Fine Particle Maps Derived from Regional PM2.5 and Visibility Data

Contents:

• Background and Rationale
• Approach
• Fine Mass Concentrations
• Visibility as a Fine Particle Surrogate
• Visibility Data Processing
• Bext-Fine Mass Relationship
• Data Flow for Fine Particle Mass Map Preparation
• Resulting Fine Mass Concentration Maps
• Limitations and Possible Improvements to Fine Particle Maps
• Appendix I
• References

Fine particles are known contributors of visibility degradation, acid deposition, and regional climate change. Recent epidemiological studies also indicate fine particles as causal factors in human health effects. For this and other reasons, the Environmental Protection Agency is considering the promulgation of ambient concentration standards for fine particles. In this report, the term fine particles refers to the mass concentration of particles below 2.5 µm (nominal). In other contexts, the fine-coarse particle size cut may be anywhere between 1 and 2.5 µm.

The spatial and temporal pattern and trends of fine particles over the U.S. is not well established since systematic fine particle monitoring has begun only in the 1980s and at limited number of sites. On the other hand, estimating the fine particle effects on human health and other effects requires higher resolution than what is provided by existing fine particle monitoring networks. Higher spatial and time resolution particle concentrations can be estimated by several methods:

1. Integration and fusion of multiple fine particle mass concentration data sets.
2. Use of surrogate variables for spatial and temporal extrapolation.
3. Application of properly calibrated and verified regional and local atmospheric transmission models as data extrapolators.

This report applies the surrogate method 2, by utilizing the existing higher resolution visibility data for the interpolation of measured fine particle concentrations.

Back to Contents

Data Sets

Data Fusion

Data fusion applied in this context reconciles the multiple data inputs by re-projecting the data sets onto a common spatial and temporal resolution. Following such reconciliation, standard data operations, such as addition subtractions, etc. can be performed on spatially compatible data sets. The results of such manipulations are fused data. In this particular report, the fusion is between measured fine particle concentrations and the visibility derived light extinction data.

Data Flow Programming

Schematic illustration of data flow programming.

Such data flow programming is particularly suitable for graphic user interfaces where the selection of the components, the linking among the components, as well as the overall data flow between components can be laid out using drag-and-drop visual languages. The conceptual simplicity, along with the visual feedback from the data flow programming language allows end-users (scientists) to create their own data processing programs without the participation of professional programmers.

Back to Contents

Our current understanding of the US national fine mass concentration (FM) pattern arises primarily from non-urban monitoring networks, the Interagency Monitoring of Protected Visual Environment (IMPROVE) and Northeast States for Coordinated Air Use Management (NESCAUM). Additional FM data are also available through EPA's Aerometric Information Retrieval System (AIRS). However, the AIRS FM data were not used in this analysis.

The IMPROVE/NESCAUM non-urban aerosol concentrations are measured at remote sites, away from urban-industrial activities. The sites are located mostly in national parks and wilderness areas. Size-segregated aerosol mass and chemical composition data are available for 50 sites, through the IMPROVE (Eldred et al., 1987; Eldred et al., 1988; Eldred et al., 1989) and NESCAUM (Flocchini et al., 1990, Poirot et al., 1990) networks. The PM10 and PM2.5 mass concentrations are sampled and analyzed on separate filters. The sampling frequency was generally twice a week for 24-hours. The PM2.5 samples are analyzed for chemical composition which make the data sets suitable for chemical mass balance computations (e.g. Gebhart and Malm, 1987; Schichtel and Husar, 1991; Sisler et al., 1993; Eldred and Cahill, 1994; Sisler and Malm, 1994). The IMPROVE/NESCAUM aerosol data are available from 1988 through 1993.

An example of national fine mass concentration pattern is shown below.

Fine Mass Concentrations
Quarter 1 (1988-92)	Quarter 2 (1988-92)	Quarter 3 (1988-92)	Quarter 4 (1988-92)	Annual (1992)

The Figures show the location of the monitoring sites, the magnitude of annual fine mass concentration at these sites as well as the estimated contour lines. The contours drawn for the eastern US are derived from only 15-20 stations. Given the sparseness of the data, the contour lines are to be taken as guides to the eye and not as actual pattern. The coarseness of the contoured data in the Figure confirms the need to develop higher resolution fine mass concentration maps.

Back to Contents

Visual range monitoring at meteorological observation sites provides an attractive surrogate and augmentation of fine particle concentration measurements. Computerized visibility records exist since 1948 for several hundred US sites, compared to 20-50 fine particle monitoring sites. The meteorological visibility data have been evaluated for pattern and trends (Munn, 1973; Husar et al. 1976; 1979;1981; Robinson and Valente, 1982; Trijonis, 1982; Sloane, 1982a; 1982b; Husar et al. 1986; Sloane, 1984; Wolff et al., 1982; Elridge, 1966) and compared to the emission trends (Husar and Wilson, 1993).

The excellent spatial and temporal coverage of the visibility data base can be utilized only after careful site by site scrutiny for interference and anomalous behavior. The limitations of the visibility database includes threshold visibility, maximum visibility reported, lack of suitable dark targets, visual acuity of human observers, and meteorological influences (rain, fog, snow) on visual range. Nevertheless, the consistent visibility observations over the past 50 years at over 280 US sites provide a stable, long-term, well behaving data set for spatial and temporal pattern analysis.

Visibility (Extinction Coefficient - 75th Percentile)
Quarter 1 (1988-92)	Quarter 2 (1988-92)	Quarter 3 (1988-92)	Quarter 4 (1988-92)	Annual (1987-89)	Annual (1990-92)	Annual (1992)

The data quality control of the visibility data was implemented: (1) using visual inspection of daily time series for each station, (2) observing the percentile trends, and (3) comparing the time series for neighboring sites. The description of the methodologies used in visibility data pre-processing and quality control is given in the companion report "Methodology for Visibility Trend Analysis." The results of the visibility pattern and trend analysis have been presented in an earlier report, Husar et al., 1994.

Back to Contents

For purposes of spatial-temporal trend analysis, raw visibility observations were summarized as quarterly averages of extinction coefficient, but significant pre-processing was applied. The data pre-processing included a time of day and a meteorological filters. Only daily local noon observations were used. Nightime observations were ignored since they are not uniquely related to the extinction coefficient.

Meteorological parameters such as precipitation (rain, fog, and snow) and humidity tend to influence visibility and their role was minimized by the application of meteorological filters. The role of these natural obstructions to vision can be eliminated by discarding the data when these meteorological phenomena occur.

The aggregation of extinction coefficient data into monthly or quarterly averages was performed using non-parametric quantiles of the distribution function. A major advantage of quintiles in trends is that such nonparametric statistics require no assumptions about the form of the actual distribution function. Trijonis and Yuan (1978), Husar et al. (1979), and others have successfully used this approach. For each month 25th, 50th, 75th, and 90th percentiles were computed. For this work the 75th percentiles were used.

For each month and station, three different extinction coefficients were calculated: 1) The first set included all visibility data regardless of weather and pollutant conditions (BX); 2) the second group (FX) is composed of extinction coefficients excluding precipitation, fog events; and RH>90%; 3) the third set of extinction coefficients (RX) also used the precipitation and fog filter but also included a RH correction factor to compensate for water and vapor effects, RC=FX/C(RH). In effect, the role of the factor is to normalize all the extinction coefficients to 60% relative humidity. The strongest correction factor is applied at high humidities (90%). Bext values measured at RH>90% were eliminated. In the analysis below the 75th percentile FX values were used, fog and precipitation values removed.

The spatial pattern of extinction coefficients is presented as contour maps. The contouring procedure is described in Appendix 1. The resulting contours as well as the corresponding stations, Bext data (the square size is proportional to Bext at a monitoring site) are shown in Figure 3. The shades are 0.03 km-1 apart. The darkest shade (red) has an extinction coefficient (75th) percentile of >0.2 km-1 which corresponds to 1.9/0.2=9.5 km visual range. The lowest contour is set at 0.05 km-1which corresponds to 1.9/0.05=38 km visual range.

Back to Contents

Bext-Fine Mass Relationship

The visual range or visibility is a subjective concept, being a maximum distance at which an observer can discern the outline of an object. It is a measure of the quality of the atmospheric optical environment. As a measure of air pollution, visual range has the disadvantage that it is inversely related to aerosol concentration. A more suitable measure is haziness or extinction coefficient, Bext, defined as Bext =K/visual range, where K is the Koschmieder constant. The extinction coefficient is in units of km-1 and it is proportional to the concentration of light scattering and absorbing aerosols and gases.

The value of K is determined by both the threshold sensitivity of the human eye as well as by the contrast of the visible objects against the horizon sky. The original value of K proposed by Koschmieder was 3.92 which corresponds to detectable threshold contrast of 2%. In this report, we have taken K=1.9 in accordance with more recent data by Griffing 1980; Dzubay et al., 1982; Stevens et al., 1984, Ozkaynak et al., 1985.

Bext to Fine Mass Ratio
Quarter 1 (1988-92)	Quarter 2 (1988-92)	Quarter 3 (1988-92)	Quarter 4 (1988-92)	Annual (1988-92)

The Bext/FM ratio is a key variable in the present data fusion process. The use of Bext as a fine mass surrogate relies on a spatially smooth Bext/FM ratio. In general, the spatial pattern is smooth. Notable exceptions are high Bext/FM values over northern Minnesota, eastern California and a site in southern Oregon, particulalry in quarters 1 and 3.

High values of Bext/FM will arise either due to higher than expected Bext, or lower than expected FM.

Back to Contents

The procedure for creating fine particle maps using the visibility surrogate is shown the data flow chart. The Figure contains two types of components: the shaded boxes represent operators or (processors) that take input data, manipulate them, and produce an output. The oval elements represent the data inputs/outputs to/from the operators. The numbers adjacent to the data elements are sequential step output counters.

Data Flow for Fine Particle Map

The first operations involving data, step numbers 1 and 2, entail aggregating daily Bext and fine mass data to quarterly averages. The quarterly Bext and fine mass data were constructed in table formats, containing columns for site name, longitude, latitude, and associated Bext or fine particle concentration values. Additional tables containing averaged Bext data over the time periods 1987 -1989, 1990 - 1992, and 1992 was also constructed as shown as step 9. The values for the annual Bext data table are displayed as squares in the extinction coefficient maps.

The annual Bext tables were transformed to grid format using an operator called the Contourer. The inputs consist of the table, as well as user specified inputs that include a distance weight function, a radius of influence, a minimum required number of data points within the radius, a maximum number of data points used in the calculation of the grid value, and the dimensions for the output grid.

The next step, 6, was to produce a quarterly Bext/FM ratio table. This was accomplished using a ratio table producing operator. The inputs to the operator were the quarterly Bext grid and the quarterly fine mass table. Since the fine mass data were in table format, each concentration value was associated with a particular latitude and longitude. The Bext data were in the form of a 120 X 80 grid. The operator read through the two inputs and from the fine mass table it extracted the longitude, latitude, and fine mass concentration data. The operator used the longitude and latitude values from the table as search parameters in the grid. It extracted Bext values from the grid nearest the latitude and longitude from the fine particle table. The operator then divided the Bext value by the corresponding fine particle concentration value. An output table was produced containing the latitude, longitude, and Bext/FM ratio value.

The Bext /fine mass ratio table was sent through the Contourer to produce a ratio grid. This process is identical to that described in the conversion of Bext tables to grids. The resulting Bext/fine mass ratio grid was used with the quarterly Bext grid as inputs to the multiplier operator which produced annual fine particle grids. The multiplier read through the two input grids and multiplied the Bext values by the inverse of the ratio values to calculate a visibility fine particle concentration grid.

The resulting grid is shown in the visibility corrected fine mass maps. The display and printing of all the maps was accomplished through the MapEdit program.

Back to Contents

The resulting annual fine mass maps are shown below.

Visibility Corrected Fine Mass
Quarter 1 (1988-92)	Quarter 2 (1988-92)	Quarter 3 (1988-92)	Quarter 4 (1988-92)	Annual (1987-89)	Annual (1990-92)	Annual (1992)

A comparison of the fine mass concentration maps using only the fine mass data with the visibility-augmented FM maps shows significantly more detail for the latter. The fused fine mass concentrations show spatial heterogeneity over both the eastern and western US. It is likely, however, that in reality the spatial gradients are much stronger than depicted, particularly over urban centers and topographically confined valleys of the western states.

The comparison of the three maps representing annual averages for 1992, 1990-92, 1987-89, show substantial year to year and regional variation. For example, the Northeast and Mid-Atlantic states exhibit highest concentration in 1987-89. On the other hand, the fine mass estimates for the Gulf states are highest for 1990-92.

The quarterly maps show the fine mass concentration peaks in quarters 2 and 3. The visibility surrogate decreased concentrations from intial fine mass map in large areas of the west, such as the Oregon,Nevada, and Northern California area and the region extending from Montana southward to Colorado. The east shows more defined concentration contours with increases in concentrations in some areas, parts of Northa Carolina and Virginia in quarter 3, for instance, and decreases in others, like the Appalaican area of Virginia.

A verification of the fine mass maps was accomplished by comparing the measured fine mass concentrations at about 50 IMPROVE/NESCAUM sites to the reconstructed fine mass following the data fusion. Ideally, the measured and reconstructed data should be identical. For most monitoring stations the measured and fused values are virtually identical. However, for some stations the fused data are significantly less than the measured values. This is due to the smoothing of data surfaces during the gridding of Bext/FM ratios. Very high data values, such as those in eastern California and Oregon tend to be reduced by the inclusion and weighing of other neighboring sites. Another cause of error in matching the measured concentrations occurs when creating a grid from the point concentrations. The grid resolution (120X80) provides for a grid cell of approximate dimensions 45kmX30km. The value associated with a cell is based on the centriod of the cell. In most cases, fine mass sites were not located at the center of the cells so that the cells in which they were located did not have the same concentration as the site itself. As a result, when concnetrations were extracted from the grid they were different than the measured fine mass, especially in high station density areas where multiple sites significanlty influenced the value of a single grid cell.

Back to Contents

The fine particle maps of the conterminous US presented here have known limitations. The causes of these limitations and possible remedies in future work are stated below.

Inadequate PM2.5 concentration data

An improvement of the coverage of the local hot spot concentrations could be achieved by the gathering and proper inclusion of more fine particle concentration data in the spatial analysis work. The candidate data sets for this purpose include fine particle data in the national AIRS system, as well as numerous short term and geographically limited fine particle data sets that have been gathered over the US, e.g. Philadelphia, St. Louis, Salt Lake City, Los Angeles, Grand Canyon region, Oregon, and Washington. Prudent fusion of these data sets would provide a direct sensory input for more spatially resolved fine particle maps.

Use of visibility as a surrogate and interpolator

A possible improvement would arise from the combination, i.e. data fusion of several different fine particle surrogates in addition to visibility. Such surrogates may include concentrations of PM10 (available at over 1000 sites), concentration of aerosol sulfate or other chemical species and the possible use of aerosol remote sensing by geostationary and polar orbiting meteorological satellites.

Spatial interpolation schemes

An improvement in the interpolationpolation could be achieved through the consideration of other physical or chemical factors that inhibit or enhance the spatial dispersion of air pollutants. The most obvious of these factors is the terrain elevation. The use of elevation in the contouring process would necessitate certain assumptions regarding the elevation dependence of aerosol concentration. Using this approach for example, high elevation mountain peaks would be assigned low concentrations even though they may be in the vicinity of high concentration valley sites.

Physico-chemical modeling

Back to Contents

An important data transformation processes used in this work is contouring. Most of the spatial patterns are presented on contour maps. The contours were derived from the point observations using a spatial extrapolation scheme. In the first step, the data from the random locations were projected to a uniform grid with 120x80 nodes that covers the conterminous U.S. The gridding used inverse distance squared (1/r2) as the station weighing factor. The extrapolations outside the U.S. boundaries were trimmed to eliminate spurious values.

Contourer Pseudocode

Contour
	Functionality:	transform data from table format into grid format
	Input:	Table - a set of data points
		WeightFunc - distance weight function
		Radius - distance constraint
		MinPoints - minimum required number of data points within Radius
		MaxPoints - maximum number of data points used in calculation of a grid cell
	Output:	Grid - uniformly distributed set of data points
function Contour(out Grid, in Table, in WeightFunc, in Radius, in MinPoints, in MaxPoints)
{
	for each Cell in Grid
		CalcCell
}

CalcCell
	Functionality:	calculate value of a grid cell given a table of data points around it
function CalcCell(in out Cell, in Table, in WeightFunc, in Radius, in MinPoints, in MaxPoints)
{
	PointList = an empty list of table points
	for each point in Table that is not null
		if the distance between the point and the cell is less than Radius then
			add point to the PointList, sorted by distance
	Cell value = Interpolate PointList
}

Interpolate
	Functionality:	interpolate over a sorted list of data points
function Interpolate(in PointList, in WeightFunc, in MinPoints, in MaxPoints)
{
	if number of points in PointList < MinPoints return null value

	TotalWeight = 0
	Sum = 0
	WeightExp = 0 if WeightFunc = 1
		    1 if WeightFunc = 1/r
		    2 if WeightFunc = 1/r2, etc
	
	for first MaxPoints in PointList
		Dist = point distance from cell
		Weight = 1 / (Dist^WeightExp)
		TotalWeight = TotalWeight + Weight
		Sum = Sum + Weight * point value
		
	if TotalWeight = 0 return null value
	Sum = Sum / TotalWeight
	return Sum

Abidi M.A. and Gonzalez R.C. (eds) Data Fusion in Robotics and Machine Intelligence, Academic Press, Inc. Boston (1992).

Dzubay T.G., Stevens R.K., Lewis, C.W. Visibility and aerosol composition in Houston, Texas, Environ Sci. Technol., 16, 514-525 (1982).

Eldred R.A., Feeney P.J., Cahill T.A., and Malm W.C. Sampling techniques for fine particle/visibility studies in the National Park Service network. In: Visibility Protection Research and Policy Aspects, P.S. Bhardwaja, Ed. pp 386-396 (1987).

Eldred R.A., Cahill T.A., Pitchford M., and Malm W.C. IMPROVE-a new remote area particulate monitoring system for visibility studies, APCA Annual Meeting, Dalla, TX (1988).

Eldred R.A., Cahill T.A., Wilkinson L.K., Feeney P.J., Chow J.C., and Malm W.C. Measurement of fine particles and their chemical components in the IMPROVE/NPS networks. In: Visibility and Fine Particles, C.V. Mathai, Ed. pp 187-196 (1989).

Eldred R.A. and Cahill T.A. Trends in elemental concentrations of fine particles at remote sites in the United States of America, Atmospheric Environment, 28, 1009-1019 (1994).

Elridge, R. G., Climatic visibilities of the United States. J. Appl. Meteorol. 5, 227-282 (1966).

Flocchini R.G., Cahill T.A., Eldred R.A., and Feeney P.J. Particulate sampling in the Northeast, a description of the Northeast States for Coordinated Air Use Management (NESCAUM) network. In: Visibility and Fine Particles, C.V. Mathai, Ed. pp 197-206 (1989).

Gebhart K.A. and Malm W.C. Source apportionment of particulate sulfate concentrations at three National Parks in the Eastern United States. In: Visibility Protection Research and Policy Aspects, P.S. Bhardwaja, Ed. pp 898-913 (1987).

Griffing G.W. Relationships between the prevailing visibility, nephelometer scattering coefficient, and sunphotometr turbidity coefficient. Atmos. Environ. 14, 577-584 (1980).

Husar, R.B., Gillani, N.V., Husar, J.D., Paley, C.C., and Turcu, P.N., Long-range transport of pollutants observed through visibility contour maps, weather maps and trajectory analysis. Preprint volume: Third Symposium on Turbulence, Diffusion and Air Pollution, American Meteorological Society, Reno, NV pp. 344-347 (1976).

Husar, R.B., Poll, D.E., Holloway, J.M., Wilson, W.E. and Ellestad, T.G., Trends of eastern U.S. haziness since 1948. Preprint Volume: Fourth Symposium on Turbulence, Diffusion and Air Pollution, American Meteorological Society, Boston, MA pp. 249-256 (1979).

Husar, R.B., Holloway, J.M., Poll, D.E. and Wilson, W.E., Spatial and temporal pattern of eastern U.S. haziness: a summary. Atmos. Environ. 15, 1919-1928 (1981).

Husar R.B. Haze Climate of the United States, Final Report to EPA Cooperative Agreement CR810351, 1986.

Husar R.B. and Patterson D.E. Haze Climate of the United States. U.S. Environmental Protection Agency, EPA-600/S3-86-071, Research Triangle Park, NC 1986.

Husar R.B. and Wilson W.E., Haze and sulfur emission trends in the Eastern United States, Environ Sci. Technol., 27, 12-16 (1993).

Husar R.B., Elkins J.B. and Wilson W.E., US. visibility trends, 1960-1992, regional and national, Presented at the 87th Annual Air & Waste Management Meeting, Cincinnati, OH 1994.

Munn, R.E., Secular increases in summer haziness in the Atlantic provinces, Atmosphere 11, 156-161 (1973).

Ozkaynak H., Schatz A.D., Thurston G.D., Isaacs R.G., Husar R.B. Relationships between aerosol extinction coefficients derived from airport visual range observations and alternative measures of airborne particle mass, J. Air Poll. Contr. Assoc. 35, 1176-1185 (1985).

Robinson, E. and Valente, R.J., Atmospheric turbidity over the United States, 1948-1978. Research Publication GMR-3474, Env #92, General Motors Corp. (1982).

Poirot R.L. , Floccini R.G. and Husar R.B. Winter fine particle composition in the Northeast: Preliminary results from the NESCAUM monitoring network, Paper No.90-84.5. Presented at Air & Waste Manag. Assoc. Annual Meeting, Pittsburgh, PA, June 1990.

Schichtel B.A. and Husar R.B. Apportionment of light extinction by aerosol types. Report to Phillips Laboratory, Air Force Systems Command (1991).

Sisler J.F. and Huffman D., Latimer D.A., Malm W.C., and Pitchford M. Spatial and temporal patterns and the chemical composition of the haze in the United States: An analysis of data from the IMPROVE netweork, 1988-1991.Report #ISSn No. 0737-5352-26 CIRA,CSU, Fort Collins, CO, (1993).

Sisler J.F. and Malm W.C. The Relative importance of soluable aerosols to spatial and seasonal trends of impaired visibility in the United States. Atmos. Environ. 28, 851-862 (1994).

Sloane, C.S. Visibility Trends II: Methods of Analysis, Atmos. Environ. 16, 41 (1982).

Sloane, C.S. Visibility Trends II: Mideastern United States, Atmos. Environ. 16, 2309 (1982)

Sloane, C.S., Summertime visibility declines: meteorological influences. Atmos. Environ. 17, 763-774 (1982).

Sloane, C.S., Meteorologically adjusted air quality trends: visibility, Atmos. Environ. 18, 1217-1229 (1984).

Stevens R.K., Dzubay T.G., Lewis C.W., Shaw R.W. Source apportionment methods applied to the determination of origin of ambient aerosols that affect visibility in forested areas. Atmos. Environ. 18, 261-272 (1984)

Trijonis, J., Existing and natural background levels of visibility and fine particles in the rural East. Atmos. Environ. 16, 2431 (1982).

Trijonis J. and Yuan K. Visibility in the Northeast:long-term visibility/pollutant relationship, EPA Report 600/3-78-075, US Environmental Protection Agency, RTP, NC 1978.

Wolff, G.T., Kelly, N.A. and Ferman, M.A., Source regions of summertime ozone and haze episodes in the eastern United States. Water Air Pollution. (1982).

Back to Contents