Final Report

Local and Regional Contributions of Fine Particulate Mass to Urban Areas in the Mid-Atlantic and Southwestern US

by

Bret A. Schichtel

Center for Air Pollution Impact and Trend Analysis (CAPITA)

Washington University

St. Louis, MO, 63130

Contract # 7D-0869-NAEX

Project Officer: Alan Huber

US EPA

Research Triangle Park, NC 27711

 

March 27, 1999

 

Abstract

This work examined the seasonal local and regional source contributions of PM2.5 to urban areas in the Mid-Atlantic States: Baltimore, MD, Washington, DC, and Philadelphia, PA and Phoenix, AZ in the Southwest. This was accomplished using two different methods. The first method estimated urban excesses by comparing seasonal PM2.5 trends at the urban monitors to nearby rural monitors. Washington and Philadelphia had urban excesses of ~40% in winter which decreased to ~20% during the summer. Baltimore’s excess varied considerably (0-60%) from month to month. At Phoenix, the urban excess was between 50-90% in winter and 30-70% during the summer. The second approach used a simple model based on the PM2.5 dependence on wind speed and wind direction to classify a site as being dominated by local or regional source contributions. The method also quantifies the regional contributions during high wind speed conditions. The wind vectors were derived from surface observations and air mass histories. All monitoring sites in the urban centers were dominated by local sources during the cold season (November – March), but the Mid-Atlantic urban centers were dominated by regional sources during the warm season. Regional contributions at the Mid-Atlantic urban centers during high wind speeds were between 7-12 and 7-15 mg/m3 during the cold and warm seasons respectively. The lower concentrations typically were associated with winds from the north-northeast, while the higher concentrations were associated with winds from the south-southwest. At Phoenix, PM2.5 concentrations during high wind speeds were between 2-10 mg/m3 during the cold season and 5 mg/m3 during the warm season.

 

1.0 Introduction

With the promulgation of the new fine particulate, PM2.5, national air quality standard, all non-attainment regions will need to modify emission of PM2.5 and their precursors to reach attainment. Prudent air quality management requires knowing which sources contribute to the problem and how much. Determining PM2.5 source contributions is complicated due to the fact that often more than 50% of the PM2.5 is composed of secondary species (EPA 1996, and Schichtel and Husar, 1992) masking its point of origin. In addition, PM2.5 has a lifetime on the order of several days (Husar et al., 1978), enabling sources a 1000 miles away to impact a receptor.

This work examines a subset of the source apportionment problem by providing evidence for local and regional source contributions and first order approximations of their respective seasonal contributions to urban areas in the Mid-Atlantic states (Baltimore, MD, Washington, DC and Philadelphia, PA) and Phoenix, AZ in the Southwest. This is accomplished using two different approaches. The first approach derives urban excesses by comparing seasonal PM2.5 trends at the urban monitors to nearby rural monitors. The second approach uses a simple model based on the PM2.5 dependence on wind speed and direction, where the wind vector is derived from surface meteorological observations and air mass histories. Using wind vectors to derive estimates of local and regional contributions is based upon work initially done by Gifford, and Hanna (1973) and more recently by Husar and Renard (1998) in their analysis of ozone in the Eastern US.

2.0 Fine Particulate Data

The fine particulate data used in this study came from the National PM Research Monitoring Network and other sources compiled in the North American fine particle data set (Schichtel et. al. 1999). The National PM Research Monitoring Network was established with the primary objective of providing ambient air quality data for relating health effects to chemical and/or physical properties of PM and to support emerging regulatory implementation and development issues. This network began collecting fine and coarse speciated PM data and meteorological data in Phoenix AZ in February of 1995. Monitoring platforms at Baltimore, MD and Fresno, CA were added in 1997. The monitoring platforms have a dichotomous sampler collecting 24 hour integrated samples every three days and a dual fine particle sequential sampler (DFPSS) collecting 24 hour integrated samples every day. The PM2.5 from these two samplers are compared in Appendix A.

The North American fine particle data set is an integration of 18 historical and active fine particle monitoring networks, and contains some data for ~600 urban and rural monitoring sites in the US and Canada from 1979 through 1997. For this study, data from the IMPROVE, AIRS, and California Airs Resource Board were extracted from this data set and combined with the data from National PM research Monitoring Network. These sites were selected for their proximity to Baltimore or Phoenix (Figure 1) and had at least two years of data in the 1990’s.

A B

Figure 1. Urban and rural PM2.5 monitoring sites from the North American integrated fine mass data set used in the analysis, and the National Weather Services surface observation stations.

3.0 Urban & Rural Seasonal Trends – Urban PM2.5 Excess

As a first approximation of local and regional contributions to the urban centers, the differences between the concentrations of the monthly average urban and nearby rural monitoring data were examined. This approach assumes that the PM2.5 at the rural sites are not contaminated by the urban emissions and that the same regional sources have the same impact on the rural monitors as the urban monitors. Using monthly averaged data, it is unlikely that both of these assumptions can be met. For example, the closer the rural site is to the urban site the more likely the same regional sources will influence the rural and urban sites, but the more likely the urban emission will contaminate the rural PM2.5 concentrations.

3.1 Mid-Atlantic States

In the Mid-Atlantic states the PM2.5 database contained suitable data from three urban sites, Baltimore, MD from the National PM Research Monitoring Network, Washington, DC from the IMPROVE network and Philadelphia from AIRS (Figure 1A). The seasonal trends for these three sites are presented in Figure 2. As discussed in Appendix A, the two collocated Baltimore samplers produced different seasonal trends. Therefore, the Baltimore seasonal trend was generated using the dichotomous data, and the daily DFPSS data were used to fill in missing dichotomous samples. The dichotomous data were used because both the IMPROVE and AIRS samples were collected using dichotomous samplers.

As shown in Figure 2A, Washington DC and Philadelphia have very similar PM2.5 seasonal trends with about a 60% increase in concentration from spring (15 mg/m3) to summer (25 mg/m3). The Baltimore site has lower winter and spring (November – June) PM2.5 concentrations (9-12 mg/m3), but similar summer and fall concentrations. The Baltimore data is only for 1997 while Washington and Philadelphia had up to 10 years of data. However, similar differences are seen when the 1997 data for Baltimore and Washington are compared (Figure 9B).

The IMPROVE network has seven rural monitoring sites in the vicinity of the three Mid Atlantic Urban centers (Figure 1A) ranging in elevation from 5 to 1700 m. The PM2.5 concentrations are potentially dependent on elevation (see Appendix B) so only the three lowest sites, Brigantine, NY, Jefferson NF, VA and Ringwood, NJ were used in the analysis, since the urban sites were near sea level. The seasonal trends from these sites are presented Figure 3. As shown, these sites are summer peaked with almost a factor of two increase from winter to summer. The November – April concentrations are nearly constant at ~10 mg/m3 for Ringwood and Brigantine and 12 mg/m3 at Jefferson and increase to ~ 17 and 22 mg/m3 respectively in the summer.

Figure 4 compares the urban PM2.5 concentrations to the rural sites. At Washington, DC, the excess PM2.5 concentrations over Brigantine are 5-8 mg/m3 throughout the year. The Washington urban excess over Jefferson NF is more seasonally dependent with an excess greater than 4 mg/m3 for the winter but less than 1 mg/m3 during August and September. Therefore, a first order approximation of the winter regional contribution to Washington is 10-12 mg/m3 and the local contribution is 4-8 mg/m3 or 40% of the total PM2.5. During the summer months, the regional contribution increased to 14 - 20 mg/m3 and the local contributions decreased to 1 – 8 mg/m3 or about 20% of the total mass (Table 1). The local and regional contributions for Philadelphia are similar to Washington since their seasonal trends are also very similar (Figure 2A). At Baltimore, the PM2.5 concentrations are generally less than at Jefferson NF except during June, July and September where they exceed Jefferson NF concentrations by 2-3 mg/m3 (Figure 4B). The Baltimore PM2.5 exceeds Brigantine’s PM2.5 concentrations from April to November by 2 – 10 mg/m3, but is less than or equal to Brigantine’s in February and March. Therefore, there are no apparent local contributions to the Baltimore site’s PM2.5 during February and March, and between 0-10 mg/m3 or 0- 60% from April to November.

A B

Figure 2. Average monthly PM2.5 for three Mid-Atlantic urban sites using A) all available data and B) 1997 data.

Figure 3. Seasonal trends from three rural low elevation IMPROVE sites.

A B

Figure 4. A) Washington DC seasonal PM2.5 trend compared to Brigantine, NJ and Jefferson NF, VA two rural low elevation sites. B) Baltimore seasonal PM2.5 trend compared to Brigantine, NJ and Jefferson NF, VA.

3.2 Phoenix Arizona

The seasonal trends for Phoenix, AZ and three other urban sites in Southern California are presented in Figure 5A. All four sites show pronounced winter peaks with ~50% more PM2.5 from November to January than during April and May. The two sites Calexico and El Centro, CA also have small summer peaks. At Phoenix, the monthly average PM2.5 concentrations from February to August are nearly constant decreasing from 10 to 8 mg/m3, but the PM2.5 concentration increases to 20 mg/m3 by December.

In contrast to the urban trends, the rural PM2.5 varies from 2-5 mg/m3 in the winter to 4-7 mg/m3 in the summer (Figure 5B). Figure 6 presents the comparison of the Phoenix seasonal trend to three rural sites located south (Tonto), east (Petrified Forest) and north (Hopi Point) of Phoenix. The urban excess is largest during the fall and winter at 9-18 mg/m3 from October to January. This decreases to 3 – 6 mg/m3 during the spring and summer months April – August. Therefore, the winter local contribution at Phoenix is between 50 – 90% with the largest local contributions occurring during December. During the summer months, the local PM2.5 contribution is between 30 – 70% (Table 2).

A B

Figure 5 A) The average monthly concentrations for Phoenix, AZ and three other urban sites in Southern California from the California Air Resource Board’s network. B) Five rural sites south, east, and west of Phoenix, AZ from the IMPROVE monitoring network.

Figure 6. Phoenix, AZ seasonal PM2.5 trend compared to Tonto, NM, Petrified Forest, AZ and Hopi Point, AZ.

Table 1. The local and regional contributions to the Mid-Atlantic urban centers estimated from the urban excesses.

 

Winter

Summer

 

Conc. mg/m3

% Total PM2.5

Conc. mg/m3

% Total PM2.5

Local

4 – 8

40%

1 – 8

20%

Regional

10 – 12

60%

14 – 20

80%

Table 2. The local and regional contributions to Phpenix, AZ as estimated from the urban excess.

 

Winter

Summer

 

Conc. mg/m3

% Total PM2.5

Conc. mg/m3

% Total PM2.5

Local

9 – 18

50 – 90%

3 – 6

30 – 70%

Regional

2 – 5

10 – 50%

4 – 7

70 – 30%

 

4.0 PM2.5 as a Function of Wind Speed and Direction

This section examines the local and regional contributions of fine mass to a receptor's concentration by establishing the PM2.5 dependence on wind speed and direction. The analysis is based on the premise that if the PM2.5 concentration declines steadily with increasing wind speed, i.e. ventilation, then the PM2.5 is dominated by local sources. On the other hand, if the PM2.5 concentration does not decline with wind speed, then the PM2.5 is attributable to distant regional sources. Husar and Renard (1998) successfully used this approach to examine the regional and local ozone contributions in the Eastern US.

4.1 Wind Speed and Direction Data

The analysis was conducted using surface winds and airmass history winds. The surface wind data were obtained from the National Weather Service synoptic monitoring network, consisting of about 300 sites in the conterminous US (Figure 1). Only the noon wind speed and directions were used since the surface winds at night are usually de-coupled from the upper air winds, and are not representative of the pollutant transport layer. The surface winds are able to identify the noon local dispersion characteristics at the receptor site. However, they do not account for the transport history of the airmass prior to impacting the receptor.

The airmass history winds are a regional dispersion index that takes into account three days of airmass history dispersion. They are generated from back trajectories calculated using synoptic scale wind fields. The methodology for creating the airmass history winds is presented in Appendix C. This index should be better able to account for regional impacts than the surface winds since it includes airmass dispersion over potential regional sources. It should also be better able to account for some important dispersion patterns conducive to pollutant build up and ventilation. For example, fast recirculating winds will accumulate pollutants while slower but directionally persistent winds will ventilate an area. In the first case, the airmass history winds would be slow while the surface winds would be fast, and in the second case, the airmass history wind speed would be moderate while the surface winds would be slow. Problems with the airmass history winds are that they are dependent on modeled meteorological data that are subject to errors and biases. In addition, these winds will be dependent on the maximum particle age used in the airmass history. No sensitivity testing has been conducted at this point to test the affect of the airmass history’s maximum ages on the airmass history winds and PM2.5 classification using these winds.

4.2 Method and Interpretation

The PM2.5 wind speed and direction dependence was created for each monitoring site by sorting the PM2.5 concentrations for specific wind direction and speed ranges then averaging the sorted values together. The average PM2.5 concentration was computed for each wind direction range in 90° increments and without regard to the wind direction (all wind directions). The first directional wind was between 0-90°, i.e. when the wind blew from north or northeast. This resulted in five wind directional concentration bins. The average concentrations for each directional bin was further classified by wind speed, ranging between 0-2, 2-4, 4-6, 6-8, 8-10, 10-12 m/s increments. This was done for a cold season (October – March) and a warm season (April – September). The results of the analysis are presented as charts of average concentrations as a function of wind speed and wind direction at selected locations in the Mid-Atlantic States and the Southwest.

The following discussion was largely taken from the Husar and Renard (1998) paper. In order to interpret the results it is instructive to examine the wind speed dependence of a pollutant using a simple one dimensional transport model (Figure 7). The pollutant emissions are confined to a mixing height of H[m, 1000m]. Within the mixing layer the unidirectional wind speed is U[m/s] and carries a background concentration C0[g/m3] into a source area. The source area itself has an emission density of Q[g/m2,s] as well as the source length in the direction of the wind vector, L[m]. Assuming that the local emissions are mixed instantaneously, the concentration, C[g/m3], averaged over the source region can be estimated by the classic box model expression:

C = C0 + QL/UH

The total concentration is the sum of the background and local contributions. The second term on the right side represents the local contribution. It is proportional to the source strength (QL) and is inversely proportional to the ventilation coefficient (UH). As shown in Figure 8, the local contribution is inversely dependent on the wind speed. The concentration, C, is highest at low wind speeds because the pollutants accumulate due to poor ventilation. With increasing wind speed, concentrations asymptotically approaches the regional background concentrations due to the rapid dilution of the local contributions.

Inherently, the model is only applicable near the sources where the removal processes are not significant. In addition, the background concentration entering a source area, C0, represents the sum of all PM2.5 contributions entering the domain from natural and anthropogenic sources. The role of a variable regional pollutant concentration is not incorporated.

Figure 7. Schematic illustration of a simple one-dimensional model.

Figure 8. Concentration as a function of wind speed at different local source strengths.

In the analysis below, the box model is used to qualitatively interpret the measured PM2.5 concentrations as being of local or regional in origin. Strongly declining concentrations with wind speed will be interpreted as evidence of local source contributions, since higher wind speeds cause increasing dilution of local contributions. If the concentration is invariant with wind speed, then it is taken as evidence that the local contribution is not significant, hence regional transport dominates. The general directions of the regional sources are identified through the direction analysis. This analysis also provides a quantitative estimate of the regional PM2.5 concentrations during high wind speeds, because the contribution of emissions from local sources cannot accumulate and are low. In addition, a significant fraction of the fine aerosol is made up of secondary material, over 50% in the case of Washington DC (EPA 1996). High wind speeds will remove precursor species before the formation of aerosol further suppressing the local contributions.

4.3 Results

4.3.1 Mid-Atlantic States

PM2.5 as a Function of Surface Wind Speed and Direction. The PM2.5 concentrations as a function of surface wind speed for the three urban sites (Baltimore, MD, Washington, DC, and Philadelphia, PA) and four rural sites, (Shenandoah, VA, Dolly Sods, WV, Brigantine, NJ, Jefferson, VA) in the Mid-Atlantic states are presented in Figure 9. During the cold season (November – March), the PM2.5 at all three urban sites decline sharply with increasing wind speeds, 15-25 mg/m3 to 7 mg/m3 or ~60% over the wind speed range. With the exception of Jefferson NF, VA the rural sites show smaller declines in PM2.5 with increasing wind speeds particularly at the lower wind speed. For example, the PM2.5 concentrations decrease ~40% over the wind speed range at Brigantine and Shenandoah, while at Dolly Sods the PM2.5 is practically independent of wind speed. These results are indications that the urban areas are dominated by local emissions while the rural areas are more influences by regional sources.

The inverse relationship in the urban PM2.5 concentration and wind speed decreases substantially during the warm season (Figure 9). The PM2.5 at all three urban sites is practically independent of wind speed below 5 m/s at 15-20 mg/m3. At the highest wind speeds, the Philadelphia and Baltimore concentrations decrease to 7 – 12 mg/m3, or 40-60%. The rural sites decrease only 20 – 40% over the entire wind speed range. At Jefferson NF, VA the PM2.5 concentrations actually increases from 15 to 20 mg/m3. These PM2.5 to wind speed relationships indicate that the regional sources contribute a larger fraction of the total PM2.5 during the warm season than the cold season.

At the highest wind speeds, both urban and rural sites have cold season PM2.5 concentrations ~5-7 mg/m3 (Figure 9). The fact that both urban and rural sites have the same concentrations supports the supposition that at these higher wind speeds, PM2.5 concentrations are dominated by the regional sources. During the warm season, the PM2.5 concentrations at high wind speeds varying between 7 – 12 mg/m3 for Philadelphia, Baltimore, Shenandoah, Dolly Sods and Brigantine and about 20 mg/m3 for Washington, DC and Jefferson NF.

The PM2.5 dependence on wind speed and direction for selected sites is presented in Figure 10. Appendix D presents these figures for all sites examined in this study. During the cold season, there is a small PM2.5 concentration dependence on the wind direction with the lowest concentrations associated with winds from the North 270 – 90o and the highest concentrations associated with winds from the south, 90 to 270o. During the warm season, the PM2.5 concentrations are generally larger when the wind blows from the southwest and smallest when the wind blows from the northeast. This is most pronounced at Philadelphia where at wind speeds of 7 m/s the PM2.5 concentration is 21 mg/m3 with winds from the southwest and 8 mg/m3 with winds from the northeast (Figure 10).

PM2.5 as a Function of Airmass History Wind Speed and Direction. The PM2.5 concentration as a function of airmass history wind speed and direction for the urban and rural sites are presented in Figures 11-12. An interesting feature of these figures is that during both the cold and warm seasons nearly all sites show decreases in PM2.5 concentrations with increasing wind speeds for all directions. The only exception occurs at Brigantine during the warm season (Figure 11). This is an indication that these winds are inadequate for classifying a site’s PM2.5 concentrations as dominated by local or regional sources. However, the directional dependence of PM2.5 concentrations on wind direction is much more apparent than was seen using the surface winds. The highest concentrations are associated with winds from the south-southeast and the lowest concentrations are associated with winds from the north-northeast. During the cold season, the southeast flow had ~5 mg/m3 higher than flow from the northeast (Figure 12) for all sites. This difference increased to about 10 mg/m3 during the warm season.

High airmass history wind speeds are the result of high persistent wind speeds throughout the three day pathway of the airmass prior to reaching the receptor, thus they identify periods of regional scale ventilation. Consequently, these winds should be a good indicator of the regional contributions during high wind speeds. As shown during the cold season (Figure 11 and 12) the high wind speed PM2.5 concentrations are between 7 – 12 mg/m3 at Washington and Philadelphia and the rural sites Brigantine and Jefferson. The lower concentrations are associated with flow from the northeast and the high PM2.5 concentrations are associated with flow from the southwest (Figure 12). At Shenandoah and Dolly Sods, PM2.5 concentrations are 2-5 mg/m3 during high wind speeds. During the warm season, PM2.5 concentrations under high wind speed conditions are 7 – 15 mg/m3 (Figure 11) with the lowest concentration typically associated with flow from the northeast and the highest concentration with flow from the southeast (Figure 12).

4.3.2 Southwest

PM2.5 as a Function of Surface Wind Speed and Direction. The PM2.5 concentrations as a function of surface wind speed and direction for Phoenix and the surrounding rural sites, Tonto, Saquaro, Hopi Point, and Petrified Forest are presented in Figures 13 and 14. Data for Long Beach is also include to check whether the method can properly classify urban and rural locations. Figure 1 shows the closest National Weather Station, NWS, to the Southwest monitoring sites. With the exception of Saquaro, the rural PM2.5 monitoring sites were more than 50 miles from the closest NWS surface observation stations. At these large distances and the fact that the southwest is complex terrain, it is possible that the surface winds at the NWS stations had little correspondence with the true winds at the monitoring sites. Therefore, the PM2.5 and winds at these stations should be view with caution.

As shown in Figures 13, the PM2.5 concentrations decrease rapidly with increasing wind speeds at both Phoenix and Long Beach. During the cold season the Phoenix PM2.5 decreases 60%, from 18 to 7 mg/m3 with an increase in wind speed from 1 to 5 m/s. The rate of decrease during the warm season is only 20%, 10 to 8 mg/m3, with an increase in wind speed from 1 to 5 m/s. The five rural sites show no or small decreases in PM2.5 with increased wind speeds during the cold and warm seasons (Figure 13). These PM2.5 concentrations are between 3 and 6 mg/m3. There is no evident PM2.5 relationship with surface wind direction during the cold or warm season (Figure 14).

The PM2.5 concentrations at the rural sites during high wind speeds are about 3 and 5 mg/m3 during the cold and warm season respectively (Figure 13). At Phoenix, the PM2.5 concentration under the high wind speeds is ~7 mg/m3 during the cold season and ~5 mg/m3 during the warm season. These Phoenix concentrations may still be significantly influenced by local sources since the highest wind speeds were between 5 – 7 m/s which may not have been swift enough to completely ventilate the urban center.

PM2.5 as a Function of Airmass History Wind Speed and Direction. During the cold season, the PM2.5 concentrations at the urban sites Phoenix and Long Beach decrease ~55% as the airmass history wind speed increases from 1 to 9 m/s (Figure 15). The PM2.5 concentrations at the two rural sites Hopi Point and Tonto initially increase with increasing wind speeds, but then decrease ~70% from their peaks (Figure 15). The other rural sites show no trends with increasing wind speeds. At Phoenix, the high wind speed PM2.5 concentration is ~10 mg/m3. At the rural sites, concentration varies between 2 to 4 mg/m3 depending on the site.

During the warm season, all sites in Arizona show no relationship between the PM2.5 and airmass history winds. This is highly suspect. Past analyses of the NGM winds in Arizona compared to observations have shown large biases in the NGM wind fields (Schichtel 1994). It is likely that these biases are affecting the results so that no conclusions can be drawn.

Surface Winds Surface Winds – Normalized PM2.5

Figure 9. Dependence of PM2.5 concentrations on surface wind speed for urban and rural sites in the Mid-Atlantic States during the warm and cold seasons. The normalized concentrations are PM2.5 divided by the PM2.5 concentration at wind speeds < 2 m/s.

Cold Season Surface Winds Warm Season Surface Winds

Figure 10. Dependence of PM2.5 concentrations on surface wind speed and direction for urban and rural sites in the Mid-Atlantic States during the cold (November – March) and warm (April – September) seasons.

AMH Winds AMH Winds – Normalized PM2.5

Figure 11. Dependence of PM2.5 concentrations on airmass history wind speed for urban and rural sites in the Mid-Atlantic States during the warm and cold seasons. The normalized concentrations are PM2.5 divided by the PM2.5 concentration at wind speeds < 2 m/s.

Cold Season AMH Winds Warm Season AMH Winds

Figure 12. Dependence of PM2.5 concentrations on airmass history wind speed and direction for urban and rural sites in the Mid-Atlantic States during the cold (November – March) and warm (April – September) seasons.

Surface Winds Surface Winds – Normalized PM2.5

Figure 13. Dependence of PM2.5 concentrations on surface wind speed for urban and rural sites in the Southwest during the warm and cold seasons. The normalized concentrations are PM2.5 divided by the PM2.5 concentration at wind speeds < 2 m/s.

Cold Season Surface Winds Warm Season Surface Winds

Figure 14. Dependence of PM2.5 concentrations on surface wind speed and direction for urban and rural sites in the Southwest during the cold (November – March) and warm (April – September) seasons.

AMH Winds AMH Winds – Normalized PM2.5

Figure 15. Dependence of the PM2.5 concentrations on airmass history wind speed for urban and rural sites in the Southwest during the warm and cold seasons. The normalized concentrations are PM2.5 divided by the PM2.5 concentration at wind speeds < 2 m/s.

Cold Season AMH Winds Warm Season AMH Winds

Figure 16. Dependence of PM2.5 concentrations on airmass history wind speed and direction for urban and rural sites in the Southwest during the cold (November – March) and warm (April – September) seasons.

5.0 Summary and Conclusions

This work examined the seasonal local and regional source contributions of PM2.5 to urban areas in the Mid-Atlantic States: Baltimore, MD, Washington, DC and Philadelphia, PA and Phoenix, AZ in the Southwest. This was accomplished using two different methods. The first derived urban excesses by comparing seasonal PM2.5 trends at the urban monitoring sites to nearby rural monitors. The second approach used a simple model based on the PM2.5 dependence on wind speed and direction. The wind vectors were derived from surface observations and air mass histories. The second approach was able to classify a site as being dominated by local or regional source contributions as well as estimating the regional contributions under high wind speed conditions. Neither approach is definitive in its ability to distinguish and quantify regional and local PM2.5 contributions, and are more appropriate for weight of evidence evaluations. In this work both approaches corroborated each others results providing an additional measure of confidence in the conclusions.

It was found that at the Washington, DC and Philadelphia, PA urban sites, the winter excess was 4-8 mg/m3 or 40% of the total PM2.5 concentration. This decreased during the summer period to 1-8 mg/m3 or 20% of the total PM2.5, indicating the regional contributions are more important during the winter than the summer. The same conclusion was drawn from the PM2.5 dependence on surface wind speed analysis where it was found that the cold season PM2.5 concentrations at Washington, DC decreased nearly linearly with increasing wind speeds, but were essentially independent of wind speed during the warm season. The regional concentrations during high surface wind speeds were about 7 and 13 mg/m3 during the cold and warm season respectively or about 70% of the concentrations from the regional monitoring sites.

The Phoenix, AZ winter urban excess was also greater than the summer urban excess, 9-18 mg/m3 (50-90%) compared to 3-6 mg/m3 (30-70%). The lower summer time urban impact at Phoenix, AZ compared to winter was also reflected in the PM2.5 to wind speed dependence plots. The warm season PM2.5 decreased at a lower rate with increasing wind speeds than during the cold season. The regional concentrations during high surface wind speeds were about 7 and 5 mg/m3 during the cold and warm season respectively. This is 50% greater than the winter PM2.5 concentrations at the regional sites and about equal to the PM2.5 concentrations at the regional sites during the summer.

The PM2.5 dependence on airmass history winds could not classify a site as being dominated by regional or local concentrations. However, these winds were better able to identify the PM2.5 dependence on wind direction. It was found that in the Mid-Atlantic States airmass history winds from the southeast during the cold season were associated with PM2.5 concentrations ~5 mg/m3 higher than winds from the northeast. This difference increased to about 10 mg/m3 during the warm season.

6.0 Acknowledgment.

The U.S. Environmental Protection Agency through its Office of Research and Development funded and managed the research described here under contract # 7D-0869-NAEX to Bret A. Schichtel. It has been subjected to the Agency's peer and administrative review and has been approved for publication as an EPA document.

7.0 References

Fast, J.D. and Berkowitz, C.M. 1997. Evaluation of back trajectories associated with ozone transport during the 1993 North Atlantic Regional Experiment. Atmos. Environ. 31, 825-837.

Gifford, F.A. and Hanna, S.R. 1973. Modelling urban air pollution. Atmos. Environ. 7, 131-136.

Husar, R.B. Lodge Jr., J.P. and D.J. Moore. 1978. Sulfur in the atmosphere. Proceedings of the International symposium held in Dubrovnik Yugoslavia, 7-14 September 1977. Pergamon Press, Oxford.

Husar, R.B. and Renard W.P. 1998. Ozone as a Function of Local Wind Speed and Direction: Evidence of Local and Regional Transport; Paper No. 98-A922, presented at the 91st Annual Air & Waste Management Association Meeting, San Diego, CA, June 1998.

Rolph G.D. 1992. NGM Archive. NCDC Report TD-6140, July, National Climatic Data Center.

Schichtel, B.A. and Husar, R.B. 1992. Aerosol types over the continental U.S.: spatial and seasonal patterns. Presented at the A&WMA conference in Kansas City, MO. Paper 92-60.07.

Schichtel, B.A. 1994. Verification of the NGM wind fields in the Southwestern US during 1992. CAPITA report.

Schichtel, B.A., Falke, S.R. and R.B. Husar. North American Integrated Fine Particle Data Set. Paper No. 99-398, submitted for presentation at the 92nd Annual Air & Waste Management Association Meeting, St. Louis, MO, June 1999.

Schichtel, B.A. and Husar, R.B. 1996. Regional simulation of atmospheric pollutants with the CAPITA Monte Carlo model. J. of Air & Waste Manage. Assoc. 47, 331-343.

Schichtel, B.A. and Husar, R.B. 1999. Sulfur kinetic rate coefficients over the eastern US derived using a semi-empirical approach. Accepted for publication in the J. of Air & Waste Manage. Assoc.

U.S. Environmental Protection Agency (EPA). 1996. Air Quality Criteria for Particulate Matter Volume I. Research Triangle Park, NC: Office of Air Quality Planning and Standards; EPA report nos. EPA/600/P-95/001aF.

Appendix A: Comparison of PM2.5 Collected by a Dichotomous and DFPSS Sampler in the National PM Research Monitoring Network

The National PM Research Monitoring Network's monitoring platforms had a dichotomous sampler collecting 24 hour integrated samples every three days and a dual fine particle sequential sampler (DFPSS) collecting 24 hour integrated samples every day. This appendix compares the fine particulate estimates from these collocated samplers at the Baltimore, MD and Phoenix, AZ monitoring sites.

The Baltimore, MD PM2.5 data collected using the dichotomous and DFPSS samplers are compared in Figure A-1. As shown in the scatter plot, there were days when the two PM2.5 measurements differed by more than a factor of two. These days were often associated with non standard sampling flags, but December 15, 1997 was not flagged and the dichotomous sampler measured 23 ug/m3 while the DFPSS sampler measured 11 ug/m3. Over all, the correlation between the two samplers was 0.9 with the DFPSS PM2.5 on average 1 ug/m3 or 7% less than the dichotomous PM2.5 data (Figure A-1A). The seasonal PM2.5 trends are compared in Figure A-1B. These trends were created by only including days with both dichotomous and DFPSS samples in the monthly averages. As shown, there is close agreement between the two samplers from February – July, but from August – December, the DFPSS sampler measured about 2 mg/m3 less than the dichotomous sampler did.

The scatter plot comparing the Phoenix dichotomous and DFPSS data, Figure A-2A, show better agreement than the Baltimore data. There are no outliers, the correlation coefficient is 0.96, and the overall averages differ by only 1.5%. The good correspondence is also seen in the dichotomous and DFPSS seasonal trends in Figure A-2B where the monthly averages are nearly identical.

A B

Figure A-1. A) Scatter plot comparing the Baltimore, MD dichotomous and DFPSS PM2.5 data. B) Baltimore, MD 1997 seasonal trend created from days with both dichotomous and DFPSS samples.

A B

Figure A-2. A) Scatter plot comparing the Phoenix, AZ dichotomous and DFPSS PM2.5 data. B) Phoenix, AZ seasonal trend created from days with both dichotomous and DFPSS samples.

Appendix B: The dependence of PM2.5 on Elevation in the Southern Appalachian Mountains.

The dependence of PM2.5 on elevation was examined by comparing the seasonal trends from five IMPROVE monitoring sites located along a 300 mile segment of the southern Appalachian Mountains (Table B-1). The lowest elevation site, Jefferson, NF, was located in a valley between Shenandoah and Dolly Sods to the Northeast and Smoky Mountains and Shining Rock to the southeast (Figure 1). The seasonal trends for the five locations are displayed in Figure B-1. As shown, during the summer months, May – September, the four sites (Jefferson NF, Great Smoky Mountains, Shenandoah and Dolly Sods) have virtually identical monthly averaged PM2.5 concentrations. The elevations for these sites range from 280 to 1200 m (Table B-1). Shining Rock with an elevation of 1700m, has 3-5 mg/m3 less PM2.5 than the other sites. During the winter months, the highest concentrations occur at the lowest elevation site, Jefferson NF at ~11 mg/m3, decreasing to ~7 mg/m3 at the three intermediate sites and again decreasing to 3 – 4 mg/m3 at Shining Rock. The effect of elevation during August and January at the Appalachian monitoring sites is presented in Figure B-1B. During August, the PM2.5 is independent of height up to at least 1200 m where it then decreases. During January, the PM2.5 decreases about 50% from 280 to 760 m. It is then approximately constant up to 1200 m where the concentrations again decrease ~50% from 1200 to 1700 m.

The overriding difference between the January and August elevation trends is that the concentration at Jefferson NF, the low elevation site, during January are about twice as large as at the intermediate elevation sites, but the Jefferson NF concentrations are approximately equal to those at the intermediate elevation sites during August. A plausible alternative explanation to PM2.5 dependence on elevation for the increased Jefferson NF winter PM2.5 concentrations is that this site is influenced by local sources. Jefferson NF is located in a valley and the low winter mixing heights could trap local emissions in the valley increasing the concentrations.

Table B-1. The IMPROVE monitoring sites located in the southern Appalachian Mountains and their elevation, and average August and January PM2.5 concentrations. See Figure 1 for the location of the monitoring sites.

Monitoring Site

Elevation (meters)

PM2.5 (mg/m3)
August

PM2.5 (mg/m3)
January

Jefferson NF, VA

280

22.5

12

Great Smoky Mnts, TN

760

23.5

6.6

Shenendoah, VA

1100

22.7

6

Dool Sods, WV

1160

24

7.2

Shining Rock, NC

1680

19

3.3

A B

Figure B-1. A) Seasonal trends of the five IMPROVE monitoring sites in the southern Appalachian Mountains. B) The PM2.5 dependence on elevation in the southern Appalachian Mountains. The PM2.5 concentrations are divided by the concentration at Jefferson NF’s, the lowest elevation site.

 

Appendix C: Creation of the Airmass History Wind Field

The airmass history wind field is comprised of 24-hour average wind speeds and directional frequency roses derived from airmass histories. These winds are a regional wind field taking into account the changes in wind speed and direction with height and along an air parcels pathway prior to impacting the receptor. The airmass history winds were created using three-day back trajectories for about 500 locations evenly distributed over most of North America from 1991 – 1996 (Figure C-1).

Figure C-1. The spatial, temporal , and variable domians of the airmass history wind field database.

The back trajectories were calculated using the CAPITA Monte Carlo particle dispersion model (Schichtel and Husar 1996; Schichtel and Husar 1999). This model simulates atmospheric transport and diffusion by tracking the movement of multiple "particles" released from sources forward in time or released from receptors backwards in time. The intense vertical mixing that takes place within the atmospheric boundary layer is simulated using a Monte Carlo technique. The model was driven by the synoptic scale meteorological fields from the National Meteorological Center’s Nested grid Model (NGM) (Rolph, 1992). These are synoptic scale meteorological data containing three dimensional wind vectors on an ~160 km grid (Figure C-2) with ten vertical layers up to seven kilometers at a two hour time step. The lowest layer has a height of ~160 meters.

The back trajectories were generated from a forward particle flow simulation over most of North America. The flow simulation was created by releasing three particles every two hours from 504 sources evenly distributed over most of North America (Figure C-2). The movement of these particles were then tracked at two hour intervals for seven days or until they reached the edge of the NGM grid. At each two hour interval the three dimensional position of the particles were saved into an airmass history database.

Figure C-2. The NGM meteorlogical grid the Monte Carlo Model operated on. The squares mark the position of each source that released particles in the atmospheric flow simulation..

The back trajectories for a given receptor time period and receptor volume were then created by querying the airmass history database for all particles in the receptor volume and at the receptor time which were less than 3 days old. The back trajectory then consisted of the forward particle trajectories from the receptor volume back to their original source. These back trajectories are identical in concept to a "typical" back trajectory that simply releases particles from a receptor and tracks their pathways back in time. The benefits of these back trajectories over typical ones is that they enable the creation of back trajectories anywhere in North American by simply querying the airmass history database instead of having to rerun the model. This also avoids some of the controversy concerning the validity of the typical back trajectory’s ability to track a parcel’s pathway from the receptor (Fast and Berkowitz 1997). The main draw back, is that the spatial extent of the back trajectories do not extend to the edge of the meteorological grid, but the edge of the uniform source grid, which in the case of this work extends only to the North American coast (Figure C-2).

C.1 Airmass History Wind Speed and Direction.

The airmass history wind speed is a measure of the airmass residence over the region prior to impacting the receptor. It was calculated by averaging together the straight line distances, d, of the particles from the receptor to the source that they were initially released from divided by the particle's age, t. This average is taken over all particles at a given receptor up to the maximum age T (Figure C-3). In this work, the maximum age was three days.

 

 

 

 

 

 

 

 

 

where u is the airmass history wind speed

di is the distance between the receptor and source for particle i

ti is the age of particle i at the receptor from its time of release

N is the total number of particles at the receptor with age less than T

T maximum particle age

q angle the line from the receptor to the source makes with due North

Figure C-3. Illustation and equation used to calculation the airmass history wind speed.

The airmass history wind direction is a frequency distribution of the directions that all of the particles at the receptor came from. The particle direction is the angle between north and the straight line from the receptor to the particle’s source. This is angle q in Figure C-3. In this work, the frequency distribution was created by placing each particle direction into one of eight 45o sector bins. These bins were for sectors 0-45, 45 – 90, etc. Each bin count was then divided by the total number of particles at the receptor, creating the directional frequency distribution. No bin count was incremented at those times when the receptor and source locations coincided, but these particles were counted in the total particle count.

Appendix D. PM2.5 as a Function of Wind Speed and Direction Figures

The following figures present the dependence of the PM2.5 on wind speed and directions for all Mid-Atlantic and Southwest monitoring sites used in this study. The methodology for the generation of these plots is presented in section 4.2. Note in these plots normalized PM2.5 concentrations are the PM2.5 concentrations dividing by the concentration when the wind speed was less than 2 m/s.

Cold Season Surface Winds

Figure D-1. Dependence of PM2.5 concentration on surface wind speed and direction for urban and rural sites in the Mid-Atlantic States during the cold season (November – March).

Cold Season Surface Winds – Normalized PM2.5

Figure D-2. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on surface wind speed and direction for urban and rural sites in the Mid-Atlantic States during the cold season (November – March).

Warm Season Surface Winds

Figure D-3. Dependence of PM2.5 concentration on surface wind speed and direction for urban and rural sites in the Mid-Atlantic States during the warm season (April – September).

Warm Season Surface Winds – Normalized PM2.5

Figure D-4. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on surface wind speed and direction for urban and rural sites in the Mid-Atlantic States during the warm season (April – September).

Cold Season AMH Winds

Figure D-5. Dependence of PM2.5 concentration on airmass history wind speed and direction for urban and rural sites in the Mid-Atlantic States during the cold season (November – March).

Cold Season AMH Winds – Normalized PM2.5

Figure D-6. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on airmass history wind speed and direction for urban and rural sites in the Mid-Atlantic States during the cold season (November – March).

Warm Season AMH Winds

Figure D-7. Dependence of PM2.5 concentration on airmass history wind speed and direction for urban and rural sites in the Mid-Atlantic States during the warm season (April – September).

Warm Season AMH Winds – Normalized PM2.5

Figure D-8. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on airmass history wind speed and direction for urban and rural sites in the Mid-Atlantic States during the warm season (April – September).

Cold Season Surface Winds

Figure D-9. Dependence of PM2.5 concentration on surface wind speed and direction for urban and rural sites in the Southwest during the cold season (November – March).

Cold Season Surface Winds – Normalized PM2.5

Figure D-10. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on surface wind speed and direction for urban and rural sites in the Southwest during the cold season (November – March).

Warm Season Surface Winds

Figure D-11. Dependence of PM2.5 concentration on surface wind speed and direction for urban and rural sites in the Southwest during the warm season (April – September).

Warm Season Surface Winds – Normalized PM2.5

Figure D-12. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on surface wind speed and direction for urban and rural sites in the Southwest during the warm season (April – September).

Cold Season AMH Winds

Figure D-13. Dependence of PM2.5 concentration on airmass history wind speed and direction for urban and rural sites in the Southwest during the cold season (November – March).

Cold Season AMH Winds – Normalized PM2.5

Figure D-14. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on airmass history wind speed and direction for urban and rural sites in the Southwest during the cold season (November – March).

Warm Season AMH Winds

Figure D-15. Dependence of PM2.5 concentration on airmass history wind speed and direction for urban and rural sites in the Southwest during the warm season (April – September).

Warm Season AMH Winds – Normalized PM2.5

Figure D-16. Dependence of the PM2.5 concentration, normalized by the low speed PM2.5 concentration, on airmass history wind speed and direction for urban and rural sites in the Southwest during the warm season (April – September).