2012-2013 was 8.02 µg filter -1 with a 2 sigma of 1.06 µg corresponding to 0.33 ± 0.04 µg m -3 .PM 2.5 chemical speciation sample collection was based on a 1 in 3 day schedule except for BUFF which used a 1 in 6 day schedule.Collection at WFM switched to a 1 in 6 day schedule in 2006.Measurements were made using either a Rupprecht & Patashnick Co., Inc.Partisol Speciation Sampler model 2300, Albany, NY (hereafter R&P2300) or a Met One Instruments Spiral Aerosol Speciation Sampler, Grants Pass, OR (hereafter MetOne).A summary of instrument parameters is listed in Table 2 and detailed descriptions are provided elsewhere (Solomon et al., 2003).The R&P2300 contained three individual sample modules with greased impactor PM 2.5 size selective inlets (Demokritou et al., 2001).A Teflon filter module operated at 16.7 LPM was used for PM 2.5 mass and element analysis by gravimetric and X-ray Fluorescence, respectively.Nylon and quartz filter modules which operated at 10 LPM were used to sample for major ions and carbonaceous particle material, respectively.The MetOne sampler also had three individual modules but the flow through each channel was 6.7 LPM.A sharp cut cyclone (SCC 2.141, BGI Incorporated, Waltham, MA) provided the 2.5 µm cut-point.Blank filter measurements were obtained by placing a filter into the sampler for the 24h period with no flow through the filter.Filters were shipped to RTI for analysis.Mean blank measurements for major ions (analyzed by ion chromatography) typically varied from 0.02-0.06µg m -3 for SO 4 and NO 3 and were below 0.02 µg m -3 for NH 4 .Quartz filters were analyzed based on the NIOSH 5040 method (Peterson and Richards, 2002) using Thermal Optical Transmission (TOT).Typical mean blank measurement for OC of 1.0 µg m -3 was subtracted from the raw OC mass measurements.No correction was applied to the EC fraction as the blanks were below the detection limit.
A change to the CSN sampling and analysis methods of carbonaceous particles (Rice, 2006) led to the introduction of a URG-3000N Sequential Particulate Speciation System (URG, Chapel Hill, NC) in 2007.A cyclone (Air and Industrial Hygiene Laboratory, Berkeley, CA) operated at 22 LPM (volumetrically controlled) provided the PM 2.5 size cut.The sampler used a pair of 25 mm Pallflex Tissue quartz filters placed back to back and is based on that used in the IMPROVE (Interagency Monitoring of Protected Visual Environments) network.The use of the second filter was to provide a measurement of the active adsorptive material.In addition blank filters were also collected.Mean blank measurements for OC ranging from 0.10-0.20 µg m -3 were subtracted from the reported OC mass concentrations.Quartz filters were analyzed using the IMPROVE_A temperature protocol (Chow et al., 2007) at Desert Research Institute, Reno, NV.Dual mode detectors provided for analysis using both transmission (TOT) and reflectance (TOR).

PM 2.5 Mass
Annual mean PM 2.5 mass concentrations are shown for the NYC and various sites across the state since 2000 in Figs.1(a) and 1(b), respectively.A summary of the data is presented in Table 3 along with 95% confidence intervals.Concentrations reflect population density being highest in NYC and BUFF followed by ALB and ROCH with the rural PSP and WFM sites the lowest.Across the NYC metropolitan area the highest levels are in Manhattan, followed by Bronx then Brooklyn, Queens and Staten Island in that order.A consistent decline in the annual mean PM 2.5 mass is observed throughout the region since 2000.Mann-Kendall trend tests shows these are significant at the 95% confidence interval.For example in the NYC metropolitan area annual mean PM 2.5 mass ranged from 12.4-18.5µg m -3 in 2000 versus 8-11 µg m -3 in 2014 which is a drop of 4-7 µg m -3 equivalent to a 30-40% decrease.Thus PM 2.5 mass in NYC which exceeded the 1997 annual standard of 15 µg m -3 at several sites in 2000, has dropped below the current NAAQS of 12 µg m -3 in recent years.Similar decreases are observed for upstate urban sites of BUFF (6 µg m -3 ) and ALB (5 µg m -3 ) with ROCH slightly less (~4 µg m -3 ).The corresponding reduction at the rural PSP and WFM sites is 2-3 µg m -3 equivalent to a change of 25-30%.Likewise, the 98 th percentile PM 2.5 mass has decreased (significant at the 95% interval) although with considerable year to year variations, Fig. 2. In NYC and BUFF the 98 th percentile PM 2.5 mass has decreased by 15-20 µg m -3 compared to 9-14 µg m -3 for upstate urban centers, ALB and ROCH versus 5-10 µg m -3 at the rural locations.In 2000-2001 the 98 th percentile PM 2.5 mass was approximately 35-40 µg m -3 in NYC and BUFF, 30-35 µg m -3 for ROCH and ALB versus 20-30 µg m -3 for the two rural sites.The highest daily values across the state were observed on July 7, 2002 when the Eastern US was impacted by smoke from wildfires in Quebec, Canada (DeBell et al., 2004).On that day PM 2.5 mass for the NYC area ranged from 75-85 µg m -3 versus 50-75 µg m -3 in ALB and 40-50 µg m -3 in the western part of the state.This event increased the annual mean PM 2.5 mass by approximately 0.5 µg m -3 for 2002.Although the 98 th percentile annual PM 2.5 mass values were high across the state in 2002, they were not unusually high in NYC and in fact 2003 was slightly higher.Further details of the forest fire event are discussed in the supplemental section.

Seasonal Patterns
In order to compare seasonal patterns over the time period mean seasonal PM 2.5 mass concentrations were calculated for combined calendar years 2000-2001versus 2012-2013, shown in Fig. 3. Seasonal data for 2000-2001, Fig. 3(a), show highest NYC mean concentration in winter (Dec, Jan, Feb) of 17.6 µg m -3 followed closely by summer (Jun, Jul, Aug) 16.6 µg m -3 then autumn (Sep, Oct, Nov) 15.0 µg m -3 with the lowest in spring (Mar, Apr, May) of 13.3 µg m -3 , representing a seasonal amplitude of 4 µg m -3 .Corresponding seasonal variations in upstate urban centers, ROCH and ALB are similar although at ROCH both winter and summer have equivalent concentrations as well as spring and autumn.BUFF shows highest seasonal concentrations in summer and lowest in autumn, however the seasonal amplitude is the same.These seasonal differences, except for winter and spring at ALB, are not significant within the 95% confidence limits.At the rural sites of PSP and WFM seasonal concentrations for 2000-2001 show summer is significantly higher than the other seasons with a 5.5 µg m -3 and a 3.3 µg m -3 variation (between summer and winter) at the two sites, respectively.The similarity in summer PM 2.5 mass at PSP and NYC and upstate urban centers is also reflected in the seasonal SO 4 concentrations, one of the major components of PM 2.5 (discussed below).In addition another major component, OC, is also similarly elevated in summer.This can result in an extensive pollution event across the Northeastern US lasting over a day or more.Secondary species are regional in nature which leads to high spatial homogeneity (Turner and Allen, 2008).By 2012-2013 mean seasonal PM 2.5 mass has decreased by 3.4-7.5 µg m -3 across NYC, and upstate urban centers versus 2-5 µg m -3 and 1.3-2.3µg m -3 at the rural PSP and WFM sites, respectively.Seasonal data for 2012-2013 shows summer PM 2.5 mass is significantly higher than other seasons except in NYC and ALB.NYC seasonal PM 2.5 differences are not significant and at ALB only spring is significantly lower than winter and summer.

PM 2.5 Particle Species (i) Major Ions
Annual mean SO 4 , NO 3 and NH 4 for the PM 2.5 CSN sites are shown in Figs.4(a), 4(b), 4(c), respectively.A major difference between annual SO 4 and PM 2.5 mass is the similarity in PM 2.5 SO 4 concentrations across the state with the exception of WFM which is significantly lower by about 2 µg m -3 .In fact, SO 4 is highly correlated across all sites with coefficients of determination, R 2 , varying from 0.89 to 0.98.Concentrations at the rural PSP site are within 0.5 µg m -3 of those at the urban NYC and BUFF sites.ROCH appears on the lower edge of the group, just outside the 95% confidence interval of the QUEENS data.Similar observations were reported by Dutkiewicz et al. (2004) and Schwab et al. (2004).Highest SO 4 was associated with air mass transport through the Ohio River Valley and Great Lakes Basin, regions of high SO 2 emissions (Dutkiewicz et al., 2004).This reflects the regional nature of particle SO 4 with its major precursor emission source from Electric Generating Utilities (EGUs) lying towards the south and west of the state.Over the 10 yr period between 2002-2003 and 2012-2013 SO 2 emissions have been cut by about 65-70%, Fig. S2(a), compared to decreases in SO 4 of 50-64% observed across NY state.WFM being furthest from the emissions sources shows the least decrease of 50%.There is a linear relationship between SO 2 emissions and ambient SO 4 with coefficients of determination above 0.89 for all sites.However, there is not quite a 1:1 proportional response in SO 4 to the SO 2 emissions reductions as previously observed in the Southeastern US (Blanchard et al., 2013).A change in the rate of SO 4 decrease is evident around 2005-2007, with a general levelling off in concentrations in recent years.The same pattern is observed in EGU SO 2 emissions as it took time for CAAA controls to be adopted.Elemental selenium (Se) which is associated with coal combustion in the northeastern US (Dutkiewicz et al., 2006) also shows a similar decrease around 2006 (see below).Therefore, we attribute the decreases in SO 4 to the CAAA  Between 2002Between -2003Between and 2012Between -2013 total NO x emissions have been cut by 42% or 4.2 % yr -1 which is similar to reduction rates for particle NO 3 from 3.7 to 5.1 % yr -1 across NY State.Emission controls for SO 2 on EGUs also reduced NO x .In addition, the NO x SIP call in 2004 required further NO x reductions on EGUs and major combustion sources.

Seasonal Patterns
Enhanced photochemical activity (higher solar flux and extended daylight hours compared to other seasons) leads to higher SO 2 to SO 4 conversion during summer months throughout the region (Rattigan et al., 2006).In addition  air mass transport during summer is frequently from the southwest or west and through high SO 2 emitting areas (Dutkiewicz et al., 2004) Within NYC urban area we attribute a proportion of the winter SO 4 to residential oil combustion used for space and water heating as reflected in the higher winter elemental nickel (Ni) concentrations shown in Fig. S10.On the other hand particle NO 3 arises via nitrogen oxides with large emissions from roadway vehicles particularly high traffic density metropolitan areas.Particle NO 3 in the northeastern US shows elevated concentrations from Nov to May (Wittig et al., 2004;Rattigan et al., 2006) because the partitioning between NO 3 and its precursors, HNO 3 and NH 3 , favors particle NO 3 during colder temperatures (Seinfeld and Pandis, 1998).Fig. S4 shows NO 3 seasonal pattern with highest concentrations during winter and lowest in summer.Winter concentrations are significantly higher compared to summer (but not different compared to other seasons).Highest NO 3 is observed at the urban locations with mean winter values of 3.1-4.2µg m -3 , during 2002-2003, Fig S4(a), Fig S4(a).Although MAN shows highest concentrations among the urban sites the intraurban differences for a given season are not significant.
Winter NO 3 is a factor of 2 to 5 lower (significant) at the rural locations of PSP and WFM.These urban-rural NO 3 differences also contribute to the spatial differences observed in PM 2.5 mass in Fig. 3.By 2012-2013 NO 3 has decreased by approximately 37-50% across all sites.Concentration changes ranged from 0.5-1.7 µg m -3 at urban and 0.1-0.9µg m -3 at rural sites, Fig. S4(b).NO 3 is likely dominated by local production but even if some NO 3 is transported from other locations, the amount of particle NO 3 will still be determined by the partitioning which depends on ambient temperature and relative humidity.When emissions of NO x were lowered this led to a uniform drop in the amount of NO 3 throughout the year.Since it is ambient temperature which drives the seasonal NO 3 pattern , the same seasonal pattern is retained following the NO x reductions.Seasonal patterns for NH 4 in 2002-3003 show highest concentration during summer, Fig. S5(a), although not significant.Urban areas show lowest values are in autumn with a summer/ autumn amplitude from 30% to a factor of 2. At the rural location of PSP, autumn and winter have similar minimum values of around 1 µg m -3 whereas at WFM autumn, spring and winter are all around 0.60 µg m -3 with a slight summer max of 0.80 µg m -3 .The higher summer values reflects more the seasonal pattern of SO 4 rather than NO 3 since NH 4 preferentially associates with SO 4 first, then providing there is sufficient NH 3 available it also associates with NO 3 (Seinfeld and Pandis, 1998).By 2012-2013 the seasonal pattern has switched from a summer high to a winter max, Fig. S5(b), at all sites except for WFM where all seasons are similar.The winter/summer amplitude in 2012-2013 is approximately a factor of two at NYC sites and 40-60% at BUFF, ROCH and PSP.Thus, as a consequence of the substantial reductions in SO 2 and particle SO 4 the seasonal pattern of NH 4 reflects that of NO 3 in 2012-2013 rather than SO 4 .

(ii) Carbonaceous Material
Trends in PM 2.5 carbon, Fig. 5, are less evident and the interpretation of the data is hampered due to the change in sampling and analysis methods in 2007.This impacted the organic carbon fraction the most as the fraction is sensitive to both ambient conditions and measurement methods including sampling schedule, filter face velocity and shipment and transport conditions (Dillner et al., 2009;Rattigan et al., 2011).The impact on EC is less pronounced.In order to provide a consistent trend the more recent URG-3000N carbon OC and EC data were scaled to match those of the older CSN method using collocated measurements at the IS52 and PSP sites (Rattigan et al., 2011).Additional data became available since then including collocated data from QUEENS which were included in the analysis.The scaling factor for OC was approximately a 1.5 multiplier to the URG-3000N OC (by TOR analysis) to match the older CSN OC and a small downward scaling factor of 5-8% for the EC fraction.In addition collocated measurements from other carbon samplers were also analyzed for trends.
The annual mean EC concentrations in Fig. 5(a) show a large site to site variability with close to an order of magnitude range in concentrations between the highest urban areas of the Bronx and Manhattan to the lowest at the rural WFM location.Annual mean concentrations are similar for IS52 and MAN varying from 1.1 to 1.4 µg m -3 with uniform concentrations up to 2007-2008 and concentrations declining thereafter.This is more evident in the continuous MAN data record where 2014 is significantly lower by 30% compared to the 2007-2008 data.IMPROVE measurements at IS52, which sampled on the same day as the CSN, indicate factor to lower levels after 2007.EC at ROCH appears to increase slightly by 0.2 µg m -3 between 2001 and 2007 followed by a moderate decline up to 2010 (approximate 20% decrease).However, 2011 and 2012 show higher concentrations, 2012 being elevated by approximately 0.2 µg m -3 compared to 2010.An atypical coloration was noted in the 2011-2012 ROCH carbon filters which is believed to have caused an interference in the determination of EC and OC fractions (but did not impact total carbon) at this location.Apart from these 2 years the ROCH trend is similar to that at other urban sites.At PSP EC varied between 0.20 and 0.26 µg m -3 with little indication of a trend at least up to 2006.EC in 2007 appears somewhat elevated attributed to 2 days when concentrations (2.1 and 1.5 µg m -3 , respectively) were more than a factor of 5 above the annual mean.Subsequently, EC appears to decline with concentrations 30% lower in 2012-2013.A change in the sampler occurred in 2009 (URG-3000N replaced the MetOne).In addition the analysis methods also changed to the IMPROVE_A method (Chow et al., 2007) The high 2007 data at WFM was attributed to 5 days when EC was unusually high, 0.65-0.9µg m -3 , compared to the annual mean of around 0.2 µg m -3 .Furthermore, there were 2 additional days with concentrations of approximately 0.5 µg m -3 .After 2007 EC at WFM also decreases similar to other locations, but like PSP a change in the sampling and analysis methods occurred in 2009.In addition, since EC at WFM (and PSP) is at or near the detection limit the interpretation of EC trends at these rural locations should be viewed with caution.OC annual mean data in Fig. 5(b) also show an urbanrural contrast with concentrations at MAN a factor of 3-4 higher than the remote WFM location.OC concentrations in descending order are highest at MAN with IS52 approximately 1 µg m -3 lower, then QUEENS and BUFF 0.5 µg m -3 lower with ROCH approximately 0.5 µg m -3 lower still.OC at PSP is lower than ROCH and WFM is the lowest of all the sites.There is no consistent trend in the annual OC long term record.OC at MAN shows an initial decrease up to 2006 (18% drop) but 2007 and 2008 appear somewhat elevated and subsequent OC varying between 3.9 to 4.  S7(a).At urban sites the seasonal EC differences are approximately 10-15% and not significant.At the rural sites EC seasonal differences are approximately 30% but only spring is significantly higher compared to autumn.Small seasonal differences are also observed for 2012-2013, Fig. S7(b), Fig. S7(b), which are mainly not significant in NYC, except for QUEENS spring EC being lower than winter.At BUFF and ROCH, summer EC tends to be higher than winter but it is barely significant.EC in 2012-2013 at the rural sites is higher in summer and autumn but not significant.At the large urban centers EC is dominated by local primary emissions which appear reasonably consistent throughout the year or show a slight winter maximum.At rural locations there are few if any major sources of EC and the sites are impacted more by transported pollutants (and occasional vegetative burning episodes) thus leading to different seasonal patterns and significantly lower EC.The seasonal pattern for OC for both time periods shows a consistent summer maximum at all locations (as observed previously, Schwab et al., 2004) most noticeable at ROCH, BUFF and the rural locations of PSP and WFM where it is significantly higher compared to other seasons, Fig. S8.Among the NYC sites QUEENS summer OC is significantly higher than spring and autumn but not winter and at MAN only 2012-2013 summer is significantly higher than other seasons.The summer OC maximum is attributed to enhanced secondary organic aerosol production and is most noticeable at less polluted sites which are impacted less by local primary emissions.Enhanced emissions in winter from space heating sources lead to increased production of primary OC (and EC) thus leading to a lower seasonal contrast in OC at the larger urban NYC locations.

(iii) Selected Elements
Se shows some similarities with SO 4 in that concentrations are relatively similar at all sites across the state ranging from 1-1.6 ng m -3 except WFM which is a factor of 2-3 lower, Fig. 6.A downward trend is also observed but the pattern is somewhat different to SO 4 .The largest change occurs around 2006, where a factor of 3 decrease is observed across the state (WFM showing a somewhat smaller decrease, factor of 2) between 2006 and 2008.Since Se is associated with coal combustion sources outside the region (Dutkiewicz et al., 2006) this change is attributed to the emission reductions on EGU's and large combustion sources through the CAAA and NO x SIP call.After 2008 concentrations are relatively flat and are uniform across all sites with a range of 0.2-0.5 ng m -3 , WFM being on the lower edge.The fact that Se concentrations at PSP are similar to urban areas also indicates that the major source emissions are from outside the state.Since Se is produced from direct emissions it is not influenced by photochemistry and therefore has a different seasonal pattern to secondary species like  SO 4 .There is however an indication of a winter source, likely from space heating sources such as oil, (Graham, 2010) in the 2002-2003 data as shown in the slightly higher (although not significant) winter concentrations at all sites except MAN in Fig. S9.By 2012-2013 concentrations are frequently near the detection limit and seasonal differences may not be meaningful.
Ni is characterized by showing elevated concentrations only in the NYC area with values below or at detection limits elsewhere, Fig. 7.In 2001-2003 annual mean Ni ranged from around 15-20 ng m -3 in NYC but a steady decline is observed since with concentrations reaching 3 ng m -3 by 2013 (factor of 5-6 decrease).A strong seasonal gradient is observed, Fig. S10, with higher concentrations (more than a factor of 2 between winter and summer) during colder months consistent with the major sources being residential heating oil combustion used for space and water heating (Lippmann, 2009).In addition the ratio of Ni to Co of around 12-14  (Rattigan et al., 2013) is typical of fuel oils used in the Northeastern states (Graham, 2010).The very low Ni concentrations observed at sites outside NYC is due to the predominant use of natural gas and wood rather than oil for domestic heating in these areas.A switch to alternative type heating fuel oils (NYC required the phase out of number 6 heating oil by July 1, 2015) or natural gas which have lower Ni emissions is largely responsible for the extensive Ni reductions observed in NYC.

Other Locations
In addition to the reductions in PM 2.5 mass and components reported here reductions in gas phase species, SO 2 , NO 2 and CO have also occurred (Civerolo et al., Manuscript in Preparation).Reductions in air pollutants reported here are generally consistent with that observed throughout the US (USEPA, 2012).For example the US annual mean PM 2.5 mass has shown a 24% decrease over the period 2001-2010 whereas the 98 th percentile 24h PM 2.5 mass has decreased by 28%.More recent data show a 34% decrease in the annual and daily mean PM 2.5 mass from 2000 to 2013.Corresponding decreases in the annual PM 2.5 mass reported above were 30-40% at urban New York sites and 25-30% at rural locations whereas, the 98 th percentile PM 2.5 mass decreased by 40-50%.The national emission inventories trends (USEPA, 2014) reported decreases of 68% and 42% over the period 2000-2013, for SO 2 and NO x , respectively.An analysis of PM 2.5 mass data from various New York sites over the period 1999 to 2007 by Buckley and Mitchell (2011) also showed downward trends.Long term measurements of particle SO 4 across the US (Hand et al., 2012) also show decreasing trends.Average PM 2.5 mass decreased by 13 to 54% at various sites in southern California over the period 1994-2011 (Lurmann et al., 2015).However, further improvements in air quality are required as the Los Angeles area still lies outside the NAAQS for PM 2.5 .Decreasing trends in PM 2.5 and various components have been observed across European locations (Cusack et al., 2012).These trends were attributed to pollution abatement strategies, downturn in the economy (which had a greater impact in Europe compared to the US) and unusual meteorological conditions.Ambient measurements in the southeastern US (Blanchard et al., 2013) show reductions in PM 2.5 of around 4.5 µg m -3 and SO 4 of 2.3 µg m -3 from 1999 to 2010 which is similar to that reported here.Controls on EGU and vehicle emissions is largely responsible for these decreases.However, Blanchard et al. (2013) also report large reductions in carbonaceous species, OM (2.2 µg m -3 ) which is not observed in New York.An analysis of trends in PM 2.5 OC and EC across the US (Hand et al., 2013) found the largest downward trends in the western US with generally small or insignificant changes at eastern US locations.Thus, it seems that trends in PM 2.5 carbonaceous particles are more variable and less homogeneous compared to the inorganic species which show a consistent downward trend throughout the US.The impact of the reductions on the percentage contribution to PM 2.5 mass can be seen in Fig. 8 which shows annual mean data from QUEENS.Here a multiplier of 1.6 for converting OC to organic mass was used (Bae et al.,Fig. 8. Species percentage composition of PM 2.5 in 2002-2003versus 2012-2013at QUEENS. 2006)).Metals was calculated as the sum of soil components (coarse metal oxides) and remaining trace elements.Percentages are relative to total PM 2.5 mass on the CSN speciation sample filter.Sum of species is within ± 5-10% of total mass.Uncertainties in total PM 2.5 mass and species measurements and the OC multiplier can account for the differences.In 2002-2003 the combined contribution from particle SO 4 , NO 3 and NH 4 was approximately 60%, whereas the organic mass contribution was 30%.By 2012-2013 a substantial change in the proportions are observed, with the major ion species amounting to 40% compared to approximately 45-50% for organic mass.These percentages vary somewhat from year to year, but the impact is essentially the same.Thus, in recent years the carbonaceous (OM and EC) fraction is the dominant component amounting to approximately 50-60% of the total PM 2.5 mass.Therefore, targeting carbonaceous species emissions may be required in order to achieve additional air quality improvements, particularly since trends in inorganic species mass concentrations have levelled off in recent years.

CONCLUSIONS
Analysis of an approximate 15 yr period of PM 2.5 mass and major chemical component measurement data in New York has shown a downward trend for PM 2.5 mass and most major species (except organic carbon) since 2000.For example the reductions amount to a 30-40% decrease in annual mean PM 2.5 mass since 2000.Large reductions are also observed for the major ion particle species; SO 4 , NO 3 and NH 4 and selected metals, Se and Ni.These trends are mainly attributed to emissions reductions under the CAAA, NO x SIP call and CAIR which targeted Electric Generating Utilities and large combustion facilities.The introduction of improved vehicle technology and cleaner fuels (transport and residential heating) also lowered NO 3 and Ni levels.A decrease in EC after 2007 of approximately 30% is also attributed to lower vehicle emissions.These changes have led to large improvements in ambient air quality particularly, in the New York City region, which in 2000 was out of compliance of the NAAQS annual PM 2.5 15 µg m -3 standard, to the present day conditions where it is within the newer 12 µg m -3 standard.However, the downward trend is less evident for carbonaceous particle species, particularly the OC fraction which is currently the dominant component of PM 2.5 mass.Therefore, future improvements in air quality may benefit from targeting this species.

Fig. 1 .
Fig. 1.Annual mean PM 2.5 mass for (a) NYC boroughs and (b) across NY state.In panel (a) MAN, BRONX and SI refer to the means across the Manhattan, Bronx and Staten Island sites, respectively and BKYLN refers to Brooklyn.The dashed lines in (a) show the range across individual NYC sites.The error bars represent 95% confidence limits on (a) BRONX data and (b) the NYC and WFM data.

Fig. 2 .
Fig. 2. Annual 98 percentile PM 2.5 mass for (a) NYC boroughs and (b) across NY State.In panel (a) MAN, BRONX and SI refer to the means across the Manhattan, Bronx and Staten Island sites, respectively and BKYLN refers to Brooklyn.The dashed lines in (a) and error bars in (b) represent the range across individual NYC sites.

Fig. 4 .
Fig. 4. Annual mean concentrations of (a) SO 4 , (b) NO 3 and (c) NH 4 at sites in New York.Error bars represent the 95% confidence intervals on QUEENS and WFM data.

Fig. 5 .
Fig. 5. Annual mean concentrations of (a) EC and (b) OC at sites in New York.Gaps in the data are due to instrumental problems (WFM) or no site sampling (IS52).OC was adjusted for the adsorption artifact based on respective site field blank data.Error bars show 95% confidence intervals on QUEENS and WFM data.
. The trend in the IMPROVE EC data at PSP is somewhat different compared to the CSN with IMPROVE showing uniform concentrations of 0.33-0.40µg m -3 between 2001 and 2007, Fig. S6(b).Subsequently, IMPROVE EC shows a drop of approximately 30% between 2007 and 2010 consistent with EC patterns observed at the above urban sites.At WFM although there is an indication of an initial downward trend up to 2006, EC in 2007 like at PSP appears elevated.
3 µg m -3 from 2009 to 2014.The change in sampling methods which occurred part way through 2007 and missing samples during June to August in 2009 and March to June 2010 may have impacted the annual mean values for those years.Nevertheless, QUEENS OC and a shorter record at the IS52 site are also similarly uniform and there is an indication that 2007 OC is also elevated at IS52.BUFF OC is very similar to QUEENS except that 2004 BUFF OC is unusually low (30% lower than 2003 and 2005) and concentrations drop by 18% after 2010.Annual OC at ROCH, PSP and WFM shows a slight upward trend at least until 2008 at ROCH and WFM with the trend at PSP being upward throughout.Seasonal patterns for EC and OC are shown in Figs.S7 and S8, respectively.Mean seasonal EC for combined years 2002-2003 are shown in Fig.

Fig. 6 .
Fig. 6.Annual mean PM 2.5 Se concentrations at sites in New York.Error bars show 95% confidence intervals on QUEENS and WFM data.

Fig. 7 .
Fig. 7. Annual mean PM 2.5 Ni concentrations at sites in New York.Error bars show the 95% confidence intervals on QUEENS and WFM data.

Table 1 .
Summary details of ambient monitoring sites.
a Canal Street sampling end date and Division street start date.b NYBG speciation sampling end date.c Speciation data for ALB and NYBG not included for long term analysis.

Table 2 .
Summary details of Sampling Instruments.

Table 3 .
Summary of annual PM 2.5 mass mean concentrations in µg m -3 at various sites.The values in parenthesis for NYC shows the range across the metropolitan sites.The values b for other sites are the 95% confidence intervals. a